ICEMlSymR7A of Mesorhizobium loti is an integrative and conjugative element (ICE) that confers the ability to form a nitrogen-fixing symbiosis with Lotus species. Horizontal transfer is activated by TraR and N-acyl-homoserine lactone (AHL), which can stimulate ICE excision in 100% of cells. However, in wild-type cultures, the ICE is excised at low frequency. Here we show that QseM, a widely conserved ICE-encoded protein, is an antiactivator of TraR. Mutation of qseM resulted in TraR-dependent activation of AHL production and excision, but did not affect transcription of traR. QseM and TraR directly interacted in a bacterial two-hybrid assay in the presence of AHL. qseM expression was repressed by a DNA-binding protein QseC, which also activated qseC expression from a leaderless transcript. QseC differentially bound two adjacent operator sites, the lower affinity of which overlapped the −35 regions of the divergent qseC-qseM promoters. QseC homologues were identified on ICEs, TraR/TraM-regulated plasmids and restriction-modification cassettes, suggesting a conserved mode of regulation. Six QseC variants with distinct operators were identified that showed evidence of reassortment between mobile elements. We propose that QseC and QseM comprise a bimodal switch that restricts quorum sensing and ICEMlSymR7A transfer to a small proportion of cells in the population.
Horizontal gene transfer drives the spread of virulence, antibiotic resistance, metabolic and symbiotic determinants and facilitates the rapid generation of novel genotype combinations. Bacterial chromosomes are often mosaics of ‘core’ and laterally acquired genes, with acquired regions commonly present as large genomic islands (Goldenfeld and Woese, 2007; Juhas et al., 2009; Dobrindt et al., 2010). Integrative and conjugative elements (ICEs) represent a distinct group of self-transmissible genomic islands that are widespread in prokaryotic genomes (Wozniak and Waldor, 2010; Guglielmini et al., 2011). ICEs reside within the host's chromosome, but excise to form a circular element that donates a single DNA strand for conjugative transfer to a recipient cell. Integration into the donor and recipient genomes then occurs through site-specific recombination facilitated by a phage or transposon-like recombinase (Burrus et al., 2002). The ongoing characterization of ICEs has revealed that while their recombination and mobilization determinants appear as a mosaic of phage, transposon and plasmid-like sequences, they have evolved unique mechanisms to regulate their transfer machinery (Wozniak and Waldor, 2010).
ICEMlSymR7A of Mesorhizobium loti strain R7A is a 502 kb ICE that was discovered through its ability to transfer to other bacteria in the soil and convert them into strains capable of a nitrogen-fixing symbiosis with Lotus species (Sullivan et al., 1995). Integration of ICEMlSymR7A into the single M. loti phe-tRNA gene (Sullivan and Ronson, 1998) requires the integrase IntS. IntS-mediated excision requires expression of the recombination directionality factor RdfS, which is encoded by the first gene in a predicted operon encoding conjugative proteins, presumably co-ordinating excision with transfer. (Ramsay et al., 2006). ICEMlSymR7A also encodes a relaxase RlxS and proteins constituting the Mpf (membrane pore formation) system that are required for conjugative transfer of excised ICEMlSymR7A (Sullivan et al., 2002; Ramsay et al., 2006). Stable maintenance of the excised form of ICEMlSymR7A requires RlxS (Ramsay et al., 2006), suggesting that ICEMlSymR7A can be maintained by rolling-circle replication in its excised form, as has been observed for other ICEs (Grohmann, 2010).
ICEMlSymR7A excision and transfer is positively regulated by a LuxRI-family autoinduction/quorum sensing (QS) system (Ramsay et al., 2006; 2009). TraR responds to N-(3-oxo-hexanoyl)-l-homoserine lactone (3-oxo-C6-HSL) produced by TraI1 and activates expression of traI1 and the traI2-msi172-msi171 operon. The traI2 gene encodes a cryptic AHL synthase that has no apparent effect on AHL production or excision, while the msi172 and msi171 gene products are essential for activation of excision and conjugative transfer. In wild-type strain R7A grown in broth culture, AHL production is low; however, introduction of a low-copy-number plasmid containing a cloned copy of traR (pJRtraR) into R7A leads to strong traI1 expression and a 1000-fold increase in 3-oxo-C6-HSL production. Similarly, excision of ICEMlSymR7A in wild-type cultures is low, ranging from about 0.06% of cells in exponential phase to about 6% of cells in stationary phase broth cultures, but R7A(pJRtraR) cultures show excision and extrachromosomal maintenance of ICEMlSymR7A in 100% of cells (Ramsay et al., 2009). In a traI1 mutant, these effects of pJRtraR require as little as 50 pM exogenous 3-oxo-C6-HSL, whereas in the absence of pJRtraR, the traI1 promoter remains uninduced even in the presence of as much as 50 nM exogenous 3-oxo-C6-HSL (Ramsay et al., 2009). Thus it appears that activation by TraR is repressed in wild-type cells, but this repression is overcome through introduction of additional copies of traR on pJRtraR.
TraR-activated transfer of the Agrobacterium tumefaciens Ti plasmid (Fuqua et al., 1995; Hwang et al., 1995) and the Rhizobium leguminosarum bv. viciae plasmid pRL1JI (Danino et al., 2003) is countered by the antiactivator TraM that directly binds TraR and prevents response to low amounts of AHL (Qin et al., 2007). A Ti plasmid constitutively expressing traR requires 1 nM of 3-oxo-C8-HSL for maximum induction of transfer, whereas only 100 pM is required in a traM mutant (Khan et al., 2008). Recent characterization of several novel QS antiactivators in Pseudomonas aeruginosa (Siehnel et al., 2010; Seet and Zhang, 2011; Venturi et al., 2011) affirms that antiactivation of QS systems is more common than previously appreciated. Moreover, some mathematical models predict that antiactivation is essential for an ‘off-state’ in QS regulatory circuits (Goryachev et al., 2005; Goryachev, 2011).
