Transcriptome of Pectobacterium carotovorum subsp. carotovorum PccS1 infected in calla plants in vivo highlights a spatiotemporal expression pattern of genes related to virulence, adaptation, and host response

Abstract Bacterial pathogens from the genus Pectobacterium cause soft rot in various plants, and result in important economic losses worldwide. We understand much about how these pathogens digest their hosts and protect themselves against plant defences, as well as some regulatory networks in these processes. However, the spatiotemporal expression of genome‐wide infection of Pectobacterium remains unclear, although researchers analysed this in some phytopathogens. In the present work, comparing the transcriptome profiles from cellular infection with growth in minimal and rich media, RNA‐Seq analyses revealed that the differentially expressed genes (log2‐fold ratio ≥ 1.0) in the cells of Pectobacterium carotovorum subsp. carotovorum PccS1 recovered at a series of time points after inoculation in the host in vivo covered approximately 50% of genes in the genome. Based on the dynamic expression changes in infection, the significantly differentially expressed genes (log2‐fold ratio ≥ 2.0) were classified into five types, and the main expression pattern of the genes for carbohydrate metabolism underlying the processes of infection was identified. The results are helpful to our understanding of the inducement of host plant and environmental adaption of Pectobacterium. In addition, our results demonstrate that maceration caused by PccS1 is due to the depression of callose deposition in the plant for resistance by the pathogenesis‐related genes and the superlytic ability of pectinolytic enzymes produced in PccS1, rather than the promotion of plant cell death elicited by the T3SS of bacteria as described in previous work.


| INTRODUC TI ON
In recent decades, advances in understanding how pathogenic bacteria respond to host plants and biotic or abiotic environmental factors have been proposed in many important phytopathogenic bacteria (Toth et al., 2003;Charkowski et al., 2012;Mansfield et al., 2012;Ji et al., 2016;Leonard et al., 2017). Previous research revealed the characteristics of conservation and specificity in the gene expression of bacteria interacting with environmental factors. For example, in two phylogenetically distinct Ralstonia solanacearum strains recovered after inoculation in tomato plants, approximately 70% of the common orthologous genes expressed in a similar pattern (Jacobs et al., 2012), while profound differences were found in the global transcriptome of Salmonella enterica after in vitro growth in different media (Blair et al., 2013). In our recent work, the proteomic profiles of Pectobacterium carotovoum revealed the influence of nutrients on gene expression, showing that some proteins were detected in the samples of cells recovered after being inoculated in the living host plant rather than in samples of the cells in medium supplemented with plant extracts (Wang et al., 2018b). P. carotovorum, which causes rot, wilt, and blackleg in many crops and ornamental plants, results in important economic losses worldwide and is one of the top 10 plant pathogenic bacteria based on scientific/economic importance (Ma et al., 2007;Mansfield et al., 2012;Li et al., 2018;Zhao et al., 2018). Bacterial strains from the genera Pectobacterium and Dickeya are classified as soft-rot Enterobacteriaceae (SRE). It is well known that SREs use plant cell wall-degrading enzymes (PCWDEs) as the main pathogenic determinants to successfully infect host plants, and encode all six known protein secretion systems involved in attacking host plants and competing with environmental bacteria (Hugouvieux-Cotte-Pattat et al., 1996;Bell et al., 2004;Mole et al., 2007;Charkowski et al., 2012;Nykyri et al., 2012;Joshi et al., 2016a). SREs secret PCWDEs mainly through a type II secretion system (T2SS) and digest their hosts more extensively than any other microbes. PCWDEs together with additional virulence factors, such as type III effector protein DspE and necrosis inducing protein Nip, are used to macerate plant tissue and promote plant cell death to provide nutrients for the multiplication and colonization of these necrotrophic pathogens in the course of infection (Kim et al., 2011;Charkowski et al., 2012;Haque et al., 2017). Previous research suggested that elicitation of programmed cell death by a type III secretion system (T3SS) in plant leaves can promote virulence in P. carotovorum subsp. carotovorum (Kim et al., 2011;Charkowski et al., 2012). The T3SS deletion strains of SREs and the strain naturally lacking a T3SS can attack potato stems and tubers to a similar extent to those possessing a T3SS . It is not known whether the inducement of plant cell death elicited by the T3SS is crucial to the virulence of SREs. Some regulators (RccR and HexR) controlling gene expression in response to carbon source availability in Pseudomonas fluorescens SBW25 and some genes (pycA, aroBCD, eda) encoding enzymes in the metabolic pathways of carbohydrates are crucial to the virulence of Shigella flexneri, Listeria monocytogenes, and P. carotovorum (Eisenreich et al., 2010;Chavarria et al., 2012;Campilongo et al., 2017;Wang et al., 2018b), but we know little about these in Pectobacterium.
RNA-Seq has been widely used in to help understand pathogen-plant interactions. Based on dynamic expression changes, RNA-Seq approaches have identified genes that function in pathogen infection and adaption processes, such as pathogenicity, metabolism, signalling regulation, and response to complex environmental factors (Rio-Alvarez et al., 2012;Jiang et al., 2015;Ah-Fong et al., 2017).

