We coupled LC-MS/MS technology with an in vitro assay system to obtain the time-resolved proteome data for P-activated Xoo cells (Scheme 1). The in vitro assay system recapitulated the initial interaction between Xoo cells and damaged rice leaf tissues at the site of infection. Fresh RLX were prepared by grinding the leaves of a Xoo-susceptible rice cultivar (Milyang 23) in liquid nitrogen and added to a Xoo cell culture in the mid-exponential phase. Samples for proteome analysis were collected from RLX-treated (P-activated) and untreated (control) Xoo cells at 0, 30, 60, 90, and 120 min after RLX treatment (Supplementary Table 1).

Analysis on UniProt revealed that the total 4,956 predicted open reading frames in the Xoo genome (KACC10331) corresponded to 4,382 proteins in the proteome, of which 2,589 proteins (59%) were identified for at least one time point and 2,296 proteins (52%) were detected in both replicates (Supplementary Figure 1A). Median sequence coverages for total identified proteins were 24% and 23% for each of the independent duplicate experiments (Supplementary Figure 1B). A total of 20,963 and 19,684 non-redundant peptides were identified in both replicates, with 47,025 and 41,428 peptide-spectrum matches, respectively (Supplementary Figure 1C).

Protein abundance values obtained after quantile-normalization (Supplementary Figure 2A) were used for pairwise comparisons; the Pearson’s correlation coefficients corresponding to the abundance values showed close correlations (0.98–0.99), indicating comparable cellular concentration of most proteins (Supplementary Figure 2B). The smallest correlations were observed for the RLX-treated samples at 30 min, indicating greater changes in the proteome during the initial 30 min; this was consistent with the transcriptome data. We further performed a principal component analysis to determine the relationships between the assessed samples. Supplementary Figure 2C shows that the P-activated sample at 0 min clustered closely with all control samples, whereas the P-activated samples at other time points were more spread out. The proteome of the P-activated Xoo cells at 30 min was considerably different from that of P-activated Xoo at other time points; this was consistent with the results of the pairwise multi-scatter plot (Supplementary Figure 2B).

Quantitative proteome data obtained from P-activated and control Xoo cells at every 30 min allowed the visualization of the three-dimensional protein expression data in terms of the genes, time intervals, and expression levels (Figure 1 and Supplementary Table 2). The expressed protein level of each gene from P-activated Xoo cells are divided by that of the same gene from control at each time point to calculate the fold change of time-dependent protein expression level of the specific gene (Supplementary Table 3). For all the open reading frames, approximately 93 (2.0%), 213 (4.5%), and 468 (9.9%) proteins were upregulated by more than 200% (two-fold), 50%, and 20% at 30 min, respectively, and approximately 7 (0.1%), 93 (2.0%), and 561 (11.9%) proteins were downregulated to less than 25% (two-fold), 50%, and 80% at 30 min, respectively.

In case of the duration of expressed proteins in P-activated Xoo, only 8 (0.2%), 32 (0.7%), and 75 (1.6%) proteins were upregulated for entire 120 min by more than 200% (two-fold), 50%, and 20%, respectively, and 0 (0%), 1 (0.02%), and 87 (1.8%) proteins were downregulated for the same 120 min to less than 25% (two-fold), 50%, and 80%, respectively (Supplementary Table 4), indicating that most proteins were temporarily upregulated or downregulated in the in vitro assay.

The proteins at 30 min presented the highest change at the expression level within 120 min. More than 90 proteins (1.9%) were upregulated by more than two-fold at 30 min in each dataset, whereas 27 proteins were upregulated in both datasets from the duplicate experiments (Supplementary Table 5). Almost half of the 27 upregulated proteins were related to cell motility and ion uptake, seven were related to chemotaxis and motility, and five were related to transporters and pumps. At the same 30 min, more than 90 proteins were downregulated by more than 50%, 17 proteins were downregulated in both datasets (Supplementary Table 6), including transcription-related proteins—such as sigma-54 modulation protein and MetE/MetH family transcriptional regulator—and cell division- and cell cycle-related proteins.

To study the global gene expression pattern based on gene function, we superimposed time-dependent protein expression levels as per the functional categories in the Clusters of Orthologous Groups of proteins (COGs) database (Figure 1), which were grouped into three classes (red, yellow, and blue) depending on the observed pattern. To facilitate the comparison of time-dependent protein expression levels, the expression level at 0 min was set as the reference level (=1) for each gene, and the fold change in the protein expression level was calculated at each time point, as in case of the RNA-Seq data analysis (Supplementary Table 2; Kim et al., 2016).

