A review of redox signaling and the control of MAP kinase pathway in plants

Mitogen-activated protein kinase (MAPK) cascades are evolutionarily conserved modules among eukaryotic species that range from yeast, plants, flies to mammals. In eukaryotic cells, reactive oxygen species (ROS) has both physiological and toxic effects. Both MAPK cascades and ROS signaling are involved in plant response to various biotic and abiotic stresses. It has been observed that not only can ROS induce MAPK activation, but also that disturbing MAPK cascades can modulate ROS production and responses. This review will discuss the potential mechanisms by which ROS may activate and/or regulate MAPK cascades in plants. The role of MAPK cascades and ROS signaling in regulating gene expression, stomatal function, and programmed cell death (PCD) is also discussed. In addition, the relationship between Rboh-dependent ROS production and MAPK activation in PAMP-triggered immunity will be reviewed.

groups based on the metal cofactors: iron SOD (Fe-SOD) is predominantly localized in the chloroplast, manganese SOD (Mn-SOD) is localized in the mitochondria and peroxisomes, and copper/zinc SOD (Cu/Zn-SOD) is localized in the cytosol and chloroplast [13][14][15]. Nonenzymatic ROS scavenging mechanisms consist of soluble molecules which can be oxidized by ROS. Glutathione (GSH) and ascorbate (ASC) are the major non-enzymatic systems by which H 2 O 2 can be reduced to H 2 O and O 2 [16]. Maintenance of the reduced and oxidized glutathione ratio (GSH/GSSG) is critical for ROS homeostasis.
In contrast to their potentially harmful effects, ROS play an important role in regulating plant growth and development, and response to biotic or abiotic stresses. It has long been observed that a virulent pathogens, successfully recognized by plants via R proteins, induce a biphasic ROS production with a low-amplitude and transient first phase (ROS burst) followed by a sustained phase of much higher magnitude [17,18]. In abiotic stress, the biphasic increases of ROS production can also be observed, for instance, during ozone or wounding stress [18,19]. It has been demonstrated that, in particular tissues of Arabidopsis thaliana, cold, heat, high light, wounding, or salinity triggers AtRbohD-dependent ROS production. This local production of ROS can initiate rapid and long-distance signals traveled from the affected tissue to the entire plant, which has been shown to be accompanied by apoplastic ROS production and depends on AtRbohD [20]. Further research indicates a model that ROS induced ROS release (termed the ROS wave [21]), coupled with calcium signaling and other systemic signals (e.g. abscisic acid (ABA), hydraulic waves, and electric signals), plays an important role in abiotic stresses [22][23][24]. Furthermore, in response to Pseudomonas syringae pathovar tomato (Pst) DC3000, it was demonstrated that AtRbohD could also facilitate a long-distance signal for activation of defense response at distal sites [25].
The mitogen-activated protein kinase (MAPK or MPK) cascade is an intracellular pathway conserved among eukaryotes. A MAPK cascade is composed of three kinases, MAPK, MAPK kinase (MKK), and MAPKK kinases (MAPKKK or MEKK) [26]. The three kinases form the basic module of a MAPK cascade, MAPKKK-MKK-MAPK, in which MAPKKK phosphorylates and activates MKK, and the activated MKK phosphorylates and activates MAPK [27]. MAPK cascades are involved in a many aspects of plant physiology, including cell division [28], plant growth and development [29], plant resistance to pathogens [30] and insect herbivores [31], and plant response to abiotic stresses [32,33].
It has been observed that, on one hand, exogenous application of H 2 O 2 or ozone activates components of MAPK cascades, and initiation of ROS signaling can lead to changes in MAPK cascades (Table 1; [34,35]), on the other hand, manipulating MAPK cascades results in initiation of ROS responses (Table 2; [34,35]). These observations indicate that MAPK cascades are complicated in both ROS signaling and responses.

Activation and regulation of MAPK cascades by ROS
Experimental results indicate that O 2 deprivation leads to mitochondria-dependent ROS production in A. thaliana [36]. Under O 2 deprivation, as well as application of cell respiration inhibitors antimycin A (AA) or potassium cyanide (KCN), resulted in ROS production and transient activation of AtMPK6 and AtMPK3 [36]. The authors suggested that low O 2 stimulates mitochondrial ROS that moves to the cytoplasm to activate MAPKs, leading to retrograde signaling between mitochondria and nucleus. As the authors point out, however, the O 2 deprivation treatment also removes carbon dioxide. In addition, the inhibitors AA and KCN may affect photosynthesis. Therefore, it cannot be excluded that, in green leaves, signaling from the chloroplast may also participate in activating AtMPK6 and AtMPK3. Recently, it has been shown that, when A. thaliana plants shifted from low-light to high-light 8-800 μmol quanta m −2 s −1 ), AtMPK6 can be activated in retrograde signaling between chloroplast and nucleus. In this signaling, AtMPK6 regulates gene expression of several key transcriptional factors, e.g. ERF6 and ERF104, possibly prior to H 2 O 2 accumulation [37].