ICEMlSymR7A does not encode an identifiable TraM homologue, nor does the M. loti R7A genome (J. Ramsay, unpubl. data). In this study we describe the identification of a novel antiactivator of AHL production and ICEMlSymR7A excision and transfer, QseM. The expression of qseM was found to be controlled by a DNA-binding protein QseC, which bound the region containing the divergent qseC and qseM promoters and also positively regulated its own expression from a leaderless mRNA. The qseM and qseC genes were found in a similar configuration on 26 other putative ICEs, suggesting a conserved mechanism of regulation on these elements.
qseM encodes a negative regulator of AHL production, ICEMlSymR7A excision and conjugative transfer
The divergently transcribed genes msi170 and msi169 (renamed qseM and qseC in this study) are located downstream of the traR traI2-msi172-msi171 gene cluster (Fig. S1). QseM is a predicted protein of 83 amino acids that has no similarity to characterized proteins. However, QseM homologues are widespread; a blastp search identified approximately 100 proteins that showed at least 40% amino-acid identity over at least 90% of the protein. QseC is 69 amino-acid protein that is a member of the Xre family of helix–turn–helix DNA-binding proteins. An examination of traR-traM plasmid-encoded loci revealed that similar DNA-binding protein genes were also often located adjacent to traM (Fig. S1). The similarity of the traR-qseC genetic context to traR-traM loci prompted us to investigate a role for qseM in the control of ICEMlSymR7A excision and AHL production.
A qseM markerless deletion mutant R7AΔqseM (Table S1) was constructed and analysed for the production of AHLs using the bioassay strain Chromobacterium violaceum CV026 (Fig. 1). Supernatants from R7AΔqseM cultures induced a zone of violacein production comparable to that of R7A(pJRtraR) culture supernatants, which accumulate 1000-fold more 3-oxo-C6-HSL than R7A supernatants (Ramsay et al., 2009). This indicated a role for qseM in the repression of AHL production. Next, the level of ICEMlSymR7A excision in strain R7AΔqseM was analysed by quantitative polymerase chain reaction (QPCR), using an assay that detects the excision products attP and attB relative to the total number of chromosomes present in the sample (Ramsay et al., 2006). attP and attB were present at only 0.06% in log-phase and 1% in stationary-phase R7A cultures, compared with 40–70% of cells in log-phase growth and 100% of cells in stationary phase in R7AΔqseM cultures (Fig. 2). Complementation of both these phenotypes was achieved through introduction of plasmid pJRqseM (Table S2), which contained a copy of qseM expressed from its native promoter (Figs 1 and 2). Assays revealed that the efficiency of transfer of ICEMlSymR7A to the recipient R7ANS (a derivative of R7A cured of ICEMlSymR7A) was increased over 1000-fold in R7AΔqseM relative to R7A (Table 1). In summary, these data established qseM as a negative regulator of AHL production, excision and conjugative transfer and the gene was therefore named qseM for ‘quorum sensing and excision modulator’. The adjacent gene msi169 was named qseC for ‘quorum sensing and excision controller’ for reasons described below.
Table 1. Transfer frequency of ICEMlSymR7A from M. loti R7A to R7ANS
aValues are the average of two representative biological replicates. Mating experiments for each pair of donors/recipients were carried out at least three times and showed a consistent trend.
2.5 × 10−7
6.8 × 10−7
4.4 × 10−4
7.6 × 10−4
1.5 × 10−9
3.1 × 10−9
8.0 × 10−9
3.8 × 10−9
Activation of AHL production and excision in R7AΔqseM is traR-dependent
The phenotypes exhibited by R7AΔqseM were similar to those observed for R7A(pJRtraR), suggesting that they resulted from increased traR expression or TraR activity. To test the dependence of AHL production in R7AΔqseM on traR, a double mutant R7AΔqseMΔtraR was constructed. Supernatants from cultures of this strain did not induce violacein production in CV026, indicating that activation of AHL production in R7AΔqseM was traR-dependent (Fig. S2A). Similarly, this strain did not show the large increase in excision seen in R7AΔqseM (Fig. S2B), indicating this increase was also TraR-dependent. In order to test if overexpression of qseM repressed AHL production in strain R7A(pJRtraR), plasmid pNqseM that expressed qseM constitutively from the nptII promoter was constructed. Introduction of pNqseM into R7A(pJRtraR) extinguished AHL production detectible in CV026 assays, indicating that overexpression of qseM overcame the effect of increased traR copy number provided by pJRtraR (Fig. 1). Furthermore, analysis of excision levels in R7A(pJRtraR)(pNqseM) showed that ICEMlSymR7A was only excised in ∼ 1% of cells in exponential phase and ∼ 4% in stationary phase, a level reduced from that of R7A(pJRtraR) (Fig. 2).
Mutation of msi172 or msi171 completely abolishes excision, whereas mutation of traR has little effect on excision levels in R7A cultures (Ramsay et al., 2009) (Fig. 2). This suggests that in R7A and R7AΔtraR, a basal or uninduced level of expression of traI2-msi172-msi171 promotes excision in a minority of cells. Therefore, if QseM acted solely by repressing TraR activity or traR expression, overexpression of qseM would result in levels of excision similar to those observed in R7AΔtraR. Instead excision in R7A and R7AΔtraR cells carrying pNqseM was reduced to barely detectible levels (Fig. 2), indicating that the basal level of excision was further repressed in these strains. Therefore, while the activation of AHL production and excision in R7AΔqseM was clearly traR-dependent, it seemed that QseM was able to repress excision to levels observed for msi172, msi171, intS and rdfS mutants (Ramsay et al., 2006; 2009).
