Genomic and Functional Dissections of Dickeya zeae Shed Light on the Role of Type III Secretion System and Cell Wall-Degrading Enzymes to Host Range and Virulence

ABSTRACT Dickeya zeae is a worldwide destructive pathogen that causes soft rot diseases on various hosts such as rice, maize, banana, and potato. The strain JZL7 we recently isolated from clivia represents the first monocot-specific D. zeae and also has reduced pathogenicity compared to that of other D. zeae strains (e.g., EC1 and MS2). To elucidate the molecular mechanisms underlying its more restricted host range and weakened pathogenicity, we sequenced the complete genome of JZL7 and performed comparative genomic and functional analyses of JZL7 and other D. zeae strains. We found that, while having the largest genome among D. zeae strains, JZL7 lost almost the entire type III secretion system (T3SS), which is a key component of the virulence suite of many bacterial pathogens. Importantly, the deletion of T3SS in MS2 substantially diminished the expression of most type III secreted effectors (T3SEs) and MS2’s pathogenicity on both dicots and monocots. Moreover, although JZL7 and MS2 share almost the same repertoire of cell wall-degrading enzymes (CWDEs), we found broad reduction in the production of CWDEs and expression levels of CWDE genes in JZL7. The lower expression of CWDEs, pectin lyases in particular, would probably make it difficult for JZL7 to break down the cell wall of dicots, which is rich in pectin. Together, our results suggest that the loss of T3SS and reduced CWDE activity together might have contributed to the host specificity and virulence of JZL7. Our findings also shed light on the pathogenic mechanism of Dickeya and other soft rot Pectobacteriaceae species in general. IMPORTANCE Dickeya zeae is an important, aggressive bacterial phytopathogen that can cause severe diseases in many crops and ornamental plants, thus leading to substantial economic losses. Strains from different sources showed significant diversity in their natural hosts, suggesting complicated evolution history and pathogenic mechanisms. However, molecular mechanisms that cause the differences in the host range of D. zeae strains remain poorly understood. This study carried out genomic and functional dissections of JZL7, a D. zeae strain with restricted host range, and revealed type III secretion system (T3SS) and cell wall-degrading enzymes (CWDEs) as two major factors contributing to the host range and virulence of D. zeae, which will provide a valuable reference for the exploration of pathogenic mechanisms in other bacteria and present new insights for the control of bacterial soft rot diseases on crops.

Although isolated mainly from monocotyledonous hosts in nature, D. zeae is so far known to infect 24 types of dicots and 23 types of monocots, indicating a broad host range (9). In recent years, D. zeae has caused considerable losses in crop yields in China, especially on rice and banana (13)(14)(15)(16). At the same time, D. zeae strains from diverse sources also display considerable variations in their pathogenesis. For instance, the strains EC1 (isolated from rice) and MS2/MS3 (isolated from banana) showed different virulence on various monocot and dicots plants and exhibited 2-fold difference in their ability to inhibit rice seed germination (9,17). Furthermore, consistent with these phenotypic variations, there are also key genomic differences between EC1 and MS2/ MS3, including the zms gene cluster encoding the phytotoxic zeamines, one of the most important virulence factors unique to and shared by rice strains isolated from geographically distant regions (9,18,19), the biosynthetic gene cluster (C1O30_ RS04995 to C1O30_RS05185) encoding a novel phytotoxin in MS2, and the pipR and pipA that are present only in MS1 and MS2 (17).
Importantly, we have recently isolated from the ornamental plant Clivia miniata three D. zeae strains which are, for the first time, found to specifically infect monocots (9). These strains were pathogenic on nearly all tested monocots (e.g., rice, banana, clivia), although their virulence was usually lower than that of other D. zeae strains, and could not infect any of the nine dicots tested (e.g., potato, tomato, radish, cabbage) (9). A distinctive characteristic of these clivia-isolated strains is the remarkably lower activity of cell wall-degrading enzymes (CWDEs) compared with that of other D. zeae (9). However, the molecular basis of the dramatic alteration in host range remains largely unknown.
To investigate determinants of the host range of D. zeae, we sequenced the genome of JZL7, one of the monocot-specific strains isolated from C. miniata, and performed comparative genomic as well as functional analyses of JZL7 and other D. zeae strains that infect both monocots and dicots. The findings of this study will facilitate our understanding of the nature and the evolution of the pathogenicity mechanisms of D. zeae.

RESULTS AND DISCUSSION
General genomic features of D. zeae strain JZL7. The genome of D. zeae strain JZL7 was sequenced on both the PacBio RSII and the Illumina HiSeq X Ten platforms, producing 196,305 long sequencing reads (N 50 : 8,071 bp; mean read length: 5.356 Kb; total size: 1.051 Gb) and 5.44 million short sequencing reads (PE150; total size: 1.789 Gb), respectively. A hybrid de novo assembly using both the short and long sequencing reads resulted in a high-quality and gapless assembly consisting of a single circular chromosome of 4,925,859 bp in size; no plasmids were detected (Fig. 1). BUSCO evaluation of the JZL7 genome detected 438 of the 440 genes (99.5%) in the enterobacterales_odb10 data set, suggesting that the assembly is highly complete. The JZL7 genome contains 4,268 protein-coding genes, 75 tRNA genes, and 22 rRNA genes which belong to seven complete rRNA loci, including the unusual 16S-23S-5S-5S operon previously observed in other Dickeya genomes (Table 1) (18). JZL7 has both the largest genome and the largest repertoire of protein-coding genes among all sequenced D. zeae strains; it contains 167 more genes than Ech586, which has the second largest genome, and 443 more genes than EC1, which has the smallest genome (Table 1). On the other hand, the G1C content of JZL7 genome (53.68%) is highly similar to that of the other D. zeae genomes (53.34% to 53.65%) ( Table 1).
