DspA/E-Triggered Non-Host Resistance against E. amylovora Depends on the Arabidopsis GLYCOLATE OXIDASE 2 Gene

DspA/E is a type three effector injected by the pathogenic bacterium Erwinia amylovora inside plant cells. In non-host Arabidopsis thaliana, DspA/E inhibits seed germination, root growth, de novo protein synthesis and triggers localized cell death. To better understand the mechanisms involved, we performed EMS mutagenesis on a transgenic line, 13-1-2, containing an inducible dspA/E gene. We identified three suppressor mutants, two of which belonged to the same complementation group. Both were resistant to the toxic effects of DspA/E. Metabolome analysis showed that the 13-1-2 line was depleted in metabolites of the TCA cycle and accumulated metabolites associated with cell death and defense. TCA cycle and cell-death associated metabolite levels were respectively increased and reduced in both suppressor mutants compared to the 13-1-2 line. Whole genome sequencing indicated that both suppressor mutants displayed missense mutations in conserved residues of Glycolate oxidase 2 (GOX2), a photorespiratory enzyme that we confirmed to be localized in the peroxisome. Leaf GOX activity increased in leaves infected with E. amylovora in a DspA/E-dependent manner. Moreover, the gox2-2 KO mutant was more sensitive to E. amylovora infection and displayed reduced JA-signaling. Our results point to a role for glycolate oxidase in type II non-host resistance and to the importance of central metabolic functions in controlling growth/defense balance.


Introduction
During evolution, plants have developed a sophisticated and multi-layered immune system that enables them to perceive and counteract potential invaders. The basal line of defense relies on the recognition of structural microbial components, i.e., bacterial flagellin, fungal chitin and lipopolysaccharides, commonly designated as microbe-or pathogen-associated molecular patterns (MAMPs or PAMPs) [1], by specific cell surfacelocalized receptors named Pattern Recognition Receptors (PRRs) [2,3]. A classic example of a PAMP-PRR interaction is the recognition of the conserved twenty-two amino acid epitope of flagellin, defined as FLG22, by the leucine-rich repeat receptor-like kinases (LRR-RLKs) Flagellin-sensitive 2 (FLS2) [4]. This line of defense, defined as PAMP-Triggered Immunity (PTI), consists in a complex signaling network mediated by mitogen-activated

Identification of DspA/E Suppressor Mutants
We previously generated and characterized two transgenic lines, 13-1-1 and 13-1-2, expressing DspA/E under the control of an estradiol-inducible promoter [25]. Estradiolinduction of DspA/E in A. thaliana is toxic for plants ( Figure 1a) and ultimately leads to plant cell death. In order to identify the mechanism(s) by which DspA/E triggers plant cell death, we subjected line 13-1-2 to EMS, which induces random point mutations in the genome. Approximately 125,000 M0 seeds of 13-1-2 were treated overnight with either 0.2% or 0.4% EMS, rinsed in water, sterilized and grown in vitro for two weeks. Viable seedlings were transferred to soil and the M2 progeny was recovered (Figure 1b). M2 seedlings were sown in vitro on 10 nM estradiol, a condition in which original 13-1-2 seeds do not germinate due to DspA/E toxicity (Figures 1a and 2a). Each putative mutant was backcrossed to the parental line and the progeny was sown on estradiol to identify recessive mutations ( Figure S1a).  [25]) were sown on MS medium (1% sucrose) with the indicated concentration of estradiol. Pictures were taken two weeks after sowing; (b) Outline of DspA/E suppressor mutant selection. The 13-1-2 transgenic line was EMS mutagenized, and the F2 progeny was sown on 10 nM estradiol to identify DspA/Esuppressor mutants. Further molecular characterization was performed to identify mutants in which DspA/E was expressed at least at the same level as in the parental 13-1-2 line. Created with Biorender (https://biorender.com/ accessed on 1 February 2022).