Exogenous application of H 2 [38], and AtMKK4-AtMPK6 and AtMKK3-AtMPK7 may be two pathways involved in H 2 O 2 responses [39]. H 2 O 2 also activates an oxidative signal inducible1 (OXI1) kinase that is required for full activation of AtMPK6 and AtMPK3 [40]. Exposure of acute ozone (high dose within a short time frame, e.g. 250-700 nl L -1 within 24 h) activates AtMPK3/ AtMPK6 in A. thaliana [41] and SIPK/WIPK in tobacco [42], which has been shown to be important for ozone sensitivity.
One way that changes in apoplastic redox status trigger intracellular signaling is likely by H 2 O 2 ·H 2 O 2 has the longest lifetime among the various ROS and can be transported across the plasma membrane [43]. Although H 2 O 2 could simply diffuse across the plasma membrane, it is suggested that H 2 O 2 may preferentially enter the cell by aquaporin channels [44][45][46]. This regulated entry may provide another way of adjusting local concentrations of H 2 O 2 . If H 2 O 2 can be channeled to an intended intracellular compartment, it is speculated that this could provide a layer of specificity in ROS signaling. Inside the cell, while it is possible that H 2 O 2 may initiate MAPK cascades directly, it is indicated that the prominent role of H 2 O 2 is to disturb ROS homeostasis via a redox relay mechanism [47]. As a result of these disturbances, intracellular levels of ROS may rapidly rise, transferring the oxidation to target MAPK cascades. Indeed, this mechanism has been demonstrated in mammalian cells·H 2 O 2 -induced oxidation of thioredoxin (Trx) leads to the conformational change of Trx and disruption of interaction between Trx and apoptosis signal-regulating kinase 1 (ASK1), resulting in ASK1 activation and subsequent phosphorylation of its substrate p38 MAPK [48]. In plants, activation of MAPKKKs, e.g. ANP1, OMTK, or MEKK1, by H 2 O 2 has been observed (Table 1). It remains to be seen if a redox sensor could activate the MAPKKK in a redox-dependent manner. In Nicotiana tabacum (tobacco), recent study showed that infiltration of either reduced glutathione (GSH) or the oxidized form (GSSG) could activate SIPK and WIPK [49]. Transgenic tobacco expressing the bifunctional glutathione biosynthetic enzyme from Streptococcus thermophilus (StGCL-GS) showed increased levels of glutathione accumulation and displayed activation of SIPK and to a lesser extent of WIPK [49]. These suggest that the oxidative shift in cytosolic redox potential could activate MAPK signaling (Fig. 1).
In addition, apoplastic ROS are involved in cell wall integrity (CWI) signaling [65]. Potential CWI sensors include the members of wallassociated kinases (WAKs) and Catharanthus roseus RLK1-like kinases (CrRLK1Ls). It is possible that apoplastic ROS may activate MAPK signaling by CWI sensors, e.g. in cell wall metabolism and development [65,66]. Apoplastic ROS may also directly change the activity of extracellular proteins by redox-dependent modifications [67]. In response to ozone, it is hypothesized that cysteine-rich receptor-like kinases (CRKs) are good candidates for relaying extracellular ROS signaling to intracellular MAPK activation [8] (Fig. 1).

