Abstract Pathogen attack can increase plant levels of reactive oxygen species (ROS), which act as signaling molecules to activate plant defense mechanisms. Elucidating these processes is crucial for understanding redox signaling pathways in plant defense responses. Using an iodo-tandem mass tag (TMT)-based quantitative proteomics approach, we mapped 3362 oxidized cysteine sites in 2275 proteins in rice leaves. Oxidized proteins were involved in gene expression, peptide biosynthetic processes, stress responses, ROS metabolic processes, and translation pathways. Magnaporthe oryzae infection led to increased oxidative modification levels of 512 cysteine sites in 438 proteins, including many transcriptional regulators and ribosomal proteins. Ribosome profiling (Ribo-seq) analysis revealed that the oxidative modification of ribosomal proteins promoted the translational efficiency of many mRNAs involved in defense response pathways, thereby affecting rice immunity. Our results suggest that increased oxidative modification of ribosomal proteins in rice leaves promotes cytosolic translation, thus revealing a novel function of post-translational modifications. Furthermore, the oxidation-sensitive proteins identified here provide a valuable resource for research on protein redox regulation and can guide future mechanistic studies. Key words: ROS, protein oxidation, translation, infection, immunity, rice __________________________________________________________________ This study improves our understanding of oxidation-sensitive plant proteins. Oxidation levels of rice proteins are increased substantially upon fungal infection, and the increased oxidative modification of ribosomal proteins in leaves promotes cytosolic translation, revealing a novel function of post-translational modifications. Introduction Protein functions are modulated not only by the amino acid sequence of the protein but also by post-translational modifications (PTMs) such as methylation, acetylation, ubiquitination, and phosphorylation. PTMs play important roles in gene expression as well as protein localization, stability, and interactions, adding layers of complexity and greater flexibility to cellular signaling ([37]Millar et al., 2019). PTMs are widely involved in regulating, and especially fine tuning, plant responses to pathogen infection ([38]Withers and Dong, 2017; [39]de Vega et al., 2018; [40]Chen et al., 2021). Redox regulation is a general and fundamental mechanism involved in transcription, metabolic processes, epigenetic processes, and cell signaling ([41]Mikhed et al., 2015; [42]Moore et al., 2016). Redox regulation often occurs in the thiol moieties of cysteine residues and includes thiol/disulfide transitions, S-glutathionylation (S-SG), sulfinylation, S-nitrosylation (S-NO), sulfonylation, and sulfenylation ([43]Lindermayr et al., 2005; [44]Hägglund et al., 2008; [45]Fares et al., 2011; [46]Couturier et al., 2013; [47]Puyaubert et al., 2014; [48]Akter et al., 2015). Reactive oxygen species (ROS; e.g., hydrogen peroxide [H[2]O[2]]) function as signaling molecules mainly by oxidatively modifying redox-sensitive proteins that act as molecular thiol switches ([49]Finkel, 2011; [50]Roos and Messens, 2011; [51]Cezary et al., 2015). This reversible oxidation process can control molecular thiol switches that regulate protein activity, stability, conformational changes, interactions, and cellular location ([52]Sokolov et al., 2006; [53]Tada et al., 2008). Changes in ROS production and processing result in transient or permanent changes in the redox status of the cell ([54]Cezary et al., 2015; [55]He et al., 2018). Pathogen attack can induce an increase in ROS in infected plants, thereby activating defense responses ([56]Apel and Hirt, 2004). Therefore, identifying the targets of plant ROS and determining how the corresponding signal is transduced are crucial for deciphering the redox signaling pathways involved in defense responses. As a staple food for over half the world’s population, rice (Oryza sativa) is crucial for global food security ([57]Bhullar and Gruissem, 2013). The filamentous fungus Magnaporthe oryzae causes rice blast, a highly destructive disease of cultivated rice that threatens food production around the world ([58]Wilson and Talbot, 2009; [59]Dean et al., 2012). M. oryzae is a model organism used to study the physiological and pathogenic molecular mechanisms of plant pathogenic fungi ([60]Wilson and Talbot, 2009). An intricate physiological redox balance is essential for the interaction between M. oryzae and rice ([61]Kou et al., 2019). In the early stage of this interaction, rice produces copious ROS to help resist infection ([62]Kou et al., 2019). This leads to oxidation of proteins, which may alter their structure and molecular function. However, the precise role of ROS in signaling and the regulation of specific targets in plant cells are largely unknown. To address these questions, we used an iodo-tandem mass tag (TMT) proteomics approach to characterize dynamic changes in the oxidation levels of proteins in rice leaf cells during M. oryzae infection. Our analysis revealed that increased ROS levels during pathogen infection resulted in oxidation of many proteins, including metabolic enzymes, protein kinases, and transcription factors. In particular, we found that ribosomal proteins were also oxidized by ROS generated during the plant–pathogen interaction, thus regulating cytosolic translation. This study provides the first in-depth, comprehensive analysis of oxidation-sensitive proteins in plant–pathogen interactions and the effect of ribosomal protein oxidation on translation in crops, expanding our understanding of PTM functions. The identified oxidation-sensitive proteins in rice provide a critical resource for understanding plant–pathogen interactions. Results Identification of oxidation sites and proteins in rice leaves We used an iodo-TMT-based proteomics approach to quantify cysteine oxidation in rice leaves. In brief, proteins from rice leaves were extracted with 25 mM iodoacetamide (IAM) to block free thiols (S-H) and then digested with trypsin. We specifically reduced reversible oxidized thiols (S-NO, S-SG, and S-SP) to thiols with tris(2-carboxyethyl) phosphine (TCEP) and labeled proteins in the samples with different TMT labels. We enriched the final multi-labeled peptides with an anti-TMT antibody and analyzed them using liquid chromatography–tandem mass spectrometry (LC–MS/MS) ([63]Figure 1A). Figure 1. [64]Figure 1 [65]Open in a new tab Site-resolved analysis of the oxidation status of proteins in rice leaves. (A) Experimental workflow for identification of oxidized Cs using a TMT-based approach. R-SH, reduced thiols; S-NO, S-SG, and S-SP represent oxidized thiols in different