Phenotypical Characterization of Suppressor Mutants
We identified three putative recessive mutants (I-18, II-36 and I-40) in which the DspA/E transgene was expressed at a level similar to the parental 13-1-2 line (Figure 2b). The three mutants were crossed with each other, and progeny were tested for their resistance to DspA/E. The progeny of crosses from mutants I-18 and II-36 were all resistant to DspA/E toxicity since they germinated on estradiol ( Figure S1), indicating that these two mutations constitute a complementation group. The progeny of all crosses with the third mutant, I-40, were sensitive to estradiol, indicating that this mutant did not belong to the same complementation group as mutants I-18 and II-36 ( Figure S1b). The two mutants belonging to the same complementation group, I-18 and II-36, were selected for further analysis.
The 13-1-2 transgenic line displays several phenotypes associated with the expression and the toxicity of DspA/E including inhibition of de novo protein synthesis and inhibition of root elongation [25]. In order to determine which aspects of DspA/E toxicity were affected in the two selected suppressor mutants, I-18 and II-36, we compared their phenotypes with that of the 13-1-2 parental line following estradiol-induced DspA/E expression. Root tips of one-week-old seedlings were treated with estradiol for 10 min and root growth was measured 24 h after treatment. As described previously, the 13-1-2 line showed a complete inhibition of root growth per day following the induction of DspA/E expression (Figure 3a). Following estradiol treatment, both suppressor-mutants showed root growth similar to mock-treated control conditions (without estradiol), indicating that these mutants were resistant to the toxicity of DspA/E for root growth (Figure 3a). We then analyzed the capacity of DspA/E to suppress de novo protein synthesis using 35S-labeled methionine. As previously described [25], when DspA/E expression was induced in the parental 13-1-2 line, de novo protein synthesis was stopped 3 h following the estradiol treatment ( Figure 3b). In both suppressor mutants, de novo protein synthesis was not blocked following the induction of DspA/E expression, indicating that these mutants are resistant to the toxic effect of DspA/E on de novo protein synthesis. Altogether, our data show that DspA/E toxicity and associated cellular phenotypes are abolished in both suppressor mutants identified. (a) Dose-response of root growth inhibition in the parental 13-1-2 line and in the two selected suppressor mutants. Root growth of mutants is not affected by dspA/E expression in mutants I-18 and II-36. Bars represent standard deviation (SD), n = 20 seedlings; similar results were observed in at least three independent experiments. Stars indicate significant difference to untreated control according to Mann-Whitney (p-value < 0.05); (b) Translation is not inhibited by DspA/E in suppressor mutants. Ten-day-old seedlings were treated with DMSO (-) or 5 µM estradiol (+) for 3 h and labelled with 35S methionine. Total protein extracts (10 µg per sample) were separated by SDS-PAGE and stained with Coomassie blue (bottom panel); 35S methionine incorporation was detected by autoradiography (top panel). Similar results were obtained for two biological replicates. The picture corresponds to a single gel.

