An Operon of Three Transcriptional Regulators Controls Horizontal Gene Transfer of the Integrative and Conjugative Element ICEclc in Pseudomonas knackmussii B13

The integrative and conjugative element ICEclc is a mobile genetic element in Pseudomonas knackmussii B13, and an experimental model for a widely distributed group of elements in Proteobacteria. ICEclc is transferred from specialized transfer competent cells, which arise at a frequency of 3-5% in a population at stationary phase. Very little is known about the different factors that control the transfer frequency of this ICE family. Here we report the discovery of a three-gene operon encoded by ICEclc, which exerts global control on transfer initiation. The operon consists of three consecutive regulatory genes, encoding a TetR-type repressor MfsR, a MarR-type regulator and a LysR-type activator TciR. We show that MfsR autoregulates expression of the operon, whereas TciR is a global activator of ICEclc gene expression, but no clear role was yet found for MarR. Deletion of mfsR increases expression of tciR and marR, causing the proportion of transfer competent cells to reach almost 100% and transfer frequencies to approach 1 per donor. mfsR deletion also caused a two orders of magnitude loss in population viability, individual cell growth arrest and loss of ICEclc. This indicates that autoregulation is an important feature maintaining ICE transfer but avoiding fitness loss. Bioinformatic analysis showed that the mfsR-marR-tciR operon is unique for ICEclc and a few highly related ICE, whereas tciR orthologues occur more widely in a large variety of suspected ICE among Proteobacteria.


Introduction
Comparisons between ever-increasing numbers of sequenced genomes reveal the large extent to which prokaryotic genomes have undergone horizontal gene transfer (HGT) [1][2][3][4][5]. HGT has traditionally been viewed as the consequence of natural transformation, or of the action of mobile elements such as conjugative plasmids and phages [6,7]. During the last decade, however, other types of mobile genetic elements such as integrative and conjugative elements (ICEs) have been recognized, which are widespread and thus may significantly contribute to HGT [8][9][10][11][12][13]. In contrast to phages and plasmids, however, we still know little about the life styles of the diverse ICE types, their modes of self-transfer and regulatory pathways controlling self-transfer. Like temperate phages, ICEs mostly exist in an integrated form at one or more specific sites in the host's chromosome (often in genes for tRNA), and are vertically transmitted to daughter cells by chromosome replication and segregation [8,10,14]. In order to transfer horizontally, ICEs excise themselves by site-specific recombination (attL and attR, Figure 1). This produces a circular double-stranded DNA molecule, which can transfer by conjugation to a new recipient cell, where it can reintegrate [14]. Autonomous plasmid-like replication of the excised form may occur [15][16][17], but is not required for the transfer itself.
The regulatory mechanisms that control the switch from integrated to excised state vary widely among different ICE types insofar as this has been studied. In several ICEs, this switch is the consequence of a cascade of a variety of regulatory factors, such as PhrI/RapI and ImmR/ImmA in ICEBs1 [18], SetR/SetCD in ICESXT [19,20], KorSA/Pra in pSAM2 [21][22][23] or QseM/TraR in ICEMlSym R7 [24,25]. Most wild-type ICEs transfer at low frequencies (i.e., less than 1 per 10 3 donors), suggesting that the regulatory cascades keep extremely tight control and allow only a small subset of cells in a population to follow the path of ICE excision and transfer, but the need for such tight control is a priori unclear. This bistability is most pronounced and well-studied for a model ICE named ICEclc in Pseudomonas [26,27], which is evolutionary very distinct from the afore-mentioned ICEs [8,28]. ICEclc is originally found in two copies in Pseudomonas knackmussii B13 and is member of a family of ICEclc-like elements widely distributed among proteobacterial species [29]. ICEclc is integrated at the 39-end of tRNA Gly genes but can excise itself by the action of the IntB13 integrase encoded on the element ( Figure 1A). Expression of intB13 in the integrated form is under control of the promoter P int , which by single cell reporter gene analysis was shown to become active only in 3-5% of a bacterial population during stationary phase [26]. Direct single cell visualization further confirmed that only cells which express reporter gene fused to P int above a threshold are capable of transferring ICEclc to new recipients, a bistable state which we recently named ''transfer competence'' (tc) [30]. Irrespective of the success of ICEclc transfer, tc cells can only divide a few times once they re-enter exponential phase before they arrest growth. We recently showed that this is due to the expression of the ICEclc genes shi and parA [30]. Expression of intB13 is dependent on a variety of factors, most notably a gene named inrR ( Figure 1A), which itself is also bistably expressed [26]. Both inrR and intB13 expression are dependent on the abundance of the stationary phase sigma factor RpoS, with cells having highest RpoS levels being more likely to activate P int and P inR [29]. RpoS and InrR are important for activating ICEclc excision and transfer, but are not sufficient. Therefore, we hypothesized that additional factors are necessary for the tc state to develop [29].
In this study, we report a locus of three consecutive regulatory genes on ICEclc, which is essential for controlling its transfer. The locus was uncovered by random transposon mutagenesis, and further studied by creation of deletion mutants and complementation. The effect of mutations was studied at the level of ICEclc expression through microarray hybridizations, RT-PCR and reporter gene-based single cell fluorescence microscopy, and further in ICEclc transfer assays. Fitness of mutants compared to wild-type was examined in growth assays and individual cell fates were followed by microscopy. Bioinformatics was used to analyze the configuration of the ICEclc regulatory locus within this ICE family, and to possibly reconstruct the steps that may have led to selection of the specific regulatory control mechanism of ICEclc. The results of our study help to explain why a careful balance has to be maintained between ICE transfer frequency and fitness loss.