The most prominent changes in protein expression in Xoo cells were observed for proteins associated with cell motility (N) and inorganic ion transport and metabolism (P), which were placed in the red class (Figure 2). In category N, two major expression peaks, indicating more than two-fold upregulation, were detected at 30 and 90 min. In category P, upregulated proteins peaked at around 30 min. Proteins grouped in the yellow class exhibited moderate changes in their expression level, and were divided into nine functional categories, including signal transduction (T), intracellular trafficking, secretion, and vesicular transport (U), energy production and conversion (C), transcription (K), replication (L), cell wall/membrane/envelop biogenesis (M), and protein turnover (O). The blue class included proteins associated with other nine categories, the expression levels of which exhibited little change when compared with those at 0 min, albeit except for a few proteins.

We integrated two different molecular levels of mRNAs and proteins to investigate the time-resolved systematic analysis of both transcriptome and proteome in P-activated Xoo cells. Both transcriptome and proteome data were obtained for up to 120 min after RLX treatment (Scheme 1). The experimental procedure used to prepare protein samples for LS-MS/MS required at least 30 min, whereas that for mRNA samples required at least 5 min; the shortest interval of time for the proteome data was set to 30 min, whereas that for transcriptome data was set to 5–30 min. Among the protein-coding genes observed in both proteome and transcriptome, we selected 1,868 genes which were detected more than three times in both omics layers. The integrative expression matrix was generated by merging two separate expression z-score matrices of mRNAs and proteins.

After performing the hierarchical clustering analysis to the integrative expression matrix, we found out two major proteomic patterns of upregulated and downregulated protein levels (pink and blue boxes, respectively, in Figure 3). The two lists of proteins were used for STRING analysis, which produced two interaction networks (Supplementary Figures 3A,B). The two groups of proteins can be sub-grouped into several smaller patterns by the time-resolved mRNA expression. Therefore, we identified the five sub-patterns with the genes in the pink and blue boxes (Figure 3 and Supplementary Table 7). In detail, 343 genes, downregulated in protein level (P_D), were sub-grouped into two patterns of 187 and 156 genes in P_D + R_D and P_D + R_U, respectively, whereas 311 genes, upregulated in protein level (P_U), were sub-grouped into three patterns of 64, 159, and 88 genes in P_U + R_EarlyU, P_U + R_LateU, and P_U + R_D, respectively. Among the five sub-patterns, three sub-patterns of P_U + R_EarlyU, P_U + R_LateU, and P_D + R_D showed similar expression patterns in both mRNAs and proteins, and other patterns of P_U + R_D and P_D + R_U) showed opposite expression patterns (Figure 3). Although both P_U + R_EarlyU and P_U + R_LateU showed the increased proteins and mRNAs levels in P-activated conditions, P_U + R_EarlyU showed the upregulation of mRNA in early time points of 0–15 min, in contrast, P_U + R_LateU showed the upregulation of mRNA in late time points of 30–120 min.

We then profiled the temporal fold changes of mRNAs and proteins between P-activated and control Xoo cells (Figure 3). As the proteome datasets do not include protein levels in time points of 5, 10, 15, and 45 min, in which the protein levels were extrapolated based on the protein levels at other time points using non-linear regression.

Xanthomonas oryzae pv. oryzae (Xoo) strain KACC10331, consisting of 4,941,439 nucleotides and 4,733 open reading frames, without any apparent autonomous plasmids, was obtained from the Korean Agricultural Collection Center (KACC) (Lee et al., 2005). The bacteria were cultured in nutrient broth (Difco, Detroit, MI, United States) or Yeast Glucose Cm Agar (YGC) [1% yeast extract, 2% D-(+)-glucose, 2% CaCO₃, and 1.5% agar] at 28°C.

The effector genes XoAvrBs2 (Xoo0168) and XoAvrBs3—including the promoter region (from −149 to −750 bp to the start codon of the respective gene)—were amplified by PCR and ligated into the pGEM-T Easy Vector (Promega). The cloned sequences were verified and then digested with KpnI and SacI, and the products were ligated into the pHM1-XTAP-T_gap vector. The recombinant vectors were purified and introduced into Xoo strain KACC10331 by electroporation, using Gene Pulser II (Bio-Rad, Hercules, CA, United States) with a 0.2 cm-gap cuvette at 2.5 kV cm^–1. Xoo cells were then diluted immediately with 1 mL Super Optimal Broth (SOC) medium and incubated at 28°C with agitation for 2 h. Cells were then recovered from the culture medium and plated on nutrient broth agar plates containing 50 μL mL^–1 spectinomycin and incubated at 28°C for 4 days.

