Catalase and NO CATALASE ACTIVITY1 Promote Autophagy-Dependent Cell Death in Arabidopsis^,,

Catalase directly interacts with and detoxifies reactive oxygen species. This work identifies catalase-deficient mutants in a screen for suppression of cell death and finds that promotion of cell death associated with the plant hypersensitive response requires catalase, suggesting that catalase could act as a direct molecular link between reactive oxygen species and cell death signaling.

Abstract

Programmed cell death often depends on generation of reactive oxygen species, which can be detoxified by antioxidative enzymes, including catalases. We previously isolated catalase-deficient mutants (cat2) in a screen for resistance to hydroxyurea-induced cell death. Here, we identify an Arabidopsis thaliana hydroxyurea-resistant autophagy mutant, atg2, which also shows reduced sensitivity to cell death triggered by the bacterial effector avrRpm1. To test if catalase deficiency likewise affected both hydroxyurea and avrRpm1 sensitivity, we selected mutants with extremely low catalase activities and showed that they carried mutations in a gene that we named NO CATALASE ACTIVITY1 (NCA1). nca1 mutants showed severely reduced activities of all three catalase isoforms in Arabidopsis, and loss of NCA1 function led to strong suppression of RPM1-triggered cell death. Basal and starvation-induced autophagy appeared normal in the nca1 and cat2 mutants. By contrast, autophagic degradation induced by avrRpm1 challenge was compromised, indicating that catalase acted upstream of immunity-triggered autophagy. The direct interaction of catalase with reactive oxygen species could allow catalase to act as a molecular link between reactive oxygen species and the promotion of autophagy-dependent cell death.

INTRODUCTION

Programmed cell death (PCD) is a necessary part of the life of multicellular organisms. It is associated with normal development and with immune responses aimed at pathogen clearance and destruction of otherwise harmful cells. Plant and animal innate immune systems detect conserved microbe-associated molecular patterns, such as flagellin-derived peptides, using pattern recognition receptors (Ausubel, 2005). To circumvent this detection system, adapted plant pathogens inject effectors directly into plant cells. In resistant plants, plant immune receptors known as R proteins detect the presence of effectors. This recognition leads to effector-triggered immunity, which often culminates in a hypersensitive response (HR) leading to accumulation of reactive oxygen species (ROS) and PCD (Jones and Dangl, 2006).

Since the discovery more than two decades ago of HR-associated ROS accumulation (Doke, 1983), also known as the oxidative burst, its causal relationship with PCD has been an active area of research. Over the years, evidence has accumulated that ROS trigger PCD and are not merely cell death byproducts. For instance, pharmacological treatments with ROS antagonists block induction of HR-associated PCD (Lamb and Dixon, 1997), transgenic plants with lower levels of antioxidative enzymes display aggravated PCD responses (Van Breusegem and Dat, 2006), and superoxide induces runaway cell death in Arabidopsis thaliana lesions simulating disease resistance1 mutants (Jabs et al., 1996). These observations highlight the importance of understanding how the plant interprets ROS signaling to induce PCD. Modulation of ROS levels during a HR could result in oxidative modification of countless proteins, emphasizing the complex nature of this topic and the potential difficulties in characterizing the nature of the ROS signal that induces PCD.

Since ROS signaling events resulting in PCD occur downstream of pathogen effector recognition by plant R proteins, genetic identification of signaling components required for R protein–dependent PCD is one possible avenue of investigation. It was revealed that autophagy components are involved in PCD execution downstream of the R protein RPM1 using an avrRpm1-expressing strain of the hemibiotrophic bacterium Pseudomonas syringae pv tomato (Pst) in a reverse genetics approach (Hofius et al., 2009). Furthermore, accumulating evidence suggests the existence of autophagy-dependent cell death pathways across the eukaryotic kingdom (Hofius et al., 2011). Autophagy facilitates catabolism of cellular components and requires the action of autophagy-related genes (ATG) that coordinate membrane rearrangements (Mizushima, 2007). It comprises both unselective bulk degradation of cytoplasmic contents and selective autophagy of specific organelles or proteins. In Arabidopsis, selective autophagy is mediated by the recently identified cargo receptor NEXT TO BRCA1 GENE1 (NBR1), which is itself degraded in the vacuole upon increased autophagic activity (Svenning et al., 2011).

In addition to autophagic components, metacaspases also play an important role in execution of cell death triggered by RPM1 (Coll et al., 2010). Furthermore, a diverse group of PCD signaling components, including an leucine-rich repeat kinase, an Asp protease, and a uridine diphosphate-glucosyl transferase, seems to operate downstream of another R protein from Arabidopsis, RPP4 (Wang et al., 2011). In tomato (Solanum lycopersicum), E3 ubiquitin ligases were required for PCD downstream of Cf-9 (González-Lamothe et al., 2006). However, none of the PCD signaling components identified so far appear to be directly associated with ROS, and it remains an open question how ROS signaling induces PCD in plants.

Originally aimed at understanding the mechanisms of PCD induced by the antineoplastic drug hydroxyurea, we conducted a forward genetic screen for hydroxyurea resistance in Arabidopsis. The screen appeared to select specifically for mutants with compromised activity of the antioxidative, hydrogen peroxide (H₂O₂)–decomposing enzyme catalase, as we recovered 17 independent cat2 mutant alleles. We showed that hydroxyurea and catalase interacted directly, indicating that catalase was involved in drug toxification and offering a likely explanation for the catalase specificity of the screen (Juul et al., 2010).

