Excision and transfer of the Mesorhizobium loti R7A symbiosis island requires an integrase IntS, a novel recombination directionality factor RdfS, and a putative relaxase RlxS


*E-mail Clive.Ronson@stonebow.otago.ac.nz; Tel. (+64) 3479 7701; Fax (+64) 3479 8540.


The Mesorhizobium loti strain R7A symbiosis island is an Integrative Conjugative Element (ICE), herein termed ICEMlSymR7A, which integrates into a phetRNA gene. Integration reconstructs the phetRNA gene at one junction with the core chromosome, and a direct repeat of the 3-prime 17 bp of the gene is formed at the other junction. We show that the ICEMlSymR7AintS gene, which encodes an integrase of the phage P4 family, is required for integration and excision of the island. Excision also depended on a novel recombination directionality factor encoded by msi109 (rdfS). Constitutive expression of rdfS resulted in curing of ICEMlSymR7A. The rdfS gene is part of an operon with genes required for conjugative transfer, allowing co-ordinate regulation of ICEMlSymR7A excision and transfer. The excised form of ICEMlSymR7A was detectable during exponential growth but occurred at higher frequency during stationary phase. ICEMlSymR7A encodes homologues of the traR and traI genes of Agrobacterium tumefaciens that regulate Ti plasmid transfer via quorum sensing. The presence of a plasmid with cloned island traR traI2 genes resulted in excision of ICEMlSymR7A in all cells regardless of culture density, indicating that excision may be similarly regulated. Maintenance of ICEMlSymR7A in these cells depended on msi106 (rlxS) that encodes a putative relaxase. Transfer of the island to non-symbiotic mesorhizobia required intS, rlxS and rdfS. The rdfS and rlxS genes are conserved across a diverse range of α-, β- and γ-proteobacteria and identify a large family of genomic islands with a common transfer mechanism.


Comparative analysis of prokaryotic genomes has revealed that related but phenotypically distinct chromosomal lineages often share a common core of genetic information interspersed with strain-specific regions (Ochman et al., 2000). These regions often represent DNA acquired by horizontal transfer, introduced through the actions of mobile elements including prophage, plasmids, transposons and genomic islands (Frost et al., 2005). Genomic islands are located on the chromosome and often carry genes that endow a selective advantage on the host bacterium in a specific environment. They can be further categorized into subtypes including pathogenicity islands (PAIs) which encode virulence determinants, catabolic islands which contain xenobiotic degradation pathways, resistance islands which confer antibiotic resistance and symbiosis islands which provide genes required by a bacterium to establish a symbiotic interaction with a eukaryotic host (Burrus et al., 2002; van der Meer and Sentchilo, 2003; Burrus and Waldor, 2004; Dobrindt et al., 2004). Consistent with their external origin, many genomic islands contain phage-related integrases and are integrated adjacent to stable RNA (tRNA or tmRNA) genes, with their boundaries demarcated by direct repeat sequences comprising part of the RNA gene. However, the majority of genomic islands appear to have become permanently anchored in the host genome, having lost the ability to transfer as a result of reductive evolution since their acquisition (Dobrindt et al., 2004).

Recently it has been suggested that a subset of mobile genetic elements including conjugative transposons, integrative plasmids and mobile genomic islands should be grouped under a new classification termed Integrative and Conjugative Elements or ICEs (Burrus et al., 2002). ICEs are defined as elements that excise from their host chromosome in a site-specific manner, leading to formation of a circularized element that is generally transient. Conjugative transfer initiating from the circularized form is followed by integration of the element into the recipient chromosome. Integration and excision are facilitated by an integrase or recombinase encoded on the element and a tRNA gene is often utilized as the integration target site (Burrus et al., 2002). Mobile genomic islands that may be considered ICEs include the SXT element of Vibrio cholerae which encodes resistance to several antibiotics (Hochhut and Waldor, 1999; Burrus et al., 2006), the clc element of Pseudomonas strain B13 which contains genes required for chlorocatechol degradation (Ravatn et al., 1998; Gaillard et al., 2006) and related islands in Haemophilus influenzae (Mohd-Zain et al., 2004), and the symbiosis island of Mesorhizobium loti strain R7A which contains genes required to form a nitrogen-fixing symbiosis with the legume Lotus corniculatus (Sullivan and Ronson, 1998; Sullivan et al., 2002).

The M. loti strain R7A symbiosis island, herein termed ICEMlSymR7A in accordance with the nomenclature proposed by Burrus et al. (2002), is a 501.8 kb ICE which is transferable to non-symbiotic mesorhizobia. The island integrates into a phenylalanine-tRNA (phetRNA) gene leading to reconstruction of the phetRNA gene at the left junction between R7A core chromosome and ICEMlSymR7A and the formation of a direct repeat of the 3′-terminal 17 bp of the phetRNA gene at the right junction (Sullivan and Ronson, 1998). Annotation of the ICEMlSymR7A sequence revealed 414 genes with predicted roles in nodulation, nitrogen fixation, secretion, metabolism and island transfer among others (Sullivan et al., 2002).

An integrase-encoding gene intS is the first ORF located in from the left end of ICEMlSymR7A and is likely to mediate integration of the island into the phetRNA gene. Sequence analysis indicates that IntS is a tyrosine recombinase that belongs to the satellite bacteriophage P4 integrase family (Esposito and Scocca, 1997; Sullivan and Ronson, 1998), a family of integrases that usually target the 3′ end of tRNA genes for integration (Williams, 2002). Integrases catalyse site-specific recombination between short identical DNA sequences, allowing integration of a circular DNA element through single-crossover. They simultaneously bind short ‘core’ sequences at the site of recombination and flanking ‘arm’ sequences within regions termed attachment sites (att). Core sequences are found within the attachment site (attP) of the circular element, the integration site on the host chromosome (attB) and as a direct repeat flanking the integrated element at the element/chromosome junctions attL and attR (reviewed by Groth and Calos, 2004).