Regulation of ROS-related genes by MAPK cascades
In A. thaliana, MEKK1-AtMKK1/AtMKK2-AtMPK4 cascade is involved in ROS homeostasis. Transcriptional analysis showed that some redox-related genes are misexpressed in mekk1 or atmpk4 mutant plants [62,63]. The microarray analysis concerning MEKK1-AtMKK1/AtMKK2-AtMPK4 cascade revealed that, on one hand, there is a great deal of overlap among ROS-related genes in mekk1, atmkk1/ atmkk2, and atmpk4 mutants. On the other hand, some ROS-related genes, e.g. CSD1 (encoding Cu/Zn-SOD1), CAT1, and APX1, are differentially expressed between mekk1, atmkk1/atmkk2, and atmpk4 mutants [68]. Among the genes encoding ROS scavenging enzymes that are crucial for redox homeostasis, CAT2 expression is downregulated in mekk1 and atmpk4 mutants, but unaffected in atmkk1/ atmkk2 double mutants. Furthermore, the steady-state transcripts of CAT1 and CAT3 are up-regulated in mekk1 mutants and in atmkk1/ atmkk2 double mutants, but not in atmpk4 mutants [68]. These results suggest that, in regulating ROS scavenging genes, MEKK1, AtMPK4, or AtMKK1/AtMKK2 may function independently of the MEKK1-AtMKK1/AtMKK2-AtMPK4 cascade in response to specific ROS signaling or in different cell-compartments. Furthermore, in atmpk4 mutants or atmpk4/ics1 double mutants, it was shown that most of the analyzed genes encoding ROS scavenging enzymes, e.g. GR2 (encoding glutathione reductase 2), FSDs (encoding Fe-SODs), and APXs, are down-regulated, whereas CSD2 can be up-regulated [69]. The enzyme analysis demonstrated Cu/Zn-SOD and APX activities were increased whereas Fe-SOD activity was decreased in atmpk4 mutants or atmpk4/ ics1 double mutants [69] (Fig. 2).
Previous studies indicated taht AtMKK1 induce CAT1 expression by triggering H 2 O 2 production in response to ABA, drought, and salt stress [70]. Further analysis showed that AtMKK1-AtMPK6 could regulate CAT1 expression in ABA-induced H 2 O 2 production [71]. Recently, Xing et al. suggested that AtMKK5-AtMPK6 regulate genes encoding Cu/Zn-SOD (CSD1 and CSD2) in high light ((800 mmolm-2 s −1 ; [14]) and genes encoding Fe-SODs (FSD2 and FSD3) in salt stress [15]. It is likely that the MAPK-regulated gene expression of ROS scavenging enzymes can be regulated by stress-induced ROS production. This regulation may be crucial for redox homeostasis and stress tolerance (Fig. 2).
Studies focusing on general stress response (GSR) in A. thaliana revealed that GSR could modulate stress-responsive genes partly by the cis-regulatory element CGCGTT known as rapid stress response element (RSRE; [72]). In mekk1-5 mutants, wounding induced an earlier peak time of RSRE monitored by luciferase in vivo when compared with the parent plants, suggesting that, genetically,  MEKK1 inhibits the rapid expression of stress-responsive genes possessing CGCGTT in wounding [73]. Pharmacologically, PD98059 (a broad-spectrum inhibitor of MKK) can inhibit the activity peak and can also delay the peak time of RSRE in response to flg22 (a 22-amino acid synthetic polypeptide that corresponds to a highly conserved epitope of the Pseudomonas PAMP flagellin), indicating MAPK cascades may regulate the rapid transcription of genes possessing CGCGTT regulatory elements during plant immune responses [73]. The reports also support the idea that RSRE activation is likely regulated by ROS, as (1) the earlier peak time of RSRE is consistent with the over-accumulation of ROS in mekk1-5 mutants; and (2) MVinduced 1 O 2 could activate RSRE [73]. In A. thaliana, several transcription factors have been identified that act downstream of MAPKs in ROS responses. These include the MYB-transcription factor 44 (MYB44; [74,75]), the heat shock transcription factor A4a (HSFA4A; [76]), and the ethylene response factor 6 (ERF6; [77,78]). Particularly, AtMPK6-ERF6 have been demonstrated to regulate ROS-responsive gene via GCC box (AGCCGCC) in response to biotic and abiotic stresses [77,78]. Interestingly, AtMPK6-ERF6 can also bind the GCC box to regulate jasmonic acid/ethylene (JA/ET)-responsive genes in response to necrotrophic fungal pathogen Botrytis cinerea [79][80][81][82]. It is unclear whether JA/ET and ROS signaling can partly become merged in regulating genes with GCC box in the promoters, e.g. in response to B. cinerea. In rice, saltresponsive ERF1 (SERF1) is a transcription factor. Expression of SERF1 gene can be induced by salt and H 2 O 2 treatment. Analysis of SERF1 loss-of-function plants showed that SERF1-dependent genes are H 2 O 2 responsive. Further analysis indicated that SERF1 is a substrate of MAPK5 (OsMPK6). SERF1 regulates H 2 O 2 -responsive genes, e.g. MAP3K6, MAPK5, DREB2A, ZFP179, and SERF1, by DREB-specific cis-element GCCGAC or ACCGAC [83] (Fig. 2).