| Symptoms and bacterial population in the course of Pectobacterium PccS1 infection in planta
To visualize transcriptome analyses at different time points after PccS1 inoculation, calla lily plant symptoms were continuously observed from inoculation of PccS1 onto the petioles of plants until the petioles had nearly fallen ( Figure S1), and the quantity of PccS1 cells in planta was measured as previously described (Jiang et al., 2017). The results revealed that PccS1 multiplied in planta slowly in the early infection stage (before 4 hr after inoculation, HAI). The population was nearly 5 × 10 6 cfu at 12 HAI, and then increased quickly ( Figure S2), similar to previous results that demonstrated that the threshold of bacterial population should be 10 6 -10 7 cfu to effectively elicit expression of the genes participating in a variety of cell density-dependent physiological processes (Toth et al., 2003;Ng and Bassler, 2009;Hawver et al., 2016). More DEGs were identified from the data set when PccS1 in MM was used as the control than when PccS1 in LB was used as the control.

| DEGs in PccS1 recovered from inoculated plants
Between the data sets with different controls, the quantity of differences of DEGs in PccS1 recovered at 4, 8, 12, and 16 HAI were 8.78%, 15.98%, 14.37%, and 11.81% of the total annotated gene number, respectively (Table S1.1). Our results agree with previous research that showed that the growth medium of bacteria is an important consideration during experimental design for any experimental protocol (Blair et al., 2013). For example, a total of 621 genes were found to be differentially expressed when S. enterica SL1344 was grown in MOPS MM compared to growth in LB (Blair et al., 2013).
Based on the stringency of the ratio of log 2 -fold ≥ 2.0 and a threshold FDR of 0.05 or less, the sDEGs in PccS1 infection in planta at different time points were obtained (Tables S1.2 and S1.3), and the numbers of sDEGs in each sample were calculated (Figure 1a). There were 391 and 399 sDEGs present at all four time points compared with the in vitro controls in LB and MM, respectively (Figure 1b-e).
Most of the genes in the T3SS and T6SS clusters, and some genes for pectate lyase and protease (Tables S1.2 and S1.3), were included in the consistently up-regulated sDEGs in PccS1 infection. Some of these genes have been demonstrated to be important for the virulence of SRE (Koo et al., 2012;Bondage et al., 2016;Pédron et al., 2018). In some closely related SRE, genes in the T6SS cluster and for pectate lyase were also found to be up-regulated during infection (Bellieny-Rabelo et al., 2019).

| Validation of DEGs by quantitative reverse transcription PCR
To verify the results of DEGs identified from Illumina sequencing data, a total of 30 genes were selected randomly from the sDEGs for quantitative reverse transcription PCR (RT-qPCR) analysis, as previously described (Allie et al., 2014;Wang et al., 2015).
The gene expression trends of the selected genes in each sample were analysed using RT-qPCR and the results were mainly consistent ( Figure 2), although variations in the exact fold changes were observed between the results of RT-qPCR and RNA-Seq, possibly due to differences in the sensitivity and specificity between these two approaches. The data demonstrate that the RNA-Seq data accurately reflect the response of Pectobacterium PccS1 to the establishment of infection in Zantedeschia odorata plants, as seen in previous work (Allie et al., 2014;Wang et al., 2015;Skorupa et al., 2016). Meanwhile, to evaluate gene F I G U R E 1 The number of significantly differentially expressed genes (ratio of log 2 -fold ≥ 2) (sDEGs) in Pcetobacterium carotovorum subsp. carotovorum PccS1 recovered at four different time points after inoculation in Zantedeschia odorata plants compared with cells in Luria Bertani medium (LB) and minimal medium (MM). (a) Total number of sDEGs in PccS1 recovered compared with the LB and MM controls. (b)-(e) Venn diagrams showing number of sDEGs up-regulated and down-regulated in PccS1 recovered compared with the controls expression, the expression levels of some housekeeping genes, such as recA, gyrB, infB, and rpob, were compared in the samples; the results showed no significant expression changes in each sample of either PccS1 in planta or the in vitro controls, indicating that the RNA-Seq data are comparable with previous work (Shen et al., 2016).

