Proteome-Wide Analysis of Cysteine Reactivity during Effector-Triggered Immunity^,

Increase in the level of reversible oxidation states during the plant immune response affects cysteines in a wide range of proteins, follows a wave pattern with a maximum at eight hours after pathogen infection, and is modulated by the oligopeptidases TOP1 and TOP2.

Abstract

A surge in the accumulation of oxidants generates shifts in the cellular redox potential during early stages of plant infection with pathogens and activation of effector-triggered immunity (ETI). The redoxome, defined as the proteome-wide oxidative modifications of proteins caused by oxidants, has a well-known impact on stress responses in metazoans. However, the identity of proteins and the residues sensitive to oxidation during the plant immune response remain largely unknown. Previous studies of the thimet oligopeptidases TOP1 and TOP2 placed them in the salicylic acid dependent branch of ETI, with a current model wherein TOPs sustain interconnected organellar and cytosolic pathways that modulate the oxidative burst and development of cell death. Herein, we characterized the ETI redoxomes in Arabidopsis (Arabidopsis thaliana) wild-type Col-0 and top1top2 mutant plants using a differential alkylation-based enrichment technique coupled with label-free mass spectrometry-based quantification. We identified cysteines sensitive to oxidation in a wide range of protein families at multiple time points after pathogen infection. Differences were detected between Col-0 and top1top2 redoxomes regarding the identity and number of oxidized cysteines, and the amplitude of time-dependent fluctuations in protein oxidation. Our results support a determining role for TOPs in maintaining the proper level and dynamics of proteome oxidation during ETI. This study significantly expands the repertoire of oxidation-sensitive plant proteins and can guide future mechanistic studies.

Plant stress is responsible for most crop yield loss and is also one of the greatest challenges to overcome in modern agriculture (Dubé et al., 2012). It is critical to understand how plants sense and integrate diverse stress signals to elicit physiological responses for survival and, in the long-term, to reach a predictive understanding of the adaptation and tolerance mechanisms.

Under normal physiological conditions, reactive oxygen species (ROS) produced in various cellular compartments such as chloroplasts and mitochondria, are essential to the plant growth and development; cellular ROS concentration is maintained at low levels due to the tight regulation of their production and degradation through enzymatic and nonenzymatic processes (Noctor and Foyer, 2016). Infection with pathogens perturbs ROS homeostasis and lead to oxidant accumulation at above-physiological levels followed by localized shifts in the reduction-oxidation (redox) potential of the apoplast, cytosol, and intracellular compartments (Suzuki et al., 2012; Han et al., 2013). Cellular oxidants have a strong impact on the proteome, whereby reactive protein cysteines are oxidized—a process well documented in metazoans (Yang et al., 2016), and in the initial stages of being evaluated and understood in plants (Akter et al., 2015b). Under physiological conditions or stress Cys thiols (R-SH) may deprotonate to reactive thiolates (R-S⁻) and become modified by ROS to reversibly oxidized products such as sulfenic acid (R-SOH) and disulfides (R-SS-R′). Further oxidation of sulfenic acids may lead to irreversible products such as sulfinic acid (R-SO₂H) and sulfonic acid (R-SO₃H). Following oxidation of Cys residues, a protein’s functional and structural characteristics can change significantly; thus, ROS-mediated modification of cysteines provides an important protein post-translational regulatory mechanism. It is considered that ROS-sensing mechanisms and signal transduction rely on such oxidative modifications of redox sensor proteins (Noctor and Foyer, 2016).

The study of oxidant agents and their effects on proteins in the context of plant interactions with pests and pathogens has identified proteins that act as sensors and enhanced our understanding of redox signaling and the molecular mechanisms driving it (Spoel and Loake, 2011). Plant infection with avirulent pathogens triggers the synthesis of salicylic acid (SA) and strong localized oxidative bursts in multiple cellular compartments—events considered central to the plant immune response (Mou et al., 2003). A positive feedback loop between SA and ROS synthesis is maintained until a threshold is reached for the activation of SA signaling and the effector-triggered immunity (ETI; Rustérucci et al., 2001). Several components in the SA-mediated pathways are regulated by changes in the cellular redox homeostasis (Couturier et al., 2013). Investigations from our and other groups of two Arabidopsis (Arabidopsis thaliana) thimet oligopeptidases—the organellar TOP1 and cytosolic TOP2—revealed their roles as participants in redox processes and as likely subjects of redox regulation in the context of the ETI and chloroplast-originating oxidative stress (Polge et al., 2009; Kmiec et al., 2013; Moreau et al., 2013; Westlake et al., 2015). Both oligopeptidases are required for plant defense to avirulent strains of Pseudomonas syringae through the activation of the resistance proteins RPS2 or RPS4, and necessary to regulate the programmed cell death (PCD). In a current model, TOPs sustain interconnected organelle and cytosol proteolytic pathways that regulate the ETI oxidative burst and plant resistance to pathogens (Kmiec et al., 2013; Westlake et al., 2015).

Mass spectrometry has a great potential to allow the systematic exploration of the plant redoxome. Several recent studies report proteome-wide mining of hydrogen peroxide (H₂O₂)-sensitive cysteines in several plant species (Alvarez et al., 2011; Wang et al., 2012b; Muthuramalingam et al., 2013; Liu et al., 2014; Slade et al., 2015), and collectively, highlight the broad impacts of ROS on plant proteomes. However, the composition of the plant redoxome in response to plant-produced ROS, the identity of redox sensors, and their contribution to plant defense and adaptive pathways remain largely unknown. Mass spectrometry-based methodologies are needed to permit the quantitative and site-specific mapping of redox-mediated modification of protein amino acid residues. Moreover, applying proteome-wide mining to biological material undergoing physiologically relevant oxidative stress responses would help to determine the biological functions of thiol sensitivity in particular proteins or cellular pathways.

In this study, we enriched proteins with oxidative post-translational modification (oxPTM) for the identification and quantification of reversible Cys oxidation in the Arabidopsis proteome and characterized the thiol redoxomes of Col-0 wild-type and the top1top2 null mutant plants at the early stages of the ETI response.

RESULTS

Strategy for Quantitative Profiling of Reversible Cys Oxidation in Plants Undergoing the Effector-Triggered Immune Response

We have developed a strategy to study changes in protein Cys oxidation occurring during the ETI “oxidative burst” (Fig. 1A). The null mutant top1top2 is known to be more susceptible than the wild-type Col-0 to the bacterial pathogen P. syringae DC3000 pv. tomato (Pst) carrying the avirulence gene avrRpt2 (Moreau et al., 2013). To assess the cellular redox stress and the progression of the hypersensitive response (HR) cell death in top1top2, we quantified ion leakage in leaf tissue of plants inoculated with Pst avrRpt2 (Fig. 1B). Infiltration of Pst avrRpt2 at high inoculum density (10⁶ CFU/mL) into leaves of Col-0 produced maximum conductivity values at 8 h postinoculation (hpi), which were sustained at 12 hpi, as previously shown (Katagiri et al., 2002). By comparison, top1top2 plants had significantly reduced levels of ion leakage at 8 hpi and increased levels at 12 hpi. The results suggest a defect in the progression of the ETI oxidative burst in top1top2. No discernable increase in conductivity was recorded in the rps2 control, or any of the genotypes analyzed when infiltrated with MgCl₂ (Supplemental Fig. S1A). Next, to characterize the protein oxidation events triggered by the ETI oxidative burst and gain insight into the probable causes of top1top2 susceptibility, we inoculated Col-0 and top1top2 with Pst avrRpt2 (10⁵ CFU/mL), the condition under which the difference in pathogen growth between Col-0 and top1top2 was found as statistically significant (Moreau et al., 2013). The pathogen- and mock (buffer)-inoculated leaf tissue was collected for processing and mass spectrometry analysis at 8 hpi and 12 hpi in biological triplicates.

Figure 1.

