Distribution and Replication of the Pathogenicity Plasmid pPATH in Diverse Populations of the Gall-Forming Bacterium Pantoea agglomerans

Abstract

Pantoea agglomerans has been transformed from a commensal bacterium into two related gall-forming pathovars by acquisition of pPATH plasmids containing a pathogenicity island (PAI). This PAI harbors an hrp/hrc gene cluster, type III effectors, and phytohormone biosynthetic genes. DNA typing by pulsed-field gel electrophoresis revealed two major groups of P. agglomerans pv. gypsophilae and one group of P. agglomerans pv. betae. The pPATH plasmids of the different groups had nearly identical replicons (98% identity), and the RepA protein showed the highest level of similarity with IncN plasmid proteins. A series of plasmids, designated pRAs, in which the whole replicon region (2,170 bp) or deleted derivatives of it were ligated with nptI were generated for replicon analysis. A basic 929-bp replicon (pRA6) was sufficient for replication in Escherichia coli and in nonpathogenic P. agglomerans. However, the whole replicon region (pRA1) was necessary for expulsion of the pPATH plasmid, which resulted in the loss of pathogenicity. The presence of direct repeats in the replicon region suggests that the pPATH plasmid is an iteron plasmid and that the repeats may regulate its replication. The pPATH plasmids are nonconjugative but exhibit a broad host range, as shown by replication of pRA1 in Erwinia, Pseudomonas, and Xanthomonas. Restriction fragment length polymorphism analyses indicated that the PAIs in the two groups of P. agglomerans pv. gypsophilae are similar but different from those in P. agglomerans pv. betae. The results could indicate that the pPATH plasmids evolved from a common ancestral mobilizable plasmid that was transferred into different strains of P. agglomerans.

The bacterial species Pantoea agglomerans (Beijerinck 1888) comb. nov. belonging to the Enterobacteriaceae family was proposed by Gavini et al. (18) to include strains of Erwinia herbicola, Enterobacter agglomerans, and the Erwinia milletiae complex. P. agglomerans is widely distributed in nature and has been isolated from numerous ecological niches, including plants, water, humans, and animals. It is frequently associated with plants as an epiphyte (49) or an endophyte (27) but is rarely associated with plants as a pathogen. However, some isolates of this bacterium have been reported to be tumorigenic pathogens, inducing galls on gypsophila (10), beet (7), Douglas fir (14), wisteria (45), and cranberry (51). Gall formation on gypsophila and beet is caused by two related pathovars; P. agglomerans pv. gypsophilae is pathogenic only on gypsophila (10) but elicits a hypersensitive response on beet (15), whereas P. agglomerans pv. betae is pathogenic on both beet and gypsophila (7). The pathogenicity of both pathovars is determined by the presence of an indigenous plasmid, designated pPATH_Pag and pPATH_Pab for P. agglomerans pv. gypsophilae and P. agglomerans pv. betae, respectively (35). pPATH_Pag of P. agglomerans pv. gypsophilae strain 824-1, which has been extensively investigated, is 135 kb long and harbors a pathogenicity island (PAI) that is approximately 75 kb long. This PAI carries the hrp/hrc gene cluster containing a functional type III secretion system (T3SS), a cascade of regulatory genes, and a gene encoding harpin (41, 43, 44). Additionally, six experimentally confirmed T3SS effectors, a cluster of indole-3-acetic acid and cytokinin biosynthetic genes, and numerous highly diverse insertion (IS) elements are also present in this PAI (9, 15, 21, 31, 32, 50). Simultaneous inactivation of the pathways for biosynthesis of indole-3-acetic acid and cytokinins in P. agglomerans pv. gypsophilae 824-1 significantly reduced the gall size but did not prevent gall development (37). In contrast, any mutation in the T3SS or hrp regulatory genes of P. agglomerans pv. gypsophilae or P. agglomerans pv. betae completely eliminated gall formation (43, 44), indicating that T3SS effectors rather than phytohormones secreted by the pathogen may act as primary pathogenic factors in gall formation. Moreover, studies of the T3SS effectors HsvG, HsvB, and PthG of P. agglomerans pv. gypsophilae and P. agglomerans pv. betae strongly suggest that they play a role in both the pathogenicity of P. agglomerans and its divergence into two pathovars (4, 16, 42).

The PAI concept has emerged to describe genomic regions of pathogens which carry virulence genes together with loci whose presence strongly indicates horizontal gene transfer between species or even genera (26). Moreover, PAI structures strongly reflect different stages of evolution (22). The PAI in pPATH_Pag of P. agglomerans pv. gypsophilae 824-1 has been shown to comply with all the major criteria in the PAI definition (21, 34). Additionally, many residual sequences of known genes from various bacteria are interspersed within the PAI (21). This information, together with the PAI's large size, multiple IS elements, and location on a plasmid, supports the premise that this PAI is currently in the early stages of evolution (22). In contrast, chromosomal hrp PAIs, such as that of Pseudomonas syringae pv. tomato DC3000, exhibit a compact tripartite mosaic structure that provides an example of an evolutionarily advanced PAI (2). It is noteworthy that the difference in evolutionary stages between the PAIs of P. agglomerans pv. gypsophilae and P. syringae pv. tomato seems to have an impact on the relative contributions of T3SS effectors to virulence. Thus, the pPATH_Pag-borne PAI accommodates nine putative T3SS effectors, and mutagenesis of six of these effectors almost completely or significantly reduced gall formation (5). In contrast, more than 45 T3SS effectors have been characterized in the model strain P. syringae pv. tomato DC3000. However, only a very few mutations in these effectors had a slight effect on virulence, apparently due to functional redundancy (1). The PAIs in pPATH_Pag and pPATH_Pab share identical IS elements, which are present only in these plasmids and were presumably instrumental in their evolution (21, 32, 34). These observations may suggest that the two PAIs had a common origin. However, distinct morphologies of galls induced by P. agglomerans pv. gypsophilae and P. agglomerans pv. betae on their common host gypsophila have been reported (7), which suggests that there are genetic differences.