Recombination by integrases is highly directional and favours integration (the formation of attL and attR) in the absence of additional factors. In most cases the excision function of an integrase is stimulated by the presence of a recombination directionality factor (RDF), also termed an excisionase, that binds and bends DNA within the attachment sites to promote excisive recombination (Sam et al., 2004). RDFs are small (generally < 100 amino acids) proteins that are usually basic, and some contain a helix–turn–helix DNA-binding motif. They exhibit considerable variation in amino acid sequence, including between members belonging to the same family (Lewis and Hatfull, 2001).

P4-family integrases have been shown to be essential for integration and excision of several ICEs including the clc element (Ravatn et al., 1998), the high-pathogenicity island (HPI) of Yersinia enterocolitica and Escherichia coli (Lesic et al., 2004), the she and SRL PAIs of Shigella flexneri (Luck et al., 2001; Sakellaris et al., 2004) and several PAIs in the uropathogenic E. coli strain 536 (Hochhut et al., 2006). RDFs that act in association with P4-family integrases to promote excision have been discovered for phage P4 (Cali et al., 2004), cryptic phage CP54 (Kirby et al., 1994), and recently the she, SRL and HPI PAIs (Lesic et al., 2004; Luck et al., 2004; Sakellaris et al., 2004). These results suggest that P4-family integrases have associated RDFs, but RDFs have not been identified for the clc element or the PAIs in E. coli strain 536.

The aim of this study was to identify the genes required for the integration and excision of ICEMlSymR7A. The requirement of IntS for integration and excision was confirmed and a novel RDF, Msi109 (RdfS), identified. Evidence was obtained that excision of ICEMlSymR7A was increased in stationary phase and was under quorum-sensing control. Finally a putative novel relaxase Msi106 (RlxS) was found to be required for maintenance of the island in its excised form.


Excision of ICEMlSymR7A increases in stationary phase and is dependent on IntS

Integrative and Conjugative Elements are predicted to excise from the chromosome to form a circular intermediate prior to conjugative transfer to recipient cells (Burrus et al., 2002). We confirmed by PCR that ICEMlSymR7A excised to give a circular form and that the host genome was repaired upon excision, using primers pairs LE1/RE2, and LE2/RE1 (Table S1) that face outward from the integrated island or anneal to chromosomal DNA flanking the island insertion site respectively (Fig. 1A). PCR products were obtained for both primer pairs (Fig. 2, strain R7A) and sequence analysis confirmed the expected structures of attP and attB (data not shown).

Figure 1.

Location of primers used to detect ICEMlSymR7A excision and integration. ICEMlSymR7A DNA is shown as a black bar or circle, chromosomal DNA is shown by white bars and vector DNA is hatched. Attachment sites are illustrated by black or white boxes representing their composite structure. Location of primers used for standard PCR are shown by small arrows, while QPCR amplicons are shown as dumbbells above the target sequence. Genes are shown by thick arrows. Maps not to scale.
A. Model for ICEMlSymR7A excision and integration.
B. Map of pJR201 used as a control template for QPCR data.
C. Map of mini islands pJJ608 and pJJ609.

Figure 2.

PCR amplicons (indicated left) obtained from wild-type and mutant strains (named above). An attL amplicon was not obtained from strain R7AΔintS as the ΔintS deletion removed the priming site for LE1. The msi109 amplicon obtained with strain R7AΔrdfS is smaller than that obtained with strain R7A, due to the 135 bp deletion in the mutant. Note that rdfS was previously called msi109 (see text).

To measure the frequency of excision of ICEMlSymR7A, quantitative PCR (QPCR) assays using TaqMan® probes were developed to detect the attB (unoccupied integration site) and attP (circular form of ICEMlSymR7A) sequences within DNA samples (Table S1, Fig. 1A and B). The results from both assays were normalized using results from a third assay that detected melR, the chromosomal gene located immediately upstream from attB, to measure the number of chromosomes within each DNA sample. Amplification efficiencies, determined using linearized DNA of the plasmid pJR201 that contains a single copy of each of the target sequences (Fig. 1B, Table 1) as a template, were 1.92, 1.90 and 1.81 for the attB, attP and melR amplicons respectively. These values are close to the optimal efficiency of 2.00.

