The Apoplastic Oxidative Burst Peroxidase in Arabidopsis Is a Major Component of Pattern-Triggered Immunity

In plants, reactive oxygen species (ROS) associated with the response to pathogen attack are generated by NADPH oxidases or apoplastic peroxidases. Antisense expression of a heterologous French bean ( Phaseolus vulgaris ) peroxidase ( FBP1 ) cDNA in Arabidopsis thaliana was previously shown to diminish the expression of two Arabidopsis peroxidases (peroxidase 33 [PRX33] and PRX34), block the oxidative burst in response to a fungal elicitor, and cause enhanced susceptibility to a broad range of fungal and bacterial pathogens. Here we show that mature leaves of T-DNA insertion lines with diminished expression of PRX33 and PRX34 exhibit reduced ROS and callose deposition in response to microbe-associated molecular patterns (MAMPs), including the synthetic peptides Flg22 and Elf26 corresponding to bacterial ﬂagellin and elongation factor Tu, respectively. PRX33 and PRX34 knockdown lines also exhibited diminished activation of Flg22-activated genes after Flg22 treatment. These MAMP-activated genes were also downregulated in unchallenged leaves of the peroxidase knockdown lines, suggesting that a low level of apoplastic ROS production may be required to preprime basal resistance. Finally, the PRX33 knockdown line is more susceptible to Pseudomonas syringae than wild-type plants. In aggregate, these data demonstrate that the peroxidase-dependent oxidative burst plays an important role in Arabidopsis basal resistance mediated by the recognition of MAMPs.


INTRODUCTION
Plants have evolved sophisticated surveillance systems triggered by recognition of microbe-associated molecular patterns (MAMPs), molecules such as eubacterial flagellin or peptidoglycan (PGN) or fungal chitin, which plants use to identify the presence of potentially harmful microbes. The plant defense response has now been dissected into several tiers of preexisting and/or activated defenses, many of which seem to involve the production of reactive oxygen species (ROS) (Chisholm et al., 2006;Jones and Dangl, 2006;He et al., 2007). The primary ROS studied to date include superoxide, hydrogen peroxide, and nitric oxide, which function either directly in the establishment of defense mechanisms or indirectly via synergistic interactions with other signaling molecules, such as salicylic acid (reviewed in Bolwell and Daudi, 2009). The major sources of ROS during most biotic stress responses seem to be plasma membrane-localized NADPH/NADH oxidases that generate superoxide, or cell walllocalized peroxidases that generate hydrogen peroxide, or both systems operating in tandem (Bolwell, 1999). To elucidate the underlying biochemical and physiological mechanisms involved in the plant defense response to pathogen attack, it is important to dissect the relative contributions of NADPH oxidases and peroxidases in the production of ROS in both basal resistance and in the so-called hypersensitive response (HR) Martinez et al., 1998;Bolwell, 1999; J.J. M. Grant et al., 2000;Torres et al., 2002;Bolwell and Daudi, 2009). By definition, basal resistance or pattern-triggered immunity (PTI) refers to defense responses activated by the perception of MAMPs. The class of responses encompassing gene-forgene resistance and the hypersensitive response, now collectively termed effector-triggered immunity (ETI), includes defenses activated by recognition of microbial virulence factors (effectors) that target PTI or other key host functions.
Well-studied MAMP elicitors of ROS production or the oxidative burst include the synthetic peptides Flg22 and Elf26, which correspond to conserved epitopes of bacterial flagellin and elongation factor Tu (EF-Tu), respectively. Hydrolytic enzymes, including chitinases, xylanases, or polygalacturonases, may also elicit ROS by generating chitin oligomers, b-1,3-glucans, or oligogalacturonides (OGs), respectively. Partially purified extracellular matrix preparations derived from fungal pathogens, such as Fusarium oxysporum, can also be used to elicit PTI. After interaction of MAMPs with the relevant host plant receptor, a series of well-defined and conserved events occurs in all plant species examined, among which an increase in cytosolic Ca 2+ , often measured within seconds of elicitation, is believed to be a primary signal essential for several downstream events (Chandra et al., 1997;M. Grant et al., 2000). In addition to this early calcium influx, a rapid efflux of K + and Cland extracellular alkalinization of cell cultures has been measured in response to MAMP elicitation (Fellbrich et al., 2000).
NADPH oxidases, also referred to as respiratory burst oxidases, have been implicated in biotic interactions, abiotic stress responses, and development in different plant species and have been studied in detail in Arabidopsis thaliana . Among the members of the 10-gene family of RBOH genes encoding homologs of the mammalian NADPH oxidase gp91 phox (Keller et al., 1998;Torres et al., 1998), RBOHD and RBOHF were found to be required for the production of a full oxidative burst in response to avirulent strains of the bacterial and oomycete pathogens Pseudomonas syringae and Hyaloperonospora arabidopsidis, respectively (Torres et al., 2002).
Although RBOHD and RBOHF seem to play crucial roles in Arabidopsis in the generation of ROS in response to pathogen attack, rbohD mutants do not appear to be more susceptible to a virulent strain of P. syringae, whereas rbohF mutants exhibit a modest level of enhanced susceptibility (Chaouch et al., 2011).  suggest that instead of generating ROS that act directly as antimicrobial agents or in the activation of defense response pathways, the major role of NADPH oxidases may be to limit the spread of salicylic acid-mediated cell death in cells surrounding an infection site . This conclusion was based on the observation that a lesion stimulating disease1 (lsd1)-rbohD-rbohF triple mutant exhibits spreading lesions. In the absence of the rbohD and rbohF mutations, lsd1 mutants form spontaneous lesions in the absence of pathogen attack, but these lesions remain localized.  also showed that NADPH oxidases need to be activated by an independent source of ROS to generate their own oxidative burst. It is thus possible that the ROS that activates RBOHD and RBOHF is generated by peroxidases during the basal (PTI) defense response, in which case NADPH oxidases may function mainly in the HR. On the other hand, a relatively limited role for NADPH oxidases in the HR has been observed in tobacco (Nicotiana tabacum), where rbohD-mediated hydrogen peroxide production does not seem to be essential for the development of the HR or systemic acquired resistance (SAR) (Rouet et al., 2006;Lherminier et al., 2009). SAR is a whole plant resistance response that is normally triggered by an HR and spreads systemically from a localized site of infection to distal parts of the plant, mediated by salicylic acid.
The apoplastic peroxidase-dependent oxidative burst was first proposed by Bach and coworkers (Bach et al., 1993), who showed that the cell wall is required for a full oxidative burst in carrot (Daucus carota) cultured cells. This mechanism has now been validated in French bean (Phaseolus vulgaris) Bolwell, 1999), Arabidopsis Davies et al., 2006), Capsicum annum (Choi et al., 2007), Lactuca sativa (Bestwick et al., 1998), and Gossypium hirsutum (Martinez et al., 1998). Although pharmacological inhibitor-based studies have implicated peroxidases as key components in ROS production in the defense response (J.J. Soylu et al., 2005), there has been relatively little investigation of peroxidase-dependent ROS production in Arabidopsis. We previously identified an Arabidopsis sodium azide-sensitive but diphenylene iodonium-insensitive apoplastic oxidative burst that generates H 2 O 2 in response to a Fusarium oxysporum cell wall preparation Davies et al., 2006). Inhibition of an oxidative burst by diphenylene iodonium with an I 50 of ;0.2 mM is indicative of the involvement of an NADPH/ NADH oxidase, whereas inhibition by sodium azide with an I 50 greater than 50 mM suggests a peroxidase-dependent mechanism Frahry and Schopfer, 1998).
Because the Arabidopsis genome encodes at least 73 class III peroxidases (Welinder et al., 2002;Oliva et al., 2009), which complicates genetic analysis, we constructed transgenic Arabidopsis plants expressing an antisense cDNA encoding a French bean class III peroxidase (FBP1) with the goal of knocking down expression of key Arabidopsis peroxidases involved in ROS generation during the defense response . Importantly, transgenic FBP1 plants exhibited an impaired oxidative burst and were more susceptible than wild-type plants to both fungal and bacterial pathogens. Genome-wide transcriptional profiling and RT-PCR analysis showed that the antisense FBP1 transgenic Arabidopsis plants had reduced levels of two class III peroxidase mRNAs corresponding to the Arabidopsis genes At 3g49120 (PCb; PRX34) and At 3g49110 (PCa; PRX33). These data suggested that peroxidases play a significant role in generating H 2 O 2 during the Arabidopsis defense response and in conferring resistance to a wide range of pathogens.
Because our previous work with the Arabidopsis antisense FBP1 transgenic plants showed that functional peroxidases were required for the production of ROS in response to a fungal cell wall elicitor, we hypothesized that peroxidases may catalyze ROS production during basal resistance triggered by recognition of MAMPs and that this initial oxidative burst may be essential for the activation of basal defenses associated with PTI. This might explain why the FBP1 antisense lines exhibited susceptibility to a broad range of pathogens. In this article, we further examine the role of apoplastic peroxidases in Arabidopsis PTI. We find that several independent lines with reduced expression of two key peroxidase genes, PRX33 or PRX34, exhibit many defects in PTI in mature plants. An rbohD mutant is also impaired in some PTIrelated responses, but to a lesser degree than the peroxidase knockdown lines.