F I G U R E 2
Ratios of the expression levels of 30 genes between Pectobacterium PccS1 recovered and the cultures in Luria Bertani (LB) and minimal medium (MM) using quantitative reverse transcription PCR (RT-qPCR) and RNA-Seq analyses. Bars without standard errors indicate ratios of transcript abundance changes between PccS1 recovered and the controls according to RNA-Seq data (log 2 -fold ratio ≥ 2 for at least one time point in PccS1 recovered versus the controls). Bars with standard errors represent the ratios of gene expression between PccS1 recovered and the controls determined by RT-qPCR

| Functional characterization of the sDEGs in PccS1 recovered after inoculation
To understand the biological significance of the sDEGs in the course of PccS1 infection, we characterized the functions of the proteins encoded by the sDEGs at four time points by mapping the genes to the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (http:// www.genome.jp/keeg/) ( Figure 3). The results reveal that the functional categories of the proteins encoded by the sDEGs were mainly involved in three KEGG pathways (membrane transport, carbohydrate metabolism, and energy metabolism) in PccS1 recovered at the four time points, as compared with the in vitro controls cultured in either LB or MM (Figure 3). The expression of the sDEGs for these categories of protein showed a consistent pattern of regulation in the course of F I G U R E 3 Distribution of significantly differentially expressed genes among the functional categories of the KEGG pathway annotation in Pectobacterium PccS1 recovered compared with the controls from the cultures in Luria Bertani (LB) and minimal medium (MM) infection: most of the sDEGs encoding proteins in membrane transport were up-regulated, those for carbohydrate metabolism were downregulated, and approximately half of those for energy metabolism were up-regulated and the other half were down-regulated ( Figure 3). In some protein categories, more sDEGs were observed in the data set of PccS1 infection compared with growth in MM than for PccS1 infection compared with growth in LB, such as those belonging to carbohydrate metabolism, translation, signal transduction, cell motility, and nucleotide and amino acid metabolism (Figure 3). Previous research showed that genes in amino acid biosynthetic pathways were up-regulated in S. enterica and Escherichia coli grown in MM as compared with the bacteria grown in LB (Tao et al., 1999;Blair et al., 2013). The proteins encoded by the sDEGs with increased expression in PccS1 infection are probably involved in bacterial-plant interactions and bacterial multiplication, representing genes involved in the successful colonization and adaption of PccS1 to different nutrient levels.

