Nucleotide sequence, functional characterization and evolution of pFKN, a virulence plasmid in Pseudomonas syringae pathovar maculicola



Pseudomonas syringae pv. maculicola strain M6 (Psm M6) carries the avrRpm1 gene, encoding a type III effector, on a 40 kb plasmid, pFKN. We hypothesized that this plasmid might carry additional genes required for pathogenesis on plants. We report the sequence and features of pFKN. In addition to avrRpm1, pFKN carries an allele of another type III effector, termed avrPphE, and a gene of unknown function (ORF8), expression of which is induced in planta, suggesting a role in the plant–pathogen interaction. The region of pFKN carrying avrRpm1, avrPphE and ORF8 exhibits several features of pathogenicity islands (PAIs). Curing of pFKN (creating Psm M6C) caused a significant reduction in virulence on Arabidopsis leaves. However, complementation studies using Psm M6C demonstrated an obvious virulence function only for avrRpm1. pFKN can integrate and excise from the chromosome of Psm M6 at low frequency via homologous recombination between identical sequence segments located on the chromosome and on pFKN. These segments are part of two nearly identical transposons carrying avrPphE. The avrPphE transposon was also detected in other strains of P. s. pv. maculicola and in P. s. tomato strain DC3000. The avrPphE transposon was found inserted at different loci in different strains. The analysis of sequences surrounding the avrPphE transposon insertion site in the chromosome of Psm M6 indicates that pFKN integrates into a PAI that encodes type III effectors. The integration of pFKN into this chromosomal region may therefore be seen as an evolutionary process determining the formation of a new PAI in the chromosome of Psm M6.


Pathogens require an arsenal of pathogenicity factors to infect a host. Genes important for pathogenicity are often grouped in regions of the chromosome or on plasmids. These regions are referred to as pathogenicity islands (PAIs) (for a review, see Hacker and Kaper, 2000). PAIs typically carry virulence genes and sequences related to  DNA mobility, such as genes encoding integrases or transposases, and insertion sequences (IS). It is very likely that PAIs are acquired through horizontal transfer (Hacker and Kaper, 1999). This is suggested by the fact that their G+C content often differs from that of the core genome. Moreover, in many cases, PAIs are flanked by tRNA genes. Some tRNA loci are known to be integration targets for DNA of bacteriophages or plasmids (Hou, 1999). The VPI (Vibrio pathogenicity island), for example, results from the integration of the bacteriophage VPIΦ (Karaolis et al., 1999) into the chromosome of Vibrio cholerae. In other cases, the PAI seems to have arisen from the integration of a plasmid. Indeed, Rajakumar et al. (1997) proposed that a 99 kb chromosomal element in Shigella flexneri 2a YSH6000 carrying a multiantibiotic resistance locus arose after the integration of an NR1-like plasmid.

The mosaic-like structure of numerous PAIs suggests that they are generated by a multistep process. This process may involve various mobile elements and multiple genomic rearrangements. For example, the SPI2 (Salmonella PAI 2) PAI contains separate 25 kb and 14.5 kb regions that differ from each other in their G+C content, their codon usage, the type of functions they encode and their respective distribution in other strains of Salmonella (Hensel et al., 1999). Similarly, the comparison of five locus of enterocyte effacement (LEE) PAIs from three rabbit-specific strains of enteropathogenic Escherichia coli (REPEC 83/39, 84/110-1 and RDEC-1) and human enteropathogenic or enterohaemorrhagic E. coli strains (E2348/69 and EDL933) shows that the five PAIs originated from the same source (Tauschek et al., 2002). However, each of them has evolved and diverged from the PAIs of the other strains through DNA insertions and potential deletions (Tauschek et al., 2002). On some occasions, a PAI may excise from the chromosome. The mechanism of excision is sometimes the same as the mechanism of integration. For example, an integrase encoded on the HPI (high pathogenicity island) of Yersinia pseudotuberculosis and Yersinia pestis is responsible for the integration of this PAI into its attachment site on the chromosome (Rakin et al., 2001). Rakin and coworkers showed that, when this integrase is expressed, the HPI could be excised.

In the plant pathogen Pseudomonas syringae, several PAIs have been identified (Alarcon-Chaidez et al., 1999; Jackson et al., 1999; 2000; Alfano et al., 2000; Lu et al., 2002). They carry genes encoding the type III secretion system apparatus (called the hrp cluster), type III effector proteins that serve as virulence factors or proteins catalysing the production of toxins. Thus, PAIs carry genes representing the major components of P. syringae pathogenicity. The hrp cluster is necessary for pathogenicity (Lindgren et al., 1986; 1988) and encodes a type III pilus (Jin and He, 2001).

The P. syringae type III secretion system delivers type III effector proteins to the plant cell (Cornelis and Van Gijsegem, 2000; Jin and He, 2001; Li et al., 2002). Although some type III effectors have demonstrable roles in virulence, very little is known about their host targets (Nimchuk et al., 2001). Target processes affected by P. syringae type III effectors were described recently (Mackey et al., 2002; Van der Hoorn et al., 2002). Some type III effectors are recognized inside the plant cell by the plant R gene system, and these trigger a disease resistance-associated programmed cell death known as the hypersensitive response (HR) (Dangl and Jones, 2001). Because the recognition of some type III effector proteins causes the plant to be resistant to a particular pathogen strain, most of the genes encoding P syringae type III effectors were originally named avirulence (avr) genes.

The P. syringae hrp PAI also encodes the HrpL alternative sigma factor. This factor regulates the expression of both the genes encoding the type III delivery pilus and those encoding type III effectors (Xiao et al., 1994). The promoters of these genes contain a consensus hrp box (Innes et al., 1993). Hence, many genes encoding putative type III effectors have been identified recently using sequence analysis methods on the whole genome of the Pst DC3000 strain and sequences of DNA fragments from other pathovars (Jackson et al., 1999; Alfano et al., 2000; Arnold et al., 2001; Boch et al., 2002; Fouts et al., 2002; Petnicki-Ocwieja et al., 2002).

The type III effector avrRpm1 (Ritter and Dangl, 1995) contributes to virulence on Arabidopsis plants that cannot recognize it (host genotype rpm1). It is encoded on a plasmid in several P. syringae pv. maculicola (Psm) strains, hereafter named pFKN. Several P. syringae plasmids have been partially or fully characterized. They often carry genes encoding virulence factors or functions enabling the bacteria to survive in given environments (for a review, see Vivian et al., 2001). We describe the analysis of the pFKN sequence and suggest that it carries at least one PAI. We investigated the mechanism by which the plasmid integrates and excises into and out of the chromosome of Psm M6, as well as the chromosomal region involved in the process. We also found that a piece of the pFKN mosaic structure is a transposon carrying an allele of a second type III effector, avrPphE. Finally, we describe the genomic context of this transposon in several P. syringae strains. Our work provides insights into the evolutionary process leading to the formation of new PAIs in P. syringae, and how they may be transferred from strain to strain.