Genomic differences between JZL7 and other D. zeae strains. To determine the evolution relationship between JZL7 and other D. zeae strains, all sequenced D. zeae and the closely related D. oryzae ZYY5 (8) genomes were used for phylogenetic analysis based on the 3,110 single-copy core genes (present in all strains). Results showed that the 15 D. zeae strains can be divided into two clusters (Fig. 2); the average nucleotide identity (ANI) values are above 95% for all pairwise comparisons within each cluster but below 95% for nearly all comparisons between clusters (Fig. 2). The first cluster consists of seven strains, including DZ2Q, EC1, EC2, ZJU1202, and ZYY5, the five ricepathogenic strains, as well as CSL_RW192 and NCPPB_3531. Our results thus corroborate the recent proposal by Wang et al. (8) to classify this clade as a novel species, Dickeya oryzae.
The second clade contains JZL7, CE1, and Ech586, as well as five other strains that are highly similar to each other (ANI values of .97.85%), namely, MS1, MS2, MK19, NCPPB_2538, and NCPPB_3532 (Fig. 2). We have recently shown that EC1 (first clade) and MS2 (second clade) are pathogenic toward both monocot and dicot hosts, whereas JZL7 (second clade) can infect only monocots (9). Taken together, our results suggest that the broad host range may have been the ancestral state to both D. zeae  and D. oryzae, while JZL7 became specialized on monocots during recent evolution.
To characterize the genomic differences between JZL7 and other D. zeae and D. oryzae strains that might underlie their divergent host specificities (Table S1), we first constructed whole-genome alignment between JZL7, CE1, Ech586, EC1, EC2, and MS2, all of which have completely sequenced genomes. Results showed that the six strains are largely conserved in their overall genome structures, with the only exceptions being a large inversion of ;1.4 Mb in the genome of Ech586 and a few smaller translocations in the genomes of EC1 and EC2 (Fig. 3). At the same time, insertions/deletions up to several tens of kilobases are frequently observed and distributed throughout the six genomes (Fig. 3).
We then performed pangenome analyses on the six strains to further examine differences in their gene contents. In total, 4,966 distinct gene families were identified, the majority of which (3,112 families; 62.67%) are shared by all six strains (Fig. 4), indicating a relatively stable core genome. On the other hand, 209 ortholog group family genes are JZL7 specific (Table S4), whereas 9 gene families are absent in JZL7 but shared by all the other five strains, which may contain genes that are critical for the pathogenicity to dicots.
Among the unique genes in the JZL7 genome, an additional set of type IV secretion system (T4SS) that consists of 14 trb genes was detected (Table S5). This extra T4SS in JZL7 shares high levels of nucleotide sequence similarity to the T4SS-encoding gene sets in Ralstonia solanacearum strains FQY_4 and GMI1000, Pseudomonas aeruginosa F30658, and Pectobacterium carotovorum subsp. brasiliense strain SX309. T4SS usually functions in the translocation of nucleic acids or proteins from donor to recipient by conjugation and DNA release/uptake (20). The two sets of T4SSs encoded in the JZL7 genome (Table S5) may have contributed to the abundant strain-specific genes of JZL7 ( Fig. 1 and 3, Tables S6 and S7). However, given that strain JZL7 has more restricted host range than the strains EC1 and MS2 (9), we conjectured that this extra T4SS is not important for the pathogenesis of D. zeae.
The loss of T3SS in JZL7 partially explains its restricted host range. The more restricted host range of JZL7 (Table S1) might be due to the loss of some vital determinants of virulence or host specificity during its evolution. Accordingly, we found in our pangenome analysis 85 gene families that are shared by both EC1 and MS2 but absent in JZL7 (Fig. 4, Table S8), including a set of 30 genes that are colocalized in the genome; they comprise the majority of the dsp gene cluster, which encodes multiple T3SS secreted proteins (e.g., DspE, HrpZ, and HrpW) as well as their chaperone (DspF), and the hrp/hrc gene cluster, which encodes a T3SS (Fig. 5, Table S8). The hecA-hecB genes between the dsp and hrp/hrc clusters, however, were conserved in JZL7; they encode the type V secretion system (T5SS), which contributes to the adherence to hosts (21). Exactly the same organization in this genomic region was found in CE1, the closest relative of JZL7 (Fig. 5). Interestingly, the D. oryzae strain EC2 also lost its dsp and hrp/hrc clusters, but it has a substantially different set of remaining genes.
T3SS forms a syringe-like structure to directly translocate type III secreted effectors (T3SEs) from bacterial cells into host cells and is considered a key determinant for virulence in many bacteria. It has been reported in a number of plant bacterial pathogens (e.g., Erwinia pyrifoliae, Ps. syringae, and Xanthomonas campestris pv. campestris) that the loss of the T3SS would severely reduce the elicitation of host defenses and thus lead to attenuated disease symptoms (22)(23)(24). Moreover, T3SS was demonstrated to be required for the full virulence of D. dadantii 3937 (25,26).