forms; TCEP, tris(2-carboxyethyl) phosphine; IAM, iodoacetamide. (B) Overlap of oxidized C sites and proteins identified in three biological replicates and the proportion of sites/proteins quantified in at least two of the three biological replicates. The mass errors for oxidation-containing peptides were less than 5 ppm, indicating the high accuracy of our MS data ([66]Supplemental Figure 1A). Most identified peptides varied from 7–21 amino acids in length ([67]Supplemental Figure 1B). We identified 4292 unique cysteine-containing peptides in 2808 proteins ([68]Additional Supplemental File 1), and the 3362 sites from 2275 proteins that were quantified in at least two of three biological replicates ([69]Figure 1B) were considered the total number of discovered sites ([70]Additional Supplemental File 1) and were used for further analysis. Of the 2275 oxidized proteins, 1563 had only one oxidation site, 696 had 2–5 oxidation sites, and 16 had 6 or more oxidation sites ([71]Supplemental Figure 2A and 2C). The most oxidized protein identified here was AUXIN TRANSPORT PROTEIN (encoded by Os09g0247700) ([72]Supplemental Figure 2B). To determine whether oxidized proteins had a common sequence motif, we aligned the amino acid sequences surrounding each oxidation site and compared them with the corresponding sequences in all rice backgrounds. With the positions of the oxidized cysteine (C) residues defined as the 0 positions, we established that lysine residues were enriched from positions −9 to −6 and from positions +7 to +9. Similarly, C was overrepresented at positions −3 and +3, and aspartic acid was enriched at the −5, +1, +2, and +5 positions ([73]Supplemental Figure 3). Cs often occurred in vicinal dithiol structures with two other amino acids (X); such CXXC motifs in various subcellular compartments ([74]Supplemental Figure 3), in which the two reactive sulfhydryl groups are close together and form disulfides or metal clusters, are well-conserved functional elements of proteins. Collectively, these results suggest that oxidation sites are common, broadly distributed, and most often found near relatively polar amino acids. Subcellular localization of oxidized proteins To establish the subcellular localization of oxidized proteins in rice leaves, we examined their predicted subcellular compartments. Of the 2275 oxidized proteins, most localized to the chloroplast (41.8%), cytoplasm (22.6%), or nucleus (14.2%) ([75]Figure 2A). We also analyzed the subcellular localization of oxidized proteins with different numbers of modification sites. In the nucleus and mitochondria, the proportion of proteins with more than five modification sites was lower than that of proteins with one modification site ([76]Supplemental Figure 4). However, we observed the opposite trend in the cytoplasmic membrane. This diversity in the subcellular localization of oxidized proteins suggests that oxidation may have key roles in a variety of cellular processes. Figure 2. [77]Figure 2 [78]Open in a new tab Identification of subcellular localization and functional characterization of C-oxidized proteins in rice leaves. (A) Pie chart of subcellular localization of C-oxidized proteins. (B) Gene Ontology (GO) enrichment analysis of C-oxidized proteins for the biological process, molecular function, and cellular component categories. (C) Biological pathway enrichment analysis of C-oxidized proteins identified in this study. Interaction networks of oxidized proteins To determine the interaction landscape among oxidized proteins in rice leaves, we generated a protein interaction network of all oxidized proteins based on the Search Tool for the Retrieval of Interacting Genes (STRING) database. Most oxidized proteins formed a highly interconnected protein–protein interaction (PPI) network. Some of the complexes and cellular functions defined prominent and highly connected clusters, including the ribosome, proteasome, spliceosome, ubiquitin-mediated proteolysis, glycolysis/gluconeogenesis, aminoacyl-tRNA biosynthesis, oxidative phosphorylation, and protein processing in the endoplasmic reticulum ([79]Supplemental Figure 5). These data suggest that oxidation preferentially occurs in specific protein complexes or functional clusters. Functional annotation of oxidized proteins To elucidate the potential functions of oxidation, we examined the Gene Ontology (GO) functional classifications of all oxidized proteins identified in rice leaves. In the biological process category, the oxidized proteins were mostly enriched in cellular processes (41.3%) and metabolic processes (32.8%). Many oxidized proteins were associated with responses to stimuli, signaling, and the immune system ([80]Figure 2B). In the molecular function category, oxidized proteins were enriched in catalytic activity (57.9%), binding (27.4%), structural molecule activity (4.8%), molecular function regulator activity (1.9%), and translation regulator activity (1.6%) ([81]Figure 2B). In the cellular component category, oxidized proteins were mostly enriched in cellular anatomical entity (78.4%) and protein-containing complexes (21.6%) ([82]Figure 2B). Further pathway enrichment analysis revealed that gene expression, translation, detoxification, responses to stress, metabolic processes, and biosynthetic processes were significantly enriched ([83]Figure 2C), suggesting that protein oxidation might have important regulatory roles in these metabolic pathways under normal conditions. Characterization of oxidized transcription factors and chromatin regulators C oxidative modifications function in many physiological pathways by modulating the activities of key signaling proteins and regulating the function of transcription factors and chromatin regulators to influence gene expression ([84]Poole and Nelson, 2008; [85]Roos and Messens, 2011; [86]Tian et al., 2018; [87]Shen et al., 2020; [88]Zhou et al., 2021a, [89]2021b). GO pathway enrichment analysis revealed that oxidized proteins play a key role in gene expression. A few studies have indicated that C oxidative modifications regulate the function of transcription factors and chromatin regulators to modulate gene expression ([90]Tian et al., 2018; [91]Yang et al., 2021; [92]Zhou et al., 2021a, [93]2021b), suggesting the functional importance of such modifications. In this study, we identified 56 oxidized proteins that are transcription factors or chromatin regulators ([94]Supplemental Table 1). Among these, 46 had 1 oxidation site, 8 had 2 oxidation sites, and 2 had 5 oxidation sites ([95]Figure 3A). Oxidized proteins included many members of the WRKY, v-myb avian myeloblastosis viral oncogene homolog (MYB), SQUAMOSA promoter-binding protein-like (SPL), Plant homeodomain (PHD), homeobox, and zinc-finger families of transcription factors as well as the