Table 1.  Bacterial strains and plasmids.
 R7AField reisolate of ICMP 3153; wild-type symbiotic strainSullivan et al. (1995)
 R7ANSNon-symbiotic derivative of R7A; lacks ICEMlSymR7AThis study
 CJ4Non-symbiotic Mesorhizobium strain; field isolateSullivan et al. (1996)
 N18Non-symbiotic Mesorhizobium strain; field isolateThis laboratory
 R7AΔintSΔintS::nptII, NmR; gene replacement deletion mutant of intSThis study
 R7AΔmsi109Δmsi109; markerless in-frame deletion mutant; mutant renamed as R7AΔrdfS when function of Msi109 shownThis study
 R7Amsi106pFUS2 insertion mutant of msi106; mutant renamed as R7ArlxS when function of Msi106 shownThis study
E. coli
 S17-1pro recA RP4-2(TcS::Mu)(KmS::Tn7); Mob+Simon et al. (1983)
 pFAJ1700Broad-host-range plasmid, TcRDombrecht et al. (2001)
 pFAJ1708pFAJ1700 containing nptII promoterDombrecht et al. (2001)
 pFUS2oriCColE1oriTRK2lacZ transcriptional reporter; suicide vector, GmRAntoine et al. (2000)
 pLAFR1Broad-host-range plasmid, TcRFriedman et al. (1982)
 pPH1JIIncP plasmid, GmR CmR SmR SpRHirsch and Beringer (1984)
 pPR691Low-copy-number cosmid derived from pSC101, KmR SpR SmRJiang et al. (1987)
 pJQ200SKSuicide vector containing sacB gene, GmRQuandt and Hynes (1993)
 p239pUC19 containing nptII gene from Tn5Sullivan et al. (2001)
 pJJ607pPR691 with cos sites removed and replaced with RK2 oriT from pFUS2This study
 pJJ608pJJ607 containing attP and intS amplified from R7A genomic DNA using primers CIB5 and CIB3 and cloned as BamHI fragmentThis study
 pJJ609pFUS2 containing attP-intS BamHI fragment from pJJ608This study
 pJJ610pFUS2 containing BamHI-PstI attP fragment from pJJ608. The PstI site is 4 bp upstream of the intS start codon.This study
 pJJ611pFAJ1708 containing intS amplified using primers IntnptIIL and IntnptIIR and cloned as BamHI-EcoRI fragment downstream of nptII promoterThis study
 pJS101pUC19 containing 6.145 kb EcoRI fragment containing attL and intSThis study
 pJS102pJS101 ΔintS::nptIIThis study
 pJS103pLAFR1 containing insert from pJS102This study
 pJS112pFUS2 containing 364 bp intragenic region of msi106 amplified using primers msi106L and msi106RThis study
 pJK301pJQ200SK containing two PCR products which flank msi109 (rdfS) amplified using primer pairs 109LL, 109LR and 109RL, 109RRThis study
 pJR201pUC8 containing PCR products spanning attP, attB and a region of melR amplified using primer pairs RE2/LE1, LE2/RE1, and ML/MR, respectivelyThis study
 pJR202pFAJ1700 containing msi109 (rdfS) and promoter region amplified using primers 109L and 109CN3This study
 pJR204pFAJ1708 containing msi109 (rdfS) amplified using primers 109CN5 and 109CN3This study
 pJR206pFAJ1700 containing traR traI2 amplified using primers traRIL and traRIRThis study

The effect of growth phase on the excision of ICEMlSymR7A was investigated by QPCR of DNA prepared from broth cultures that were grown with shaking and sampled at 8 h intervals until late-stationary phase. Excision products were present at low frequency in exponentially growing cultures (0.08% and 0.04% for attP and attB respectively), but increased approximately 10- to 100-fold during stationary phase (Fig. 3).

Figure 3.

Effect of growth phase on level of ICEMlSymR7A excision. QPCR was used to determine the percentage of attP and attB sequences resulting from excision of ICEMlSymR7A. The percentages of cells containing attP are shown as black bars and attB as white bars. Triplicate measurements were obtained for each sample, and the mean ± standard deviation are shown for each assay. The growth curve is shown as a continuous line.

An intS deletion mutant designated strain R7AΔintS was constructed via marker exchange (see Experimental procedures). Template DNA extracted from cultures of R7AΔintS failed to produce amplification products in both the attP and attB PCR and QPCR assays (Figs 2 and 4), showing that excision required intS. The effect of constitutive production of IntS on excision was examined by introducing pJJ611 (nptIIp-intS) into strains R7A and R7AΔintS. Strain R7A(pJJ611) showed levels of excision in exponentially growing and stationary-phase cells similar to those of R7A containing the pFAJ1700 vector alone (Fig. 4). Surprisingly the levels of excision in R7AΔintS(pJJ611) were about 10-fold higher than in R7A in both phases of growth (Fig. 4). This result confirmed that excision was restored in R7AΔintS by providing intS in trans and also suggested that the ΔintS mutation may have some regulatory effect on excision.

Figure 4.

ICEMlSymR7A excision in various mutant and complemented strains. Triplicate measurements were obtained for each sample, and the mean ± standard deviation are shown for each assay. Percentage attP and attB at 24 h (exponential growth) are shown by black and white bars respectively, while percentage attP and attB at 64 h (stationary phase) are shown by black bars/white hatching and white bars/black hatching. Asterisks indicate samples with significant deviation from the wild-type data (*P < 0.05, **P < 0.01). Neither target was detected in R7AΔintS, and no attB was detected in R7AΔrdfS. At t = 24 for R7AΔrdfS, attP was detected in only a single sample and so no standard deviation or significance value was obtained.

intS and attP are required and sufficient for integration

To determine whether intS and attP were sufficient to direct integration of ICEMlSymR7A, the ability of a ‘mini-island’ pJJ608 (Fig. 1C), comprising a suicide plasmid containing just the attP region and intS, to integrate into the chromosome of the non-symbiotic Mesorhizobium sp. strain CJ4 was examined. pJJ608 was introduced into strain CJ4 by conjugative transfer from E. coli strain S17-1 and 15 exconjugant clones were analysed by Southern hybridization using an intS PCR product as probe. The results showed that pJJ608 had integrated into the same position on the chromosome in all cases (data not shown). Analysis of these strains by PCR using primers LE1 and LE2 (Fig. 1) and sequencing of the product confirmed that pJJ608 had integrated at the phetRNA locus in CJ4, as previously observed with ICEMlSymR7A (Sullivan and Ronson, 1998).