The transformants were cultured in 100 mL of nutrient broth up to the mid-exponential phase (OD₆₀₀ = 0.5). Cells were harvested by centrifuging 1 mL of the cell culture at 12,000 rpm and 4°C for 5 min. Harvested cells were washed once with phosphate-buffered saline (PBS) at pH 7.2, resuspended in 200 μL PBS, and then sonicated. Protein samples were serially diluted using 2 M urea in PBS in 96-well plates and then transferred to a polyvinylidene difluoride membrane (PVDF; 0.2 μm, Bio-Rad) using a 96-well vacuum dot-blotter (Bio-Rad). The membrane was then washed thrice with PBS, blocked with 5% skim milk for 30 min, and subjected to a one-step immuno-affinity reaction using the rabbit peroxidase-anti-peroxidase soluble complex antibody (Sigma-Aldrich, St. Louis, United States). The membrane was developed, and bound antibodies were detected by chemiluminescence.

Oryza sativa L. cv. Milyang 23, a Xoo-susceptible rice cultivar, was used for performing proteome analysis. Rice plants were grown in a paddy field at Jeonju in South Korea (35°49′52.0″N 127°03′55.6″E) until panicle initiation (approximately 8–9 weeks). Forty clumped rice leaves were harvested and homogenized with liquid nitrogen using a mortar and pestle. One-gram aliquots of the resulting RLX were transferred to Eppendorf tubes and stored at −80°C. Xoo was cultured (100 mL) in nutrient broth up to the mid-exponential phase (OD₆₀₀ = 0.5) in a shaking incubator at 28°C and 200 rpm, and RLX (2 g) was then added to the culture medium. The culture (100 mL) was subjected to sequential filtration through a gauze, 40-μm nylon cell strainer (FALON, New York, United States), and 5-μm syringe filter (Sartorius, Germany) to remove RLX (0, 30, 60, 90, and 120 min after RLX addition). The filtered culture (100 mL) was centrifuged at 10,000 × g and 4°C for 10 min. Duplicate samples were obtained for each time point from two independent experiments.

The harvested samples were lysed in a lysis buffer containing 9 M urea prepared in 20 mM HEPES (pH 7.5), supplemented with protease inhibitor cocktail (Complete mini, Roche) and phosphatase inhibitor (PhosSTOP, Sigma-Aldrich) and sonicated on ice. The exact amount of proteins in each sample was determined using the bicinchoninic acid assay. Protein integrity was confirmed by SDS-PAGE and 200 μg of protein from each sample was used for analysis. The disulfide bonds were reduced by treatment with 10 mM dithiothreitol for 1 h, and incubation with 30 mM iodoacetamide (30 min in the dark) was performed to alkylate free sulfhydryl functional groups. Samples were diluted with triethylammonium bicarbonate buffer (pH 8.0) in a manner such that the final urea concentration was 1.5 M. Proteins were digested using MS grade trypsin (Thermo Fisher Scientific) at a protein to enzyme ratio of 50:1 for 12 h at 37°C. The reaction was quenched by lowering the sample pH (<3) using trifluoroacetic acid. The obtained peptides were desalted using a C18 spin column (Harvard) to remove salts and other contaminants, and the purified peptides were dried. Then, they were isotopically labeled using the 10-plex tandem mass tag (TMT, Thermo Fisher Scientific), as per the manufacturer’s protocol. The reaction was allowed to continue for 2 h at room temperature and TMT-labeled samples were subsequently dried in a SpeedVac concentrator. Chemical labeling with TMT was confirmed by liquid chromatography-tandem mass spectrometry (LC-MS/MS), and the samples were pooled and fractionated using a basic reverse phase liquid chromatography (RPLC) system. The pooled TMT-labeled peptide mixture was resuspended in 10 mM ammonium formate and fractionated into 12 fractions using a C18 column (C₁₈, 5 μm pore size, 4.6 mm × 250 mm, XBridge, Waters). The fractionated peptides were dried and stored at −80°C until LC-MS/MS analysis.

Each fractionated peptide sample was analyzed using an Orbitrap Fusion^TM Lumos^TM Tribrid^TM Mass Spectrometer coupled with the Easy-nLC 1200 nano-flow liquid chromatography system (Thermo Fisher Scientific). The dried peptides were reconstituted using 0.1% formic acid and loaded on a C18 trap column. Peptides were resolved using a linear gradient solvent B (0.1% formic acid in 95% acetonitrile) and analyzed by high resolution mass spectrometry in the data-dependent acquisition mode. MS1 and MS2 were acquired for the precursor and the peptide fragmentation ions, respectively. MS1 scans were measured at a resolution of 120,000 and an m/z of 200. MS2 scans were acquired following the fragmentation of precursor ions by high-energy collisional dissociation (HCD) and were detected at a mass resolution of 50,000 and an m/z of 200. Dynamic exclusion was used to reduce redundant fragmentation of the same ions. The obtained mass spectrometry data were analyzed using the MaxQuant software (Tyanova et al., 2016a). Raw MS data were searched against the Xoo proteome in UniProt database. Carbamidomethylation of cysteine and 10-plex TMT modification of lysine and N-terminals were set as static modifications, whereas oxidation of methionine was set as a variable modification. False discovery rates at the levels of protein and peptide-spectrum matches were set at 0.01. The raw MS data and MaxQuant search results have been submitted to ProteomeXchange (project accession: PXD020135, reviewer access with username: reviewer34070@ebi.ac.uk and password: mzn76I1O).