Generation of Transgenic Arabidopsis Plants Expressing
Antisense FPB1 cDNA We previously described the generation of a transgenic Arabidopsis line expressing a heterologous FBP1 cDNA in an antisense orientation . In these previous experiments, it was very challenging to maintain the transgenic antisense FBP1 plants, because they succumbed to opportunistic infections. We were only able to recover a single viable line "asFBP1.1 H 4 ", which we refer to in this article as asFBP1.1. The transgenic line asFBP1.1 exhibited a reduced oxidative burst in response to a Fusarium cell wall elicitor preparation, reduced expression of two cell wall peroxidase-encoding genes (PRX33; At 3g49110 and PRX34; At 3g49120), and enhanced susceptibility to infection by bacterial and fungal pathogens . Besides its hypersusceptibility to infection, asFBP1.1 does not exhibit any obvious adverse phenotypes. Interestingly, however, the leaves of asFBP1.1 at all stages of development after the formation of the rosette are ;50% larger than ecotype Columbia (Col-0) wild-type leaves, and asFBP1.1 transgenic plants exhibit a delayed senescence phenotype of 7 to 10 d.
To overcome the difficulties of generating transgenic plants that constitutively express FPB1 antisense cDNA, we also constructed transgenic Col-0 Arabidopsis plants in which the antisense FBP1 cDNA was expressed under the control of a dexamethasone (dex)inducible promoter (see Methods). In contrast with asFBP1.1, the dex-inducible lines did not exhibit any discernable developmental differences compared with wild-type Col-0 plants, including leaf size and time to senescence. Although it proved difficult to work with these dex-inducible lines, because there was a significant amount of variation from one experiment to another in the levels of PRX33 and PRX34 mRNAs after dex treatment, in several independent dex-inducible lines, dex-treated FBP1 plants exhibited reduced levels of PRX33 and/or PRX34 mRNA compared with Col-0 plants (see Supplemental Figure 1A online) and enhanced susceptibility to P. syringae pv tomato strain DC3000 (see Supplemental Figure 1B online).
As described in Methods, PCR analysis of asFBP1.1 revealed that the FBP1 antisense cDNA construct is inserted into the second intron of gene At 5g01075 (on chromosome 5), which is annotated on The Arabidopsis Information Resource (www. Arabidopsis.org) as encoding a b-galactosidase. No evidence was found in the literature to suggest that At 5g01075 plays a role in pathogen defense signaling or basal resistance. Moreover, Genevestigator analysis (Hruz et al., 2008) revealed that At5g01075 did not show any significant differential expression in response to treatment by several common biotic stresses or MAMPs (see Supplemental Figure 2 online). Finally, SALK line_084805C, which contains a homozygous T-DNA insertion in At 5g01075 (the insertion site of the FBP1 cDNA in asFBP1.1), exhibited a wild-type phenotype with respect to ROS accumulation after treatment with the Fusarium cell wall preparation and did not exhibit enhanced susceptibility to P. syringae pv tomato strain DC3000. Thus, these data indicate that the enhanced disease susceptibility phenotype of asFBP1.1 is most likely not a consequence of the disruption of the gene into which the asFBP1 transgene is inserted or to other off-target effects.
Some T-DNA Insertion Lines That Affect the Expression of PRX33 and PRX34 Mimic the Phenotype of asFBP1.1 To confirm the conclusion that the immunocompromised phenotype of asFBP1.1 and the dex-inducible FBP1 antisense lines is a consequence of diminished expression of PRX33 and/or PRX34, we tested three T-DNA insertion lines (see Methods) in which the expression of PRX33 and/or PRX34 is affected. PRX33 and PRX34, which are contiguous genes, show nearly 95% homology at the protein level, but their promoter and intronic sequences are highly divergent (Valé rio et al., 2004). Quantitative RT-PCR (qRT-PCR) analysis showed that the basal level of PRX34 mRNA is 128-fold higher than the basal level of PRX33 mRNA in wild-type Col-0 plants. The prx33:T-DNA knockdown line (ecotype Wassilewskija [WS]) contains an insertion in intron 1 of At 3g49110 660 bp downstream of the ATG initiation codon and could potentially be a hypomorphic mutant. The prx34: T-DNA knockdown line (ecotype Col-0) contains an insertion in the promoter region of At 3g49120 200 bp upstream of the ATG initiation codon and is most likely a hypomorphic mutation. The prx33:T-DNA;prx34:RNA interference (RNAi) double knockdown line was obtained by transforming the prx33:T-DNA line with a prx34:RNAi construct (Passardi et al., 2006). Similarly to asFBP1.1, the prx33:T-DNA, prx34:T-DNA, and prx33:T-DNA; prx34:RNAi lines have larger leaves than Col-0 or WS (but smaller than asFBP1.1) and exhibit delayed senescence. Figure 1A shows that the prx34:T-DNA line expresses low basal levels of PRX34 mRNA but that PRX33 expression is not affected. However, the T-DNA insertion in PRX33 affects both PRX33 and PRX34 expression ( Figure 1A). The double knockdown line, prx33: T-DNA;prx34:RNAi, expresses low levels of both PRX33 and PRX34 mRNAs. All three of these PRX33 and/or PRX34 knockdown lines exhibited highly diminished 3,39-diaminobenzidine (DAB) staining after elicitation with a Fusarium oxysporum cell wall elicitor preparation (FoCWE) (Figures 2A and 2B; see Supplemental Figures 3 and 4 online). An image analysis-based scoring system using CellProfiler software (Carpenter et al. 2006) was devised to quantify the DAB staining data as shown in Figure  2. DAB staining is caused by an H 2 O 2 -dependent polymerization reaction that is catalyzed by peroxidases (Thordal-Christensen et al., 1997). However, it is unlikely that the diminished levels of DAB staining in the PRX33 and PRX34 knockdown lines are solely a consequence of diminished levels of total peroxidase activity rather than diminished ROS levels. The Arabidopsis genome encodes at least 73 class III peroxidases (Welinder et al., 2002;Oliva et al., 2009), and we showed previously in the asFBP1.1 line that soluble peroxidase activity is not diminished and that bound peroxidase activity is only diminished ;50% . Moreover, we also showed that ROS production in asFBP1.1 is diminished using a xylenol orange assay . Similarly, we have found a reduction in ROS production using a xylenol orange assay in transgenic tissue culture lines knocked down for PRX33 and PRX34 expression (J.A. O'Brien, unpublished data).
Similarly to asFPB1.1, the prx33:T-DNA and the prx33:T-DNA; prx34:RNAi lines were significantly more susceptible to P. syringae pv tomato strain DC3000 than the parent wild-type WS ecotype ( Figure 3B). By contrast, prx34:T-DNA plants were not significantly more susceptible to P. syringae pv tomato strain DC3000 than Col-0 plants ( Figure 3A), although in some experiments, they did exhibit a modest level of enhanced susceptibility.