Besides regulating H 2 O 2 -responsive genes, MAPK signaling has also been implicated in regulating genes encoding Rboh in ROS production. In wounding response, AtMPK8 negatively regulates AtRbohD expression. In atmpk8 mutants, AtRbohD expression was higher compared with wild type. Further analysis showed that AtMPK8 negatively regulates ROS accumulation via repressing AtRbohD expression [84]. In response to flg22, AtRbohD is up-regulated in ERF104-overexpressing plants and ERF104 has been identified to be a substrate of AtMPK6 [85]. In erf6 mutants, it seems that AtRbohD expression can be up-regulated [77]. However, although AtMPK6-ERF6 and AtMPK6-ERF104 have been demonstrated to regulate stress-responsive genes via GCC box, it is unlikely that they directly regulate AtRbohD expression via GCC box, since no GCC box was found in the upstream 2500 bp region of AtRbohD promoter. The precise role of MAPK signaling in regulating AtRbohD expression needs to be elucidated in future studies. In N. benthamiana, NbRbohB gene, the orthologue of Arabidopsis AtRbohD, was shown to be important for ROS production in immunity. It has been shown that NbRbohB expression was induced by MAPK signaling (possibly by NbMKK2-SIPK and NbMKK1-NTF6 cascades) in response to INF1, the elicitor produced by the oomycete Phytophthora species [86,87]. The recent discovery demonstrated that NbMKK2-mediated MAPK signaling regulates cis-element (TTTGGTCAAAC) of the NbRbohB promoter by phosphorylating WRKYs (e.g. WRKY7/WRKY8/WRKY9/WRKY11) in response to INF1 or R3a/AVR3a (the potato R protein/oomycete effector pair) but not flg22 [88] (Fig. 2). In addition, Maize MAPK5 (ZmMPK6-2) is one of the orthologues of Arabidopsis AtMPK6 [89,90]. Based on pharmacological evidence, it is proposed that the expression of ZmRboh genes could be regulated by MAPK5 in maize leaves in response to ABA, H 2 O 2 , and brassinosteroids [91,92].
The mechanisms by which MAPK signaling controls gene expression remain to be elucidated. Experimental data showed that components of MAPK cascade can be detected in the nucleus [93]. AtMPK4 [94], AtMPK6 [85], and SIPK [95] have been implicated in ROS signaling and responses. In addition, in the early stages of O 3 treatment, it is indicated that activated AtMPK3 and AtMPK6 are translocated to the nucleus where they are thought to phosphorylate substrates and regulate gene expression [41]. Furthermore, AtMPK3/ AtMPK6 can phosphorylate ERF6. The stabilization and accumulation of ERF6 were reported to initiate transcriptions of stress genes [82]. It is possible that the mechanisms underlying ROS-related gene expression by MAPK cascades might be different in response to different types of stimuli or depending on the physiological context. ROS may directly regulate transcription factors, e.g·H 2 O 2 can regulate transcription factors with redox-sensitive Cys residue [96]. The regulation of transcription factors by ROS may influence the conformational changes, nucleocytoplasmic shuttling, proteolysis, or interaction with other regulators [96]. Although it has not been demonstrated, it is possible that some MAPK substrates may be regulated by ROS directly.
In this scenario, ROS may lead to simultaneous multisite modifications of the same pathway, which may guarantee efficient or specificity of signaling transduction.

ROS and MAPK cascades in HR-like cell death
The incompatible plant-pathogen interaction can induce plant programmed cell death (PCD) at the infection sites known as hypersensitive response (HR; [97]). HR is often characterized by discrete cellular lesions and preceded by an ROS burst. HR and the accompanying ROS is thought to act against hemi-biotrophic and biotrophic pathogens that obtain their nutrition from living host cells. Initiation of HR may further lead to activation of systemic acquired resistance to avoid the death of the entire plant.
Previous studies revealed that transgenic A. thaliana with inducible AtMKK4 DD (the constitutively active mutants of AtMKK4 protein) or AtMKK5 DD exhibit cell death when AtMKK4 DD or AtMKK5 DD are induced. The cell death is similar when an HR is induced by infection by the bacterial pathogen Pst DC3000 [98]. The HR-like cell death is preceded by H 2 O 2 generation and AtMPK3/AtMPK6 activation. In tobacco leaves, transient expression and induction of AtMKK4 DD , AtMKK5 DD , or NtMKK2 DD also leads to PCD with prolonged activation of SIPK/WIPK, the orthologues of AtMPK6/AtMPK3. There are also reports that in N. tabacum and N. benthamiana, SIPK-mediated MAPK cascade regulate harpin-induced cell death [99] and INF1induced cell death [100]. Furthermore, Zhang et al., 2012a showed that many bacterial, fungal, and oomycete species secrete necrosis and ethylene-inducing peptide 1 (Nep1)-like proteins (NLP) that trigger PCD. Interestingly, the MAPKKKα-NbMKK2-WIPK cascade might regulate Nep1Mo-activated H 2 O 2 accumulation and contribute to Nep1Mo-triggered cell death, as virus-induced gene silencing WIPK but not SIPK affects Nep1Mo-induced H 2 O 2 accumulation [101].