Sequencing of pFKN

The sequence of pFKN was determined to an average of ≈ 3.1× coverage using a shotgun approach. The complete nucleotide sequence of pFKN consists of 39 554 bp. Its overall G+C content is 53.53%. Twenty-nine putative open reading frames (ORFs) were identified, representing 75.7% of the sequence. To assign possible functions to these ORFs, we compared the sequence of each with available databases, using blastn and blastx. Twenty-seven ORFs had homology with genes present in the databases. We searched the translated products of each ORF for known functional domains and motifs and could assign a putative function to most of them (Table 1).

Table 1. . ORFs and non-coding elements identified in P.s. pv. maculicola pFKN. a . We identified 4 DNA fragments that were conserved between pFKN and the genome of Pst DC3000 (>85% nucleotide identity). The ORFs that have no homologues in the genome of Pst DC3000 are noted as ‘none’, the fragments are numbered 1–4. b . The expression of each gene carried by the plasmid was assessed in KB and MM (hrp-inducing media) by RT-PCR, and by using GFP-translational fusions in planta for 7 of them (not shown). When the gene is expressed in one given condition, the table exhibits ‘+’. A lack of expression is marked by ‘–’. NA, not available.Thumbnail image of

We organized the ORFs into various functional groups (Fig. 1). The plasmid encodes two type III effectors (avrRpm1 and an allele of avrPphE). Five non-contiguous regions on the plasmid may be mobile elements or remnants thereof. Three other groups of genes were defined, according to the putative function that they encode.

Figure 1.

Graphical map of pFKN and its 29 identified ORFs. The proposed direction of transcription is indicated by arrows. Red boxes indicate genes encoding for type III effectors. A black arrow line at the 5′ end of any genes indicates a possible hrp box. Genes related to those of mobile elements are represented by green boxes. The avrPphE transposon and the IS801 homologue are framed. Transposon repeats surrounding the avrPphE transposon are marked by green lines. The avrRpm1 gene is surrounded by repeats similar to the repeats of the transposon Tn501; these repeats are also marked by green lines. Purple boxes indicate genes with replication and partition functions. Genes encoding for putative transporters are indicated by blue boxes. The putative transcriptional regulators are indicated by orange boxes. The grey boxes represent genes that do not fall into any of the categories listed. The number of each ORF is noted in bold on the outside of the circle.

Plasmid replication and maintenance.

ORF1 shares high homology with genes that encode the replication proteins of plasmids belonging to the PT23A family in the P. syringae species (Sesma et al., 1998). Homologous replication proteins are found in other genera, such as the replication proteins from the ColE2 family in E. coli (Sesma et al., 1998). The product of ORF10 shares similarity with ParA, a protein involved in partitioning working in concert with a protein encoded by ParB. No ParB gene was found on pFKN. Therefore, it is not clear how the plasmid copies (around five per cell; data not shown) are distributed efficiently after cell division. ORF28 may encode a protein involved in the transfer of pFKN. The localization of the protein encoded by ORF28 is predicted to be in the bacterial periplasmic space. It has three domains. The N-terminus shows high homologies to VirB1 of Brucella suis (a mammalian pathogen) named after its homology to VirB1 of Agrobacterium tumefaciens. The central region shows homology to VirB4 of the same organism. The C-terminus shares homology with different DNA-binding proteins (DNA nickase, primase). However, no other gene on the plasmid seems to encode a function related to transfer. It is highly unlikely that this unique protein would allow transfer of pFKN. Therefore, ORF28 encodes an unknown function or could represent the remnant of a transfer system encoded on the plasmid before genomic rearrangements leading to its loss.

Type III effectors.

The promoter regions of two genes on the plasmid contain an hrp box, suggesting that they are part of the type III regulon. ORF2 is 80% identical to the avrPphE allele carried by P. s. syringae (Psy) (Alfano et al., 2000). The avrPphE gene had originally been identified in P. s. phaseolicola (Pph) and encodes a protein delivered to the host cell (Mansfield et al., 1994). Since then, several alleles have been identified in other pathovars of P. syringae (Fouts et al., 2002; Guttman et al., 2002). They all contain an hrp box; in ORF2 of pFKN, this is found 51 bp upstream of the ATG at nt 1561 as 5′-GGAACTGA-N13-CGACATA-3′ (Innes et al., 1993). Based on the homologies and the presence of the hrp box in the promoter, we hypothesized that the AvrPphE-like protein encoded by ORF2 is also a type III effector (see below). AvrRpm1 (numbered ORF5) is identical to that defined in the strain Psm M2 (Dangl et al., 1992).

Mobile element-related genes.

ORF3 (TnpA) encodes a putative truncated transposase. Importantly, avrPphE and TnpA are located between nearly identical inverted repeats of 66 bp (Fig. 1), thus defining a probable transposable element remnant. The sequences surrounding avrRpm1 (ORF4 and ORF6) are homologous to a sequence identified in P. s. pisi by Arnold et al. (2001), with 96–97% nucleotide identity. The DNA fragment carrying these sequences in P. s. pisi may also be a transposon or a remnant. It is therefore likely that avrRpm1 is also located on the remnant of a transposon.

ORF13 and ORF14 are related to ORFs found in the prophage 933 L on the LEE of the enterohaemorrhagic E. coli O157:H7, L0015 and L0014 respectively. These two ORFs are thought to belong to an IS element from the IS66 family, primarily identified in Sinorhizobium meliloti, that includes IS elements from animal and plant pathogens. These two genes are usually associated with a third one, L0013 in the LEE, that is absent in pFKN (the sequence upstream of ORF14 does not contain any ORF and does not show homology to L0013). Moreover, the 5′ end of ORF14 is also missing. Therefore, the sequence encoding ORF13 and ORF14 may represent the remnant of an IS element belonging to the IS66 family, not previously identified in P. syringae.

Transcription factors.

Four pFKN genes share homology with transcriptional regulators. Some of these may function during interactions with eukaryotic hosts. For example, the closest homologue to ORF24 is a putative transcriptional factor from V. cholerae. ORF24 is also related to putative proteins of unknown function (along the whole length of the protein) found in organisms as diverse as E. coli (50% identity/67% homology), Caenorhabditis elegans (41%/57%) Homo sapiens (42%/58%), Drosophila melanogaster (40%/58%) and Arabidopsis thaliana (29%/42%). A domain characteristic of FAD-dependent oxidoreductase (in Pfam) is detected in ORF24. Therefore, the function of the translation product of this ORF may require electron transfer through FAD.

The protein encoded by ORF27 is a member of the small RfaH family. RfaH controls the transcription of E. coli and Salmonella operons that direct the synthesis, assembly and export of the lipopolysaccharide core, exopolysaccharide, F-conjugation pilus and haemolysin toxin (for a review, see Santangelo and Roberts, 2002). RfaH is a specific regulator of transcript elongation; its loss increases transcription polarity in these operons without affecting initiation from the operon promoters. In addition, in one case, the RfaH protein controls the transcription of a single gene encoding an outer membrane haemin receptor, ChuA in pathogenic E. coli, as well as other genes involved in pathogenicity of the strain (Nagy et al., 2001; 2002). The promoter of the genes or operons controlled by RfaH contains a conserved sequence called JUMPStart (Hobbs and Reeves, 1994). No JUMPStart sequence was detected on pFKN.