Peroxidase Knockdown Lines Are Compromised in MAMP-Elicited Immune Responses
The observation that the oxidative burst elicited by the The WS ecotype, however, did respond to the other three well-characterized MAMPs (Elf26, PGN, and OG) that were tested as well as to FoCWE.
Col-0 wild-type leaves exhibited robust but different patterns of H 2 O 2 accumulation in response to treatment with the different elicitors tested ( In addition to the activation of an oxidative burst, another wellstudied MAMP-elicited response in Arabidopsis is the deposition of callose, a b-1,3-glucan polymer that is believed to strengthen and plug weak or compromised sections of plant cell walls. Its deposition at the site of pathogen attack has been extensively studied. FoCWE, Flg22, Elf26, PGN, and OG all elicited callose deposition to varying extents and in varying patterns in wild-type Col-0 leaves, and all of these MAMPs, with the exception of Flg22, also elicited callose deposition in wild-type WS leaves. By contrast, asFPB1.1 and the PRX33 and MAMP-elicited callose deposition has recently been shown to be dependent on the synthesis of indole glucosinolates (IGSs) (Clay et al., 2009). In turn, IGS synthesis requires the IGS biosynthetic enzymes CYP79B2 and CYP81F2 and the transcription factor MYB51. All three of the corresponding genes, CYP79B2, CYP81F2, and MYB51, are upregulated by MAMPs (Clay et al., 2009). The basal level of expression of these three mRNAs ( Figure 5A), as well as the Flg22-or Elf26-elicited levels ( Figure 5B), was significantly lower in asFBP1.1 and the PRX33 and PRX34 T-DNA lines. These data are consistent with the callose deposition data shown in Figure 4 and Supplemental Figures 6A, 7, and 8 online. A fourth MAMP-elicited gene Data represent the average 6 SD. All quantitative gene expression measurements were performed using technical and biological triplicates. The experiments were repeated at least two times with similar results. Values for the peroxidase T-DNA and antisense knockdown lines were found to be significantly different from rbohD with a P value < 0.05 as determined by Student's t test (A). Values for the peroxidase T-DNA and antisense knockdown lines treated with Flg22 were found to be significantly different from Col-0, WS, and rbohD with a P value < 0.05 as determined by Student's t test (B). prx34, prx34: prx33,prx34:RNAi. (CYP71A12) that is involved in the biosynthesis of camalexin was also expressed at lower levels after Flg22 or Elf26 treatment ( Figure 5B) in the PRX33 and PRX34 knockdown lines. Finally, as shown in Figure 1B, the transcription of the PRX33 and PRX34 genes themselves is activated by Flg22 or Elf26, but this activation is abrogated in the PRX33 and PRX34 knockdown lines. After Flg22 treatment, the level of PRX34 mRNA was approximately 32-fold higher than PRX33 mRNA in wild-type Col-0 plants.
Finally, we tested whether MAMP-elicited responses are completely abrogated in the PRX33 and PRX34 knockdown lines or whether these lines retain some residual ability to respond to MAMPs. For example, we found that whereas wildtype leaves responded to infiltration of 10 nM of Flg22 with callose deposition, no callose deposition was observed in asFBP1.1 or prx34:T-DNA leaves in response to either 10 nM or 100 nM of Flg22 (see Supplemental Figure 6B online). However, after infiltration of 1 mM of Flg22, callose deposition was observed in both of the peroxidase knockdown lines as well as in the wild type (see Supplemental Figure 6B online). These data show that asFBP1.1 and prx34:T-DNA leaves can respond to Flg22, but only at Flg22 concentrations ;100-fold higher than those required to elicit a response in wild-type plants.

