In an independent study (M. Añorga, unpublished data), we observed the spontaneous generation of autonomously replicating deletion derivatives of pPsv48C lacking the RepA-PFP replicon (Figure 1B). The smallest derivative contains five putative coding sequences (CDSs), whose annotation is not related to plasmid replication, and a 661 nt fragment that appears in two nearly identical copies in pPsv48C (Figure S1). The second copy of this fragment precedes gene repA, matching the typical organization of RepA-PFP replicons, and was previously shown to be essential for replication of PFP plasmids (Murillo and Keen, 1994; Sesma et al., 1998; Gibbon et al., 1999; Sundin et al., 2004).

By cloning diverse PCR fragments into the E. coli vectors pK184 or pKMAG (Figure S2), which do not replicate in Pseudomonas, we determined that a 1,375 nt fragment (coordinates 29,386–30,760 in the pPsv48C sequence, accession no. FR820587; fragment B-F, Figure 1A and Figure S1), comprising around half of the 661 nt repeated fragment, contained all the essential elements for autonomous replication in the plasmidless strains P. syringae pv. syringae B728a (Figure 1B) and P. syringae pv. savastanoi UPN912. This fragment did not contain any obvious direct repetitions reminiscent of iterons, but was rich in palindromic structures and could adopt a complex folding structure. We could distinguish two well-defined structural regions in this minimal replicating fragment; based on their conservation and functionality (see below), we have defined these regions as plasmid replicon exchangeable (REx) modules: the Rex-C module contains the putative replication control system, whereas the Rex-R module comprises the replication system.

The absolute requirement of SaL 1 and 2 for replication can be overcome when repJ is transcribed from the strong Plac promoter, suggesting a plausible role of these structures in directing repJ transcription. We thus examined transcription of this gene by RT-PCR in clones lacking SaL 1, 2 and 3. Using clone AG in pKMAG (Figure 1C, upper gel), we observed a long repJ transcript that extended at least to the annealing site for primer C, but not to that for primer B. This indicates transcription from a promoter situated between the annealing sites for primers A and C (Figure S1), overlapping the putative antisense RNA, and possibly involved in transcription of repI and repJ. Amplification from smaller clones showed that repJ was transcribed even in the absence of SaL 1, 2 and 3 (fragments CG and DG, Figure 1C lower gel). Since the vectors used contain a T4 transcriptional terminator upstream of the cloned fragments, this shows that there is an additional active promoter immediately upstream of repJ. These results indicate that the REx-C module contains at least two functional promoters for the transcription of gene repJ and that the failure of fragments CG and DG to replicate is not due to a lack of transcription of repJ, suggesting an additional role for SaL 1 and 2.

Blast comparisons revealed that a region of up to 2,785 nt containing the minimal RepJ replicon (coordinates 28,783–31,067 in FR820587) is syntenic, with very high identity, in diverse genomoespecies of the P. syringae group as well as, with less identity, in a few other pseudomonads (not shown). Most of the homologs are from draft genomes and it is not possible to clearly determine if they localize to the chromosome or to plasmid sequences. Sequence variation among the homologs from P. syringae was not distributed randomly (Figure S3): whereas the stems from SaL 1, 2, and 3 were identical in all sequences, there was a high sequence variation in the loop of SaL 2, in the 3′ end of gene repJ and in SaL 4 and 5, downstream of this gene. As it occurs with the repA gene (Gibbon et al., 1999), the nucleotide variation in repJ (Figure S3) leads to a higher degree of variation in the C-terminal end of the deduced product; the phylogeny of a selection of these products is shown as clade I in Figure 2.

Using blastp, we found RepJ homologs only within members of Pseudomonadales, mainly in species of Pseudomonas, and in a few Desulfovibrio spp. strains. An ML tree with selected sequences (Figure 2) grouped homologs in four well-defined clades; from this tree we can also infer that the RepJ replicon has recently moved horizontally among pathovars and species of the P. syringae complex and that certain strains contained two repJ homologs (e.g., P syringae. pv. fraxini CFBP5062 in clade I). The minimal RepJ replicon is moderately conserved and syntenic among homologs from clades I and II (Figure S4). Of note, the REx-C modules from the RepJ replicons from clade I were more similar among themselves than to the modules from RepA-PFP replicons, indicating that these RepJ replicons are long-time inhabitants of pseudomonads. Conversely, the REx-C module is not conserved among members in clades III and IV whereas the region downstream of repJ shows a degree of conservation (Figure S4); additionally these homologs show a genetic organization that is conserved among clade members, but different than that from members of clades I and II.

