Contribution of cell wall peroxidase‐ and NADPH oxidase‐derived reactive oxygen species to Alternaria brassicicola‐induced oxidative burst in Arabidopsis

Summary Cell wall peroxidases and plasma membrane‐localized NADPH oxidases are considered to be the main sources of the apoplastic oxidative burst in plants attacked by microbial pathogens. In spite of this established doctrine, approaches attempting a comparative, side‐by‐side analysis of the functions of extracellular reactive oxygen species (ROS) generated by the two enzymatic sources are scarce. Previously, we have reported the role of Arabidopsis NADPH oxidase RBOHD (respiratory burst oxidase homologue D) in plants challenged with the necrotrophic fungus Alternaria brassicicola. Here, we present results on the activity of apoplastic class III peroxidases PRX33 (At3g49110) and PRX34 (At3g49120) investigated in the same Arabidopsis–Alternaria pathosystem. ROS generated by Arabidopsis peroxidases PRX33 and PRX34 increase the necrotic symptoms and colonization success of A. brassicicola. In addition, the knockdown of PRX33 and PRX34 transcript levels leads to a reduced number of host cells showing an extracellular burst of ROS after inoculation with A. brassicicola. Our results also reveal an age‐dependent transcript distribution of ROS‐producing peroxidase and NADPH oxidase enzymes, and some potential new components of the RBOHD, PRX33 and PRX34 signalling networks.


SUMMARY
Cell wall peroxidases and plasma membrane-localized NADPH oxidases are considered to be the main sources of the apoplastic oxidative burst in plants attacked by microbial pathogens. In spite of this established doctrine, approaches attempting a comparative, side-by-side analysis of the functions of extracellular reactive oxygen species (ROS) generated by the two enzymatic sources are scarce. Previously, we have reported the role of Arabidopsis NADPH oxidase RBOHD (respiratory burst oxidase homologue D) in plants challenged with the necrotrophic fungus Alternaria brassicicola. Here, we present results on the activity of apoplastic class III peroxidases PRX33 (At3g49110) and PRX34 (At3g49120) investigated in the same Arabidopsis-Alternaria pathosystem. ROS generated by Arabidopsis peroxidases PRX33 and PRX34 increase the necrotic symptoms and colonization success of A. brassicicola. In addition, the knockdown of PRX33 and PRX34 transcript levels leads to a reduced number of host cells showing an extracellular burst of ROS after inoculation with A. brassicicola. Our results also reveal an age-dependent transcript distribution of ROS-producing peroxidase and NADPH oxidase enzymes, and some potential new components of the RBOHD, PRX33 and PRX34 signalling networks.

