ROS homeostasis during development: an evolutionary conserved strategy

Abstract

The balance between cellular proliferation and differentiation is a key aspect of development in multicellular organisms. Recent studies on Arabidopsis roots revealed distinct roles for different reactive oxygen species (ROS) in these processes. Modulation of the balance between ROS in proliferating cells and elongating cells is controlled at least in part at the transcriptional level. The effect of ROS on proliferation and differentiation is not specific for plants but appears to be conserved between prokaryotic and eukaryotic life forms. The ways in which ROS is received and how it affects cellular functioning is discussed from an evolutionary point of view. The different redox-sensing mechanisms that evolved ultimately result in the activation of gene regulatory networks that control cellular fate and decision-making. This review highlights the potential common origin of ROS sensing, indicating that organisms evolved similar strategies for utilizing ROS during development, and discusses ROS as an ancient universal developmental regulator.

Introduction

The prevailing life forms today are aerobic organisms which rely on O₂ for their energy production. Paradoxically, this life-essential gas is toxic in nature, and all aerobes including plants, humans or bacteria experience molecular damage when exposed to high concentrations of O₂ [1]. Moreover, even under ambient O₂ concentration, aerobic organisms suffer from continuous oxidative damage [2]. The ability to use O₂ has greatly stimulated evolution of complex multicellular life forms [3], and is thought of as being a prerequisite to this evolutionary step [4, 5]. Still, in the anoxic deep sea, anaerobic multicellular metazoan life forms lacking mitochondria have recently been discovered [6], indicating that anaerobic metabolism does not preclude the development of complex life.

The evolution of the photosynthetic apparatus capable of splitting H₂O into O₂, protons and electrons was the main cause for the introduction of O₂ into an anaerobic biosphere 2.2 billion years ago [7]. The release of O₂ at that time put organisms under strong selection pressure to effectively develop new redox energy couples. Thus, organisms concomitantly evolved antioxidant strategies to mitigate the effects of highly toxic O₂ derivates, i.e., superoxide, hydroxyl radical and hydrogen peroxide, commonly called reactive oxygen species (ROS) [8]. In the relative absence of O₂ and under the strong reducing conditions that existed on the primordial Earth’s surface, ancient unicellular life forms relied on metabolic pathways that use electron acceptors such as CO₂ and SO₄ [9]. Oxygen (O₂) is a much more effective electron acceptor resulting in the largest free energy release per electron transfer, yielding about fourfold more energy per molecule of oxidized glucose as compared to the most efficient anaerobic pathways [10]. Still, many enzymes involved in aerobic metabolism have remained O₂/redox sensitive, which is especially the case for those containing iron-sulfur (Fe–S) clusters and/or cysteine residues [11–13]. In part, organisms have solved this problem by evolving efficient Fe–S cluster biogenesis machinery that allows for rapid reactivation of these enzymes during oxidative stress conditions [14]. Next to that, signaling and regulation processes can benefit from redox-sensitive proteins. Systematic analysis of the evolution of enzymes and the development of biochemical networks revealed that O₂ is one of the most utilized molecules, superseding NAD and ATP [8], although it constitutes a major threat to cell viability by the inevitable generation of ROS. Reactive oxygen species are toxic in nature, however, in the past decade, it has become apparent that, in addition, they act as indispensable signaling molecules for numerous developmental and adaptation processes in plants, animals, bacteria and fungi [15–18].

Since the oxygenation of the atmosphere occurred at a time during which mainly unicellular organisms populated the Earth, the adaptation to oxidative stress has evolved alongside with organism complexity. ROS by-products are released from a whole palette of aerobic reactions, and therefore cells needed to be able to monitor and control these toxic products during the adaptation towards aerobic growth. The development of complex cells and multicellularity brought with it a requirement for more sophisticated signaling systems to enable communication between different organelles and cell types. In contrast to the classical animal and plant hormones, ROS might represent an ancient shared signal between kingdoms. If this is true, the ways in which ROS are received and through which they affect cellular functioning and development should partially be conserved at the mechanistic level. In this review, we aim to provide evidence for a conserved role of ROS during the development of bacteria, fungi, animals and plants.