An intriguing question is how P. agglomerans evolved from a commensal bacterium present on many different plants into a host-specific gall-forming bacterium. The answer to this question is obviously complex and requires understanding both the evolution of the plasmid-borne PAI and its distribution among P. agglomerans populations. Much of the lateral gene transfer among bacteria occurs through the action of conjugative plasmids that encode all of the functions necessary for their hosts to transmit them to recipient cells, or alternatively, they may be transferred by mobilizable plasmids and can utilize Tra functions of a promiscuous self-transmissible plasmid (48). The pPATH plasmid provides a framework for evolution of the PAI and may serve as a vehicle for its transfer to other bacteria.

The present study focused on the pPATH plasmids of P. agglomerans pv. gypsophilae and P. agglomerans pv. betae and sought answers to the following questions. Are the pPATH plasmids distributed in clonal or diverse populations of P. agglomerans? How conserved are PAI structures of different pPATH plasmids? Do the pPATH plasmids of different pathogenic strains share the same replicon? Is the pPATH plasmid a conjugative plasmid? Answers to these questions may contribute to greater insight into the power of genetic exchange in de novo emergence of new phytopathogens or phytopathogenic variants.

MATERIALS AND METHODS

Bacterial strains, growth conditions, and pathogenicity tests.

All bacterial strains used in this study are listed in Table 1. Escherichia coli and P. agglomerans were grown in LB medium (47) at 37 and 28°C, respectively. The P. agglomerans strains included strains of P. agglomerans pv. gypsophilae (10), the beet pathovar P. agglomerans pv. betae (7), P. agglomerans pv. milletiae pathogenic on Wisteria sinensis (45), P. agglomerans strains associated with cranberry stem galls, and nonpathogenic strains, most of which were isolated from gypsophila. Pantoea stewartii subsp. stewartii pathogenic on corn was also included in the genomic analyses. The identity of the Pantoea strains was confirmed by 16S rRNA gene sequencing (54) based on the Ribosomal Database Project (http://rdp.cme.msu.edu). Stereotype groups of P. agglomerans pv. gypsophilae strains were determined as described previously (36). Genetic manipulations were carried out in E. coli DH5α (Table 1). Antibiotics were used at the following concentrations: ampicillin, 150 μg/ml; kanamycin (Km), 30 μg/ml; spectinomycin (Spec), 50 μg/ml; streptomycin (Sm), 50 μg/ml; and rifampin, 150 μg/ml. Pathogenicity tests with cuttings of Gypsophila paniculata cv. perfecta were carried out as described by Valinsky et al. (50). Pathogenicity tests with table beet cubes were performed as described by Ezra et al. (15).

TABLE 1.

Bacterial strains and plasmids used in this study

Strain(s) or plasmid	Relevant characteristics	Source or reference(s)
Strains
E. coli DH5α	lacZΔM15 Δ(lacZYA-argF)U169 gyrA96	Invitrogen
P. agglomerans 299R, BRT98, 23-9, 3-1, 717-2, 340, 163-5, and 4/4	Nonpathogenic strains isolated from gypsophila and other sites	6, 36
P. agglomerans pv. gypsophilae 3-1a, 13, 420, 300, 6-2, 102, D1, PD713, 53, D3011, Z1, 24-2, 824-1, M1, 350-1, 350-2, 350-4, 615, 611, 441, and PD459	Pathogenic on gypsophila	36; this study
P. agglomerans pv. gypsophilae 824-1r	Rifr, wild-type pathogenic strain	36
P. agglomerans pv. gypsophilae 824-1Mx27	Kmr, marker exchange mutant in pthG	15
P. agglomerans pv. betae 4188, 2188, 4430, 4303, 1188, and 3188	Pathogenic on gypsophila and beet	7; this study
P. agglomerans isolated from cranberry strains 13-99, PA4/99, and Pa5-99	Pathogenic on cranberry	39
P. agglomerans pv. milletiae 1 and 800	Pathogenic on milletiae (wisteria)	45
P. stewartii subsp. stewartii SS104	Pathogenic on corn	40
E. amylovora 238 and 244	Isolated from pear tree	S. Manulis collection
E. caratovora subsp. atroseptica 1843-b/LE	Isolated from potato	S. Manulis collection
E. caratovora subsp. carotovora EC1	Isolated from potato	S. Manulis collection
Erwinia chrysanthemi 6453	Isolated from potato	S. Manulis collection
Xanthomonas campestris pv. campestris CR5	Isolated from cauliflower	S. Manulis collection
P. syringae pv. tomato DC3000	Isolated from tomato	G. Sessa, Tel-Aviv University
Plasmids
pGreenII0029	Kmr, binary vector for Agrobacterium transformation containing the nptI gene	25
pGEM-T Easy	Apr, T-vector for PCR cloning	Promega
pRA1	Kmr, 2.17 kb of pPATH replicon region from strain P. agglomerans pv. gypsophilae 824-1 ligated to nptI	This study
pRA2	Kmr, 1,750 bp of pPATH replicon from strain P. agglomerans pv. gypsophilae 824-1 ligated to nptI	This study
pRA3	Kmr, 1,200 bp of pPATH replicon from strain P. agglomerans pv. gypsophilae 824-1 ligated to nptI	This study
pRA4	Kmr, 1,899 bp of pPATH replicon from strain P. agglomerans pv. gypsophilae 824-1 ligated to nptI	This study
pRA5	Kmr, 1,479 bp of the pPATH replicon from strain P. agglomerans pv. gypsophilae 824-1 ligated to nptI	This study
pRA6 (base replicon)	Kmr, 929 bp of the pPATH replicon from strain P. agglomerans pv. gypsophilae 824-1 ligated to nptII	This study
pRep824	Apr, 2.17 kb of pPATH replicon from strain P. agglomerans pv. gypsophilae 824-1 cloned into pGEM-T Easy	This study
pRep350	Apr, 2.17 kb of pPATH replicon from strain P. agglomerans pv. gypsophilae 350-1 cloned into pGEM-T Easy	This study
pRep4188	Apr, 2.17 kb of pPATH replicon from strain P. agglomerans pv. betae 4188 cloned into pGEM-T Easy	This study