The relationship between ROS and AtMPK6/AtMPK3 (SIPK/ WIPK) remains elusive. It is suggested that harpin-induced ROS accumulation is required for activation of SIPK and WIPK [99]. In another study, however, it was shown that activation of the SIPK/Ntf4/ WIPK cascade by pathogens Phytophthora cryptogea promotes the ROS generation in chloroplasts, which plays an important role in executing HR-like cell death [102]. Further studies indicated a rapid inhibition of photosynthetic CO 2 fixation after NtMKK2 DD -induced SIPK/Ntf4/WIPK activation, leading to the generation of ROS in chloroplasts under illumination [103]. In the dark, cell death is only delayed but not blocked [103]. Therefore, it seems that the chloroplast is not the sole organelle to execute HR-like cell death. It has been shown that SIPK/WIPK function downstream of HSP90 to transduce the cell death signal to mitochondria in TMV resistance gene Nmediated HR-like cell death [104].
L-myo-inositol 1-phosphate synthase 1 (MIPS1) is the rate-limiting enzyme in myo-inositol biosynthesis. Mutation in MIPS1 results in spontaneous cell death [111,112]. It is possible that MIPS1 controls its own transcription by the AtMPK4 pathway, as MIPS1 expression is down-regulated in atmpk4 mutant [113]. ROS have been involved in the spontaneous cell death in mips1 mutants. For instance, mips1 mutants show increased sensitivity to ROS stress and reduced ascorbic acid levels [112], and a decrease in myo-inositol accumulation is necessary for the spontaneous cell death in catalase-deficient cat2 mutants [114,115]. Furthermore, the cleavage and polyadenylation

Y. Liu, C. He
Redox Biology 11 (2017) 192-204 specificity factor 30 (CPSF30), encoding a polyadenylation factor subunit homolog, is required for spontaneous cell death in mips1 or atmpk4 mutants [116]. It was shown that CPSF30 is a general controller of the salicylic acid (SA) pathway rather than a direct regulator of ROS production or scavenging in the spontaneous cell death [116]. However, in the process of spontaneous cell death, CPSF30 may be connected with ROS signaling and responses by unknown mechanisms, as CPSF30 mutant is more tolerant than wild type in ROS stress [117].

ROS and MAPK cascades in acute ozone-induced PCD
Exposure to acute ozone can induce ROS burst leading to cell death that shares similarities with the HR [19,118,119]. Acute ozone induces MAPK cascades (Table 1). Genetic manipulation of MAPK cascades affects the cell death phenotype observed in O 3 responses ( Table 2). Mutations of radical-induced cell death1 (rcd1) show more ozonesensitive and prolonged activation of AtMPK3/AtMPK6 compared with the wild-type plants [120], indicating that prolonged activation of MAPK signaling may participate in the induction of cell death process in response to acute ozone.
It is possible that regulation of Rboh-dependent ROS production is the initial step in ozone signaling. It was reported that, in A. thaliana, exposure of acute ozone induced biphasic ROS generation that is associated with the pathogen response [19]. Recent reports showed that in the JA receptor mutant coi1-16, ozone-induced cell death was suppressed by impairment of AtRbohF [121]. This observation is reminiscent of reports showing that AtRbohD/AtRbohF-dependent ROS can suppress SA-dependent HR in cells around infection sites in response to Pst DC3000, with AtRbohD having a prominent role [122].