To evaluate whether strain JZL7 could elicit hypersensitive response (HR) as strain MS2 harboring T3SS, we performed HR assay on Nicotiana tabacum K326 (27) and Nicotiana benthamiana (28). Results indicated that strain JZL7 could not cause HR on either of the tobacco leaves, the same as the negative-control LS5 medium, whereas strain MS2 elicited a typical HR quickly in the early stage of inoculation (12 hpi) (Fig. 6). To further verify the importance of T3SS, we respectively deleted the dsp cluster, the hrp/hrc cluster, and both in MS2. HR induction of Ddsp was substantially reduced in the early stage of inoculation compared with that of wild-type MS2, but little difference was found between them after 24 hpi (Fig. 6), suggesting that the T3SEs DspE, HrpZ, HrpW, and HrpK function in the initial infection of plant hosts. The T3SS structure deletion mutant Dhrp/hrc almost could not induce HR both in the early and later stages of infection, while the whole T3SS deletion mutant Ddsp/hrp/hrc significantly attenuated the HR reaction at 12 hpi but recovered part of it at 24 hpi (Fig. 6). HR is a form of programmed cell death (PCD) at the site of pathogen infection commonly controlled by interactions between pathogen avirulence gene products and plant resistance genes. Although the T3SS has been deleted from strain MS2, some avirulence genes, like the known avrL (C1O30_RS14255), are still present in the genome; thus, reduced HR could be observed by mutant infiltration. Furthermore, the components of the T3SS and the T3SEs are not the only pathogen-associated molecular patterns (PAMPs) that plant defenses are known to recognize. Flagellin, peptidoglycan, and lipopolysaccharides also contribute to triggering host innate immune response through recognizing by plant pattern recognition receptors (PRRs). To quantify the function of the T3SS, we infiltrated tobacco leaves with controlled inoculum size. Results showed that the threshold of 10 6 MS2 CFU elicited visible HR at 12 h (Fig. 6E), which may help distinguish between true HR and other types of PCD.
To further investigate the role of T3SS in the pathogenicity and host specificity of D. zeae pathogens, we tested the pathogenicity of Ddsp, Dhrp/hrc, and Ddsp/hrp/hrc mutants in strain MS2 on monocotyledonous banana stems and clivia leaves, as well as dicotyledonous potato and radish slices. We found that the Dhrp/hrc showed 30% reduced virulence on banana and clivia seedlings ( Fig. 7A and B) and 22% and 24% reduced virulence on potato and radish slices, respectively, compared with wild-type MS2 ( Fig. 7C and D). The Ddsp mutant similarly showed lower (35% on potato and 36% on radish) virulence on dicots, whereas its virulence on monocots was 20% reduced on banana seedlings and 7% reduced on clivia seedlings, not significantly different from that of MS2. Notably, the deletion of the whole T3SS (Ddsp/hrp/hrc) dramatically attenuated the virulence of MS2 by more than 60% on monocots ( Fig. 7A and B), to a level that is comparable to that of JZL7, and by more than 65% on dicots ( Fig. 7C and D). Altogether, our results indicate that T3SS has a major role in the virulence of D. zeae. Previous studies have revealed that tissue maceration requires type II secretion (T2SS) of CWDEs and does not involve T3SS (18). Our study revealed the contribution of T3SS to the development of soft rot symptoms, which probably results from the action by the T3SEs regulated by HrpL, as shown in Fig. S1. Unlike JZL7, however, the Ddsp/hrp/ hrc mutant of MS2 still exhibited considerable virulence on dicots, suggesting that there are likely other factors contributing to the host specificity of D. zeae pathogens in addition to T3SS.
HrpL regulates most predicted T3SEs in MS2. In addition to the syringe structure of T3SS, genes in the dsp and hrp/hrc clusters also encode multiple transcription FIG 6 The HR of the T3SS mutants of MS2 and strain JZL7 on Nicotiana tabacum K326 (A, C, and E) and N. benthamiana (B and D) after 12 and 24 hpi. Bacterial cultures (OD 600 of 1.0 in LS5 medium) were dipped by a puncher (5 mm), pressed on the back of tobacco leaves, and incubated at 28°C. The area of HR lesions was measured using Image J 1.52a after 12 and 24 h. **, P , 0.001; ***, P , 0.0001; ****, P , 0.0001; ns, not significant (Student's t test, n = 3 independent experiments). (E) Quantitative determination of the T3SS functions in hypersensitive response indicated that the threshold of 10 6 CFU of MS2 elicits HR at 12 h. regulators and effectors. We generated single-gene knockout mutants for selected regulator (HrpL HrpS, HrpX, and HrpY) or effector (DspE, HrpK, HrpN, HrpW, and HrpZ) coding genes in MS2 and tested their pathogenicity on potato slices. Results showed that three regulator gene mutants (DhrpL, DhrpX, and DhrpY) and three effector gene mutants (DhrpK, DhrpN, and DhrpZ) exhibited significantly reduced virulence compared with that of MS2 wild type (Fig. S1). In particular, the DhrpL mutant showed the most dramatic reduction in virulence, which is comparable to the knockout of the entire T3SS (Ddsp/hrp/hrc) (Fig. S1).
HrpL is a sigma factor activating T3SS at different levels, including gene transcription, mRNA stability, and enzymatic activity in many plant-pathogenic bacteria, such as Erwinia, Pseudomonas, Ralstonia, and Xanthomonas (25, 29-34). In Dickeya bacteria, FIG 7 The virulence of the T3SS mutants of MS2 and strain JZL7 on monocotyledonous and dicotyledonous plants. (A) Strain inoculation on banana seedlings. Every 100 mL of bacterial culture (OD 600 of 1.2 in LB medium) was injected into the basal stem and incubated at 30°C with 12-h alternating light-dark cycles for 7 days. The disease was assessed using the virulence scoring method described previously (17,80). Red lines indicate the mean virulence score of the strains. (B) Strain inoculation on clivia leaves. Every 100 mL of bacterial culture was injected into the base of clivia leaf. (C and D) Strain inoculation on potato and radish slices. Bacterial cells of 2 mL were applied to the center of the tuber slices. Same volume of LB medium was inoculated as a negative control. All plants were kept in a growth chamber with controlled conditions of 28 6 2°C, 75% 6 15% relative humidity and 24-h white light (7,350 lx) illumination until symptoms appeared. Visible macerate areas on clivia and potato and radish slices were measured using Image J 1.52a. Statistical analysis was performed on each group of data, and significantly different values (ANOVA P , 0.05) are indicated by different letters. however, only one effector, DspE, has been characterized alongside two harpins, HrpN and HrpW (35,36). To investigate the regulatory role of HrpL, we first used a combination of four state-of-the-art bioinformatic tools (i.e., Bastion3, BEAN2, DeepT3, and pEffect) to identify candidate T3SEs in the genomes of MS2 and JZL7. As a result, we predicted 34 T3SEs that are present in both strains, as well as 16 and 10 T3SEs that are unique to MS2 and JZL7, respectively ( Table 2). Given that hrpL is absent in JZL7, we measured the expression of all T3SE genes in MS2 wild type and DhrpL mutant using quantitative reverse transcriptase PCR (qRT-PCR). Results showed that the expression of 5 shared T3SE and 8 MS2-unique T3SE genes was significantly downregulated in the DhrpL mutant, including dspE, hrpA, hrpK, hrpS, hrpW, and hrpZ, which are all members of the dsp and hrp/hrc clusters (Fig. 8A). Chi-square tests on the fold change values of shared versus unique T3SEs showed that HrpL regulates the expression of the MS2unique T3SE genes more strongly than the shared ones.