Ancient oxygenic minimal signal transduction systems

A common theme for aerobic life forms is the evolution of various ROS-responsive systems. Although O₂ is the most energetic electron acceptor, many unicellular organisms are only facultative aerobes. With some exceptions, eukaryotes depend on O₂ respiration and organic carbon as fuel, whereas prokaryotes use a whole array of electron acceptors and energy sources [9]. Therefore, it is not surprising that eukaryotes have more sophisticated systems to not only signal but also actively generate ROS such as superoxide during signaling responses to a variety of environmental factors or developmental cues. The simplest signaling system involves the direct interaction of the signal with a transcriptional regulator, the so-called one-component system (Fig. 1). In fact, the best understood ROS-responsive transcription factors (TFs) in terms of their molecular mechanism are those from bacteria and fungi [19]. In Escherichia coli, the helix-turn-helix (HTH) OxyR TF is directly oxidized by hydrogen peroxide to form an intramolecular disulfide bond, resulting in the transcriptional activation of hydrogen peroxide-responsive genes [20, 21]. Similarly, the HTH OhrR transcriptional repressor in Bacillus subtilis is oxidized by peroxides at a single cysteine residue within its DNA binding domain releasing it from the target promoter. To date, the vast majority of ubiquitous so-called “single-Cys” and “two-Cys” redox sensors identified in bacteria are known or projected to essentially function in this way [22]. Bacteria also contain transcriptional superoxide sensors, with the E. coli HTH SoxR TF being the best characterized protein that is activated by oxidation of Fe–S clusters [23]. SoxR activates the expression of the gene SoxS in response to exposure to superoxide-generating agents which leads to increased expression of several antioxidant genes, e.g., superoxide dismutase and glucose-6-phosphate dehydrogenase, which provide reducing equivalents in the form of NADPH. The yeast bZIP domain containing TF AP-1-like transcription factor 1 (Yap1) represents the first H₂O₂-sensing mechanism in eukaryotes. Hydrogen peroxide activates the TF by oxidation of two cysteine residues resulting in the induction of antioxidant genes during oxidative stress [24]. The mammalian NF-E2-related factor 2 (Nrf2) bZIP TF is redox-regulated and a known inducer of antioxidant genes by directly binding to the antioxidant responsive element (ARE) in the promoter of target genes, which was originally described as the binding site for activator protein-1 (AP-1) [25, 26]. Activator protein-1 is a transcription factor complex that consists of either a Jun–Jun homodimer or a Jun–Fos heterodimer. Both forms induce the expression of a diverse range of genes involved in proliferation, differentiation and defense. Although no direct homologous of OxyR, SoxR or Yap1 proteins have been found in plants, the known redox sensing mechanisms are conserved between species albeit different molecular players are involved.

Fig. 1.

Schematic overview of the one-component and two-component signal transduction paradigm. The simplest transduction system is the one-component system (left) which is mainly found in unicellular organisms and, e.g., includes the OxyR and SoxR TFs, depicted as response regulators (RR) whose cysteine residue or Fe–S cluster, respectively, is oxidized by ROS. In the classical two-component pathway (left), the HK domain of the sensor kinase autophosphorylates and subsequently transfers a phosphoryl group to the aspartate residue of the RR. The sensory domain connected to the HK domain varies widely to allow receiving the various signals. Phosphorelay systems (second from right) are a common extension of the minimal two-component signaling cascade which in plants is involved in cytokinin signaling. Perception of a stimulus activates autophosphorylation and the phosphoryl group is intramolecularly passed to a C-terminal receiver domain. Subsequently, an additional histidine phosphotransferase (HPT) shuttles the phosphoryl group from the hybrid kinase to a soluble RR protein. Alternatively, as in plant-specific ethylene signaling or in some other higher eukaryotic organisms (rats), the phosphorelay feeds forward onto the MAPK cascade (right), which subsequently activates an RR protein via Ser/Thr residues. Taken together, one- and two-component pathways enable cells to sense and respond to stimuli by inducing changes in transcription

Interestingly, in Arabidopsis thaliana, the redox-regulated NONEXPRESSOR OF PATHOGENESIS-RELATED PROTEINS 1 (NPR1) is present in an inactive oligomeric form established through intermolecular disulfide bonds which are released upon perturbation of the redox balance [27]. NPR1 functions as a transcriptional coregulator that, upon activation, localizes to the nucleus and interacts with members of the D group of Arabidopsis bZIP factors (TGA) [28]. Moreover, the TGA factors bind the oxidative stress and salicylic acid responsive as-1 cis-element found in promoters of glutathione transferases and other antioxidant genes [29]. Interestingly, both the mammalian ARE motif and the plant as-1 element share a TGAC core sequence which is the characteristic binding site for a subset of bZIP TFs [25, 29]. In addition, TFs of the plant-specific oxidative stress-responsive WRKY family can bind to a cis-element with a TGAC core motif [30]; however, no evidence for redox regulation of these TFs has so far been demonstrated [31].

Due to the simplicity of the one-component TFs, it is unlikely that they are sufficient to fully control gene expression in the highly compartmentalized eukaryotic cell. Especially for the transduction of signals from the extracellular to the intracellular space, other signaling cascades are required to act upstream of the transcription factors. Similar to WRKY proteins in plants, Fos and Jun in animal cells are targets of the mitogen-activated protein kinase (MAPK) cascade and are stabilized and activated by phosphorylation [32, 33].