Identification of Causal Mutations in the Glyoclate Oxidase 2 Gene
In order to identify the causal mutations leading to suppression of DspA/E toxicity, we adapted the Shoremap strategy described previously [26]. For this, we crossed each suppressor mutant with the parental 13-1-2 line. The F1 progeny of this cross was then selfpollinated and the F2 progeny was analyzed to identify individuals resistant to DspA/E toxicity. After sowing on 100 nM estradiol, 100 F2 individuals that germinated on estradiol, and thus resistant to DspA/E toxicity, were sampled and pooled for genomic sequencing. This was performed in parallel for both the I-18 and the II-36 suppressor mutants.
Illumina sequencing of the pools of F2 individuals bearing the I-18 and II-36 suppressor mutations was performed at a 40× depth. Sequence analysis was performed using the CLC genomics workbench software. In parallel, the parental 13-1-2 line was sequenced in order to eliminate single nucleotide polymorphisms (SNPs) already present. We then identified the SNPs present in the F2 pool of each suppressor mutant. We found a region of chromosome 3 in which the percentage of SNPs in both suppressor mutants was between 80 and 100% ( Figure S2). This region contained five candidate genes and detailed analysis of the SNPs present in both suppressor mutants showed that only one gene had SNPs leading to non-synonymous mutations in both suppressors; the At3g14415 gene encoding GLYCOLATE OXIDASE 2 (GOX2). In the II-36 mutant the identified SNP led to a Ser27Phe mutation in the GOX2 protein, while in the I-18 mutant the identified SNP led to an Ala96Val mutation (Figure 4a,b). A multiple sequence alignment analysis showed that both mutations altered conserved amino acids of GOX enzymes from different plant species, including a GOX enzyme from Malus domestica (Figure 4c). In order to confirm that mutations in the GOX2 gene were responsible for the loss of DspA/E toxicity in A. thaliana seedlings, we crossed both the parental 13-1-2 line and the suppressor mutants with the gox2-2 T-DNA KO line described previously [27]. When crossed with the parental 13-1-2 line, the F1 progeny were unable to germinate on estradiol, as expected since the gox2-2 mutation is recessive. When the gox2-2 mutant was crossed with the I-18 suppressor mutant, the F1 progeny was able to germinate on estradiol ( Figure 5a). Similar results were obtained when the gox2-2 mutant was crossed with the II-36 suppressor. Thus, the gox2-2 KO mutant was unable to functionally complement the suppressor mutants for the susceptibility to DspA/E toxicity. This result is consistent with the fact that DspA/E toxicity in A. thaliana requires GOX2. Figure 5. gox2-2 mutants are allelic to suppressor mutants and mutations in GOX2 suppress DspA/E toxicity in A. thaliana. (a) Seedlings were grown for 7 days directly on 5 µM estradiol. The control line (wt) and the three parental lines are shown on the left. As expected, the parental 13-1-2 line does not germinate on estradiol while the three other lines do. The F1 progeny of the 13-1-2 x gox2-2 cross does not germinate on estradiol while the F1 progeny of the I-18 x gox2-2 cross is able to germinate on 5 µM estradiol. All genotypes were able to germinate on MS without estradiol. Each cross was performed at least in triplicate and 2 to 5 seeds per cross were tested. A representative seedling is shown for the DspA/Eresistant genotypes; (b) GOX activity increases following infection of 5-week-old rosette leaves inoculated with wild-type E. amylovora (Ea). This increase is not observed in response to the Ea dspA/E-deficient mutant (d) and is not observed in the A. thaliana gox2-2 mutant. m: mock; Ea: wild-type E. amylovora; d: E. amylovora dspA/E-deficient mutant. For each condition, four pools of three rosette leaves were analyzed; (c) GOX activity in the suppressor mutants. GOX activity was measured in vitro. It was strongly reduced in the gox2-2 mutant and in the II-36 suppressor mutant. In the I-18 we observed a slight reduction that was not significant. For each condition, four pools of 20 seedlings were analyzed; (d): GOX activity of GOX2 recombinant proteins bearing mutations found in I-18 (A96V) and II-36 (S27F) mutants. (b-d): * indicates significant difference with the wild-type (Mann-Whitney, p-value < 0.05).
In A. thaliana, GOX enzymes, involved in photorespiration, have been attributed to a family of five genes; however, only three of them have been shown to convert glycolate to glyoxylate; GOX1-3 [28]. The GOX2 protein has been shown to be involved in nonhost resistance [27], and this role is believed to be linked to ROS production since GOX enzymatic activity generates H 2 O 2 [29,30]. We tested the global GOX activity in A. thaliana leaf extracts inoculated or not with E. amylovora. We found that in response to E. amylovora infection, there was an increase in the extractable GOX activity of leaf extracts (Figure 5b). Since A. thaliana rosettes contain both GOX1 and GOX2 [28], it was not possible to determine whether the observed increase in GOX activity is essentially due to GOX2 alone. However, smaller increase in GOX activity in the gox2-2 following inoculation with E. amylovora could indicate that this is the case (Figure 5b). We then compared leaf GOX activity in wild-type plants and the gox2-2, I-18 and II-36 mutants. As expected, the gox2-2 KO mutant showed a 47% reduction in GOX activity compared to the control wild-type line (Figure 5c). This was also true for the II-36 suppressor mutant in which a 30% reduction in GOX activity was observed. In the I-18 suppressor mutant the leaf GOX activity was 15% lower than in wild-type; however, this difference was not significant, suggesting that the I-18 mutation might not affect global GOX activity.
2.5. GOX2 Is Required to Induce JA-Dependent Defense and Confers Non-Host Resistance to E. amylovora GOX2 has been previously shown to be involved in non-host resistance [27]. We therefore analyzed the response of the gox2-2 mutant to E. amylovora infection. As expected, gox2-2 was more susceptible to E. amylovora (Figure 6a). This was correlated with a reduction in electrolyte leakage in the mutant in response to E. amylovora infection (Figure 6b), which is consistent with a reduction in defense response reminiscent of type II non-host. We then analyzed the expression of two defense-signaling marker genes, CHI-B and PR1, both previously shown to be induced in wild-type plants in response to E. amylovora infection [19]. We found that the SA-dependent PR1 gene was unaffected in the gox2-2 mutant while the JA-dependent CHI-B gene was less induced in the gox2-2 mutant background ( Figure 6c). Furthermore, GOX enzymes catalyze reactions that oxidize different 2-hydroxy acid substrates and all produce H 2 O 2 . In the case of GOX2, H 2 O 2 production occurs in response to infection by non-adapted pathogens [27]. However, we did not find any difference in response to E. amylovora in terms of H 2 O 2 accumulation detected by DAB staining at the level of gox2-2 leaves when compared to wild-type leaves (Figure 6d).
T3Es are injected into the plant cytoplasm; however, their final localization within the plant cell can be in the cytosol and/or in different organelles. Motif search in the DspA/E protein sequence showed the presence of several organelle-targeting signals including a peroxisome targeting signals (PTS) [31]. However, previous works trying to establish the intracellular localization of DspA/E were unsuccessful, probably due to the toxicity of this protein for eukaryotic cells [25,32,33]. Since GOX2 suppresses DspA/E toxicity, it was decided to check the subcellular localization of the GOX2 protein. Its revpresumed peroxisomal localization is based on the presence of a typical C-terminal tripeptide required to address proteins into peroxisomes [34] and its presence in peroxisomal proteomes [35] but its subcellular localization has not been confirmed by other means. We constructed a GOX2::GFP fusion under the native GOX2 promoter and transformed protoplasts obtained from a stable transgenic line bearing a peroxisomal RFP marker [36]. Our results confirm that GOX2 is indeed peroxisomal (Figure 7).