H 2 O 2 Rescues the Callose Deposition-Deficient Phenotype of Peroxidase Knockdown Lines
To determine whether the abrogated responses to MAMPs in the peroxidase knockdown lines are a direct consequence of the lack of production of hydrogen peroxide, we attempted to rescue the callose deposition-deficient phenotype of the peroxidase knockdown lines after Flg22 treatment by adding exogenous H 2 O 2 . Indeed, H 2 O 2 or Flg22 treatment in the absence of the other did not elicit any detectable callose in asFBP1.1 or prx34: T-DNA leaves, but cotreatment with H 2 O 2 and Flg22 together resulted in callose deposition levels similar to those observed in wild-type plants treated only with Flg22 ( Figure 6). The concentrations of H 2 O 2 (5 or 15 mM) used in these experiments seem to be physiologically relevant based on the levels of H 2 O 2 measured in Arabidopsis plant tissue challenged with P. syringae (M. . These data indicate that the lack of H 2 O 2 per se in the PRX33 and PRX34 knockdown lines results in lack of callose deposition and suggest that a lack of H 2 O 2 also causes the lack of DAB staining and defense gene expression in response to MAMPs as observed in Figures 2 and 5, respectively. Hydrogen peroxide production detected by DAB staining was quantified as the percentage and intensity of leaf area stained. Approximately 0.1 mL of 0.5 mM Flg22, 0.5 mM Elf26, 100 mg/mL FoCWE, 100 mg/mL OG, or 100 mg/mL PGN were infiltrated into mature rosette leaves of 23-to 25-dold Arabidopsis plants. The infiltrated leaves were detached at 2 h after MAMP treatment and then stained with DAB as described in Methods. At least six independent plants were used as biological replicates, and three rosette leaves were sampled from each plant. The experiment was repeated at least two times. Representative leaves for each condition were selected for Supplemental Figures 3 to 5 online, and these leaves were used for the quantitative analysis shown in this figure. A combination of tools from Adobe Photoshop and Cell Profiler (Carpenter et al. 2006) was used to establish the threshold of DAB staining in the leaves and distinguish the staining from the background. The final measurement used to quantify the DAB staining was the area of the stain divided by the total area of the selected representative leaf. Col-0, prx34, asFBP1.1, and rbohD (A); WS, prx33 and prx33-34-(B); Col-0, WS,asFBP1.1,rbohD,prx33,prx34,

Comparative Roles of RBOHD and Peroxidases in MAMP-Mediated Responses
To determine the relative roles of peroxidases and NADPH oxidases in MAMP-mediated responses, we compared an rbohD T-DNA insertion mutant with asFBP1.1 and the PRX33 and PRX34 T-DNA lines. Similar to previously reported results (Torres et al., 2002), MAMP-elicited DAB staining was absent in the rbohD mutant, similar to the results obtained with the peroxidase T-DNA and antisense knockdown lines (Figure 2A; see Supplemental Figures 3 and 5 online). Unexpectedly, we found that the rbohD mutant was consistently modestly more susceptible to P. syringae pv tomato strain DC3000 than wild-type plants, although not as susceptible as asFBP1.1 ( Figure 3A). Although Chaouch et al. (2011) did not observe enhanced susceptibility of an rbohD mutant to Pto DC3000, because the rbohD mutant was only modestly more susceptible in our experiments, the conflicting results are probably due to differences in growth conditions. MAMP-elicited callose deposition was also modestly diminished As described in Methods, ?0.1 mL of a P. syringae tomato strain DC3000 suspension at OD600 = 0.0005 was infiltrated into mature rosette leaves of 23-to 25-d-old Arabidopsis plants, leaves were harvested at 2, 24, 48, and 72 h after infiltration, and bacterial counts were determined. Data represent the average 6 sd. Values for all four lines at 72 h after infiltration were found to be significantly different from each other except Col-0 and prx34:T-DNA with a P value < 0.05 as determined by Student's t test (A). Values for prx33:T-DNA and prx33:T-DNA;prx34:RNAi at 72 h after infiltration were found to be significantly different from WS, with a P value < 0.05 as determined by Student's t test (B). At least four biological replicates from independent plants were used for each time point, and the entire experiment was repeated with similar results. prx34, prx34: T-DNA; prx33, prx33:T-DNA; prx33-34-, prx33:T-DNA;prx34:RNAi. cfu, colony-forming units. in the rbohD mutant, especially following elicitation with the FoCWE (Figure 4; see Supplemental Figures 7 and 8 online). Finally, the rbohD mutant was also modestly compromised for the Flg22-mediated activation of MYB51, CYP79B2, or CYP81F2, which are involved in callose deposition ( Figure 5B). The expression of PRX33 and PRX34 was not abrogated in the rbohD mutant before or after Flg22 treatment ( Figure 1A).

DISCUSSION
In this article, we provide several lines of evidence that show that Arabidopsis cell wall peroxidases encoded by PRX33 and PRX34 play important roles in PTI elicited in response to a variety of MAMPs. A transgenic line constitutively expressing an antisense The observation that the prx34:T-DNA line did not consistently exhibit enhanced susceptibility to P. syringae pv tomato strain DC3000, whereas the asFBP1.1, prx33:T-DNA, and prx33:T-DNA; prx34-RNAi lines were consistently susceptible in many experiments, is in agreement with the data in Figure 1A, which shows that the T-DNA insertion in PRX33 affected basal expression of both PRX33 and PRX34, whereas PRX33 is expressed normally in the prx34:T-DNA line. Indeed, it is likely than none of the PRX33/ PRX34 knockdown lines, including asFBP1.1, is completely deficient in peroxidase activity. Figure 1 shows that there were measurable levels of PRX33 and PRX34 mRNA in all of the PRX33/PRX34 knockdown lines. Moreover, the T-DNA insertion in the prx34:T-DNA line is located upstream of the PRX34 coding region. Similarly, the T-DNA insertion in prx33:T-DNA is in an intron, and it is likely that the mutated gene can be translated to yield at least a low level of active peroxidase. These considerations and the fact that the Arabidopsis genome encodes at least 73 class III peroxidases (Welinder et al., 2002;Oliva et al., 2009) make it difficult to estimate whether there is any residual activity encoded by PRX33 and PRX34 in the mutant lines and how much of the total peroxidase activity is accounted for by PRX33 and PRX34.
The double prx33:T-DNA;prx34:RNAi knockdown line exhibited the lowest levels of PRX33 and PRX34 mRNA (Figure 1). This line is completely viable and does not exhibit any obvious growth or developmental defects, with the exception of larger leaves and delayed senescence. Because this double prx33:T-DNA;prx34: RNAi knockdown line does not seem to have as severe a phenotype as asFBP1.1, even though it has lower levels of PRX33 and PRX34 mRNA, it suggests that in asFBP1.1 in addition to PRX33 and PRX34, peroxidases that may have redundant functions are also knocked down.