The contaminant and reverse identified proteins were removed from the MaxQuant data. Proteins identified in both replicates were pooled for quantile normalization. The normalized values for the replicates were subjected to supervised hierarchical clustering and principal component analysis, using Perseus (Tyanova et al., 2016b), and the results were depicted in the form of a multi-scatter plot.

In addition to the previously obtained RNA-Seq data for P-activated and control Xoo cells, RNA-Seq data at 0, 90, and 120 min were obtained to correspond with the proteome data at these time points. RNA-Seq and data analysis were performed as previously described (Kim et al., 2016). Briefly, total RNA from samples was used to generate sequencing libraries, from which ribosomal RNA was removed using MICROBExpress Bacterial mRNA Enrichment Kit (Ambion, Austin, TX, United States), and enriched mRNA was prepared using Illumina TruSeq RNA Sample Preparation Kit (Illumina, San Diego, CA, United States). The RNA obtained after fragmentation was used to generate cDNA fragments, which were sequenced using Illumina Genome Analyzer IIx and mapped to the reference genome sequence¹ using CLC Genomics Workbench 4.0 (CLC bio, Aarhus, Denmark). Relative transcript abundance was calculated based on the number of reads per kilobase per million mapped sequence reads (RPKM).

The RPKM values of the mRNAs in the transcriptome and TMT intensities of the proteins in the proteome corresponded to the observed expression level of each gene at a given time point. To facilitate the comparison of gene expression levels, the observed expression level at each time point was converted to fold change in gene expression, by dividing the expression level at a given time point by the initial expression level (0 min) of the same gene. The fold change in the time-resolved expression levels of a given gene during the 2 h following the RLX treatment was fitted to a curve and analyzed using non-linear regression by GraphPad Prism (version 3.02 for Windows, GraphPad Software, San Diego, CA, United States)², and the continuous time-dependent changes in the mRNA and protein expression levels were determined using the fitted curve.

For the simple comparison of the mRNA and protein expression levels in P-activated Xoo cells, the expression levels were corrected by comparing with those in the control cells at each time point. The fold change in mRNA and protein expression in P-activated Xoo cells at a given time point was divided by that of the control cells at the same time point. The resulting control-corrected fold change values of mRNAs and proteins were considered as the relative gene expression levels of mRNAs and proteins at each time point.

For the comparison of mRNA and protein datasets, we first generated the integrative expression matrix by merging the two omics expression matrices. In detail, each matrix was processed via following steps. For proteome expression matrix, two TMT expression matrices were log2-quantile normalized and z-transformed independently. For transcriptome expression matrix, both RPKM matrices of dataset set1 for 0, 5, 10, 15, 30, 45, and 60 min and dataset set2 for 0, 90, and 120 min were log2-quantile normalized. The values in set2 were normalized by that at set2 0 min, centered by that at set1 0 min, and merged into set1 sequentially. The merged RPKM matrix was z-transformed. Among the detected protein-coding genes in proteome and transcriptome, we selected genes with observation of more than 3 times in both omics matrices for further analysis and made the integrative expression matrix of TMT set1, TMT set2, and merged RPKM. Then, we performed the hierarchical clustering analysis (distance: average; method: Pearson’s method) to investigate the gene expression patterns.

Based on the hierarchical clustering result, we selected the optimal sub-pattern number for upregulated and downregulated proteome patterns (three sub-groups for upregulated genes and two sub-groups for downregulated genes) and performed the k-means clustering to each proteome pattern.

The list of proteins in the upregulated and downregulated patterns from the hierarchical clusters was uploaded on the STRING database³ to analyze the protein interaction maps. The clusters of proteins in the upregulated and downregulated classes, which show the similar time-resolved expression patterns, were analyzed, and the list including the gene names with the selected organism was inputted in the multiple proteins search setting. The number of nodes and edges were automatically calculated based on Xoo genes with a PPI enrichment p-value of 2.09E-9. Figures were downloaded in the PNG file format for visualization.