DspA/E-Triggered Cell Death Is Associated with Strong Metabolite Changes
In order to better understand the mechanisms underlying both DspA/E triggered cell death and its suppression by the two suppressor mutants, we performed a metabolomic analysis by GC-MS. Seedlings of the 13-1-2 parental line and of the two suppressor mutants were either mock-treated (without estradiol) or treated with increasing estradiol concentrations. The results show that DspA/E expression in the 13-1-2 parental line leads to depletion in metabolites associated with the TCA cycle, such as fumarate and citrate ( Figure 8). In the 13-1-2 parental line an accumulation of metabolites associated with cell death and defense such as pipecolate and stigmasterol was observed ( Figure 8). Other metabolites showed an altered accumulation in response to DspA/E expression such as salicylate and several organic acids ( Figure 9). Interestingly, TCA cycle intermediates accumulated more in the suppressor mutants than in the 13-1-2 parental line. On the contrary, cell-death and defense-associated metabolites showed levels in the suppressor mutants comparable to the control line (Figures 8 and 9). Altogether, our data show that expression of DspA/E in A. thaliana seedlings leads to important modifications of the metabolome and that part of these modifications of metabolite levels in response to DspA/E expression were lost in both suppressor mutants (Figures 8 and 9).

Discussion
The pathogenicity of E. amylovora depends on its T3SS which injects type three effectors (T3Es) inside the plant cell; among them, the T3E DspA/E plays a major role since a DspA/E-deficient mutant is non-pathogenic [23]. In host plants, DspA/E causes a rapid oxidative burst associated with disease [38,39]. In the model plant A. thaliana, E. amylovora is able to multiply transiently in a T3SS-dependent manner [40]. However, A. thaliana can restrict bacterial growth by triggering an active type II non-host resistance that includes callose deposition, cell death and expression of defense signaling pathway genes [19]. In a previous study, we generated the transgenic A. thaliana line 13-1-2 expressing dspA/E under an estradiol-inducible promoter and showed that DspA/E plays an important role in triggering non-host resistance in A. thaliana [25]. However, the exact mechanisms by which DspA/E triggered cell death in the line 13-1-2 remained to be elucidated.
In the present study, we EMS mutagenized the 13-1-2 line and identified two independent suppressor mutants, I-18 and II-36, that were resistant to DspA/E-inhibition of seedling germination. Further characterization of these two mutants showed that they were rescued for all phenotypes associated with DspA/E expression that were tested: inhibition of germination, inhibition of root elongation, and inhibition of de novo protein synthesis. To further understand the mechanisms at play, we performed a metabolomic analysis of the 13-1-2 line and of the two suppressor mutants (Figure 8). Interestingly, we observed that the TCA cycle was strongly affected in the 13-1-2 line when DspA/E was expressed, which could explain the lack of germination of transgenic seeds in these conditions [41,42]. In contrast, both suppressor mutants showed higher accumulation of TCA cycle intermediates such as fumarate, 2-oxoglutarate, malate and succinate (Figure 8), thus perhaps explaining their capacity to germinate while expressing DspA/E. Indeed, there is increasing evidence of that metabolites such as TCA cycle intermediates can act as signaling intermediates [43]. Another key feature of the metabolomic response of A. thaliana seedlings to the expression of DspA/E was the accumulation of several metabolites associated with cell death and defense. For example, pipecolate, a catabolite of lysine, that can induce resistance to P. syringae pv. maculicola [44], accumulated strongly in response to DspA/E expression. Stigmasterol, known to reduce membrane fluidity and permeability also accumulated strongly in response to DspA/E expression [45]; DspA/E expression triggers strong electrolyte leakage and it is possible that stigmasterol accumulation is a means for the plant to control this leakage. All of these metabolites accumulated less in the two suppressor mutants, indicating that these metabolites could play a key role in the signaling of DspA/E-triggered defense and/or in the coping with damage to the plant cells associated with DspA/E expression.