Why Are the Peroxidase Knockdown Plants More Susceptible to Pathogens?
Because the peroxidase knockdown lines exhibit reduced ROS generation and callose deposition as well as aberrant activation (A) Basal levels of CYP71A12, CYP79B2, CYP81F2, and MYB51 mRNAs were measured in 23-to 25-d-old plants as described.
(B) Levels of CYP71A12, CYP79B2, CYP81F2, and MYB51 mRNAs were measured at 2 h after infiltration of 0.1 mL of 0.5 mM Flg22 or Elf26. Data represent the average 6 SD. All quantitative gene expression measurements were performed using technical and biological triplicates. The experiments were repeated at least two times with similar results. Values for CYP71A12 expression in rbohD and prx33 were found to be significantly different from all other expression values, with a P value < 0.05 as determined by Student's t test (A). Values for the gene expression levels of each of the four genes in Col-0 and WS, and all genes except CYP79B2 in rbohD, were found to be significantly different from the gene expression levels of all four genes in the peroxidase knockdown lines, with a P value < 0.05 as determined by Student's t test (B). prx34, prx34: prx33,prx34:RNAi. of MAMP-elicited transcription of defense-related genes, the simplest explanation for why these plants are more susceptible to P. syringae is that they are impaired in PTI. An alternative explanation, which is not mutually exclusive with a defect in PTI, is that PRX33 and PRX34 play a role in the cross-linking of cell wall or defense-related polymers. Interestingly, the peroxidasedownregulated lines exhibit larger leaves than wild-type plants, indicating that peroxidases also play a role in leaf expansion. In roots, PRX34 has been associated with cessation of elongation (Dunand et al., 2007). The role of peroxidases in this case is peroxidative cross-linking of cell wall polysaccharides, particularly pectin. By extrapolation, the leaf cell walls in the peroxidase knockdown plants may be less cross-linked. In response to pathogen challenge, a second type of peroxidative cross-linking occurs in papillae, targeted at Hyp-rich glycoproteins and Prorich proteins (Bradley et al., 1992;Wojtaszek et al., 1995;Brown et al., 1998). A decrease in either type of cross-linking in the peroxidase knockdown lines may contribute to enhanced susceptibility to pathogen attack.

Relative Roles of PRX33 and PRX34
It is clear from the data presented here and in Bindschedler et al. (2006) that PRX33 and PRX34 play important roles in Arabidopsis PTI responses, but the respective roles of PRX33 and PRX34 are not completely clear. The FBP1 antisense line asFBP1.1 exhibits decreased levels of both basal and Flg22-elicited PRX33 and PRX34 mRNA. Interestingly, both the prx33:T-DNA and prx34: T-DNA lines exhibit reduced levels of both PRX33 and PRX34 mRNA after Flg22-elicitation, and the basal mRNA levels of both genes are lower in the prx33:T-DNA line. The PRX33 and PRX34 genes (At 3g49110 and At 3g49120) are adjacent in the Arabidopsis genome, but why the T-DNA insertion in At 3g49110 affects expression of At 3g49120 and vice versa under some conditions is not clear. In any case, because the expression of both peroxidase genes is affected in all the lines used in our experiments after Flg22 treatment, and because they are expressed at different levels, it is difficult to precisely sort out the relative roles of these two peroxidases.

Relative Roles of Peroxidases and NADPH Oxidases in Pattern-Triggered Immunity
Previous work (Bolwell, 1999;J.J. Grant et al., 2000;Torres et al., 2002) and the data shown here in Figure 2 and Supplemental Figures 3 to 5 online indicate that both plasma membranelocalized NADPH oxidases and cell wall peroxidases are required for a MAMP-elicited oxidative burst in mature plants. This lack of ROS in the early stages of exposure to the various MAMPs was accompanied by a lack of callose deposition after 18 h in mature leaves in the various lines with reduced levels of PRX33 and PRX34, whereas MAMP-elicited callose deposition seemed to be at least somewhat independent of the RBOHD NADPH oxidase-mediated oxidative burst (Figure 4; see Supplemental Approximately 0.1 mL of 0.5 mM Flg22 was infiltrated into mature rosette leaves of 23-to 25-d-old Arabidopsis plants. At least six independent plants were used as biological replicates, and three rosette leaves were sampled from each plant. The experiment was repeated at least two times. Representative leaves are shown. The infiltrated leaves were detached at 2 h after MAMP treatment and then placed in 12-well sterile plates overnight in 1 mL solution containing either 5 mM, 15 mM H 2 O 2 , or sterile dH 2 O. The leaves were then harvested and stained for callose as described in Methods. Hydrogen peroxide was applied in solution at the time of Flg22 treatment. prx34, prx34:T-DNA. Bar = 20 mM. Figures 7 and 8 online). Similarly, MAMP-elicited expression of selected defense-related genes also seemed to be more dependent on PRX33 and PRX34 than on RBOHD ( Figure 5). Finally, lines in which both PRX33 and PRX34 are substantially knocked down are more susceptible to P. syringae than an rbohD mutant (Figure 3) or an rbohF mutant (Chaouch et al., 2011). These data suggest that the oxidative bursts generated by peroxidases and NADPH oxidases are not functionally equivalent and that peroxidases and NADPH oxidases play distinct roles in the plant defense response. ROS inhibitors tested on transgenic Arabidopsis cell culture lines in which the expression of PRX33 and PRX34 is downregulated by antisense expression of FBP1 cDNA show that peroxidases generate about one-half of the H 2 O 2 produced in response to MAMP treatment, whereas NADPH oxidases and other sources, such as mitochondria, account for the remaining ROS production (J.A. O'Brien, unpublished data).
One important caveat with our conclusions concerning the relative roles of apoplastic peroxidizes and NADPH oxidases is that our experiments only investigated the role of RBOHD in PTI, whereas previous work has suggested that RBOHD and RBOHF may have redundant roles in plant defense. An rbohD;rbohF double mutant, but not rbohD or rbohF single mutants, failed to exhibit an oxidative burst when infected with P. syringae or H. arabidopsidis (Torres et al., 2002). Unfortunately, under our growth conditions, the rbohD;rbohF double mutant exhibited a dramatic dwarf phenotype and could not be used in the experiments reported here. In addition, we attempted to cross asFBP1.1 with rbohD and tested several recombinants but found that, in all cases, the expression of PRX33 and PRX34 was no longer reduced. Further experimentation will be required to fully sort out the relatively roles of peroxidases and NADPH oxidases in the plant defense response.

