For preparing culture filtrate (CF), the mycelia and sporangia of a P. infestans isolate 88069 were cultured in the Plich liquid medium (0.5 g KH₂PO₄, 0.25 g MgSO₄·7H₂O, 1 g Asparagine, 1 mg Thiamine, 0.5 g Yeast extract, 10 mg β-sitosterol and 25 g Glucose in 1l) for four weeks at 16°C in dark. Then the liquid was used for plant treatment.

N. benthamiana plants were grown from seeds and maintained in the controlled climate chamber under a 16h/8 h light and dark cycle at 24°C. The fully expanded leaves of 4-5-week-old plants were used for CF infiltration with a needleless syringe. And samples were taken at 0 h (immediately after CF infiltration), 8 h, 12 h, 24 h, and 48 h, respectively. Three infiltrated leaves were taken at each time point as three biological replicates.

Approximately 1.5 g of N. benthamiana leaves for each sample was ground in a mortar with liquid nitrogen. The tissues were homogenized by 10 ml of SDS Lysis Buffer (60 g sucrose, 4 g SDS and 3.084 g DTT in 200 ml) containing protease inhibitor and centrifuged at 20 000 rpm at 4°C for 10 min. The supernatant was taken to a freshly pre-cooled 50 ml centrifuge tube and mixed with an equal volume of Tris saturated phenol (pH 8.0). Afterward, the mixture was incubated at 37°C on a shaker for 5 min. After centrifugation at 20 000 rpm at 4°C for 10 min, the mixture was divided into three layers, with protein in the upper layer. It was then carefully taken out and mixed with 5× volumes of pre-cooled 0.1 M ammonium acetate in methanol for overnight protein precipitation at -20°C. After centrifuged at 12 000 rpm for 10 min at 4°C, the supernatants were removed, and the pellets were washed with 10 ml of pre-cooled methanol, and then further washed with 10 ml of pre-cooled acetone. Finally, the protein pellets were dissolved in the 4 ml urea buffer (8 M urea in 0.05 M ammonium bicarbonate) with ultrasonic for 6 min (ultrasonic for 2 s, stop for 3 s). After centrifugation at 20 000 rpm for 10 min at 4°C, the supernatant was carefully aspirated into a freshly pre-cooled 15 ml centrifuge tube for subsequent protein quantification. Protein concentration was determined using the Pierce™ bicinchoninic acid (BCA) Protein Assay Kit (Thermo Scientific, USA) with bovine serum albumin (BSA; 2mg/ml) as a standard for quantification.

Then each protein sample (5 mg) was reduced by 20 μl of 1 M dithiothreitol (DTT) at 56°C for 30 min and alkylated by 120 μl of 0.5 M iodoacetamide at room temperature in dark for 40 min. Then 40 μl of 1 M DTT was added into the protein materials in dark for 20 min to quench the extra iodoacetamide. Afterward, each sample was diluted into 8× volumes of 50 mM ammonium bicarbonate solution including 1 M urea and 400 μl of trypsin (enzyme/protein, 1:25 w/w) for protein digestion overnight at 37°C. Peptides were generated by specific trypsin cleavage at the C-terminal at the arginine and lysine sites of proteins. The acquired peptides were acidified by 10% trifluoroacetic acid (TCA) and desalted using oasis^® HLB solid-phase extraction (SPE; Agilent, USA). Finally, the desalted peptide samples were dried in a speed-vacuum concentrator and stored at -80°C or directly dissolved in 0.1% formic acid for LC-MS/MS analysis.