Label-free quantitative analysis of the in vivo oxidation status of the Arabidopsis proteome. A, Diagram of the experimental strategy used in this study. For the determination of the in vivo oxidation status of Cys residues in the Arabidopsis proteome, plants grown under long-day conditions (16 h light/8 h dark) were inoculated with buffer-only (Mock) or with Pst avrRpt2 and tissue samples were collected at 8 hpi and 12 hpi. Following protein extraction, peptides with reversible redox modifications at cysteines were enriched (as described in C), and peptides were identified based on fragment ions detected in LC-MS/MS. B, Development of the HR in Arabidopsis plants undergoing the effector-triggered immune response. Various plant genotypes (Col-0, rps2, and top1top2) were inoculated with Pst avrRpt2 and development of cell death during the HR was monitored by measuring the electrolyte leakage from inoculated tissue at multiple time points (hpi). Asterisk (*) represents significant differences (P < 0.01) in comparison to that in Col-0 at each respective time point (Student’s t test parameters: two-tailed distribution, two-sample unequal variance). Four replicates were analyzed for each infiltration group per time point per line. C, Workflow for the protein-level enrichment of reversible Cys oxidation. Free thiols were blocked by IAM during cell lysis. Blocked proteins were reduced, and nascent protein thiols were then covalently bound to the TPS6B resin. D, Workflow for processing Cys-bound redox proteins for LC-MS/MS analysis. Enriched Cys-peptides were eluted from the resin using DTT, analyzed via LC-MS/MS, and processed through an informatics pipeline for site-specific, global profiling of the reversibly oxidized proteome. RT, retention time.

Successful redox proteomics experiments are founded on the steps taken to preserve the in vivo thiol status of cellular proteins. During cell lysis and sample preparation, redox-active enzymes and artificial oxidation can perturb thiol-disulfide equilibria and lead to misinterpretation of results (Hansen and Winther, 2009). To address this issue, free thiols were blocked by iodoacetamide (IAM), an irreversible alkylating agent, under denaturing conditions during cellular lysis to prevent artificial oxidation of in vivo reduced cysteines (Fig. 1C). Blocked proteins were reduced using dithiothreitol (DTT) before buffer exchange for the removal of the excess reducing agent. Nascent protein thiols, formerly in vivo reversibly oxidized cysteines, were then covalently bound to the Thiopropyl Sepharose 6B (TPS6B) resin by a thiol-disulfide exchange reaction. To ensure complete alkylation, we incubated blocked proteins with TPS6B resin before DTT reduction to bind any left-over reduced Cys and analyzed the proteins by SDS-PAGE (Supplemental Fig. S1B). This negative control was observed to have appreciably low background, especially compared to samples reduced with DTT before enrichment, thereby demonstrating efficient alkylation of in vivo reduced Cys. Proteins bound were on-resin digested with trypsin, and enriched Cys-peptides were eluted using DTT and analyzed via liquid chromatography–mass spectrometry (LC-MS/MS; Fig. 1D).

The ETI-Triggered Reversible Cys Redoxomes of Col-0 and top1top2

LC-MS/MS data were processed through a statistical pipeline for label-free site-specific global profiling of the reversibly oxidized proteome (Supplemental Fig. S2A). The data analysis pipeline consists of: (1) data normalization and variance-stabilization, (2) data filtering, (3) missing data imputation, (4) linear model analysis (limma), (5) differential analysis using limma output and fold change criteria, (6) multiple imputation and binomial testing, (7) time-series clustering analysis, (8) structure-reactivity clustering analysis of Cys residues, and (9) functional Gene Ontology (GO) annotation analysis. The first peptide analysis steps follow the implementation of the peptide-centric proteomics pipeline described in Berg et al. (2019), which has demonstrated very good performance for analysis of reversible oxidized cysteines on a benchmark dataset. An important result from our previous study was the significant performance improvement obtained when using a linear model in combination with multiple data imputations for comparative analysis of large-scale reversible Cys oxidation datasets. For missing values, we performed 100 multiple imputations generating 100 unique datasets. Data imputation was performed by sampling values from a normal distribution with parameters robustly estimated from the entire dataset. We use the limma method (Ritchie et al., 2015) with a significance threshold of 0.05 and a log fold change of ±1 to identify differentially oxidized cysteines together with multiple imputations to decide on statistically significant changes in proteome quantitation due to missing data (Supplemental Table S1). Comparisons among data sets were performed at 8 hpi and 12 hpi with the pathogen for the two genotypes under treatment (T) or control conditions (C), to detect changes in Cys oxidation. Pearson correlation was calculated for each unique combination of samples (Supplemental Fig. S2B) demonstrating the high reproducibility of the label-free quantitation (LFQ) methodology across biological replicates. Each comparison distinguished oxidized Cys-sites with significantly changing abundance (i.e. the proportion of proteins bearing a particular Cys becoming reduced or oxidized with decreasing or increasing abundance, respectively).

We detected significant changes in Cys oxidation between and within genotypes. Oxidized cysteines relative abundance spanned six orders of magnitude, demonstrating a wide range in the level of oxidation of individual residues within the proteome and the sensitivity of LFQ. Overall, in Col-0, we identified 2,113 oxidation-sensitive sites and mapped them to 913 Universal Protein Resource (UniProt) identifiers; for top1top2, 1,942 oxidation-sensitive sites were mapped to 823 UniProt identifiers; 338 peptides mapping to 226 proteins had significantly changed abundance in top1top2 (Fig. 2A). Col-0 and top1top2 were most similar at 12 hpi as shown by both Cys- and protein-centric analyses; the largest differences were calculated for 8h-C versus 8-hpi and 8h-C versus 12-hpi sets (Supplemental Fig. S2, C and D). Further analysis compared the proportion of the not-significant cysteines (first bars in Supplemental Fig. S2, C and D) to the 10 biggest sets of significant cysteines (bars 2–11 in Supplemental Fig. S2, C and D) with the equivalent sets in the proteins, and indicated that oxidation-sensitive cysteines were dispersed over a large number of proteins rather than aggregated onto a few. Overall, most peptides contained only one site of oxidation (88%); likewise, 60% of identified proteins had just one unique site.

Figure 2.

Col-0 and top1top2 Cys redoxomes. A, Venn diagrams representing the overlap of peptides with experimentally identified sensitive cysteines in Col-0 and top1top2 plants undergoing the effector-triggered immune response, and oxidized peptides with statistically significant differential accumulation (“Differential”) in controls and after inoculation with the pathogen. The numbers represent numbers of peptides with potentially sensitive Cys residues. B, Comparative analysis of oxidation-sensitive proteins from this study with published sets. The bar plot shows the 40 largest sets (x axis) containing common genes unique to the selected datasets (y axis) represented as shaded (black) dots. The inset bar plot shows the total number of proteins identified in each study.

Finally, we compared our redoxome with sets from related studies (Fig. 2B). A comparison with the H₂O₂ Cys-sensitive redoxomes generated found that 40% of the datasets from Akter et al. (2015a) and 35% of the datasets from Wang et al. (2012b) were identified in the Col-0 and top1top2 redoxomes. Our data sets also included proteins, albeit at lower percentages, from the Arabidopsis stress-induced s-Glutathionylated dataset (Dixon et al., 2005), thiol-disulfide proteome of the chloroplast revealed by differential electrophoresis (Ströher and Dietz, 2008), OxiTRAQ redox-sensitive proteome (Liu et al., 2014), seed oxidized proteome (Job et al., 2005), sulfenome (Waszczak et al., 2014), and the s-nitrosylated proteome (Lindermayr et al., 2005). The somewhat limited overlap with published data may reflect differences in detection methodology, but also the type of stress applied—with the published data sets obtained following exogenous application of hydrogen peroxide. In total, considering all pathogen-treated and control samples, this study contributes 830 new putative redox-sensitive proteins associated with the ETI, and 548 redox-sensitive proteins not associated with the ETI.