Role of Hydrogen Peroxide in Callose Deposition
In French bean, callose accumulation is accompanied by production of H 2 O 2 by the FBP1 peroxidase at sites of bacterial infection . Here we show in Figures 4 and 6 and Supplemental Figures 6 to 8 online that callose formation is most likely dependent on the production of hydrogen peroxide generated by a peroxidase-mediated apoplastic oxidative burst. Importantly, the lack of callose deposition in the peroxidase knockdown lines asFBP1.1 and prx34:T-DNA could be rescued by the exogenous administration of physiologically relevant H 2 O 2 concentrations. Consistent with the lack of callose deposition, the mRNA levels of several genes previously shown to be required for callose deposition, including the transcription factor gene MYB51 and two cytochrome P450-encoding genes required for the synthesis of methoxylated IGSs (Clay et al., 2009;Millet et al., 2010), are downregulated in the peroxidase knockdown lines.

Peroxidase-Dependent Gene Expression and Prepriming of the Basal Defense Response
Initial ROS production, because of its rapid production, is likely performed by existing enzymes and does not require new protein synthesis. Nevertheless, there is a significant increase in the levels of PRX33 and PRX34 transcription after Flg22 elicitation, which is not significantly affected in an rbohD mutant ( Figure 1B). These data suggest that PRX33 and PRX34 transcription may be autoregulated by the ROS generated by PRX33 and PRX34. The basal levels of several well-studied MAMP-activated genes are also downregulated in unchallenged leaves of the prx33 and prx34 knockdown lines. Similar downregulation of MAMP-activated genes is also observed in transgenic Arabidopsis cell culture lines that are downregulated in the expression of PRX33 and PRX34 (J. A. O'Brien, unpublished data). This study also shows that several MAMP-activated proteins are depleted in CaCl 2 extracts of the PRX33/PRX34 downregulated cell culture lines. These data suggest that a low level of apoplastic ROS production may be required to preprime basal resistance. These class III peroxidases are unique, because they can generate H 2 O 2 at normal physiological pH 7, whereas several other heme proteins, including catalase, require pH greater than 8 to generate H 2 O 2 (Bolwell and Daudi, 2009).

Conclusions
We previously speculated that a peroxidase-generated oxidative burst may play an essential role in PTI as well as playing a role in activating NADPH oxidase during an effector-triggered HR . The data presented in this article support this conclusion, showing that peroxidase knockdown lines are impaired in some key aspects of PTI. Overall, we have shown that the peroxidase-dependent oxidative burst generates H 2 O 2 , which is required for wild-type levels of several PTIassociated responses, including callose deposition and the transcriptional activation of several PTI-associated genes (MYB51, CYP79B2, CYP81F2) required for callose deposition. ROS generated by RBOHD is also required for some PTI responses, including an oxidative burst, but interestingly does not seem to be absolutely required for callose deposition. The latter result is consistent with the observation that the rbohD mutant is not compromised for Flg22-mediated activation of MYB51, CYP79B2, and CYP81F2. However, an important caveat is that ROS generated by either RBOHD or RBOHF may be required for callose deposition. Somewhat unexpectedly, we found that an rbohD mutant is more susceptible to P. syringae than wild-type plants, at least under the particular growth conditions used in our experiments. Because the occurrence of a peroxidase-dependent oxidative burst is a widespread mechanism that has been demonstrated in many genera, our data suggest a general role for the apoplastic peroxidase-dependent oxidative burst in the plant defense response to pathogen attack.

Plant Material and Growth Conditions
The transgenic Arabidopsis thaliana line asFBP1.1 (H 4 ) expressing a French bean (Phaseolus vulgaris) peroxidase in antisense orientation was described previously . Arabidopsis T-DNA insertion lines in ecotypes WS and Col-0 exhibiting diminished expression of At 3g49110 (prx33; SALK_CS10885) or At 3g49120 (prx34; SALK_051769), respectively, or ecotype WS with diminished expression of both At 3g49110 and At 3g49120 (line 4.1; prx33-34-), were kindly provided by C. Penel at University of Geneva, Switzerland (Passardi et al., 2006). A knockdown line altered in At 5g47910 (rbohD) was provided by J. Dangl at University of North Carolina at Chapel Hill (Torres et. al., 2002). A T-DNA insertion line corresponding to At 5g01075 (SALK_084805C) was obtained from the Nottingham Arabidopsis Stock Centre. Arabidopsis ecotypes Col-0 or WS were used as wild-type controls and were germinated on Murashige and Skoog salts with vitamins (MS) solid media (Duchefa) after 24 h of stratification at 48C. Arabidopsis transgenic seeds were germinated on MS solid media containing 50 mg/mL kanamycin. Twelve-d-old seedlings were transferred to autoclaved soil and grown in climate-controlled chambers under long day conditions (16 h light/8 h dark cycle) at 238C.