A 2,808 nt fragment (positions 35,079–37,886, accession no. KB644113) containing the repJ homolog from the chromosome of strain NCPPB 3335 (PSA3335_1080, clade III; see Figure S4) cloned in pKMAG did not generate any transformant after electroporation into P. syringae pv. syringae B728a or P. syringae pv. savastanoi UPN912, whereas in the same experiments we obtained hundreds of clones using the minimal RepJ replicon from pPsv48C. These results suggest that the chromosomal homolog from strain NCPPB 3335 is not able to sustain autonomous replication.

Blastn comparisons showed that the REx-C module is present, with varying degrees of conservation, preceding the rep genes from at least four non-homologous replicon families from Pseudomonas (Figure 3; Table S1), whereas nucleotide identity is rapidly lost shortly after the CDS start codon. These families include the RepA-PFP family from P. syringae and other bacteria (exemplified by repA from pPsv48C) (Bardaji et al., 2011), the RepJ family (among others, repJ from pPsv48C and pA506) (Bardaji et al., 2011; Stockwell et al., 2013), and what we designated the RepA-RA2 family (named after pRA2) (Kato and Mizobuchi, 1994) and the RepA-Pa family (named here after the P. antarctica PAMC 27494 plasmid pP27494_2). The highest level of sequence conservation of the REx-C module among representatives of the four families is in a 175–194 nt fragment that spans SaL structures 1–3 (Figure S5). The stem sequences from SaL 1 (the putative transcriptional terminator for the antisense RNA), as well as the stretch of adenines in either side of them (Gibbon et al., 1999), are almost perfectly conserved and changes in one arm of the stems are usually compensated with complementary changes in the other arm (Figure S5) (Gibbon et al., 1999). The other structures are also well-conserved among the four replicon families; however, the number and position of palindromes, which could form SaL structures, are variable among replicons (Figure 3; Figure S5). Therefore, the stark conservation of the REx-C module suggests that it contains features universally essential for plasmid replication in species of Pseudomonas. Remarkably, we did not find REx-C sequences conserved in the control region of ColE2 replicons, which contain a replication initiator protein homologous to that from RepA-PFP replicons, or in any other plasmid outside of the genus Pseudomonas, indicating a diversity of REx-C modules among homologous theta replicons.

As described above, the REx-C modules accompanying genes repA and repJ from pPsv48C are nearly identical (Figure S1), although there is a gradient of conservation of this module along homologs of their corresponding gene families (Figure 3, Figures S3, S4) (Gibbon et al., 1999; Sesma et al., 2000; Stavrinides and Guttman, 2004). Additionally, the REx-C module preceding repJ from pA506 shows a very high degree of identity to that preceding repA from pMP-R124 (a PFP plasmid), whereas they are less similar to REx-C modules from replicons of their same family (Figure 3). These results show that there is frequent horizontal exchange of REx-C modules among RepA-PFP and RepJ replicons.

The results of multiple blast comparisons indicate that this exchange of REx-C modules also occurs with other replicon families. To illustrate this, we did a blast comparison against the non-redundant nucleotide collection (December 2016) of the 600 nt fragment immediately preceding the start codon for gene repA from the PFP plasmid pMP-R124, which comprises the REx-C module. The rep genes found immediately downstream of the first eight homologous sequences retrieved belong to the RepA-PFP, RepJ and RepA-RA2 families (Table S2). As before, and as an example, the REx-C module from pMP-R124 (RepA-PFP) shows a higher identity to pA506 (RepJ) than to the REx-C module from pPT14-32 (RepA-PFP) (Figure 3; Table S2). Incidentally, pMP-R124 and plasmids from the RepA-RA2 and RepA-Pa families also have a shorter REx-C module lacking the putative leader peptide gene sequence (Figure 3 and Figure S5). Together, these results indicate that the REx-C modules from these four families evolved vertically within the family, but were also subjected to horizontal exchange between members of other replicon families. Likewise, they indicate that plasmids from these four families have similar mechanisms of initiation of replication and its control.