Here, we describe the characterization of other mutants recovered from the hydroxyurea resistance screen and how they implicate catalase and autophagy components in PCD. Although the vast majority of the hydroxyurea-resistant mutants were catalase deficient, we identified a single resistant mutant with normal catalase activity carrying a loss-of-function allele of the autophagy gene ATG2. We found that atg2 and other autophagy component mutants showed reduced sensitivity to PCD induced by both hydroxyurea and avrRpm1. This dual loss of sensitivity, and the apparent specificity of the hydroxyurea resistance screen for catalase-related events, suggested that modulation of catalase activity might also play a role in avrRpm1-induced PCD. To test this hypothesis, we selected mutants with extremely low catalase activities but without lesions in CATALASE2 (CAT2) and showed that NCA1 is required for the activities of all three catalase isoforms in Arabidopsis. Both cat2 and nca1 mutants displayed loss of both avrRpm1 and hydroxyurea sensitivity. They also accumulated the selective autophagy target NBR1 and ubiquitinated proteins on avrRpm1 challenge, indicating that catalase activity is required to promote autophagy-dependent PCD.

RESULTS

Autophagy Is Required for Cell Death Induction by avrRpm1 and Hydroxyurea

We previously recovered 17 independent CAT2 loss-of-function alleles in a screen for resistance to hydroxyurea-induced cell death (Juul et al., 2010). To identify cell death signaling components acting downstream or independently of catalase, we screened for resistant mutants with normal catalase activity by quantifying the catalase activity levels of all isolated mutants. A single such mutant was found, and SHOREmap (Schneeberger et al., 2009) was used to identify a candidate premature stop mutation at amino acid 626 in the autophagy ATG2 (AT3G19190) gene (see Supplemental Table 1 online). We designated this mutant atg2-3. Confirming correct identification of the causal mutation, the T-DNA line atg2-1 (Yoshimoto et al., 2009, SALK_076727) was also resistant to hydroxyurea (Figure 1). In addition, atg4a atg4b, atg5, and atg7 mutants displayed hydroxyurea resistance (Figure 1), showing that the resistance was not a specific property of the atg2 mutants but rather depended on autophagy deficiency in general. By contrast, ENHANCED DISEASE SUSCEPTIBILITY1 and NON-RACE-SPECIFIC DISEASE RESISTANCE1, which are immune regulators acting downstream of R proteins, were not required for hydroxyurea sensitivity (Figure 1), indicating that the hydroxyurea signal acts independently or downstream of these signaling hubs.

Figure 1.

Hydroxyurea Resistance of Autophagy and Immunity Mutants.

Seedlings germinated in the presence of 3 mM hydroxyurea. Unfolded green cotyledons indicate resistance to cell death induced by hydroxyurea. Ws, Wassilewskija.

[See online article for color version of this figure.]

NCA1 Is Required for Catalase Activity

The atg5 and atg7 mutants were previously reported to show attenuated RPM1-dependent PCD in response to avrRpm1 challenge (Hofius et al., 2009). Since the hydroxyurea resistance screen appeared highly selective for catalase-related events, the reduced hydroxyurea/avrRpm1 sensitivity of the autophagy mutants suggested that abnormal regulation of catalase activity might also interfere with avrRpm1 induced PCD. To test this hypothesis, we screened our collection of hydroxyurea resistant mutants for potential regulators of catalase activity. By sequencing the catalase genes of all mutants with reduced catalase activity, we identified two putative catalase regulator mutants with intact catalase genes but extremely low catalase activities, which we designated nca1-1 and nca1-2.

In Arabidopsis, CAT2 activity is required for dissipation of photorespiration-derived H₂O₂, and the apparent growth retardation of cat2 mutants is alleviated at high CO₂ concentrations as a result of repression of photorespiration (Queval et al., 2007). Similarly to cat2 mutants, the nca1 mutants were resistant to hydroxyurea (Figure 2A) and showed reduced growth at ambient CO₂ levels (Figure 2B), which could be rescued by increasing the CO₂ concentration (Figure 2C). This classified the nca1 alleles as photorespiratory mutants and suggested that their growth retardation phenotype could be explained by catalase deficiency.

Figure 2.

Characterization of the nca1 Mutant Phenotype.

(A) to (C) Wild-type (wt), nca1, and cat2 seedlings grown on one-half-strength MS supplemented with 3 mM hydroxyurea (A), one-half-strength MS at ambient CO₂ level (B), and one-half-strength MS at high CO₂ level (C).

(D) to (G) Material from 12-d-old seedlings. Student's t test (two means) or one-way analysis of variance (ANOVA) and Dunnett's multiple comparison test was used for P value calculations.

(D) Comparison of catalase activity levels. All mutants differed significantly from the wild type (P < 0.0001). Both nca1 mutants differed significantly from cat2-2 (*P < 0.05 and **P < 0.01).

(E) Comparison of catalase transcript levels. White, Col-0 wild type; light gray, cat2-2; dark gray, nca1-1. nd, not detected. Only CAT1 expression in the nca1 background was significantly different from the wild type (*P < 0.05). Error bars indicate se.