chromatin regulators histone methyltransferase (SDG or JMJ), acetyltransferase, and histone deacetylase (HDAC) ([96]Figure 3B). These oxidized transcription factors and chromatin regulators form a highly interconnected PPI network ([97]Figure 3C). We therefore hypothesize that oxidation may play a role in regulating the function of transcription factors and chromatin regulators to modulate gene expression. Figure 3. [98]Figure 3 [99]Open in a new tab Characterization of oxidized transcription factors and chromatin regulators. (A) Distribution of the number of oxidized C sites per identified protein. (B) Oxidized transcription factors and chromatin remodelers identified in rice leaves. The colors represent the number of C oxidation sites. (C) PPI network of oxidized transcriptional regulators. M. oryzae infection leads to hyperoxidation of proteins To test whether M. oryzae infection affects oxidation, we performed a multi-parallel LC–MS/MS analysis of Mock and Infection samples. To eliminate the possibility that the oxidation changes may have resulted from changes in protein abundance, we normalized the oxidation proteome data to the corresponding proteome data ([100]Additional Supplemental File 2). Principal component analysis showed good repeatability across three biological replicates of Mock and Infection samples, supporting the robustness of the data ([101]Supplemental Figure 6). Heatmap and boxplot analyses showed that M. oryzae infection significantly increased protein oxidation levels ([102]Figure 4A and 4B). Of the 2690 quantifiable oxidation sites across all samples, 68.3% (1837 sites) showed an increase in oxidation levels in the Infection samples ([103]Figure 4C; fold change > 1, false discovery rate [FDR] < 0.05). Scatterplot analysis revealed that 512 sites from 438 proteins had higher levels of oxidation after infection, whereas 4 sites from 2 proteins had lower levels of oxidation ([104]Figure 4D; [105]Supplemental Table 2; fold change > 1.5, FDR < 0.05). To further confirm that M. oryzae infection leads to increased oxidation, the thiol-modifying reagent 4-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid (AMS) was used to examine the redox state of two selected proteins, ACTIN and GAPDH. AMS can react with reduced C and slow protein mobility on an SDS–PAGE gel, enabling determination of protein redox status as a band shift. The assays revealed that large fractions of the two tested proteins, ACTIN and GAPDH, were present in a reduced state under normal conditions, whereas M. oryzae infection led to their significant oxidation ([106]Supplemental Figure 7A). To further test whether the increased protein oxidation in Infection samples was caused by higher levels of ROS, we examined the oxidation proteome of H[2]O[2] (2 mM)-treated and Mock leaves ([107]Additional Supplemental File 1). Data analysis revealed that 79% (>1-fold) of sites (n = 512) with significantly increased C oxidation in M. oryzae–infected samples also displayed increased oxidation levels in H[2]O[2]-treated samples ([108]Supplemental Figure 7B and 7C), suggesting that hyperoxidation of proteins in Infection samples was caused by an ROS burst. To study the functions of these hyperoxidized proteins in Infection samples, we first analyzed the subcellular localization of the 438 proteins and found that most were localized to the chloroplasts (42.0%) and cytoplasm (31.5%). The rest were localized to the nucleus (9.4%), mitochondria (5.0%), plasma membrane (4.1%), vacuolar membrane (1.6%), endoplasmic reticulum (0.9%), Golgi apparatus (0.2%), or peroxisomes (0.2%) or were extracellular (2.7%) ([109]Supplemental Figure 7D). Although the proportion of oxidized proteins in the rice cytoplasm was 22.6% ([110]Figure 2A), the relative ratio increased to 31.5% after M. oryzae infection ([111]Supplemental Figure 7D). Because the ratio change (8.9%) in the cytoplasm was higher than that in other subcellular fractions, we hypothesized that the oxidized proteins in the cytoplasm might be important for the rice response to M. oryzae infection. Most of these proteins (86.3%) had one oxidation site that showed changes in oxidation level, whereas only three proteins had four oxidation sites with increased oxidation ([112]Supplemental Figure 7E). GO pathway enrichment analysis revealed that proteins with higher oxidation were enriched in aerobic respiration, the tricarboxylic acid cycle, ribonucleotide metabolic processes, purine-containing compound metabolic processes, and photosynthesis. We also noted the enrichment of GO terms for translation, peptide biosynthetic processes, and peptide metabolic processes ([113]Figure 4E), suggesting that oxidation might be involved in the regulation of rice translation during pathogen infection. Figure 4. [114]Figure 4 [115]Open in a new tab M. oryzae infection leads to hyperoxidation of proteins in rice leaves. (A) Heatmap of oxidized C sites on proteins in rice leaves inoculated with M. oryzae (Infection) and control leaves (Mock). (B) Boxplots showing the changes in protein oxidation levels in Infection and Mock rice leaves. (C) Scatterplot of the fold change in the number of oxidized C sites in Infection and Mock rice leaves. (D) Volcano plot of significantly upregulated (fold change > 1.5, FDR < 0.05) and downregulated (fold change > 1.5, FDR < 0.05) oxidized C levels in Mock and Infection rice leaves. (E) GO pathway enrichment analysis of proteins with significantly increased (fold change > 1.5, FDR < 0.05) C oxidation levels. M. oryzae infection leads to increased oxidation of ribosomal proteins We identified 30 translation factors (44 sites) and 44 ribosomal proteins (62 sites) that were oxidized in rice leaf cells under normal conditions ([116]Supplemental Table 3). Among them, 33 ribosomal proteins (10 of which were significantly increased) and 17 translation factors (5 of which were significantly increased) showed increased oxidation levels in Infection compared with Mock leaves ([117]Figure 5A and 5B). To further investigate the hyperoxidation levels of ribosomal proteins in infected leaf cells, we used the ribosomal protein RPL38 for verification. First, we carried out a biotin-conjugated IAM (BIAM) labeling assay to determine whether RPL38 contains a redox-sensitive C residue (C46) that can be oxidized. BIAM and H[2]O[2] selectively and competitively react with redox-sensitive Cs and can thus be used to test the redox sensitivity of a protein. In this assay, we incubated purified maltose-binding protein (MBP)-tagged RPL38 with H[2]O[2] and BIAM and detected BIAM-labeled RPL38 with an anti-BIAM antibody ([118]Figure 5C). We established that MBP-RPL38, but not the MBP control, could be labeled with BIAM and that labeling decreased with an increased amount of H[2]O[2], as expected ([119]Figure 5C). Furthermore, we incubated purified MBP-RPL38^C46A (C mutated to alanine) with H[2]O[2] and BIAM, and BIAM-tagged RPL38^C46A was not detected using an anti-BIAM antibody ([120]Figure 5C). These findings indicate that RPL38 contains H[2]O[2]-sensitive C(s) that might be redox modified. To further confirm that M. oryzae infection leads to increased oxidation of RPL38, we performed AMS labeling of RPL38. AMS is an N-ethylmaleimide variant that reacts specifically with reduced thiols, increasing the size of the labeled protein ([121]Figure 5D). In the presence of AMS, two bands were detected for RPL38. The AMS-labeled band was more pronounced, suggesting that most of the RPL38 was present in a reduced state under normal conditions ([122]Figure 5D). Upon infection, a weak AMS-labeled RPL38 band was detected in Infection leaves, suggesting that a large fraction of RPL38 was in an oxidized state. Collectively, these results indicate that M. oryzae infection leads to increased oxidation levels in ribosomal proteins. Figure 5. [123]Figure 5 [124]Open in a new tab Characterization of identified oxidized transcriptional regulators and ribosomal proteins. (A) Heatmap of oxidized C sites on ribosomal proteins in rice leaves inoculated with M. oryzae (Infection) and control leaves (Mock). (B) Scatterplot of the fold change in the number of oxidized C sites on ribosomal proteins in Mock and Infection rice leaves. (C) Flowchart and detection of OsRPL38 oxidized by H[2]O[2] in a BIAM labeling assay. Increasing amounts of H[2]O[2] led to lower reduction of MBP-OsRPL38 proteins. Purified MBP-OsRPL38 proteins were treated with BIAM and H[2]O[2], which competitively react with reduced Cs that are susceptible to oxidation. S-BIAM represents the thiol labeled with biotin. BIAM-labeled MBP-OsRPL38 was detected by an anti-biotin antibody. Purified MBP protein was used as a control. (D) Flowchart of the AMS shift assay used to detect the redox state of OsRPL38. AMS reacted specifically with reduced thiols to increase protein (OsRPL38) mass by ∼0.5 kDa. In the AMS shift assay, the reduced form of OsRPL38 was labeled by AMS, leading to a gel shift because of increased protein size. M. oryzae infection promotes cytosolic ribosome translation Previous studies have shown that PTMs such as acetylation, phosphorylation, and ubiquitination of ribosomal proteins affect ribosome function ([125]Simsek and Barna, 2017; [126]Xu et al., 2021). Cytosolic translation can be studied using Ribosome profiling (Ribo-seq), which involves sequencing mRNA segments that are protected by ribosomes from nuclease digestion ([127]Andreev et al., 2017). To determine whether increased oxidation of ribosomal proteins in Infection samples alters the ribosome translation process, we performed RNA sequencing (RNA-seq) and Ribo-seq ([128]Supplemental Figure 8A) and analyzed ribosome translational differences between Mock and Infection samples. Three biological replicates were used for RNA-seq and Ribo-seq analysis ([129]Supplemental Figure 9). The lengths of Ribo-seq reads were 20–32 nt, with a peak at 26–28 nt ([130]Supplemental Figure 8B), similar to those from other organisms ([131]Andreev et al., 2017; [132]Xu et al., 2021). About 88% of the mapped reads were localized to annotated exons or coding DNA sequences (CDSs) ([133]Supplemental Figure 8C and 8D). Triplet periodicity is a unique feature of read density distributions of ribosomes because they advance 3 nt at a time during translation. Here, we observed a strong 3-nt periodicity when scanning after the start codon and before the stop codon in the Mock and Infection leaves ([134]Supplemental Figure 8E), indicating that most of the protected mRNAs were being translated by ribosomes. The distribution of aligned Ribo-seq reads over the translated CDS depends on the kinetics of the translation process ([135]Andreev et al., 2017). Metagene analysis of ribosome-binding profiling in all translating mRNAs revealed abrupt peaks of read density surrounding the translation start sites and translation end sites in Mock and Infection leaves ([136]Figures 6A and 6B). This is consistent with previous results indicating that translation initiation and termination may be the rate-limiting steps of translation in rice leaves. In Infection samples, 1690 mRNA species had significantly higher translation levels (>2-fold, P < 0.05) and 895 mRNA species had significantly lower translation levels (>2-fold, P < 0.05) compared with those in Mock samples ([137]Figure 6C; [138]Supplemental Table 4). Almost twice as many genes were upregulated as downregulated at the translational level, indicating that oxidation of ribosomal proteins may promote cytosolic translation. Increased translation might be caused by higher mRNA levels; therefore, we analyzed transcriptional changes in genes due to infection using RNA-seq. We detected 959 genes that were upregulated (>2-fold, P < 0.05) and 683 genes that were downregulated (>2-fold, P < 0.05) at the transcriptional level ([139]Supplemental Figure 10A; [140]Supplemental Table 5). There was a relatively low correlation (r = 0.31) between translational and transcriptional changes ([141]Supplemental Figure 10B). To exclude any translational changes that might actually be caused by transcription, we calculated the translational efficiency (TE) by normalizing the ribosome binding data (Ribo-seq) to mRNA abundance (RNA-seq). Genome-wide data analysis showed that 1011 mRNA species had significantly higher TE (>2-fold, P < 0.05) and 690 had significantly lower TE (>2-fold, P < 0.05) ([142]Figure 6D; [143]Supplemental Table 6) after infection. The transcripts with higher TE corresponded to the most highly expressed genes in the genome ([144]Figure 6E). In addition, their 5′ untranslated regions (5′ UTRs) were longer than the genomic average ([145]Figure 6F) and contained CCT-rich sequences ([146]Figure 6G). GO pathway analysis revealed that these genes were most enriched in defense responses and responses to stimuli ([147]Figure 6H). Collectively, these data suggest that M. oryzae infection results in significantly increased translation of a subset of high-abundance genes that are involved in defense and contain CCT-rich sequences in their 5′ UTRs. Figure 6. [148]Figure 6 [149]Open in a new tab Increased translation in response to M. oryzae infection and characteristics of mRNAs with upregulated TE in infected rice leaves. (A and B) Metagene analysis of RFP density in the coding regions of all rice genes in control (Mock; A) and M. oryzae–inoculated (Infection; B) leaves. RPKM, reads per kilobase per million mapped reads; TSS, translation start site; TES, translation end site. (C) Number of translationally upregulated and downregulated (fold change > 2, P < 0.05) genes in Infection