I N T RO D U C T I O N
Plant cells respond to invading microbial pathogens with a series of changes in the activity of their various metabolic pathways. One of the most apparent cellular reactions on pathogen attack is a timely, coordinated accumulation of reactive oxygen species (ROS) in the apoplast and in some cellular compartments, which is often referred to as the 'oxidative burst'. ROS that are released during the course of the oxidative burst include the free radicals superoxide (O 2 •¯) , hydroxyl radicals ( • OH) and nitric oxide ( • NO), with unpaired valence electrons, as well as the nonradicals hydrogen peroxide (H 2 O 2 ) and singlet oxygen ( 1 O 2 ) (Groß et al., 2013;Mignolet-Spruyt et al., 2016;Mittler, 2017). Several enzymatic systems have been recognized as cellular sources of the oxidative burst in the apoplast of plant cells. These include NADPH oxidases, cell wall peroxidases (class III secretory plant peroxidases), amine and polyamine oxidases (AOs and PAOs), oxalate oxidases and quinone reductases (Kärkönen and Kuchitsu 2015). Endeavours to decipher the roles of NADPH oxidases and cell wall peroxidases in plant diseases and immunity have led to significant discoveries, but the functions of AOs, oxalate oxidases and quinone reductases are still largely elusive.
The most extensively studied sources of extracellular ROS in plants challenged by pathogenic microbes are respiratory burst oxidase homologue (RBOH) NADPH oxidases localized in the cellular plasma membrane (Frederickson Matika and Loake, 2013;Liu and He, 2016). The genome of Arabidopsis thaliana (hereafter referred to as Arabidopsis) encodes 10 RBOH proteins. Of these 10 isoforms, RbohD and RbohF are transcribed at the highest levels in root and leaf tissues (Torres et al., 1998). The contribution of Arabidopsis NADPH oxidases RBOHD and RBOHF to the apoplastic oxidative burst, host immunity and/or cell death regulation has been reported in a wide range of Arabidopsis-pathogen interactions. These interactions include virulent or avirulent bacterial, oomycete or necrotrophic and biotrophic fungal pathogens of Arabidopsis (Marino et al., 2012). Concerning the significance of apoplastic Arabidopsis class III peroxidases in plant-pathogen interactions, PRX33 and PRX34 (encoded by loci At3g49110 and At3g49120, respectively) have been described as sources of an extracellular oxidative burst when plants/cell suspensions are challenged by avirulent strains of Pseudomonas syringae (Bindschedler et al., 2006) or treated with microbe-associated molecular pattern (MAMP) elicitors O'Brien et al., 2012). The proportions of ROS released by NADPH oxidases and peroxidases were evaluated by a pharmacological approach. In Arabidopsis cell suspensions treated with various MAMP elicitors, at least 50% of the H 2 O 2 produced could be credited to peroxidases, predominantly to PRX33 and PRX34, and the remaining 50% or less was attributed to NADPH oxidases and intracellular sources . Peroxidases PRX33 and PRX34 were found to contribute to resistance against virulent and avirulent strains of P. syringae, as well as against infections caused by Botrytis cinerea and powdery mildew fungi (Bindschedler et al., 2006). Interestingly, an oxidative burst triggered by the treatment of Arabidopsis plants with a cell wall elicitor prepared from the fungus Fusarium oxysporum could be abolished by knocking down the transcript levels of PRX33 and PRX34 through the transgenic expression of a French bean class III peroxidase cDNA (FBP1) in an antisense orientation . This genotype has been referred to as asFBP1.1. The depletion of the PRX33 mRNA level in an Arabidopsis T-DNA insertion line, however, resulted in reduced susceptibility to F. oxysporum, indicating that PRX33 promotes susceptibility to this particular fungus (Lyons et al., 2015).
Peroxidases PRX33 and PRX34 have also been shown to act in salicylic acid (SA) signalling, as asFBP1.1 plants are impaired in the expression of several SA-responsive genes and in SA-induced NPR1 monomerization (Mammarella et al., 2015). The formation of NPR1 monomers is dependent on the cellular redox state and is also a crucial component of the SA-mediated signalling pathway (Mou et al., 2003;Tada et al., 2008). Two studies on the cell wall proteome of Arabidopsis leaves also drew attention to PRX33 or PRX34 in responses to pathogen infection and redox imbalances. PRX34 is over-represented in the apoplastic proteome of Verticillium longisporum-infected leaves (Floerl et al., 2012), and PRX33 and PRX34 are both more abundant in cell wall protein extracts of ascorbate-deficient Arabidopsis leaves relative to those of wild-type plants (Sultana et al., 2015).
In addition to NADPH oxidases and apoplastic peroxidases, AOs and PAOs have been proposed as alternative sources of extracellular ROS accumulation. Polyamines are catabolized by copper-containing AOs (CuAOs) and FAD-dependent PAOs. PAOs catalyse the oxidation of spermine (Spm), spermidine (Spd) and/or their acetylated derivatives at the secondary amino groups (Angelini et al., 2010). In Arabidopsis, five PAO isoforms (AtPAO1-AtPAO5) have been identified, and the main physiological role of these proteins has been linked to compartment-specific H 2 O 2 synthesis in different phases of development and differentiation, as well as in the course of defence mechanisms against pathogens and abiotic stress (Andronis et al., 2014;Fincato et al., 2012;Takahashi et al., 2010). Arabidopsis plants inoculated with an avirulent strain of P. syringae pv. tomato DC3000 (avrRpm1) accumulate polyamines, and infiltration of polyamines into the apoplast of leaves causes an oxidative burst and subsequent cell death (Yoda et al., 2009).
The Arabidopsis thaliana-Alternaria brassicicola pathosystem has been widely used to study plant defence and fungal pathogenesis strategies (Schenk et al., 2003;Su'udi et al., 2011). We have reported previously that the functional Arabidopsis NADPH oxidase RBOHD (respiratory burst oxidase homologue D) is required for an extracellular oxidative burst when plants are spray inoculated with a conidial suspension of the necrotrophic fungus A. brassicicola. An rbohd knockout line showed reduced susceptibility to the pathogen and increased spread of cell death, indicating that RBOHD contributes to the success of the colonization process and is also engaged in the regulation of A. brassicicola-induced host cell death. Finally, an interplay between RBOHD, ethylene and SA was demonstrated, and diverse roles of RBOHD-dependent ROS were suggested between cells that were affected directly by the fungus and cells that were in neighbouring positions (Pogány et al., 2009).
The necrotrophic fungus A. brassicicola causes black spot disease and is an economically important pathogen of Brassicaceae species (Thomma, 2003). The interaction between A. brassicicola strain MUCL 20297 and Arabidopsis thaliana ecotype Col-0 (both used throughout this work) is considered to be incompatible (Narusaka et al., 2005;Thomma et al., 1998). Nevertheless, our experimental conditions (which were excessively favourable for the fungus) enabled considerable pathogen growth (Pogány et al., 2009).
The various enzymatic sources of extracellular ROS in diseased (or wounded) plants are typically examined separately and in different pathosystems. Sometimes, when these sources are investigated together, enzyme inhibitors are used to distinguish between NADPH oxidase-, class III peroxidase-or PAO-derived ROS (Dmochowska-Boguta et al., 2013;O'Brien et al., 2012;Roach et al., 2015). As an extension of our work on the cellular functions of the NADPH oxidase RBOHD (Pogány et al., 2009), in this study, we provide results on the contribution of peroxidases PRX33 and PRX34 to a pathogen-induced apoplastic oxidative burst. The experiments are performed with the same Arabidopsis-Alternaria pathosystem as in the RBOHD study, using knockout and knockdown insertion mutant and genesilenced plant samples. Some likely candidates of the RBOHD, PRX33 and PRX34 signalling networks are also exposed.

Elicitation of Arabidopsis leaf cells by A. brassicicola activates PRX33 and PRX34 transcription
To explore the contribution of apoplastic peroxidases and PAOs in the oxidative burst, we monitored mRNA abundance for peroxidase genes PRX33 and PRX34 and PAO isoforms PAO1, PAO2, PAO3, PAO4 and PAO5 in our pathosystem. Expression of the two cell wall peroxidases was clearly induced after 12 h, and the highest expression levels were observed at 24 h after inoculation (hai) for both PRX genes in comparison with mockinoculated plants (Fig. 1). In contrast, transcription of PAO1, PAO2, PAO3 and PAO4 remained unaffected and PAO5 mRNA levels were repressed by elicitation with A. brassicicola at 24 hai, indicating that they are not involved in the A. brassicicolainduced oxidative burst and other host responses (Fig. S1, see Supporting Information).