Characterization and Comparative Analyses of the Col-0 and top1top2 Redoxomes

To uncover specific effects of TOP1 and TOP2 on redox-mediated processes during the ETI, we performed a comparative functional analysis of the Col-0 and top1top2 redoxomes (Supplemental Table S2). A GO analysis showed strong enrichment for metabolic processes in both Col-0 and top1top2. GO biological processes (GOBPs) such as “oxidation-reduction process,” “generation of precursor metabolites and energy,” and “translation” were enriched in both genotypes (Fig. 3A). Chloroplast-specific GOBPs including pigment metabolism, heme biosynthesis, and thylakoid membrane organization were enriched only in Col-0; however, GOBPs related to the organization and synthesis of precursors for the cell wall (CW), endoplasmic reticulum protein degradation, and pathogen-induced cell death were uniquely enriched in top1top2 (Fig. 3A). An analysis of the protein groups (GO Molecular Function) revealed commonalities but also marked dissimilarities among datasets (Fig. 3B). As such, in both redoxomes “oxidation-reduction activities” and proteins with “FeS cluster binding” properties, known for their role in the oxidation-reduction reactions of chloroplastic and mitochondrial electron transport (Dai et al., 2000), were significantly over-represented. However, whereas Col-0 was enriched in glutathione-disulfide reductases and phosphotransferases, the top1top2 redoxome had a high abundance of chitinases, peptidases, hexokinases, and thioredoxin-disulfide reductases. The quantitative and qualitative differences between Col-0 and top1top2 redoxomes were further revealed by a pathway-enrichment analysis (Fig. 3C). Whereas both genotypes showed an over-representation of the pathways for proteasome degradation, glycolysis, and amino acid biosynthesis, they differed in the over-representation of distinct amino acid biosynthesis pathways.

Figure 3.

Comparison between Col-0 and top1top2 reversible Cys redoxomes during the effector-triggered immune response. A to C, Enrichment of GOBP terms. GOBPs (A), GO molecular functions (GOMF; B), and Pathways (C) in Col-0 (gray) and top1top2 (brown). Text in black shows GO terms/pathways common for Col-0 and top1top2, whereas text in red shows terms unique for each genotype. D, The frequency of annotated proteins with PTMs in Col-0 and top1top2 redoxomes. E and F, Overlap and specificity in the Col-0 and top1top2 redoxomes. Numbers represent the numbers of proteins in each category; colored text indicates proteins unique to top1top2 (brown) or with differential oxidation in top1top2 (blue). E, Common and unique proteins associated with the GO terms “defense to bacteria,” “incompatible interaction,” “hypersensitive response/PCD,” “SA signaling,” and “systemic acquired resistance.” F, Common and unique proteins associated with the GO terms “oxidative stress response” and “cellular redox homeostasis.” G, Predicted cellular pathways with defective protein Cys oxidation in top1top2 plants. Nodes in brown are oxidized proteins unique for, or with, differential oxidation in top1top2. OxRed, oxidoreductase activity; Fru 6P, Fru 6-P tEF, translation elongation factor activity; Chl, chloroplast.

We queried the UniProt database to retrieve protein functional and structural information. We found that greater than one-quarter of Col-0 and top1top2 datasets (26% and 29%, respectively) contains proteins with annotated redox-associated PTMs (oxPTMs; Supplemental Fig. S3A). Among the proteins with annotated PTMs, 38% in Col-0 and 53% in top1top2 contained annotated redox-active Cys or structural disulfide bridges, thiol glutathionylation, and nitrosylation. Other types of PTMs with high representation in our datasets included N-acetylation and Ser/Thr phosphorylation. We asked next whether the position of sensitive cysteines within proteins corresponds to the annotated sites of Cys oxPTM (Fig. 3D). For the Col-0 dataset, annotations were available for 54 Cys sites in 18 proteins. We found that 15 annotated sites (in 13 proteins) were identified as oxidation-sensitive in Col-0 and an additional 11 sites (in eight proteins) were new sensitive sites identified. Altogether, almost 50% of the annotated redox-sensitive cysteines were discovered in the Col-0 dataset. For top1top2, annotations were available for 94 sensitive sites in 15 proteins. Specifically, we identified 17 annotated sites (located in 12 proteins), and seven additional sites (located in six proteins) without a UniProt annotation. Direct comparison of Col-0 and top1top2 coverage is difficult due to the relatively low overlap between the two datasets; nevertheless, it appears that fewer annotated sensitive sites are oxidized in top1top2 relative to Col-0.

Because many of the annotated redox-sensitive PTMs present in the redoxomes belong to well-known redox sensors or redox-modulated proteins, we took a closer look into the apparent disparity in the number of oxidized sites in Col-0 versus top1top2 (Table 1). In the 20 proteins surveyed, 37 and, respectively, 29 cysteines were identified as oxidized in Col-0 and top1top2, indicating a 22% decrease in the number of oxidized sites in the mutant. The most affected proteins in top1top2 were: the SA-binding BCA1 with functions in the ETI, chloroplastic magnesium-chelatase subunit ChlI-2, glutaredoxin C1, NADPH-dependent thioredoxin reductase3 for chloroplast protection against oxidative damage (Marty et al., 2009), chloroplastic thioredoxin M1 with roles in the oxidative stress, disulfide-isomerase PDI5-1, and chloroplastic dihydrolipoyl dehydrogenase2. Notably, the chloroplastic redox sensor 3'(2'),5'-bisphosphate nucleotidase 1 was identified as well; Cys-167 from the disulfide bridge 167-190, described as inhibitory under oxidizing conditions (Chan et al., 2016), was oxidized in Col-0 but not in top1top2.

Table 1. Proteins with annotated oxidation-sensitive cysteinesInformation about the position of disulfide bridges and Cys PTM were retrieved from UniProt and relevant literature (cited in the main text).

Symbol	Protein Name	Disulfide Bond/PTM	Cys Oxidized in top1top2	Cys Oxidized in Col-0	Effects of Oxidation
BCA1/ SABP3	Beta carbonic anhydrase1, chloroplastic (Salicylic Acid Binding Protein3)	Cys-280 S-nitrosylated	230-257-277-280	167-173-230-277-280	S-nitrosylation at Cys-280 is up-regulated during nitrosative burst and blocks binding of SA and BCA1 activity
AMY3	Alpha-amylase3, chloroplastic	DISULFID 499-587	None	363	Redox-regulated, with the highest activity under reducing conditions. The disulfide between 499 and 587 inhibits catalysis
CDSP32	Thioredoxin-like protein, chloroplastic	DISULFID 219-222	219-222	219-222-253	Redox-active
CHLI2	Magnesium-chelatase subunit ChlI-2, chloroplastic	DISULFID 96-187	390	390	Redox regulation; active in reducing conditions, inactive in oxidizing conditions. TRX F and M mediate the reversible reductive activation of oxidized CHLI2
		DISULFID 348-390
CP121	Calvin cycle protein CP12-1, chloroplastic	DISULFID 68-77	110	110	Only the oxidized protein is active in complex formation. The C-terminal disulfide is involved in the interaction with GAPDH. The N-terminal disulfide mediates the binding of PRK with this binary complex
		DISULFID 110-119
CP122	Calvin cycle protein CP12-2, chloroplastic	DISULFID 75-84	117	126
		DISULFID 117-126
CYP20-3/ ROC4	Peptidyl-prolyl cis-trans isomerase, chloroplastic	DISULFID 131-248	None	206	PPIase activity is optimal in reduced form and minimal in oxidized form. Reduction of the oxidized form is mediated by TRX M
DSP4/ AtSEX4	Phosphoglucan phosphatase DSP4, chloroplastic (Dual Specificity Protein Phosphatase4; Protein StarchExcess4)	ACT_SITE 198	198-374	198-374	198 Phospho-Cys intermediate in the enzyme active site
FTRC	Ferredoxinthioredoxin reductase catalytic chain, chloroplastic	DISULFID 87-117	104-106-115-117	104-106-115-117	Redox-active
GRXC1	Glutaredoxin-C1 (AtGrxC1)	DISULFID 39-42	None	39-42	Redox-active
IF4E4/ LSP1	Eukaryotic translation initiation factor isoform 4E (Loss Of Susceptibility To Potyvirus1)	DISULFID 97-138	97	97	According to the redox status, the 97-138 disulfide bridge may have a role in regulating protein function by affecting its ability to bind capped mRNA
MSRB8	Peptide Met sulfoxide reductase B8	DISULFID 75-128	128	128	Redox-active
PDI51	Protein disulfideisomerase5-1	DISULFID 55-58	51-55-85	51-55-85	Redox-active
PGR5	Protein PROTON GRADIENT REGULATION5, chloroplastic	DISULFID 11-105	105	105	DISULFID11-105 bridge is probably involved in disulfide bridges with PGRL1A
PLPD2	Dihydrolipoyl dehydrogenase2, chloroplastic	DISULFID 122-127	122-127	122-127-397	Redox-active
SAL1	SAL1 phosphatase, chloroplastic	DISULFID 167-190	167	None	Oxidative stress sensor; disulfide bridge 167–190 inhibitory under oxidizing conditions
STR9	Rhodanese-like domain-containing protein9, chloroplastic; Sulfurtransferase9	ACT_SITE 145	145	145	Cys-145 in the active site is a persulfide intermediate
TRXB3	NADPH-dependent thioredoxin reductase3	DISULFID 217-220	None	454-457	Both disulfide bridges are redoxactive
		DISULFID 454-457
TRXF1	Thioredoxin F1, chloroplastic (Thioredoxin F2)	DISULFID 99-102	99-102	99-102	DISULFID99-102 Redox-active; Glutathionylation at Cys-126 decreases its ability to be reduced by FTRC and reduces its efficiency in activating target chloroplastic enzymes
		Cys-126 S-glutathionyl Cys transient
TRXM1	Thioredoxin M1, chloroplastic	DISULFID 104-107	104-107	104-107	Redox-active