Our comparative analyses of extant sequences (Figure 3) suggest that REx-C modules are freely exchanged among different replicons. However, previous works postulated that compatibility of coexisting RepA-PFP replicons was due to specificity between the C-terminal part of RepA and the loop sequence of SaL 2 (Figures S3, S6), which are highly variable and could coevolve for complementarity (Murillo and Keen, 1994; Gibbon et al., 1999; Stavrinides and Guttman, 2004; Ma et al., 2007). Additionally, the same pattern of variation is seen in a comparison of RepJ replicons (Figure S3). We therefore tested this putative specificity by swapping the respective SaL structures from the REx-C module (SaL fragment in Figure 4) and the rep fragments (partial repI plus REx-R, see Figure 4) from plasmids p1448A-B (RepA-PFP) and pPsv48C (RepA-PFP and RepJ) (Figure 4 and Figure S6). These two RepA-PFP initiator proteins are 88% identical (92% similar; Table S1 and Figure S6), while they are not homologous to RepJ (Table S1). Likewise, the three replicons show a different loop in SaL 2 and 1–3 nt changes within SaL 3 (Figure S6). In spite of the differences in sequence, all the analyzed chimeras were able to sustain autonomous replication in the plasmidless strain P. syringae pv. syringae B728a (Figure 4), indicating a lack of specificity between the SaL structures from the REx-C module and the rest of the replicon.

To test whether or not sequence variations in SaL 1 (Figures S1, S6) would be specific for each Rep protein, we evaluated chimeric clones containing the SaL fragment from plasmid pPsv48A and the rep fragments of the RepA-PFP or RepJ replicons from pPsv48C. These two RepA-PFP proteins, from pPsv48A and pPsv48C, are 97% identical (Table S1 and Figure S6), but the pPsv48A REx-C module (and consequently the SaL fragment tested) is comparatively shorter and with several changes in SaL 1 and 2. In spite of the differences, both chimeras replicated autonomously in strain B728a (Figure 4), indicating that sequence variations in the loop sequence from SaL 1, and in the sequence upstream of this structure, do not significantly impact replication ability.

Finally, none of the four SaL fragments cloned in pKMAG generated any transformants when transferred to strains B728a or 1448A (the latter containing two plasmids that could supply the RepA initiator in trans), suggesting that the REx-C module cannot sustain replication by itself.

The REx-C module showed no apparent specificity with the associated Rep protein and is readily exchanged among replicons (Figures 3, 4), serving as a portable putative replication control region. Therefore, it is feasible that REx-C modules could move within bacterial genomes, co-opt genes with the appropriate characteristics and thereby directly create new autonomous replicons. To broadly test this, we examined the effect of the SaL structures from the RepA-PFP and RepJ replicons from pPsv48C on the replication ability of gene PSA3335_1080. This gene is located in the chromosome of P. syringae pv. savastanoi NCPPB 3335 and its deduced product shows 72.7% identity (81.2% similarity) with that from repJ (Figure 2 and Figure S6; Table S1), but does not replicate autonomously (see above).

We therefore amplified and cloned a 1,772 nt fragment containing gene PSA3335_1080 plus the upstream 114 nt, to preserve the RBS and a similar spacing to the SaL structures as with repJ, and 815 nt downstream, spanning sequences homologous to SaL 5 (Figure 4). This fragment was ligated in the proper orientation to a partial fragment of the REx-C module containing the SaL structures preceding either repA or repJ from pPsv48C (279 nt), or preceding repA from pPsv48A (203 nt). The three constructions replicated in P. syringae pv. syringae B728a with very high efficiency (Figure 4). Disruption of the PSA3335_1080 reading frame by filling-in an internal restriction site abolished the replication ability of the clone containing the SaL fragment from the RepA-PFP replicon of pPsv48C. These results indicate that acquisition of a REx-C module can immediately confer autonomous replication ability to sequences containing PSA3335_1080, and that replication is dependent on the activity of this gene.