| Classification of the Pectobacterium PccS1 sDEGs by the expression pattern in the processes of infection
There were 2,281 sDEGs in the trancriptome pools from PccS1 recovered at the four time points after inoculation versus the reference genes from the cells in either LB or MM. The 2,281 sDEGs were classified into five types to visualize the different types of gene expression change under different nutrient levels (Table 1) (c) rich-nutrient-induced genes were differentially expressed in the bacteria recovered from plants compared to the LB control, but were similar to the MM control; (d) nutrient-level susceptible genes in the recovered bacteria were not only differentially expressed compared with both the control media, but also exhibited different regulation (up and down) in comparison with the different media (MM and LB); and (e) special genes were expressed in the recovered bacteria with different change tendencies at different infection stages compared with the two controls. PccS1_00151 and PccS1_00152 were classified as special genes (Tables 2, S1.2, and S1.3) as their expression levels were linked to the infection stage. Genes in each category were further identified as a subtype of positive or negative by the tendencies of expression change as compared with the controls, except the two special genes (Table 1).
Thirty genes with a log 2 -fold ratio larger than 10 for at least one recovered time point versus the reference genes are listed in Table 2. None of the rich-nutrient negatively induced genes had a log 2 -fold ratio of expression change larger than 10 (Table S1.2 and S1.3).
In previous work, we identified "infection-induced regulation genes" in the transcriptome data of bacterial infection compared with the in vitro control in LB medium (Pédron et al., 2018;Bellieny-Rabelo et al., 2019). We propose the classification of the sDEGs based on the expression changes in PccS1 in planta versus the two controls. If the sample in MM had not been used as a control, the sDEGs that were classified as "basal-nutrient-induced genes" would not have been obtained, highlighting the importance of media choice. For example, pfk (Table 2) did not present as one of the DEGs identified in the wild-type D. dadantii interaction with A. thaliana versus that of the cultures in LB only (Pédron et al., 2018).
In the experiments on PccS1, 682 sDEGs were identified that exhibited differential expression relative to MM (Table 1). If the samples in LB had been used as the sole control, "plant-induced," "rich-nutrient-induced", and "nutrient-level susceptible" genes would have been identified as "infection-induced regulation genes," although they are not the same based on the expression changes in PccS1 infection versus in MM. This example represents an advantage of a gene expression experiment based on the comparison of RNA-Seq data with two controls. By using a two-control gene expression experiment, we obtained DEGs including not only infection-induced genes but also nutrient-adaption genes, although the classification of genes with different expression patterns in PccS1 infection in different controls requires further support from studies on other pathogenic bacteria.  In our recent work, we obtained a transposon library of PccS1 with kanamycin resistance dependent on the host plant in the mimic test (MM supplemented with 0.3% Zantedeschia elliotiana extract) (Jiang et al., 2017), a host plant that is closely related to Z. odorata used in this work. Only one transposon insertion mutant showed attenuated virulence in the host plants; the gene at the insertion site of the mutant was characterized as rplY (Jiang et al., 2017), which exhibited a similar expression pattern in planta and in the mimic test.
However, in this study, the expression pattern of rplY in PccS1 in planta should be classified as "nutrient-level susceptible" (Figure 2 and Table 1). To some extent, this reveals the effects of host nutrients on bacterial adaptation. Similarly, genome comparison has revealed a predicted benzoic acid/salicylic acid carboxyl methyltransferase that relates to life in planta and other specific environmental conditions in Pectobacterium wasabiae SCC3193 (Nykyri et al., 2012).
Three P. wasabiae strains, including SCC3193, collected from potato, constituted a separate clade from the original P. wasabiae strain from Japanese horseradish using multilocus sequence analysis. The separate clade was validated by DNA-DNA hybridization and genome average nucleotide identity, and the three potato strains were transferred to a proposed new species called Pectobacterium parmentieri (Khayi et al., 2016). Our data confirm that global transcriptomes from bacterial cells grown in different conditions are profoundly different, and that choice of medium should be considered carefully during experimental design (Blair et al., 2013). Gene characterization and further study of their functions based on nutrient utilization will enhance our understanding of bacterial adaptation to the host (Anderson and Kendall, 2017).

| Expression of genes related to the main virulence determinants during Pectobacterium PccS1 infection
The PCWDEs, including pectinase, cellulase, and protease, have been demonstrated to be the main virulence determinants in SRE that cause extensive plant tissue maceration (Toth et al., 2006;Wang et al., 2018b). The Pectobacterium PccS1 genome contains 27 pectinase-encoding genes (Table 3) (Table 3).
Three cellulose genes (cel, celV, and a gene for cellulose precursor) and five protease genes (prtC, D, E, F and inh) are positively plant-induced genes, and celB , like rplY, is a nutrient-level susceptible gene (Table 3) In our recent work, we demonstrated that genes crucial to virulence, such as rplY, eda, hfq, and flgK, significantly affected the activities of PCWDEs at both the transcriptional and translational levels in PccS1 (Yang et al., 2012;Jiang et al., 2017;Wang et al., 2018aWang et al., , 2018b.
Our transcriptome data revealed that the longer the time of PccS1 infiltration in the host, the higher the expression level of most of the genes for PCWDEs during the course of infection (