We considered whether the differences in oxidation patterns extend to the proteins categorized in GO as associated with defense or redox homeostasis in our datasets. An analysis of identifiers associated with “immune response,” “SA signaling,” and “programmed cell death” (Fig. 3E) and “cellular redox homeostasis” and “oxidative stress response” (Fig. 3F), identified overlapping and genotype-specific proteins. Common oxidized proteins include components of the Resistance-gene signaling (Century et al., 1995), SA signaling (MPK4, ROC1, and pathogenesis-related proteins [PR]1, PR2, and PR4; Sharma et al., 1996; Brodersen et al., 2006), and the SA-regulated catalase-1 and catalase-2 (Durner and Klessig, 1996). Known defense or oxidative stress markers, including the NPR1-dependent OPR1 (Blanco et al., 2005), PR4, and the cell death antagonist KTI1 (Li et al., 2008), had significantly different oxidation levels in the top1top2 redoxome. To test the possibility that proteins in Figure 3, D and E, are connected functionally, information on their known and curated protein–protein interactions (PPI) was extracted from public repositories (Szklarczyk et al., 2016); the PPI network generated (Supplemental Fig. S3B; Supplemental Table S3) had significant PPI enrichment P values, indicating that most of these proteins are biologically connected. We used the software PathLinker (Huang et al., 2018) to assemble the PPIs in a putative signaling network (Supplemental Fig. S3C), from which a subnetwork was retrieved including only the components with altered levels of oxidation in top1top2 (Fig. 3G). In Figure 3G, 36% of the nodes are chloroplastic and mitochondrial oxidoreductases—including the thioredoxin ACHT2, peroxidases (SAPX, TAPX, and WCRKC1), the redox-regulated and redox shuttling malate dehydrogenase mMDH1 (Yoshida and Hisabori, 2016), the NADPH-dependent thioredoxin reductase C for protection against oxidative damage (Puerto-Galán et al., 2015), the superoxide dismutase CSD2, and the GOX1 oxidase that mediates ROS signal transduction in plant resistance (Rojas et al., 2012).

Overall, we identified known redox-regulated proteins in Col-0 and top1top2 redoxomes, suggesting that they are active during the ETI; comparative analyses revealed the identity of the proteins and potential pathways with altered oxidation in top1top2, implicating oxidoreductases and defense components as a possible major disparity between genotypes.

Progressive Differences in Thiol Oxidation between Genotypes

We investigated the time dynamics of the Col-0 and top1top2 redoxomes. When comparing oxidation levels at 8 hpi between the control and infection conditions, we identified 550 cysteines that mapped to 458 proteins in Col-0, and 444 cysteines that mapped to 375 proteins in top1top2 (Supplemental Fig. S4A). In Col-0, 521 (95%) peptides increased, whereas 29 (5%) decreased in oxidation at 8 hpi relative to the 8 h control; by comparison, in top1top2, 393 (89%) increased, and 51 (11%) decreased in their relative oxidation levels. When comparing 8 hpi with 12 hpi, we found 596 peptides (485 proteins) in Col-0 and 555 significant peptides (465 protein identifiers) in the top1top2 mutant with significantly changed oxidation state. In Col-0, 178 (30%) increased and 418 (70%) decreased at 12 hpi, whereas in top1top2, 212 (38%) increased and 343 (62%) decreased at 12 hpi, showing an overall decreased oxidation trend in both Col-0 and top1top2. This was the condition where Col-0 and top1top2 exhibited the greatest difference regarding their Cys oxidation dynamics. In the control samples at 8 h, we identified 109 significant reversible oxidation changes between the genotypes; 66 (61%) were increasing in Col-0, and 43 (39.4%) were increasing in top1top2. A similar analysis was performed with the peptides showing time- or genotype-specific significant differences. Between Col-0 with top1top2 at 8 hpi, 85 peptides were significantly different; 50 (58.8%) showed an increase in Col-0 whereas 35 (41.2%) were higher in top1top2. Finally, at 12 h, out of the 144 significantly modified cysteines identified, 70 (48.6%) increased in Col-0 and 74 (51.4%) in top1top2. We summarized the above-described changes in oxidation and plotted the net change at all time points (Fig. 4A). Overall, the redoxomes of Col-0 and top1top2 showed a similar time-dependent dynamics—low oxidation at 8 h control, high increase in oxidation at 8 hpi, followed by a sharp decrease in the number of oxidized cysteines at 12 hpi. Notably, both the positive and negative fluctuations in net oxidation levels were larger in Col-0 than in top1top2 at both 8 and 12 hpi.

Figure 4.

Reversible Cys oxidation during the plant immune response shows time-dependent dynamics. A, Net changes in protein Cys oxidation in Col-0 and top1top2 plants during the ETI. Net oxidation was calculated by subtracting the number of peptides with reduced oxidation from the number of peptides with increased oxidization at each time point: Ctrl (8h-C), and 8 hpi and 12 hpi. B, Time-dependent oxidation dynamics in Col-0 and top1top2 protein clusters. Average change was calculated between 8 hpi and 12 hpi, respectively, with P. syringae avrRpt2 and buffer control, respectively. Red/green colors indicate, respectively, a higher/lower LFC; orange shows neutral change (fold change of 1). C, GO enrichment in Col-0 clusters 1, 6, 9, 14, and 23 showing increased oxidation at 8 hpi and decreased oxidation at 12 hpi. D, GO enrichment in top1top2 clusters 3, 5, 16, and 18 showing increased oxidation at 8 hpi and decreased oxidation at 12 hpi. E, GO enrichment in top1top2 clusters 7, 12, and 22 showing decreased oxidation at 8 hpi and increased oxidation at 12 hpi. GOCC: GO cellular localization (shaded terms); OxRed: oxidoreductase activity.