Their frequent exchange (Figure 3), and our experiments with chimeras (Figure 4), indicate that the REx-C modules have a low, or no specificity for their cognate REx-R module and that they probably confer a strong selective advantage. Among other possibilities, we speculated that the exchange of these modules could be a way to reduce or evade incompatibility, especially for coexisting plasmids carrying related replicons, such as the PFP group. The ColE2 replicon, related to the RepA-PFP replicons (Gibbon et al., 1999), contains two incompatibility determinants, corresponding to the control region and to the origin of replication (Tajima et al., 1988; Hiraga et al., 1994). Importantly, the RepA-PFP replicons contain sequences highly similar to the origin of replication of ColE2 replicons (Figure S7) (Yagura et al., 2006), including the primer RNA sequence (AGA), and located either within the RepA coding sequences or, for pMP-R124, situated some 300 nt after the rep gene stop codon.

We thus evaluated possible changes in the incompatibility behavior of chimeric replicons by transforming strain P. syringae pv. phaseolicola 1448A with the native RepA-PFP replicons from plasmids p1448A-B (B) and pPsv48C (C), and with their two corresponding chimeras of SaL and rep fragments (combinations B-C and C-B of the SaL-rep fragments; see Figures 4, 5). Strain 1448A contains naturally the native RepA-PFP plasmids p1448A-A (132 kb) and p1448A-B (52 kb). As expected, the native replicon from pPsv48C did not show any obvious incompatibility, producing a large number of transformants and coexisting with the two native plasmids from strain 1448A (Figure 5, lane C–C). The same results were observed with the C–B chimera (Figure 5, lane C–B), containing the SaL structures from plasmid C and the rep fragment from plasmid B. Conversely, the native replicon from plasmid B (B–B in Figure 5) and the B–C chimera generated about half the number of transformants than the two other replicons tested, taking double the time to reach colonies of the same size. Incompatibility mediated by the native replicon from plasmid B and the B-C chimera was evident in plasmid profile gels, where they either cointegrated with p1448A-B or induced its loss, or appeared with an apparently reduced copy number (Figure 5). These results indicate that strong replicon incompatibility between RepA-PFP replicons is associated to the REx-C module and, unlike what happens with ColE2 replicons (Tajima et al., 1988; Yasueda et al., 1994), not to the Rep protein or the origin of replication and that it can be overcome by only a few nucleotide changes in this module.

Bacterial strains and plasmids used in this work are detailed in Table S3. E. coli strains DH10B and GM2929 (dcm⁻, dam⁻), when unmethylated DNA was needed, were used for DNA manipulations and were grown in LB at 37°C. Strains P. syringae pv. phaseolicola 1448A (Joardar et al., 2005), pv. syringae B728a (Feil et al., 2005), and pv. savastanoi NCPPB 3335 (Rodríguez-Palenzuela et al., 2010) and UPN912, which derives from strain NCPPB 3335 and is cured of its three native plasmids (M. Añorga, unpublished results), were propagated using King's medium B (King et al., 1954) at 25°C. When necessary, media were supplemented with 100 μg ml⁻¹ ampicillin or 25 μg ml⁻¹ kanamycin. We used a mixture of plasmids pME6031 (8.3 kb), pBBR1MCS-2 (5.4 kb), pKMAG-C (4.3 kb) and pBlueScript II (3.0 kb) (Table S3), purified from E. coli using the Illustra plasmidPrep Mini Spin kit (GE Healthcare, UK), as size markers in plasmid profile gels.