ROS and MAPK cascades in fumonisin B1-and sphingolipidinduced PCD
Fumonisin B1 (FB1) is a specific inhibitor of ceramide synthase. FB1 application can generate an accumulation of LCBs and it has been used to investigate the functions of sphingolipid metabolism in plants [123,124]. FB1-induced cell death in A. thaliana is associated with ROS production, callose deposition, and PR gene expression, which is similar to HR [125]. The mutant fumonisin B1 resistant11-1 (fbr11-1) is defective in a long-chain base1 subunit of serine palmitoyltransferase and is compromised in ROS generation and subsequent induction of PCD when challenged by FB1 [123]. Direct feeding experiments showed that dihydrosphingosine (d18:0), phytosphingosine (t18:0), and sphingosine (d18:1) can induce ROS generation followed by cell death, whereas the phosphorylated form of d18:0 (d18:0-P, dihydrosphingosine-1-phosphate) could block the ROS generation and PCD, indicating that homeostasis of free LCB and its phosphorylated form may determine the cell fate [123]. AtMPK6 plays a regulatory role in PCD induced by LCBs and FB1 [124]. In the atmpk6 seedling, exogenous application of d18:0 did not lead to severe PCD. The PCD phenotype result from the increase of endogenous LCBs induced by FB1 was attenuated in atmpk6 seedlings [124]. In another study, it is reported that t18:0-P rapidly and transiently formed in A. thaliana upon chilling (4°C). The activity of AtMPK6 (0-30 min) was stimulated by exogenous t18:0-P [126]. The authors showed that d18:0-P or t18:0 cannot induce AtMPK6 activity at 22°C or 4°C, in contrast with the observations that unphosphorylated LCB t18:0 can trigger AtMPK6 activation [124], possibly reflecting differences in the biological material used and in the treatment procedure. It is speculated that AtMPK6 acts downstream of homeostasis between free LCBs and the phosphorylated derivatives. Activation of AtMPK6 may lead to response to chilling stress in the presence of elevated t18:0-P. However, the prolonged activation of AtMPK6 may promote ROS accumulation and PCD in the presence of elevated d18:0.

ROS and MAPK cascades in self-incompatibilily-induced PCD
Self-incompatibility (SI) is a genetic mechanism used by many angiosperm species to prevent inbreeding. Studies in Papaver rhoeas showed that recognition of incompatible pollen can trigger signaling to initiate PCD in pollen tubes. This signaling includes activation of a 56 kDa MAPK [127,128]. The activation of 56 kDa MAPK was shown at 5 min, peaking at 10 min, and lasting up to 30 min after SI induction [127]. Because around 10 min after SI induction may play a pivotal role in determining irreversible PCD of incompatible pollen [127][128][129], it seems that a MAPK cascade is involved in initiating SI-induced PCD. Increases in ROS, however, can be detected at 2 min and can last up to 15 min, indicating that ROS signaling acts upstream of the 56 kDa MAPK activation [128,129]. The possible role of the 56 kDa MAPK cascade is to stimulate nitric oxide (NO) signaling to activate caspase-3-like activity, as increases in NO occurred at 15-40 min after SI induction and pharmacological application of U0126 (the specific MKK inhibitor) can prevent activation of caspase-3-like activity [128,129].

Rboh-dependent ROS production and MAPK cascades in PAMP-triggered immunity
Plants use two layers of immune systems to defend against pathogens. The first layer is based on perception of pathogen-associated molecular patterns (PAMPs) by pattern recognition receptors (PRRs) in the cell surface, leading to the PAMP triggered immunity (PTI). The second layer is based on recognition of microbial effectors secreted in cells by resistance (R) proteins, leading to the effector triggered immunity (ETI; [130]). In A. thaliana, perception of bacterial PAMPs flagellin and EF-Tu by the corresponding PRRs, flagellinsensing 2 (FLS2) and EF-Tu receptor (EFR), activates early common responses such as ROS production and MAPK activation ( [131] and references therein). In response to flg22 or Pst DC3000, the rapid and transient ROS production (about 0-30 min) is dependent on AtRbohD [132][133][134][135], one of the ten Rboh proteins in A. thaliana [9]. Quantitative phosphoproteomic analysis showed that phosphorylation of AtRbohD plays an important role in AtRbohD-dependent ROS production [132,136]. Recent reports demonstrated that the membrane-anchored Botrytis-induced kinase 1 (BIK1), a multifunctional receptor-like cytoplasmic kinase, directly regulates AtRbohD-dependent ROS production in response to flg22 or elf18 [137,138].