Knockdown of PRX33 and PRX34 transcript levels reduces symptoms and colonization success of A. brassicicola
In a process to select Arabidopsis genotypes with adequate PRX33 and PRX34 knockdown mRNA levels, T-DNA insertion mutants prx33 (SALK_062314C) and prx34 (SALK_051769C) (Passardi et al., 2006), together with the line asFBP1.1 expressing a French bean class III peroxidase in an antisense orientation (Bindschedler et al., 2006), were initially included in this work. Transcript levels of peroxidase gene PRX33 were significantly reduced in the prx33 insertion line and in asFBP1.1, but mRNA levels for the peroxidase gene PRX34 were not sufficiently suppressed in our hands in the prx34 insertion line or in the asFBP1.1 line. Therefore, we created a Tobacco rattle virus (TRV) vectorbased virus-induced gene silencing (VIGS) construct (Hayward et al., 2011) that provided Arabidopsis plants with markedly reduced PRX34 (and PRX33) transcript levels. This TRV-GFP-PRX VIGS construct is hereafter abbreviated as TRV-PRX. Diminished activity of PRX33 and PRX34 genes in the prx33 insertion line and in plants treated with the TRV-PRX VIGS construct is shown in Figs S2 and S3 (see Supporting Information). All subsequent pathogen and ROS detection assays aimed at the characterization of extracellular PRX33 and PRX34 functions in Arabidopsis during fungal pathogenesis were conducted with the prx33 (SALK_062314C) line and with plants carrying the TRV-PRX VIGS construct. A previously characterized rbohd NADPH oxidase knockout line (SALK_070610C) was also included in some experiments (Pogány et al., 2009).
Down-regulation of PRX33 and PRX34 yielded reduced fungal symptoms on Arabidopsis leaves. The percentage of leaf chlorosis and necrosis triggered by A. brassicicola in plants with diminished PRX33 and PRX34 transcript levels was only 32%-36% of that in plants showing wild-type PRX33/PRX34 transcript levels (Fig. 2). The fungal biomass of A. brassicicola was also lower in prx33/prx34 knockdown plants, representing 68% (prx33) and 51% (TRV-PRX) of that of corresponding inoculated control plants with wild-type PRX33/PRX34 transcript levels (Fig. 3).

H 2 O 2 accumulation
H 2 O 2 was detected by a non-fluorescent 3,3′-diaminobenzidine (DAB) histochemical staining procedure, which is based on an H 2 O 2 -dependent peroxidase-catalysed polymerization reaction (Thordal-Christensen et al., 1997), and a fluorescent 2′,7′-dichlorofluorescein diacetate (DCFH-DA) method. DCFH-DA also reacts with H 2 O 2 in the presence of peroxidases, yielding a fluorescent DCF product (Bozsó et al., 2005). Inoculated leaves were observed microscopically for the DABstained samples and visualized with UV light for the DCFH-DA-stained leaves, where the intensity of the fluorescence signal released by individual leaves was detected.
Knockdown of PRX33 and PRX34 transcript levels led to decreased H 2 O 2 formation in Arabidopsis leaves challenged with A. brassicicola (Fig. 4). This altered response was quantitative because the apoplastic burst was not completely abolished, but the number of cells exhibiting a burst of extracellular H 2 O 2 was reduced ( Fig. 4C). This is in contrast with the fully inhibited apoplastic ROS accumulation in the rbohd mutant (Pogány et al., 2009). We quantified the number of cells showing extracellular H 2 O 2 accumulation at 2 days after inoculation with A. brassicicola for 20 randomly selected infection sites, observing leaves from two independent experiments. This evaluation revealed that infection sites in wild-type (Col-0) plants were composed of an average of 42.6 ± 5.7 cells showing ROS accumulation relative to 16.0 ± 2.5 cells in prx33 plants. Likewise, infection sites in TRV-GFP plants with normal PRX33/PRX34 transcript levels comprised an average of 34.4 ± 3.8 cells exhibiting H 2 O 2 accumulation relative to 17.0 ± 1.2 cells in prx33/prx34 under-producing TRV-PRX plants. Fig. 1 Alternaria brassicicola-induced activation of PRX33 and PRX34 genes in wild-type Arabidopsis plants. Five-to six-week-old Arabidopsis plants (whole rosettes) were spray inoculated with A. brassicicola conidial suspension at a concentration of 5 × 10 5 conidia/mL distilled water. Data represent the mean of three independent biological samples with three technical replicates each. Statistical analysis was performed using Mann-Whitney U-test. Asterisks indicate statistically significant differences (*α = 0.05, ***α = 0.001).

Fig. 2
Symptoms of Alternaria brassicicola infection are suppressed in Arabidopsis plants with reduced PRX33/PRX34 mRNA levels. Leaves of A. brassicicolainfected Col-0, prx33 knockdown mutant, rbohd knockout mutant, TRV-GFP and TRV-PRX plants are shown 10 days after inoculation. Five-to six-week-old Arabidopsis plants (whole rosettes) were spray inoculated with A. brassicicola conidial suspension at a concentration of 5 × 10 5 conidia/mL distilled water. Detached leaves in middle positions (leaf levels 5-8) were evaluated. Results are presented as the chlorotic and necrotic leaf area compared with the total surface area of leaf blades analysed by ImageJ (1 is equal to 100% leaf surface), and represent the means of two experiments (n = 30 for each genotype/ treatment) ± standard error (SE). Different letters indicate statistically significant differences between genotypes/treatments using Tukey's post hoc test. Increased symptoms of A. brassicicola on leaves of the rbohd mutant are included as a reference. The adaptation of a quantifiable, fluorescent, ROS detection method suitable for whole leaves gave us similar results on the accumulation of ROS in our A. brassicicola-infected genotypes/ treatments, confirming the lower levels of ROS in prx33/prx34 under-producing Arabidopsis plants (Fig. 4D). ROS formation, represented by the intensity of fluorescence signals, in leaves of infected prx33 plants was 59% of that in leaves of the corresponding infected wild-type plants, and in leaves of infected TRV-PRX plants was 74% of that in leaves of infected TRV-GFP plants, measured 2 days after inoculation with A. brassicicola. The fluorescence signal intensity in leaves of NADPH oxidase mutant rbohd plants was also detected as a reference, and a marked reduction in ROS accumulation was observed. In rbohd plants, the ROS-related fluorescence signal intensity was 44% of that of wild-type Col-0 plants.