To capture additional details on the temporal changes in oxidation and ascribe a functional identity to oxidized proteins, we utilized the time-series clustering package TMixClust (Golumbeanu and Beerenwinkel, (2018); R Development Core Team, 2018) to group peptides according to their oxidation dynamics patterns. We found 23 and 24 clusters, respectively, to be optimum settings for Col-0 and top1top2 (Supplemental Fig. S4B). A cluster oxidation gradient measuring the average increase or decrease in the oxidation level of protein identifiers at 8 hpi and 12 hpi within each Col-0 or top1top2 cluster was visualized using a heat map (Fig. 4B). A comparative inspection of the heat map revealed many clusters showing the wave pattern of oxidation described in Figure 4A—sharp increase in oxidation at 8 hpi followed by a decrease at 12 hpi. We quantified the range of increase/decrease in protein oxidation level and confirmed that it was higher in Col-0 (−8.4 to 4.8 Log₂ fold-change [LFC]; range: 13.2) than in top1top2 (−6.0 to 3.9 LFC; range: 9.9). In Col-0, proteins associated with response to stress, including various peroxidases and proteins localized in chloroplast stromules and apoplast, had increased Cys oxidation at 8 hpi and decreased oxidation at 12 hpi (Fig. 4C). Clusters with a similar wave dynamics in top1top2 were enriched in oxidoreductases, porphyrin and chlorophyll metabolic enzymes, plastid-, and apoplast-localized proteins (Fig. 4D). Furthermore, cluster analysis identified top1top2 proteins (69 identifiers grouped in clusters 7, 12, and 22) with significantly reduced oxidation at 8 hpi and increased oxidation at 12 hpi; this group was enriched in proteins associated with defense against bacteria and systemic acquired resistance localized in the apoplast, extracellular matrix, and chloroplast stroma (Fig. 4E).

Our results demonstrate that protein oxidation during ETI follows distinct time-dependent patterns; cysteines with similar dynamic profiles belong to proteins having common localizations, molecular functions, and participating in similar biological processes. Although Col-0 and top1top2 manifested similar dynamics in protein oxidation, the amplitude of the oxidation wave was lower in top1top2 than in Col-0; both fewer oxidized proteins and fewer oxidized cysteines per protein were identified in the mutant.

Structural Determinants of Cys Oxidation

Biochemical studies with metazoan proteins have identified several factors that determine the susceptibility of protein thiols (-SH) to oxidation, including thiol’s acid dissociation constant (pK_a) and position within the native protein structure. At physiological pH, thiols with pK_a < 7 exist as thiolates (S⁻) that have increased reactivity toward electrophiles such as oxidants. Exposure of thiolates on protein surface increases their proximity to the oxidant source and determines responsiveness to cellular redox environment. To identify possible specificity-determinants for Cys oxidation, we searched the Protein Data Bank with the complete list of oxidized proteins and retrieved complete 3D structures for 48 proteins (Supplemental Table S4) containing 518 cysteines of which, 104 (20%) were identified as oxidation-sensitive in the ETI redoxome.

To investigate the relationship between cysteines exposure on the protein surface, acidity, and their susceptibility to oxidation, we calculated the surface accessibility (“buried” B coefficients) and pK_a values of their cysteines using the software PROPKA v2.1 (Rostkowski et al., 2011). Overall, 34% of cysteines were defined as “exposed” (B ≤ 60) and 66% as “buried” (B > 61). Although similar percentages among the buried (25%) and exposed cysteines (21%) were found as oxidized, the cysteines with low solvent-exposure predominated in the oxidized cohort (Fig. 5A). An analysis of the distribution of the B values in oxidized versus nonoxidized residues showed that oxidized cysteines segregated into two distinct groups. Among the “buried” cysteines, the oxidized sites were significantly less solvent-exposed than the nonoxidized sites; on the other hand, among the solvent-exposed cysteines, the oxidized sites had significantly higher solvent-exposure than the nonoxidized sites (Fig. 5B). A similar analysis performed using the pK_a values of cysteines determined that surface-exposed sites had significantly lower predicted pK_a values (average pK_a = 9.7) than buried cysteines (average pK_a = 12.7); however, we found no relevant differences between the oxidized and nonoxidized cysteines (Fig. 5C). We considered the possibility that potential differences in the reactivity of the oxidized cysteines are hidden in this analysis due to the relatively large number of sites per protein and the large variation in their pK_a values. To account for these variations, the pK_a values for each protein were normalized. An inspection of the normalized pK_a values in proteins with “buried” or “exposed” oxidized cysteines, revealed patterns recurring in at least half of the proteins analyzed. Fig. 5D shows 10 proteins where the “exposed” oxidized cysteines have the lowest pK_a values within the respective proteins. This group includes Cys-65 in the sugar sensor hexokinase1 (HXK1; Jang et al., 1997), Cys-136 in the o-acetyl-Ser (thiol) lyase OASC, Cys-211 in the hydrolase DWARF14, Cys-169A in the oxidoreductase OPR3, aCys309 in the oxidoreductase At4g09670, and Cys-197 in the Cys- and His-rich domain of RAR1 (Kadota et al., 2008). On the other hand, we found 11 proteins (seven of them shown in Fig. 5E) where the “buried” oxidized cysteines have the highest pK_a values within each protein. This set included Cys-48 in the chloroplastic NADPH-dependent aldo-keto reductase, Cys-40 and Cys-41 in the phylloquinone biosynthesis hydrolase DHNAT1, Cys-167 in SAL1, and Cys-235 in the chlorophyll catabolic enzyme HCAR. In these proteins, buried cysteines may be oxidized following interactions with ROS-sensors such as peroxidases.

Figure 5.

Structural determinants of Cys oxidation in the Arabidopsis proteome. A, Protein Cys residues having low or high solvent exposure in a subset of the Col-0 and top1top2 redoxomes, found to be oxidized (ox-Cys) or unmodified (nox-Cys). B, Distribution of ox-Cys and nox-Cys according to their average of solvent exposure. “All Cys” shows the distribution of average B coefficients/protein for all cysteines analyzed. The asterisks show statistical relevance (P < 0.05, Student’s t test parameters: two-tailed distribution, two-sample unequal variance) between the exposed and buried cysteines. C, The distribution of pK_a values for oxidized cysteines with high (Exposed Cys) and low solvent-exposure (Buried Cys). Statistically, significant differences (P < 0.01) are shown between exposed and buried cysteines for both oxidized and nonoxidized Cys groups. D, The distribution of pK_a values within proteins in which solvent-exposed cysteines (B < 61) were found to be sensitive to oxidation. E, The distribution of pK_a values within proteins in which cysteines with low solvent-exposure were identified as oxidation sensitive. In (D) and (E), values above bars are the B coefficients of cysteines. The analyses were performed on 514 Cys residues in 48 proteins. The asterisks show statistical relevance (P < 0.05, Student’s t test parameters: two-tailed distribution, two-sample unequal variance). The position of each Cys residue is shown on the x axis, and the residues were arranged in the order of increasing pK_a value. (Red) Oxidized residues; (left) UniProt (or Araport) protein identifiers.

Overall, our results demonstrate that both surface-exposed and internal cysteines are sensitive to oxidation; exposed cysteines with the lowest pK_a values or buried cysteines with the highest pK_a values within a protein are susceptible to oxidation, suggesting a remarkable specificity in protein thiol oxidation and a possible mechanism.

DISCUSSION

A Mass Spectrometry Label-Free Approach to Characterize Plant Redoxomes—Methodological Considerations

To study the dynamic reversible oxidation of protein cysteines in Arabidopsis, we used a tag-switch assay to enrich modified proteins, wherein in vivo reduced thiols were blocked by IAM before reduction of all reversibly oxidized sites and enrichment using TPS6B resin. This methodology has been used extensively in the study of reversible oxidation with considerable success in diverse biological systems (Forrester et al., 2009; Paulech et al., 2013; Guo et al., 2014a, 2014b). Although generally susceptible to false-positive identifications of reversibly oxidized sites due to incomplete alkylation of reduced cysteines, we performed a control experiment to ensure that the reaction was sufficiently complete (Supplemental Fig. S1A). Label-free approaches rely on the mass spectrometric signal intensity of peptide precursor ions across separate LC-MS/MS acquisitions for relative quantification. Whereas isobaric labeling techniques provide higher precision and faster acquisition when comparing multiple sample types, label-free workflows allow enhanced proteome coverage and improved linear range of quantitation at a greatly reduced cost to the researcher (Wang et al., 2012a; Dong et al., 2015). Therefore, we opted to use a workflow incorporating LFQ utilizing our protein-level TPS6B enrichment. Our robust LFQ measurements following this enrichment technique allowed differential analysis of the redoxome with increasing duration post-bacterial challenge in Col-0 and top1top2. We discovered that Cys reactivity in both redoxomes was strongly correlated with previously annotated sites of reactive Cys residues part of redox-active sites, disulfide bridges, or cysteines modified by reversible oxPTM including glutathionylation and nitrosylation. It is important to notice that our method may not have identified all sensitive cysteines because of their inaccessibility to label and low cellular abundance. These changes were used to elucidate both temporal and genotype-specific Cys oxidation to the biotic stress as a first principle, albeit we recognize that changes in protein abundance could falsely skew the measured fold changes attributed to Cys oxidation. We, therefore, take the findings of this study as a necessary guide for researchers to potentially responsive Cys-sites, with the caveat that the actual protein abundance may change.

