A Conjugative 38 kB Plasmid Is Present in Multiple Subspecies of Xylella fastidiosa

A ∼38kB plasmid (pXF-RIV5) was present in the Riv5 strain of Xylella fastidiosa subsp. multiplex isolated from ornamental plum in southern California. The complete nucleotide sequence of pXF-RIV5 is almost identical to that of pXFAS01 from X. fastidiosa subsp. fastidiosa strain M23; the two plasmids vary at only 6 nucleotide positions. BLAST searches and phylogenetic analyses indicate pXF-RIV5 and pXFAS01 share some similarity to chromosomal and plasmid (pXF51) sequences of X. fastidiosa subsp. pauca strain 9a5c and more distant similarity to plasmids from a wide variety of bacteria. Both pXF-RIV5 and pXFAS01 encode homologues of a complete Type IV secretion system involved in conjugation and DNA transfer among bacteria. Mating pair formation proteins (Trb) from Yersinia pseudotuberculosis IP31758 are the mostly closely related non-X. fastidiosa proteins to most of the Trb proteins encoded by pXF-RIV5 and pXFAS01. Unlike many bacterial conjugative plasmids, pXF-RIV5 and pXFAS01 do not carry homologues of known accessory modules that confer selective advantage on host bacteria. However, both plasmids encode seven hypothetical proteins of unknown function and possess a small transposon-associated region encoding a putative transposase and associated factor. Vegetative replication of pXF-RIV5 and pXFAS01 appears to be under control of RepA protein and both plasmids have an origin of DNA replication (oriV) similar to that of pRP4 and pR751 from Escherichia coli. In contrast, conjugative plasmids commonly encode TrfA and have an oriV similar to those found in IncP-1 incompatibility group plasmids. The presence of nearly identical plasmids in single strains from two distinct subspecies of X. fastidiosa is indicative of recent horizontal transfer, probably subsequent to the introduction of subspecies fastidiosa to the United States in the late 19th century.


Introduction
Horizontal gene transfer plays a critical role in bacterial adaptation and evolution. On average, 81% of the genes in a typical bacterial genome have been involved in a horizontal transfer event at some point in the past [1]. One of the most common mechanisms for DNA exchange is via conjugative plasmids that encode type IV secretion systems (T4SS), a broad class of macromolecular translocation machinery. There are three main types of T4SS: i) conjugation systems that transfer DNA and, in some instances, DNA-binding proteins; ii) effector translocator systems that deliver proteins and other effectors to eukaryotic cells during bacterial infection of eukaryotic hosts; and iii) DNA uptake or release systems that move DNA between the interior of the cell and the extracellular environment [2]. Conjugation systems are found both on self-transmissible or conjugative plasmids, circular DNA molecules that replicate independently of the bacterial chromosome, and on integrative and conjugative elements that integrate into the chromosome, excising and forming a circular intermediate prior to translocation [2]. One well characterized conjugation system is VirB/D4 from Agrobacterium tumefaciens, which is composed of a cell-envelope spanning secretion channel and an extracellular pilus that contacts the recipient cell [3]. For VirB/D4, the translocation system consists of VirB2-11 proteins forming the secretion channel, the VirD4 substrate receptor or type IV coupling protein (T4CP), and proteins for pilus formation and DNA substrate processing [2]. Many conjugative plasmids also contain accessory modules encoding cargo proteins which act as virulence factors, confer resistance to antibiotics/heavy metals, or catabolize toxic organic substances [4].
Xylella fastidiosa is a fastidious, xylem-limited Gram-negative bacterial phytopathogen causing numerous vascular occlusion and water stress diseases including Pierce's disease of grape, almond leaf scorch, oleander leaf scorch, and other diseases of perennial crops and landscape plants [5]. Four subspecies of X. fastidiosa have been identified based on a multi-locus sequence typing (MLST) phylogeny [6,7]. Subspecies fastidiosa contains strains of low genetic diversity that cause Pierce's disease and sometimes almond leaf scorch in the U. S. and diverse strains from Central America; subsp. fastidiosa is thought to have been introduced to the U. S. in the late 19 th century [8]. Subspecies multiplex is an endemic North American clade capable of infecting numerous hosts (but generally not grapevine) [9][10][11]. Subspecies pauca contains South American strains causing citrus variegated chlorosis and coffee leaf scorch [12]. Subspecies sandyi consists of closely related strains isolated from oleander in California and Texas and is thought to have been introduced to the United States approximately 30 years ago [6]. Currently, 5 fully sequenced and annotated X. fastidiosa genomes are available (pauca strain 9a5c [13], fastidiosa strain Temecula [14], multiplex strain M12 [15], fastidiosa strain M23 [15], and fastidiosa strain GB514 [16]). Two additional sequences (multiplex strain Dixon, GenBank accession number NZ_AAAL00000000.2, and sandyi strain Ann-1, GenBank accession number NZ_AAA M00000000.3) are incomplete, unassembled shotgun sequences; the Ann-1 sequence may have been derived from a mixed culture and appears to be contaminated with sequences from a multiplex strain [17].
Here, we characterize two closely related 38kB conjugative plasmids of X. fastidiosa. Plasmid pXF-RIV5 was isolated from the Riv5 strain of X. fastidiosa subspecies multiplex [18]; complete sequence of pXF-RIV5 was determined in this current work. Plasmid pXFAS01 is known only as a circular contig discovered during the complete genome sequencing of X. fastidiosa subspecies fastidiosa strain M23 [15]. While minimal annotation of pXFAS01 accompanies the sequence in GenBank, no analysis of the type IV secretion system, origin of transfer or origin of replication from pXFAS01 has been presented previously. This work presents analyses of gene complement and phylogeny of these two closely related plasmids to reveal i) an evolutionary history of recombination among divergent sources to generate the mosaic backbone shared by pXF-RIV5 and pXFAS01, and ii) evidence of recent translocation of plasmid DNA via conjugation among distinct subspecies of X. fastidiosa.