Whole genome sequencing of the two suppressor mutants allowed us to determine that they each carried different missense mutations in the A. thaliana GOX2 gene, which encodes a glycolate oxidase involved in photorespiration [28]. GOX2 belongs to a family of five genes: GOX1, GOX2, GOX3, HAOX1 and HAOX2 [30]. Previous analysis of single knock-out lines for GOX1 and GOX2 and an artificial microRNAi line targeting both genes in Arabidopsis showed that these two genes were the major photorespiratory genes in leaves [28]. Conversely, other works showed that GOX3 encoded for a lactate oxidase involved in roots metabolism during hypoxia [46], while HAOX1 and HAOX2 encoded for long-chain 2-hydroxy acid oxidases mainly expressed in seeds [47]. Nevertheless, all GOX genes seem to play a role in the non-host resistance against P. syringae pv. tabaci in Arabidopsis leaves, since GOX3, HAOX1 and HAOX2 were significantly induced after 24 h of infection with this pathogen [27]. Here, we showed that, following E. amylovora infection, overall leaf GOX activity can be significantly induced in a DspA/E-dependent manner in Col-0 plants while it was not the case in the gox2-2 mutant (Figure 5b). Interestingly, the two suppressor lines harboring different mutations in the coding region of GOX2 genes had also a lower overall leaf GOX activity (Figure 5c).
GOX2 has been identified in the proteome of peroxisomes [34,35], and here confocal microscopy using a peroxisomal RFP marker and a GOX2::GFP construct allowed us to confirm that GOX2 is indeed a peroxisomal protein. Both suppressor mutations modified a different conserved amino acid of the GOX2 protein. Interestingly, GOX has already been shown to be involved in non-host resistance against P. syringae pv. tabaci [27]. In this study, ROS accumulation in the leaf blade in response to a non-adapted pathogen was lower in knock-out gox mutants. GOX2 is part of a five member gene family comprising GOX1, GOX2, GOX3, HAOX1, and HAOX2 and perhaps surprisingly all GOX family knockout mutants exhibited a reduction in ROS accumulation in response to the non-adapted pathogen P. syringae pv. tabaci [27]. We found no major difference in macroscopic accumulation of ROS in the gox2-2 background using the non-adapted pathogen E. amylovora, suggesting that in this case, there were GOX2-independent sources of ROS. Indeed, we showed previously that ROS production in response to E. amylovora in A. thaliana is strongly dependent on the RBOHD NADPH oxidase [48]. The KO gox2-2 mutant showed a higher susceptibility to E. amylovora, showing the involvement of this enzyme in type II non-host resistance against a necrotrophic pathogen. Interestingly, we found normal expression of the SA-dependent PR1 gene in the gox2 mutant background but low expression of the JA-signaling marker gene CHI-B. Since JA biosynthesis is in part localized in the peroxisome [49], it could be speculated that perturbation of peroxisomal functions in the gox2 mutant alters the capacity of the plant to synthesize JA and/or compromises JA signaling upon pathogen attack. Therefore, we show that in addition to being involved in non-host resistance against hemi-biotrophic pathogens, GOX2 is also involved in type II non-host resistance against necrotrophic pathogens.
In this study, we show that A. thaliana gox2-2 mutants are resistant to the toxicity of DspA/E and more susceptible to E. amylovora. This suggests that, in the non-host context, DspA/E toxicity is correlated with defense activation. This is interesting as the role of the toxicity of DspA/E in plant cells during the infection process is still unclear to date. However, the precise role of GOX2 in both the toxicity of DspA/E and the susceptibility to E. amylovora remains to be determined. The gox2-2 mutant showed reduced expression of CHI-B, a JA-signaling marker gene, indicating reduced defense activation in this genetic background, which could explain the increased susceptibility to E. amylovora. We also observed that in response that in response to E. amylovora infection there is an increase in total GOX activity, which was strongly reduced in the gox2-2 mutant. This result suggests that the peroxisome could be targeted during the infection process.
Finally, GOX2 is an enzyme involved in photorespiration, an energy-costly process, which can limit crop yield, especially under low CO 2 conditions that can be produced by biotic-stress related stomatal closure. It is well known that there is a trade-off between growth and defense in plants [50]. In the presence of DspA/E and in response to E. amylovora infection, the growth-defense equilibrium could be altered. One explanation for the suppression of DspA/E toxicity by the mutation of the GOX2 enzyme could be an alteration of growth-defense equilibrium. In this scenario, loss of GOX2 and associated defense reactions could re-equilibrate the defense/growth balance in favor of growth. Forward genetics is a powerful strategy to uncover unexpected links between different cellular processes. Central metabolic functions and specialized metabolites are increasingly shown to play a crucial role in plant-pathogen interactions. Further studies will be necessary to uncover the precise links between these processes.