Plasmids and DNA manipulation.

The cloning vectors and plasmids used and constructed in this study are listed in Table 1. DNA isolation, agarose gel electrophoresis, electroporation, PCR, and other DNA manipulation procedures were performed according to standard protocols (3, 47) or as recommended by the supplier of the materials. DNA fragments were amplified by PCR with Taq polymerase (Super-Therm polymerase JMR-801; Roche, Mannheim, Germany), and synthetic oligonucleotides (see Table S1 in the supplemental material) were synthesized according to the manufacturer's specifications (Sigma-Aldrich, Rehovot, Israel). Transformation of E. coli DH5α and Pantoea strains was performed by electroporation with a Gene Pulser apparatus (Bio-Rad Laboratories, Hercules, CA) used according to the manufacturer's instructions. Southern hybridization was performed with the ECL direct nucleic acid labeling system (Amersham Biosciences, Uppsala, Sweden) as described by the manufacturer.

PCR-amplified DNA and plasmid DNA were purified with an AccuPrep plasmid extraction kit (Bioneer, San Francisco, CA) prior to sequencing or cloning into the pGEM-T Easy vector (Promega, Madison, WI). Automated sequencing with Taq DNA polymerase was carried out at the Biological Services Laboratories of Tel-Aviv University with an ABI Prism 3100 genetic analyzer (Applied Biosystems, Foster City, CA). When necessary, the sequence was completed with custom primers. Analyses of the sequence data for the DNA and deduced protein sequences were performed mainly as described in the program manual for the Wisconsin Package, version 11 (Genetic Computer Group, Madison, WI). The additional programs used for searching in the GenBank were Motif Scan (ExPASy Tools), BlastW or BlastN (NCBI), and ClustW (EMBO).

For deletion analysis of the pPATH replicon, a series of pRA plasmids were constructed by PCR amplification with suitable primers and P. agglomerans pv. gypsophilae 824-1, P. agglomerans pv. gypsophilae 350-1, or P. agglomerans pv. betae 4188 cells as templates. The DNA fragments obtained were then blunt end ligated into a Km^r cassette (nptI) to obtain the pRA plasmids. nptI was generated by PCR amplification from pGreenII0029, using primers NptI5pgreen and NptI3pgreen (see Table S1 in the supplemental material). The pRA1, pRA2, and pRA3 plasmids were generated by using primers Rep25101_3, Rep24681_3, and Rep24131_3, respectively, for the 3′ end and primer Rep22950_5 for the 5′ end (see Table S1 in the supplemental material). pRA4, pRA5, and pRA6 were generated by using primers Rep25101_3, Rep24681_3, and Rep24131_3, respectively, for the 3′ end and primer Rep23220_5 for the 5′ end. Similarly, pRA7 was constructed with primers Rep23220_5 and Rep24100_3, whereas pRA8 was generated with primers Rep 24131_3 and Rep23270_5.

To test whether pPATH_Pag is a self-transmissible plasmid, spontaneous mutants resistant to Spec and Sm of nonpathogenic P. agglomerans strains 3-1, 717-2, and 23-9 were obtained by repeated growth on LB agar supplemented first with one and then with both antibiotics. The mutant P. agglomerans pv. gypsophilae 824-1Mx27 containing a Km^r cassette in the pthG gene (17) was used as an antibiotic tag for pPATH_Pag in the rifampin-resistant strain P. agglomerans pv. gypsophilae 824-1. This mutation caused a 50% reduction in the gall size in gypsophila and made P. agglomerans pv. gypsophilae pathogenic on beet instead of eliciting a hypersensitive response (16, 17). Thus, transfer of the mutated pPATH_Pag plasmid into the nonpathogenic strains should have made them pathogenic on beet and gypsophila. Mixtures containing the Spec^r Sm^r nonpathogenic strains described above in different combinations with P. agglomerans pv. gypsophilae 824-1Mx27 (10⁷CFU/ml per strain) were used for both in vitro and in vivo conjugation tests involving repeated growth on LB agar and inoculation into gypsophila cuttings and beet cubes, respectively. Following 15 transfers on LB agar the bacteria were screened for resistance using Spec, Sm, and Km. Similarly, extracts of 3-week-old galls on gypsophila and beet cubes were screened for resistance to antibiotics, as described above. All attempts to demonstrate conjugation by means of these procedures were negative.