QseM does not regulate traR transcription or mRNA levels
Next we investigated the possibility that QseM might repress traR expression at the transcriptional level. To measure expression from the traR promoter, the 743 bp region that precedes traR was amplified and fused to lacZ in pFJX to produce pFJXtraR. β-Galactosidase activity from R7A(pFJXtraR) was low (similar to that of the empty vector) and was unaltered in R7AΔqseM(pFJXtraR) (Table S3). pFJXtraR was then introduced into the ICEMlSymR7A-cured strain R7ANS, along with either pNPqseM or the vector-only control pPR3, to determine the effect of overexpression of qseM on the traR promoter. Again, β-galactosidase assays showed that expression of traR was low and unaltered by the presence of qseM (Table S3). To test if QseM had an effect on traR mRNA levels, we analysed the abundance of selected mRNA transcripts in RNA extracted from R7A and R7AΔqseM strains using qRT-PCR. As expected on the basis of our previous phenotypic data, the levels of traI1, traI2 and rdfS transcripts were increased in R7AΔqseM. However, the level of traR mRNA in R7AΔqseM was at most only slightly increased over that observed in R7A (Table 2).
Table 2. Effects of inactivation of qseM and qseC on mRNA abundance relative to R7A as determined by RT-qPCR
aValues are the average of three biological replicates with the standard deviation in brackets.
QseM interacts directly with TraR in the presence of 3-oxo-C6-HSL
We hypothesized that, by analogy to TraM, QseM might function as a TraR antiactivator. To test if TraR and QseM interacted directly, we used the Escherichia coli-based two-hybrid system Bacteriomatch II (Agilent Technologies). In this system, protein fragments are fused to the RNA polymerase α-fragment (RNAPα) and the λ cI protein. When the fusions are juxtaposed by protein–protein interaction, they activate expression of the yeast HIS3 gene, leading to histidine prototrophy and the ability to grow in the presence of 3-amino-1,2,4-triazole (3-AT). traR and qseM were cloned into fusion vectors pBT or pBTL [a derivative of pBT carrying a (Gly4Ser)3 linker (Erler et al., 2004)] and pTRG, which resulted in the fusion of cI or RNAPα to the N-terminus of TraR and QseM respectively. The four plasmid constructs (pBTLtraR, pBTqseM, pTRGtraR and pTRGqseM) were then used in combination with each other and the parent plasmids pBTL and pTRG in cotransformation assays, with or without 1 μM 3-oxo-C6-HSL included in the selective media (Table S4). A strongly positive interaction was found for the pTRGtraR and pBTqseM plasmid cotransformation in the presence of 3-oxo-C6-HSL, indicated by high colony numbers on media containing 3-AT. Colony numbers on media without 3-oxo-C6-HSL were similar to those produced from cotransformations using the control plasmids pBTL and pTRG (Table S4). The requirement for AHL may indicate that the QseM–TraR interaction is specific for TraR-3-oxo-C6-HSL, or that the stability or folding of TraR is AHL-dependent. The pTRGqseM and pBTLtraR cotransformations did not indicate a positive interaction, likely due the formation of one or more non-functional protein fusions or steric hindrance. The positive interaction of pTRGtraR and pBTqseM indicates that QseM directly interacts with TraR, establishing it as a novel QS antiactivator protein.
QseC controls excision and AHL production through regulation of qseM expression
The positioning of qseC immediately adjacent to qseM (Fig. S1) suggested it might have a role in the control of qseM expression. Therefore, a mutant strain R7AΔqseC::Ω-Km, in which qseC was replaced with the Ω-Km fragment from pHP45Ω-Km, was constructed. Excision of ICEMlSymR7A in cultures of R7AΔqseC::Ω-Km was significantly reduced relative to R7A and comparable to that seen in R7A(pNqseM). Likewise, transfer efficiency from donors R7AΔqseC::Ω-Km and R7A(pNqseM) to R7ANS was reduced 167-fold and 32-fold respectively, relative to transfer from R7A (Table 1). Next we investigated if the repression of excision in R7AΔqseC::Ω-Km was overcome by introducing pJRtraR. Unlike R7A(pJRtraR) supernatants, supernatants from R7AΔqseC::Ω-Km(pJRtraR) cultures did not induce violacein production in CV026 assays (Fig. 1). Excision of ICEMlSymR7A in R7AΔqseC::Ω-Km(pJRtraR) was 0.6% in log-phase and 40% in stationary-phase cultures, a significant increase relative to R7A but lower than the 100% frequency of excision observed in both growth phases for R7A(pJRtraR) (Fig. 2). A plasmid containing qseC expressed from its native promoter (pJRqseC) restored excision in R7AΔqseC::Ω-Km (Fig. 2).
To elucidate the hierarchy of control between qseC and qseM, a mutant strain R7AΔqseMC was constructed in which the entire qseC-qseM region was deleted. R7AΔqseMC supernatants produced zones of violacein induction identical to those of R7AΔqseM in CV026 assays. Levels of ICEMlSymR7A excision in R7AΔqseMC were also identical to those seen for R7AΔqseM (Fig. 2). Introduction of either pJRqseMC (carrying the entire qseM-qseC region) or pJRqseM into R7AΔqseMC repressed AHL production, whereas introduction of pJRqseC into R7AΔqseMC had no effect (Fig. 1). Thus the phenotypes observed for R7AΔqseMC were identical to those of R7AΔqseM, suggesting that the repression of AHL production and excision observed in R7AΔqseC::Ω-Km was due to increased qseM expression.