ORF18 encodes a probable transmembrane protein that is highly homologous to sugar transporters. One is the product of mae1, a gene of the yeast Schizosaccharomyces pombe, encoding a protein involved in the uptake of l-malate, succinate and malonic acid. The product of ORF18 may provide the bacteria with a fitness advantage in certain conditions.

Other putative functions encoded on the plasmid.

Other genes may encode proteins providing an advantage to the bacteria during the infection process. For example, the product of ORF12 displays all the characteristics of methyl-accepting chemotaxis proteins. The genes encoding the flagellar unit and the chemotaxis functions are often organized in clusters, which is not the case on pFKN. In this particular case, the flagellar unit may be provided in trans.

pFKN is composed of different fragments that indicate a multistep process leading to its formation

The G+C content of the pFKN ORFs is highly variable along the plasmid (Table 2). Yet several contiguous gene sets on the plasmid have similar G+C contents. This probably indicates that these genes were acquired together from an original source. The G+C content might allow us to track the multistep acquisition process resulting in pFKN. For example, ORFs 7, 8 and 10 share nearly identical G+C content (50.5%), but are separated by an IS element with a significantly different G+C content (60.3%). To determine the distribution of pFKN ORFs in other pathovars, and how this distribution is related to G+C content by ORF group, we performed sequence comparisons with the nucleotide sequence of the complete genomes of P. syringae pv. tomato DC3000 (http:tigrblast.tigr.orgufmg) and P. syringae pv. syringae B728a (http:www.jgi.doe.govJGImicrobialhtmlindex.html). DNA fragments carrying from two to seven contiguous pFKN ORFs were detected in both genome sequences (85–100% identity). The four DNA fragments from P. s. tomato DC3000 carrying linked pFKN ORFs were numbered and added to Table 2. Of these, Pst DC3000 fragment 2 was also detected in Psy B728a (98% identical to pFKN). In addition, sequences homologous to repA and ORF13 were also detected in this Psy B728a. Table 2 shows that ORFs belonging to the same contiguous DNA fragment have a similar G+C content. For example, genes belonging to fragment 4 have a relatively high G+C content. Genes belonging to fragments 1 and 2 have similar G+C contents, and genes in fragment 3 exhibit a lower G+C content. Thus, the different fragments constituting pFKN were probably acquired by the three genomes in different steps, followed by subsequent sequence deletions and/or rearrangements with respect to Psm M6, as the entire collinear pFKN is not detected in either Pst DC3000 or Psy B728A.

Table 2. . pFKN ORFs sorted by descending G+C content. a . Some pFKN ORFs are contiguous on the plasmid and are collinear in this table, indicating that they possess a similar G+C content. When this is the case, an empty row separates them from the next ORF in the table. b . The DNA fragments from the Pst DC3000 genome that contain sequences homologous to ORFs from pFKN were colour-coded (#1: green, #2: red, #3: blue, #4 pink). The table indicates that fragment #4 has a relatively high G+C content, whereas fragment #3 has a relatively low content.Thumbnail image of

Full virulence of PsmM6 requires pFKN and is essentially caused by avrRpm1

To determine the contribution to virulence of the plasmid pFKN, we cured Psm M6 of pFKN (see Experimental procedures) to generate strain Psm M6C. There was no obvious macroscopic difference in visible disease symptoms after inoculation of either wild-type or Psm M6C onto rpm1-3 Arabidopsis leaves (unable to recognize avrRpm1). At high levels of inoculation (5 × 106 cfu ml−1), the inoculated leaves exhibited water soaking and tissue collapse starting at 24 h after inoculation. Inoculation at 5 × 105 cfu ml−1 led to chlorosis between 48 and 72 h after inoculation. Using this low-density inoculum, we measured pathogen growth over time in rpm1-3 leaves. Psm M6C grew 10-fold less than the wild-type strain (Fig. 2A). Therefore, the cured strain is less virulent than the wild-type strain. We performed complementation experiments using various pFKN subclones in Psm M6C to determine which pFKN ORFs would partially or fully restore virulence. Only clones carrying avrRpm1 could restore full virulence (Fig. 2A). This result is consistent with insertional mutation of avrRpm1 in the Psm M2 strain (Ritter and Dangl, 1995). No other clone conferred enhanced virulence (data not shown).

Figure 2.

avrRpm1 encodes the determining virulence factor on pFKN.

A. The growth of Psm M6C on rpm1-3 plants was compared with the growth of the wild-type Psm M6 strain and Psm M6C strains complemented with different fragments from pFKN. The pFKN subclone carrying avrRpm1 restored full virulence to Psm M6C (the other subclones did not restore any virulence and are not shown). Leaves were hand inoculated with 105 cfu ml−1, and a titration of bacteria in the leaves was performed 3 days after inoculation. Data points represent the mean ± standard error of four samples.

B. The avrPphE gene was deleted from the chromosome of Psm M6C to generate Psm M6CKO. To assess the virulence function of avrPphE or its influence on the virulence function of avrRpm1, the growth of Psm M6CKO on rpm1-3 plants was compared with the growth of Psm M6C, Psm M6C + avrRpm1 and Psm M6CKO + avrRpm1. Data points represent the mean ± standard error of four samples.

The role of avrPphE in virulence could not be assessed using Psm M6C because a second copy of the gene is present on the chromosome of the cured strain [as assessed by polymerase chain reaction (PCR) amplification and sequencing of the PCR product, see below]. We replaced the avrPphE copy on the chromosome of Psm M6C with an omega fragment to create Psm M6CKO (see Experimental procedures). Psm M6CKO grew similarly to Psm M6C on rpm1-3 plants (Fig. 2B), suggesting that avrPphE plays no major role in the virulence of Psm M6. However, our recent results indicate that expression of avrPphE from the lac-nptII promoter in PsmM6CKO does enhance virulence of PsmM6CKO, suggesting that the quantity of AvrPphE protein delivered to the host cell may be important in its function (J. Shock, J. Chang, Z. Nimchuk, L. Rohmer and J. L. Dangl, unpublished). As avrPphE was located in the vicinity of avrRpm1 on pFKN, we tested the potential involvement of avrPphE in the virulence function of avrRpm1. We compared the growth of Psm M6C and Psm M6CKO when each carried avrRpm1 on a plasmid. No difference in growth was observed (Fig. 2B), indicating that avrPphE does not affect avrRpm1 virulence function.

Identification of additional genes putatively involved in pathogenicity based on their expression pattern

The expression of known P. syringae virulence factors is specifically induced in planta. Minimal media (MM) plus a sugar source seems to mimic the plant apoplast, leading to the expression of the type III pilus genes and effectors (Huynh et al., 1989; Innes et al., 1993). We compared the expression of the pFKN ORFs in MM and in rich medium (KB) using reverse transcription RT-PCR methods. When the expression of an ORF was observed in both media, we assumed that the gene was constitutively expressed. Expression of a gene in MM but not in KB may indicate a role for this gene during interaction with the host. Most of the pFKN genes are constitutively expressed. However, the expression of both avrRpm1 (Ritter and Dangl, 1995) and avrPphE was detected in MM, but not in KB. Additionally, ORF26 is expressed in MM but not in KB. ORFs 14 and 24 were not expressed in either of the two conditions. The expression of ORF8 and ORF11 could not be assessed using this technique despite many attempts.