(F) Comparison of catalase protein levels as determined by immunoblotting. Three biological replicates are shown for each genotype.

(G) Zymogram showing activities of the different catalase isoforms. Three biological replicates were used, and a representative gel is shown.

(H) NCA1 gene structure. Exons are shown as boxes. Untranslated regions are shown in gray. Mutations are indicated by triangles. See also Supplemental Figures 1 and 2 and Supplemental Tables 2 and 3 online.

[See online article for color version of this figure.]

The nca1 mutants showed catalase activities below 5% of wild-type levels (Figure 2D). However, catalase transcript levels were similar to wild-type levels (Figure 2E), particularly for CAT2, which contributes the majority of catalase activity in green leaves. The amount of catalase protein was markedly decreased in both nca1 mutants (Figure 2F) and the enzymatic activity of all three catalase isoforms was clearly decreased (Figure 2G), indicating that NCA1 is required for the activity of all Arabidopsis catalases and acts mainly at the posttranscriptional level.

Using SHOREmap (Schneeberger et al., 2009), we identified an amino acid change at position 298 in AT3G54360 as a candidate causal mutation in nca1-1 (Figure 2H; see Supplemental Table 2 online). The nca1-2 allele contained a splice site–disrupting mutation in AT3G54360 (Figure 2H). Furthermore, introduction of the AT3G54360 gene into the nca1-1 mutant background rescued the mutant phenotype (see Supplemental Figure 1 online). We conclude that NCA1 is identical to AT3G54360.

The NCA1 gene product contains a Really Interesting New Gene domain (cd00162), and its function has so far not been described. To get an indication whether the requirement of NCA1 for catalase activity is specific to Arabidopsis or more widely conserved, we searched for closely related genes in other plant species by querying the Phytozome plant gene family database (Goodstein et al., 2012). NCA1 is a member of an uncharacterized protein family (family 23249884), with one or two copies per plant genome in land plants (see Supplemental Figure 2 and Supplemental Data Set 1 online). Sequence conservation within the family is high, with 46% amino acid identity between NCA1 homologs from Arabidopsis and from the moss Physcomitrella patens (see Supplemental Table 3 online). Validating extensive evolutionary conservation of NCA1 gene function, the P. patens NCA1 cDNA complemented the Arabidopsis nca1-1 mutant phenotype (see Supplemental Figures 1A and 1B online).

The nca1 and cat2 Mutants Display Similar Morphological Alterations, Oxidative Stress Responses, and Proteome Compositions

Because of the dramatic drop in catalase activity seen in the nca1 mutants, we hypothesized that NCA1 might directly interact with catalase. Using NCA1 and CAT2 fusion protein constructs that complemented the respective mutant phenotypes, we found that CAT2, as expected, localized to punctate structures that most likely represent peroxisomes. NCA1, by contrast, localized to the cytosol and nucleus, making a direct interaction with CAT2 unlikely (Figure 3A). Since plant catalases localize to peroxisomes and their import depends on C-terminal peroxisomal targeting signals (Mullen et al., 1997), a second possibility was that NCA1 could influence peroxisomal integrity and thereby reduce catalase activity. However, we detected signals from the peroxisomal marker PTS1-GFP (for green fluorescent protein; Mano et al., 1999, 2002) in the nca1-1 mutant background that appeared similar to those in the wild type, indicating that peroxisomes remained functional, at least with respect to PTS1-guided protein import (Figure 3A).

Figure 3.

NCA1 and CAT2 Protein Localization and Mutant Phenotypes.

(A) Localization of the fusion proteins indicated in the top left corner in the genetic backgrounds shown in the bottom left corner. The NCA1-eYFP and CAT2-eYFP fusion proteins were driven by their native promoters, and PTS1-GFP was expressed from the cauliflower mosaic virus 35S promoter. Bars = 10 µm.

(B) Four-week-old plants grown in short days on soil.

(C) Response to photorespiration-promoting growth conditions. The three leftmost panels show maximum quantum efficiency of photosystem II (Fv/Fm) on a heat map scale where red represents the lowest values. The experiment was performed in six replicates, and a representative plate is shown.

(D) Expression of oxidative stress marker genes. Fold changes with respect to the untreated Col-0 sample are shown. Gray bars indicate expression prior to photorespiratory stress treatment. Blue bars show expression after 24 h of treatment. Three replicates were used, and error bars indicate se. Two-way ANOVA and Dunnett's multiple comparison test were used for calculating P values indicating the probability that an expression value was identical to the Col-0 value at the given time point (*P < 0.05 and **P < 0.01).

(E) Log₂ fold changes in protein abundances as determined by iTRAQ proteomics are plotted. Fold change values are based on three biological and four technical replicates. All quantified proteins were included in the plot. wt, wild type.