Nevertheless, the results from the literature and database mining indicate that a large percentage of the reactive cysteines we identified are located in proteins with known oxPTM or redox-active sites. Even though additional validation of the oxidized cysteines discovered in this study is required, we have drastically increased the current repository of reversible oxPTMs on cysteines and substantiated previous studies of oxidation-sensitive proteins. We found that most proteins have one or few sensitive Cys sites and sensitive sites are present in a large number and diverse classes of proteins, suggesting the extensive oxidation of the ETI proteome.

How do TOP1 and TOP2 Contribute to the Redox Proteome Phenotype?

Data presented in this article brings information on the impact of TOP1/TOP2 on the composition of the Cys-oxidized proteome in the early stages of the ETI and highlights the contributions of TOP1/TOP2 to achieving and maintaining proper levels of oxidation during the oxidative burst. Plants contain numerous proteases, and many are known to participate in immune signaling and HR initiation or execution (van der Hoorn, 2008). Barring a few exceptions (Breitenbach et al., 2014; Zhang et al., 2014), the coordination of pathogen defense pathways with the action of proteases is insufficiently understood (van der Hoorn, 2008). Several mechanisms could contribute to the impaired regulation of ETI redox homeostasis in top1top2. Akin to FtsH proteases (Yoshioka-Nishimura et al., 2014), TOP1 oligopeptidase activity may ensure the optimal productivity of the photosystem complexes as part of the chloroplast quality control machinery and may function in the light-dependent pathway that controls the magnitude of cell death (Kangasjärvi et al., 2014; Stael et al., 2015). On the other hand, TOP2 appears to operate in parallel, or the aftermath of, the chloroplastic oxidative burst. TOP2 may participate in the degradation of irreversibly oxidized proteins downstream of the 20S proteasome to decelerate cell death and spatially restrict the HR. A more precise function can be envisioned whereby TOP2 may generate signaling peptides to suppress cell death and activate immune responses through the H₂O₂-induced signaling that controls redox homeostasis (Nakagami et al., 2006) or ETI-associated pathways (Fu et al., 2012; Spoel and Dong, 2012) known to function in a proteasome-dependent manner.

Variation in net protein oxidation and number of oxidized cysteines observed between or within the genotypes could be a consequence of several mechanisms. The increase in the oxidation levels of a protein can result from an upsurge in its oxidation when the protein maintains its steady-state abundance, from an increase in its abundance with steady-state oxidation, or a combination of both. Conversely, the decrease may originate from either steady-state oxidation on a proteolytically degraded protein, reduction of cysteines on a protein in steady-state abundance, irreversible oxidation of cysteines on a protein in steady-state abundance, or combination of the three. Each possibility would increase/decrease relative abundance measured in our assays. Notably, loss of both TOP1 and TOP2 has no significant effects on plant development, growth, and reproduction under physiological conditions (Moreau et al., 2013). Thus, rather than having generalized proteolytic housekeeping functions, TOPs are likely to have specialized roles that manifest under plant stress, akin to the chloroplastic proteases from the Clp system found to have a small and specialized repertoire of substrates (Nishimura and van Wijk, 2015). A more precise answer regarding TOP1/TOP2 contributions to the redoxome phenotypes, additional knowledge is needed on their function to plant proteostasis and the identity of their substrates.

So far, based on our results, we anticipate that pathways and processes activated in the plant during the immune response, including signaling and metabolic pathways, are the probable subjects of regulation through oxPTMs. Significant changes occur in top1top2 in the composition of the redoxome and the total number of Cys sites oxidized. To expand on this hypothesis, let us examine our findings from this perspective. The molecular players activated as part of the innate immune response are fairly well known. In resistant plants, the avrRpt2-mediated degradation acts as a signal for RPS2 activation. A critical component is the cyclophilin ROC1 (Aumüller et al., 2010) that activates P. syringae’s putative Cys-protease AvrRpt2 (Coaker et al., 2005). Whereas in Col-0, both Cys-69 and Cys-122 of ROC1 were oxidized, only Cys-122 was detected in the mutant; also, the related ROC4 was found oxidized exclusively in the Col-0. The oxidation-sensitivity of ROCs is consistent with reports ascribing oxPTMs and redox-regulation to cyclophilins (Shapiguzov et al., 2006; Campos et al., 2013). Protein-folding machinery is required for transduction of R-gene mediated resistance (Holt et al., 2005). Cys-43 in the CHORD domain of the cochaperone RAR1 was identified in both Col-0 and top1top2; however, the infection-triggered chaperone HSP70 was differentially oxidized. Furthermore, PLTP2, a PCD regulator from the PR14 family (Sels et al., 2008) has four disulfide bonds; the sensitive Cys-100 was oxidized in both Col-0 and top1top2. The SA-inducible PR1, PR4, and PR5 were also identified as sensitive to oxidation. The potential redox-regulation of PRs is supported by the finding that PR5 is a possible target of thioredoxins (Marchand et al., 2006) and that PR1 accumulation is under the regulation of glutathione (Senda and Ogawa, 2004). Peptidases implicated in immune-related processes at the CW and extracellular matrix were identified as possible redox-modulated proteins. The SA-inducible Gly-rich protein3 participates in CW repair; four residues (Cys-82/86/87/94) in Gly-rich protein3 Cys-rich motif required for the oxidative cross linking to CW (Domingo et al., 1999) were oxidized in both genotypes. ASPG1, an aspartyl protease induced locally and systemically during systemic acquired resistance (Breitenbach et al., 2014); out of the six structural disulfides, Cys-271 was found differentially oxidized in top1top1. Cys-292 in the PCD-regulator XCP2 peptidase (Zhang et al., 2014) was oxidized in both genotypes. Lastly, Cys peptidases from the metacaspase family were found to be oxidation-sensitive. In MC4, a positive PCD regulator (Watanabe and Lam, 2011), Cys-22 and the catalytic Cys-139 were oxidized in both genotypes; s-nitrosylation of Cys-139 regulated the activity of MC9 (Vercammen et al., 2004; Belenghi et al., 2007).

Primary metabolism, especially pathways necessary for producing energy and metabolites, are critical to the execution of plant defense responses. Transcriptional down-regulation of genes involved in photosynthesis and chlorophyll biosynthesis accompanies responses to avirulent pathogens (Apel and Hirt, 2004; Rojas et al., 2014). Considering the significant over-representation in our datasets of carbohydrate metabolism, photosynthesis, and photorespiration pathways, we postulate that changes in the redox homeostasis during the ETI are liable to be involved in the regulation of enzymatic activities acting in these pathways. The link between carbohydrate metabolism and defense responses is best exemplified by the activity of HXK1, a sugar-sensor that controls the transcriptional induction of PR1 and PR5 (Xiao et al., 2000). Cys-159, in the proximity of the HXK1 sugar-binding site, was found oxidized in both Col-0 and top1top2. In top1top2, BCA1/SABP3, required for photosynthesis and CO₂ signaling and known to be inactivated by oxidation or s-nitrosylation of Cys-280 (Wang et al., 2009; Nienaber et al., 2015), was found underoxidized relative to Col-0. Several enzymes with roles in chloroplast biogenesis and photosynthetic pigment synthesis were differentially oxidized in top1top2. ACD1, an FeS cluster oxygenase that functions in chlorophyll catabolism and the ETI (Yao and Greenberg, 2006), was oxidized only in Col-0, whereas CHLI2, an Mg-chelatase that catalyzes chlorophyll biosynthesis and active only under reducing conditions (Kobayashi et al., 2008), was oxidized at higher levels in top1top2. Altogether, the altered patterns of oxidation in top1top2 of components of resistance as well as metabolic pathways suggest the role of the TOP oligopeptidases in regulating their catalytic activity and, possibly, accumulation during the ETI.