The six genes (avrRpm1, avrPphE, ORF8, ORF11, ORF14 and ORF24) for which no constitutive expression was observed by RT-PCR were selected for further study. We constructed transcriptional fusions between the 5′ end of the genes of interest and a gene encoding green fluorescent protein (GFP) and chose ORF10 as a constitutive control. The fluorescence of PsmM6C carrying the transcriptional fusions was assessed in KB, MM and in planta by exposing the bacteria grown in these conditions to blue light (488 nm) and observing them by fluorescence microscopy. A strain carrying promoterless GFP did not exhibit any fluorescence in the conditions tested, whereas all cells carrying the ORF10–GFP fusion did. Only the avrPphE, avrRpm1 and ORF8 GFP fusions were active in plants (Table 1).

pFKN can integrate into the Psm M6 chromosome

One colony of Psm M6, derived during routine maintenance, exhibited unusual features. Although this strain triggered recognition (measured by HR) on RPM1 plants and disease symptoms on rpm1 plants, no plasmid could be isolated from it. AvrRpm1 was still present in total DNA, as assayed by DNA blot (Fig. 3A). We therefore concluded that avrRpm1 was now located on the Psm M6 chromosome. To determine whether additional portions of pFKN were also present on the variance chromosome, total DNA from the variant Psm M6 was digested with EcoRI, HindIII, EcoRV, SacI, PvuII, SmaI, BglII and NsiI and probed with the DNA from different parts of the pFKN plasmid. The DNA of Psm M6 containing pFKN was used as a positive control. This analysis revealed a polymorphism between the avrRpm1 gene in this strain (e.g. an 8.0 kb BglII fragment) compared with pFKN (e.g. an 8.8 kb BglII fragment; Fig. 3A). All restriction fragments from pFKN were detected in the DNA of Psm M6integron. Therefore, pFKN has integrated into the chromosome. For the purpose of clarity, the strain bearing avrRpm1 in the chromosome will be referred to as Psm M6integron.

Figure 3.

pFKN can integrate into the Psm M6 chromosome.

A. Total DNA from Psm M6 (lanes F) and Psm M6integron (lanes I) was digested with EcoRI, HindIII, EcoRV, SacI, PvuII, SmaI, BglII and NsiI and probed with various pFKN fragments. Here, the DNA blot was probed with the avrRpm1 ORF. The two DNA samples exhibit a similar pattern except for a polymorphism (e.g. EcoRI, BglII and SmaI) near the site of pFKN integration into the chromosome.

B. The integration of the plasmid alters the overall fitness of Psm M6integron. The growth of Psm M6integron in KB (empty symbols) and MM (filled symbols) was compared with the growth of Psm M6 and Psm M6C. Bacteria were inoculated at 104 cfu ml−1, and cultures were titrated at 0, 6, 12, 29, 32, 54 and 60 h after the inoculation. Psm M6pFKN is indicated by squares, Psm M6C is indicated by triangles, and rhombi indicate Psm M6integron. Data points represent the mean ± standard error of four samples (some error bars are not visible because too small). Psm M6pFKN and Psm M6C grew equally well in both minimal medium and KB. Psm M6integron grows slower than Psm M6pFKN and Psm M6C in both media.

We have, on occasion, isolated a replicon from individual colonies of Psm M6integron (see next section) The plasmid purified from these bacteria always exhibited a restriction profile identical to that of pFKN. Therefore, pFKN can excise from the chromosome, and Psm M6integron can revert to Psm M6. Interestingly, the overall fitness of Psm M6 is affected by the integration status of pFKN into the chromosome. Psm M6integron grows in media (KB and MM) more slowly than either the original Psm M6 strain or a revertant from Psm M6integron that is cured of the plasmid (Fig. 3B). Because Psm M6 (with a pFKN replicon) and the plasmid-cured derivative of Psm M6integron both grew equivalently, we can conclude that it is the integration of pFKN per se that results in this growth impairment. This is not uncommon, as several plasmid replicons are known to slow growth of the organism when integrated (see Discussion).

The pFKN integration site in Psm M6integron is the avrPphE transposon

We noticed from the DNA blot that the genomic environment of avrRpm1 is different in Psm M6 and Psm M6integron. This is also the case for both copies of avrPphE (data not shown), suggesting that the pFKN integration site is near avrRpm1 and avrPphE, and that the chromosomal location into which the plasmid integrates is near the chromosomal copy of avrPphE. We first defined that the chromosomal avrPphE is carried on the same transposon found in the sequence of pFKN, using PCR amplification on total DNA of Psm M6C, with primers specific to sequence on the avrPphE transposon (Table 5). The PCR products were sequenced and compared with the corresponding sequenced regions of pFKN. As shown in Fig. 4A, the chromosomal copy of the transposon contains two sequences not present on pFKN. One 1244 bp sequence (region A) is located between avrPphE and the coding region of TnpA. It carries an ORF with a sequence that is nearly identical to various genes encoding resolvases of Tn3 transposons. Additionally, a portion of 661 bp of this region encodes the 5′ portion of TnpA. Another region of 271 bp unique to the chromosomal copy is located within the coding region of the transposase (region C), and is 3′ from a region shared by both pFKN and chromosomal copies of TnpA (region B). The addition of regions A and C to the chromosomal copy results in a complete TnpA ORF correcting the frameshift at the 3′ end of TnpA carried on pFKN.

Table 5. . Primers used in this study.
Primer nameSequence (5′−3′)Purpose
fkp1405 <GTGAAATCGCTTAGCGTACTTTTAIL-PCR on the left border of the avrPphE transposon
fkp1444 <AGCGGCAACAGCGCGGGGAAATAIL-PCR on the left border of the avrPphE transposon
fkp1509 <GCCATAGCCGCTGAAGCAACATTTAIL-PCR on the left border of the avrPphE transposon
fkp1559 <TGTCGTCATTTGGCCGGGTCTAIL-PCR on the left border of the avrPphE transposon
fkp4637 >TGGGCGAAATCCGAGACCGTAGCTTAIL-PCR on the right border of the avrPphE transposon
fkp4802 >TGTCGCCGCTGGGCTGGGTAIL-PCR on the right border of the avrPphE transposon
fkp6431 >CGCCTCGGTCGTGACTGCATTGTAIL-PCR on the right border of the avrPphE transposon
K110TCGGCACTTTGTAGCCAGTSequencing of the avrPphE transposon
K113GTCTTGCCCTGGCACCACSequencing of the avrPphE transposon
Larp2AATAGGTTTACACCTGTTTCACGSequencing of the avrPphE transposon
Larp3TTTCACGCCCTTGGCCTTGGSequencing of the avrPphE transposon
AD1NTCGASTWTSGWGTTDegenerated primers for TAIL-PCR
AD4NCTAGWASTWGSTTGDegenerated primers for TAIL-PCR
Figure 4.

The pFKN integration site is a second copy of the avrPphE transposon.