Although the MAPK cascade is a well-defined integrator of ROS signals in plants, yeast and animals it is not present in prokaryotes [34–36]. In contrast, transmembrane signal transduction represents a common feature of all eukaryotic and prokaryotic cells [37]. As MAPK cascades act as mediators between sensor proteins and effector proteins, they likely evolved at a later stage in evolution. A more ancient evolutionary conserved signaling pathway is the two-component system which consists of a sensor histidine kinase (HK) and a response regulator (RR) with a DNA-binding domain (Fig. 1) [38]. In complex organisms, one or more phospho-relay system(s) and/or MAPK cascade(s) expanded the original HK-RR system, indicating that mainly the signaling mechanisms rather than individual proteins are conserved. In the case of phospho-relay, the HK autophosphorylates and transfers the phosphoryl group to an internal receiver domain rather than to a separate RR protein. Subsequently, the phosphoryl group is shuttled to a histidine phosphotransferase (HPT) and finally to a terminal RR protein (Fig. 1). The His–Asp phosphorelay systems were originally thought of as being exclusive to prokaryotes [39]; however, it was shown in the 1990s that fungi and plants also use these systems [40, 41]. Moreover, signaling in eukaryotes is not limited to His–Asp phosphotransfer as seen in the high-osmolarity glycerol (HOG) pathway in yeast (Saccharomyces cerevisiae and S. pombe), which forwards its signal to the MAPK cascade pathway [42]. The majority of bona fide two-component elements in plants play a central role in cytokinin signaling and involve HKs, HPTs and RRs [43]. The Arabidopsis genome encodes nine HKs, of which three, AHK2, AHK3 and AHK4, have been implicated in cytokinin signaling [44]. While the cytokinin receptors are located at the endoplasmic reticulum, AHK5 localizes to the plasma membrane and has been shown to be involved in H₂O₂-dependent closure of stomata [45]. Besides AHK5, ETHYLENE RESPONSE 1 (ETR1) contains an HK domain and functions in H₂O₂ signaling [46]. Though ETR1 does provide HK activity, such activity does not appear to be necessary for ethylene responses [46]. Two-component signaling systems, which in part also feed forward onto the MAPK cascade, have been confirmed as H₂O₂ sensors in yeast [47–50]. Here, the SLN1 transmembrane protein is part of a typical two-component signal transduction system that senses environmental changes leading to the activation of its HK domain. The phosphoryl group is transferred to a histidine residue in Ypd1p, a small cytoplasmic protein, and subsequently transferred to an aspartate residue in the receiver domain of the RR protein SSK1 [47]. Loss of the SLN1–SSK1 system results in H₂O₂ susceptibility (not other oxidants) that can be relieved by introducing ETR1, indicating conserved mechanisms for sensing hydrogen peroxide among yeast and plants [46]. Also, in cyanobacteria, HKs functioning as H₂O₂ sensors have been found [51]. In conclusion, the two-component system appears to be a conserved strategy for the perception of ROS in plants, fungi and bacteria. Animals lack HK two-component-like sensors, but nevertheless, histidine phosphatases, which are involved in the cytokinin signaling pathway in plants [44], and histidine–aspartate phosphorelay mechanisms as in E. coli [52] have been isolated in rats [53].

Interestingly, the potential plant ROS sensors overlap with known signaling pathways for phytohormones like salicylic acid, cytokinin and ethylene, suggesting that ROS is interconnected with these development-regulating hormones.

Role of ROS during cell proliferation and differentiation

During the evolution of multicellularity, the omnipresent oxygen radicals were potentially harnessed as second messengers coordinating cellular metabolism and growth [53]. The major advantage of redox sensing is the ability to respond rapidly and adequately to environmental stimuli and to integrate all cellular metabolic activities [54]. This is exemplified in the social amoeba Dictyostelium discoideum, which straddles the boundary between a unicellular and a multicellular organism. At the onset of its development shift from unicellular to multicellular, D. discoideum utilizes superoxide as a signaling intermediate [19]. Cellular decision-making, environmental sensing and cell-to-cell communication are key processes underlying pattern formation and development from microbes to plants and mammals. Note that the term “cellular decision-making” may be misleading, as the decisions actually occur at the level of gene regulatory networks such as the ones controlled by outputs of the one-component and two-component systems [55]. Cells merely provide microscopic meeting places for the real key players: genes connected into regulatory networks [56]. Cellular decision-making is frequently based on communication networks with multiple nested feedback loops. Studying the dynamics of the multiple feedback loops and their role in differentiation and development has revealed novel insights, as recently shown for plants and microbes [57–59]. Since the expression of many genes is affected by redox changes irrespective of the organism, ROS appear to have progressed as important inputs and outputs of cellular decision-making [60, 61].

One crucial step for life is reproduction, which in its most basic form is represented by a mother cell dividing into two daughter cells. In human cells, it has been found that very low levels of H₂O₂ or superoxide stimulate the proliferation of smooth muscle cells [62, 63], macrophages [64], and vascular endothelial cells [65], among others (Fig. 2). In addition, overexpression of catalase and/or superoxide dismutase, which lower H₂O₂ or superoxide levels, respectively, inhibit proliferation of vascular smooth muscle cells in response to epidermal growth factor (EGF), which is accompanied by a reduction in extracellular-signal-regulated kinase (ERK) 1 and 2 phosphorylation events [66, 67]. However, higher concentrations of hydrogen peroxide can temporarily arrest growth, followed by a transient adaptive response at the expression level for oxidant protection and DNA repair [68]. At even higher hydrogen peroxide concentrations, mammalian fibroblasts are not able to adapt; instead, they enter a state resembling cellular replicative senescence or die in an organized manner resulting in apoptosis [68]. In recent years, the implications of ROS on the cell cycle have been further refined by the discovery of a redox-dependent signaling pathway that controls cell cycle proteins [69]. Distribution of specific ROS acts as an important signal at the mRNA and protein levels during cell cycle progression in animal cells [70], both in healthy tissues and during disease [71]. In mammalian cells, a transient increase in cellular oxidant levels in the G₁ phase [70] is required for entry into the S phase, which is also observed for plant cells [72]. Also, in yeast, a redox cycle that links the metabolic cycle with the budding phase has been discovered [73].