compared with Mock leaves. (D) Volcano plot of genes for which translational efficiency (TE) significantly increased (fold change > 2, P < 0.05) and decreased (fold change > 2, P < 0.05). (E) Analysis of transcript levels of genes with upregulated TE compared with genome-wide gene transcript levels. Genes were divided into three ranks: high 25% (transcripts per kilobase of exon model per million mapped reads [TPM] > 38.3), middle 25%–75% (4.88 < TPM ≤ 38.2), and low 25% (1 ≤ TPM ≤ 4.87). Genes with TPM < 1 were excluded. Significant differences were determined using a two-tailed Student’s t-test. (F) Comparison of 5′ untranslated region (5′ UTR) lengths between Infected genes with upregulated TE and all rice genes. Significant differences were determined using a two-tailed Student’s t-test. (G) MEME motif of the 5′ UTR of genes with upregulated TE during infection. (H) GO pathway analysis of genes with upregulated TE in (D). Increased ribosome binding may lead to translational pause ([150]Li et al., 2012; [151]Das Sharma et al., 2019). An analysis of ribosome binding frequency at individual positions relative to the average mRNA population (using PausePred software; [152]Kumari et al., 2018) revealed an overall lower ribosome pausing score in Infection compared with Mock leaves ([153]Supplemental Figure 11A). The 1011 mRNA species with significantly higher TE in the mutants displayed no difference in pausing score between Mock and Infection samples ([154]Supplemental Figure 11B), suggesting that hyperoxidation of ribosomal proteins did not lead to ribosome stalling in Infection leaves; however, an indirect effect cannot be excluded. To refine the analysis, we analyzed the codon occupancy profiling of the Ribo-seq data. The analysis revealed that there were no clear increases (>1.2) in occupancy at codons within the A site in Infection leaves compared with Mock leaves ([155]Supplemental Figure 11C), suggesting that hyperoxidation of ribosomal proteins indeed promoted protein translation in the cytosol. To further test the above hypothesis, we analyzed proteomics data from Infection and Mock leaves. Comparative analysis between the two groups revealed a higher overall proteomic level in Infection than in Mock leaves ([156]Figure 7A). Further data analysis revealed that 300 proteins showed higher protein levels (>1.2-fold, FDR <0.05), whereas only 100 showed lower protein levels (>1.2-fold, FDR <0.05) ([157]Figure 7B; [158]Supplemental Table 7) in Infection leaves compared with Mock leaves. Furthermore, the protein levels corresponding to the 1011 mRNA sequences with higher TE were significantly higher in Infection than in Mock leaves ([159]Figure 7C), supporting the hypothesis that increased oxidation in ribosomal proteins leads to higher TE. Immunoblotting of proteins that showed significantly increased/decreased abundance further supported this hypothesis ([160]Figure 7D). Figure 7. [161]Figure 7 [162]Open in a new tab Proteome analysis of rice leaves inoculated with M. oryzae (Infection) and control leaves (Mock). (A) Boxplot of protein levels in Mock and Infection leaves. (B) Numbers of significantly upregulated and downregulated proteins (fold change > 1.2, FDR < 0.05) in Infection leaves compared with Mock leaves. (C) Levels of proteins encoded by mRNAs with higher TE in Mock and Infection leaves. (D) Immunoblot detection of EIF3C, CAT2, and RuBisCO protein levels in Mock and Infection leaves. Discussion In this first report of the rice thiol proteome, we present a comprehensive resource of the in vivo thiol oxidation status of about 3362 distinct C peptides in 2275 proteins under normal growth conditions. We discovered that the vast majority of these proteins had one oxidized C residue, consistent with earlier research on organisms ranging from plants to animals ([163]McConnell et al., 2019; [164]Menger et al., 2015; [165]Topf et al., 2018). The oxidation sites were discovered on 2275 protein groups from various subcellular compartments, and specific consensus amino acid motifs were discovered surrounding the C oxidation sites. We also found that exposing rice plants to pathogens resulted in an overall increase in the in vivo oxidation status of the thiol proteome, consistent with previous findings in Arabidopsis ([166]McConnell et al., 2019). In general, our study greatly expands the current repository of oxidative post-translational modifications (oxPTMs) on Cs and validates prior investigations of oxidation-sensitive proteins. The plant immune system comprises two layers: effector-triggered immunity and pathogen-associated molecular pattern (PAMP)-triggered immunity (PTI) ([167]Jones and Dangl, 2006; [168]Wang et al., 2020). The pattern recognition receptors in the plant plasma membrane recognize the PAMPs of pathogens and stimulate the plant PTI response, which can inhibit the growth and spread of most pathogens ([169]Dodds and Rathjen, 2010; [170]Dou and Zhou, 2012; [171]Zipfel, 2014). There are two types of pattern recognition receptors in plants: receptor-like proteins and receptor-like kinases (RLKs) ([172]Wu and Zhou, 2013). In rice, chitin binds to the LysM-glycosylphosphatidylinositol (GPI)-anchored protein OsCEBiP and recruits OsCERK1 (chitin elicitor receptor kinase) to form a complex that initiates PTI signaling through autophosphorylation and dephosphorylation of its kinase domain ([173]Tariqjaveed et al., 2021). We found that many proteins involved in PTI signaling are oxidized, including calcium-dependent protein kinases (OsCDPK8, OsCDPK16, OsCDPK20, OsCDPK24, and OsCDPK28), mitogen-activated protein kinases (OsMPK1, OsMPK5, OsMPK12, OsMPK13, and OsMPK14), and chitin elicitor-binding protein (OsCEBiP). Chitinase stimulates the PTI pathway in plants by degrading fungal chitin, which acts as a PAMP ([174]Miya et al., 2007; [175]Shimizu et al., 2010), and we found that numerous chitinases (OsCht1, OsCht4, OsCht5, OsCht6, OsCht8, and OsCht10) were oxidized. We also found RLKs that were oxidized, including OsFLS2 (flg22 receptor), brassinosteroid leucine-rich repeat (LRR) receptor kinases (OsBRI1, OsBSK1-2, and OsBSK3), Wall-associated kinase (WAK) receptor-like cytoplasmic kinase (OsWAK10a), and SNF1-related protein kinase (OsSAPK5). Therefore, the role of oxidation in the PTI pathway deserves further study, and the datasets generated here provide a valuable resource. Transcription factors and chromatin regulators play key roles in regulating gene expression ([176]Vachon et al., 2018), and C oxidative modifications regulate the functions of transcription factors and chromatin regulators to modulate gene expression ([177]Poole and Nelson, 2008; [178]Roos and Messens, 2011; [179]Tian et al., 2018; [180]Shen et al., 2020; [181]Zhou et al., 2021a, [182]2021b). For example, in