A. The region in the Psm M6 chromosome that hybridizes to the avrPphE and TnpA probes was amplified and sequenced. The sequence of the chromosomal region is 100% identical to the transposon in the plasmid, but contains two additional regions (regions A and C). TnpA encodes a transposase and res a resolvase. Based on homology, the chromosomal avrPphE transposon could be complete and functional. The avrPphE transposon carried on the plasmid lacks the 5′ end of TnpA and res (region A), and the absence of region C results in the truncation and a frameshift (marked FS) of the gene encoding the transposase. The primers used for PCR amplification (Fig. 4B) are indicated by arrows, and their names are in bold.

B. To investigate where regions A and C were distributed after integration of the plasmid into the chromosome, PCRs were performed on the DNA of Psm M6integron and Psm M6 using Larp3 and K110, and K110 and K113 (see Fig. 4A). The product amplified from Psm M6integron is run in lane I. The product amplified from Psm M6 is run in lane F. The product amplified from Psm M6C is run in lane C. The difference in size observed between the PCR products from Psm M6integron and Psm M6 using Larp3 and K110 indicates that regions A and C are not carried on the same putative transposon in the chromosome of Psm M6integron (in contrast to the chromosome of Psm M6). The PCR products from Psm M6integron and Psm M6 using K113 and K110 show that neither region A nor region C has been lost in the chromosome of Psm M6integron. The identical size of the PCR products from Psm M6integron and Psm M6 using K113 and Larp2 indicates that region C is on the right border of pFKN when integrated into the chromosome. The region A is on the left border, as it is carried by the other putative transposon.

C. After integration of pFKN into the chromosome, TnpA on the left border contains region A, but not region C. TnpA on the right border contains region C. The primers used for the PCR amplification are indicated by arrows, and their names are in bold (in Fig. 4A).

TnpA, avrPphE and the putative resolvase fragment are located between nearly identical inverted repeats of 66 bp on both the plasmid and the chromosome. Their organization is reminiscent of Tn3-like transposons (Sherrat, 1989). Moreover, this TnpA is 81% identical to those of the Tn21 subfamily of Tn3 elements (Diver et al., 1983). As we suspected that the integration site is in the vicinity of avrPphE in the chromosome and the plasmid (i.e. in the two transposons), we analysed by PCR the two transposons in the genome of Psm M6integron. Figure 4B shows that the unique regions A and C are no longer located on the same transposon, but are instead separated by the entire plasmid sequence (Fig. 4C). This indicated that the integration site of the plasmid is located in the region between region A and region C, i.e. within the homology provided by region B (Fig. 4).

We used this sequence data to investigate the rate of excision occurring in Psm M6integron grown in different conditions by PCR. The rate of excision was estimated at around 10−4 (data not shown). No difference in rate of excision could be detected between bacteria grown in KB, MM and on resistant or susceptible plants.

The genomic context of avrPphE is different in various P. syringae strains

We investigated the regions surrounding the avrPphE transposon on the chromosome of Psm M6 (Fig. 5) by TAIL-PCR. The left border shows 97% nt identity to the 5′ end of the insertion sequence ISPsy3 (AB063176) from P. syringae pv. eriobotryae. This element is interrupted by the avrPphE transposon. The translation product of the interrupted ORF found on this border is homologous to the IS801 transposase (79% identity and 86% homology), and IS801 has been associated with other type III effectors in various P. syringae pathovars (Kim et al., 1998). On the right border, the 200 bp sequence following the terminal repeat of the avrPphE transposon encodes a protein homologous (81% nucleotide identity) to the 3′ end  of an unrelated putative transposase gene from Pseudomonas aeruginosa. The next 1062 bp sequence downstream shows high nucleotide homology to sequence identified in P. syringae pv. pisi (Arnold et al., 2001). This DNA fragment may be a transposon remnant and carries the avrPpiA1 allele of avrRpm1 from P. syringae pv. pisi, among other genes. In the chromosome of Psm M6, this sequence shows homology to the P. s. pisi genes ORFD (94% identity), encoding a putative transposase, and ORF3, encoding a putative type III effector (100% identity). Thus, this region in Psm M6 is a remnant of the putative transposon found in P. s. pisi.

Figure 5.

Sequences surrounding the avrPphE transposon define four different genetic contexts. The figure is a graphical representation of the four genomic environments in which the avrPphE transposon was detected. The regions represented were characterized by sequence comparison using the blastx algorithm. It is not known whether the avrPphE transposon is located in the chromosome or on a plasmid in the genome of Psm M5 and Pst DC3000.

We detected the avrPphE transposon in other strains of P. syringae pv. maculicola and in the sequence of the genome of P. s. tomato DC3000. In Psm M2, Psm M8, Psm M12, Psm m90 and Psm m10831, the borders are the same as in Psm M6 (data not shown). On the other hand, they differ in Psm M5 and Psm M9, and these all differ from Pst DC3000. In Psm M5 (Fig. 5) and Psm M9 (not shown), the left border consists of the 5′ end of a gene sharing 99% identity to the replicase of pPS10, a plasmid found in P. syringae pv. savastanoi (Nieto et al., 1992). This replicase is not related to that of pFKN (a PT23A-like plasmid; see above). The right border sequence is a remnant of a gene encoding a putative ABC transporter (23% identity and 43% homology to a putative sugar uptake ABC transporter ATP-binding protein from Sinorhizobium meliloti).

In the genome of Pst DC3000, we looked at 30 kb surrounding the transposon (Fig. 5). The proteins encoded on this 30 kb region share homology to either DNA mobility functions (transposases, plasmid partitioning proteins, replicases) or type III effectors (identical to avrPpiB; Cournoyer et al., 1995). The region also contains other sequences often found associated with plasmids and genes involved in pathogenicity, such as rulA and rulB (Sesma et al., 2000). Sequences directly neighbouring the transposon on the left side are homologous to a gene encoding the transposase of the IS100 elements from Yersinia pestis, Ralstonia solanacearum and Salmonella enterica. On the right-hand side of the avrPphE transposon, the first homology is to a fragment of a putative ABC transporter, identical to the one identified in Psm M5. This suggests that this part of the two genomes is more recently diverged than either is from Psm M6.

No common sequence was identified in the four different environments surrounding the avrPphE transposon that would help to identify the integration target of the avrPphE transposon, although the proximal right borders in Psm M5 and Pst DC3000 are nearly identical. This suggests that the avrPphE transposon and this bordering sequence were acquired together by these two strains, and that subsequent events at both the distal right flank and on the left border are responsible for the differences in genomic context. In contrast, the insertion sites of the avrPphE transposon in Psm M5 and Psm M6 are completely different. In sum, the features of these integration sites are consistent with the model that PAIs evolve by repeated insertion and deletion into pre-existing PAIs (Hensel et al., 1999; Tauschek et al., 2002).