It seems that AtMPK3/AtMPK6 activation by elf18 is also dispensable for elf18-induced ROS production. Normal function of EFR requires endoplasmic reticulum (ER) quality control systems. Signaling defects in ER quality control of EFR biogenesis affect the full ROS production in response to elf18 [147][148][149][150]. In psl1-1 (encoding calreticulin3 in ER quality control signaling) mutants, elf18-induced ROS production (0-40 min) is retained at intermediate levels, whereas MAPK activation is much reduced [147,150]. In psl5-1 (encoding α-subunits of endoplasmic reticulum resident glucosidase II) mutants, elf18-induced ROS production (0-30 min) is strongly reduced below detectable levels, whereas an increase of MAPK activity is observed [148]. In the Wassilewskija (WS) ecotype A. thaliana, elf26triggered ROS production (30 min) is not reduced in both atmpk6 mutants and atmkp1/atmpk6 (WS) double mutants [53]. Recently, it has been shown that in anp2anp3 (WS) double mutants or β-estradiolinducible anp1anp2anp3 (WS) conditional triple mutants, elf18-induced production of ROS was compromised but activation of AtMPK3/ AtMPK6 was normal [151]. Collectively, these data demonstrate that, in the early stage of PTI (about the first 30 min), PAMP-activated AtMPK3/AtMPK6 may occur independent of the PAMP-triggered ROS production.
In PTI, AtMPK4 can be activated by flg22, and it is assumed that AtMPK4 pathway positively regulates PTI [106]. The role of AtMPK4 in regulating ROS production in PTI is unclear. Experimental results suggest that flg22-induced ROS production was enhanced in atmpk4 mutants [143], and A. thaliana expressing a constitutively active AtMPK4 showed a reduced ROS production in response to flg22 [152]. These results indicate that AtMPK4 negatively regulates ROS production in PTI. In another study, Kong et al. identified that the function of AtMPK4 is dependent on MEKK2. ROS production induced by flg22 was comparable in wild type and mekk2 mutants [107], indicating that disruption of AtMPK4-MEKK2 signaling pathway may not change AtRbohD-dependent ROS production in PTI. It is possible that mutations in the MEKK2 gene and disinhibition of AtMEKK2 activation (in atmpk4 mutants) can lead to different conclusions about the role of AtMPK4 in regulating ROS production in PTI.

Regulation of stomatal function by MAPK cascades in ROS signaling
MAPK cascades have been shown to play a role in stomatal function in response to H 2 O 2 . AtMPK3 antisense plants are less sensitive to exogenous H 2 O 2 , both in inhibition of stomatal opening and in promotion of stomatal closure, suggesting a positive role of AtMPK3 in H 2 O 2 signaling in stomata [153] (Fig. 4). AtMPK3 also regulates stomatal closure in response to Pst DC3000 [154] or Xanthomonas campestris [153,155], likely acting downstream of H 2 O 2 but indepen-  dent of ABA. AtMPK6 is a positive regulator of flg22-triggered stomatal closure [152], although it is not clear if this function is related to H 2 O 2 . In the atrbohD mutant, flg22-induced AtMPK3 and AtMPK6 activation is not affected [106,140], suggesting that AtMPK3 and AtMPK6 may not function downstream of AtRbohD-dependent ROS production in stomata. It has been shown that AtMPK6 is not involved in ABAmediated stomatal closure [154]. In ABA inhibition of stomatal opening, AtMPK6 seems to participate in ABA-triggered H 2 O 2 production downstream of AtMKK1 [71].
In N. tabacum, NtMPK4-silenced plants showed enhanced sensitivity to ozone with impaired regulation of stomatal closure [156]. In N. attenuate, silencing of NaMPK4 gene impaired stomatal closure induced by exogenous H 2 O 2 [157]. In seedlings of A. thaliana, AtMPK4 is predominantly expressed in stomata [27,158]. In immunity, AtMPK4 seems does not regulate stomatal closure in response to pathogens, as stomata in AtMPK4-overexpressing plants responded normally to Pst DC3000 or flg22 [152]. Loss-of-function of MEKK1 resulted in H 2 O 2 accumulation in stomata [62,63], suggesting a role of MEKK1 in regulating ROS homeostasis in stomata (Fig. 4).
It is reported that the AtMPK9 and AtMPK12 act redundantly as positive regulators of ROS-mediated stomatal closure in response to ABA [159], yeast elicitor [160], or methyl jasmonate [161]. However, AtMPK9 and AtMPK12 may act via a different mechanism in stomata. It has been shown that AtMPK12 but not AtMPK9 can interact with MAPK phosphatase IBR5 [162]. Recent research indicates that AtMPK9 can be activated by autophosphorylation that may be independent of any upstream MAPKKKs or MKKs [163]. Furthermore, AtMPK12 allele with an amino acid substitution alone could impair the ABA inhibition of opening [164].