Plant Material and Growth Conditions
In vitro growth was performed on 1× Murashige and Skoog (MS) medium supplemented with 1% sucrose in growth chambers under a long day light regime (16 h light/8 h dark) at 25 • C (day) and 20 • C (night).
The 13-1-2 A. thaliana transgenic line, described previously, bears the dspA/E gene under the control of an estradiol-inducible promoter [25]. The gox2-2 mutant bears a T-DNA insertion in the 5 UTR region of the AT3G14415 gene and has been shown to be a knock-out mutant [27]. All the plant material used in this study is in the Col-0 accession.

Bacterial Strains and Pathogen Infection
The bacterial strains used in this study are the wild-type CFBP1430 strain of Erwinia amylovora and the M81 mutant defective for DspA/E [25].
For pathogen infections, rosette leaves of 5-week-old plants were infiltrated with E. amylovora CFBP1430 or the M81 dspA/Emutant using a needleless syringe. Bacterial suspensions were prepared in sterile water (10 7 CFU/mL; OD 600 = 0.1). Twenty-four hours post infection (hpi), we performed bacterial counting by grinding infected leaves using glass beads in a TissueLyser (Qiagen/Retsch, Hilden, Germany). The bacterial suspensions were used to prepare serial dilutions, which were plated on an LB medium, and after 1 or 2 days the colonies formed were counted to evaluate the initial number of bacteria.