Fig. 2.

Output of ROS on cellular proliferation and differentiation, depending on the strength and type of the signal. A transient increase in ROS levels stimulates cell division, while a higher level can result in cell differentiation. Even higher levels, as experienced during cellular stress, will result in growth inhibition and subsequent adaptation. However, when ROS levels further increase and cells can no longer adapt, they either go into a state resembling cellular replicative senescence or die in an organized manner resulting in cell death. The strength of the ROS signal is indicated by the weight of the arrow

The finding that ROS homeostasis regulates the transition from proliferation to differentiation has been nicely demonstrated by Benfey and co-workers in the root of the model plant species Arabidopsis thaliana (Fig. 2) [74]. Superoxide and H₂O₂ are differentially distributed within a plant root [75], while superoxide mainly accumulates in dividing and expanding cells of the meristem and H₂O₂ accumulates in the elongation zone; an overlap of both ROS types is observed within the so-called “transition zone”. The basic helix-loop-helix (bHLH) transcription factor UPBEAT1 (UPB1) shows increased expression in the root transition zone and directly represses the expression of peroxidase genes. Loss of UPB1 results in longer roots, while overexpression decreases root organ size. The elongation zone of upb1 mutant roots shows decreased and that of UPB1 overexpressor roots increased levels of H₂O₂. In contrast, superoxide increased in the elongation zone in upb1 roots and was reduced in plants overexpressing UPB1. Taken together, the position of the transition zone is determined by the coincidence of the gradients formed by H₂O₂ in the elongation zone (required for differentiation) and by superoxide in the meristem (to maintain cellular proliferation) [74]. The functional characterization of UPB1 revealed the first insight into the gene regulatory networks that link ROS signals with the developmental program in plants. Of note, ROS not only guide cell differentiation in plants: in Drosophila melanogaster, ROS prime hematopoietic progenitors for differentiation [76]. Multipotent hematopoietic progenitor cells show increased basal levels of ROS under physiological conditions, which are lowered upon differentiation. Decreasing ROS levels by scavengers retards their differentiation while treatment with moderate ROS levels triggers precocious differentiation into all three mature blood cell types found in D. melanogaster. Reactive oxygen species have long been known to induce apoptosis and negatively influence neuronal cell fate; intriguingly, they appear to be essential during neurogenesis [16]. Although the exact mechanisms involved in ROS-mediated neurogenesis remain unclear, it has been established that endogenously produced H₂O₂ is perceived by a MAPK pathway resulting in the activation of redox-sensitive TFs [16]. In the fungus Neurospora crassa, a hyperoxidant state occurs at the start of each cell differentiation step leading to conidia development [77]. The hyperoxidant state is defined as an unstable, transient state in which ROS surpass the antioxidant capacity of the cell [78]. It is known that, in several of the examples mentioned above, antioxidants inhibit development, which is consistent with the idea that ROS are required for the developmental process to occur. On the other hand, it is expected that a lack of antioxidant enzymes would result in higher ROS levels and precocious or enhanced cell differentiation.

The process of cell death induced by very high ROS levels is an active process in many species, and often part of their developmental and defense program (Fig. 2). Interestingly, the cell death response is conserved between unicellular and multicellular organisms, although its appearance comes in many flavors. In animals, the cell death types are divided into two categories: apoptosis and necrosis [79]. Discrimination between these two categories is based on the presence or absence of specific biochemical and molecular hallmarks, such as DNA laddering, cytochrome c release, caspase involvement, ATP depletion, cytoplasmic swelling, and loss of membrane integrity [80]. Next to eukaryotes, even bacteria harness ROS to undergo programmed death by activating the suicide module mazEF [81]. The reason for single cell organisms to commit suicide lies in the fact that most of them live in large colonies that may act as a multicellular organism [55]. Apoptosis-like elimination of defective cells in S. cerevisiae suggests that all unicellular life forms have evolved altruistic programmed death that serves a variety of useful functions [82]. In plants, the process of cell death is either passive (necrotic cell death) or active (programmed cell death), and it has been shown to be caused by high ROS levels during development [83].

Taken together, from the above reasoning, it becomes clear that, in unicellular and multicellular organisms, ROS appear to act in a similar manner on development. At a first glance, ROS seem to be rather unspecific regulators; however, to distinguish the message contained within the ROS signal four types of ‘information packaging’ evolved, in addition to the identity of the ROS involved, the intensity of the signal (burst), its duration, and the (sub)-cellular location are important.