Materials and Methods
Culture and MLST of X. fastidiosa subspecies multiplex strain Riv5 Isolation of X. fastidiosa strain Riv5 from ornamental plum (Prunus cerasifera) was described previously [18]. Strain Riv5 cultures were grown in liquid periwinkle wilt (PW) medium for 7-10 days at 28C and used to inoculate plates containing solid PW medium. After 7-10 days of growth at 28C, bacterial colonies were washed from 10 PW plates and extracted for total genomic DNA [11]. Genomic DNA was used as template for PCR amplification of seven housekeeping genes (cysG, gltT, holC, malF, leuA, nuoL, petC) and pilU [7]; consensus sequences were determined for each amplified region based on sequences of three independent clones per PCR product. Consensus sequences for each amplified region were concatenated into a single sequence. MLST was performed as described [7] with concatenated Riv5 sequences aligned with the corresponding concatenated sequences from multiple strains representative of each X. fastidiosa subspecies available in GenBank using CLUSTALX. Phylogenetic placement of strain Riv5 was determined based on a neighbor-joining tree (1000 bootstrap replications) using the multiple alignment of concatenated MLST sequences as input data.

Plasmid DNA isolation and sequencing
Previously, strain Riv5 was shown to harbor a large plasmid (designated here as pXF-RIV5), yielding multiple products when digested with HindIII [18]. Purification of pXF-RIV5 DNA was as described [18] from Riv5 cultures grown under the same regime as that used to extract genomic DNA. Purified plasmid DNA was digested with HindIII; each resulting fragment was gel purified and ligated into HindIII digested pGEM7Zf+ (Promega, Madison, WI). Ligation products were transformed into Escherichia coli JM109; recombinant plasmids bearing HindIII inserts of pXF-RIV5 were sequenced using a combination of universal (M13 forward and reverse) and custom primers. As preliminary sequences of insert termini obtained with universal primers indicated a very close relationship with pXFAS01, custom primers were based on the known sequence of pXFAS01 (GenBank Accession NC_010579.1) associated with the X. fastidiosa subspecies fastidiosa strain M23 genome sequence [15]. The complete nucleotide sequence of pXF-RIV5 (GenBank Accession JX548317) was assembled using the sequence of pXFAS01 as a scaffold.

DNA sequence analysis and annotations
Open reading frames were identified based on similarity to pXFAS01 and using ORF Finder (http://www.ncbi.nlm.nih.gov/ gorf/gorf.html). Annotations are derived from similar sequences identified using BLAST [19] and NCBI conserved domain searches [20]. The origin of conjugative transfer (oriT) was identified by similarity to the oriT regions of pRP4 and pR751 [21]. The tandem repeats in oriV were found using Repfind (http://zlab.bu.edu/repfind/index.shtml). A map of pXF-RIV5 was drawn using GENtle (http://gentle.magnusmanske.de/) open software package.