Microbial Elicitors
The MAMPs or microbial elicitors used in this study were all prepared fresh on the day of use from stock solutions stored at 2808C. The microbial elicitors included the synthetic peptides Flg22 (Felix et al., 1999) and Elf26 (Kunze et al., 2004), which were synthesized by the Massachusetts General Hospital Peptide Synthesis and Sequencing Core Facility, Boston, Massachusetts, and used at concentrations of 0.5 mM unless otherwise indicated. A FoCWE was prepared as reported previously  and used at a concentration of 100 mg/mL. OGs and PGN were also used as microbial elicitors, both at concentrations of 100 mg/mL diluted in sterile deionized water. OGs were prepared as described in Ferrari et al. (2007). PGN, derived from Staphylococcus aureus, was acquired from Sigma-Aldrich. Diluted microbial elicitor solutions were stored on ice until required, but were brought to room temperature prior to inoculation to avoid cold stress in leaves. Leaves were also sprayed with water 30 min prior to inoculation to ease inoculum uptake and to reduce physical damage.

Dex-Inducible Constructs
Dex-inducible FBP1 lines were generated by cloning a FBP1  into pTA7001, a binary vector with a dex-inducible promoter system (McNellis et al., 1998). pTA7001 was obtained from N.-H. Chua at Rockefeller University, New York. Primers used to subclone the FBP1 cDNA were: 59-ACTAGTGTTGTGGGGGTTGTGCTT-39 (forward primer containing a SpeI site) and 59-CTCGAGTAACATAGGAAACAACATCAT-GAAAA-39 (reverse primer containing a XhoI site). After PCR of FBP1 cDNA using the original construct from Bindschedler et al. (2006) as the template, the resulting PCR product was cloned into pCR2.1 TOPO TA. After verifying the presence of the insert and confirming the sequence, the FBP1 cDNA was excised and cloned into pTA7001 by digestion and ligation of the XhoI and SpeI restriction sites. pTA7001-FBP1 DNA was transformed into Agrobacterium tumefaciens GV3101, and GV3101 (pTA7001-FBP1) was used to transform actively budding bolts of Col-0 plants as described elsewhere (Clough and Bent, 1998). F 1 transformants were selected from the resulting seed by germinating on MS media plates with 0.8% phytagar and 20 mg/mL hygromycin. F 2 progeny were collected from several F 1 hygromycin-resistant (hyg r ) individuals and screened for 3:1 segregation of hyg r . In turn, F 3 progeny seed was collected from ;10 F 2 individuals from each line and tested for 100% hyg r to identify F 2 individuals that were homozygous for the transgene cassette. A total of five dex-inducible FBP1 lines, selected from dex-asFBP1i.1 to dex-asFBP1i.12, were examined for a dex-inducible decrease in FBP1 mRNA and for dex-inducible susceptibility to Pseudomonas syringae pv tomato strain DC3000.

Dex Treatments
An aqueous 10 mM solution of dex (Sigma-Aldrich) was sprayed onto mature 4-week-old plants. Control plants were sprayed with sterile dH 2 O, and all plants were then kept under separate domes at high humidity for the duration of the experiment. For bacterial infections, dex was sprayed onto the plants 3 h prior to bacterial infiltration.
Mapping the FBP1 Insertion Site in asFBP1.1 Mapping of the FBP1.1 insertion site was performed by PCR-based genome walking adapting a previously described protocol (Siebert et al., 1995). Briefly, total Arabidopsis leaf genomic DNA (gDNA) was extracted using the cetyltrimethylammonium bromide method (Ausubel et al., 1994), and 2.5 mg of the extracted gDNA was digested overnight with EcoRV, ScaI, or HincIII. The digested gDNA fragments were purified by phenol:chloroform and resuspended in 20 mL of dH 2 O. An adaptor was designed to ligate to the gDNA fragment ends to create fragments that would serve as templates for nested PCR. The sequence of the long adaptor was 59-GTAATACGACTCACTATAGGGCACGCGTGGTCGACG-GCCCGGGCTGGT-39, and the sequence of the short adaptor was 59-PO 4 -ACCAGCCC-H 2 N-39. The adaptor was incubated for 2 min at 1008C and ligated to 5 mL of the digested gDNA with concentrated T4 DNA ligase (Biolabs). The reaction was incubated overnight at 168C. After ligase inactivation, the sample was diluted 10 times in TE buffer (10 mM Tris pH 7.5, 1 mM EDTA) to a final volume of 100 mL.
Next, nested PCR was performed to amplify FBP1-specific fragments that would be present on the adaptor-ligated gDNA fragments after specific restriction digestion dependent on restriction sites in the FBP1 sequence. For PCR amplification of specific sequences in the genome, four sets of primers were used: adaptor primer 1, nested adaptor primer 2, gene-specific primer 1, and gene-specific primer 2, described in Supplemental Table 1 online. The amplification of the sequence of interest was performed by two consecutive nested PCR reactions. For the first reaction, 1 mL of the diluted gDNA adapter sample was used with the adaptor primer 1/gene-specific primer 1 combination using the following thermal cycler program: seven cycles of 25 s at 958C followed by 3 min at 728C, then 32 cycles of 25 s at 958C and 3 min at 678C, and finally one cycle of 7 min at 728C. The first PCR reaction was then diluted 1:50 and used for the nested reaction with the nested adaptor primer 2/genespecific primer 2 combination using the following thermal cycler program: five cycles of 25 s at 958C followed by 3 min at 728C, then 20 cycles of 25 s at 958C and 3 min at 678C, and finally one cycle of 7 min at 728C for DNA strand extension. The products of the nested PCR were analyzed using 1% agarose gel electrophoresis and then purified from the gel using the QIAquick Gel Extraction Kit (Qiagen) and sequenced (Eurofins MWG Operon).

RT-PCR and Real-Time PCR
RT-PCR was performed using a Primus 25 thermocycler (Peqlab), and qRT-PCR was performed using a Corbett Rotor-Gene (Corbett) or a Rotor-Gene Q (Qiagen). The final concentration of the RT-PCR reaction was: 0.20 mM primers, 0.2 mM deoxynucleotide triphosphate, 1.5 mM MgCl 2 , and 2 mL 10 ng/mL cDNA in 53 reaction buffer together with 0.625 units of DNA Taq polymerase (GoTaq; Promega) in a total reaction volume of 50 mL. The reaction commenced at 948C for 2 min, followed by 25 to 35 cycles of 40 s at 948C, 40 s at 558C, and 40 s at 708C. The cycling reaction was completed with a final extension for 7 min at 708C. Primer sequences used for the reactions are shown in Supplemental Table 1 online. Primers were designed using Primer 3 (www.wi.frodo.mit.edu) with gene sequences obtained from The Arabidopsis Information Resource (www. Arabidopsis.org).
For qRT-PCR, primers were designed to generate product sizes of 60 to 100 bp spanning the first intron of each target gene. Stretches of nucleotide G or C were limited to less than three where possible, and two of the last five bases on the 39 end were selected as either G or C. Primer efficiency was tested with 0.25 mM primers, 2 mL of 10 ng/mL cDNA, and 10 mL iQ SYBR-Green Supermix (Bio-Rad) or SYBR Green JumpStart Taq ReadyMix (Sigma-Aldrich) in a total reaction volume of 20 mL. The reaction commenced at 958C for 3 min, followed by a cycling stage of 10 s at 958C, 30 s at 558C, and 30 s at 728C. qRT-PCR was performed similarly using GAPDH, TUB9, and EIF4A primers as the housekeeping reference genes. The PCR reaction set up was automated to reduce handling errors using a QIAgility PCR automation robot (Qiagen).