EMS Mutagenesis and Screening
EMS mutagenesis of the 13-1-2 line was performed as described previously [51,52]. Briefly, approximately 125,000 seeds (2.5 g) were treated with 0.2 or 0.4% EMS overnight, rinsed in water, sterilized and grown in vitro for two weeks. Viable seedlings were transferred to soil in growth chambers (16 h light/8 h dark) at 25 • C (day) and 20 • C (night) and the M1 progeny was recovered in 150 bulks of approximately 10 plants each. The screening was performed by sowing on MS supplemented with 1% sucrose and 10 nM estradiol. Seedlings able to germinate in these conditions were transferred to fresh MS without estradiol for a few days and then transferred to soil to collect their progeny.

Genome Sequencing of Suppressor Mutants and Analysis
The selected suppressor mutants were backcrossed with the parental 13-1-2 line and the F2 progeny was analyzed. For each suppressor mutant, 100 F2 individuals able to germinate on 10 nM estradiol were sampled and pooled. For each pool, DNA was extracted and whole genome Illumina ® sequencing at a 40× depth was performed. The sequences were mapped to the A. thaliana Col-0 reference genome using the CLC genomics workbench software. In parallel, the parental line 13-1-2 was sequenced to identify the SNPs it carried, these SNPs were then eliminated from our analysis of the genomes of the suppressor mutants.

Recombinant GOX2 Production and Site-Directed Mutagenesis of AtGOX2
To produce recombinant GOX2 proteins, the previously described pET28a-AtGOX2 expression plasmid [53] was used as a template to introduce the desired point mutations by PCR using specific primer pairs (Table S1) and the QuikChange ® II XL site-directed mutagenesis kit (Agilent ® , Les Ulis, France), according to the manufacturer's instructions. This strategy generated AtGOX2-S27F and AtGOX-A96V mutated proteins. All constructions were subsequently verified by DNA sequencing using T7 and T7-term primers (Table S1).
The N-terminal His-tagged GOX2 proteins were purified by affinity chromatography as previously described [28]. The purity of each recombinant GOX2 protein was checked by SDS-PAGE (10% acrylamide) stained with Coomassie Brilliant Blue [54].
Recombinant GOX2 activities were measured using 5 µg of purified recombinant

Glycolate Oxidase Activity In Vitro Assay
Glycolate oxidase activity was measured in vitro on protein extracts as described previously [28,53]. Briefly, leaf samples were ground in liquid nitrogen using a TissueL-yserII (Qiagen ® , Retsch, Hilden, Germany) and resuspended in Tris-HCl (50 mM, pH 8). Protein extracts were purified using a NAP-5 Sephadex G-25 DNA grade column (GE Healthcare ® , Uppsala, Sweden). Enzyme activity was measured in Tris-HCl (50 mM, pH 8) with glycolate (10 mM) and 300 µg of purified total protein by an enzyme-coupled reaction at 30 • C. Glycolate-dependent H 2 O 2 production was quantified in the presence of 0.4 mM o-dianisidine and 2 units of horseradish peroxidase by measuring the ∆A440 nm using a Varian Cary 50 spectrophotometer.