Determinants of Cys Sensitivity and Specificity in Protein Oxidation

Systematic and thorough identification of cysteines undergoing oxPTMs under various conditions will help recognize possible general criteria in Cys sensitivity and specificity, and define redox switches and mechanisms of signal transmission. Our analysis focused on the chemical and structural properties of cysteines in a subset of proteins from the Arabidopsis reversible Cys redoxome. We aimed to use information on the solvent-exposure of thiols and their predicted pK_a values to identify commonalities among oxidized cysteines. In agreement with previous findings on thiol reactivity (Conte and Carroll, 2013), a set of the experimentally identified oxidized cysteines had high solvent-accessibility and the lowest predicted pK_a values within the respective proteins.

Interestingly, we also identified oxidized cysteines with low solvent-exposure and high pK_a values. Considering that thiols with higher pK_a values are strong nucleophiles and, thus, more reactive (Fernandes and Ramos, 2004), it is possible that they are targets of oxidation in vivo. Several caveats have to be mentioned here. First, because pK_a values of protein cysteines vary with the pH and subcellular localization, oxidation of proteins is modulated by its localization and the type of ROS that oxidize it (Foyer and Noctor, 2016). Second, the in vivo reactivity of protein thiols may differ from our predictions. Third, denaturing conditions in our protocol may have unavoidably exposed buried thiols that otherwise may not be susceptible to oxidation. Nevertheless, overall, our results strongly support a high specificity in plant thiol oxidation that cannot be solely the result of overall thiol reactivity. Potential mechanisms of protein oxidation may involve interactions between ROS sensors such as thiol peroxidases with protein substrates in thiol oxidizing relays (Sobotta et al., 2015), and self-sensitization mechanisms where the initial oxidation of an exposed thiol may cause local structural changes in the native protein structure to expose buried thiols.

MATERIALS AND METHODS

Plant Growth

Seeds of Arabidopsis (Arabidopsis thaliana; ecotype Columbia) and a double-knockout mutant of both TOP1 (AT5G65620) and TOP2 (AT5G10540) genes (Moreau et al., 2013) were sown individually in 24-hole plug trays containing soil. After stratification for 2 d at 4°C in the dark, plants were cultivated in a growth chamber for four weeks under medium-day conditions (12 h light/12 h dark) at 22°C and 60% relative humidity with 150 µmol/m²s photon flux density.

Plant Infection

Pseudomonas syringae pv. tomato (Pst) avrRpt2 was freshly streaked from a glycerol stock onto King’s B medium plates containing 100 mg/L rifampicin and 25 mg/L kanamycin and incubated at 28°C for 3 d. Bacterial cells were scraped from the plate, resuspended in 0.1 m sucrose in 10 mm MgCl₂ solution, and incubated at 28°C until OD₆₀₀ of 0.1 was achieved. Three- to four-week-old plants were syringe-infiltrated to saturation with 1 × 10⁵ CFU/mL Pst DC3000 avrRpt2 suspended in 10 mm MgCl₂. Six leaves from each plant were syringe-infiltrated with either 10 mm MgCl₂ (mock) or bacteria (treatment), and plants were then moved back to the growth chamber. Plants mock-inoculated with 10 mm MgCl₂ solution were harvested after 8 h to serve as experimental controls. Otherwise, after 8 hpi or 12 hpi, leaves were excised from plants by cutting the petiole with a razor blade and immediately frozen under liquid nitrogen before storage at −80°C.

Ion Leakage Measurements

Three- to four-week-old plants were syringe-infiltrated, five leaves per plant, to saturation with 1 × 10⁶ CFU/mL Pst DC3000 avrRpt2 to elicit the oxidative burst and HR. Conductivity was used as a quantitative measure of ion leakage due to PCD. For each replicate, five leaf discs (1 per each of the five inoculated leaves) were collected from noninfiltrated leaf halves at 0, 4, 8, and 12 hpi and floated abaxial side up on MQH₂O. Leaf discs were shaken at 60 rpm at room temperature for 2 h before measuring the conductivity of the bathing solutions. Plants infiltrated with 10 mm MgCl₂ served as controls for ion leakage caused by MgCl₂ and wounding during infiltration. Total ion content was measured for each replicate after freezing and thawing the leaves within their respective bathing solutions. Four replicates were analyzed for each infiltration group per time point per line. Significant differences in comparison to Col-0 plants after Pst-infiltration were calculated using Student’s t test parameters: two-tailed distribution, two-sample unequal variance.

Protein Extraction

Three biological replicates were used for 8-h mock-inoculated control, 8-hpi, and 12-hpi conditions for each genotype (e.g. Col-0 and top1top2 mutants). Proteins were extracted using a Tris-buffered phenol method described in Slade et al. (2015). Briefly, leaf tissue (0.2 g) was ground under liquid N to a fine powder before homogenization in 2 mL of buffer containing 50 mm Tris (pH 8), 1 mm EDTA, 0.9 m sucrose, protease inhibitors (EDTA-free; Roche), and 100 mm IAM before adding SDS to a 1% (v/v) final concentration. The homogenate was incubated at room temperature for 90 min on a Techne rotator at 10 rpm to effectively block all in vivo reduced thiols. Proteins were then extracted from the aqueous phase by adding an equal volume (2 mL) of Tris-buffered phenol, pH 8 and mixing for 1 min before centrifugation (20,000g, 5 min). The top layer of phenol was separated to a prechilled tube, and an equal volume of Tris-buffered phenol was added to back-extract the aqueous phase. Proteins were precipitated from the combined phenolic layers using five volumes of cold 100 mm ammonium acetate in methanol and incubated at −80°C for 1 h before centrifugation (2,000g, 5 min). The resulting pellet was washed twice with 100 mm ammonium acetate in methanol and once with 70% (v/v) ethanol. The precipitate was resuspended in 50 mm Tris (pH 8.0), 0.5% (v/v) SDS, and 8 m urea and remaining cellular debris was removed by centrifugation (2,000g, 5 min). Protein concentration in the supernatant was determined using the 2D-Quant Kit (GE Healthcare) against BSA standard. Reversibly oxidized thiols were reduced using 10 mm DTT and incubated in a Thermomixer for 1 h at 37°C and 850 rpm. Excess DTT was removed by loading protein sample to an Amicon Ultra-0.5 centrifugal filter (EMD Millipore) with 10 kD MWCO and centrifuging for 10 min at 4°C and 10,000g. Washing buffer was then added containing 50 mm Tris, pH 8 with 1 mm EDTA and 0.5% SDS in two passes. Protein was recovered from the filter using an equal volume of washing buffer and estimated for concentration. Subsequent enrichment of nascent thiols was performed using 500 µg of protein.

Protein-Level Cys Enrichment

Detailed evaluation of the steps used in the enrichment of reversible Cys oxidation was performed as described in Berg et al. (2019). Briefly, each replicate and 50 mg of TPS6B (GE Healthcare Bio-Sciences) resin slurry were combined and incubated in a Thermomixer for 2 h at 30°C and 850 rpm to allow protein binding via Cys thiols. Nonspecifically bound proteins were removed by washing the resin twice with 10 bed volumes each of washing buffer, washing buffer with 2 m NaCl, 70% (v/v) acetonitrile with 0.1% (v/v) trifluoroacetic acid (TFA), and finally 50 mm Tris, pH 8. On-resin digestion of Cys-bound proteins was performed by resuspending protein-resin slurry in 500 µL of 50 mm Tris, pH 8 before adding 10 µg of Trypsin Gold (Promega). The mixture was incubated in a Thermomixer for 3 h at 37°C and 850 rpm before separating peptide flow-through from bound Cys-proteins via brief centrifugation. Samples were then washed with 10 bed volumes each of washing buffer with 2 m NaCl, 70% (v/v) acetonitrile with 0.1% (v/v) TFA, and 50 mm Tris, pH 8. Bound Cys-containing peptides were eluted from the resin twice using 250 µL of 10 mm Tris, pH 8 with 50 mm DTT for 15 min.