Since these localization experiments failed to identify an intimate2 association between catalase and NCA1, beyond the strongly reduced catalase activity in the nca1 mutants, we performed a more detailed phenotypic analysis of the nca1 and cat2 mutants to evaluate their level of similarity. When grown on soil in short days, the nca1 and cat2 mutants were morphologically very similar, both displaying hyponastic leaves and reduced growth compared with the wild type (Figure 3B). Since both mutants were catalase deficient, we compared their oxidative stress responses. We imposed oxidative stress using a bioassay in which photorespiration-dependent H₂O₂ accumulation is induced by restricted gas exchange and simultaneous exposure to constant light conditions in Petri dishes. In contrast with the wild type and atg2, both nca1 and cat2 displayed decreased maximum quantum efficiency of photosystem II (Fv/Fm) within 24 h and died after 8 d of treatment (Figure 3C). Consistent with this aggravated response, both nca1 and cat2 displayed a stronger induction of expression of the oxidative stress marker genes FACTOR EXCHANGE FOR SSA1 PROTEIN1A and URIDINE DIPHOSPHATE GLYCOSYLTRANSFERASE 74E2 (Tognetti et al., 2010; Inzé et al., 2012) than did the wild type and atg2 after 24 h of oxidative stress treatment (Figure 3D). The hypersensitivity to oxidative stress for both mutants is consistent with the rescue of their growth retardation phenotype by increased CO₂ levels, which suppress photorespiration (Figures 2B and 2C).

These strong similarities indicated that the morphological aberrations and oxidative stress responses of the nca1 mutant could be explained by catalase deficiency. However, it was still possible that NCA1 could affect the abundance of other proteins in addition to catalase, which would result in protein level perturbations exclusive to the nca1 mutant. To investigate this possibility, we compared the proteomes of short-day-grown wild-type, nca1, and cat2 plants using isobaric tags for relative and absolute quantitation (iTRAQ) labeling. We detected and quantified a total of 933 proteins (see Supplemental Data Set 2 online). For comparison with the catalase immunoblots and activity measurements (Figures 2D, 2F, and 2G), we first looked more closely at the peptides unique to the three catalase isoforms. We identified no unique CAT1 peptides. CAT2 peptides were not detected in the cat2 mutant and were found at ∼6% of wild-type levels in nca1 mutants, whereas CAT3 peptides were detected at reduced levels in both nca1 and cat2 mutants (see Supplemental Tables 4 and 5 online). The CAT2 peptide data matches the in-gel CAT2 activity data, showing reduced activity in nca1 and no detectable activity in cat2 (Figure 2G). The reduction in CAT3 peptide abundance matches the reduced activity seen in nca1, whereas cat2 appeared to have no reduction in the catalase activity derived from CAT1 and CAT3 (Figure 2G). This apparent discrepancy could be explained either by a large contribution from CAT1, which we were unable to quantify using our proteomics approach or by different levels of activity of the residual CAT3 protein in the two mutants. To evaluate the overall similarity of the nca1 and cat2 proteomes, we then compared their global changes in protein abundance with respect to the wild-type control. The fold changes in the two mutants were highly correlated (Figure 3E) with a Pearson correlation coefficient of 0.66. Inspecting the proteins displaying the highest fold changes for each mutant, we found that ∼50% of the top 80 regulated proteins overlapped between nca1 and cat2, further underlining the similarity of the two mutant proteomes (see Supplemental Data Set 3 online).

Catalase Activity Is Required for Cell Death Induction by avrRpm1 and Hydroxyurea

Based on morphology, oxidative stress response, and proteome characteristics, the nca1 mutant phenotype could be explained by catalase deficiency. NCA1 was thus specifically required for the activity of all catalase isoforms in Arabidopsis. These properties make the nca1 mutant well suited for testing our hypothesis that catalase activity is required for sensitivity to both hydroxyurea and avrRpm1-induced cell death. We therefore quantified the avrRpm1 cell death response in short-day-grown plants and found that both cat2 and nca1 mutants showed an attenuated response. The nca1 mutant displayed a stronger reduction in the PCD response, consistent with its lower catalase activity levels (Figure 4A). Overexpression of NCA1 from the cauliflower mosaic virus 35S promoter in the nca1 mutant background partially rescued the PCD response, but it did not result in cell death levels higher than those seen in the wild type (see Supplemental Figure 1 online). Restriction of catalase activities through transcriptional repression of NCA1 does therefore not appear to play a role during avrRpm1-induced PCD. We also found a strongly attenuated cell death response for the atg2 mutant (Figure 4), confirming that it had lost both its sensitivity to hydroxyurea and its ability to induce cell death in response to avrRpm1.

Figure 4.

Suppression of Pst DC3000 avrRpm1–Mediated Cell Death in Catalase- and Autophagy-Deficient Mutants.

Ion leakage measurements of the wild type (Col-0), rpm1-3, cat2-2, nca1-1, and atg2-1 after inoculation with a (OD₆₀₀ = 0.2) culture of avirulent Pst DC3000 expressing avrRpm1.

(A) Four-week-old short-day-grown plants.

(B) Four-week-old short-day-grown plants shifted to long days 3 d prior to infection. Mean and se were calculated from four leaf discs per treatment, with four replicates per experiment.

Although the attenuated avrRpm1 cell death response of the catalase-deficient mutants agrees well with their identification in our screen for resistance to hydroxyurea-induced cell death, it is controversial because the opposite phenotype, exaggerated cell death responses, has previously been reported for the cat2 mutant (Queval et al., 2007; Chaouch et al., 2010). According to these studies, the cat2 exaggerated cell death phenotype is daylength dependent. Since the plants used for our cell death assay were grown in short days, but the cat2 cell death phenotype manifests itself in longs days (Queval et al., 2007; Chaouch et al., 2010), we investigated whether a difference in light conditions could explain the discrepancy. Indeed, shifting the plants to long days 3 d prior to avrRpm1 challenge resulted in an aggravated response in the cat2 mutant, while cell death remained suppressed in nca1 (Figure 4B).