We sequenced pFKN, an ≈ 40 kb plasmid that contributes to the virulence of Psm M6. The plasmid carries two type III effectors on remnants of transposons, contiguous to ORF8, itself neighbouring an element closely related to IS801. avrRpm1 encodes a known virulence factor and avrPphE, a type III effector. The expression of ORF8 is induced in planta and, therefore, ORF8 may play a role in the interaction with the host plant. The portion of the plasmid carrying avrRpm1, avrPphE and ORF8 may thus be considered as a PAI. Interestingly, the G+C content of the ORFs carried on this region are very heterogeneous, and the three genes are flanked by sequences related to mobile elements and, therefore, are likely to have different origins. This is consistent with the idea that the formation of PAIs requires a multistep process (Hensel et al., 1999; Alfano et al., 2000; Tauschek et al., 2002). It is also consistent with the findings of Kim et al. (1998) showing that most of the known type III effectors are associated with this type of sequence. pFKN (and therefore the PAI that it carries) sometimes integrates into and can excise from the chromosome, as demonstrated by the discovery of the variant of the strain Psm M6. We identified the region of the chromosome into which the integration took place. This region is a transposon carrying a second copy of avrPphE.

Alleles of the avrPphE gene have been detected in many P. syringae. pathovars (Mansfield et al., 1994; Alfano et al., 2000; Fouts et al., 2002; Guttman et al., 2002), suggesting that it provides an important virulence function on a variety of host plants. Our findings are the first to associate avrPphE with a transposon. In Psy B728a and in Pph 1302A, for example, avrPphE resides directly next to the hrp locus, in a region called the ‘exchangeable effector locus’ or EEL (Alfano et al., 2000), and is not associated with a transposase. In fact, in Psy B728a, avrPphE is predicted to be the third gene in an operon controlled by the hrpK promoter (Alfano et al., 2000). This striking finding illustrates the dynamism of PAIs.

We determined by complementation assay that the partial loss of virulence in the strain cured of pFKN results from the absence of avrRpm1. The virulence function of avrRpm1 has already been demonstrated in the strain Psm M2 (Ritter and Dangl, 1995). Recently, AvrRpm1 has been demonstrated to interact with RIN4, a protein involved in the regulation of some of the plant defence responses (Mackey et al., 2002). Mackey and coworkers proposed that the role of AvrRpm1 is to reduce the level of plant defence response triggered by the attack of the pathogen. This hypothesis might be consistent with the fact that only quantitative differences in growth were observed between Psm M6C and Psm M6. No difference in growth was observed between Psm M6C and Psm M6CKO (lacking avrPphE). Similar results were observed with the avrPphE allele of P. s. phaseolicola (Stevens et al., 1998). This observation could be explained by either a functional redundancy with another gene in the genome of the Psm M6 or the fact that the assay is inappropriate to assess the virulence function of this type III effector.

The sequence and the integration mechanism of pFKN into the Psm M6 chromosome provide insights on the evolutionary dynamics of the genomes of P. syringae pathovars. First, the analysis of the G+C content of fragments of pFKN suggests that they may have different origins. This hypothesis is further supported by the presence of some defined fragments from pFKN in the genomes of Pst DC3000 and Psy B728a. The fragment 3 (see Table 2) encodes one putative transcriptional regulator, proteins predicted to be located in the bacterial inner or outer membrane by psort (inorganic pyrophosphatase, sugar transporter, putative protein) and a putative integrase/recombinase homologous to ripX from Bacillus subtilis. The presence of this latter gene suggests that fragment 3 may be the remnant of a bacteriophage.

The integration and excision mechanism is very probably a homologous recombination in the 1.5 kb sequence region B of TnpA carried on the avrPphE transposon (Fig. 6). It is not clear whether the integration/excision process relies on general recombination or on a site-specific recombination, dependent on an integrase/recombinase. As indicated by the sequence of the two transposons before and after integration of pFKN, no nucleotide sequences were altered at the site of exchange, and not a single nucleotide was lost or gained (this excludes a mechanism of transpositional site-specific recombination). General recombination requires a large DNA region, whereas the site-specific recombination requires as few as 240 bp for the lambda phage for example (Thompson and Landy, 1989). The rate of general recombination is unknown for P. syringae, but has been estimated at 10−3 in Pseudomonas tolaasii (and, when recA is inactivated, at 10−4; Sinha et al., 2000). The rate of excision of pFKN has been estimated between 10−3 and 10−4 and would therefore be consistent with a mechanism of general recombination. Additionally, the region B is 1.5 kb long, enough to ensure general recombination. Older reports describe the integration of pMMC7501 into the chromosome of P. s. pv. phaseolicola (Curiale and Mills, 1982). An imprecise excision of pMMC7501 led to the formation of eight derivative plasmids, all sharing the same origin of replication (Szabo and Mills, 1984a). This excision occurred through recombination events between sets of repeats in the chromosome exhibiting sequence homology with IS elements (Szabo and Mills, 1984b; Poplawsky and Mills, 1987).

Figure 6.

The integration of pFKN may have occurred via homologous recombination between the regions B common to the two TnpA genes. The TnpA genes are constituted of four DNA sequence regions: (1) before the integration, the chromosomal copy of TnpA contains all four regions (A, B, C and D). The TnpA on the plasmid is missing the 5′ end of the gene (region A) and a fragment towards the 3′ end (region C), which results in a frameshift (FS). After the homologous recombination between two identical regions (region B) in the coding region of the TnpA genes (2), the region A is found in the TnpA of the left border and region C in the TnpA of the right border (3).

The integration of a virulence gene into the chromosome may be a first step towards its stabilization in the genome. Indeed, lack of selection can lead to the stochastic loss of the plasmid, and many steps in the pathogen life cycle probably do not require the presence of virulence functions. The loss of a plasmid carrying the type III effector avrPpiB was observed in P. syringae pv. pisi (Bavage et al., 1991). We propose that the integration of pFKN could lead to the stabilization of avrRpm1 in the genome of Psm M6. Interestingly, a DNA segment of at least 42 kb carrying avrPphB in P. syringae pv. phaseolicola can excise from the chromosome, but it is not maintained in the strain as a plasmid. This region may represent an ancient plasmid integration event, followed by successive genomic rearrangements deleterious for plasmid maintenance (Jackson et al., 2000). The fact that we can still observe the excision of pFKN from the chromosome of Psm M6integron shows that the newly generated genetic unit is not rapidly stabilized in the genome, and that further events might be necessary to prevent excision. Furthermore, the integration of pFKN into the chromosome slightly impairs the fitness of Psm M6integron. The presence of the functional origin of replication of pFKN may interfere with the replication of the chromosome. Integration of multicopy plasmids is commonly lethal to E. coli unless the replication origin of the integrated element is suppressed (Yamaguchi and Tomizawa, 1980; Froehlich et al., 1983).

pFKN is not the only genetic unit potentially stabilized by integration. The chromosomal avrPphE transposon that acts as the insertion site for pFKN may be active before integration. The chromosomal TnpA has a full ORF, and we observed transcription in both KB and MM using RT-PCR (data not shown). The pFKN integration event disrupts the chromosomal avrPphE transposon, thus preventing autonomous transposition. As transposition can be detrimental, the stabilization of the avrPphE transposon could be advantageous for the bacterium.