Previous studies in Er. amylovora, Pantoea stewartii, and Ps. syringae have characterized a conserved binding motif for HrpL, which is called the "hrp box" (GGAACC/T-N15/16-C/T/ GCACNNA) (32,33,36). We analyzed the promoter sequences of all predicted T3SE genes and found the typical hrp box in six of them, including the known effector DspE (36), two harpins, HrpN and HrpW, secreted through the T3SS (31), two putative effectors, HrpZ and HrpK, and a T3SS gatekeeper, HrpJ (Table 2). Three other candidate T3SE genes, C1O30_RS12170, C1O30_RS16815, and C1O30_RS19500, encoding hypothetical proteins, also contain the hrp box in their promoters, except the distances between the two conserved modules are abnormally long (69, 56, and 37 nucleotides, respectively) ( Table 2). Additionally, 400 ng of HrpL was demonstrated to bind to the promoter of hrpA encoding a T3SS substrate in D. dadantii 3937 with the help of core RNA polymerase (RNAP) (34). In MS2, HrpA was not predicted as a T3SE but contains a conserved hrp box in its promoter (GGAACC-N15-CTACTTA). To verify the affinity of the HrpL to the above-mentioned hrp box-containing promoters, we carried out electrophoretic mobility shift assays (EMSAs) and found that a high concentration (5 mg) of HrpL could directly bind to the promoters of hrpA, hrpN, hrpZ, hrpK, dspE, hrpX, C1O30_RS12170, C1O30_RS16815, and C1O30_RS19500 in an hrp box-dependent manner without adding core RNAP in vitro (Fig. 8B). This is similar to the phenomenon observed in Bacillus subtilis primary sigma s A , which by itself is able to interact with promoter DNA at the concentration of 10 mM without the assistance from core RNAP (37). The observed binding of HrpL in high concentrations to the promoters of C1O30_RS12170, C1O30_RS16815, and C1O30_RS19500 containing unusual hrp-boxes in the absence of core RNAP is not likely to be biologically relevant or evidence of function, since these atypical hrp-boxes are not functional HrpLdependent promoters, and the expression of these three genes is not significantly affected by HrpL (Fig. 8A). To verify this, low concentrations of HrpL in addition to RNAP were also used to test the affinity to these three promoters with atypical hrp boxes. No bound band was observed for the atypical promoters, while bound bands were formed by incubation of 0.2 mg HrpL protein with hrpA promoter, and 0.4 mg HrpL protein with dspE promoter, with RNAP (Fig. 8C).
The expression of CWDE genes was broadly reduced in JZL7. Plant cell wall is a major barrier for the invasion of pathogenic bacteria and is composed of mainly cellulose, hemicellulose, and pectin polymers. To breach this barrier, D. zeae bacteria produce a full set of cellulases (Cels), pectinases (Pels), polygalacturonases (Pehs), and proteases (Prts), along with a glucuronoxylanase (XynA) and two rhamnogalacturonanate lyases (RhiE1/2) ( Table 3). Our previous study showed that JZL strains produced significantly smaller amounts of all four types of CWDEs compared to EC1 and MS2 when they were all cultivated at the same, relatively high cell density (optical density at 600 nm [OD 600 ] = 1.8) (9).
In this study, we further compared the CWDE activities of strains JZL7 and MS2 at three different cell densities (i.e., OD 600 of 0.5, 1.0, and 1.5) and observed a similar pattern that JZL7 produced a considerably smaller amount of CWDEs (Cels: 41.3%; Pehs: 50%; Pels: 34.6%; Prts: 57.2%) than MS2 at all densities (Fig. 9A). The lower CWDE  activities of JZL7 might be due to either a much smaller repertoire of CWDE encoding genes or a greatly reduced expression of the same set of genes. Our genome comparison revealed a highly conserved set of 34 CWDE genes shared by both JZL7 and MS2 (84.96% to 98.69% DNA sequence identity) and very few strain-specific genes, including two beta-glucosidase-encoding genes (bglC and bglD) in JZL7 and one pectin lyase-encoding gene (pnl) in MS2 (Table 3). Interestingly, recent studies showed that pnl was drastically upregulated during early disease development in Pectobacterium carotovorum but not so in D. dadantii (38). Here, in D. zeae, we found that neither the complementation of pnl MS2 in JZL7 nor the deletion of pnl in MS2 altered the virulence of respective strains on dicotyledonous tissues (i.e., potato and radish slices) (Fig. S2).   (18,39,40). The expression level of T1SS genes in strain JZL7 is 0.66 of those in strain MS2 (Fig. 9B). Furthermore, 15 of the 34 shared CWDE genes had significantly lower expression levels in JZL7 than in MS2, including lfaA, celH, celY, celZ, pehK, pehX, rhiE1, xynA, pelD, pelL, pelN, pelX, pemA, prtX, and prtC (Fig. 9B).