Conservation of ROS producing and scavenging proteins and their role in development

As indicated above, ROS-employing signaling networks influencing development exist in distant species. To further explore this common theme, we highlight below several classes of ROS-related proteins that appear to be present in all kingdoms and to play essential roles during development.

Thioredoxins and glutaredoxins

The thioredoxin (TRX) super family is characterized by a TRX fold and catalyzes oxidoreductase reactions by a dithiol/disulfide exchange mechanism involving two redox-active cysteine residues. Besides thioredoxins, the family of TRX fold proteins also includes glutaredoxins (GRXs); both types of proteins are considered as reductants and can be found in bacteria, fungi, mammals and plants [84]. The major function of TRXs and GRXs is to reduce disulfide bonds within target proteins [85]. Oxidized target proteins can be monomeric, homo- or hetero-dimeric and possibly glutathionylated proteins [84, 86]. Although TRXs and GRXs have similar structures and activities, they show little similarity regarding their amino acid sequence. Furthermore, GRXs are non-enzymatically reduced by glutathione (GSH), whereas TRXs are enzymatically reduced by TRX reductases (TR) (Fig. 3). E. coli encodes two TRX, three GRX and two GRX-like proteins, which are individually dispensable, whereas GRX4 is involved in Fe–S cluster assembly in E. coli and yeast [84, 87, 88]. Interestingly, lethality caused by simultaneous inactivation of TR and glutathione reductase in E. coli can be rescued by dithiothreitol treatment [89]. As mentioned before, the function of redox sensing is to adjust growth and development which in turn depend on metabolism. In E. coli, it was found that TRX and GRX proteins are involved in many housekeeping reactions, e.g., the synthesis of deoxyribonucleotides by activation of ribonucleotide reductase (RNR) [90] and sulfate assimilation by interacting with 3′-phosphoadenosine-5′-phosphosulfate reductase (PAPS) [91]. Furthermore, oxidative damage repair via methionine sulfoxide reductase (MSR), that rescues proteins whose methionine residues have been oxidized to the sulfoxide state (MetSO) [92], depends on TRX. Moreover, it was shown that plastidial MSR activity allows for maintaining vegetative growth of plants during environmental constraints through the preservation of photosynthetic antennae [93]. Furthermore, the OxyR one-component sensor described above is inactivated by TRXs and GRXs when oxidative stress is alleviating [94]. Yeast (S. cerevisiae and S. pombe) provides a similar set of TRX, GRX and GRX-like proteins as E. coli, but one member of each TRXs and GRXs is localized in mitochondria and the GSH-dependent system is essential [84].

Fig. 3.

Two common thiol redox pathways that set redox homeostasis in cells. Under oxidative conditions, disulfide bonds are formed within proteins that can either activate or inactivate them. The major pathways for reducing the disulfide bonds in proteins rely on thioredoxins (TRX) or glutaredoxins (GRX). GRXs are reduced non-enzymatically by glutathione (GSH), whereas TRXs are reduced enzymatically by TRX reductase (TR). Oxidized GSH is recycled by glutathione reductase (GR). Both, TR and GR use NADPH as a substrate during the reduction reactions, except for the plastidial TR proteins which rely on ferredoxin (FTR) instead of NADPH for reduction