RNA Isolation and qRT-PCR
Total RNA was extracted from 100 mg of frozen ground leaves or seedlings using Trizol ® reagent (Invitrogen Life Technologies, Saint-Aubin, France). The RNA quality was evaluated by electrophoretic run on 1% agarose gel. First-strand cDNA was synthesized using Superscript reverse transcriptase SSII (Invitrogen, Saint-Aubin, France) from 1 µg of DNase-treated (Invitrogen, Saint-Aubin, France) total RNA in a 20 µL reaction volume. qPCR reactions were performed using SYBR ® Selected MasterMix 2x (Applied Biosystem, Villebon Sur Yvette, France), following the manufacturer's protocol. The cycling conditions consisted of an initial 5 min at 95 C, followed by 40 three-step cycles at 94 • C for 15 s, 60 • C for 30 s, and 72 • C for 30 s. Melting curve analysis was performed after cycle completion to validate amplicon identity. Relative expression levels were calculated following the standard curve-based method [37]. Expression of the Adenosine Phosphoribosyl Transferase 1 (APT1; AT1G27450) reference gene was used for normalization [55]. For each treatment, three biological replicates, corresponding to a pool of 4 leaves from a single plant or to 20 seedlings, were analyzed and each qRT-PCR reaction was carried out in duplicate; the complete experiment was conducted twice independently, and one representative experiment is presented. The gene-specific primers used in this analysis are described in [56].

Detection of ROS
Diaminobenzidine (DAB) staining was used to detect intra-and extracellular hydrogen peroxide (H 2 O 2 ). Five-week-old leaves were collected 2 h after inoculation and vacuuminfiltrated with DAB (Sigma Aldrich, Saint Louis, MI, USA) (1 mg/mL, pH 3.7). Leaves were then placed in a wet Petri dish overnight and the staining was stopped at 16 hpi using ethanol to discolor them. 2 ,7 -Dichlorofluorescein diacetate (DCFH-DA) staining was used to detect intracellular hydrogen peroxide. A 30 mM DCFH-DA (Sigma Aldrich, Saint Louis, MI, USA) solution was prepared in DMSO and diluted 100 times in deionized water. Inoculated or mock-treated leaves were collected 16 h post-inoculation (hpi) and vacuum-infiltrated with DCFH-DA. Leaves were immediately put on a microscope slide and fluorescence emission was observed under a binocular magnifier with a GFP filter (510 nm).

GOX2::GFP Construct
Plasmid pda09796 containing the GOX2 gene sequence was obtained from the RIKEN institute. The GFP coding sequence was cloned in 3 of the GOX2 coding sequence using the gateway cloning system. The primers used to perform the cloning are detailed in Table S1 (start gox2-2-F; stop gox2-2-R; end gox2-2-R).

Protoplast Production and Transfection
Protoplasts were prepared from 14-day-old A. thaliana seedlings stably transformed with a peroxisomal marker [36] as previously described [57]. A. thaliana mesophyll protoplasts were transfected with 2.5 µg of the GOX2::GFP construct. The protoplasts were imaged by confocal laser scanning microscopy after 24 h of incubation in the dark at room temperature.

Confocal Laser Scanning Microscopy
Images of fluorescent protoplasts were obtained with a Leica TCS-SP2-AOBS spectral confocal laser scanning microscope equipped with a Leica HC PL APO lbd.BL 20.0x 0.70 water immersion objective. GFP and chloroplasts were excited with the 488 nm line of an argon laser (laser power 40%) while the RFP was excited with the 543 nm line (laser power 30-40%). Fluorescence emission was detected over the range 495 to 540 nm for the GFP construct, 590 to 640 nm for the RFP construct and 670 to 730 nm for chloroplast autofluorescence. Images were recorded and processed using LCS software version 2.5 (Leica Microsystems). Images were cropped using Adobe Photoshop CS2 and assembled using Adobe Illustrator CS2 software (Abode, http://www.abode.com accessed on 1 February 2022).

Metabolite Analysis
Metabolome analysis was performed as previously described [58]. Four pools of 20 seedlings per condition were used for metabolome analysis. Approximately 30 mg of the ground frozen seedling samples was analyzed by an Agilent 7890A gas chromatograph (GC) coupled to an Agilent 5975C mass spectrometer (MS). Standards were injected at the beginning and end of the analysis. Data were analyzed with AMDIS (http://chemdata. nist.gov/mass-spc/amdis/ accessed on 5 December 2013) and QuanLynx software (Waters Corp., Milford, MA, USA).

Structural Model of AtGOX2
The structural model presented in Figure 4b is based on the 3D-structure of S. oleracea GOX (PDB 1AL7) [59].