Phylogenetic analysis of pXF-RIV5
Phylogeny of three proteins encoded by pXF-RIV5, representing three distinct genetic modules resident on pXF-RIV5 (and pXFAS01), were selected for examination. RepA represents a protein involved in DNA replication; TraI is a relaxase homologue representing the tra module (conjugative transfer functions); and TrbG is predicted to be an outer membrane protein and represents the trb module (mating pair formation functions).
Taxa selected for inclusion were based on results of BLAST P searches of the GenBank nonredundant protein database using the corresponding homologues of RepA, TraI and TrbG encoded by pXF-RIV5 as queries. A neighbor-joining tree (1000 bootstrap replicates) for each protein was constructed based on a multiple alignment of amino acid sequences generated using CLUSTALX. Nodes bearing ,70% bootstrap support were considered unreliable and collapsed to polytomies.

Riv5 is a strain of subspecies multiplex
Genomic DNA sequences used for MLST of strain Riv5 were deposited as GenBank Accessions JX679700-JX67907. Phylogeny of concatenated sequences of the eight genes examined by MLST indicated that strain Riv5 clustered with X. fastidiosa strains of subspecies multiplex, including the fully-sequenced multiplex strain M12 (data not shown). Phylogeny of each individual gene used for MLST also clustered strain Riv5 with strains of subspecies multiplex (data not shown). These results are consistent with phylogenetic placement of strain Riv5 based on 16S-23S rRNA spacer sequences [18].

pXF-RIV5 contains a type IV secretion system
The complete nucleotide sequence of pXF-RIV5 is 38,297 bp in length with a G+C content of 49.2%, which is similar to the 51-52% G+C content of sequenced X. fastidiosa genomes [13,15]. As detailed in Table 1 and Figure 1, pXF-RIV5 has 33 ORFs encoding proteins similar to characterized proteins of known function from other organisms while seven ORFs encode hypothetical proteins for which functions of homologues identified in GenBank are unknown. The two largest groups of genes are the conjugative transfer (tra) and mating pair formation (trb) modules. Together, these two genetic modules encode homologues of all proteins necessary for a functional T4SS [2]. Genes for plasmid replication (repA, kleE, and ssBP) and partition (parA and parB) also are present. Two genes, orfA and orfB, encode proteins similar to transposon-associated recombinase and transposase, respectively; no other transposon-like elements on pXF-RIV5 were identified. Unlike many conjugative plasmids from animal pathogens or Relationship of pXF-RIV5 to other X. fastidiosa plasmids The sequence of pXF-RIV5 is almost identical to pXFAS01 from X. fastidiosa subsp. fastidiosa strain M23 [15]. For consistency, nucleotide coordinates of pXF-RIV5 were assigned to correspond to nucleotide coordinates designated for pXFAS01. Alignment of pXF-RIV5 and pXFAS01 revealed polymorphism, all of which are transitions, at only six nucleotide positions over the entire ,38 kB length. Two transitions are located in intergenic regions at nt 9,556 (between trbN and ORF11) and at nt 36,281 (between ssBP and trbA). The other four transitions are located in the traI gene encoding a conjugative relaxase homologue. Two transitions (nts 24,789 and 24,921) in the traI gene were synonymous substitutions that did not alter predicted protein sequence. The remaining two transitions in the traI gene were nonsynonymous substitutions that altered the codon for amino acid 462 (nt 25,019) from proline (pXFAS01) to serine (pXF-RIV5) or altered the codon for amino acid 386 (nt 25,250) from serine (pXFAS01) to proline (pXF-RIV5). Although both nonsynonymous substitutions are not in highly conserved portions of TraI, potential alteration of function cannot be excluded. Clustering of four out of six polymorphic sites in less than 500 bp of a 38 kB plasmid raises the possibility that all substitutions in traI were introduced by a single recombination event between pXF-RIV5 or pXFAS01 and a closely related plasmid. The plasmid (pXF51) of X. fastidiosa subsp. pauca strain 9a5c encodes a partial trb module [22] which is 96% identical at the nucleotide level over almost 9 kB of pXF-RIV5 and pXFAS01, spanning trbE through trbN (nts 481 -nts 9401). Strain 9a5c also has an extensive cluster of trb genes on the chromosome [22] sharing 97% nucleotide sequence identity with ,10 kB of pXF-RIV5 and pXFAS01 (nts 37612 -nts 9417). A small region of pXF-RIV5 and pXFAS01 (140 bp; nts 9276 -nts 9414) shares 89% nucleotide sequence identity to 25 kB IncP-1 plasmids (pXF-RIV11, pXF-RIV16, pXF-RIV19, and pXF-RIV25) from mulberry-infecting strains of X. fastidiosa [18]. In pXF-RIV5 and pXFAS01, the homologous region is within the trbN gene; however, in the IncP-1 plasmids, the complete trbN gene is not present.