The expression levels of traR, traI1, traI2, rdfS and qseM in R7AΔqseC::Ω-Km relative to levels in R7A were analysed using qRT-PCR. This revealed a ∼ 7-fold increase in qseM mRNA levels (Table 2) in the absence of qseC, confirming that QseC negatively regulated qseM expression. Despite the fact that R7AΔqseC::Ω-Km showed considerably reduced ICEMlSymR7A excision relative to R7A, mRNA levels of the other genes analysed, including rdfS, showed little alteration from wild-type levels. This likely reflected the fact that in R7A, excision is already repressed in the vast majority of cells (∼ 99.94% in log phase) so repression of these genes will be negligible when averaged across the population.
QseM and RdfS define a distinct family of ICEs that show evidence of QseC sub-family shuffling
ICEMlSymR7A and Tn4371-like conjugative transposons are members of a family of ICEs that all encode homologues of RdfS, RlxS and Mpf proteins (Toussaint et al., 2003; Ramsay et al., 2006; 2009). Further inspection revealed that Tn4371-like elements did not encode QseM homologues, so we defined a subset of ICEs that encoded QseM. Parallel PSI-blast searches performed using RdfS and QseM sequences identified 26 gene clusters (excluding ICEMlSymR7A and ICEMlSymMAFF) with both sequences (Fig. S3). The clusters encoded a conserved set of 12 predicted proteins including Msi172 and Msi171, in addition to the Mpf proteins. The PSI-blast searches also revealed that QseM showed weak similarity to Msi171 (18.6% identity to the last 70 amino acids of Msi171) and that several QseM homologues were annotated as members of the conserved domain family DUF2285, as is Msi171. In Parvibaculum lavamentivorans DS-1 the entire cluster of conserved genes was present in a contiguous segment, possibly representing the archetypal arrangement for these elements. Interestingly, outside of the mesorhizobia, genes encoding TraR or TraI homologues were only associated with the clusters in Sphingomonas sp. SKA58, indicating that QS regulation may be a recent addition to this regulatory circuit. QseC-like DNA-binding proteins were encoded adjacent to qseM genes on all clusters, although their sequences aligned into six distinct groups (labelled 1–6, Figs S3 and S4). The alignments also confirmed that the first of three ATG codons in the M. loti R7A qseC sequence was the only conserved start codon (Fig. S5).
Operator sequences are often identifiable upstream of genes encoding Xre proteins (Sorokin et al., 2009). MEME analysis (Bailey et al., 2009) of the DNA region upstream of each qseC gene revealed that there were distinct sequence motifs containing dyad symmetry located upstream of each gene and that these motifs were specific for each of the six QseC groups (Fig. S4). These motifs exhibited strong similarity to the paired operator sites bound by the control (C) proteins of Type II restriction–modification systems. The conserved dyad repeats present upstream of qseC (labelled OL and OR) were identical to the putative operator sequences of the C.PvuII-like family (specifically motif 4) identified in an in silico screen for novel C proteins (Sorokin et al., 2009), suggesting that QseC is a member of the C.PvuII-like C-protein family.
Comparison of the QseC groups in Fig. S4 with the annotations of putative ICEs in Fig. S3 revealed evidence of evolutionary substitution of QseC family members, as seemingly closely related ICEs carried QseC sequences from different groups. Furthermore, additional copies of QseC homologues were identified downstream of those adjacent to QseM on several ICEs. These may represent remnants from recombination or gene replacement events occurring at these sites (Fig. S3). To visualize the inheritance pattern of both the QseC and QseM homologues simultaneously, a tanglegram (a diagram which minimizes the number of crossing branches between trees to highlight instances of incongruence) (Scornavacca et al., 2011) was constructed from the QseC tree and a tree constructed from QseM sequences (Fig. S6). This analysis highlighted several instances of phylogenetic incongruence. For example, the QseM sequence of P. lavamentivorans DS-1 ICE1 clustered with that of ICEMlSymR7A, but their associated QseC sequences clustered with groups 1 and 6 respectively (Figs S3 and S4). A similar trend was observed when a tree was constructed from TraM-associated QseC sequences, which showed these proteins clustered with groups 1, 3 and 6 of the QseM-associated QseC sequences (Figs S1 and S7). Together these analyses indicated that the qseC-qseM and indeed the qseC-traM arrangement was a conserved configuration, but that the QseC genes have been frequently replaced with different homologues together with their cognate DNA-binding regions.
qseC expression is positively autoregulated from a leaderless mRNA
We used 5′ RACE to map the transcriptional start sites of qseC and qseM. The transcriptional start site of qseM was positioned 62–63 bp upstream of the qseM translational start codon, while qseC was expressed as a leaderless mRNA. Sequences resembling the rhizobial σ70 promoter −10 consensus sequence 5′-CTATAT-3′ (MacLellan et al., 2006) were identified upstream of the transcriptional start sites of both qseC and qseM, while the −35 regions of both promoters were located within OR (Fig. 3).
C proteins often positively regulate their own expression (Sorokin et al., 2009). To investigate the role of QseC in expression from the qseC promoter, a qseC promoter-lacZ transcriptional fusion was constructed in pFJX. β-Galactosidase activity was then assayed in R7A and R7AΔqseC::Ω-Km. The qseC promoter was strongly expressed in R7A in both log phase [28.6 relative fluorescent units (RFU) min−1 per OD600, standard deviation (SD) = 9.0] and stationary phase (23.8 RFU min−1 per OD600, SD = 9.8). However, only basal-level expression (log-phase: 5.1 RFU min−1 per OD600, SD = 0.9; stationary-phase: 3.3 RFU min−1 per OD600, SD = 2.2) was observed in R7AΔqseC::Ω-Km, indicating that qseC was positively autoregulated.