Cell Death and Bacterial Resistance Are Uncoupled in nca1, cat2, and atg2 Mutants

Since the catalase- and autophagy-deficient mutants showed an attenuated cell death response, plant immunity to bacterial infection could also be compromised. However, measuring bacterial growth after infiltration showed that this was not the case. As expected, RPM1 was required for immunity, but the nca1, cat2, and atg2 mutants did not display increased bacterial growth compared with the wild type (Figure 5). Reminiscent of the Arabidopsis type I metacaspase mutant mc1 (Coll et al., 2010), the nca1, cat2, and atg2 mutants therefore uncouple PtoDC3000 (avrRpm1)–induced cell death from induced resistance.

Figure 5.

Pst DC3000 avrRpm1 Resistance Characteristics of Catalase and Autophagy-Deficient Mutants.

Growth of Pst DC3000 avrRpm1 in leaves of 5-week-old plants at 0 and 2 d after infiltration with 1 × 10⁵ CFU mL⁻¹. Log-transformed values are means ± sd (n = 3). Two-way ANOVA and Dunnett's multiple comparison test were used for calculating P values indicating the probability that a CFU value was identical to the Col-0 value at the given time point (****P < 0.0001).

Catalase Acts Upstream of Autophagy in the avrRpm1 Response

To determine the relative positions of catalase and autophagy within the PCD signaling network, we first investigated the autophagy properties of catalase mutants, and vice versa, to confirm that PCD was suppressed due to defects in distinct processes in the two groups of mutants. Autophagy is induced during starvation and senescence, and lack of autophagic components leads to hypersensitivity to nutrient limitation (Doelling et al., 2002). Leaf detachment induces senescence (Weaver et al., 1998), whereas exposure to prolonged darkness leads to carbon starvation (Thompson et al., 2005). In contrast with the atg2-1 mutant, nca1-1 and cat2-2 did not show exaggerated senescence in response to leaf detachment or prolonged darkness (Figures 6A and 6B). This argues that autophagy starvation and senescence responses are not compromised in the nca1 and cat2 mutants and that the reduced cell death response on avrRpm1 challenge is not caused by a general lack of autophagy competence in these mutants. Conversely, the atg mutants all showed catalase activities at or above wild-type levels (Figure 6C), allowing us to rule out that the atg mutants displayed reduced hydroxyurea and avrRpm1 sensitivity as a result of catalase deficiency. Autophagy competence and catalase activity therefore both appear to be required for full sensitivity to hydroxyurea and avrRpm1-induced cell death.

Figure 6.

Catalase and Autophagy Interactions.

(A) Prolonged darkness assay. Left panel, prior to dark treatment; middle panel, after 10 d in darkness; right panel, after 10 d in darkness and 10 d of recovery in light. wt, the wild type.

(B) Detached leaves incubated on moist filter paper for 3 d.

(C) Catalase activities of immune effector and autophagy mutants. Error bars show se, and three replicates were used. P values indicating the probabilities that the activity levels were identical to the wild-type level were calculated using a one-way ANOVA followed by Dunnett's test for multiple comparisons. (*P < 0.05 and ****P < 0.0001).

(D) NBR1 immunoblot on extracts from plants prior to and 7 h after inoculation (hpi) with Pst DC3000 avrRpm1. Arrowhead indicates the position of the NBR1 band. The bottom panel shows the Ponceau-stained membrane.

(E) Ubiquitin immunoblot on extracts from plants prior to and 7 h after inoculation with Pst DC3000 avrRpm1. The top panel shows ubiquitinylated proteins. The top and bottom arrowheads indicate sizes of 55 and 60 kD, respectively. The bottom panel shows the Ponceau-stained membrane.

[See online article for color version of this figure.]

To investigate if catalase acts upstream of autophagy in the avrRpm1 response, we examined the immunity-induced autophagy response of the catalase-deficient mutants. Arabidopsis NBR1 is a target of selective autophagy that accumulates in autophagy-deficient mutants (Svenning et al., 2011). Prior to avrRpm1 challenge, the NBR1 levels were similar in the wild type, nca1, and cat2 mutants, whereas the atg2 mutant showed strong NBR1 accumulation (Figure 6D). This reflects that even in the absence of starvation or senescence challenge, a basal level of autophagy is needed to prevent NBR1 accumulation. This basal activity is absent in the atg2 mutant but still present in other genotypes, confirming the autophagy competence of the nca1 and cat2 mutants. After avrRpm1 challenge, the NBR1 level remained low in protein extracts from wild-type plants 7 h after infection (Figure 6D). In marked contrast, NBR1 had accumulated to high levels in protein extracts from nca1, cat2, and atg2 mutants (Figure 6D). To confirm the aberrant immunity-induced autophagy phenotype of the catalase-deficient mutants, we assayed the abundance of ubiquitinated proteins, which accumulate in autophagy-deficient mutants under heat stress (Zhou et al., 2013). In contrast with the wild type, we found that nca1, cat2, and atg2 all displayed accumulation of ubiquitinated proteins 7 h after infection (Figure 6E). The differential accumulation of both NBR1 and ubiquitinated proteins shows that the autophagic degradation triggered by RPM1 is severely inhibited in the catalase-deficient mutants and indicates that catalase acts upstream of immunity-triggered autophagy during promotion of PCD (Figure 7).