DNA replication elements and T4SS components of pXF-RIV5 have distinct evolutionary histories
Most proteins encoded by pXF-RIV5 and pXFAS01 are homologues of proteins encoded by numerous bacterial taxa. As shown in Figure 2, neighbor-joining phylogenetic trees were constructed for representative proteins (TraI and TrbG) from the two T4SS modules and for the replication protein RepA. The trees for TraI and TrbG have similar but not identical topology. The most closely related homologues (excluding those from X. fastidiosa) for both TraI and TrbG are from Yersinia pseudotuberculosis IP31758 and E. coli PA14. Because Tra and Trb proteins must work together to form a functional T4SS, it is not surprising that both T4SS gene clusters have similar phylogeny.
The phylogenetic history inferred for RepA is quite different from that of the T4SS (Figure 2). Most RepA homologues from X. fastidiosa constituted a clade distinct from RepA homologues encoded by all other taxa identified in BLAST P searches. RepA from X. fastidiosa strains Dixon, EB92.1 and Ann-1* were either identical, or nearly identical to that encoded by pXF-RIV5 and pXFAS01. Dixon and Riv5 are subspecies multiplex strains whereas M23 (host of pXFAS01) and EB92.1 are strains of subspecies fastidiosa. It is noted that two distinct RepA sequences are associated with the Ann-1 genome. One RepA sequence from Ann-1 (designated Ann-1*) is identical to that of pXF-RIV5, pXFAS01, and strain Dixon. The second RepA sequence from Ann-1 (designated Ann-1) shared only 53.2% amino acid sequence identity with RepA of pXF-RIV5 and pXFAS01 and clustered in a different clade with RepA from pRSC35 of Ralstonia solanacearum CMR15 as the most closely related homologue identified. As the Ann-1 strain genome sequence is known to be contaminated with a subspecies multiplex genome sequence [17], the simplest interpretation of these results is that RepA Ann-1* represents the multiplex contaminant and that RepA Ann-1 is the divergent homologue resident in the ''true'' Ann-1 genome of subspecies sandyi. Confirmation of this interpretation will require sequencing of genomes of additional strains of subspecies sandyi.
The presumptive origin of transfer (oriT) of pXF-RIV5 and pXFAS01 (nts 26869-26921) are similar to the experimentally verified oriT [21] from two E. coli plasmids (pRP4 and pR751). The stem-loop inverted repeat structure that forms the TraK binding site of pRP4 oriT shares 89% nucleotide sequence identity (56 of 63 nts identical) with the oriT homologue of pXF-RIV5 and pXFAS01 ( Figure 3A). Two nucleotide substitutions in the oriT sequences between the X. fastidiosa plasmids and pRP4 are compensatory changes in the stem, preserving secondary structure ( Figure 3A); the other five substitutions are in loop regions. The oriT region from pR751 shares less nucleotide sequence identity (79.3%; 50 of 63 bp) with pXF-RIV5 and pXF-AS01 and has one additional stem base pair relative to pXF-RIV5, pXFAS01 and pRP4.
The presumptive vegetative origin of replication (oriV) of pXF-RIV5 and pXFAS01 (nts 34039-34504) consists of an approximately 300 bp A+T rich region followed by four tandem repeats ( Figure 3B). While the combination of an A+T rich region followed by tandem repeats is present in oriV of many plasmids, more common arrangements have a larger number of repeats (16 to 20) with the 19 nt core of the repeat forming a perfect 10 base pair palindrome [23]. The 19 nt core repeat sequence in pXF-RIV5 is a partial palindrome with only 6 base pairs possible.