| Expression of the genes related to the regulators of virulence
The regulatory networks of the main virulence determinants in SREs have been examined, focusing on the genes related to acyl-homoserine lactone (AHL) (such as carI and expR) and on those for regulation of pectin catabolism (kdgR, rsmA, rsmB) .
Pathogenic bacteria possess specific transcriptional regulators to sense sophisticated changes (osmotic, acidic, and anaerobic stress, redox potential, etc.) in planta and in natural environments, which in turn affect interaction with plants or microbes nearby Reverchon and Nasser, 2013;Broberg et al., 2014;George et al., 2018). Based on these previous studies, we identified 41 genes encoding regulatory proteins of virulence in the PccS1 genome, and compared their expression levels in planta to the references (Table 4).
The results revealed that the expression patterns of the seven genes related to construction of RNA are different (Box 1 in Table 4): rsmA (PccS1_00073 and PccS1_00521), rsmB (PccS1_00733 and PccS1_02891), rsmC (PccS1_01829 and PccS1_01902), and rpoS (PccS1_00230). It is well known that RsmA is an RNA-binding protein responsible for response to bacterial metabolic status . It negatively regulates PCWDE production and represses tissue maceration (Vakulskas et al., 2015). The expression pattern of rsmA (PccS1_00073) suggests that RsmA might be a switch for cellular processes including infection. The result that the rsmA deletion mutant showed an increase in both virulence (Figure 5d) and PCWDE activities ( Figure S3) supports previous work (Vakulskas et al., 2015). It has been demonstrated that SREs possess the rpoS gene to produce the alternate σ factor, and that rpoS − strains are more sensitive to hydrogen peroxide, carbon starvation, and acidic pH, and produce more PCWDEs (Mukherjee et al., 1998). Similarly, our results revealed that rpoS in PccS1 in planta was down-regulated (Table 4) and the virulence of ΔrpoS on Z. odorata in vitro and in vivo was increased (Figure 5d). This confirms that rpoS negatively TA B L E 3 Log 2 -fold ratios of the genes encoding exo-enzymes in Pectobacterium PccS1 recovered from Zantedeschia odorata at different times after inoculation compared with that of the cells grown in the Luria Bertani medium (LB) and minimal medium (MM)

TA B L E 4 (Continued)
regulates PCWDE production in PccS1, which is validated by the PCWDE activity assays ( Figure S3).  (Table 4). In our recent work, it was found that the expression of kdgR in the PccS1 Δhfq mutant, which had completely lost maceration ability, was the same as the wild type at the transcriptional and translational level (Wang et al., 2018a). Here, we deduced that the comparable expression pattern of kdgR and carI might be because their expression is similar regardless of cellular processes due to their global regulatory function.
Of the four genes for sensing nitrate/nitrite or nitrogen metabolism (narX-narL and ntrC-ntrB), three were differentially expressed based on nutrient levels, with gene expression levels and nutrient levels inversely correlated, while narX was expressed at a similar level under the three different nutrition levels (Table 4). In previous work, the genes for nitrogen metabolism (ntrC-ntrB) were undetected, although narX-narL expressed at a log 2 -fold change of 1-2 (Bellieny-Rabelo et al., 2019). Together, these results emphasize the notion of diversity in nitrogen metabolism in SREs, which might relate to host adaptation.
In this work, we also found that some regulatory genes were slightly differentially expressed in PccS1 infection, including genes in response to acidic pH (phoQ and phoP), transcriptional regulation (marR) and anaerobic regulation (fnf, norR, and arcA). Two genes of undetected expression change (oxyR and ohrR) are linked to hydroperoxide (Table 4) Together with the expression changes in PCWDEs, these data explore an expression programme between the key and accessory virulence determinants in SREs.

| Expression pattern of genes encoding nucleotide-related proteins during Pectobacterium PccS1 infection
Some nucleotide-related proteins have been revealed to have regulatory function in virulence Reverchon and Nasser, 2013;Kusmierek and Dersch, 2018 (Reverchon and Nasser, 2013). Our data reflect this programme with gene expression changes (Tables S1.2 and S1.3) and show that the expressions of crp, hexR, and hexA were all negatively regulated in planta versus the references (Box 4 in Table 4). CRP, the cyclic AMP receptor protein, has been identified as the regulator of the pectinolysis gene in D. dadantii (Nasser et al., 1997), and as a transcriptional master regulator of numerous noncoding RNAs in the regulatory architecture linking nutritional status to virulence in Yersinia pseduotuberculosis (Nuss et al., 2015). Transcriptional factor HexR has been characterized as a global regulator of the central carbohydrate metabolism genes in various groups of proteobacteria (Leyn et al., 2011). In our recent work, it was demonstrated that the expression of hexA was significantly increased at the transcriptional and translational level in the absence of hfq, which revealed the negative role of hexA in regulating the virulence of PccS1 (Wang et al., 2018a), similar to previous reports (Mukherjee et al., 2000;Tobias et al., 2017) (Table 4).
The 33 genes in the genome having amino acid motifs associated with cyclic nucleotide metabolism and binding are listed in Table S2.
They encode proteins in categories, including catalytic and regulatory domains of diguanylate phosphodiesterases, diguanylate cyclases, and cyclic di-GMP regulator. One third of these genes were down-regulated in the course of infection, and only three genes were temporarily up-regulated (two at the initial stage and one at successful infection stage) compared with the controls. Our data support previous research that showed that constitutively elevated c-di-GMP levels are detrimental for acute infections in many animal bacterial pathogens, such as Vibrio cholerae and Brucella melitensis (Romling et al., 2013), although the molecular regulatory mechanisms underlying c-di-GMP are awaiting further study in Pectobacterium.