QseC binds both operator sequences OL and OR
C-protein operator regions often consist of two adjacent operator sites that can each accommodate a C-protein dimer. The C.AhdI protein, for instance, preferentially binds the OL operator as a dimer and activates ahdICR expression. As its concentration increases, a second dimer binds cooperatively to OR on the opposite face of the DNA helix, blocking access to the −35 region and repressing ahdICR expression (McGeehan et al., 2006). Our observation that the −35 regions of both the qseC and qseM promoters were located within OR led us to suspect a similar mechanism could explain the regulation of both qseC and qseM by QseC.
The plasmid p6HQseC was used to express N-terminal hexahistidine-tagged QseC (6H-QseC) in E. coli BL21(DE3), and the protein was purified by Ni-affinity chromatography. Complementary 43 bp oligonucleotides encompassing both OL and OR (OL + OR, Fig. 4A) were annealed and labelled with digoxigenin (DIG) and used in electrophoretic mobility shift assays (EMSA) (Fig. 4B and C). This analysis revealed that 6H-QseC retarded the migration of OL + OR and formed two discrete complexes. The faster-migrating complex formed at lower concentrations of 6H-QseC, likely corresponding to binding of a 6H-QseC dimer to a single operator site. The slower-migrating complex only formed with higher concentrations of 6H-QseC, suggesting that both operator sequences were bound by 6H-QseC dimers. Binding was specific for the OL + OR sequence, as the addition of 500-fold excess of unlabelled OL + OR outcompeted the labelled OL + OR DNA, while 500-fold excess of unlabelled, randomized oligonucleotide (RDM) did not (Fig. 4B).
To confirm that the two complexes observed were formed as a result of binding to each operator sequence OL and OR, the sequences of each operator were individually shuffled (Fig. 4A) and the resulting double-stranded DNA oligonucleotides were used in EMSA assays. The wild-type sequence OL + OR again formed two distinct complexes, but the OL sequence (coupled with a shuffled OR) only formed a single complex that migrated at the same rate as the faster-migrating complex formed with OL + OR. In contrast, almost no complex formation was observed for the OR sequence coupled with a shuffled OL (Fig. 4C). No further complex formation was observed even when as much as 170 nM 6H-QseC was added (data not shown). This indicated that the binding affinity of QseC for OR was much lower than that for OL.
Excision of ICEMlSymR7A normally occurs at low frequency in wild-type R7A cultures. However, the introduction of a plasmid-borne copy of traR increases TraI1-dependent AHL production and results in excision in 100% of cells and increased conjugative transfer through expression of the excision-activating genes msi172 and msi171. Therefore, ICEMlSymR7A excision and transfer are positively regulated by TraR in the presence of AHL generated by TraI1, but traR expression or TraR activity appears to be repressed under normal laboratory conditions (Ramsay et al., 2006; 2009). These observations and the traM-like genomic context of qseM led us to view QseM as a potential repressor of TraR-mediated activation. This prediction was confirmed by the findings that mutation of qseM resulted in activation of AHL production, excision of ICEMlSymR7A in 100% of cells and a greater than 1000-fold increase in conjugative transfer to R7ANS. Overexpression of qseM repressed AHL production, excision and transfer. Furthermore, bacterial two-hybrid assays demonstrated QseM directly interacted with TraR, confirming that it was a novel antiactivator of TraR. Therefore, the role of QseM appears analogous to that of the TraM protein encoded on various TraR-regulated agrobacterial and rhizobial plasmids, although QseM shares no apparent sequence homology with TraM. Interestingly, the TraR/QseM two-hybrid interaction required the presence of 3-oxo-C6-HSL whereas cognate AHL is not required for the A. tumefaciens TraR/TraM interaction (Hwang et al., 1999; Luo et al., 2000), suggesting that antiactivation by QseM may occur through a distinct mechanism.
While TraR is required for normal conjugative transfer rates (Ramsay et al., 2009), the basal level of ICEMlSymR7A excision observed in wild-type R7A cultures is not dependent on TraR. Mutants deficient in traR or traI1 exhibit a level of excision only slightly lower than seen in wild-type cultures (Ramsay et al., 2009), and excision remains growth-phase dependent. This basal level of excision is abolished in msi172, msi171 and rdfS mutants, suggesting that low-level expression of these genes occurs independently of TraR (Ramsay et al., 2009). Surprisingly, overexpression of qseM reduced excision to near undetectable levels rather than to levels found in R7AΔtraR. Furthermore, TraR and TraI homologues were not identified on the majority of related putative ICEs identified by bioinformatic analysis in this study, suggesting that QseM evolved independently of TraR on the majority of ICEs. These observations suggest that QseM has other target(s) in addition to TraR. The weak similarity covering conserved domain DUF2285 observed between the excision-activator Msi171 and QseM may provide clues as to the other role(s) of QseM, as it suggests that these proteins may have evolved from a common ancestor, or may interact with a common target or possibly each other.
The expression of qseM was regulated by the helix–turn–helix protein QseC, which bound two conserved operator sites OL and OR. The OR site overlapped the −35 regions of the divergent qseM and qseC promoters. Expression of qseM was ∼ 7-fold upregulated in R7AΔqseC::Ω-Km, indicating that QseC repressed qseM expression. This was reflected by markedly reduced excision and conjugative transfer rates and by lack of induction of AHL production upon introduction of pJRtraR into R7AΔqseC::Ω-Km. In addition, the expression of qseC was positively autoregulated by QseC. Therefore, QseC functions as both a transcriptional activator of qseC and a repressor of qseM expression.