Discussion
As with many bacteria, X. fastidiosa harbors a variety of plasmids. Several have been characterized, including an IncP-1 plasmid [18], a small rolling-circle replicon [24], and pXF51 [22]. Figure 2. Phylogeny of RepA, TraI, and TrbG proteins encoded by pXF-RIV5. Neighbor-joining trees (1000 bootstrap replicates) presented are based on a multiple alignment of amino acid sequences. Bootstrap support values are shown for nodes with values .70%; nodes bearing #70% bootstrap support were collapsed to polytomies. Scale bar at lower left of each tree indicates a genetic distance of 0.1. Taxa names beginning with ''p'' are known to be encoded by plasmids, with plasmid designation preceding bacterial host nomenclature; replicon (plasmid versus chromosome) of other taxa is not specified. Taxa in red are encoded by strains of Xylella fastidiosa; multiple taxa assigned to a single branch are 100% identical. X. fastidiosa Ann-1 sequences identical to subspecies multiplex homologues are designated with an asterisk (Ann-1*) and may be derived from sequences contaminating the available Ann-1 strain genome sequence. Protein ID numbers are shown in Table S1. doi:10.1371/journal.pone.0052131.g002 Figure 3. Structure of oriT and oriV. A) oriT inverted repeats from pXF-RIV5, pRP4, and pR751 are shown as stem-loop structures to emphasize potential base parings. Unique bases are shown in red; bases conserved in at least two structures are shown in black. Red arrows indicate experimentally determined cleavage sites in pRP4 and pR751 [21]. B) Tandem repeats in pXF-RIV5 oriV are aligned and the 19-bp core is shown in red. Residues that vary from the consensus are shown in gray. Nucleotides shown (34333 -34503) are contiguous. doi:10.1371/journal.pone.0052131.g003 Here, we have described distinct attributes of pXF-RIV5 and pXFAS01, resident in strains representing multiple subspecies of X. fastidiosa. Given the probable conjugative abilities of these plasmids, it is possible that similar plasmids may be found in other as yet uncharacterized strains of X. fastidiosa or even in other bacterial species that share ecological niches with X. fastidiosa.
While many broad host range conjugative plasmids belong to the IncP-I incompatibility group and contain trfA homologues for vegetative replication [2,25], replication of pXF-RIV5 and pXFAS01 uses a RepA-dependent process. Therefore, pXF-RIV5 and pXFAS01 are not readily assigned to a classic incompatibility group. Nonetheless, tra and trb genetic modules of pXF-RIV5 and pXFAS01 do cluster ( Figure 2) with a subgroup typified by E. coli pRP4 and other IncPa group plasmids [2,4]. It is likely that an ancestral recombination event occurred between an IncPa group plasmid similar to pRP4 and a plasmid with RepAdependent vegetative replication to create the plasmid backbone (e.g., modules controlling replication and conjugative transfer) found in pXF-RIV5 and pXFAS01.
In addition to backbone genetic modules, many conjugative plasmids contain accessory modules encoding host-beneficial functions [26]. These accessory modules are often located at the ends of the tra and trb modules and/or near transposon or resolvase genes. No accessory modules with identifiable function were encoded by pXF-RIV5 and pXFAS01, nor by any other characterized plasmid of X. fastidiosa. Interestingly, ORFs for five of seven hypothetical proteins encoded by pXF-RIV5 and pXFAS01 are located downstream of both tra and trb modules, and in close proximity to a resolvase homologue (Figure 1). Whether these ORFs of unknown function constitute an accessory module conferring selective advantage to X. fastidiosa remains to be determined.
pXF-RIV5 and pXFAS01 are the only characterized plasmids of X. fastidiosa encoding all known factors (e.g., a complete T4SS) required for transfer of DNA from recipient to donor cells via conjugation. DNA transfer among strains/subspecies of X. fastidiosa has occurred, as evidenced by massive introgression events leading to the origin of mulberry-infecting [27] and citrus/ coffee-infecting [28] strains of X. fastidiosa. In these cases, the mechanism of DNA transfer leading to homologous recombination appears to be transformation, as lengths of recombinant regions were generally small, albeit numerous. Indeed, recent evidence suggests that X. fastidiosa is naturally competent for acquisition of foreign DNA with intrinsic transformation efficiency higher than that of many other bacterial species [29]. Thus, the identification of pXF-RIV5 and pXFAS01, bearing all known hallmarks of a conjugative plasmid, suggests that X. fastidiosa, a plant pathogen of significant economic concern, also may transfer large segments of DNA via conjugation. Indeed, the presence of almost identical plasmids in two separate subspecies of X. fastidiosa (pXF-RIV5 in multiplex and pXFAS01 in fastidiosa) implies a recent inter-subspecies translocation event. Subspecies multiplex is relatively diverse and, therefore, likely has been present in the U. S. for a considerable time; subspecies fastidiosa in the U. S. exhibits limited genetic diversity. It has been hypothesized that all strains of subspecies fastidiosa in the U. S. are derived from a single introduction from Central America that occurred circa 1880 [8]. If so, the inter-subspecies plasmid translocation event responsible for host associations of pXF-RIV5 and pXFAS01 occurred more recently than 1880. Collectively, these observations suggest that the introduction of exotic subspecies of X. fastidiosa further complicates disease management, as newly introduced X. fastidiosa subspecies not only may cause disease(s) previously not known to occur in a region, they also provide a wealth of genetic diversity to be shared with endemic subspecies.

Supporting Information
Table S1 GenBank protein ID numbers for all proteins appearing in phylogenetic trees ( Figure 2). (DOCX)