Figure 7.

Tentative Model for Catalase Integration of PCD Signals.

In our screen for hydroxyurea resistance, the death of wild-type plants at germination coincides with ROS generation as a byproduct of β-oxidation of fatty acids during seed storage oil mobilization. These ROS activate catalase, which interacts directly with hydroxyurea to produce an autophagy-dependent PCD response. AvrRpm1 challenge causes a HR comprising an oxidative burst. The ROS produced activate catalase, which reacts with unknown compounds to generate an autophagy-dependent PCD response. Resistance to Pst DC3000 avrRpm1 is induced independently of the PCD response mediated by catalase and autophagy proteins. Cell death inducers (red) and genetic components (blue) investigated in this study are highlighted.

DISCUSSION

Autophagy: Promotion or Attenuation of PCD?

It was the identification of the autophagy atg2-3 mutant in our screen for resistance to hydroxyurea-induced cell death combined with its reduced sensitivity to avrRpm1-elicited PCD that led us to suspect a similar dual loss of sensitivity for the catalase-deficient mutants recovered from the same screen. Other autophagy components are required for PCD downstream of R protein effector recognition (Hofius et al., 2009) and display hydroxyurea resistance (Figure 1). However, autophagy is also important for removal of cellular waste products, and autophagy deficiency leads to premature senescence accompanied by cell death (Yoshimoto et al., 2009). Interestingly, limiting the spread of pathogen-induced PCD appears to depend on autophagy too, suggesting a negative, rather than positive, role in PCD regulation (Liu et al., 2005; Patel and Dinesh-Kumar, 2008; Yoshimoto et al., 2009). The uncontrolled spread of HR PCD in autophagy mutants was observed only in older leaves of 7- to 8-week-old plants and not in younger tissues (Yoshimoto et al., 2009). It could thus reflect that a tissue already poised for senescence is pushed over the edge by a pathogen-induced HR. These observations suggest that different ages of the plant material used in the cell death assays offer a likely explanation for the apparent discrepancies in the reports on autophagy-dependent regulation of PCD.

Our identification of an autophagy component in a forward genetic screen for resistance to hydroxyurea-induced cell death and the strong suppression of avrRpm1-induced PCD seen in the atg2 mutant, which also shows a stronger premature senescence phenotype than other autophagy mutants (Yoshimoto et al., 2009), emphasizes the role of autophagy in promotion of PCD.

Catalase Acts as Both an Antioxidative Enzyme and a PCD Effector

In contrast with autophagy, catalase activity has so far exclusively been associated with attenuation of cell death and not with its promotion. Since ROS can induce PCD and damage cellular proteins and membranes, loss of antioxidative enzyme activity, including that of catalase, could be detrimental to the plant and result in aggravated cell death responses. There is a large body of evidence to support this view. In Arabidopsis, this includes observations that cat2 mutants display a photoperiod-conditional cell death phenotype, which is dependent on salicylic acid production and manifests itself under long-day conditions (Queval et al., 2007; Chaouch et al., 2010). These results indicated that catalase deficiency induces an HR-like PCD under growth conditions favoring accumulation of peroxisomal H₂O₂, which led the authors to suggest that suppression of catalase activity may play a role in establishing the oxidative burst and executing PCD during pathogen defense (Chaouch et al., 2010). Light-dependent lesions were also observed in earlier studies using tobacco (Nicotiana tabacum) catalase antisense lines. In addition, aggravated cell death responses and increased resistance to pathogens was observed (Chamnongpol et al., 1998; Mittler et al., 1999; Dat et al., 2003).

Our results now reveal that catalase has an additional, and opposite, role in promoting PCD. This conclusion is supported both by the identification of mutant alleles of catalase and NCA1 in our genetic screen for resistance to hydroxyurea-induced cell death and by the suppression of avrRpm1-induced cell death observed for these mutants. The PCD-promoting role of catalase has likely been masked in previous studies by the use of high-stress light conditions. As we have shown here, altering the daylength during plant growth is sufficient to change the nature of the cat2 cell death response (Figure 4).

Catalase requires activation by one molecule of H₂O₂ to enter its activated state, where it can either react with a second molecule of H₂O₂ or with other substrates (Chance, 1947). We suggest that during a HR, catalase is activated by high levels of H₂O₂ and reacts with an unknown substrate to generate a PCD-activating signal (Figure 7). If catalase levels are insufficient, the PCD signal is suppressed. According to this working model, the cat2 mutant, which retains some residual catalase activity, would still be able to induce PCD in response to pathogen challenge, and the response would be compounded by combination with severe oxidative stress (Figure 4). By contrast, the very low residual catalase activity in the nca1 mutant would be insufficient to cause immunity-induced PCD even in combination with high oxidative stress levels (Figure 4B). The residual catalase activity in the cat2 mutant is derived from CAT1 and/or CAT3, which reside in a tandem repeat on Arabidopsis chromosome 1. Based on available microarray data, CAT3 expression appears to overlap with that of CAT2 in leaves and seeds, whereas CAT1 shows weaker expression in these tissues but is strongly expressed in pollen (Schmid et al., 2005; Winter et al., 2007). One or both of these other catalase isoforms likely also contribute to the catalase-dependent PCD response. This conclusion is based on the overall similarity of the nca1 and cat2 phenotypes combined with the clear differences in catalase activities in the two mutants. However, we cannot rule out that the nca1 mutant displays more strongly reduced avrRpm1 sensitivity than cat2 because of the regulation of other unknown targets in addition to the catalases.