| Expression pattern of the genes encoding the components for the secretion systems related to pathogenicity
It is well known that several secretory systems in bacterial pathogens secrete enzymes or toxins into host cells or the surrounding environment (Izore et al., 2011;Douzi et al., 2012;Koo et al., 2012;Bondage et al., 2016). Genes annotated in the genome of PccS1 were checked for components of the secretory systems. Both in the cluster and in the solitary loci outside the cluster, there were 25 genes encoding for the T2SS, 36 genes for the T3SS, and 29 genes for the T6SS, as well as the gene for the T3SS effector (dspE), the related genes (dspF), and a gene for the chaperone of HrpW. The gene expression pattern showed that the clusters of both T3SS and T6SS were all significantly up-regulated in planta compared with the ref- Tables S3 and S4), as were most of the genes outside the T6SS cluster (Table S4) (Table S5), again similar to that in D. dadantii (Pédron et al., 2018). The T2SS is known as the out system, and is responsible for secreting PCWDEs in most SRE pathogens .

erences (boxes in
Based on the characterization of T2SS assembly and the model of exoprotein delivery (Douzi et al., 2012), the expression pattern of the T2SS genes in PccS1 recovered after inoculation indicated that the genes for the Tat export pathway (tadB, C) and motility related to the pilus assembly crossing the inner membrane (tadA and rcpA) were more susceptible to nutrient conditions than those in the cluster encoding Gsp proteins for assembling the secretion tunnel. We deduced that the T2SS secretion tunnel might be assembled for secretion at any cellular processes regardless of growth conditions in Pectobacterium PccS1. The genes with functions for pathogenicity, including those for the Tat export pathway and pili for motility in T2SS, are activated only in the presence of host plants.

| Expression pattern of the genes for the enzymes in carbon metabolic pathways
It is well known that pectinolysis is carried out by several pectinolytic enzymes in SRE (Hugouvieux-Cotte-Pattat et al., 1996;Joshi et al., 2016b). Glucose is produced in plants and converted into sucrose transported in a sieve tube. Sucrose/glucose could be utilized as nutrients when PccS1 was inoculated into the petioles of Z. odorata plants through a wound (Truesdell et al., 1991).
We summarize the expression pattern of the genes involved in the pathways of pectinolysis and carbohydrate metabolism in Figure 4, and the genes for the PCWDEs differentially expressed at a log 2fold ratio over 2 versus the references are discussed in Table 3. It is notable that expression of sacA, pfk, and rpiB were significantly activated in the presence of sucrose or glucose, similar to the activated PCWDEs in infection, although they belong to different types of gene expression. Based on the spatiotemporal expression pattern of the genes for the enzymes in the tricarboxylic acid cycle, fermentation, and the glyoxylate cycle, presented in Figure 4, it is suggested that the metabolic pathways from sucrose/glucose and pectin to acetyl-CoA via pyruvate are the main carbon metabolism pathways in PccS1 during infection, and these pathways might be subsequently pushed by the interconversion between succinate and fumarate ( Figure 4). Meanwhile, the expression of the genes for the enzymes in other pathways (pyruvate fermentation to produce lactate, acetate, and formate, and conversion to phosphoenolpyruvate) rarely changed in any the infection samples compared to the controls (Figure 4). Previous research has indicated that the limitation of acid-production from pyruvate is helpful to activate PCWDE production (Reverchon and Nasser, 2013). Thus, we can deduce that the substrates taken from the host by PccS1 are mainly metabolized to produce more energy and suitable to the establishment of infection through PCWDE activation.