DNA-based typing procedures.

The protocol used for pulsed-field gel electrophoresis (PFGE) was adapted from the protocol of Ribot et al. (46). Cell suspensions were prepared by removing the cells from the surfaces of LB agar plates that had been incubated overnight and suspending them in an Eppendorf tube containing 250 μl of suspension buffer (100 mM Tris, 100 mM EDTA; pH 8). The optical density at 600 nm of each cell suspension was adjusted to 1.33, and the suspension was mixed gently with 12.5 μl of 20 mg/ml proteinase K (Sigma-Aldrich) prior to addition of 250 μl of SeaKem Gold agarose (F50152; FMC, Rockland, ME) in TE buffer (10 mM Tris, 1 mM EDTA; pH 8) at 50°C. The agarose-cell suspension mixture was dispensed immediately into the wells of reusable plug molds (Bio-Rad Laboratories, Hercules, CA) and allowed to solidify at room temperature for 15 min. The plugs were transferred to 50-ml polypropylene tubes containing 2.5 ml of cell lysis buffer (50 mM Tris, 50 mM EDTA [pH 8], 1% sarcosine, 12.5 μl of a 20-mg/ml proteinase K solution). Lysis was allowed to occur for 4 h at 50°C with agitation in an orbital shaker. The plugs were then washed twice in 5 ml double-distilled water and four times in TE buffer at 50°C. The water and TE buffer were prewarmed to 50°C before each washing step. The plugs were either used for restriction digestion or stored in TE buffer at 4°C for up to 6 months. A 2-mm-wide slice of each plug was subjected to restriction digestion with 30 U of either XbaI, AvrII, or PsvXI restriction enzyme in 150 μl buffer for 16 h at 37°C. The plug slices were loaded into appropriate wells of a 1% SeaKem Gold agarose gel prepared in 0.5× Tris-borate-EDTA buffer (catalog no. T4415; Sigma). Lambda concatemer (catalog no. 1703635; Bio-Rad Laboratories) was employed as a size marker in the range from 50 to 1,000 kb. Electrophoresis was performed with the CHEF-DR III system (Bio-Rad). The electrophoresis conditions comprised an initial switch time of 0.2 s, a final switch time of 54.2 s, and application of a gradient of 6 V/cm at an angle of 120° at 14°C for 22 h. After electrophoresis, the gels were stained with ethidium bromide for further analysis. The PFGE patterns were analyzed with the Molecular Analysts Fingerprinting II software package, version 3 (Bio-Rad). Matching and dendrogram unweighted-pair group method using average linkage analysis of the PFGE patterns were performed by using the Dice coefficient with a 1 to 1.5% tolerance window.

Repetitive extragenic palindromic PCR was performed as described by Malathum et al. (33) with primers REP1R-Dt and REP2-Dt (see Table S1 in the supplemental material). Amplified fragment length polymorphism was carried out as described by Vos et al. (53).

Nucleotide sequence accession numbers.

The rRNA gene sequences of P. agglomerans pv. gypsophilae 824-1 and 350-1 and P. agglomerans pv. betae 4188 have been deposited in the GenBank database under accession numbers EF173382, EF173380, and EF173381. The sequences of pRep824, pRep350, and pRep4188 have been deposited in the GenBank database under accession numbers EF173387, EF173386, and EF173388.

RESULTS

Pathogenic strains of P. agglomerans are genotypically diverse.

It has been reported that P. agglomerans (previously E. herbicola) pathogenic on gypsophila (i.e., P. agglomerans pv. gypsophilae strains) can be divided into two serotype groups, groups S1 and S2 (36). Forty-one strains of P. agglomerans, including 23 P. agglomerans pv. gypsophilae strains, six P. agglomerans pv. betae strains, two P. agglomerans pv. milletiae strains, three P. agglomerans strains associated with cranberry stem galls, and seven nonpathogenic P. agglomerans strains, as well as one strain of P. stewartii subsp. stewartii (Table 1), were analyzed to determine the DNA type by PFGE (Fig. 1). Depending on the level of genetic relatedness, the following major clusters were identified for pathogenic and nonpathogenic P. agglomerans strains: P. agglomerans pv. gypsophilae group S1 (21 strains, including 5 nonpathogenic strains) with 90% similarity; P. agglomerans pv. gypsophilae group S2 (six strains) with 95% similarity; and the P. agglomerans pv. betae group (six strains) with 87% similarity. The two P. agglomerans pv. gypsophilae clusters, i.e., groups S1 and S2, showed approximately 60% similarity, whereas the similarity between these two clusters and the P. agglomerans pv. betae group was about 55%. It is noteworthy that the two serotype groups, groups S1 and S2, also were genotype groups. Sequencing of the 16S rRNA gene with primers BSF8/20 and BSR1114/16 was employed to verify the taxonomic identification of the following representative strains belonging to the three major clusters: P. agglomerans pv. gypsophilae 824-1 (group S1) and 350-1 (group S2) and P. agglomerans pv. betae 4188. Based on the Ribosomal Database Project, all three strains were identified as P. agglomerans (>96% base pair identity).