Leaf senescence clearly affects RbohD, PRX33 and PRX34 mRNA levels
The results of several studies have indicated functional interactions between the Arabidopsis NADPH oxidase RBOHD and the senescence hormone ethylene or its precursor 1-aminocyclopropane-1-carboxylic acid (ACC) (Bouchez et al., 2007;Pogány et al., 2009;Mersmann et al., 2010;Yao et al., 2017). This observation prompted us to investigate the effect of leaf age on the activity of genes (RbohD, PRX33 and PRX34) encoding crucial ROS-generating proteins in Arabidopsis, before and after elicitation with A. brassicicola. Sampling was performed by dividing true leaves growing on one plant evenly into three age groups: oldest leaves (in leaf positions 1-4), middle leaves (in leaf positions 5-8) and youngest leaves (in leaf positions 9-12). The basal level of transcript abundance examined in mock-inoculated wild-type plants was highest in the oldest bottom leaves for all three ROS-producing proteins, and non-senescent or younger leaves exhibited moderate or lower basal transcript abundance for RbohD, PRX33 and PRX34 genes (Fig. 5). Twenty-four hours after inoculation with A. brassicicola, however, non-senescent middle leaves responded to fungal infection with the highest induction of RbohD, PRX33 and PRX34 transcript levels (Fig. 5).

Searching for new components in the PRX33/PRX34 and RBOHD signalling networks
In a pursuit to find unknown elements of the RBOHD or PRX33/ PRX34 cellular interaction networks, two approaches were employed. First, T-DNA insertion lines with perturbed activity of key Arabidopsis immune regulators were inoculated with A. brassicicola. Mutants corresponding to seven crucial determinants [BIK1, NPR1, MPK6, NIA2, EIN2, JAR1 (JASMONATE RESISTANT1) and NHO1 (NON-HOST RESISTANCE TO P. SYRINGAE PV. PHASEOLICOLA 1)] showed altered host responses (in comparison with wild-type plants) after inoculation with A. brassicicola. Six were further investigated. BIK1 was excluded because its role in the regulation of RBOHD activity has been confirmed (Kadota et al., 2014;Li et al., 2014). Mutants affected in the activity of two of these pivotal plant immunity factors, JAR1 and NHO1, exhibited consistently altered PRX33 and PRX34 transcript levels ( Fig. 6), but the other four mutants showed wild-type transcript levels for PRX33 and PRX34 genes. Interestingly, no apparent differences in RbohD mRNA levels could be detected between any of the immune mutants and the wild-type, before or after inoculation with A. brassicicola.
In a second approach, T-DNA insertion lines affected in the activity of 11 selected target proteins were analysed. Earlier, these proteins had been reported to establish a physical interaction with RBOHD (Geisler-Lee et al., 2007;Jones et al., 2014). Mutants with associated insertions in the following genes were included: UDP-Galactose Transporter 3, Annexin 1, Thioredoxin H-type 7, Calmodulin 4, Glutathione S-Transferase TAU 19, Quantitative Resistance to Plectosphaerella 1, Cell Elongation Protein, ADP-Ribosylation Factor C1, Membrane-Associated Progesterone Binding Protein 3, At4g37445 without functional prediction and NDR1/HIN1-like 3 (Table S1, see Supporting Information). Two mutants in which the insertions could be linked to the same locus, At2g26330 encoding ERECTA (Quantitative Resistance to Plectosphaerella 1), gave spreading cell death phenotypes on inoculation with A. brassicicola, similar to those seen on the rbohd mutant (Fig. 7). The disease responses of the mutants representing the other 10 RBOHD interactors were the same as in the wild-type. One of the erecta T-DNA insertion lines, SALK_066455C, was also assayed for ROS formation after inoculation with A. brassicicola. The accumulation of H 2 O 2 , visualized by DAB staining, was highly compromised in whole leaves of the erecta mutant at 48 hai with A. brassicicola (Fig. 8).

D I SC U SS I O N Alternaria brassicicola infection activates PRX33 and PRX34 genes
The mRNA expression of genes encoding ROS-producing proteins was examined in A. brassicicola-infected Arabidopsis plants. PRX33 and PRX34 mRNA levels were elevated even at 12 hai, reaching their climax at 24 hai. The analysis of NADPH oxidase RbohD transcript levels, investigated in the same pathosystem, revealed a similar transcriptional response (Pogány et al., 2009). The induction of PRX34 gene expression was also reported in Sclerotinia sclerotiorum-infected Brassica napus plants at 12-48 hai (Yang et al., 2007). PAO1, PAO2, PAO3, PAO4 and PAO5 encoding genes, however, did not show A. brassicicolainduced activation at 24 hai with the fungus (Figs 1 and S1).
These data suggest that the apoplastic peroxidases PRX33 and PRX34 play important roles in the response of Arabidopsis to A. brassicicola, but the significance of PAO1, PAO2, PAO3, PAO4 and PAO5 was not proven.
PRX33 and PRX34 contribute markedly to the A. brassicicola-induced accumulation of H 2 O 2 Impaired RBOHD function leads to the abolishment of apoplastic ROS accumulation in Alternaria-infected Arabidopsis cells, Fig. 6 Transcript levels of PRX33 and PRX34 genes in nho1 and jar1 T-DNA insertion lines relative to wild-type plants. The level of gene expression was analysed before and after inoculation with Alternaria brassicicola [24 h after inoculation (hai)]. Dysfunctions in NHO1 and JAR1 activity affect PRX33 and PRX34 mRNA levels. Five-to six-week-old Arabidopsis plants (whole rosettes) were spray inoculated with A. brassicicola conidial suspension at a concentration of 5 × 10 5 conidia/mL distilled water. The results show the average of two experiments, each comprising three biological samples (each sample composed of a pool of three Arabidopsis rosettes), analysed in three technical replicates. Statistical analysis was performed using Student's t-test. Asterisks indicate statistically significant differences (*α = 0.05, **α = 0.01, ***α = 0.001).