| Virulence of the strains with a mutation in one of the sDEGs
To further understand the functions of the sDEG in the course of infection on Pectobacterium PccS1 virulence, 33 sDEGs were selected for single gene deletion, and the virulence of the mutants was determined in Brassica rapa subsp. pekinensis and Z. odorata plants. The results indicated that the effects of the sDEGs on PccS1 virulence are different (Figure 5a,d).
Five strains with a mutation in a gene for T6SS effectors, regulation, and structure (vgrGs, hcps, impL, clpB, and impJ) presented a similar level of maceration on the hosts as the wild type ( Figure 5a).
These results are different to those for some plant and animal pathogens that use T6SS as a weapon to interact with the hosts, such as Acidovorax citrulli AAC00-1 (Tian et al., 2015), Erwinia amylovora NCPPB1665 (Tian et al., 2017), Pantoea ananatis LMG 2665 T (Shyntum et al., 2015), Pectobacterium atrosepticum SCRI1043 (Bell et al., 2004;Liu et al., 2008), Pseudomonas aeruginosa PAO1 (Hachani et al., 2016), Vibrio cholera V52 (Miyata et al., 2011), and Burkholderia pseudomallei E8 (Hopf et al., 2014). Our results agree with those from E. amylovora CFBP1430 (Kamber et al., 2017) and P. atrosepticum SCRI1043 (Mattinen et al., 2007), which showed no differences in lesion development and nonhost hypersensitive response elicited between the mutants and the wild type (Figure 5a,b). It is well known that VgrG and Hcp act as both effector and structural protein in the T6SS in pathogenic bacteria (Basler et al., 2012;Ho et al., 2014). Significantly activated expression of these genes for the components of T6SS in Pectobacterium PccS1 in planta (Table S4) revealed that these genes in the T6SS in PccS1 might participate in pathogenicity indirectly. How the T6SS in PccS1 affects virulence requires further work to scan each T6SS component.
The strains with a mutation in one of the T3SS structural and regulatory genes (hrpN, hrpA, hrpI, hrpQ, and hrcN) showed no differences in lesion development compared with the wild-type PccS1 (Figure 5a), although they could not elicit a nonhost plant hypersensitive response (Figure 5c). This is different to the previous reports that P. carotovorum WPP14 used T3SS to induce plant cell death to promote leaf maceration in Nicotiana benthamiana and Erwinia chrysanthemi in African violet varieties (Yang et al., 2002;Kim et al., 2011). The results in Figure 5 and Table S3 show that Pectobacterium PccS1 uses T3SS to induce plant cell death rather than to macerate plant tissue directly. This strongly supports the statement that the strains naturally lacking a T3SS use other genes to compensate during attack of potato stems or tubers (Charkowski et al., 2012).
The increased virulence of the strains with a mutation in rsmA (PccS1_00073) or rpoS (Figure 5d) confirms the function of the negative regulators, similar to previous work (Vakulskas et al., 2015), and also agrees with the significantly depressed expression of these genes observed in the transcriptome profiles of PccS1 in planta versus the in vitro controls (Table 4).
Interestingly, the strain with a mutation in sDEG PccS1_03557, a tor, the two phenotypes were unchanged (Figure 6a,b,c). Meanwhile, these mutants caused maceration in N. benthamiana leaves in vitro the same as the wild-type 12 HAI (Figure 6d). The results revealed that PccS1_03557 did not influence growth in LB medium ( Figure   S4). Callose rarely deposited in the in vitro N. benthamiana leaves infiltrated with ΔPccS1_03557 and ΔPccS1_03557 (pBBR) showed evidence that the leaves lost the ability of resistance to the invasion of the bacterial strains, but more callose was deposited in the leaves in vivo after inoculation of these two mutants (Figure 7). Our results indicated that PccS1_03557 participates in depressing the plant's ability to deposit callose for host resistance, and the deposition of F I G U R E 4 Spatiotemporal expression pattern of the genes encoding the enzymes in carbon metabolic pathways in Pectobacterium PccS1 recovered after inoculation at four time points versus that of the cells grown in Luria Bertani medium (LB) and minimal medium (MM) callose for host resistance is a systematic process in the plant. This reveals the importance of using living plants, rather than mimics supplemented with plant extracts, in approaches for bacteria-plant interaction studies, and this was also demonstrated in our previous work (Wang et al., 2018b).
The results of RT-qPCR analysis showed that the expression of T3SS structural and regulatory genes (hrpN, hrpA, hrpI, hrpQ, and hrcN) was down-regulated when PccS1_03557 was impaired in PccS1 ( Figure 8). We have shown that the strains with a mutation in one of these genes could macerate the host at the wild-type level and lose the ability of elicitation hypersensitive response in the leaves of the nonhost (Figure 5a,

| Assays of bacterial virulence and population dynamics in the host plants, and hypersensitive response in the nonhost plants
Virulence assays were performed as previously described (Jiang et al., 2017)  Bacterial population dynamics in the host after inoculation were assayed as previously described (Andrade et al., 2008;Wang et al., 2018b). Hypersensitive response was assayed by infiltrating bacterial strains into 6-week-old leaves of greenhouse-grown tobacco plants of the two varieties (N. tabacum 'Samsun' and N. benthamiana) or the detached leaves of N. benthamiana using a needleless syringe as in previous work (Fan et al., 2011;Guo et al., 2012).