FIG. 1.

PFGE patterns of pathogenic and nonpathogenic P. agglomerans strains after macrorestriction with XbaI (center panel). An unweighted average linkage dendrogram resulting from the cluster analysis based on PFGE patterns is shown on the left. The numbers on the scale bar indicate percentages of similarity as determined by the Dice coefficient. The data on the right indicate the isolate designation, the serotype group, the pathogenicity on gypsophila cuttings (G), the pathogenicity on beet (B), and the presence of repA as determined by PCR amplification with primers Rep23220_5 and Rep25101_3 (see Table S1 in the supplemental material). Pss, P. stewartii subsp. stewartii; Pa, nonpathogenic P. agglomerans; Pag, P. agglomerans pv. gypsophilae; Pam, P. agglomerans pv. milletiae; Pac, P. agglomerans strain associated with cranberry stem galls; Pab, P. agglomerans pv. betae; ND, not determined.

The three P. agglomerans strains isolated from cranberry were clonal and constituted a separate cluster, whereas the two P. agglomerans pv. milletiae strains were widely divergent. Interestingly, although the corn pathogen P. stewartii subsp. stewartii was clearly separated from the majority of P. agglomerans strains, it shared approximately 67% similarity with P. agglomerans 299R (Fig. 1), which has been characterized as an efficient epiphyte (6). The distribution pattern obtained by PFGE was also confirmed by amplified fragment length polymorphism and repetitive extragenic palindromic PCR (results not shown).

Diversity of the PAI structure in the pPATH plasmids.

Previous reports indicated that the sizes of pPATH plasmids may differ even within a group (23, 36). However, the conservation of the PAI within or between the different groups has not been studied yet. Restriction fragment length polymorphism (RFLP) has been employed to examine the genetic diversity of the PAIs in representatives of the three major clusters, namely, groups S1 and S2 and P. agglomerans pv. betae (Fig. 1). The genomic DNA was digested with PstI and hybridized with PAI-specific probes targeting pthG, hrpS, and hopAK1 (Table 1). These probes are based on three genes of the characterized PAI in P. agglomerans pv. gypsophilae 824-1, and their locations on the PAI are separated by 25 to 40 kb (5). Figure 2 shows that the RFLP patterns obtained with each of the three probes were identical for the isolates in P. agglomerans pv. gypsophilae groups S1 and S2 but that they differed from those for the P. agglomerans pv. betae group. It is noteworthy that the intact gene coding for PthG was present in the PAIs of the members of the P. agglomerans pv. gypsophilae S1 and S2 groups, but only a partial gene was detected in the P. agglomerans pv. betae PAI (15). PthG was considered to be crucial for separation of pathogenic strains into P. agglomerans pv. gypsophilae and P. agglomerans pv. betae (4, 34).

FIG. 2.

Comparative RFLP patterns of PAIs obtained for the three major groups in the pathogenic P. agglomerans population. Three pathogenic isolates from each group (groups S1 and S2 and P. agglomerans pv. betae) were examined. The genomic DNA of each strain was digested with PstI and subjected to Southern analysis using three PAI-specific probes targeting pthG, hrpS, and hopAK1. The probes were selected from different regions of the PAI (see text for details). Lane M contained a λ BstEI DNA digest. Pab, P. agglomerans pv. betae; Pag(S2), P. agglomerans pv. gypsophilae group S2; Pag(S1), P. agglomerans pv. gypsophilae group S1.

The diversity of the PAIs was also compared by macrorestriction following fractionation by PFGE (Fig. 3A). Digestion with PspXI followed by hybridization with the hopAK1 probe resulted in identical band sizes for P. agglomerans pv. gypsophilae 824-1 (group S1) and P. agglomerans pv. gypsophilae 350-1 (group S2); in contrast, P. agglomerans pv. betae 4188 produced two bands that were close to each other (Fig. 3B). These results imply that the PAIs of the P. agglomerans pv. gypsophilae strains are different from that of P. agglomerans pv. betae. However, digestion with AvrII could distinguish between the PAIs of the two P. agglomerans pv. gypsophilae strains. Taken together, these results (Fig. 2 and 3B) suggest that the PAIs of the P. agglomerans pv. gypsophilae strains in the two clusters are highly conserved but different from that of the P. agglomerans pv. betae strains. Nevertheless, there may be some differences in restriction sites within the plasmid-borne PAIs of P. agglomerans pv. gypsophilae.

FIG. 3.

Comparative RFLP patterns of PAIs from strains P. agglomerans pv. gypsophilae 824-1 (S1), P. agglomerans pv. gypsophilae 350-1 (S2), and P. agglomerans pv. betae 4188 (P. agglomerans pv. betae group) (Pab) following macrorestriction. (A) Total DNA of each isolate was digested with either AvrII or PspXI and subjected to PFGE. Lane M contained a lambda concatemer. (B) Southern blot of the PFGE gel shown in panel A hybridized with the PAI-specific probe targeting hopAK1. (C) Southern blot of the PFGE gel shown in panel A hybridized with repA. Details are described in the text.