rice, GPX1-mediated oxidation of bZIP68 enhances its transcriptional activity and positively regulates abscisic acid (ABA)-independent osmotic stress signaling ([183]Zhou et al., 2021a, [184]2021b). In this study, GO pathway analysis revealed that oxidized proteins are mostly involved in gene expression. We found that many transcription factors are oxidized, including basic region leucine zipper (bZIP), WRKY, MYB, PHD, and CCHC- and C[2]H[2]-type zinc-finger transcription factors. One oxidized rice protein, OsWRKY13, directly or indirectly regulates the expression of genes upstream and downstream of jasmonic acid (JA) and salicylic acid by inhibiting the JA signaling pathway and activating the salicylic acid signaling pathway, thus participating in the regulation of broad-spectrum disease resistance ([185]Qiu et al., 2007). OsWRKY45-2 transcriptionally activates OsWRKY13, and OsWRKY13 directly inhibits OsWRKY42, indicating that these three WRKY transcription factors form a sequential transcriptional regulatory cascade. OsWRKY42 can negatively regulate the response of rice to M. oryzae by inhibiting JA signaling–related genes, and OsWRKY13 can transcriptionally inhibit OsWRKY42 to regulate rice resistance to M. oryzae ([186]Tao et al., 2009). We also identified chromatin regulators that are oxidized, such as histone methyltransferase (SDG or JMJ), acetyltransferase, and HDAC. OsHDA705 is an HDAC that negatively regulates broad-spectrum resistance to rice pathogens ([187]Chen et al., 2021). Therefore, we speculate that oxidative modifications may affect the activity of transcription factors or chromatin regulators to control gene expression, thereby affecting rice immunity. Systematic analysis of oxidized transcription factors and chromatin regulators provides potential candidates for further study of the functional regulation of transcriptional regulators by C oxidation in plant immune responses. Ribosomes, complex ribonucleoprotein-based molecular machines, have a basic housekeeping role in protein translation in cells ([188]Shcherbik and Pestov, 2019). Because plants are sessile, the adjustment of protein synthesis is crucial for plant acclimation to ever-changing environmental cues ([189]Moore et al., 2016). Ribosomal proteins can undergo different post-translational modifications, such as acetylation and phosphorylation ([190]Xu et al., 2021; [191]Zhen et al., 2022), which may alter ribosome function. Here, we found that ribosomal proteins can also be oxidized under normal conditions in rice leaves; we identified 56 ribosomal proteins with C oxidation, suggesting that oxidation of ribosomal proteins might be necessary for the formation of subunits (40S and 60S) or the translation process. Cytosolic translation can be regulated by viral infection in animal cells ([192]Sonenberg and Hinnebusch, 2009). Here, we found that cytosolic translation might also be modulated by pathogen infection in plants because M. oryzae–infected plants showed hyperoxidation of ribosomal proteins. To test this hypothesis, we performed ribosomal footprint (RFP) sequencing and found that hyperoxidation of ribosomal proteins promoted global cytosolic translation, especially that of mRNAs involved in plant defense responses such as RLK (Os01g0113400), WRKY42 (Os02g0462800), and WAK-LIKE KINASE (Os03g0225700). Previous studies have revealed that many biotic responsive genes can be induced under pathogen infection, thereby regulating the disease response in planta ([193]Shimono et al., 2007; [194]Hu et al., 2017; [195]Jin et al., 2018; [196]Liu et al., 2018). However, our study revealed that, in addition to inducing gene transcription, pathogen-derived oxidative stress also promotes translation of biotic responsive genes via ribosomal oxidation. Most genes with increased translation were not those whose transcription was promoted: the correlation between translation and transcription was relatively low (r = 0.3). This result suggests that plant cells use not only the transcriptional apparatus but also the translational apparatus to rapidly respond to changes in phytopathogen invasion. Ribosomes may act as a nexus by integrating information from metabolism and the environment to fine-tune plant growth and defense in response to biotic challenges. Collectively, our results identified a novel mechanism by which ribosomal proteins sense redox regulation in plant translation, which involves ROS signaling during pathogen infection in planta. We identified many transcription factors and chromatin regulators as oxidized proteins. In particular, OsWRKY13 (Os01g0750100) and OsHDA705 (Os08g0344100) have been reported to participate in rice blast resistance ([197]Tao et al., 2009; [198]Chen et al., 2021). We hypothesize that oxidation may play a role in regulating the function of transcription factors or chromatin regulators to modulate rice resistance. Moreover, we found that M. oryzae infection can promote oxidative modification of translation factors and ribosomal proteins that are involved in protein translation. Ribo-seq data suggested that oxidative modification of ribosomal proteins significantly increases the TE of many defense-related genes. These findings indicate that rice resistance may be induced by oxidative modification of ribosomal proteins. A direct link between these oxidized proteins and blast resistance should be explored further through extensive functional studies. In conclusion, our comprehensive study reveals the intricate nature of the thiol proteome produced by M. oryzae in rice, and the dataset offered here provides a framework for decoding protein oxidation activities in plant immune responses. The resulting omics databases provide not only a comprehensive landscape of plant genomes but also valuable opportunities for crop development. Methods Rice sampling and protein extraction and digestion The M. oryzae strain P131 was used for pathogenicity assays. Rice (japonica cultivar Lijiangxintuanheigu) seedlings at the fifth-leaf stage were sprayed with M. oryzae P131 conidial suspensions (5 × 10^4 conidia ml^−1) mixed with 0.25% Tween 20, and samples were collected 24 h post inoculation (hpi) when the level of ROS in host cells was high ([199]Liu et al., 2020a, [200]2020b). Rice plants injected with water served as controls. Control rice leaves (Mock) were sprayed with 2 mM H[2]O[2] for 6 h (H[2]O[2]), and infected rice leaves (Infection) were sampled and used for proteome, oxidome, and transcriptome (RNA-seq) analyses and translatome (Ribo-seq) analyses. Frozen leaf samples (approximately 0.5 g for each replicate) were ground to a fine powder using liquid nitrogen, and 4 volumes (compared with the volume of powder) of lysis buffer (0.15 M NaCl, 1% Triton X-100, 2 mM EDTA, 0.1% SDS, 0.02 M Tris-HCl [pH 8.0], 10 mM DDT, 