SCX peptide fractionation was performed according to previously reported (Lu et al., 2019). Analysis of SCX peptide fractions was performed on a Nanospray III source and a TripleTOF 5600+ (AB SCIEX) mass spectrometer (Andrews et al., 2011). The peptide fractions (2μg each) were resuspended in 0.1% formic acid (FA) and trapped separately into an Ekspert TM nanoLC 415 (AB SCIEX) with a nanoLC trap (ChromXP C18-CL 3 μm 120 Å, 350μm × 0.5 mm) at 2 μL/min for 7 min. Then, the pump flow was split to obtain a flow rate of 300 nL/min on the nanoLC trap and the nanoLC column (75 μm × 15 cm, 3C18-CL-120, 3 μm, 120 Å) was used for separation. The mobile phases consisted of solvent A containing water: acetonitrile: formic acid at a ratio of 98:2:0.1 (V/V/V), and solvent B containing water: acetonitrile: formic acid at a ratio of 2:98:0.1 (V/V/V). A nonlinear gradient of 5% to 12% B for 0.5 min, 12% to 24% B for 59.5 min, 24% to 35% B for 35 min, 35% to 50% B for10 min, 50% to 80% B for 1 min, 80% B for 5min, 80% to 5% B in 1 min, and 5% B for 8 min was employed. The mass tolerance was set to 50mDa, The spray voltage was set to 2.5 kV, and the temperature of the heated capillary was set to 275°C. The mass spectrometer was operated in positive ion mode with a nanoion spray voltage of 2.5 kV.

Data-dependent acquisition (DDA) was first performed to obtain the SWATH-MS spectral ion library. A survey scan of 0.248 s in the range 350−1250 m/z was performed to collect the MS1 spectra. The top 50 precursor ions with charge state from +2 to +5 were selected for fragmentation with an accumulation time of 50 ms per MS/MS experiment for a total cycle time of 2.3 s, and MS/MS spectra were collected in the range 100−1500 m/z. Selected ions and their isotopes were dynamically excluded from further MS/MS fragmentation for 18 s.

In the SWATH analysis, the same peptide samples were analyzed by the cyclic data-independent acquisition (DIA) mode using similar methods as described above. For DIA analysis, MS/MS proteome profiling was analyzed by the same LC-MS/MS system. A survey scan of 50 ms was performed and all precursors were subjected to fragmentation. MS/MS experiments were conducted with an accumulation time of 100 ms per 25 Da swath (total swath 36) for a total cycle time of 3.6 s.

All DDA mass spectrometry data were used to generate a reference spectral library by searching against the Nicotiana benthamiana protein database v04.4 (36151 proteins, 2010 release, Sol Genomics Network) using the ProteinPilot 4.5 software (Sciex). The parameters included digestion, alkylation, and biological modification. A false discovery rate of < 1% was accepted as the criteria for the peptide assignments and protein identification. Subsequently, the DIA data and spectral library were loaded into PeakView v.1.2 software (Sciex) under restricted criteria and settings: six peptides, six transitions, 99% peptide confidence and ion library mass tolerance (75 ppm). The output of quantified proteins and corresponding peptides by the MarkerView (Sciex) generated three quantitative files including the extracted peak area for individual fragment ions, the sum of the fragment ion areas for each peptide and the sum of peptide areas for each protein. Differentially expressed proteins (DEPs) were analyzed by ANOVA analysis and Tukey’s HSD multiple comparison (P ≤ 0.05) and then were updated from reference genome Niben v0.44 to Niben v1.01. Information of the identified DEPs were shown in Table S1 .

TBtools (Chen et al., 2020) was used to perform gene ontology (GO) enrichment for DEPs of cluster 7. GO background data was downloaded from the AgriGO website. E-value cut-off: 1.0E-3, and GO terms of Biological Process with a false discovery rate (FDR) 0.05 were obtained and included in Table S2 . They were ranked according to their P-values (ascending), and the top 15 terms for Biological Process belonging to level 3 were then selected to draw a bubble diagram by R package ggplot2 (https://github.com/tidyverse/ggplot2).