Fig. 7
Alternaria brassicicola-infected Col-0, rbohd, erecta SALK_04410 and erecta SALK_066455C mutant Arabidopsis plants. Fungal symptoms were photographed at 7 days after inoculation. T-DNA insertion lines with mutations in the genomic sequence of RbohD (At5g47910) and ERECTA (At2g26330) genes were used. Necrotic fungal symptoms are more intense in the two erecta mutants, as in the rbohd mutant, in comparison with the Col-0 wild-type. Five-to six-week-old Arabidopsis plants (whole rosettes) were spray inoculated with A. brassicicola conidial suspension at a concentration of 5 × 10 5 conidia/mL distilled water. whereas down-regulation of PRX33/PRX34 activity results in a reduced number of leaf cells exhibiting the burst of extracellular ROS (Fig. 4). These results reveal that the RBOHD and PRX33/ PRX34 ROS-generating systems contribute simultaneously to the Alternaria-induced extracellular oxidative burst in Arabidopsis. The reduced number of leaf cells showing H 2 O 2 accumulation indicates a role of the apoplastic peroxidases PRX33/PRX34 in the spread of intercellular ROS signalling initiated by RBOHD (Gilroy et al., 2014(Gilroy et al., , 2016. Some crucial Arabidopsis stomatal responses mediated by the plant hormone cytokinin have been recently linked to the activity of peroxidases PRX33 and PRX34, together with certain other apoplastic peroxidases (Arnaud et al., 2017). It was shown that cytokinin-induced stomatal closure, ROS accumulation in guard cells and stomatal immunity to coronatine-deficient P. syringae pv. tomato DC3000 bacteria were dependent on the activity of the extracellular peroxidases PRX33 and PRX34.

Necrotic symptoms of the fungal infection are enhanced by PRX33/PRX34-derived ROS
Down-regulation of PRX33 and PRX34 led to reduced necrotic fungal symptoms on Arabidopsis leaves (Fig. 2) and the relative fungal biomass of A. brassicicola was lower in prx33/prx34 knockdown plants (Fig. 3). These results suggest that, similar to RBOHD, functional PRX33 and PRX34 and their ROS products support the growth of the fungus. However, PRX33/PRX34 activity leads to increased tissue necrosis, in contrast with functional RBOHD, which suppresses tissue necrosis.
It is unclear whether the reduced necrotic symptoms in our prx33/prx34 under-producers are the result of partially compromised cell death induction or simply the consequence of limited necrotrophic pathogen growth. Mammarella et al. (2015) reported that asFBP1.1 prx knockdown plants retained the ability to mount a hypersensitive cell death response to attacks by avirulent P. syringae strains, suggesting that the apoplastic peroxidases PRX33/PRX34 are dispensable in pathogen-induced cell death in Arabidopsis.
The inhibition of RBOHD activity in a T-DNA mutant resulted in more intense tissue necrosis (increased spread of cell death) after inoculation with A. brassicicola. Fungal colonization, however, was suppressed in rbohd knockout plants relative to the wild-type. Thus, RBOHD regulates host cell death on A. brassicicola infection and functional RBOHD supports the colonization of Arabidopsis by this necrotrophic fungus (Pogány et al., 2009).

Leaf senescence modulates the transcript levels of ROS-producing enzymes
Leaf senescence has been associated with elevated ROS levels and reduced antioxidant capacity (Barna et al., 2012). In our studies, senescence was connected with an increased basal transcript level of pivotal ROS-producing enzymes, but also with a reduced ability to develop a full transcriptional response of RbohD, PRX33 and PRX34 genes following the attack of a necrotrophic fungus (Fig. 5).
This is in accord with our microscopic observations, where cells in older leaves of Arabidopsis plants exhibited a mild, but uniform, induction of apoplastic basal H 2 O 2 accumulation, but younger (non-senescent) leaves responded to A. brassicicola infection with a more pronounced accumulation of extracellular H 2 O 2 . Characteristic spatio-temporal mRNA expression patterns of NADPH oxidase isoforms RbohD and RbohF have been reported recently in unchallenged and pathogen-elicited Arabidopsis plants (Morales et al., 2016).