To confirm that the attP-dependent integration required the action of IntS, a suicide plasmid pJJ610 (GmR) containing the attP region but not intS (Table 1) was used. When E. coli strain S17-1 containing pJJ610 was used as the donor in a mating with strain CJ4 as recipient, no GmR exconjugants were detected. However, exconjugants were obtained when strain CJ4 containing pJJ611, which expresses intS constitutively, was used as recipient. Fifteen exconjugants were analysed by Southern hybridization using an intS-specific probe, as well as by PCR using primers LE2 and LE3. The data obtained confirmed that pJJ610 had integrated into the phetRNA gene of CJ4. Taken together, these results confirm the requirement for both the attachment site region attP and intS for integration at attB.

Efficient excision is dependent on the presence of a novel recombination directionality factor Msi109 (RdfS)

Most integrases require the presence of a RDF to stimulate excision. In order to find candidate RDFs on ICEMlSymR7A, we carried out blast searches of the symbiosis island sequence using the sequence of the bacteriophage P4 RDF, Vis, as the query. No obvious candidates were found using this approach so we used psi-blast (Altschul et al., 1997) to identify more distantly related RDF homologues from the NCBI database. The Vis sequence was used as the query, and sequences of between 70 and 150 residues were selected for submission in iterations. After the third iteration, the search results included the ICEMlSymR7A ORF msi109 that encodes an 89-amino-acid protein with limited similarity to Vis and related RDFs (Fig. 5). Vis is a member of the SLP1 subfamily of RDFs, all of which are associated with P4-family integrases (Lewis and Hatfull, 2001).

Figure 5.

clustalw alignment of the RdfS amino-acid sequence with similar RDFs that interact with P4-family integrases: Rox (gi:40949919) from the Shigella flexneri 2a she PAI, Vis (gi:1589630) of Enterobacteria phage P4 and Hef (Yp43) (gi:4106643) from HPI of Yersinia pestis.

msi109 is a member of a gene cluster msi110 – msi106 (Fig. 6) that Sullivan et al. (2002) speculated was involved in conjugative transfer on the basis of the similarity of msi108 (traF) and msi107 to genes known to have a conjugative role. A near-identical cluster is present on the putative symbiosis island of M. loti strain MAFF303099 except for the presence of a transposon inserted between the msi107 and msi106 orthologues in MAFF303099 (Kaneko et al., 2000; Sullivan et al., 2002). To determine if msi109 had a role in excision of ICEMlSymR7A, we constructed an in-frame markerless deletion mutant R7AΔmsi109 (see Experimental procedures). QPCR analysis of the mutant showed that attP was only present sporadically at low frequencies (< 0.003%) in independently derived templates extracted from exponential or stationary-phase cultures, while attB was not detected (Fig. 4). Complementation analysis using pJR202 that contains msi109 and the preceding 210 bp region (Table 1) was then carried out to confirm the role of msi109. The plasmid did not affect the excision frequencies relative to those observed with strain R7A(pFAJ1700), but produced excision frequencies approximately 100-fold higher than those observed for R7A when in the R7AΔmsi109 background, during both phases of growth (Fig. 4). Based on these results, we renamed msi109 as rdfS to reflect its role as the ICEMlSymR7A RDF.

Figure 6.

Organization of the msi110 – msi106 gene cluster on ICEMlSymR7A (data from Sullivan et al., 2002). msi109 and msi106 were renamed as rdfS and rlxS, respectively, during this work.

Bioinformatic analysis of RdfS

Potential orthologues of RdfS revealed in a blastp search were screened for ORFs with similar genetic context to that found in ICEMlSymR7A. Nucleotide sources of the 50 matching protein sequences with an E-value less than 10−3 were examined for the presence of homologues of traF (msi108) and msi106 within a 6 kb region. Homologues of both genes were found downstream of and in the same orientation as the rdfS orthologue in 40 cases. Of the remaining 10 excluded proteins, six were part of contigs from incomplete genomes that contained homologues of conserved hypotheticals found in either ICEMlSymR7A or Tn4371, while two homologues were found directly downstream of predicted P4-family integrases (gi:99078558, gi:84687547).

The 40 RdfS orthologues were aligned using clustalw (Fig. S1) and the alignment used to create a bootstrapped neighbour-joining tree (Fig. 7). The tree grouped RdfS with sequences from the north Atlantic Ocean isolate Sphingomonas SKA58 and Sargasso Sea isolate Parvularcula bermudensis HTCC2503 and these formed a distinct clade. The six RdfS orthologues from the putative genomic islands previously identified by their similarity to the conjugative transposon Tn4371 (Toussaint et al., 2003) formed a tight cluster together with several new additions (Fig. 7). The group of sequences ranged in length from 75 to 123 aa (average 93 aa) and all had a predicted basic isoelectric point (average 10.70 ± 0.72), consistent with the properties of other characterized and predicted RDFs.

Figure 7.

Neighbour-joining tree based on a clustalw amino-acid sequence alignment of RdfS homologues. The percentage of trees supporting each node are indicated (1000 trees total).