Replicon of the pPATH plasmids.

During sequencing of pPATH_Pag an open reading frame (ORF) showing high similarity to repA was discovered (5). Nucleotide sequences of the DNA replication region (2,170 bp) of the pPATH plasmids of P. agglomerans pv. gypsophilae 824-1, P. agglomerans pv. gypsophilae 350-1, and P. agglomerans pv. betae 4188, which represent the three major clusters (Fig. 1), were PCR amplified with primers Rep22950_5 and Rep25101_3 (see Table S1 in the supplemental material) and were cloned into the pGEM-T Easy vector to obtain pRep824, pRep350, and pRep4188, respectively (Table 1). The three cloned sequences showed 98% identity at both the nucleotide and amino acid levels, suggesting that the pPATH plasmids of the various P. agglomerans pv. gypsophilae and P. agglomerans pv. betae strains share the same replicon.

The RepA protein of the pPATH plasmids is composed of 228 amino acids and has a predicted molecular mass of 26.4 kDa (Fig. 3). This size falls within the range of sizes of plasmid Rep proteins, which is usually 25 to 40 kDa (13). Further analyses with the Predict Protein Server programs PHD and PHDsec (http://predictprotein.org/newwebsite/docs/methodsPP.html) revealed that the RepA protein contains three helix-turn-helix motifs between amino acid positions 8 and 73, 117 and 153, and 167 and 198, which suggests a potential DNA binding capability. Helix-turn-helix motifs have been identified in various RepA proteins, and they were found to be essential for RepA binding to iteron DNA sequences (13, 24). Although leucine zipper-like motifs have been detected in the N-terminal region of several Rep proteins (13, 19), no apparent leucine zipper motif could be detected in the RepA of the pPATH plasmids. Blast analysis of the RepA protein of the pPATH plasmids indicated that the highest scores were the scores with RepA of the IncN plasmids. Thus, a ClustalW multiple-sequence alignment (Fig. 4) revealed that RepA proteins of the pPATH plasmids shared 98% identity with each other and 70% identity with RepA proteins of the IncN plasmids pMUR050 and pCU1 (20, 28). These results suggest that the pPATH plasmids can be classified in the IncN incompatibility group.

FIG. 4.

ClustalW multiple-sequence alignment of RepA amino acid sequences obtained from the pPATH and IncN plasmids. The RepA protein sequences are in the following order (from top to bottom): P. agglomerans pv. gypsophilae 350-1 (accession number EF173386), P. agglomerans pv. betae 4188 (EF173388), P. agglomerans pv. gypsophilae 824-1 (EF173387), and RepA from IncN plasmids pMUR050 (YP_724520) and pCU1 (NP_040397). Black and gray backgrounds indicate conserved amino acids in all and some of the proteins, respectively. Asterisks indicate intervals of 10 amino acids.

To identify the minimal sequence that allows plasmid replication (base replicon), deletion analysis of the 2,170-bp replicon region of P. agglomerans pv. gypsophilae 824-1 was performed by generating a series of pRA plasmids. Figure 5 shows that pRA6 (929 bp) was the minimal sequence required for replication in E. coli and in nonpathogenic P. agglomerans strains when organisms were grown on LB with Km. This plasmid was shown to be stable when organisms were grown for 20 generations on LB without Km. In contrast, pRA6 could not maintain growth on Km when it was introduced into P. agglomerans pv. gypsophilae or P. agglomerans pv. betae harboring pPATH_Pag or pPATH_Pab. In addition to the repA ORF, the pRA6 plasmid includes a purine-rich DNA sequence corresponding to a putative Shine-Dalgarno sequence (−9 bp) and a putative DnaA binding box (−65 bp) upstream of the start codon. Although the sequence of oriV in the pPATH plasmids has not been characterized yet, this sequence is generally known to be located adjacent to repA (13) and presumably is present in pRA6 within 150 bp downstream of the repA ORF. Upstream or downstream deletions of the base replicon (e.g., pRA7 or pRA8) arrested plasmid replication. pRA1 could also replicate in P agglomerans pv. milletiae, P. stewartii, Erwinia amylovora, Erwinia carotovora subsp. carotovora, P. syringae pv. tomato DC3000, and Xanthomonas campestris pv. chrysanthemi (data not shown), indicating that it has a broad host range, like the broad host range that generally characterizes the IncN plasmids (29).

FIG. 5.

Analysis of the pPATH_Pag replicon. Details of the structure of the replicon region, as revealed by generating the pRA plasmid series, are described in the text. Bacterial growth was tested on LB agar plates containing Km. A plus sign indicates colony formation after 3 days. SL, stem loops. The solid bar indicates the repA ORF, and the cross-hatched bar indicates a repeat region. Pa, nonpathogenic P. agglomerans; Pag, P. agglomerans pv. gypsophilae; Pab and pab, P. agglomerans pv. betae.