The broadly reduced expression of CWDE genes observed here was consistent with the aforementioned lower CWDE activities in JZL7 and may provide a basis for the strain's lack of virulence on dicots. Specifically, monocots and dicots differ in the composition and structure of their primary cell walls; the former contain mainly cellulose and rarely pectin, while the latter consist of both (41). In this study, we showed that several pectin lyase-encoding genes were either lost (e.g., pnl) or expressed at remarkably lower levels (e.g., pelD, pelL, pelN, pelX, and pemA) in JZL7 (Fig. 9B). The resulting reduction in Pel activity might considerably compromise the ability of JZL7 to degrade cell wall and invade dicot hosts, which, together with the loss of T3SS, might abolish the pathogenicity of JZL7 on dicots. To verify whether the significantly lower expression of CWDEs is due to the lower expression of some important regulators controlling CWDE activity, we measured the expression of the genes encoding CWDE regulators, such as Fis, SlyA, VfmE, PecS, PecT, and KdgR, in both MS2 and JZL7 by RT-PCR. Results showed that none of these genes was differentially expressed between MS2 and JZL7 (Fig. S3).
Other 13 genes were newly found to encode pathogenicity-related proteins in D. zeae MS2. In order to identify additional factors related to the virulence of D. zeae pathogens on dicots, we obtained knockout mutants in MS2 for most of the genes absent in JZL7 but shared by MS2 and EC1 (apart from the T3SS and T3SE genes, which have already been examined above; Table S8). Pathogenicity tests showed that deletion of C1O30_RS01370-01390, C1O30_RS02130, C1O30_RS02370, C1O30_RS02510, C1O30_RS03460, C1O30_RS04230, C1O30_RS04475, C1O30_RS05105, C1O30_RS06880, C1O30_RS07840, C1O30_RS08725, C1O30_RS12525, and C1O30_RS14540 significantly reduced the virulence of MS2 on potato slices (Fig. 10). Among them, C1O30_RS01370-1390, C1O30_RS02510, C1O30_RS03460, C1O30_RS04230, C1O30_RS04475, C1O30_ RS05105, C1O30_RS07840, C1O30_RS08725, C1O30_RS12525, and C1O30_RS14540, respectively, carry a dimethyl sulfoxide reductase anchor subunit gene cluster (dmsC, dmsB, dmsA, dmsD, and a SDR family oxidoreductase) and encode a GNAT family N-acetyltransferase, LysE family transporter, a PAS domain-containing protein (helix-turn-helix transcriptional regulator), an SDR family oxidoreductase, an HAMP domain-containing protein, an FMN-binding negative transcriptional regulator, a glycosyltransferase, DUF2335 domain-containing protein, and a diguanylate cyclase, while all the others encode hypothetical proteins (Table S8). All these genes were first reported to be involved in Dickeya pathogenesis and, thus, represent prominent targets of future studies.
Recently, an saxA gene encoding isothiocyanate hydrolase was found to enable a potato-specific Pectobacterium parmentieri strain to increase the ability to macerate Arabidopsis, suggesting its significant role in defining the host range (42). We searched the homologs of this interesting gene in the genomes of D. zeae and found 4 and 6 copies of saxA, respectively, present in the genomes of MS2 and JZL7. Thus, we ruled out the possibility of SaxA as a host range determinant.
Pathogenic mechanism of Dickeya and other soft rot Pectobacteriaceae. In this study, we sequenced the genome of D. zeae JZL7 and carried out in-depth comparative and functional analyses with other D. zeae strains. The results not only provide insights into the unique pathogenicity and host specificity of JZL7 but also help to better understand the pathogenic mechanism of soft rot Pectobacteriaceae pathogens (e.g, Dickeya, Erwinia, and Pectobacterium) in general.
Xanthomonas, Ps. syringae, and many other bacterial phytopathogens adopt the "stealth" model of pathogenicity whereby they make use of an extensive battery of type III secreted effector proteins and phytotoxins for successful infection. These effectors act to suppress or manipulate host defenses as the bacterial population grows to numbers that are sufficient to induce disease symptoms (43)(44)(45)(46)(47). Additionally, many studies have demonstrated that T3SEs could alter the physiology of host cells in a way that is beneficial for the pathogens. In Xanthomonas, a mutation in the T3SS impairs the ability to inject T3SEs in the host plant and, as a consequence, abolishes pathogenicity and multiplication in planta (48). Among pathogenicity determinants shown to display heterogeneous distribution between strains are T3SEs. Furthermore, in Xanthomonas and Ps. syringae, diversity in T3SE repertoires from different hosts revealed determinants of host specificity (45,49,50).
Soft rot Pectobacteriaceae pathogens also possess in their genomes the molecular machineries required for the stealth mode of infection, such as T3SS, T4SS, phytotoxins, and so on. Functional studies have also demonstrated the importance of T3SS in soft rot Pectobacteriaceae (SRP) pathogens. For instance, T3SS is important for the full virulence of D. dadantii 3937 (25,26) and required for pellicle formation and cell aggregation (51,52). However, T3SS or T3SE has not been reported in SRP as a determinant of host range, and the role of T3SS in the virulence of necrotrophic SRP and other bacterial pathogens appears quite complex.