| Recovery of bacterial cells from the inoculated-plant and the media
Pectobacterium PccS1 were recovered using centrifugation from the petiole segments detached from the calla plants after inoculation at different time points, as previously described (Jacobs et al., 2012;Meng et al., 2015).  Sequencing was carried out using a 2 × 125 paired-end configuration. In order to obtain clean reads, raw reads were first filtered using statistical software Trimmomatic v. 0.30. Adapter sequences and low-quality sequences were removed. Low-quality sequences included reads of base number less than 75, reads with N percentage (the percentage of the nucleotides that could not be sequenced in the read) over 5%, and those with Q-value of both the 5′ and 3′ ends lower than 20 (Q-value of 20 means the percentages of the incorrect sequenced bases in the reads were lower than 1%). The sequences were then re-evaluated using software FastQC v. 0.10.1 and the clean reads were assembled.

| Identification of DEGs
The reads were mapped back to the transcriptome of PccS1 using the alignment software bowtie2 v. 2.1.0. The number of mapped clean reads for each unigene was then counted and normalized into a fragment per kilobases per million reads (FPKM) value (Mortazavi et al., 2008), which was calculated by RSEM v. 1.2.4 software. Data that passed quality controls were analysed using Bioconductor edgeR software. DEGs were identified using statistical analysis among the libraries as described previously (Shen et al., 2012). The p value threshold in multiple tests was determined by the false discovery rate (FDR) (Benjamini et al., 2001;Liu et al., 2015). In the present work, differentially expressed unigenes between the samples of cells recovered from living plants and cultured in LB or MM were screened with a threshold of FDR ≤ 0.05 and an absolute value of log 2 -fold ≥ 1 as in previous studies Wang et al., 2015;Hu et al., 2016). GO and KEGG pathway enrichments were compared between up-regulated and down-regulated unigenes. All the unigenes were identified by BLAST comparison with the complete genome of Pectobacterium PccS1.

| RT-qPCR
To confirm the DEGs identified by RNA-Seq in the course of PccS1 infection, RT-qPCR assays were performed with bacterial cells collected at the time points in the processes of infection as previously described (Kersey et al., 2012;Wang et al., 2018b). The gene-specific primers are listed in Table S8.

| Gene knockout and complementation
Gene knockout mutants were constructed and complemented as previously described (Wang et al., 2018b). Plasmids pEX18Gm carried the 300-600 bp fragment cloned from upstream or downstream of the target genes using the relevant primers (Table S6). To construct complementation strains, the target gene and its promoter region were amplified and cloned into pBBR1-MCS5 (Table S7). All constructs were verified by PCR and sequencing.

| Callose deposition assay
The assay of callose deposition in the leaves of N. benthamiana and N. tabacum 'Samsun' was performed as previously described (Kim et al., 2011;Wang et al., 2018a). P. syringae pv. tomato DC3000 and sterilized water were used as controls.

| Statistical analysis
Each assay described above was repeated at least three times with three to five replicates in each.
The hypothesis test of percentages (Duncan's multiple range test, α = 0.05 or 0.01) was used to determine significant differences in the assays described above. Institutions (PPZY2015B157). We acknowledge the reviewers whose comments and suggestions helped us improve the manuscript.

CO N FLI C T O F I NTE R E S T
The authors have declared no conflicts of interest.
F I G U R E 8 Bars represent relative mRNA levels of the genes from the T3SS in the wild-type PccS1 and mutant Δ03557. The mRNA levels of these genes in PccS1 were set to 1 to calculate the relative expression ratio. The housekeeping gene recA was used as an endogenous control for assessing expression. The experiments were repeated three times with at least three internal replicates in each. The data are shown as averages ± SD. *p < .05, **p < .01 versus the wild type, Duncan's multiple range test

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information may be found online in the Supporting Information section.