Assay for Callose Formation
In situ detection of callose was performed using aniline blue staining adapted from methods used previously (Jacobs et al., 2003;Clay et al., 2009). Briefly, three rosette leaves from 23-to 25-d-old Arabidopsis plants were syringe-infiltrated with ;0.1 mL of a diluted microbial elicitor solution (0.5 mM Flg22, 0.5 mM Elf26, 100 mg/mL OG, 100 mg/ mL PGN, or 100 mg/mL FoCWE) and placed under high humidity (;85% humidity) for 18 h. At least six independent plants were used as biological replicates, and three rosette leaves were sampled from each plant. The experiment was repeated at least two times. After infiltration and overnight incubation, leaves were harvested and placed in sterile 12-well plates. Acetic acid:ethanol (1:3) was added over 8 h with two changes to destain the chlorophyll from the leaf, followed by ethanol (50% v/v) for 1 h, ethanol (30% v/v) for 1 h, and finally dH 2 O for 2 h with two changes. During the washing steps, the 12-well plates containing the leaves were placed on a constant shaker at 120 rpm to ensure even destaining and rehydration of the leaves. After that, the leaves were stained with 5 mg/mL aniline blue in 150 mM sodium phosphate (pH 7.0) for at least 30 min in the dark. Leaves were mounted in glycerol (50% v/v) and examined under fluorescence microscopy. For experiments involving exogenous H 2 O 2 application, leaves were detached from the plant at 2 h after Flg22 treatment and placed in a sterile 12-well plate overnight in 1 mL of solution containing either the specified concentrations of H 2 O 2 or sterile dH 2 O. Callose intensity was quantified by counting the number of callose spots in images acquired from at least three leaves from each of three to six different plants (representative images shown in Figure 4 and Supplemental Figure 7 online) per line/treatment. On the y axis, a score of 10 was equivalent to saturated callose accumulation, 8 represented > 200 spots or too many to count, 5 represented ;100 spots, 3 represented ;50 spots, 2 represented five to 25 spots, and 1 represented less than five spots. The leaves included in this analysis were from at least six independent plants (three leaves each; see above).

Detection of Reactive Oxygen Species
In situ detection of hydrogen peroxide was performed by staining with DAB (Sigma-Aldrich) using an adaptation of a previous method (Thordal-Christensen et al., 1997;Bindschedler et al., 2006). Briefly, ;100 mL of a diluted microbial elicitor solution (0.5 mM Flg22, 0.5 mM Elf26, 100 mg/mL OG, 100 mg/mL PGN, or 100 mg/mL FoCWE) was infiltrated individually into mature rosette leaves from 23-to 25-d-old Arabidopsis plants using a needleless 1-mL syringe and infiltrated under gentle vacuum with 1 mg/mL DAB containing Tween 20 (0.05% v/ v) and 10 mM sodium phosphate buffer (pH 7.0). At least six independent plants were used as biological replicates, and three rosette leaves were sampled from each plant. The experiment was repeated at least two times. Infiltrated leaves were sampled and stained after 2 h. The staining reaction was terminated 5 h after DAB infiltration, and leaves were fixed in ethanol:glycerol:acetic acid 3:1:1 (bleaching solution) placed in a water bath at 958C for 15 min. Leaves were reimmersed in bleaching solution until chlorophyll was completely depleted and then were visualized under white light and photographed. A combination of tools from Adobe Photoshop and CellProfiler (Carpenter et al. 2006) was used to establish the threshold of DAB staining in the leaves and to distinguish the staining from the background. The final measurement used to quantify the DAB staining was the area of the stain divided by the total area of the leaf.
Inoculation and in Planta Growth Determination of P. syringae P. syringae pv tomato strain DC3000 was cultured on Kings B agar (Sigma-Aldrich) containing 40 mg/mL rifampicin. Single colonies grown for 48 h at 288C were isolated from agar plates and transferred to Kings B or Luria-Bertani broth (Sigma-Aldrich) and grown overnight. Bacterial suspensions at OD 600 0.2 were washed and diluted in 10 mM MgSO 4 to OD 600 0.0005. Mature leaves of 23-to 25-d-old Arabidopsis plants were then infiltrated with 0.1 mL of bacterial inoculum using a needleless 1-mL syringe, and the plants were placed under high humidity (;85% humidity) for the duration of the experiment. Leaves were harvested at 24, 48, and 72 h to test for levels of bacterial growth. Control leaves were harvested 2 h after inoculation to determine the initial dose. At least four biological replicates from independent plants were used for each time point. Each biological replicate was a pooled sample of two rosette leaves from the same plant. Leaf discs of 0.5 cm diameter were homogenized using stainless steel beads (VWR) in 100 mL of 10 mM MgSO 4 in a Qiagen TissueLyser (Qiagen) for 1.5 min at a frequency of 30/sec. Serial 10-fold dilutions were streaked on Kings B agar plates supplemented with 40 mg/mL of rifampicin, and colonies were counted. Growth experiments shown in Figure 3 were blinded. The experimenter received the seeds of the various Arabidopsis lines in coded tubes from the greenhouse manager Jenifer Bush at Massachusetts General Hospital, and the code was not revealed until the in planta growth data for each line had been determined.

Statistical Analyses
Student's t test was performed to calculate whether the differences between distributions of data were significant using PRISM v4.0 (GraphPad Software). A P value of <0.05 was considered statistically significant.

Accession Numbers
Sequence data from this article can be found in the EMBL/GenBank data libraries under accession numbers At 3g49110 (PRX33), At 3g49120 (PRX34), At 5g47910 (RBOHD), and At 5g01075 (b-GAL).

Supplemental Data
The following materials are available in the online version of this article.