Concluding remarks and perspectives
ROS production can be either harmful or beneficial to plants. Different  ) and subcellular production sites (plastidic, cytosolic, peroxisomal, or apoplastic) may determine the physiological, biochemical, and molecular responses. The concentration of ROS and its interaction with other signaling components may also determine the specificity of ROS response. The complex role of MAPK cascades in ROS signaling and responses has been revealed in regulation of ROS-related genes, executing plant cell death, and modulating stomatal function. How ROS activates MAPK cascades still remains unclear. It is possible that plants not only use MAPK cascades to transduce ROS signaling to gene expression and sometimes cell death, but also initiate the negative feedback regulation by MAPK cascades to maintain ROS homeostasis. The different combinations of the three tiers of kinases, distribution, timepointdependent activation, strength, duration, and availability of substrates of MAPK cascades may determine the feed-forward or feed-back outcomes (Fig. 5).
Accumulating evidence indicates that, in A. thaliana, AtMPK3/ AtMPK6 activation and AtRbohD-dependent ROS production are two independent pathways in the early stage of PTI (Fig. 3). Most recently, however, it was found that MKKK7, a MAPKKK, attenuates AtMPK6 activity and suppresses ROS production in response to flg22, through direct modulation of the FLS2 complex [165]. In response to wounding, AtMPK8 negatively regulates ROS production by repressing AtRbohD expression [84]. These results suggest a complicated relationship between MAPK signaling and AtRboh-dependent ROS production. Furthermore, both MAPK signaling and AtRboh-dependent ROS production have been implicated in O 3 [121]. Further studies are needed to unravel the whole picture about the role of MAPK signaling in Rboh-dependent apoplastic production of ROS.
A large body of research demonstrates that exogenous H 2 O 2 can activate MAPK signaling (Table 1). It seems that exogenous application of H 2 O 2 alone is not enough to induce the long-distance signaling manifested by AtRbohD-dependent ROS-triggered ROS waves [20,22,24]. Therefore, it is likely that activation of MAPK signaling by exogenous application of H 2 O 2 is a relatively localized event. The molecular mechanism by which apoplastic ROS activate intracellular MAPK pathway needs further to be investigated.
Whereas ROS have been implicated in PCD in plant developmental processes and stress responses, the sub-cellular location and the mechanism of ROS generation during PCD are not clear. Compartment-specific ROS signals and communication between compartments may determine the outcome of ROS responses [11]. For instance, disturbance of the ROS homeostasis in chloroplasts, mitochondria, or peroxisomes tends to switch signaling to result in PCD [103,104]. However, Rboh-dependent ROS appear to play a protection role in limiting the spread of PCD, e.g. in response to O 3 or Pst DC3000 [121,122]. It is speculated that low concentrations and restricted distribution of ROS mediates the role as signaling components to enhance cell survival and proliferation, whereas high concentrations of ROS (over a certain threshold) throughout the cell mediate the role as a death signal. MAPK cascades may act upstream of chloroplasts, mitochondria, or peroxisomes to regulate ROS homeostasis, or act downstream of these organelles to signal ROS response. In resistance to biotrophic or hemibiotrophic pathogens, MAPK cascades and ROS may elicit the escalated signaling at the interface of plant-pathogen interactions to induce HR that restricts pathogen access to water and nutrients. In response to necrotrophic pathogens or abiotic stresses, however, it is not yet known weather activation of MAPK cascades is simply to amplify the death signal. Furthermore, it remains to be seen how plants can discriminate between pathogens and beneficial microorganisms to initiate different MAPK activation and ROS responses.
PCD is a complex process that involves many signaling pathways. A strong interplay of ROS and MAPK cascades with other signaling molecules exists during plant PCD. It has been suggested that the HR is triggered only by balanced production of NO and ROS [166]. MAPK cascades are also involved in regulating NO signaling [87,167]. ROS and SA act synergistically to drive HR [168], and the phenotype of

Y. Liu, C. He
Redox Biology 11 (2017) 192-204 spontaneous cell death in defect of MEKK-AtMKK1/2-AtMPK4 cascade result from accumulation of both ROS and SA [105,158]. Ethylene signaling plays an important role in cell death that involves MAPK cascades [169][170][171]. Furthermore, crosstalk of ROS and MAPK cascades with calcium and lipid signaling fine-tunes the biological responses. The hierarchy of these events remains to be further elucidated. Although many questions still remain unanswered, further research will increase our understanding of the role of MAPK cascades in ROS signaling and responses.

Conflict of interest
No conflicts of interest to declare. . We would like to thank Professor Victor M. Darley-Usmar for providing helpful comments. We apologize to all investigators whose results could not be cited due to space limitations.