Virus-induced gene silencing (VIGS) constructs were made by cloning ∼300 bp PCR fragments from candidate genes into the TRV2 vector, respectively (Ratcliff et al., 2001). The inserted fragments were selected with the help of a VIGS tool in the genome website of Solanaceae (https://vigs.solgenomics.net/). Primers for VIGS of candidate genes shown in Table S3 . A TRV2 construct expressing GFP was used as a control (Du et al., 2017). Agrobacterium tumefaciens strains GV3101 containing TRV1 and each candidate gene VIGS construct at an OD600 of 0.3 were mixed at a 1:1 ratio and then infiltrated into the leaves of the four-leaf-stage N. benthamiana plant. Systemic leaves were detached and analyzed by qRT-PCR.

Total RNA was extracted from the leaves of N. benthamiana plants after infiltration of CF and VIGS using a Total RNApure Kit (Zoman Biotechnology Co. Ltd.) according to the manufacturer’s instructions. First-strand cDNA was synthesized using the 5X All-in-One RT MasterMix Reverse Transcription Kit (ABM). qRT-PCR was carried out using SYBR green as described previously (Du et al., 2017). Primers for qRT-PCR are shown in Table S3 and Actin was used as an internal control. For VIGS experiment, at least one primer was designed outside the region of cDNA targeted for silencing.

Luminol-based assays were used to evaluate ROS bursts. After three weeks of VIGS, at least nine leaves of N. benthamiana plants for each candidate gene were excised into leaf discs of 0.09 cm². To eliminate the wounding effect, leaf discs were incubated in the 96-well plate with 200 µl of ddH₂O overnight, which was replaced by 100 µl of reaction solution as previously described with a difference in the concentration of flg22 (Sang and Macho, 2017). Instead of 5 μl flg22 stock solution (100 μM) added to 10 ml of ultra-pure distilled H₂O, 10 μl flg22 stock solution (100 μM) was added in this study. Measurements were taken immediately after adding the solution with a 1 min interval reading time for a period of 60 min with a luminometer. The measurement values for ROS production from 24 leaf discs per treatment were indicated as means of RLU (Relative Light Units).

To test the PAMP-triggered immune responses by P. infestans culture filtrate (CF), it as well as a negative control culture medium (CM) were infiltrated into the same N. benthamiana leaves, respectively. Then the infiltrated leaves were observed and sampled at 0 (immediately after CF infiltration), 8, 12, 24 and 48 hpi. Under natural light, no obvious phenotypes were observed at 0, 8, 12 and 24 hpi, while at 48 hpi, a cell death was visible at the CF-infiltrated site but not at the CM-infiltrated site ( Figure 1A ). Interestingly, under UV light, little cell death responses could be observed as early as at 24 hpi at the CF-infiltrated site but not at the CM-infiltrated site ( Figure 1A ). Then samples taken at 0, 8, 12, 24 and 48 hpi were analyzed for the expression of three PTI marker genes WRKY8, PTI5 and ACRE31 (Nguyen et al., 2010; Wang et al., 2018) by qPCR. The expression of all three genes was significantly upregulated at 8 h after CF infiltration, however, their expression changes were diverse during 48 hpi ( Figure 1B ). After 8 hpi, the expression of WRKY8 kept steady until 48 hpi, while the expression of PTI5 and ACRE31 peaked at 12 and 24 hpi, respectively, and then fell down ( Figure 1B ).

In order to identify novel differentially expressed proteins (DEPs) induced by CF in N. benthamiana, only CF-infiltrated leaves taken at 0, 8, 12, 24 and 48 h were sent for the subsequent SWATH-MS analysis ( Figure 2 ). For each time point, three infiltrated leaves were taken for independent tests and each sample was also tested three times for SWATH-MS as technical repeats. As a result, nine data sets were obtained for each time point. With data from all the five time points, a reference database containing a total of 38,683 unique peptides corresponding to 5714 proteins was generated using the data-dependent acquisition (DDA) model, with a false discovery rate (FDR) less than 1%. Compared with our reference database, a total of 4401 proteins were identified at all five time points with the Process in SWATH 2.0 software analysis based on important information including retention time, precursor ion mass, and fragment ion mass-to-charge ratio.