Constitutive expression of rdfS results in loss of ICEMlSymR7A

To provide constitutive expression of rdfS in trans, the rdfS coding sequence was cloned into pFAJ1708 to give pJR204, which was introduced into strain R7A by electroporation. Relatively few colonies appeared and the majority of these took 2 days longer to grow to 1 mm diameter on GRDM plates than colonies maintaining pFAJ1708 (which grew at the same rates as R7A colonies), suggesting that constitutive expression of rdfS caused a reduced growth rate. However, a small number of colonies (≤ 3 per plate) grew at the same rate as R7A(pFAJ1708) colonies. A similar result was observed when pJR204 was introduced into R7AΔrdfS. Introduction of pJR204 into R7AΔintS by either electroporation or conjugation resulted in only a small number of colonies that grew at the R7A rate and no slower-growing colonies. As colonies with wild-type growth rate were in the minority in each electroporation with pJR204 (excluding R7AΔintS), we tested the hypothesis that these cells contained mutated copies of pJR204 not expressing rdfS. Primers FAJ5 and FAJ3 (Table S1), which flank the nptII promoter and MCS in pFAJ1708, were used to amplify DNA from six R7A(pJR204) isolates. One had a deletion in the nptII promoter, another contained a mutated rdfS gene, and no product was obtained from the other four strains, thus supporting our hypothesis.

The above results showed that pJR204 caused a reduction in growth rate when introduced into wild-type or ΔrdfS cells, but was lethal in the R7AΔintS background. We hypothesized that lethality of pJR204 in R7AΔintS cells was due to the inability of the strain to excise ICEMlSymR7A and thus that the slow-growing R7A and R7AΔrdfS isolates had excised ICEMlSymR7A. This hypothesis was tested by QPCR analysis of DNA isolated from the slow-growing R7A cells containing pJR204. Unoccupied attB sites were present on 100% of chromosomes [Ct(attB) = Ct(melR)] and attP was undetectable, suggesting that these cells had lost ICEMlSymR7A.

Isolates cured of pJR204 were obtained as tetracycline-sensitive colonies with wild-type growth rates on G/RDM. DNA derived from one isolate (R7ANS) was screened by PCR for the presence of ICEMlSymR7A genes intS, nodB, thiC and rdfS, the attP site and the ICEMlSymR7A–chromosome junctions attL and attR. None of these assays produced PCR products, whereas PCR products of the expected size were obtained for melR and attB amplifications (Fig. 2 and data not shown). Phenotypic studies showed that R7ANS, in contrast to wild-type R7A, was unable to nodulate Lotus corniculatus or utilize succinate as a sole carbon source, and was auxotrophic for thiamin, biotin and nicotinate. Genes essential for the synthesis of these three vitamins and the dct genes required for transport of C4-dicarboxylic acids are located on ICEMlSymR7A (Sullivan and Ronson, 1998; Sullivan et al., 2001; 2002). These results confirmed that R7ANS was cured of ICEMlSymR7A.

Efficient excision requires only intS and rdfS

To determine if intS and rdfS were sufficient for island excision, the gentamicin-resistant mini-island pJJ609 (Table 1) was introduced into strain R7ANS. As found for pJJ608 in strain CJ4, pJJ609 integrated at high frequency into attB in strain R7ANS to give strain R7ANS::pJJ609 (Fig. 2). QPCR of DNA extracted from stationary-phase broth cultures revealed that excision was only observed sporadically in a low number of cells, with attP being present in less than 0.01% of cells, and attB in less than 0.003%. To test if constitutive expression of intS induced excision of pJJ609, pJJ611 was introduced into R7ANS::pJJ609. QPCR analysis revealed that pJJ611 increased the presence of attP to 0.12%, but attB was detected at 9.1%, suggesting that some loss of the mini-island was occurring in these cells. To test if expression of rdfS induced excision of pJJ609, pJR204 was introduced into R7ANS::pJJ609 cells by electroporation and the cells plated on medium containing tetracycline, with or without gentamicin. Colonies did not form on plates containing both antibiotics, but colonies were obtained on plates containing only tetracycline. PCR analysis showed that these isolates had lost pJJ609. Hence symbiosis genes other than rdfS and intS are not required for island excision, confirming that RdfS is the island RDF. Introduction of pJR202 into R7ANS::pJJ609 did not induce excision of the mini-island, suggesting that expression of rdfS from its native promoter may require activation by other genes encoded on ICEMlSymR7A.

Transfer of ICEMlSymR7A requires intS, rdfS and msi106 (rlxS)

To test whether R7AΔintS and R7AΔrdfS were able to transfer ICEMlSymR7A to non-symbiotic mesorhizobia, conjugation experiments were carried out using these strains as donors and the non-symbiotic Mesorhizobium sp. strain N18 containing the non-transferable plasmid pFAJ1700 (TetR) as recipient. Medium containing succinate as carbon source and lacking vitamins was used to select exconjugants that had received ICEMlSymR7A, and tetracycline was used to counterselect the donors. Exconjugants were also readily distinguished from R7A by growth rate. Wild-type R7A transferred the island at a frequency of between 9 × 10−5 and 3 × 10−4 exconjugants per donor, while no transfer of the island was detected from R7AΔintS or R7AΔrdfS (transfer frequency less than 10−10 exconjugants per donor).

The island gene msi106 was annotated as encoding a possible relaxase required for initiation of the rolling-circle replication involved in conjugative transfer, based on very limited similarity (25% identity over 22% of Msi106; E =  0.16) of its gene product to VirD2 from Agrobacterium tumefaciens, the relaxase required for T-DNA transfer (Pansegrau et al., 1993). The mutant strain R7Amsi106 was also unable to transfer ICEMlSymR7A to strain N18(pFAJ1700) at detectable frequency. To reflect its putative relaxase function, we have renamed msi106 as rlxS.