The QseC amino-acid sequence and the operator sites OL and OR identified on ICEMlSymR7A showed strong similarity to control (C) proteins of Type II restriction modification (RM) cassettes and their associated operator sites. Additionally, qseC was transcribed from a leaderless mRNA, a feature common to several C protein genes including those encoding C.PvuII and C.AhdI (Knowle et al., 2005; Bogdanova et al., 2008; Sorokin et al., 2009). Indeed, a high proportion of members of the C.PvuII family of C proteins are predicted to be expressed from leaderless mRNAs, due to a high conservation of a 17–18-nucleotide spacer [also conserved for qseC (Fig. 3)] between the end of the operator motif and the translational initiation codon (Sorokin et al., 2009). The C proteins of RM systems act as both transcriptional activators and repressors to control expression of the C protein, the restriction endonuclease (RE) and often the methylase (M). DNA methylation by M occurs prior to RE expression, preventing digestion of host DNA. For the C proteins C.PvuII, C.AhdI and C.Esp1396I (McGeehan et al., 2006; 2008; Mruk et al., 2007), this is accomplished through differential binding of C dimers to two adjacent operator sequences that contain inverted repeats, OL and OR, that are located immediately upstream of the C-RE locus. In the absence of C following entry into a naïve host, weak C-RE gene expression occurs while M expression is active. As C concentration increases, preferential binding of a C dimer to OL that is located adjacent to the −35 region of the C-RE promoter activates transcription of C-RE through interaction with RNA polymerase. Upon further increase in C concentration, cooperative binding of another dimer to OR blocks the −35 region and transcription of C-RE (McGeehan et al., 2006; 2012; Ball et al., 2012).
Our EMSA data demonstrated that QseC preferentially bound OL which is adjacent to the −35 region of the qseC promoter. OR was only bound very weakly in the absence of a functional OL. This suggested that binding to OR was cooperative with binding of OL. Comparison of our data with the C.AhdI and C.Esp1386I C-protein systems described above allowed us to propose the following model for regulation by QseC (Fig. 5A). In conditions where QseC is absent, e.g. following entry of ICEMlSymR7A into a naïve host, expression of qseC is weak and expression of qseM is strong, repressing QS, excision and transfer genes and stimulating integration of ICEMlSymR7A into the new host [Fig. 5A(i)]. QseC concentration eventually reaches a threshold whereupon the protein dimerises and binds to OL, further activating qseC expression [Fig. 5A(ii)]. In a few cells, QseC concentration reaches a higher threshold and OR is then also bound by a QseC dimer, repressing both qseC and qseM promoters by preventing binding of RNA polymerase holoenzyme to the −35 regions of both promoters [Fig. 5A(iii)]. Falling QseM levels then lead to activation of QS, excision and transfer. Thus QseM and QseC comprise a bimodal molecular switch that restricts transfer to the minority of cells in the population that are repressed for qseM expression.
A qseC homologue was found adjacent to qseM in the 26 putative ICEs identified in this study. These proteins were clustered into six groups, each preceded by a specific putative operator motif. Members of these QseC families and their operator motifs were also found adjacent to traM genes on several plasmids (Figs S5 and S7), suggesting QseC proteins may also control traM expression. In contrast the A. tumefaciens Ti plasmid lacks a QseC homologue and traM expression is regulated by TraR (Fuqua et al., 1995). Thus it appears that there has been active evolutionary recruitment and exchange of members of the QseC/C-protein family between RM cassettes, TraR/TraM-regulated plasmids and QseM-regulated ICEs, suggesting that the QseC-like proteins confer a common mechanism of regulatory control that is advantageous in each setting.
The phylogeny of the QseC and QseM homologues on each element was often incongruent, suggesting that there has been frequent substitution of distinct QseC homologues among the elements. A similar observation has been made for RM loci, where frequency-dependent selection acts to favour loci with different C protein variants through apoptotic mutual exclusion of RM elements with identical operators (Nakayama and Kobayashi, 1998). It has been demonstrated that entry of a unique RE gene into a host carrying a matching C protein causes premature activation of RE expression and digestion of host DNA. The QseC homologue shuffling observed in this study could be explained by a similar model, with establishment or long-term maintenance of the mobile element in a new host being disrupted by the presence of an identical resident QseC protein and/or RM locus. In this manner QseC proteins could restrict the host range of their cognate element, indirectly leading to selection for acquisition of elements carrying QseC members with different operator sequences.
The integration of QS control into a pre-existing regulatory pathway has been observed in other bacteria (Lerat and Moran, 2004; Malott et al., 2005; Coulthurst et al., 2006) and presumably reflects positive selection for acquisition of the regulatory system. For ICEMlSymR7A, it appears that through an evolutionary fusion of the traI2 gene and its promoter to the msi172-msi171 operon, TraR has been recruited to positively reinforce expression of msi172-msi171, presumably to bring activation of excision and transfer under population-level control (Fig. 5B). The further tier of regulation provided by QseM–QseC ensures that only low levels of AHL are produced by the population and that most cells remain blind to this signal. However, the low basal frequency of transfer that occurs using wild-type cultures requires TraR (Ramsay et al., 2009). When TraR levels are not limiting [as in R7A(pJRtraR)], extremely low AHL concentrations (50 pM or about one molecule/cell in stationary phase) induce TraR-dependent activation of transcription and transfer (Ramsay et al., 2009). This indicates that the low level of AHL produced by R7A cultures is likely to be physiologically relevant to cells in which qseM expression is repressed. Therefore, it seems plausible that the QseM–QseC switch facilitates bimodal induction of QS, excision and transfer in the population, allowing the co-ordination of transfer from the sub-population of qseM-repressed cells. Other physiological or environmental factors, such as the availability of suitable recipients or presence of a host plant, may introduce conditionality on the QS system and/or repression by QseM by modulating the proportion of cells receptive to induction by QS. This mechanism could allow ICEMlSymR7A to maximize its potential for propagation by both vertical and horizontal descent, by appropriately partitioning a subpopulation of cells for transfer in response to environmental conditions.