In our study, we knocked out the whole T3SS gene clusters dsp/hrp/hrc in D. zeae MS2, which resulted in significantly reduced virulence on both monocotyledonous and dicotyledonous hosts (Fig. 7), demonstrating its role as an important virulence factor in the course of disease. However, Ddsp/hrp/hrc was not virulence free on either monocots or dicots, suggesting that T3SS alone does not determine host range in D. zeae. In addition, Er. pyrifoliae, Pe. carotovorum, Pectobacterium wasabiae, and Ps. syringae strains lacking a functional T3SS have been reported to infect plants (22,23,53,54). In Ps. syringae, nonpathogenic isolates are separate from the pathogenic ones due to the deficiency of T3SS and T3SEs (23). In contrast, T3SS-deficient Pe. wasabiae strains are still virulent (53). More strikingly, human-pathogenic Ps. aeruginosa isolates from patients with chronic lung infections are typically T3SS deficient, even though 90% of environmental isolates carry a T3SS (55). These T3SS-deficient pathogens may have alternative ways of modifying the host plant at the initiation of infection.
It has often been assumed that the mechanisms of infection used by SRP are distinctly different from those used by other bacteria like Xanthomonas and Ps. syringae (56). In addition to the above-mentioned stealth model, SRP pathogens also share a "brute-force" model of pathogenicity; they produce many CWDEs to physically attack the plant cell walls and their surrounding apoplast, which thus promotes soft rotting (45,57). It is unquestionable that brute force has made the SRP highly successful pathogens, and the CWDEs are critical factors in both the pathogenicity and host range of SRP. We compared the concentrated enzymatic activities of soft rot D. zeae MS2 and JZL7 and found that broadly reduced expression of CWDE genes might be a major cause for the significantly lower virulence of JZL7 on dicots (Fig. 9). Importantly, the brute-force and stealth models seem to operate in parallel in SRP pathogens, as the production of CWDEs remained unaffected in the T3SS mutants of MS2 (Fig. S4), which might explain their remaining soft rot symptoms on potato and radish slices (Fig. 7C  and D).
Bacterial phytopathogens employ various approaches to infect and kill their host(s). Plant-pathogen interaction is a multifaceted process, mediated by the pathogen-and plant-derived molecules. Secreted as well as translocated molecules, derived from the pathogens, are the key factors, which determine their pathogenicity and allow successful colonization of pathogens on the host. On the other hand, plant-derived molecules are involved in the recognition of pathogen and triggering of the defense response. In the future, to learn more about the latent stage of Dickeya infection and host range, we should consider factors including both CWDEs and T3SEs.

MATERIALS AND METHODS
Bacterial strains and growth conditions. The bacterial strains and plasmids used in this study are listed in Table S1. Escherichia coli strains were grown at 37°C in Luria-Bertani (LB) medium. D. zeae and its derivatives strains were cultivated at 28°C in LB medium, minimal medium (MM), or LS5 medium (58,59). Antibiotics were added to the media at the following final concentrations when required: streptomycin 50 mg/mL, kanamycin 50 mg/mL, tetracycline 15 mg/mL, and ampicillin 100 mg/mL. Genome sequencing and assembly. Genomic DNA was extracted from D. zeae strain JZL7 in LB medium culture using MasterPure DNA purification kit (EPICENTRE Biotechnologies, Madison), which was then subjected to quality control by agarose gel electrophoresis and quantified by Qubit. The genome sequencing was performed at Health Time Gene (Shenzhen, China) using both the PacBio RS II and the Illumina Hiseq X Ten platforms.
For the PacBio sequencing, genomic DNA was treated into fragments in 10 kb by g-TUBE first. The fragments were damage-repaired and end-repaired. Both sides of the DNA fragments were, respectively, connected with hairpin adapter to get a dumbbell structure, which is known as SMRTbell. After annealing, the SMRTbell was fixed at the bottom of the ZWM polymerase and was sequenced last. Adapter trimming and quality filtering were performed by using fastp (version 0.20.0) (60) with default parameters.
For the Illumina sequencing, genomic DNA was randomly broken, and 350 bp fragments were purified and end-repaired using T4 DNA polymerase, Klenow DNA polymerase, and T4 PNK. The fragments were added with an "A" base at the 39 terminal and ligated with the adaptor with a "T" base at its terminal. The library was sequenced on the Hiseq X Ten platform to generate PE150 reads. A hybrid genome assembly with both the PacBio long-reads and Illumina short-reads was carried out using Unicycler v0.4.7 (default parameters) (61).
Generation of in-frame deletion mutants. For generation of in-frame deletion mutants of D. zeae MS2, triparental mating was performed using the methods described previously (15). For instance, to knock out the hrp/hrc gene cluster of T3SS in strain MS2, the upstream and downstream fragments of hrp/hrc gene cluster were amplified using the primer pairs of hrp/hrc-1 accompanied with hrp/hrc-2, and hrp/hrc-3 accompanied with hrp/hrc-4, respectively (Table S3), and purified with AxyPrep DNA gel extraction kit (Axygen Biotech Co., Hangzhou, China). The fragments were ligated with the BamHI and Spe digested suicide plasmid pKNG101 using ClonExpress MultiS kit (Vazyme Biotech Co., Nanjing, China) and transformed into E. coli CC118 competent cells. In triparental mating, the donor and receptor cells were mixed with the helper strain E. coli RK2013 (grown in LB medium containing kanamycin) in a ratio of 2:1:1 on LB plate and incubated at 28°C overnight. The transformants were grown on minimal medium (MM) agar plates (13) containing streptomycin sulfate. Single colony culture was then spread on MM agar plate containing 5% sucrose to exclude the suicide plasmid. The resultant deletion mutants were confirmed by PCR using the detection primer pair of hrp/hrc-F and hrp/hrc-R and DNA sequencing. In the same way, we correspondingly deleted the important T3SS transcriptional regulation genes hrpX, hrpY, hrpS, and hrpL and the candidate T3SE genes dspE, hrpZ, hrpW, hrpK, and hrpN, the pectin lyaseencoding gene pnl, and, compared to JZL7, some genes specific to MS2. In addition, the Ddsp/hrp/hrc mutant was also generated by deleting the dsp gene cluster based on the obtained Dhrp/hrc mutant. The primers are listed in Table S3.