DNA was amplified using a high fidelity enzyme (PrimeStar HS, Takara Bio Inc., Japan) and cloned using the CloneJET PCR Cloning Kit (Thermo Scientific) or the pGEM-T Easy Vector System (Promega), following the manufacturer's instructions. For plasmid profile gels, DNA was purified by alkaline lysis and separated by electrophoresis in 0.8% agarose gels with 1xTAE as described (Murillo et al., 1994). Plasmids were transferred to P. syringae by electroporation (Choi et al., 2006). For RT-PCR analysis, DNA-free RNA was obtained from bacterial cultures grown overnight in medium B using TriPure Isolation Reagent (Roche Diagnostics) and Ambion TURBO DNA-free Kit (Life Technologies). Concentration and purity of RNA were determined spectrophotometrically, and its integrity confirmed by electrophoresis in agarose gels. cDNA was synthesized from RNA using the ImProm-II reverse transcriptase system (Promega), following the manufacturer's recommendations. Primer Int_R, 5′-GCCGGTGCAGAGATACCC-3′, specific for the sense transcript of repJ, was used for the reaction: at 25°C for 5 min for primer annealing, 60 min at 42°C for reverse transcription, and 15 min at 70°C for enzyme inactivation. cDNA was amplified using primers B (5′-CGATGTAGATTCACGAATCGCAG-3′), C (5′-CTGATTATGGCGTTCACTGC-3′), or D (5′-TGCAAGCTGTCTAAAGTGAAGC-3′) together with 1R (5′-GCTGTTGTTCAGAGAGATGACG-3′), to determine the size of the transcript, or primer pair Int_F (5′-GAGAAGTTTCTGGCCATCGAG-3′) and Int_R (5′-GCCGGTGCAGAGATACCC-3′) to analyze the transcriptional activity of gene repJ (see Figure 1). The program used comprised 30 cycles (94°C for 30 s, 58°C for 30 s, and 72°C for 30 s) plus a final extension step of 6 min at 72°C. Control reactions included PCR amplification of pure extracted RNA, to verify the absence of contaminating DNA; amplification of purified DNA, to verify the reaction conditions, and amplification of an internal fragment of gene gyrB, to confirm the synthesis of cDNA.

Multiple-sequence alignments using Muscle, determination of the optimal substitution model, and Maximum likelihood phylogenetic tree construction using the JTT matrix-based model with a gamma distribution with 5 categories were done using MEGA7 (Kumar et al., 2016); confidence levels of the branching points were determined using 500 bootstraps replicates. Searches for sequence similarity in the NCBI databases were done using the BLAST algorithms (Hubbard et al., 2008) and sequences were aligned on-line using the MULTALIN program (Corpet, 1988) or the tools in the EMBL-EBI server (http://www.ebi.ac.uk/Tools/msa/). Search for protein motifs and fold recognition was done using the InterPro (Mitchell et al., 2015) (http://www.ebi.ac.uk/interpro/) and the Phyre2 (Kelley et al., 2015) web servers. Genome and nucleotide sequences were visualized and manipulated using the Artemis genome browser; when necessary, blast comparisons were visualized with ACT (Carver et al., 2008). Oligonucleotide primers were designed using Primer3plus software (Untergasser et al., 2012). Promoters were predicted using the online BPROM server (http://www.softberry.com). DNA or RNA Folding predictions were done using the Mfold web server (Zuker, 2003) using the default settings, except for a folding temperature of 25°C. Patterns of nucleotide polymorphism were calculated using DnaSP v 5.10.01 (Librado and Rozas, 2009).

The E. coli pKMAG vector and the E. coli-Pseudomonas pKMAG-C vector were constructed to examine, respectively, the replication ability of cloned fragments and the expression of repJ in these fragments. To construct pKMAG, a PCR-amplified fragment from the pK184 vector (positions 71-2,064; accession no. U00800) (Jobling and Holmes, 1990), retaining the kanamycin resistant gene and the origin of replication and lacking the lacZ promoter, and a fragment from pME6041 (Heeb et al., 2000) (positions 3,809–4,255; accession no. AF118812), containing its polylinker and the T4 transcription terminator, were ligated together resulting in the new vector. The RepA-PFP replicon from pPsv48C (positions 41,791–1,428; accession no. FR820587) was then cloned into the unique AscI site of pKMAG, resulting into pKMAG-C. Construction details are included in Figure S2.