The ICEMlSymR7AtraR traI2 genes are involved in the regulation of island excision

The ICEMlSymR7A genes msi174 (traR) and msi173 (traI2) show similarity to the traR and traI quorum-sensing genes of A. tumefaciens that regulate Ti plasmid transfer in response to population density (Fuqua and Winans, 1994; Sullivan et al., 2002). As the frequency of island excision increased at high population density, it seemed possible that these genes were involved in the regulation of excision. To test this idea, pJR206 (Table 1) that contains the island traR traI2 genes and the preceding 696 bp was electroporated into strain R7A. The resultant strain caused intense purple colouration of the N-acyl homoserine lactone indicator strain Chromobacterium violaceum CV026 (McClean et al., 1997) when plated adjacent to it, whereas strain R7A caused only a faint thin line of purple colouration (data not shown). This suggested that strain R7A(pJR206) overproduced an N-acyl homoserine lactone compared with strain R7A, as might be expected if the traI2 gene was under autoregulation. QPCR assays of DNA from exponential and stationary-phase cultures of strain R7A(pJR206) showed that the island was excised in 100% of cells [Ct(attB) = Ct(melR)] under both growth conditions and that attP was present at a ratio of approximately 1.5:1 to attB (Fig. 4).

ICEMlSymR7A may replicate by rolling-circle replication in its excised form

As well as indicating that island excision was regulated by traR traI2, the above results suggested that ICEMlSymR7A must be able to replicate autonomously in its excised form. This was surprising as the island lacks obvious plasmid replication genes (Sullivan et al., 2002). To determine if rlxS (the putative island relaxase required for rolling-circle replication) was required for the maintenance of ICEMlSymR7A in its excised state, pJR206 was electroporated into strain R7ArlxS. Small colonies were obtained on media containing gentamicin and tetracycline, selecting for both ICEMlSymR7A(rlxS) and pJR206, but these could not be subcultured. Normal-sized colonies were obtained on medium with tetracycline alone. Analysis of one such isolate by PCR indicated that it had lost ICEMlSymR7A, and this was confirmed by phenotypic analysis (data not shown). These results indicate that maintenance of ICEMlSymR7A in excised circular form in the presence of pJR206 requires the rlxS gene product, consistent with RlxS being required for episomal replication of the island.


Excision of ICEMlSymR7A from the chromosome resulted in the formation of a circular form that likely acts as an intermediate for horizontal transfer via conjugation. We have shown that both integration and excision of ICEMlSymR7A are dependent on intS and that the rdfS gene product acts as a RDF required for efficient IntS-mediated excision. The chromosome was repaired upon excision, resulting in formation of the unoccupied attachment site attB and reconstitution of the phetRNA gene. The excised form of ICEMlSymR7A was detectable during exponential growth but was present at higher frequency during stationary phase. Evidence that excision may be regulated in a cell density-dependent manner was obtained by the finding that introduction of cloned island traR traI2 genes resulted in excision of ICEMlSymR7A in all cells. Maintenance of the excised form of ICEMlSymR7A in these cells was dependent on the rlxS gene product, the putative island relaxase, and transfer of the island required intS, rlxS and rdfS.

A phylogenetic analysis of homologues of RdfS identified by blastp analysis showed that they formed a distinct grouping from other RDF families defined by Lewis and Hatfull (2001). Within this grouping was a clade of 14 closely related proteins found in both γ-proteobacteria (e.g. Pseudomonas aeruginosa) and β-proteobacteria (e.g. Ralstonia species), indicating recent and possibly multiple horizontal transfer events between these two clades of proteobacteria. Other orthologues of RdfS comprise more divergent proteins from α-proteobacteria, β-proteobacteria and γ-proteobacteria. The α-proteobacterium Sphingomonas SKA58 contained the closest match to RdfS and this strain also encodes nearby a match to the ICEMlSymR7A integrase IntS with 60% amino acid identity, the highest similarity to IntS outside mesorhizobia. Like ICEMlSymR7A, this putative genomic island is integrated adjacent to a phetRNA gene and encodes a trb operon and several other conserved hypotheticals. A second more distantly related homologue of rdfS is found in Sphingomonas SKA58 within a second putative island located directly downstream of the first. This island also encodes copies of rlxS, traF and the trb gene cluster, indicating a tandem integration of two islands from different branches of the tree (data not shown).

rdfS falls within a cluster of genes (msi110 – msi106) on ICEMlSymR7A (Sullivan et al., 2002) and members of this cluster are also found in close proximity to rdfS homologues on the chromosomes of several proteobacteria (Fig. 7). The clustered genes on these chromosomes include homologues of traF (msi108), which encodes a protease that acts on the TrbC pilin protein, and rlxS, defined in this work as the likely ICEMlSymR7A relaxase. Some but not all of the clusters also contained homologues of msi107, the product of which has similarity to murein lytic transglycosylases. Many of these chromosomes also contained within reasonable proximity to the rdfS cluster, a trb gene cluster of the same organization (including homologues of conserved hypothetical genes msi031 and msi021) as that found on ICEMlSymR7A (Sullivan et al., 2002; Toussaint et al., 2003). We propose that, together, these gene clusters define a large family of conjugative genomic islands present in a wide range of environmental proteobacteria that share a common transfer mechanism, of which ICEMlSymR7A is the prototype as it was the first member of the family sequenced. As found for other families of genomic islands, the conserved transfer genes are interspersed with island-specific modules that encode adaptive functions. Our identification of RdfS as a RDF and RlxS as a relaxase will now allow the annotation of many currently hypothetical genes, and the identification of many putative genomic islands.