Measurement of bacterial growth curves. Bacterial strains were grown in LB medium overnight at 28°C, adjusted to an OD 600 of 1.5, and diluted into fresh LB and LS5 medium in a 1:100 ratio, respectively (9). Aliquots of 500-mL dilutions were transferred into 2.0-mL tubes. Bacteria were grown with shaking at 200 rpm under 28°C, and cell density was measured every 2 h. The experiment was repeated three times in triplicate.
HR assay. Bacteria were grown in LS5 medium (pH 5.5) (59) until they reached an OD 600 of 1.0. Leaves of Nicotiana tabacum variant K326 (27) and N. benthamiana (28) were, respectively, inoculated on the back by pressing a sterilized puncher (5 mm) dipped with bacterial culture and incubated at 28°C. The area of HR lesions on leaves was measured using Image J 1.52a 12 and 24 h postinoculation (hpi) (79). To quantify the function of the T3SS, we infiltrated leaves of N. tabacum K326 with 100 mL of MS2 bacterial dilutions in LS5 medium (OD 600 of 0.5, containing approximately 10 8 CFU/mL), and 100 mL each of LS5 medium, Ddsp/hrp/hrc, and JZL7 (10 7 CFU/mL) was also infiltrated as a negative control. Each assay was repeated three times in triplicate.
Pathogenicity tests on monocotyledonous and dicotyledonous plants. Strains were cultured to the logarithmic phase until an OD 600 of 1.2 was reached in LB medium, and pathogenicity tests were carried out on monocotyledonous and dicotyledonous hosts. For monocots, every 100 mL of bacterial culture was injected into the basal stem of banana and the base of clivia leaf. For dicotyledonous plants, potato and radish tubers were surface-sterilized with 70% ethanol, cut evenly about 5 mm in thickness, and then placed onto moistened filter paper in a tray. Bacterial cells of 2 mL were applied to the center of the tuber slices. All plants were kept in a growth chamber with controlled conditions of 28 6 2°C, 75% 6 15% relative humidity, and 24 h white light (7,350 lx) illumination until symptoms appeared, except that the bananas were incubated at 30°C with 12-h alternating light-dark cycles for 7 days and then disease was assessed using a modified virulence scoring (17,80). Same volume of LB medium was inoculated as a negative control. Visible macerate areas on clivia leaves and potato and radish slices were measured using Image J 1.52a. Each assay was repeated three times in triplicate.
RNA purification. The bacterial cultures used for RNA extraction were the ones described below for CWDE activity measurement. Since the same strain shared basically the same enzymatic activity in different cell densities (OD 600 of 0.5, 1.0, and 1.5), the bacterial cultures of strains MS2 and JZL7 at an OD 600 of 1.5 were selected for RNA extraction. On the other hand, to determine whether the predicted T3SE expression is induced by the HrpL regulator, we grew strains MS2 and DhrpL in LS5 medium (pH 5.5) (59) for RNA extraction until an OD 600 of 1.0 was reached. The RNA was extracted using the SV total RNA isolated system kit (Promega, Madison, WI, USA), further purified using the RNA clean kit (Qiagen, Hilden, Germany), and treated with DNase I to degrade any possible DNA contamination. Quantity of RNA was first measured using a NanoDrop 2000c (Thermo Fisher Scientific, MA, USA), and the integrity of RNA was detected by agarose gel electrophoresis.
RT-PCR analysis. For each RNA sample for CWDE activity measurement, two dilutions (5 and 50 ng) were reverse transcribed into cDNAs using FastKing gDNA dispelling RT mix (Tiangen Biotech, Co., Ltd., Beijing, China), and the concentration of each resultant cDNA was quantitatively equilibrated according to the expression quantities of reference gene infB of strains MS2 and JZL7. PCR was performed using the primers listed in Table S3, and gene expression (signal intensity) of each gene was determined using the software Image Lab (Bio-Rad, USA). The experiment was repeated three times in triplicate.
qRT-PCR analysis. In qRT-PCR analysis, an aliquot of 300 ng RNA sample was used for genomic DNA elimination and cDNA synthesis by a FastKing RT kit (with gDNase) (Tiangen Biotech, Co., Ltd., Beijing, China) following the manufacturer's protocol. The housekeeping gene atpD was used as a reference. The PCR efficiency of each gene was determined using five DNA standards at different concentrations (10, 1, 0.1, 0.01, and 0.001 g/mL). The qRT-PCR was conducted on a Quantstudio 6 Flex system using ChamQ Universal SYBR qPCR master mix (Vazyme Biotech Co., Nanjing, China) with the following cycle profile: 1 cycle at 95°C for 30 s, followed by 40 cycles at 95°C for 10 s, 60°C for 30 s, and then 1 cycle at 95°C for 15 s, 60°C for 60 s, and 95°C for 15 s. The experiment was repeated three times, each time in triplicate. Data were analyzed using the 2 2DDCT method as described previously (81).