Vertebrate genomes encode at least one cytosolic and one mitochondrial set of TRX, GRX and GRX-like proteins [84]. Similar to yeast and bacteria, TRX function is essential in animals as indicated by embryo lethality upon loss of TRX1 in mice [96]. In addition, some targets of the system are also conserved between the kingdoms, including RNR and MRS [97]. Peroxiredoxins (PRXs) represent another group of common target proteins potentially functioning as H₂O₂ sensors and scavengers which are themselves dependent on the TRX and/or GRX systems as discussed below [84, 95]. Furthermore, the activation of the AP-1 TFs (Fos and Jun) that bind and recognize the ARE element is partially regulated by this redox system [25, 98]. Finally, the TRX and GRX systems are conserved in plants but operate mechanistically different. First of all, the eukaryotic plant cell contains plastids and mitochondria, adding an additional layer of information processing to the system. The reduction of TRX in the cytosol and mitochondria of plant cells is performed by a bacterium-like NADPH-dependent TRX reductase (NTR), while in plastids the reduction is carried out by ferredoxin [99]. Plants have very complex thioredoxin systems, as revealed by sequencing of the Arabidopsis genome resulting in approximately 40 genes encoding putative TRX and TRX-related proteins. NADPH-dependent TRX reductases are key regulatory enzymes determining the redox state of the thioredoxin system. Arabidopsis encodes at least two NTRs (NTRA and NTRB). Surprisingly, the double knock-out mutant ntra ntrb is viable and fertile [100]. Thus, in contrast to the other biological systems, neither cytosolic nor mitochondrial NTRs are essential in plants. It has been shown that a glutathione-dependent pathway for reducing TRX is utilized in NTRA/B-deficient plants. Therefore, in contrast to bacteria and animals, in which the NTR/TRX and GSH/GRX systems operate separately, these pathways are interconnected in the cytosol of cells of land plants [84]. That GSH has a major role in plant development became evident after map-based cloning of the ROOT MERISTEMLESS 1 (RML1) gene, which encodes γ-glutamylcysteine synthetase (γ-ECS) that mediates the first step in glutathione biosynthesis [101]. Within the root meristem, the cell cycle gets blocked in the G₁ phase, indicating a requirement for GSH during cell proliferation in plants. Later studies in both animals and plants demonstrated that GSH is specifically recruited into the nucleus during the early phase of cell proliferation to adjust whole-cell redox homeostasis [102, 103]. Crossing ntra ntrb with rml1 resulted in the complete inhibition of both shoot and root growth, indicating that at least NTR or the glutathione pathway is required for postembryonic activity in the apical meristem [100]. In the search for Arabidopsis mutants impaired in cell-to-cell transport via plasmodesmata (PD), the seedling lethal mutant gfp-arrested trafficking (gat1) was identified, which affects unloading of GFP from the phloem into the root meristem causing a plant size reduction [104]. GAT1 encodes the plastidial thioredoxin TRX-m3. Its effect on cellular redox status seems to play a role in the regulation of PD permeability, as indicated by wild-type plants subjected to oxidative conditions that phenocopied the gat1 trafficking defects. Most TRX proteins are specifically localized in plastids and encoded by nine small gene families in Arabidopsis: TRX-m, TRX-f, TRX-x, TRX-y, CDSP32, HCF164, WCRKC, APR, and Lilium with four, two, one, two, one, one, two, three, and five members, respectively [84]. So far, only TRX-o1 has been shown to be specific for mitochondria, while another ten members lacking a transit peptide are heterotrophically localized (h-type) [84]. The importance of plastidial TRXs lies in the fact that they regulate carbon metabolism during the diel cycle. To survive at night, chloroplasts derive energy from the breakdown of carbohydrates, in part via the oxidative pentose cycle [105]. During the day, glucose 6-phosphate dehydrogenase, the enzyme catalyzing the first step of the oxidative pentose phosphate pathway, is deactivated by TRXs [106]. Recently, it was demonstrated that chloroplastic TRXs are circadian regulated through the core clock, which might be related to their redox regulation of light-dependent protein targets [107]. Still, one of the most interesting proteins that depends on TRXs for its activity is the transcription factor PHAVOLUTA (PHV) which is a central regulator of plant development by controlling radial patterning [108, 109]. PHV only binds to DNA under reducing conditions. The protein contains a CxxxC motif that has been implicated as a redox-switch [110]. Worthy of note is the fact that the two cysteines involved in redox modulation are conserved in all members of the HD-Zip III family to which PHV belongs and which are known for their role in plant development [109]. Therefore, it would be of great interest to further analyze the redox regulation of this kind of TFs in relation to plant growth. The NPR1 activation through monomerization (see above) is achieved by the activity of TRX-h3 and TRX-h5 in the cytosol [111]. Thus, NPR1 represents a genuine target of the ROS sensors. The number of GRX genes is particularly high in plants, with at least 31 representatives in, e.g., Arabidopsis [112], which are tightly linked to the GSH network. GSH is a highly abundant antioxidant that can be present in millimolar concentrations in eukaryotic cells. A picture has emerged in the past that shows a role of GRXs in plant development. A screen for floral development mutants in Arabidopsis revealed a mutant defective in the initiation of petal primordia and subsequent development called roxy1 [113]. ROXY1 encodes a plant-specific CC-type GRX, 1 of the 21 members of the CC-type family [114]. Subsequently, the authors looked at the closest homologue of ROXY1, i.e., ROXY2, which does not cause an obvious developmental phenotype when knocked out. However, the double roxy1 roxy2 mutant is unable to develop normal anthers and lacks pollen [115]. The ROXY proteins hold another surprise: protein–protein interaction studies identified the TGA bZIP TFs as target proteins [116]. Interestingly, one of these interacting TGA factors is PERIANTHIA (PAN), which when mutated causes a characteristic transformation of tetrameric to pentameric flowers [117]. Moreover, PAN directly binds to the promoter of the MADS box transcription factor AGAMOUS (AG) to enhance its expression, indicating that redox-controlled growth has to be implicated in the ABC model of floral development [118]. TGA9 and TGA10 are two redundant factors that interact with ROXY1 and ROXY2. They play an essential role in anther development, but act downstream of the MADS-box-like TF SPOROCYTELESS/NOZZLE (SPL/NZZ) [119]. The cytosolically located class I GRX proteins are essential for plant development as indicated by the double mutant grxc1 grxc2, which is embryo lethal [120]. GRXS17 in Arabidopsis regulates polar auxin transport and thereby controls root development [121]. Root cells appeared to precociously enter differentiation, indicating that GRXS17 is required for cell cycle progression and proliferation. Next to that, loss of GRXS13 results in the accumulation of superoxide and reduced plant growth [122], indicating that the imbalance in redox homeostasis is directly translated into a growth phenotype. In conclusion, the cross-kingdom conserved GRX and TRX proteins seem to be encoded by highly specialized developmental genes in plants. This specialization became possible by the evolutionary establishment and diversification of the relative large number of family members. It will be of great interest in the coming years to identify and characterize the cellular targets regulated by each TRX and GRX.