Strains, plasmids and growth conditions
Strains and plasmids used in this study are listed in Table S1. Escherichia coli and M. loti strains were cultured in Luria–Bertani (LB) and TY media respectively, which were supplemented with antibiotics when required (Ronson et al., 1987; Ramsay et al., 2006; 2009). Chromobacterium violaceum AHL reporter assays were carried out as described (McClean et al., 1997) except that 20 × 20 cm LB agar plates were used and 200 μl of filter-sterilized supernatant from 64 h TY broth cultures of M. loti were assayed. Plasmids were introduced into M. loti either by electroporation or by biparental matings using either E. coli strain S17-1 (Simon et al., 1983) or β2163 (supplemented with 0.3 mM 2,6-diaminopimelic acid) (Demarre et al., 2005) as a donor.
General molecular biology methods and strain construction
DNA extractions, electrophoresis, PCR, Southern hybridization and DNA sequencing were carried out as previously described (Ramsay et al., 2006; 2009). Primers used in this study are listed in Table S2. To construct strain R7AΔqseC, two PCR products each containing 1 kb of DNA flanking qseC were amplified using primers msi169LL + msi169LR and msi169RL + msi169RR. The Ω-Km interposon was then cloned between the two PCR products in pIJ3200 to produce pJS169, which was then used for marker exchange, as described for construction of R7AΔintS (Ramsay et al., 2006). For construction of in-frame markerless deletion mutants R7AΔqseM and R7AΔqseMC, the deletions were first constructed by overlap extension PCR as described (Ramsay et al., 2006), using primers msi1705F + msi1705R (5′ region) and msi1703F + msi1703R (3′ region) for R7AΔqseM and msi1705F + msi1691705R (5′ region) and msi1691703F + msi1691703R (3′ region) for R7AΔqseMC. The resulting PCR products were cloned into pJQ200SK as XbaI fragments and used for producing markerless deletions in M. loti as described (Ramsay et al., 2006). Mutants were confirmed by PCR and sequence analysis. For construction of R7AΔqseMtraR, the qseM deletion was made in R7AΔtraR (Ramsay et al., 2009). Construction details for other plasmids are listed with their corresponding entries in Table S1.
QPCR, RNA extraction, 5′ RACE and qRT-PCR
The ICEMlSymR7A excision products attP and attB were detected using Taqman chemistry and quantified relative to the chromosomal marker melR, a gene located adjacent to the ICEMlSymR7A insertion site, as described (Ramsay et al., 2006). All data presented are the average of at least two biological and three technical replicates. RNA for 5′ RACE was extracted as described (Ramsay et al., 2009) and 5′ RACE was carried out using a Roche 5′/3′ second-generation kit according to the manufacturer's instructions. For mapping of PqseC, the SP1 and SP2 primers (see kit instructions) correspond to msi169_SP1 and msi169_SP2 respectively. For mapping of PqseM, SP1 and SP2 correspond to msi170_SP1 and msi170_SP2. Both reactions gave single PCR products and the products were directly sequenced using SP2. For qRT-PCR, RNA was extracted from 8 ml of M. loti culture following 24 h growth in TY at 28°C (OD600 of 0.8–1.2). RNA was stabilized using RNAprotect Bacteria (Qiagen) and extracted using Protocol 4 from the RNAprotect Bacteria Reagent handbook (Qiagen). DNAase treatment was carried out using the TURBO DNA-free kit (Ambion). cDNA synthesis was carried out using random hexamers (Qiagen) and Superscript III reverse transcriptase (Invitrogen) according to the manufacturer's instructions. qRT-PCR was performed on an ABI ViiA 7 machine using Fast SYBR Green Master Mix (Applied Biosystems) and analysed using ViiA 7 software version 1.2. Relative quantification was determined using primers specific for gyrA as a housekeeping gene and efficiency correction was used for the calculation of final values (Pfaffl, 2001). The amplification efficiencies for gyrA, traR, traI1, traI2, qseM and rdfS amplicons were 1.87, 1.89, 1.92, 1.80, 1.85, and 1.87 respectively. Controls without RT were included for each analysis and melting curves were used to confirm the amplification of a single product.
To monitor β-galactosidase produced from reporter–gene fusions in pFJX, 150 μl samples of culture were frozen at −70°C until required. β-Galactosidase activity was determined from 10 μl of each sample using the substrate 4′-Methylumbelliferyl-β-d-galactoside (MUG) and liberated fluorescence was recorded as previously described (Ramsay et al., 2011), using a Gemini XPS (for measurements of pJRCpro) or Infinite Tecan M200 (for measurements of pFJXtraR) microplate reader.
Conjugative transfer assays
Assays were carried out as previously described (Sullivan and Ronson, 1998; Ramsay et al., 2009), with the following modifications. Five hundred microlitres of late stationary-phase (72 h) TY broth cultures of the donor strain and the recipient R7ANS(pFAJ1700) were mixed and transferred onto 2.5 cm diameter 0.45 μm nitrocellulose filters that were then incubated on TY plates for 48 h at 28°C. The mating mixtures were then resuspended and plated on selective media. Transconjugants were identified by plating on media containing tetracycline (Tc) but lacking biotin, thiamin and nicotinate.