Quest for new PRX33/34 and RBOHD interactors
In a search for new components in the PRX33/PRX34 and RBOHD signalling networks, numerous Arabidopsis mutants were tested.
T-DNA insertion lines jar1 and nho1 consistently exhibited altered PRX33/PRX34 transcript abundance, both before and after inoculation with A. brassicicola (Fig. 6). JAR1 encodes a jasmonic acid-amido synthetase which catalyses the formation of a biologically active jasmonyl-isoleucine (JA-Ile) conjugate. This amino acid is necessary for optimal signalling in some jasmonate responses in Arabidopsis (Staswick and Tiryaki, 2004;Staswick et al., 2002). Synthesis of the JA-Ile conjugate by JAR1 and related JA-conjugating enzymes is required for the plant immune system (Ryu et al., 2004;Staswick et al., 1998) and for various abiotic stress responses (Rao et al., 2000). Functional JAR1 alleviates the oxidative burst in Arabidopsis seedlings triggered by the bacterial MAMP elicitor flagellin (Yi et al., 2014). An interaction between JAR1 and ROS has been reported during cell wall damage-induced lignin synthesis in Arabidopsis (Denness et al., 2011). NHO1 protein is a glycerol kinase, which converts glycerol to glycerol 3-phosphate and performs a rate-limiting step in glycerol metabolism (Kang et al., 2003). This protein is also known as GLI1 or GLYCEROL-INSENSITIVE 1 (Eastmond, 2004). Arabidopsis NHO1 transcript levels are induced by non-host Pseudomonas bacterial strains, and functional NHO1 is required for resistance against non-host and avirulent Pseudomonas strains, as well as against the necrotrophic fungal pathogen Botrytis cinerea. However, the NHO1 gene is ineffective against virulent P. syringae pv. tomato DC3000, and its transcription is suppressed by the DC3000 strain. Interestingly, the JA signalling pathway is specifically required for NHO1 suppression by DC3000 bacteria (Kang et al., 2003;Li et al., 2005;Lu et al., 2001). Although the fact that NHO1 is a significant component of the immune system of Arabidopsis is well established, its specific functions in immune signalling are still largely elusive (Lu et al., 2010;Maeda et al., 2010). Nonetheless, a connection between ROS production and NHO1 activity is emerging . Perturbation of the cellular redox status in plants, triggered by catalase or glycolate oxidase deficiency or by treatment with the ROS-generating compound methylviologen, significantly alters the transcript level of NHO1 (Chaouch et al., 2010;El-Maarouf-Bouteau et al., 2015;Rojas et al., 2012). Conversely, plants with disturbed glycerol metabolism or with impaired NHO1 gene expression have been shown to exhibit elevated H 2 O 2 formation Li et al., 2016). Our results, revealing enhanced PRX33/PRX34 gene activity in the nho1 knockout (in comparison with wild-type plants), before or after inoculation with A. brassicicola, indicate that these extracellular peroxidases may be partly responsible for the observed interaction between ROS and NHO1 (Fig. 6).
In a different set of experiments, Arabidopsis plants carrying mutations in the ERECTA gene were shown to display spreading necrosis symptoms on A. brassicicola infection, very similar to those observed on the rbohd mutant (Fig. 7). Inoculated leaves of the erecta mutant (SALK_066455C) also exhibited reduced H 2 O 2 accumulation (Fig. 8). ERECTA has been reported previously to establish a physical interaction with RBOHD (Geisler-Lee et al., 2007;Jones et al., 2014).
ERECTA (QUANTITATIVE RESISTANCE TO PLECTOSPHAERELLA 1, AGI locus code: At2g26330) has been connected to thermotolerance. It belongs to the protein kinase superfamily and to the serine/threonine (Ser/Thr) protein kinase family, which includes proteins that are associated with development, pathogen defence and phytohormone perception (Sánchez-Rodríguez et al., 2009;Torii et al., 1996). ERECTA plays a role in the immune response of Arabidopsis because erecta mutant plants are more susceptible than wild-type plants to a variety of pathogens (Magnaporthe oryzae, Plectosphaerella cucumerina, Verticillium longisporum) (Haffner et al., 2014;Llorente et al., 2005;Takahashi et al., 2016). Fusarium graminearum infection apparently down-regulates mRNA abundance for the ERECTA gene in Arabidopsis pistils (Mondragon-Palomino et al., 2017). Results in our study demonstrate that ERECTA is also required for ROS accumulation and the suppression of necrotic symptoms caused by A. brassicicola, a necrotrophic fungal pathogen. ERECTA not only physically interacts with RBOHD (one of the major cellular sources of apoplastic ROS) in Arabidopsis, but has been shown to be part of a signalling pathway responsible for ROS sensing and redox-mediated cortex proliferation in the roots (Cui et al., 2014). Recently, the ROS responsivity of the ERECTA gene has been exposed in methylviologen-treated Arabidopsis rosettes (Han et al., 2014).
Taken together, we consider JAR1 and NHO1 to be potential members of the PRX33/PRX34 cellular signalling system and ERECTA to be an expected component of the RBOHD interaction network in Arabidopsis leaf cells when plants are challenged by the necrotrophic fungus A. brassicicola.
In this study, we have provided information on the simultaneous and comparable contribution of class III peroxidases PRX33 and PRX34 and NADPH oxidase RBOHD to the apoplastic oxidative burst investigated in the same Arabidopsis-Alternaria pathosystem. We have also presented microscopic images of the altered pattern of ROS accumulation in Arabidopsis plants with reduced PRX33 and PRX34 activity following inoculation with A. brassicicola. In contrast with the ROS-producing NADPH oxidase RBOHD, which surprisingly inhibits the spread of cell death in pathogen-infected Arabidopsis tissues (Pogány et al., 2009), functional cell wall peroxidases PRX33 and PRX34 apparently enhance the development of necrotic symptoms triggered by infection with A. brassicicola. Fungal growth, however, is similarly stimulated by both RBOHD and the two apoplastic peroxidases. Finally, a senescence-dependent distribution of RbohD, PRX33 and PRX34 transcripts is reported. As a conclusion of this study, ROS generated by the extracellular peroxidases PRX33 and PRX34 function in Arabidopsis as susceptibility factors when plants are challenged by the necrotrophic fungal pathogen A. brassicicola.

E X PE R I M E N TA L PROC E D U R E S Plant materials and growth conditions
Wild-type (Col-0), transgenic, mutant and gene-silenced Arabidopsis plants were grown in a peat growing medium (Pindstrup Plus Orange, Pindstrup Mosebrug, Ryomgaard, Denmark) at 22 °C in a growth chamber under a 14-h light/10-h dark cycle with 80 µmol/m 2 /s irradiation. Seeds of the Arabidopsis line asFBP1.1 expressing a French bean (Phaseolus vulgaris) peroxidase sequence in antisense orientation (Bindschedler et al., 2006) were received from Professor F. M. Ausubel (Department of Genetics, Harvard Medical School, Boston, MA, USA). Arabidopsis T-DNA insertion lines were ordered from the European Arabidopsis Stock Center. The list of Arabidopsis lines used in this study is provided in Table S1. Characterization of the prx33 insertion line (SALK_062314C) is shown in Fig. S4 (see Supporting Information), whereas lines prx34 (SALK 051769C) and rbohd (SALK_070610C) have been described previously (Passardi et al., 2006;Pogány et al., 2009). An Arabidopsis line expressing a green fluorescent protein coding mGFP-ER sequence (Haseloff et al., 1997), used for all VIGS experiments, was generated by a modified floral dip method (Logemann et al., 2006).