Thiol peroxidases; peroxiredoxins and glutathione peroxidases

As indicated above, TRX and GRX proteins serve as redox transmitters within the cellular thiol/disulfide redox network to control many cellular functions [123]. Peroxiredoxins (PRX) and glutathione peroxidases (GPX) act as thiol-dependent peroxidases with high affinity for peroxide to protect protein thiols from oxidation [124]. Peroxiredoxins have a much higher affinity to H₂O₂ than catalases providing the specificity required for H₂O₂ sensing and signaling [125]. Peroxiredoxins can react with H₂O₂ at low “signaling” concentrations before this molecule is intercepted and degraded by other systems, which fits the function of H₂O₂ receptors described for thiol peroxidases in lower eukaryotes [126]. Of the thiol-based peroxidases, the PRXs are conserved across all kingdoms while the related GPXs are confined to bacteria, fungi, protozoa, insects, and plants [125]. In fission yeast, the peroxiredoxin TPX1 is essential for growth under aerobic conditions, which can be attributed to its scavenging and signaling properties [127]. TPX1 is required for the H₂O₂-dependent activation of the bZIP TF PAP1 at mild H₂O₂ levels [128]. PAP1 temporarily enters the nucleus upon oxidative stress and negatively regulates growth, and returns to the cytoplasm once cells have adapted to the stress. At a higher concentration of H₂O₂, TPX1 activates the MAPK-dependent pathway controlled by Sty1 [129], revealing an elegant molecular switch in which H₂O₂ signaling is performed by PRXs in a concentration-dependent manner. Moreover, a recent transcriptomic study of yeast (S. cerevisiae) lacking all eight thiol PRXs resulted in the lack of a transcriptional response specifically towards H₂O₂ [130]. This study indicates that thiol peroxidases sense oxidative signals and transfer them to signaling proteins and regulate transcription. Moreover, the direct interaction of H₂O₂ with other cellular proteins is of only secondary importance in yeast. Whether the transcriptional response to H₂O₂ is also performed by PRX proteins in other organisms remains to be demonstrated. In bacteria, the deletion of the PRX gene ahpC results in shortening of the generation time [131]. One potential reason for a reduced generation time is that the mutant defective in ROS scavenging more rapidly accumulates DNA damage and thereby loses viability. A more rapid cycling might help to compensate for the negative effect on DNA integrity. Mammalian PRXs potentially have a role in H₂O₂ signaling by controlling the fluxes and concentration of hydrogen peroxide which is supported by effects of PRXs on downstream signaling of growth-factor tyrosine kinases and cytokinin receptors [132]. The mouse genome encodes six PRX genes which have been studied to some extent by knock-out analysis, mainly revealing a decreased oxidative stress tolerance and an early onset of several diseases like cancer, anemia and artery problems [125]. The regulation of transcription by PRXs is still poorly understood in animal systems, but redox signaling also takes place at physiological H₂O₂ fluxes similar to other systems, and altering PRXs levels affect these pathways [133]. In Arabidopsis, the PRX family is encoded by ten genes while the GPX family is represented by eight genes [124]. The PRX proteins are divided into six classes with different cellular locations. ATPER1 is specifically expressed during seed development and is regulated by ABA INSENSITIVE 3 (ABI3) to potentially control dormancy [134]. Furthermore, ATPER1 expression during early embryogenesis depends on an antioxidant-responsive promoter element (ARE). Of the four PRXII genes that encode cytosolically located enzymes, none has so far been functionally characterized. PRXIIB, C and D are each 162 amino acids long and exhibit highly similar amino acid sequences with 93–99 % identity, suggesting a potentially redundant role [124]. On the other hand, the PRXIIA protein has an additional F-box domain and thereby differs from its counterparts. PRXIIE is a plastidial protein which has been shown to be involved in nitrosylation and probably functions as a scavenger of nitric oxide in the plastid stroma [135]. Similarly, mitochondrial PRXIIF is involved in protein nitration; knock-out plants show reduced root length under oxidative stress conditions [136]. Furthermore, the authors suggested a potential role of PRXIIF in redox signaling in plant cells, since many mitochondrial and nuclear transcripts showed an altered abundance. In addition, the chloroplast contains another four PRX proteins of which two belong to the 2-Cys PRX class that is essential for chloroplast metabolism and plant biomass production [124, 137]. The other two are PRXQ proteins, which are atypical 2-Cys PRX forms that function as monomeric peroxidases with high reactivity towards H₂O₂ and lipid peroxides [124]. Single knock-out lines of PRXQ genes in Arabidopsis exhibit no clear phenotype, while RNAi targeting of both genes results in decreased oxidative stress tolerance and altered photosynthesis [124, 138]. So far, the function of PRX in plant development is still poorly analyzed and needs additional research efforts to unravel the redox regulation as shown in yeast. The glutathione peroxidases GPX1 and GPX7 represent two chloroplast localized proteins that modulate photooxidative stress tolerance and basal plant resistance [139]. Loss of function of both genes greatly affects leaf morphology in an adaxial-specific manner, suggesting that natural photooxidative stress might help guiding leaf growth. Interestingly, GPX3 has been shown to interact with the ABI1 and ABI2 Ser/Thr phosphatases and to have a role in drought stress tolerance and development [140]. Moreover, it was demonstrated that GPX3 inhibits the PP2C activity of ABI2 in a redox-dependent manner and thereby regulates ABA signaling in Arabidopsis. This indicates that, as in yeast, GPX-like enzymes might potentially act as peroxide sensors and redox transducers in plant cells. GPX5 is essential for female gametophyte development and causes a maternal defect [141]. While the other members of the Arabidopsis GPX family remain to be analyzed in more detail, it is tempting to state that in plants a redox-sensing and output network similar to that in yeast exists.