Bacterial two-hybrid assays
Experiments to detect protein–protein interaction of TraR and QseM were performed using the Bacteriomatch II system (Agilent Technologies), according to the manufacturer's instructions, with the following modifications. Screening media contained 13.5% w/v Na2HPO4, 0.6% w/v KH2PO4, 0.1% w/v NaCl and 0.2% w/v NH4Cl and M9 media additives [0.19% w/v Yeast Synthetic Drop-out Medium Supplement (Sigma, Cat Y-1751-20G), 0.4% glucose, 0.1 mM adenine hemisulphate, 1 mM MgSO4, 0.1 mM Thiamine HCl, 0.01 mM ZnSO4, 0.1 mM CaCl2, 0.05 mM isopropyl β-d-1-thiogalactopyranoside (IPTG)]. Selective and non-selective media were supplemented with 14 μg ml−1 Tc and 10 μg ml−1 chloramphenicol (Cm). Selective media contained 5 mM 3-AT, with or without 2 μM of racemic 3-oxo-C6-HSL (1 μM effective concentration). Electrocompetent ‘BacterioMatch II Validation Reporter Competent Cells’ were prepared and used in plasmid cotransformation assays. Transformation efficiency was estimated by dilution series on non-selective medium. Positive protein–protein interactions were detected by increased colony-forming units (cfu) on media containing 3-AT compared with empty plasmid controls. Cfu were normalized to transformation efficiency on medium lacking 3-AT.
Purification of 6H-QseC
An overnight 100 ml LB (plus Cm and Ap) culture of E. coli strain BL21(DE3)(pLYS) carrying plasmid p6HQseC was used to seed 500 ml of LB supplemented with Ap and 1 mM IPTG, which was then incubated at 28°C overnight. Cells were collected by centrifugation and resuspended in 50 ml lysis buffer [50 mM sodium phosphate buffer, pH 8.0, 100 mM NaC1, 20 mM imidazole, 20% glycerol, 1× cOmplete Mini EDTA-free Protease Inhibitor Cocktail Tablet (Roche)]. The cell suspension was then processed twice through an E1061 constant cell disruption system (Mitsubishi Electric) at 6900 kPa to lyse cells. Insoluble material was removed by centrifugation at 15 000 g for 30 min. Clarified lysate was passed through a column containing 400 μl of Ni-NTA resin, pre-equilibrated with wash buffer (50 mM sodium phosphate buffer, pH 8.0, 100 mM NaC1, 20 mM imidazole, 20% glycerol). Bound protein was washed with 500 ml of wash buffer and eluted with 10 ml of elution buffer (50 mM sodium phosphate buffer, pH 9.2, 100 mM NaC1, 250 mM imidazole, 20% glycerol) into 1 ml fractions. The presence of purified 6H-QseC was confirmed by analysis on a 15% acrylamide SDS-PAGE gel and by mass spectrometry of the trypsin-digested band. Fractions containing purified 6H-QseC were combined and concentrated in a centrifugal filter unit (Amicon Ultra 10K, 0.5 ml) to a concentration of 1 mg ml−1 (estimated using the Pierce BCA Protein Assay Kit) and stored in elution buffer at 4°C prior to use in EMSAs.
Electrophoretic mobility shift assay
Annealing and DIG-11-ddUTP labelling of the 43 bp complementary oligonucleotides were carried out using a DIG Gel Shift Kit, 2nd generation (Roche), according to manufacturer's instructions. Native acrylamide gel electrophoresis was carried out in pre-run gels containing 6% 19:1 acrylamide : bis-acrylamide, 0.5× TBE buffer (1× TBE = 89 mM Tris, 89 mM Boric acid, 2 mM EDTA) and 2.5% glycerol, run in 0.5× TBE buffer. Binding reaction constituents were identical to those in control reactions described in the kit instructions, except that reaction mixtures contained 1 μg of both poly[d(I-C)] and poly[d(A-T)] and that 7.75 fmol of labelled oligonucleotide was used in each assay. Electrophoresis, electroblotting, western analysis and chemiluminescent detection of DIG-11-ddUTP labelled probes were carried out according to DIG Gel Shift Kit instructions.
Annotation of ICEMlSymR7A-like elements and sequence comparison
Parallel PSI-blastp searches of QseM and RdfS were used to screen the NCBI ‘BLAST with microbial genomes’ database (http://www.ncbi.nlm.nih.gov/sutils/genom_table.cgi) with sequences available as of October 2007 queried. Sequences containing homologues of both QseM and RdfS were retrieved and analysed for the presence of other ICEMlSymR7A protein homologues by blastp querying nearby (∼ 500 kb) predicted proteins against the ICEMlSymR7A sequence. Unannotated ORFs were identified by screening sequences using blastx. Annotations are presented in Fig. S2.
For amino-acid sequence tree construction, clustalw was used to align sequences, from which neighbour-joining (Figs S2 and S4) or maximum likelihood (Fig. S3) trees were constructed using dambe software (Xia and Xie, 2001). C.AhdI was used as an out-group for QseC proteins, while the QseM-alignable portion of Msi171 was used as an out-group for QseM homologues. To compare the inferred phylogenies of the QseC and QseM proteins identified in Fig. S2, a tanglegram was then constructed using the Dendroscope software (Huson et al., 2007), which uses the NN-tanglegram algorithm (Scornavacca et al., 2011) to minimize the number of connector crossings between the leaves of each tree in a side-by-side comparison. For Fig. S3, secondary QseC proteins encoded downstream of the copy adjacent to QseM were not included in tree construction. WebLogo was used to create the sequence similarity logo in Fig. 3 (Crooks et al., 2004).
J. P. R. thanks the University of Otago for a PhD scholarship, the University of Cambridge (UK) for a Herchel Smith Postdoctoral Fellowship and the Health Sciences Division (University of Otago) for a Career Development Postdoctoral Fellowship. Rujirek Noisangiam is thanked for her contribution in mapping the qseM promoter, Josh Lynch for assistance with the two-hybrid work and Ashley Smith for QPCR data in Fig. S2. This work was supported by a grant from the Marsden Fund administered by the Royal Society of New Zealand and by a University of Otago Research Grant.