Fragments used for the definition of a minimal RepJ replicon were amplified by PCR from strain Psv48ΔAB (Bardaji et al., 2011), a derivative of strain NCPPB 3335 containing only pPsv48C, cloned into pJET1.2 (Thermo Scientific) and then subcloned into pK184 in both directions, pKMAG and/or pKMAG-C. Primers used for amplifications were A (5′-AAAGCAGCGGATTTTGTAGG-3′), B, C, and D (described above), as forward, and E (5′-GACGCTAGGAGCCTATCCAG-3′), F (5′-TCCCTGTTTTTCCTGAAAGG-3′), and G (5′-GGTCGAACCGACCAACTG-3′), as reverse primers (see Figure 1 and Figure S1). For the construction of chimeric replicons we first cloned in pKMAG amplicons containing the RepA-PFP and RepJ replicons, including the REx-C module in a 390–391 nt upstream fragment, from plasmids pPsv48C (coordinates 41,714–1,336, for RepA-PFP, and 23,309–31,027, for RepJ, from accession no. FR820587) and p1448A-B (coordinates 51,624–1,650 from accession no. CP000060). The SaL (coordinates 78,056–78,262, from accession no. FR820585) and the rep (coordinates 78,254–1,360, from FR820585) fragments from pPsv48A were amplified by PCR and cloned separately in pJET1.2 to reconstruct an XmnI site present in many RepA-PFP replicons and missing from pPsv48A, resulting in a TC change in pos. 78,263. A double digestion XmnI-EcoRI or XmnI-XhoI of the resulting clones liberated the rep genes (rep fragment), leaving a 203–280 nt fragment attached to pKMAG that contained SaL structures 1 to 3 from the REx-C module (SaL fragment). The SaL and rep fragments were then separated by electrophoresis, purified and ligated in the appropriate combinations for the construction of chimeras. A 2,808 nt amplicon containing PSA3335_1080 (35,079–37,886 from accession no. KB644113), including 1,150 nt upstream and 815 nt downstream of the CDS, was cloned in pKMAG to evaluate replication of this gene in its native configuration. Additionally, an amplicon similar to the previous one but lacking the first 1,036 nt (36,115–37,886 from accession no. KB644113), and with an AT change in pos. 36,115, was ligated in the proper orientation to the appropriate XmnI-EcoRI clones in pKMAG generated above and containing the SaL fragments from repA or repJ from pPsv48C, or from pPsv48A. For functional analyses of replication initiation protein genes, replicative constructions containing repJ and PSA3335_1080 were digested using a unique restriction site internal to each CDS, filled-in with Klenow enzyme (New England BioLabs Inc., UK) and subsequently religated. Enzymes used for disrupting the CDSs by changing the reading frame were SexAI, adding 5 nt to the repJ CDS, and EcoNI, adding 1 nt to the CDS of PSA3335_1080. For all replication experiments, at least two amplicons from two separate amplification experiments were cloned and tested; the identity and integrity of all clones was confirmed by sequencing. Replication ability was assessed by electroporation into the plasmidless strains P. syringae pv. syringae B728a and UPN912. Plasmids were confirmed to replicate autonomously by their ability to generate antibiotic-resistant transformants and, necessarily, by their appearance as independent bands in undigested plasmid profile gels, basically as described (Murillo and Keen, 1994; Sesma et al., 1998). Experiments were repeated at least three times, with similar results.

Incompatibility between native plasmids and cloned replicons was analyzed essentially as described (Nordström, 1993; Sesma et al., 1998). Briefly, constructs in pKMAG containing the native RepA-PFP replicons from plasmids pPsv48C (strain P. syringae pv. savastanoi NCPPB 3335) and p1448A-B (strain P. syringae pv. phaseolicola 1448A) and the corresponding chimeras of SaL and rep fragments, were isolated from E. coli and individually transferred to P. syringae pv. phaseolicola 1448A, which naturally contains the native plasmids p1448A-A (132 kb) and p1448A-B (52 kb) (Joardar et al., 2005). Resulting km^R transformants of strain 1448A were cultured overnight in liquid B medium with kanamycin at 25°C with shaking, and their plasmid content visualized by electrophoresis in plasmid profile gels to determine the possible eviction of native plasmids.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.