25 mM IAM, and 1% protease inhibitor) were added. The samples were lysed ultrasonically. An equal volume of Tris-phenol was added, and the samples were centrifuged at 12 000 g for 10 min at 4°C. Proteins were precipitated from the supernatant with 0.1 M ammonium acetate/methanol (five times the volume of the supernatant) overnight. Protein pellets were washed with methanol and acetone, and the final pellet was reconstituted with 1% SDS. Protein concentrations were determined using a bicinchoninic acid kit according to the manufacturer’s instructions. Proteins (300 μg for each sample) were digested according to a previously described procedure ([201]Chen et al., 2021). First, the samples were enzymatically digested with trypsin overnight at 37°C using a 1:50 trypsin-to-protein mass ratio, followed by another 4 h of digestion using a 1:100 trypsin-to-protein mass ratio. The resulting peptide solution was acidified with 0.5% trifluoroacetic acid, desalinated using a SepPak C18 cartridge (Waters, Milford, MA, USA), and then vacuum dried. Peptides were reduced using 10 mM dithiothreitol (DTT) and simultaneously alkylated with 20 mM IAM for 45 min at room temperature in darkness, and the urea concentration was then adjusted to below 2 M using 50 mM ammonium bicarbonate. Iodo-TMT labeling Iodo-TMT labeling was performed using a TMT kit (Thermo Fisher Scientific, USA), following the manufacturer’s protocol. For total C labeling, each sample (100 μg, 1 μg/μl) was reduced for 1 h at 37°C using 10 mM TCEP and then alkylated with 0.9 mM iodo-TMT. Iodo-TMT labeling for all samples continued for 2 h with occasional shaking at 37°C. Then, 500 mM DTT was added to quench the reaction, and samples were incubated for 15 min, pooled (see [202]Supplemental methods for detailed information on mixing of the peptides), and precipitated with 4 volumes of chilled acetone overnight at −20°C. The resulting pellets were washed three times with ice-cold acetone to remove excess iodo-TMT reagents and dissolved in buffer (8 M urea, 50 mM ammonium bicarbonate [pH 8.0]). The TMT tags for Mock, Infection (IF), and H[2]O[2] were 126, 127, and 128, respectively. Fractionation and enrichment of oxidized peptides The tryptic peptides generated above were fractionated by high-pH reverse-phase high-performance LC (HPLC) using an Agilent 300 Extend C18 column (5-μm particles, 130 Å, 4.6 × 250 mm). The peptides were combined into six fractions and vacuum dried. For affinity enrichment, anti-TMT resin (100 μl/100 μg peptide) was washed three times with 1× Tris-buffered saline and incubated with peptides for 2 h at room temperature or overnight at 4°C. The supernatant was discarded, and the resin was washed five times with Tris-buffered saline and then three times with 1 column volume of water each time. Peptides were eluted using 4 column volumes of TMT elution buffer (40% acetonitrile, 5% trifluoroacetic acid) and lyophilized. The eluted peptides were desalted with a C18 ZipTip pipette tip (EMD Millipore, Billerica, MA, USA) for HPLC–MS/MS analysis. HPLC–MS/MS analysis The peptides were dissolved in solvent A (0.1% formic acid in 2% acetonitrile) and separated using an EASY-nLC 1200 ultra-performance LC system (Thermo Fisher Scientific). Solvent B was 0.1% formic acid in 90% acetonitrile. The following liquid-phase gradient was used: 0–38 min, 8%–23% B; 38–52 min, 23%–35% B; 52–56 min, 35%–80% B; and 56–60 min, 80% B. The flow rate was maintained at 500 nl/min. The resulting peptides were subjected to a nano-electrospray ionization ion source followed by Q Exactive HF-X MS (Thermo Fisher Scientific) coupled online to an EASY-nLC 1200 ultra-performance LC system (Thermo Fisher Scientific). The electrospray voltage was set at 2.0 kV. The peptide parent ions and their secondary fragments were detected and analyzed using a high-resolution Orbitrap MS. The scanning range of first-level MS was set to 350–1400 m/z, and the scanning resolution was set to 120 000. The second-level mass spectrum scanning range was set at a fixed starting point of 100 m/z, and the Orbitrap scanning resolution was set to 30 000. In data acquisition mode, a data-dependent scanning (DDA) procedure was used, in which the top 20 peptide parent ions with the highest signal strength were successively selected to enter the higher-energy collisional dissociation collider after the first-level scanning, with 28% of the fragmentation energy used for fragmentation, and the second-level MS analysis was conducted successively. To improve the effective utilization of the mass spectrum, the automatic gain control was set to 1E5, the signal threshold was set to 50 000 ions/s, the maximum injection time was set to 100 ms, and the dynamic exclusion time of tandem mass spectrum scanning was set to 10 s to avoid repeated scanning of parent ions. Database searches and bioinformatics analyses The acquired MS/MS data were searched with MaxQuant software (version 1.5.2.8) against the UniProt O. sativa subsp. japonica database (48 932 protein sequences). Trypsin/P was specified as the cleavage enzyme, and a maximum of two missing cleavage sites was allowed. The mass tolerance for precursor ions was set to 20 ppm in the first search and 5 ppm in the main search. The mass tolerance for fragment ions was set to 0.02 Da. FDR thresholds for peptides, proteins, and modification sites were less than 1%. Oxidized peptides were considered false positives and removed from our list when the peptides were identified from reverse or contaminant protein sequences, when the peptides had a score below 40, when the site localization probability was less than 0.75, or when the oxidized sites mapped to the C terminus of the peptide, unless the peptide C terminus was the end of the corresponding protein. To calculate P values of differentially expressed or oxidized proteins between Mock and Infection samples, unique peptide quantitative intensity values were first horizontally normalized (see [203]Supplemental methods for the detailed normalization method). Next, the normalized quantitative value of each modification site was divided by the corresponding relative quantitative value of the proteome and then log[2] transformed. Two-sample two-tailed Student’s t-tests were used to calculate P values and FDR values (Benjamini–Hochberg test). Proteins with an FDR value of less than 0.05 and expression ratio greater than 1.5 were regarded as upregulated, whereas proteins with an FDR value less than 0.05 and expression ratio less than 1/1.5 were regarded as downregulated. PPI and GO annotation analyses were carried out using STRING (version 10.5) and DAVID software. Sequence logo representations of significant motifs were identified and generated using Motif-X software (version 5.0.2). Pathways were classified into