Protein expression and purification. The open reading frame (ORF) that encodes HrpL protein was amplified from MS2 genomic DNA using primers pET32a-hrpL-F and pET32a-hrpL-R containing BamHI and HindIII restriction enzyme sites (Table S3). The PCR product was purified and ligated to the BamHI/HindIIIdigested pET32a vector harboring a thioredoxin (TRX)-His 6 tag at its N terminus. pET32a-hrpL was then transformed into the E. coli strain BL21(DE3) competent cells (TransGen Biotech Co., Beijing, China) and confirmed by sequencing. To express the HrpL protein, a single colony of BL21(pET32a-hrpL) was grown in LB medium containing 100 mg/mL of ampicillin overnight and transferred into 1 L of fresh LB medium in a 1:100 ratio to grow until reaching an OD 600 of 0.6. Then, 1 mM IPTG (isopropyl b-D-1-thiogalactopyranoside) was added to the culture to induce protein expression at 18°C for 16 h. Bacterial cells were harvested by centrifuging at 5,000 rpm at 4°C for 30 min and suspended in 20 mL of phosphate-buffered saline (PBS). The cells were lysed with sonication and centrifuged at 12,000 rpm at 4°C for 30 min. The supernatant was filtered with a 0.45-mm-pore filter (MilliporeSigma, St. Louis, MO, USA), and the filtrate was added into a Ninitrilotriacetic acid (NTA) column (Changzhou Smart-Life Sciences Biotechnology Co., Changzhou, China) that had been equilibrated with lysis buffer. After washing five times with lysis buffer, the column was eluted with 30 mL gradients of 10 to 500 mM imidazole prepared in wash buffer, respectively. Fractions were collected, and sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was used to verify the molecular weight of the target protein. The protein was concentrated by centrifugation (MilliporeSigma, St. Louis, MO, USA) at 4°C and then stored at 280°C.
Electrophoretic mobility shift assay. Electrophoretic mobility shift assay (EMSA) was performed as described previously (58). DNA probes were PCR amplified using primers listed in Table S3. The purified PCR fragments were labeled by biotin using the biotin 39 end DNA labeling kit (Thermo Fisher Scientific, MA, USA). Binding reactions were performed in a final volume of 10 mL using LightShift chemiluminescent EMSA kit following the manufacturer's protocol (Thermo Fisher Scientific, MA, USA). For each reaction, 1 mL of 10Â binding buffer, 0.5 mL of 1 mg/mL Poly, 1 nmol of labeled DNA fragments, and 1, 5, or 10 mg of HrpL protein placed in an ice bath were added together and incubated for 30 min at 20°C. The protein-DNA complexes and the unbound free DNA fragments were separated on 6% nondenaturing polyacrylamide (acrylamide/bisacrylamide 29: 1 [vol/vol]) gels using the electrophoresis buffer Trisborate-EDTA (TBE) and were detected by chemiluminescence (Tanon, China). A 100-fold molar excess of unlabeled DNA fragment was incubated with HrpL protein for 15 min before addition of the labeled DNA fragments to verify specific interaction of the HrpL protein-DNA fragment. In the experiment of adding core RNA polymerase (RNAP), E. coli core RNAP (0.5 U) (New England Biolabs [Beijing] Ltd., Beijing, China) (82) was incubated on ice for 20 min with 200 or 400 ng of purified HrpL protein (34). The remaining steps were as described above.
Measurement of CWDE activities. The activities of CWDEs were measured according to the methods described previously (9). Specifically, pectate lyase (Pel) assay medium (10 g/L polygalacturonic acid, 10 g/L yeast extract, 8 g/L agarose, 0.38 mM CaCl 2 , and 100 mM Tris-HCl [pH 8.5]), polygalacturonase (Peh) assay medium (5 g/L polygalacturonic acid, 2 g/L sucrose, 2 g/L (NH4) 2 SO 4 , and 15 g/L agar [pH 5.5]), cellulase (Cel) assay medium (1.0 g/L carboxymethyl ethyl cellulose, 3.8 g/L Na 3 PO 4 , and 8.0 g/L agarose [pH 7.0]), and protease (Prt) assay medium (10 g/L skimmed milk, 5 g/L Bacto tryptone, 2.5 g/L yeast extract, 5 g/L NaCl, and 15 g/L agar) were prepared, and 30 mL of each medium was added into a 10 by 10 cm square plate. Wells in 5 mm diameter were made, and 20 mL of bacterial cells (OD 600 of 0.5, 1.0, or 1.5) was applied to the wells. Plates were incubated at 28°C until Pel and Peh assay plates were treated with 1 M HCl after 14 h, and Cel assay plate was stained with 0.1% Congo red for 15 min after 14 h and decolored with 1 M NaCl twice. The protease activity was measured without any further treatment after 24 h. The experiment was repeated three times with duplicates.
Introduction of pnl gene in strain JZL7. To create the complementing plasmid, sequence of pnI open reading frame (ORF) with 19 bp before the start codon harboring a ribosome-binding site (RBS) was amplified and cloned from strain MS2 into the EcoRI/BamHI-digested pLAFR3 vector using pEASY-Uni seamless cloning and assembly kit (TransGen Biotech Co., Beijing, China) and transformed into E. coli DH5a competent cells. Plasmid construct was confirmed by DNA sequencing and introduced into strain JZL7 by conjugal triparental mating. The primers used here are listed in Table S3.
Statistical analysis. All the experiments were repeated three times in duplicate or triplicate. GraphPad Prism 8.4.1 (GraphPad Software, San Diego, California) was used to performed unpaired twotailed t test (83), and the data of D. zeae strain JZL7 were normalized to those of strain MS2. * indicates P , 0.05, ** indicates P , 0.01, *** indicates P , 0.001, and **** indicates P , 0.0001. Analysis of variance (ANOVA) was used between D. zeae MS2 and its mutants. The values are means of three replicates, and the error bars are standard deviations. Statistical analysis was performed on each group of data, and significantly different values (ANOVA P , 0.05) are indicated by different letters.
Data availability. Both the original genome sequencing data and the genome assembly were deposited in the NCBI Sequence Read Archive (SRA) database under the accession number PRJNA656647. The JZL7 genome sequence has been deposited in GenBank under accession number CP060263.1.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.