Constitutive expression of rdfS resulted in curing of ICEMlSymR7A, producing a non-symbiotic derivative of strain R7A, R7ANS. The loss of ICEMlSymR7A may be due to RdfS acting to force excisive integrase-mediated recombination and to inhibit integration, resulting in loss of the island in the absence of expression of genes required for rolling-circle replication. Unexpectedly, constitutive expression of rdfS was lethal in a ΔintS strain and caused a reduction in growth rate of the non-symbiotic strain R7ANS. RDFs such as the Cox protein of phages HP1 and P2 and the Vis protein of bacteriophage P4 also act as transcriptional regulators (Saha et al., 1987; Polo et al., 1996; Esposito and Scocca, 1997). Hence it is possible that rdfS may have a regulatory role additional to its role as a RDF and constitutive expression may lead to the expression or repression of ICEMlSymR7A genes causing a lethal effect. Alternatively, RdfS may interfere through DNA-binding with expression of the phetRNA gene which is a unique gene in strain R7A. Such binding may be augmented by the presence of an integrated copy of ICEMlSymR7A.

Complementation studies showed that the rdfS promoter was immediately upstream of the gene. The rdfS stop codon overlaps with the traF start codon, and traF and msi107 also overlap, indicating that rdfS is the first gene in a rdfS-traF-msi107 operon (Fig. 6) and hence is coregulated with genes required for conjugative transfer. This provides an elegant mechanism by which ICEMlSymR7A excision and transfer are co-ordinated. The finding that ICEMlSymR7A excision was increased in stationary phase suggested that rdfS may be regulated at least in part in response to cell density by quorum sensing mediated by traR and traI2 genes on the island. This was confirmed by the finding that provision of traR traI2 in trans on a plasmid caused excision of the island in all cells, a finding that also indicated traR traI2 expression is likely to be autoregulated. Clearly traR traI2 expression must be regulated by other factors in wild-type cells.

Unexpectedly, attP was present at a ratio of ∼1.5:1 to attB in cells containing pJR206 (traR traI2), in which ICEMlSymR7A was present only in excised form irrespective of growth phase. This implies that replication of the excised form of ICEMlSymR7A was occurring in these cells, a surprising observation given that the island lacks obvious plasmid vegetative replication genes (Sullivan et al., 2002). The island oriT required for conjugative transfer is located in the intergenic region between msi107 and rlxS (our unpublished data). Therefore it seemed possible that the island was replicating by rolling-circle replication initiated at the oriT. Strong support for this suggestion was obtained by the finding that ICEMlSymR7A was not maintained in a rlxS mutant containing pJR206. Despite showing only weak similarity to VirD2 itself, RlxS is a member of COG3843 that includes the VirD2 relaxase of the A. tumefaciens Ti plasmid. It is encoded immediately downstream of the island oriT and, as shown here, is required for island transfer. Hence it is likely to be the relaxase required for initiation of the rolling-circle replication required for conjugative transfer of ICEMlSymR7A. The finding that ICEMlSymR7A was maintained in excised form in exponential and stationary phase broth cultures of strain R7A(pJR206) but not in the rlxS mutant strongly suggests that under appropriate conditions favouring excision, the island can be maintained by rolling-circle replication uncoupled from conjugative transfer. Although not previously reported for other ICEs, it seems likely that such replication will contribute to the stable maintenance of the ICE in the host cell during the period that the cell is acting as a conjugative donor of the ICE. Such stability may also be enhanced if expression of the RDF inhibits cell growth, as indicated by our findings that constitutive expression of rdfS reduced the growth rate of cured cells and seemed to be lethal in cells containing an integrated ICEMlSymR7A.

In summary, our results indicate that ICEMlSymR7A excision and transfer are under co-ordinated and complex regulation with population density a key factor. The molecular mechanisms that result in excision occurring in only a fraction of cells remain to be determined. The finding that RdfS-dependent excision occurs in cultures at low cell density, albeit at much lower frequency, may indicate that cues other than quorum sensing play a role or that the cell population is heterogeneous with respect to activation of excision. We are currently further characterizing the regulatory network that controls ICEMlSymR7A excision, transfer and reintegration.

Experimental procedures


Escherichia coli strains were cultured at 37°C in either 2YT medium (Sambrook et al., 1989) or TY (Beringer, 1974). Mesorhizobium strains were cultured in TY or RDM media (Ronson et al., 1987) with 10 mM glucose (G/RDM) supplemented with thiamin (1 μg ml−1), nicotinate (1 μg ml−1) and biotin (20 ng ml−1), or in RDM media containing 10 mM succinate and no vitamins (S/RDM). Media were supplemented with antibiotics where required at the following concentrations: for E. coli 50 μg ml−1 kanamycin, 12.5 μg ml−1 tetracycline, 200 μg ml−1 ampicillin, 25 μg ml−1 gentamicin; for mesorhizobia 2.0 μg ml−1 tetracycline, 200 μg ml−1 neomycin, 50 μg ml−1 gentamicin, 200 μg ml−1 streptomycin.

Bacteria, plasmids and primers

The bacterial strains and plasmids used in this study are listed in Table 1. Primers are described in Table S1.

To construct strain R7AΔintS, plasmid pJS101 that contains a 6.1 kb EcoRI genomic DNA fragment including intS was digested with EcoNI and BglII and end-filled, removing the region spanning nucleotides 199–1248 of intS, and the deleted region replaced with a SmaI fragment containing the nptII gene from p239. The EcoRI fragment from the resultant plasmid was cloned into pLAFRI to give pJS103 which was introduced into R7A. Marker exchange of the nptII gene into R7A was forced by plasmid incompatibility using pPH1JI (Ruvkun and Ausubel, 1981) and the mutant confirmed by Southern hybridization. pPH1JI was then evicted from the strain by introducing pLAFR1 and an isolate that had lost pLAFR1 was selected as described (Hubber et al., 2004) to give R7AΔintS.