Strikingly, pFKN integrates into a region of the Psm M6 chromosome already encoding two type III effectors (avrPphE and ORF3 identified in Ppi). The avrPphE transposon has been detected at four different loci (chromosomal and on pFKN in Psm M6, one in Psm M5 and one in Pst DC3000). In all cases, genes related to DNA mobility were identified in the vicinity. In at least three out of four cases (Psm M6 chromosome, pFKN and Pst DC3000), additional genes encoding type III effectors were also found in the vicinity of the avrPphE transposon (Fig. 5). The sequence available for PsmM5 is the shortest, and we detected no homology to virulence genes in the flanking sequence. However, part of the sequence on the right border is 100% identical to the immediate right border identified in Pst DC3000. It is likely that the larger genomic context of the avrPphE transposon in both strains has a common ancestor (or was transferred from one strain to the other). Finally, the transposon in Pst DC3000 and Psm M5 is 100% identical to the transposon in Psm M6. This suggests that the acquisition of the transposon was rather recent. As a comparison, avrRpm1, its allele avrPpiA1 and their surrounding sequences are 97% identical in nucleotide sequence. The rather recent acquisition of avrPphE and its association with other genes related to pathogenicity and DNA mobility may suggest that the avrPphE transposon is part of a PAI in formation. The integration of pFKN into this site would be the next step in the process. The proposed mechanism of integration of pFKN suggests that this plasmid could integrate in the chromosome of any strain that carries the avrPphE transposon (provided that the strain acquires the plasmid). Transfer of genes related to pathogenicity or adaptation to various conditions may be facilitated or even explained by the integration of plasmids in the bacterial chromosome followed by the imprecise excision of plasmids. Excised plasmids would contain all or part of the integrated plasmid and regions from the chromosome. The transfer of these plasmids could explain how different strains share nearly identical regions in their respective genomes.

Experimental procedures

Bacterial strains and culture conditions

The bacterial strains used in this study are shown in Table 3 and the recombinant plasmids in Table 4. The Pseudomonas strains were grown in King's B media (KB) (King et al., 1954) shaken at 28°C or grown on KB agar plates at 28°C. The E. coli strains were grown in 2× YT broth or on Luria–Bertani (LB) broth agar plates at 37°C (Maniatis et al., 1989). For Pseudomonas strains, antibiotics were used at the following concentrations (mg l−1): rifampicin, 25–50; kanamycin, 50; carbenicillin, 100; spectinomycin, 10. Plates contained cycloheximide (50 mg l−1). For E. coli strains, antibiotics were used as follows (mg l−1): ampicillin, 100; nalidixic acid, 10; kanamycin, 100; spectinomycin 100.

Table 3. . Bacterial strains used in this study.
Species or pathovarStrain or raceDescriptionSource or accession no.
E. coli Dh5α Life Technology
P. syringae pv. maculicolaM6 LMG 5560
P. syringae pv. maculicolaM6integronVariant of Psm M6 cured of pFKN in which the plasmid pFKN has integrated the chromosomeThis study
P. syringae pv. maculicolaM6C Psm M6 cured of pFKNThis study
P. syringae pv. maculicolaM2 LMG 5071
P. syringae pv. maculicolaM5 LMG 5559
P. syringae pv. maculicolaM8 IHR(W) 793
P. syringae pv. maculicolaM9 IHR(W) 1120B
P. syringae pv. maculicolaM12 IHR(W) 1809A
P. syringae pv. maculicolaM90 From Brian Staskawicz (University of California, Berkeley)
P. syringae pv. maculicolaM10832 From Brian Staskawicz (University of California, Berkeley)
P. syringae pv. maculicolaM6CAR1 Psm M6 cured of pFKN carrying pDSKAR1, encoding an HA-tagged version of AvrRpm1This study
P. syringae pv. maculicolaM6CKO Psm M6 cured of pFKN in which the chromosomal version of avrPphE has been replaced with an omega fragmentThis study
P. syringae pv. maculicolaM6CKOAR1 Psm M6 cured of pFKN in which the chromosomal version of avrPphE has been deleted, carrying pDSKAR1This study
P. syringae pv. maculicolaM6CpLaR9 Psm M6 cured of pFKN carrying pLaR9This study
P. syringae pv. maculicolaM6CpLaR10 Psm M6 cured of pFKN carrying pLaR10This study
P. syringae pv. maculicolaM6CpLaR11 Psm M6 cured of pFKN carrying pLaR11This study
P. syringae pv. maculicolaM6CpLaR12 Psm M6 cured of pFKN carrying pLaR12This study
P. syringae pv. maculicolaM6CpLaR13 Psm M6 cured of pFKN carrying pLaR13This study
Table 4. . Plasmids used in this study.
PlasmidsDescriptionSource or reference
pFKN40 kb native plasmid in Psm M6This study, AF359557
pUC18cloning vector Norrander et al. (1983)
pFK20pFKN fragment (37174–1425) generated in the sequencing shotgun library, carrying the replication protein of pFKNThis study
pBBR1-MCS2Broad-host-range vector Kovack et al. (1995)
pRK2013Helper plasmid Figurski and Helinski (1979)
pDSKAR1 avrRpm1 gene fused to an HA-tag carried on pDSK519J. Chang, UNC
pLaR9pFKN ScaI–SmaI fragment cloned into This study
pBBR1-MCS2, carrying ORF7 to ORF11 
pLaR10pFKN ApaI–BamHI fragment cloned into This study
pBBR1-MCS2, carrying ORF10 to ORF14 
pLaR11pFKN EcoRI–BamHI fragment cloned into This study
pBBR1-MCS2, carrying ORF15 to ORF22 
pLaR12pFKN EcoRI fragment cloned into pBBR1-MCS2, This study
carrying ORF23 to ORF26 
pLaR13pFKN SacI–SacII fragment cloned into This study
pBBR1-MCS2, carrying ORF27 to ORF29 
pLaR24Omega fragment carrying the spectinomycin resistance gene (aadA+) cloned into the sequences surrounding avrPphE in the chromosome or Psm M6,1 kb on each side, carried on pBBR1-MCS2This study
p186vector for translation fusion (on the first reading frame) with the C terminus of AvrRpt2 and translational fusion with GFP3 on a pBBR1-MCS2 backboneJ. Chang (UNC)

DNA manipulations

Plasmid DNA isolations were performed by alkaline lysis extraction as described by Maniatis et al. (1989). Genomic DNA isolations were performed as described by Syn and Swarup (2000). All molecular manipulations were done via standard procedures. Enzymes were purchased from New England BioLabs or Boehringer Mannheim and used according to the manufacturer's specifications. For Southern analysis, ≈ 2 µg of DNA was loaded per lane on agarose gel. DNA was transferred to Hybond N+ membranes (Amersham Pharmacia Biotech), and hybridization was performed in hybridization solution [1.5× SSC, 7% SDS, 10% polyethylene glycol (PEG) molecular weight 8000, 100 µg ml−1 sonicated and denatured herring sperm DNA, 250 µg ml−1 heparin] at 65°C. The DNA probes were labelled with α-ATP using reagents provided in the Prime-It II random primer labelling kit (Stratagene).

Growth curves and inoculations

To assess the growth of bacteria in planta, bacteria were inoculated at a density of 105 cfu ml−1 in 10 mM MgCl2 on four leaves per plant, and four leaf discs from individual plants were collected at day 0 (to assess the original inoculum), day 1, day 3 and day 5. The discs were ground in 1 ml of 10 mM MgCl2, and serial dilutions were plated on KB plates to titrate the bacteria. Four technical repetitions were performed per experiment (Ritter and Dangl, 1995).