LC-MS/MS Analysis

Samples were desalted using 50 mg Sep-Pak C18 cartridges before resuspension in 3% (v/v) acetonitrile with 0.1% (v/v) TFA and 5 mm DTT to prevent oxidation of reduced thiols before LC-MS/MS analysis. Peptides were analyzed using a NanoAcquity UPLC system (Waters) coupled to a TripleTOF 5600 mass spectrometer (AB Sciex). Mobile phase A consisted of water with 0.1% (v/v) formic acid and mobile phase B was acetonitrile with 0.1% formic acid. Peptide mixtures were injected to a Symmetry C₁₈ trap column (100 Å, 5 µm, 180 µm × 20 mm; Waters) with a flow rate of 5 µL/min for 3 min using 99.9% A and 0.1% B. Peptides were then separated on an HSS T3 C₁₈ column (100 Å, 1.8 µm, 75 µm × 250 mm; Waters) using a linear gradient of increasing mobile phase B at a flow rate of 300 nL/min. Mobile phase B increased to 35% in 90 min before ramping to 85% in 5 min, where it was held for 5 min before returning to 5% in 2 min and re-equilibrating for 13 min. The mass spectrometer was operated in the positive polarity and high sensitivity mode. MS survey scans were accumulated across an m/z range of 350–1600 in 250 ms. For information-dependent acquisition, the mass spectrometer was set to automatically switch between MS and MS/MS experiments for the first 20 features above 150 counts having +2 to +5 charge state. Precursor ions were fragmented using rolling collision energy ±5% with an accumulation time of 85 ms. Dynamic exclusion for precursor m/z was set to an 8-s window. Automatic calibration was performed every 8 h using a tryptic digest of BSA protein standard to maintain high mass accuracy in both MS and MS/MS acquisition.

Database Searching and Label-Free Quantification

Acquired spectral files (*.wiff) were imported into Progenesis QI for Proteomics (Nonlinear Dynamics, version 2.0). A reference spectrum was automatically assigned, and total ion chromatograms were then aligned to minimize run-to-run differences in peak retention time. Each sample received a unique factor to normalize all peak abundance values resulting from experimental variation. Alignment was validated (>80% score) and a combined peak list (*.mgf) for all runs was exported for peptide sequence determination and protein inference by the software Mascot (Matrix Science, version 2.5.1). Database searching was performed against the UniProt reference proteome (39,196 canonical entries) for Arabidopsis (www.uniprot.org/proteomes/UP000006548; accessed September 2017). Sequences for common laboratory contaminants (www.thegpm.org/cRAP; 116 entries) were appended to the database. Target-decoy searches of MS/MS data used a trypsin protease specificity with the possibility of two missed cleavages, peptide/fragment mass tolerances of 15 ppm/0.1 Da, and variable modifications of acetylation at the protein N terminus, carbamidomethylation at Cys, deamidation at Asn or Gln, and oxidation at Met. Significant peptide identifications above the identity or homology threshold were adjusted to <1% peptide FDR using the embedded Percolator algorithm (Käll et al., 2007) and uploaded to Progenesis for peak matching. Identifications with a score <13 were removed from consideration in Progenesis before exporting “Peptide Measurements” from the “Review Proteins” stage.

Data Availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (www.proteomexchange.org) via the PRIDE partner repository (Vizcaíno et al., 2014) with the dataset identifiers PXD010104 (Username: reviewer56000@ebi.ac.uk, Password: 9HwxLtFp).

Data Preprocessing

Data were parsed using custom scripts written in the Python programming language. In the “Peptide Measurements” export, there were instances of rows with duplicated peak features and differing peptide identifications. Some features were matched with peptides having identical sequence, modifications, and score, but alternate protein accessions. These groups were reduced to satisfy the principle of parsimony and represented by the protein accession with the highest number of unique peptides, else the protein with the largest confidence score assigned by Progenesis. Some features were also duplicated with differing peptide identifications and were reduced to a single peptide with the highest Mascot ion score. Results were limited to only peptides with one or more Cys-sites of reversible oxidation, defined here as the absence of carbamidomethylation on at least one Cys residue in the peptide sequence. An identifier was then made by joining the protein accession of each feature with the particular site of Cys-oxidation in the protein sequence. Each dataset was then reduced to unique identifiers by summing the abundance of all contributing features (i.e. peptide charge states, missed cleavages, combinations of additional variable modifications). Each identifier group was represented by the peptide with the highest Mascot score in the final dataset.

Data Analysis

Linear Models Analysis and Missing Data Imputation

We employed the data analysis pipeline described in detail in Berg et al. (2019). The first step is the filtering of data features where all samples were outliers (Tukey’s lower fence with k = 1.5). We then stabilized the variance of datasets by performing a log₂-tranformation of all abundance values. We then perform missing data imputation method using random sampling from a normal distribution with parameters robustly estimated from the entire dataset (mean value for the imputation is calculated per Cys for all nonmissing values in each condition, or using Tukey’s lower fence with k = 2 if all values are missing in one condition; sd for the imputation is computed per condition by taking the median of sds of each Cys with nonmissing values; Berg et al., 2019). We quantify relative changes in peptide abundance between conditions using the limma R package (Ritchie et al., 2015). We applied a LFC change cut-off of ±1, and a false-discovery-rate–corrected P value cut-off of 0.05 to identify significantly altered oxidation of cysteines. Multiple imputations were used in conjunction with a binomial test to decide on statistically significant changes in proteome quantitation due to missing data. Each dataset was run through limma using lmFit with the method = “robust” flag and then eBayes using the default settings (Phipson et al., 2016). LFC was then calculated for each feature, comparison, and dataset. We modeled the outcome of the data imputation using a binomial distribution and performed a right-tailed binomial test (R’s core feature binom.test; R Development Core Team, 2018) to identify the peptides with an outcome > 0.5, at a significance level 0.05.

Time-Series Clustering

We used the TMixClust time-series clustering R package. We chose a cluster size corresponding to the highest likelihood indicated by the stability analysis; we used the silhouette technique to validate cluster consistency. Each cluster size was run 20 times (nb_clustering_runs), and the number of clusters was iteratively incremented starting at two. Each run had a limit of 20 Monte Carlo resamplings using mc_em_iter_max, and 2,000 iterations of the expectation-maximization algorithm, to increase the precision of each run.

Functional Annotation

We used Araport (https://apps.araport.org/thalemine) annotations, Panther GO (http://pantherdb.org/) tools, and a statistical overrepresentation test (Fisher’s Exact with false-discovery-rate multiple test correction) to perform functional analysis of significantly differentially oxidized Cys residues.

Structure-Reactivity Analysis

We used the software PROPKA 2.1 (http://nbcr-222.ucsd.edu/pdb2pqr_2.0.0/) to predict pKa and B values of cysteines in selected proteins with 3D structure information from the Protein Data Bank (https://www.rcsb.org/).

Set Analysis

To compare our redoxomes with public datasets we used the R package UpSetR to visualize the intersecting sets and their properties (Conway et al., 2017). We also used UpSetR to plot the unique common sets of proteins or Cys identifiers from selected combinations of pairwise comparisons.

Accession Numbers

Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers listed in Supplemental Table 1.

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. Supplementary material for Figure 1.

• Supplemental Figure S2. Supplementary material for Figure 2.

• Supplemental Figure S3. Supplementary material for Figure 3.

• Supplemental Figure S4. Supplementary material for Figure 4.

• Supplemental Table S1. Results of the label-free quantification of redox Cys-site identifiers and comparisons among genotypes and treatments.

• Supplemental Table S2. GO overrepresentation tests in Col-0 and top1top2 redoxomes.

• Supplemental Table S3. The PPI network of proteins classified in GO terms related to plant defense and oxidative stress response.

• Supplemental Table S4. PROPKA analysis of proteins with known 3D structures from Col-0 and top1top2 redoxomes.