Is Catalase a Direct Molecular Link between ROS and PCD Signaling?

The similarities between the PCD responses induced by hydroxyurea and avrRpm1 are striking. First, NCA1, CAT2, and autophagy genes make up a common set of genetic components required for the PCD responses to both inducers. Second, as catalase interacts directly with hydroxyurea (Juul et al., 2010), it likely acts upstream of autophagy components in hydroxyurea-induced PCD, as it does in PCD induced by avrRpm1 (Figures 6D to 6F). Third, the PCD responses induced by both hydroxyurea and avrRpm1 are associated with generation of ROS. On avrRpm1 challenge, ROS are produced during the HR oxidative burst (Doke, 1983). In the hydroxyurea resistance screen, the death of the wild-type plants at ∼2 d after germination coincides with ROS production during the respiratory burst accompanying the β-oxidation of fatty acids typical for storage oil mobilization in oilseed plants (Graham, 2008). After germination, Arabidopsis plants are much less sensitive to hydroxyurea, as evidenced by their survival on medium containing 25 mM hydroxyurea in the hydroxyurea uptake assay described by Juul et al. (2010).

Although hydroxyurea and catalase interact directly, and hydroxyurea acts as a competitive inhibitor of H₂O₂ decomposition (Juul et al., 2010), hydroxyurea does not induce PCD by overloading the system with ROS through catalase inhibition, since loss-of-function catalase mutations are resistant to hydroxyurea. The most likely explanation is instead that the combination of catalase activation by H₂O₂ and the presence of hydroxyurea generate a signal, which activates autophagy-dependent PCD. Considering the similarities between PCD induced by hydroxyurea and avrRpm1, the catalase-dependent signal activating autophagy-dependent PCD upon avrRpm1 challenge could well be similar in nature to the signal resulting from the interaction between catalase and hydroxyurea. Validation of this working model (Figure 7) will require identification of the putative signaling compounds, which is challenging because of the radical chemistry involved. However, the association of hydroxyurea, catalase, and autophagy with PCD paves the way for further advances in our understanding of the causal relationship between the oxidative burst and PCD in plant defense.

METHODS

Plant Materials and Growth Conditions

The hydroxyurea resistance screen was performed as described (Juul et al., 2010), screening for germination at a concentration of 3 mM hydroxyurea. Plants for catalase assays, transcript quantification, and immunoblots were grown on plates as previously described (Juul et al., 2010) and harvested 12 d after germination. Plants for ion leakage and detached leaf assays were grown on soil as described (Hofius et al., 2009). Long-day growth conditions were equivalent to the short-day conditions except for the 16-h daylength. Plants for CO₂ rescue were grown in growth chambers at 22°C, 16/8-h light/dark cycles, and 150 µmol m⁻²s⁻¹ photosynthetic photon flux at either 0.04% (ambient) or 0.4% (elevated) CO₂ levels on one-half-strength Murashige and Skoog (MS) medium (Duchefa) solidified with 0.9% agar. For the prolonged darkness assay, plants were grown as described (Juul et al., 2010) for 12 d, incubated for 10 d in darkness, and finally moved back to normal light conditions for 10 d.

Induction of Oxidative Stress and Fv/Fm Quantification

We used a bioassay in which photorespiration is induced by restricting gas exchange within Petri dishes. For the photorespiration-promoting conditions, plants were grown on one-half-strength MS agar medium for 2.5 weeks. Then, the plates were transferred to a continuous light regime after replacing the surgical tape (Micropore; 3M) that sealed the plates by two layers of Parafilm M (Alcan) to restrict gas exchange. The maximum efficiency of the photosystem II photochemistry (Fv/Fm) was determined using a PAM-2000 chlorophyll fluorometer.

Analysis of Catalase Activities and Transcript and Protein Abundances

Catalase activity was quantified as described (Beers and Sizer, 1952). For catalase transcript quantification, RNA was extracted from 12-d-old seedlings using the Nucleospin RNA plant kit (Macherey Nagel). Following DNase I treatment, cDNA was prepared using the RevertAid first-strand cDNA synthesis kit (Fermentas). As internal controls, the transcript levels of two previously characterized control genes At5g25760 and At1g13320 (Czechowski et al., 2005) were quantified. Primers are listed in Supplemental Table 6 online. For quantification of the oxidative stress marker genes FACTOR EXCHANGE FOR SSA1 PROTEIN1A (At3g09350) and URIDINE DIPHOSPHATE GLYCOSYLTRANSFERASE74E2 (At1g05680), RNA isolation, cDNA synthesis, and quantitative RT-PCR analyses were performed as described (Tognetti et al., 2010). Immunoblots and zymogram analyses were performed as previously described (McClung, 1997; Vandenabeele et al., 2004). For NBR1 and ubiquitin, protein was extracted in urea buffer from plants syringe-infiltrated with 2 × 10⁸ colony-forming units (CFU) mL⁻¹ of avirulent Pst avrRpm1 and detected with anti-NBR1 antibody as described (Svenning et al., 2011) or with anti-ubiquitin antibody (Zhou et al., 2013). Statistical analyses were performed using GraphPad Prism 6.0.