hierarchical categories according to the KEGG website. WoLF PSORT (version 0.2) was used to predict protein subcellular localizations. Recombinant protein expression and purification To express recombinant MBP-OsRPL38, the full-length CDS of OsRPL38 was cloned into pMALC2G. Plasmids were transformed into Escherichia coli BL21 (DE3) cells. Recombinant proteins were induced by addition of 0.1 M isopropylthio-β-galactoside, and cells were incubated at 18°C overnight until the optical density at 600 nm reached 0.8. Proteins were purified by gravity flow with amylose resin (New England Biolabs, USA) to obtain MBP-tagged proteins. BIAM labeling assays BIAM labeling and biotin-switch assays were performed as described previously ([204]Tian et al., 2018). For BIAM labeling, purified MBP-OsRPL38 proteins were treated with different concentrations of H[2]O[2] at room temperature for 20 min in the dark. Proteins labeled with BIAM were separated on a 12% SDS–PAGE gel and detected with anti-biotin and horseradish peroxidase-linked antibodies (1:5000 dilution, Cell Signaling Technology, USA). Input MBP or MBP-OsRPL38 proteins were detected using an anti-MBP antibody (1:5000 dilution, New England Biolabs, USA). AMS shift assay The AMS shift assay was performed as described previously ([205]Yang et al., 2021). In brief, 50 μl of extracted total proteins from Infection and Mock leaves were incubated in 50 μl of AMS trapping buffer (20 mM AMS [Thermo Fisher Scientific], 1 mM EDTA, 1% SDS, and 50 mM Tris-HCl [pH 7.5]) at 37°C for 1 h. Proteins were separated on a 12% non-reducing SDS–PAGE gel and detected with mouse anti-RPL38 antibody (Abcam, China). RNA-seq and data analysis RNA samples were isolated using TRIzol reagent (TransGen Biotech, Beijing, China) as described previously ([206]Chen et al., 2021). The RNA-seq libraries were generated using the Illumina TruSeq RNA sample preparation kit and sequenced on an Illumina HiSeq 4000 system (IGENEBOOK Biotechnology, China) using the PE150 method. RNA-seq data were filtered using Cutadapt (version 0.33) ([207]Martin, 2011) to remove contaminants and low-quality reads. STAR (version 2.6.1a) was used to map clean reads to the rice genome (Os-Nipponbare-Reference-IRGSP-1.0), and featureCounts ([208]Liao et al., 2014) was used to summarize the features of all RNA-seq libraries. Differentially expressed genes were identified using DESeq2 ([209]Love et al., 2014) with default parameters, a fold-change threshold of >2, and a P value threshold of <0.05. Ribo-seq and data analysis To test the effect of hyperoxidation of ribosomal proteins on translation, Mock and Infection samples were used for Ribo-seq. Isolation of ribosome-protected fragments (RPFs) and library construction were performed as described previously ([210]Su et al., 2018). Rice leaves were ground to powder in liquid nitrogen and dissolved and homogenized in 10 ml polysome extraction buffer (10% Triton X-100, 20 mM β-mercaptoethanol, 5 mg/ml cycloheximide, and 1 U/μl DNase I) on ice. Samples were then centrifuged at 12 000 g for 10 min at 4°C. The supernatant was collected and digested with RNAse I (TransGen Biotech, Beijing, China). The supernatant was filtered through Microspin S-400 columns (GE Healthcare, USA), and the ribosome pellet was collected. TRIzol reagent (Thermo Fisher Scientific) was used to extract the RPFs, and fragment size selection was performed with a NucleoSpin miRNA kit (Macherey Nagel, USA) to collect fragments smaller than 200 nt. A urea–PAGE gel (10%) was used for size selection of 24–35-nt fragments. After dephosphorylation using PNK (New England Biolabs), the RPFs were ligated to the AIR adenylated RNA linker (BIOO Scientific, USA). The ligation products were subjected to reverse transcription using Super Script III (Thermo Fisher Scientific) and circularization using CircLigase II (Illumina). Sequencing libraries were ultimately obtained through PCR amplification by Q5 polymerase (New England Biolabs). The ribosome profiling libraries were sequenced on an Illumina HiSeq 2000 system at Novogene to generate paired-end 150-nt reads. Quality filtering and adaptor clipping were carried out using FASTX-toolkit ([211]Pearson et al., 1997). rRNA sequences were removed from Ribo-seq data with Bowtie (version 0.12.9) ([212]Langmead, 2010). Mapping to the japonica Nipponbare reference genome ([213]http://rapdb.dna.affrc.go.jp/download/irgsp1.html) was performed using STAR (version 2.6.1a) ([214]Dobin et al., 2013). Alignment was performed with the following parameters: -alignEndsType EndToEnd-outFilterMismatchNmax 3. The translation abundance of each gene was calculated using featureCounts. Only reads located in the CDS were used to calculate translation abundance. DESeq2 ([215]Love et al., 2014) was used to analyze differential expression from uniquely aligned RNA-seq reads and differential RFP abundance from ribosome profiling reads that aligned uniquely to the CDS of each gene. RiboDiff ([216]Zhong et al., 2017) was used to analyze differential RFP per mRNA. Metagene analysis over start/stop codons was defined by IRGSP-1.0 protein-coding genes, and data were statistically analyzed and plotted in R (version 3.5). To visualize ribosome occupancy, all mapped reads (bam files) that overlapped each bin (bin size = 10) were calculated and then normalized to the japonica Nipponbare reference genome to obtain 1× depth of coverage (RPGC) using bamCoverage of deepTools2 ([217]Ramirez et al., 2014). A metaplot matrix was generated with deepTools2 and plotted with R. Genes with a P value <0.05 and a fold change >2 were considered to be differentially expressed. Ribosome pause and codon occupancy were analyzed using RiboToolkit ([218]Liu et al., 2020a, [219]2020b) with default parameters. Data availability Proteome data were deposited into the ProteomeXchange Dataset (PRIDE) under accession number PXD034430. RNA-seq and Ribo-seq data were deposited into the Sequence Read Archive (PRJNA817730, NCBI) and Genome Sequence Archive (CRA009462, China National Genomics Data Center). Funding This work was supported by the Fundamental Research Funds for the Central Universities of China (2021ZKPY016), the National Natural Science Foundation of China (32100465), and the Postdoctoral Science Foundation of China (2021M691183). Author contributions L.Z. conceived the project. Xiaoyang Chen conducted most of the experiments. Q.X. performed bioinformatics analysis. Y.Y. and Y.D. conducted parts of the research. H.L., Xiaolin Chen, and J.H. provided critical suggestions for structure and writing of the manuscript. Xiaoyang Chen and Q.X. wrote the manuscript. L.Z. supervised all aspects of the study and critically reviewed and revised the manuscript. All authors read and approved the contents of this manuscript. Acknowledgments