Previous efforts to cure pPATH_Pag from P. agglomerans pv. gypsophilae 824-1 by conventional means, such as growth at elevated temperatures or in the presence of acridine orange, were unsuccessful (34). Furthermore, attempts to demonstrate conjugation by methods described in Materials and Methods were also unsuccessful. Characterization of the replicon region of pPATH allowed the possibility that it could be eliminated by introducing an artificially constructed plasmid containing an identical replicon and a suitable selective marker, such as an antibiotic resistance marker (11). Figure 5 shows that only pRA1 was capable of replicating in P. agglomerans pv. gypsophilae or P. agglomerans pv. betae strains during growth on LB in the presence of Km, whereas any pRA1 deletion (e.g., pRA2 to pRA5) prevented replication. To further demonstrate that the pPATH_Pag plasmid was removed from the transconjugant P. agglomerans pv. gypsophilae 8241(pRA1), this strain and P. agglomerans pv. gypsophilae 824-1(wild type) were subjected to PFGE followed by Southern hybridization with repA or hrpJ. hrpJ is an essential component of the plasmid-borne hrp gene cluster and has been used as a marker for the presence of the pPATH plasmids (44). Figure 6A demonstrates that although the two strains showed similar band patterns, only P. agglomerans pv. gypsophilae 824-1 (wild type) produced bands that could be assigned to the pPATH plasmid and that were absent from the cured strain P. agglomerans pv. gypsophilae 824-1(pRA1). Moreover, although bands hybridized with repA were observed with each of the two strains (Fig. 6B), only the wild-type strain yielded bands hybridized with hrpJ (Fig. 6C). Finally, gall formation on gypsophila cuttings could be observed following inoculation with P. agglomerans pv. gypsophilae 824-1 (wild type) but not following inoculation with P. agglomerans pv. gypsophilae 824-1(pRA1) (Fig. 6D). These results clearly indicate that pPATH_Pag was removed from P. agglomerans pv. gypsophilae 824-1 by pRA1 and that its removal eliminated pathogenicity. Similar to what occurred in P. agglomerans pv. gypsophilae 824-1, the pPATH plasmids in P. agglomerans pv. gypsophilae 350-1 and P. agglomerans pv. betae 4188 were also cured by pRA1 (results not shown). The observation that all of the pPATH plasmids could be cured by introducing the replication region may also imply that these plasmids carry identical single replicons.

FIG. 6.

Expulsion of the pPATH_Pag plasmid by manipulation of its replicon region. The pRA1 plasmid was mobilized into P. agglomerans pv. gypsophilae 824-1 (wild type) to obtain the transconjugant P. agglomerans pv. gypsophilae 824-1(pRA1) as described in the text. (A) PFGE banding patterns of XbaI-digested total cellular DNA from P. agglomerans pv. gypsophilae 824-1 (wild type) and its isogenic transconjugant P. agglomerans pv. gypsophilae 824-1(pRA1). Lane a, lambda concatemer; lane b, P. agglomerans pv. gypsophilae 824-1 (wild type); lane c, P. agglomerans pv. gypsophilae 824-1(pRA1). The arrows indicate extra bands of the pPATH_Pag plasmid in P. agglomerans pv. gypsophilae 824-1 (wild type) that are absent in P. agglomerans pv. gypsophilae 824-1(pRA1). (B) Southern hybridization of the PFGE gel shown in panel A with the repA probe. (C) Southern hybridization of the PFGE gel shown in panel A with the hrpJ probe (indicating the presence of pPATH_Pag). Note the absence of hrpJ from P. agglomerans pv. gypsophilae 824-1(pRA1). (D) Pathogenicity on gypsophila cuttings inoculated with (cutting a) P. agglomerans pv. gypsophilae 824-1 (wild type), (cutting b) P. agglomerans pv. gypsophilae 824-1(pRA1), and (cutting c) a water control.

Since the curing ability of pRA1 appears to depend on sequences in addition to the base replicon, we hypothesized that they should be involved in the regulation of replication. Analysis of the additional sequence (271 bp) of pRA1 immediately upstream of the promoter region of pRA6 (Fig. 5) revealed five putative DNA direct repeats (12 bp) that could form five stem-loop structures. Clusters of DNA direct repeats (namely, nine 20-bp repeats, eight 22-bp repeats, and five 30-bp repeats) were also present in the additional sequence (970 bp) in pRA1 downstream of pRA6 (Fig. 5).

It has been previously observed that the replicon of pPATH_Pag is located in a DNA region (>50 kb) which is separated from the PAI and lacks IS elements (5). The diversity of the replicon region was examined by RFLP analysis with repA as the probe. Significant variation in RFLP was observed in each of the three clusters when the DNA was digested with AvrII and PspXI (Fig. 3C). This is in contrast to the conserved region of the PAIs (Fig. 2 and 3B).