SHOREmap Causal Mutant Identification

The nca1-1 and atg2-3 causal mutations were identified using the SHOREmap approach (Schneeberger et al., 2009). Landsberg erecta was used as the mapping parent, and pools of 499 (nca1-1) and 323 (atg2-2) F2 plants were sequenced to ∼7× genome coverage.

Mutant Complementation and Protein Localization

The NCA1 promoter was amplified from Columbia-0 (Col-0) genomic DNA and cloned into a pGreen0029 (Hellens et al., 2000) vector carrying a Gateway (GW) cassette using restriction sites KpnI and HindIII followed by enhanced yellow fluorescent protein (eYFP; Shaner et al., 2004) amplification and cloning into the pGreen0029-pNCA1-GW plasmid using the HindIII site to construct a pGreen0029-pNCA1-eYFP-GW (pMC-00637) plasmid. The same strategy was used for cloning the CAT2 promoter to construct the pGreen0029-pCAT2-eYFP-GW (PMC-00642) plasmid. NCA1 genomic DNA excluding promoter was amplified from Col-0 genomic DNA and recombined into the pMC-00637 plasmid to create the pGreen0029-pNCA1-eYFP-NCA1 (pMC-00638, pNCA1:NCA1) in-frame fusion expression construct. NCA1 cDNA was amplified from the U13379 clone, obtained from the ABRC, and then recombined into a pEarleyGate104 vector (Earley et al., 2006) to create the pEarleyGate-35S-YFP-NCA1 (pMC-00570, 35S:NCA1) construct. Physcomitrella patens NCA1 cDNA was amplified from RNA extracted as described (Cove et al., 2009) and recombined into pEarleyGate104 to create the pEarleyGate-35S-YFP-PpNCA1 (pMC-00588, 35S:PpNCA1) construct. CAT2 cDNA was amplified from RNA extracted as described (Cove et al., 2009) and recombined into PMC-00642 to create the pGreen0029-pCAT2-eYFP-CAT2 (PMC-00647) expression construct. Live tissue was observed under a Zeiss 510 Meta confocal microscope to image fusion protein localization.

Conductivity Measurements and Resistance Assays

Bacterial ion leakage assays following syringe infiltration of avirulent Pst avrRpm1 were performed with 2 × 10⁸ CFU mL⁻¹, and bacterial growth was determined at days 0 and 2 after infection with 1 × 10⁵ CFU mL⁻¹ as described (Mackey et al., 2003). The rpm1-3 null mutant (Grant et al., 1995) was used as control.

iTRAQ Proteomics

Tissue was collected from wild-type, cat2, and nca1 plants grown for 3.5 weeks in short days. Protein was extracted in PBS buffer, ultracentrifuged, phenol extracted, precipitated, and then redissolved. Next, disulfide bonds were reduced and protein was alkylated prior to trypsin digestion. The resulting peptides were purified and dried. After redissolving, peptides were quantified and labeled using 4-plex iTRAQ reagents (Applied Biosystems) followed by HILIC fractionation on a microHPLC system. The labeled peptides were then subjected to liquid chromatography–tandem mass spectrometry analysis using a LTQ-Orbitrap XL mass spectrometer (Thermo Fisher Scientific) for data collection. Finally, the raw data were deconvoluted and converted into peak lists by ProteomeDiscoverer software version 1.2 (Thermo Scientific). The peak lists were then used for searches against the Arabidopsis thaliana proteome using Mascot version 2.3. See Supplemental Methods 1 online for full details.

Accession Numbers

Sequence data from this article can be found in the Arabidopsis Genome Initiative database under the accession numbers listed in Supplemental Table 7 online.

Supplemental Data

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

• Supplemental Figure 1. Complementation of the nca1 Mutant Phenotype.

• Supplemental Figure 2. NCA1 Phylogeny.

• Supplemental Table 1. SHOREmap Output for the atg2-3 Mutant.

• Supplemental Table 2. SHOREmap Output for the nca1-1 Mutant.

• Supplemental Table 3. Comparison of NCA1 Protein Sequences across Species.

• Supplemental Table 4. Unique Catalase Peptides.

• Supplemental Table 5. iTRAQ Proteomics Experiment Setup.

• Supplemental Table 6. Oligonucleotide Sequences.

• Supplemental Table 7. Accession Numbers.

• Supplemental Data Set 1. NCA1 Protein Alignment.

• Supplemental Data Set 2. iTRAQ Proteomics Results.

• Supplemental Data Set 3. Top Up- and Downregulated Proteins for the nca1 and cat2 Mutants.

• Supplemental Methods 1. Phylogenetic Analysis and iTRAQ Methods.

Glossary

PCD

programmed cell death

HR

hypersensitive response

ROS

reactive oxygen species

Pst

Pseudomonas syringae pv tomato

H₂O₂

hydrogen peroxide

MS

Murashige and Skoog

CFU

colony-forming unit

Col-0

Columbia-0

ANOVA

analysis of variance