Cultivation of fungal strain and inoculation
Alternaria brassicicola strain MUCL20297, isolated from cabbage (Thomma et al., 1998), was grown on potato dextrose agar medium at room temperature in the dark. The conidial suspension contained 5 × 10 5 conidia/ml distilled water, and conidia were harvested from 1-week-old cultures. The inoculum was sprayed onto the leaf surface of 5-6-week-old Arabidopsis plants. The infected plants were incubated in 100% relative humidity in translucent plastic boxes under the same growing conditions as before. For mock-inoculated plants, distilled water was sprayed onto the leaves and they were kept under the same conditions as the inoculated plants.

Gene expression analysis
For total RNA extraction, Arabidopsis leaves were frozen in liquid nitrogen and ground with a mortar and pestle; 100 mg of plant material was used according to the manufacturer's instructions (Total RNA Extraction Miniprep Kit, Viogene, Taipei, Taiwan). The RNA content and purity were analysed with a NanoDrop-1000 (Thermo Fisher Scientific, Waltham, MA, USA) spectrophotometer.
Prior to cDNA synthesis all samples were subjected to DNase treatment (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA), DNA-free™ DNA Removal Kit).
First-strand cDNA was synthesized with a Thermo Scientific First Strand cDNA Synthesis Kit and used as a template for real-time polymerase chain reaction (PCR) analysis in ten-fold dilution. Realtime PCRs were prepared using a Bioline (Taunton, MA, USA) SensiFAST SYBR ® No-ROX Kit. The reactions were performed in a total volume of 15 μL containing 7.5 μL 2X Bioline SensiFAST qPCR Master Mix, 1.5-1.5 μL forward and reverse primers (10 μm), 2.5 μL template and 2 μL PCR-grade water. The list of primer pairs used in this work is provided in Table S2 (see Supporting Information). All real-time PCRs were carried out in a C1000 Touch Thermal Cycler equipped with a CFX96 Real-Time PCR System (Bio-Rad, Hercules, CA, USA). The following cycling conditions were selected: 95 °C for 3 min, then 40 cycles of 95 °C for 20 s, followed by 60 °C for 60 s. Finally, a melt curve analysis was performed to determine amplicon specificity with temperature increases from 65 to 95 °C in steps of 0.5 °C. Relative quantification analysis was performed using the comparative 2 −ΔΔCt method (Livak and Schmittgen, 2001). To evaluate the level of gene expression, the results were normalized using Ct values from the cDNA amplification of the constitutively expressed Arabidopsis At4g26410 gene (Czechowski et al., 2005).

Determination of fungal biomass
Genomic DNA was isolated using the method of Brouwer et al. (2003) with some modifications. Spray-infected rosettes of Arabidopsis plants (10 days after inoculation with A. brassicicola) were ground in liquid nitrogen with a mortar and pestle. To 100 mg frozen plant material, 300 µL lysis buffer [2.5 m LiCl, 50 mm Tris-HCl, 62.5 mm Na 2 -ethylenediaminetetraacetic acid (EDTA), 4% Triton X-100, pH 8.0] and an equal volume of phenol-chloroform-isoamyl alcohol (25 : 24 : 1, v/v) were added, and samples were thoroughly vortexed. After centrifugation (5 min, 16 000 g) at room temperature, the supernatant was recovered and genomic DNA was precipitated by the addition of two volumes of 100% ethanol, incubation for 15 min at −20 °C and another round of centrifugation. The DNA pellet was washed with 70% ethanol, air dried and resuspended in nuclease-free water. DNA quality and concentration were analysed with a NanoDrop spectrophotometer. Fungal biomass content was determined from total genomic DNA extracts by quantitative real-time PCR analysis, and the ratio of A. brassicicola to Arabidopsis genomic DNA was assessed using the same platform and conditions as for the gene expression analyses described above. Primer sequences specific for A. brassicicola (Pogány et al., 2009) were derived from the ribosomal ITS region of the fungus (forward, 5′-TCTCCAGTTTGCTGGAGACT-3′; reverse, 5′-GGATGCTGACCTTGGCTGGA-3′), and a primer pair specific for the sequence of Arabidopsis At4g26410 was used to determine plant biomass content in the samples.