Then principal component analysis (PCA) was performed to evaluate the data by R package factominer. The idensity of all 4401 proteins identified at five time points were taken as observation values, and samples at different time points and their biological replicates were taken as variables. The PCA biplot shows that they were distinctly separate in two principal components ( Figure 3A ), which means that proteins has experienced a significant change in N. benthamiana after CF induction. Among the above identified 4401 proteins, we compared the amount of identified proteins at the later four time points with that of 0 hpi. A total of 1429 DEPs were identified at a fold change of 1.5 with a p-value <0.05 ( Table S1 ). Specifically, 331 DEPs were identified with 129 up-regulated and 202 down-regulated at 8 hpi, 837 DEPs were identified with 343 up-regulated and 494 down-regulated at 12 hpi, 944 DEPs were identified with 515 up-regulated and 429 down-regulated at 24 hpi, 1193 DEPs were identified with 731 up-regulated and 462 down-regulated at 48 hpi ( Figure 3B ).

We used Mfuzz to cluster all 1429 DEPs according to their protein expression profiles and obtained mainly nine types of temporal patterns, which generally show four expression trends ( Figure 4A ). The first trend is overall down-regulated shown in clusters 1, 2, 3 and 8. The second trend is overall upregulated shown in clusters 4, 6 and 7. The third trend is first down-regulated then up-regulated shown only in cluster 5. The fourth trend is first up-regulated then down-regulated shown only in cluster 9. Among the upregulated clusters, the trend of cluster 7 maintains the best continuity, which may be more correlated to PTI defenses. To further analyze the function of DEPs in cluster 7, we performed GO enrichment analysis of them ( Figure 4B ). Among the top 15 GO terms for Biological Process belonging to level 3, defense-related terms were found such as response to stress, immune response, response to biotic stimulus, etc. Interestingly, some well-known PTI-related proteins such as SGT1 and BAK1 were also found in cluster 7. This indicates that more PTI-related proteins may be involved in cluster 7 since they have similar expression patterns.

To test the above hypothesis, we selected 19 DEPs belonging to “response to stress” for functional verifications by VIGS and ROS assays, including mitochondrial phosphate carrier protein (MPT) 3, protein disulfide-isomerase like (PDIL) 2, peroxisomal (S)-2-hydroxy-acid oxidase (GOX) 3, cysteine-rich receptor-like protein kinase (CRK) 2, thioredoxin reductase (NTRA) 2, heat shock 70 kDa protein BIP2 (HSP70-12), chaperone protein dnaJ (DJA) 2, calnexin homolog (CNX) 1, lysophospholipase (LysoPL) 2, PLAT domain-containing protein (PLAT) 1, ascorbate peroxidase (APX) 1, 2-hydroxyacyl-CoA lyase (HPCL), ABC transporter G family member (ABCG) 40, acidic endochitinase (CHIA), early responsive to dehydration (ERD) 9, peptidyl-prolyl cis-trans isomerase FKBP (FKBP) 15-1, vesicle-associated membrane protein (VAMP) 714, heat shock 70 kDa protein (HSP) 2, nucleosome assembly protein (NAP) 1;3 ( Table 1 ).

For VIGS, ∼300 bp PCR fragments from 19 candidate genes were amplified using the cDNA of N. benthamiana as a template. Leaves of four-leaf-stage plants were treated with TRV-candidate genes and the control TRV-GFP, respectively. After three weeks, three leaves from different plants of N. benthamiana for each candidate gene were excised, and flg22 was used to determine the ROS intensity. Compared to the control TRV-GFP, we found that the ROS production was significantly changed after plants were treated with six TRV-candidate genes. After VIGS of a candidate gene MPT3, the ROS production was significantly up-regulated, while after VIGS of five other candidate genes including VAMP714, LysoPL2, APX1, HSP70-2 and FKBP15-1, the ROS production was significantly down-regulated ( Figure 5 ). Their silencing efficiency was tested by qRT-PCR with the transcript level of candidate genes reduced below 50% ( Figure 5 ).