DISCUSSION

Analysis of P. agglomerans strains by PFGE clearly suggested that the pPATH plasmids were distributed in genetically diverse populations. Three major groups, P. agglomerans pv. gypsophilae group S1, P. agglomerans pv. gypsophilae group S2, and P. agglomerans pv. betae, were identified. In light of the RFLP analysis, the PAIs of groups S1 and S2 appeared to be highly conserved and to differ from the PAI of P. agglomerans pv. betae. Nevertheless, all the PAIs are related, as shown by their hybridization with the same PAI-specific genes (Fig. 2) and also by the presence of an identical hrp gene cluster, type III effectors, and IS elements (4, 5, 35). The pPATH_Pag plasmid of P. agglomerans pv. gypsophilae 824-1, which was extensively studied, contains, in addition to the PAI, a region harboring several genes involved in replication and maintenance of the plasmid (5). The replicon, which is located in this region, has been identified as IncN, and it was found to be identical in the pPATH plasmids of the different groups (Fig. 4). In addition to the RepA gene, other genes in this region include genes encoding ArdK, a homolog of the antirestriction protein encoded in the IncN plasmid R46 (accession number AAL13415; 56% identity); Umuc, a homolog of the DNA repair protein of Desulfovibrio vulgaris (accession number YP_012118; 73% identity); XerDC, a homolog of a site-specific recombinase belonging to the phage integrase family from P. syringae pv. tomato DC3000 (accession number NP_808665; 49% identity); HP3, a hypothetical protein homolog with a conserved antirestriction domain involved in DNA replication, recombination, and repair from Yersinia tuberculosis (accession number CAF28565; 55% identity); a hypothetical protein with a UvrD helicase conserved domain from E. carotovora subsp. atroseptica (accession number YP_048684; 66% identity); and a putative VirB4 protein of the type IV secretion pathway from E. carotovora subsp. atroseptica (accession number YP_048689; 68% identity) (5).

In contrast to the RFLP of the PAIs, which showed relative conservation among the three groups, the region of the replication and maintenance genes showed high diversity in its restriction sites (Fig. 3). In light of the differences in RFLPs we hypothesize that these two regions might have evolved independently.

The presence of the DNA repeats in the replicon region is typical of iteron plasmids such as pSC101, R6K, and RP4 (30) and strongly suggests that they may act as iteron sequences that play a regulatory role in the replication of the pPATH plasmids (30). The coupling model for regulation of iteron plasmids suggests that RepA can bind to iteron sequences of more than one plasmid at the same time, effectively coupling the plasmids or handcuffing them together (38). Higher levels of RepA proteins increase the amount of coupling between plasmids and inhibit their replication. Moreover, even the presence of iterons on an unrelated plasmid can cause coupling between the two plasmids, lowering the copy numbers of both. We hypothesize that the introduction of pRA1 into P. agglomerans pv. gypsophilae or P. agglomerans pv. betae substantially reduces the copy number of the two plasmids via binding of the RepA protein to the iterons of pPATH and pRA1. However, the selective advantage of pRA1 over pPATH, which might be expressed in its resistance to Km and its considerably smaller size, resulted in curing of the pPATH plasmid from P. agglomerans pv. gypsophilae 824-1, P. agglomerans pv. gypsophilae 350-1, and P. agglomerans pv. betae 4188, rendering these strains nonpathogenic. An additional regulatory mechanism of iteron plasmids is based on transcriptional autoregulation (52). It involves binding of RepA to its own promoter region and blocking of transcription of its own gene. Whether the RepA protein can bind to the stem-loop region upstream of pRA6 and thus inhibit its own transcription or act via the coupling model remains to be investigated. It is, however, apparent that this region is crucial for curing the pPATH plasmids with pRA1.

It is reasonable to hypothesize that an ancestor plasmid containing the replicon region, with adjacent regulatory and maintenance genes, provided the framework for the initial evolution of the pPATH plasmid and for its mobilization into P. agglomerans strains. Our attempts to detect the replicon among nonpathogenic P. agglomerans strains by PCR amplification with primers Rep22950_5 and Rep25101_3 (see Table S1 in the supplemental material) so far have been unsuccessful, which suggests that the replicon might have originated in an indigenous ancestral plasmid present in other bacteria. We further hypothesize that the initial evolution of the pathogenicity plasmid occurred by sequential acquisition of clustered and/or isolated virulence genes, forming the PAI in the ancestral pPATH. IncN plasmids are generally known to be conjugative and to have a broad host range (29). However, our efforts to demonstrate conjugal transfer of P. agglomerans pv. gypsophilae 824-1 into nonpathogenic strains under in vitro and in vivo conditions were unsuccessful. The nonconjugative nature of the pPATH_Pag plasmid was also supported by the lack of putative genes that showed significant homology to tra genes (5).

The observations described above led us to suggest that an ancestral pPATH plasmid was introduced into P. agglomerans strains as either a conjugative or a mobilizable plasmid (48). In the former case, it lost its conjugative ability following unidirectional mobilization, whereas in both cases the presence of an as-yet-unidentified oriT region is expected. Interestingly, we could not find remnants of genes associated with DNA transfer in pPATH_Pag, with the possible exception of the gene encoding VirB4, which belongs to the type IV secretion system and is implicated in transfer of the transferred DNA (12). This observation could favor the hypothesis that the ancestral pPATH plasmid was transferred as a mobilizing plasmid rather than as a conjugative plasmid. It can be speculated that the pPATH plasmid was mobilized directly into a diverse population of P. agglomerans strains. Additionally, further evolution of the PAIs has continued in P. agglomerans, leading to separation into two pathovars, P. agglomerans pv. gypsophilae and P. agglomerans pv. betae (4, 34). Another unresolved question is why the pPATH plasmid, with its broad-host-range replicon, was introduced into P. agglomerans, transforming it into a gall-forming pathogen, rather than into other plant-associated bacteria. We speculate that the efficient endophytic/epiphytic traits of P. agglomerans, as well as its high genetic plasticity, have provided a system for expression of pPATH more favorable than that in other bacteria. Additionally, the successful interaction between the plasmid-borne hrp regulon and the chromosomally encoded global regulatory systems (8) could be another important factor favoring the preferred establishment of the pPATH plasmids in P. agglomerans.