VIGS
TRV, a bipartite virus-based silencing system, was used for VIGS experiments. TRV1 contains the viral replicase, the RNA-dependent RNA polymerase and the movement protein, whereas TRV2 contains the coat protein and a multiple cloning site to insert host plant-derived fragment(s) of target sequences (Hayward et al., 2011). pTRV1 (STOCK: CD3-1039) and pTRV2 in pCAMBIA3301 (STOCK: CD3-1043) Agrobacterium plasmids were obtained from the Arabidopsis Biological Resource Center.
Afterwards, VIGS plasmids were introduced into Escherichia coli DH5α by the heat shock method (Tu et al., 2005). The identity of the clones was confirmed by DNA sequence determination. VIGS plasmids were finally purified from E. coli (Nucleospin Plasmid, Macherey-Nagel) and introduced into Agrobacterium tumefaciens MOG301 (Hood et al., 1993) by electroporation. For VIGS, Agrobacterium strains containing pTRV1 and pTRV2 vectors were grown at 28 °C overnight on LB (Lysogeny Broth) medium supplemented with the appropriate antibiotics (kanamycin sulfate, 30 µg/mL; rifampicin, 50 µg/ mL). For inoculum preparation, bacterial cells were suspended in Agrobacterium incubation buffer (1.95 g MES (2-[N-morpholino]ethanesulfonic acid), 2 g MgCl 2 . 6H 2 O in 1 L distilled water, pH 5.6) and supplemented with acetosyringone (final concentration of 150 mm). Bacterial cell densities were adjusted with a spectrophotometer to an optical density at 600 nm (OD 600 ) = 1.5. After a 3-h incubation at room temperature, the bacterial suspensions (TRV1 and TRV2), mixed in a ratio of 1 : 1 (v/v), were infiltrated into two lower leaves of 2-3-week-old, GFP-expressing transgenic Arabidopsis plants, which were grown at 22 °C in a growth chamber under a 16-h light/8-h dark cycle (Burch-Smith et al., 2006).
Three weeks after agroinfiltration, the effect of gene silencing was observed under UV light and gene-silenced plants were selected and used for mechanical inoculation as viral inoculum on new 2-3-week-old, GFP-expressing transgenic Arabidopsis seedlings which were kept under a 14-h light/10-h dark cycle. Mechanical inoculation was chosen instead of agroinfiltration because the 16-h light/8-h dark photoperiod condition required for agroinoculation did not provide Arabidopsis plants appropriate for pathological experiments with A. brassicicola (plants grown under these conditions did not develop useful rosettes). Fourteen days after mechanical inoculation, the effect of gene silencing was tested on entire plants under UV light and by real-time PCR from the UV-selected plants. The mRNA contents of PRX33 and PRX34 were calculated using the same platform and conditions as for the real-time PCR assays above (PRX33: forward, 5′-AAATTCAGCCCGAGGATTTC-3′; reverse, 5′-GAGCAGCAATGGTGAGCATA-3′; PRX34: forward, 5′-CGAGAAACCATTGTAAATGAGT-3′; reverse, 5′-CCGAGCCGAATTT GCG-3′).
The fluorescent dye DCFH-DA was also used to detect H 2 O 2 . Detached Arabidopsis leaves in the middle position (in positions 5-8) were vacuum infiltrated with 0.4 mm DCFH-DA in 10 mm sodium phosphate buffer (pH 7.4), as described by Bozsó et al. (2005). After a 10-min incubation period in the dark, leaves were photographed under 302-365-nm UV light illumination by an AlphaImager Mini gel documentation system (ProteinSimple, Santa Clara, CA, USA). For each genotype or treatment, 30 leaves (in the middle position) were assayed and, for each leaf, an average pixel fluorescence intensity was calculated by AlphaView software of AlphaImager ® Systems using the Multiplex Band Analysis function.

Statistical analysis
Experimental data were analysed using Student's t-test, the non-parametric Mann-Whitney U-test or one-way analysis of variance (ANOVA) and subsequent Tukey's honestly significant difference test. Statistical analyses were performed with IBM SPSS Statistics 20 Software (Armonk, New York, USA).

SU PP O R T I N G I N F O R M AT I O N
Additional supporting information may be found in the online version of this article at the publisher's web site:

Fig. S1
Transcript levels of five polyamine oxidase (PAO)encoding genes of wild type (Col-0) Arabidopsis thaliana. Five to 6 weeks old Arabidopsis plants (whole rosettes) were sprayinoculated with A. brassicicola conidium suspension used in a concentration of 5 x 10 5 conidia in 1 mL distilled water. Samples of mock-inoculated and A. brassicicola-infected plants (24 hai) were analyzed. Transcript levels of the five PAO isoforms are not increased as a result of the fungal infection. Data represent the mean of three independent biological samples with three technical replicates for each. Statistical analysis was performed using Student's t-test. Asterisks indicate statistically significant difference (**α = 0.01). Fig. S2 Transcript level of apoplastic class III peroxidase gene PRX33 (At3g49110) is reduced in the Arabidopsis prx33 knockdown T-DNA insertion line (SALK_062314C). Transcript levels were quantified in untreated Arabidopsis plants by real-time RT-PCR. The results show the average of two experiments each comprising three biological samples (each sample composed as a pool of 3 Arabidopsis rosettes) analyzed in three technical replicates. Statistical analysis was performed using Student's t-test. Asterisks indicate statistically significant difference (***α = 0.001). Fig. S3 Transcript levels of apoplastic class III peroxidase genes PRX33 (At3g49110) and PRX34 (At3g49120) are reduced in Col-0 Arabidopsis plants that were treated with a VIGS construct (TRV-PRX) that targets PRX33 and PRX34 mRNA sequences. Plants labelled as TRV-GFP were treated with the control construct that targets only mRNA sequences of the GFP marker gene. Silencing of transgene GFP transcript levels and a subsequent abolishment of GFP-derived green fluorescence in VIGS-treated plants was used to detect the occurrence of successful gene silencing events. Five to 6 weeks old Arabidopsis plants (whole rosettes) were spray-inoculated with A. brassicicola conidium suspension used in a concentration of 5 x 10 5 conidia in 1 mL distilled water. PRX33 and PRX34 mRNA levels were monitored in VIGS-treated Arabidopsis plants before (A) and after (B) inoculation with A. brassicicola (24 hai) by real-time RT-PCR. The results show the average of two experiments each comprising three biological samples (each sample composed as a pool of 3 Arabidopsis rosettes) analyzed in three technical replicates. Statistical analysis was performed using Student's t-test. Asterisks indicate statistically significant differences (*α = 0.05, **α = 0.01, ***α = 0.001). Fig. S4 Confirmation of T-DNA insertion by genotyping prx33 (SALK_062314C) Arabidopsis line. PCR analysis to confirm the presence of T-DNA was performed using genomic DNA of the Arabidopsis insertion line prx33 (SALK_062314C) and of wild type Col-0. The first 4 lanes show results of PCR amplifications, where primers specific for PRX33 sequence (forward 5'-ATTATAGTTGTTGTCAGCATTAGCA-3', reverse 5'-ACCATTTGTTCCTCTGAAGCA-3') were used with Col-0 and prx33 genomic DNA extracts as templates. The last three lanes exhibit PCR results where a T-DNA left border primer (LBa1:5'-TGGTTCACGTAGTGGGCCATCG -3') was combined with PRX33 sequence specific forward primer using prx33 insertion line genomic DNA extract confirming the location of T-DNA insertion. No template controls were included. PCR products were fractionated in 1% agarose gel and DNA was visualized by staining with GelRed (Fremont, California, USA). Table S1 Arabidopsis T-DNA insertion lines used in this work. Table S2 Primer sequences used in this work.