NADPH oxidase family

Whereas the TRX and PRX proteins constitute evolutionary conserved ROS sensors, ROS producing enzymes, including the specialized respiratory burst oxidase homolog (RBOH) or NADPH oxidase (NOX) enzymes, play a central role in ROS signaling and homeostasis [142]. NADPH oxidases are located within the plasma membrane where they generate a superoxide radical by using NADPH as an electron donor. At the time of evolution, during which the unique chemical/thermodynamic properties of O₂ were harnessed for aerobic respiration, cells also adopted ROS for signaling and regulation [10]. The diversity and number of NOX proteins correlate with increasing organism complexity [143]. NOX enzymes are expressed in all multicellular eukaryotes, including fungi, plants, and animals, but not in unicellular prokaryotes or lower eukaryotes like, for instance, yeast. Humans and animals in most cases have seven NOX genes, while Arabidopsis has ten predicted members [143]. On the other hand, Caenorhabditis elegans, D. melanogaster and several fungi each express two members of this gene family. Numerous studies have pointed out that NOX genes play an essential role during development in all species so far examined. In plants, the first indication for a role of NOX genes in development derived from the identification of the rhd2 mutant lacking a functional RBOHC gene, which is essential for root hair cell expansion [144]. Similarly, NADPH oxidase activity is required for pollen tube growth [145]. Others like RBOHB, RBOHD and RBOHF appear to be involved in a multitude of processes, including pathogen response, systemic signaling, lignification and seed ripening [146]. However, due to functional redundancy, it is difficult to pinpoint specific functions for each of the ten RBOH proteins in Arabidopsis. One of the challenges is the identification of downstream signaling or effector proteins that mediate the ROS signal triggered by the RBOH proteins. In mammals, it has been put forward that PRX proteins are potentially involved in mediating the signaling downstream of NOX proteins [147].

Although ROS scavenging proteins like catalase and superoxide dismutase are highly conserved between kingdoms [148, 149], we have refrained from discussing their potential role in development here, since they do not appear to be as sensitive as, for example, PRX proteins to signal ROS [126]. However, the appearance of these proteins was certainly instrumental in protecting the cellular environment from the damaging potential of ROS and for successful embedding of mitochondria and chloroplasts into eukaryotic cells.

Perspective

Compared with a decade ago, the opinion on ROS as a negative factor has been largely adjusted. It is not surprising that ROS evolved as a major signaling component in all aerobic life forms. Their co-appearance with the evolution of complex life has made them an integral part of the transcriptionally regulated developmental program. In all organisms discussed, ROS evoke a specific transcriptional response, which does not hold up for classical hormones which only function in specific kingdoms. Moreover, as depicted in the study on yeast PRX proteins [130], active ROS sensing mediates a transcriptional response rather than passive damaging of the cell. Amazingly, many parallels appear to exist in redox sensing strategies in the different organisms, which might be due to the conservation of ancient core redox sensor mechanisms. Just recently, it has been found that oxidation–reduction cycles of peroxiredoxin proteins constitute a universal marker for circadian rhythms in all domains of life [150]. Therefore, cellular time keeping appeared to have co-evolved with redox homeostasis and the oxygenation of the world.

Although it is now clear that ROS are conserved developmental regulators, many players that sense and execute the diverse transcriptional programs still need to be uncovered. Furthermore, in plants, ROS interact with virtually all phytohormone pathways; thus, from an evolutionary point of view, it would be interesting to determine when and how these pathways started to interact. For example, NPR1 is a redox- and salicylic acid-regulated transcriptional coregulator present in all higher plants; however, so far, no ortholog has been identified in algae. Still, algae do respond to salicylic acid treatment by upregulating alternative oxidase (AOX), as do higher plants [151]. Finally, further genetic and transcriptional studies are required to unravel in detail the functions of the many TRX, GRX, PRX and GPX proteins in plants, but also other organisms, to unveil how ROS have been integrated during aerobic development.