To construct strain R7AΔmsi109 (renamed as R7AΔrdfS), a PCR product containing the first 70 bp of msi109 and the upstream 931 bp was amplified using primers 109LL and 109LR and digested with XhoI and PstI. A second PCR product containing the 3′-terminal 65 nucleotides of msi109 and 948 bp of downstream sequence was amplified using primers 109RL and 109RR and digested with SpeI and PstI. The two PCR products were combined and cloned into pJQ200SK that had been digested with XhoI and SpeI, creating pJK301. pJK301 was transferred into R7A by conjugation and a gentamicin-resistant exconjugant that contained the integrated plasmid was selected. Sucrose-resistant, gentamicin-sensitive isolates were then obtained by plating on RDM containing 5% sucrose and screened by PCR using primers 110L and 109CN3 to identify a strain in which the wild-type msi109 gene had been replaced by the mutated gene. Sequencing of the PCR product confirmed that the correct mutant, with a markerless 135 bp in-frame deletion in msi109, had been obtained.

Strain R7Amsi106 (renamed as strain R7ArlxS) was constructed by insertion duplication mutagenesis (IDM) using the suicide plasmid pJS112 (Table 1). The plasmid was transferred into R7A by conjugation. Transconjugants were confirmed via Southern hybridization by probing successively with pFUS2 and the PCR product used to construct pJS112.

In order to clone a region containing attP, intS and any regions required for expression of intS, an 1897 bp region was amplified from R7A genomic DNA by PCR using primers CIB5 and CIB3. This region contained the left end of ICEMlSymR7A including intS preceded by right-terminal 436 bp of the island (Fig. 1C). The PCR product was cloned as a BamHI fragment into the low-copy-number suicide vector pJJ607 to produce pJJ608.

Bacterial matings

Biparental spot matings were carried out on TY plates using E. coli strain S17-1 as donor. Matings for transfer of ICEMlSymR7A to non-symbiont strains were carried out as previously described (Sullivan and Ronson, 1998).

DNA manipulations

Mesorhizobium DNA was prepared either as described previously (Hubber et al., 2004), or extracted from broths or colonies using the PrepMan Ultra reagent (Applied Biosystems), following the protocol of the supplier. Electroporation of Mesorhizobium strains was carried out as described (Hubber et al., 2004). PCR products were amplified using the Phusion High-Fidelity PCR kit (Finnzymes) and purified using the High-Pure PCR product purification kit (Roche). Southern hybridizations and DNA sequencing were carried out as described (Hubber et al., 2004).

Bioinformatic analysis

Alignments, alignment editing and phylogenetic tree construction were carried out using the dambe software (Xia and Xie, 2001). The clustalw amino acid alignment of RdfS homologues and Vis was carried out using the slow/accurate setting. The alignment was then truncated to remove unaligned regions, producing a 62-amino-acid alignment (Fig. S1). A neighbour-joining tree was then created from the alignment and bootstrapped 1000 times and rooted using the Vis sequence as the outgroup. Both maximum parsimony and maximum likelihood methods produced similar trees.

QPCR assays for excision

Primer Express software V. 2.0 (Applied Biosystems) was used to design primers to detect attP (attPL, attPR), attB (attBL, attBR) and melR (melRL, melRR) and to design FAM-labelled minor-groove-binding (MGB) probes for attP/attB (TAQATT) and melR (TAQMR). The same probe was used for detection of both attP and attB, as it lies within the 3′-terminal 17 bp of the phetRNA gene that is present within both attachment sites. The Applied Biosystems 7500 Fast System was used for real-time fluorescence detection of PCR products, and results were analysed with Applied Biosystems 7500 Fast System SDS software V 1.3. To prepare DNA templates for QPCR, cells from between 200 μl and 2 ml of TY broth cultures (depending on culture density) were pelleted by centrifugation and DNA was extracted using the PrepMan Ultra reagent (Applied Biosystems). Reactions were carried out in 10 μl volumes containing 5 μl of TaqMan® Fast Universal PCR Master Mix (Applied Biosystems), 900 nM of each primer, 250 nM of probe and 1 μl of template. Cycling conditions were 20 s at 95°C then 40 cycles of 3 s at 95°C and 30 s at 60°C.

The amplification efficiency of all three assays was determined using the plasmid pJR201, which contains single copies of attP, attB, and a region of melR (Fig. 1), linearized with EcoRI. Serial doubling dilutions were prepared and used as templates for QPCR to generate standard curves for each of the three PCR reactions by plotting relative DNA concentration versus log(Ct) value (Ct is the PCR cycle at which fluorescence rises beyond background levels). Amplification efficiency was determined using the equation E = 10(−1/slope) using slope values determined from the standard curves. Data from the attB or attP PCR reactions were normalized against results obtained from the melR PCR reaction with correction for differences in amplification efficiencies using the previously described equation:


where Eatt represents the efficiency of either the attB or attP PCR reaction, EmelR is the efficiency of the melR PCR reaction, ΔCt(att) represents the difference between the Ct value of the reference att PCR and the test att PCR and ΔCt(melR) represents the difference between the Ct value of the reference melR PCR and the test melR PCR (Pfaffl, 2001; Burrus and Waldor, 2003). To test for significant differences between the strains at each data point, the Student's t-test was employed, with the assumption of equal variance across groups. Prior to applying the test, the data were log transformed (log base 10) to reduce the positive association between the mean and the variance.


We thank Vincent Burrus for advice on QPCR and Mik Black for statistical advice. This work was supported by a grant from the Marsden Fund administered by the Royal Society of New Zealand.