Shotgun cloning and DNA sequencing

Purified pFKN DNA (2.25 µg) was partially digested with Tsp5091. Fragments ranging in size from 2 to 8 kb were isolated from an agarose gel with a QIAquick gel isolation kit (Qiagen). The extracted fragments were ligated into EcoRI-digested pUC18 (Pharmacia). The ligation was transformed into electrocompetent E. coli strain DH5α (Electro Cell ManipulatorR 600 from BTX, 50 µF, 129 Ω, 2.5 kV). This method has been described by Ansorge et al. (1997). The inserts of 150 randomly picked clones were sequenced using ABI dye terminator chemistry by the UNC-CH Automated DNA sequencing facility with M13F (TTCAGGGAGCCTGCG GTCC) and M13R (AACAGCTATGACCATG) primers. For gap closure, custom primers were designed from the ends of each contig and used to amplify PCR products with purified plasmid pFKN DNA as a template. The PCR products were purified using a High Pure PCR product kit (Hoffmann-La Roche) and sequenced directly by primer walking. Sequenced fragments were analysed and assembled using sequencher (Gene Codes). The coverage was on average 3.1× and minimum 2×. ORF determination was made manually and checked with glimmer (Delcher et al., 1999). The ORFs should meet the following criteria: (i) the start codon was ATG or GTG; (ii) the stop codon was TAA, TAG or TGA; (iii) the size of the ORF was over 300 bp. We may not have predicted the correct initiation codon. Where possible, we chose an initiation codon that is preceded by an upstream ribosome binding site (RBS) sequence (optimally 5–13 bp before the initiation codon). The algorithms blastn and blastx were used to compare pFKN DNA sequences with those in the GenBank database (http:www.ncbi.nlm.nih.govblast). The predicted proteins were also analysed for functionally important motifs using prosite and for func-tional and structural domains using pfam (http:www., prodom (http:protein.toulouse.inra.frprodomdocprodom.html), smart ( and blocks (http:blocks.fhcrc.orgblocksblockssearch.html). The sequence data have been submitted to the DDJ/EMBL/GenBank databases under the accession number AF359557. The sequence of pFKN was compared with the genome sequence of P. syringae pv. tomatoDC3000, available at http:tigrblast.tigr.orgufmg, and P. syringae pv. syringae B728a, available at http:www.jgi.doe.govJGImicrobialhtmlindex.html.

Plasmid curing and complementation

The subclone pFK20 from the shotgun library carries the replicase gene (ORF1) of pFKN that is also an incompatibility determinant for plasmids belonging to the PT23A plasmid family (Gibbon et al., 1999). pFK20 was electroporated into Psm M6 to cure the strain of pFKN. Single transformants were grown in three consecutive liquid cultures without selection for pFK20 (carbenicillin). The bacteria retrieved after growth no longer contained pFK20, indicating that the fragment carried by pFK20 was not sufficient for plasmid maintenance. No pFKN sequences could be detected by DNA blot.

For complementation tests, fragments of pFKN were cloned into pBBR1-MCS2 (pLaR9, pLaR10, pLaR11, pLaR12 and pLaR13; see Table 4) and transferred to the cured strain by triparental mating, using the helper plasmid pRK2013.

We deleted the avrPphE gene from the cured strain by marker exchange as described by Hendrickson et al. (2000). In brief, avrPphE was replaced by an omega fragment carrying a gene conferring resistance to spectinomycin and streptomycin (Prentki and Krisch, 1984). Sequences flanking the gene on both sides (1 kb each) were cloned into pBBR1-MCS2, carrying sacB (a lethal gene when the bacteria is grown on sucrose; Zagorec and Steinmetz, 1990). Between the two flanking regions was inserted the omega fragment carrying a gene conferring resistance to spectinomycin and streptomycin (Prentki and Krisch, 1984). The resulting plasmid (pLaR24) was transformed by triparental mating into Psm M6C using the helper plasmid pRK2013. Deletion mutants were then selected on spectinomycin (for integration of the omega fragment) and sucrose (for loss of pLaR24).

RT-PCR analysis

RNA was extracted from exponentially growing Psm M6 as described by Innes et al. (1993). The bacteria were subjected to growth in the test medium for 3 h. The RNA was then extracted using TrizolR LS reagent (Gibco BRL). RNA samples were treated with 10 U of RNase-free DNase I (Hoffmann-La Roche) for 1 h at 37°C according to the manufacturer's instructions. The enzyme was then inactivated for 5 min at 75°C. Reverse transcription reaction was performed using the RETROscriptTM kit (Ambion), according to the protocol provided by the manufacturer. PCRs included primers specific for each ORF to be tested and control primers allowing the amplification of part of the gene encoding for the glyceraldehyde-3P-dehydrogenase. The amplification with the control primers indicates that the non-amplification of the tested ORF results from the lack of expression of the gene and not a failed reaction. The primer sets used for this control were Larp17 (CAGCAACTTGCCGCTCAC) and Larp10 (ATGACTCTCCGTATCGC), giving rise to an amplification product of 1070 bp, or Larp17 and Larp11 (CTGCACCGT GAACTGGG), giving rise to a product of 406 bp. The following protocol was used: 3 min at 95°C, 30 cycles of 45 s at 94°C, 45 s at 54°C and 1 min at 72°C, and 5 min at 72°C. The primer sequences used for each ORF are available on request.

Transcriptional fusion and expression assay in planta

The 5′ ends of the genes for testing were cloned into p186 (J. Chang, personal communication) to fuse the N-termini of the genes for testing to the gene encoding GFP3. Psm M6C was transformed with the constructs by triparental mating between Psm M6C, DH5α carrying the construct or the helper plasmid pRK2013. Leaves of eight rpm1-3 plants were inoculated with each transformant at a density of 5 × 107 cfu ml−1. The bacteria were retrieved from the inoculated leaves by shaking the leaves in 10 mM MgCl2 and 0.02% Silwet. They were resuspended in glycerol and immediately observed by fluorescence microscopy (Microscope Eclipse E800; Nikon). For the GFP to be activated, the bacteria were exposed to blue light (488 nm).

TAIL-PCR method

The TAIL-PCR was a modified version of the method described by Liu and Whittier (1995). Primers used are listed in Table 5. The degenerate primers (AD1 to AD8; Table 5) were provided by P. Tornero (UNC, Chapel Hill, NC, USA). On average, three to five degenerate primers yielded one specific amplification product. The biggest products were ≈ 1 kb. PCR products were isolated on a 0.8% agarose gel (QIAquick gel isolation kit; Qiagen) and sequenced directly. The surrounding sequences of the avrPphE transposon in the chromosome of Psm M6 are available in GenBank (AF544992), as well as the sequences from Psm M5 (AF544991).


We thank Drs Jeff Chang, Petra Epple, Sarah R. Grant, Gopal Subramaniam and Ben F. Holt, III, and Zack Nimchuk for their patience and advice during the course of this work. We are especially grateful to The Institute for Genome Research (TIGR) and the DOE-Joint Genome Initiative (JGI) for making the genome sequences of P. syringae DC3000 and B728a available before publication. Our work is supported by DOE Division of Energy Biosciences grant DE-FG05-95ER20187.