Discovery of an ICEclc transfer control locus by transposon mutagenesis
In order to discover ICEclc-located factors involved in its selftransfer, a library of P. knackmussii B13 mutants was generated by using random Tn5 mutagenesis [31]. Next, we recovered ICEclc Important functional regions are depicted as grey boxes accompanied by legends. + or -, indicate the orientation of the coding strand for the respective gene(s) (the + strand being the one of intB13). clc genes, chlorocatechol degradation; amn genes, 2-aminophenol degradation. kb, kilo-basepairs. (B) Detail of the mfsR operon. Arrows, predicted open reading frames (the right-to-left orientation indicates the minus strand). Triangles, positions of the Tn5-mediated kanamycin gene insertions (nucleotide positions indicated below, according to the AJ617740.2 numbering). Regions deleted in this study are displayed as white bars with the names of the mutations noted below. The hooked arrow indicates repression by the mfsR gene product on the P mfsR promoter. (C) Detail of the fragments inserted by mini-Tn7 delivery for complementation of mfsR and tciR deletions. (D) BlastP-predicted domains for each of the three regulatory genes in the mfsR operon. C and N, carboxy and amino terminus, respectively. doi:10.1371/journal.pgen.1004441.g001

Author Summary
Integrative and conjugative elements (ICEs) are a relatively newly recognized class of mobile elements in bacteria, which integrate at one or more positions in a host chromosome, can be excised, circularized, and transfer by conjugation to a new recipient cell. Genome sequencing indicated that ICEs often carry genes with potential adaptive functions for the host. Various ICEtypes have been described and ICEclc is a useful model for a wide class of elements found in Beta-and Gammaproteobacteria. Because ICEs normally remain ''silent'' in the host chromosome and often lack selectable markers, their lifestyle is difficult to study. One of the characteristics of ICEclc is that transfer is initiated in only 3-5% of donor cells in a population during stationary phase. Here, we describe an operon of three regulatory genes, two of which control the transfer initiation of ICEclc. Our findings suggest that the low transfer rate results from the repression of an activator and that this is essential to minimize the deleterious effect of hyperactivation of transfer initiation. While the individual regulatory genes are quite common on ICEs, they rarely occur in this configuration.
elements with Km-insertions by conjugating the pool of B13 mutants en masse to Pseudomonas putida UWC1 and selecting for Kmresistant P. putida ( Figure S1). We hypothesized that mutant ICEclc with insertions in genes implicated in self-transfer could still be transferred to UWC1, when the second copy of ICEclc in the same B13 donor cell is intact and complements transfer of the mutant copy. A total of 1920 Km-resistant P. putida transconjugants was recovered and subsequently conjugated each individually with a second P. putida recipient, resistant to nalidixic acid ( Figure S1). For those conjugations in which no Km-and nalidixic acidresistant transconjugant growth was detected, the corresponding P. putida donor was recovered and the location of the Km R -gene insertion on ICEclc was mapped. A total of 18 clones was recovered, which had insertions in an ICEclc open reading frame numbered orf18502, that we renamed mfsR ( Figure 1B). Surprisingly, apart from one donor with an insertion in intB13, no other mutants with impaired ICEclc transfer were found in this screening. The Km R -gene had been inserted in four different positions in mfsR, at ICEclc nucleotide positions 19033, 18758, 18730 and 18618 ( Figure 1B, accession number: AJ617440.2). This suggests that transposon insertions in strain B13 were sufficiently frequent to cover all genes, but that the selection procedure was biased for the recovery of the mfsR insertion, which may have been due to the function of mfsR as regulator in ICEclc transfer (see below). Alternatively, it is possible that insertions in ICEclc genes needed for transfer might not be efficiently complemented by the second ICEclc copy and would thus be underrepresented in the P. putida library. Frequencies of ICEclc transfer of the strains P. putida UWC1-2961 (Km R -gene insertion at 19033) and UWC1-2962 (insertion at 18618) in a filter-based conjugation assay were 10 3 -fold and 10 2 -fold lower than of a P. putida with one integrated wild-type ICEclc copy, respectively ( Figure 2A). mfsR is part of an operon formed by three consecutive transcriptional regulators Closer inspection indicated mfsR to be the first open reading frame in a series of three consecutive transcriptional regulators, previously designated as orf18502, orf17984 and orf17162 ( Figure 1B). mfsR encodes a TetR-like regulator harboring helixturn-helix motifs TetR_N and TetR_C_7 (pfam0040 and pfam14246, respectively, see Figure 1D). The orf17984 gene overlaps with the end of the mfsR open reading frame by 4 bp and encodes a putative regulator of the MarR family (smart00347 HTH_MARR motif). The last gene of this cluster starts 24 bp downstream of the stop codon of orf17984 and is predicted to code for a LysR-type transcriptional regulator, harboring an N-terminal HTH_1 motif (pfam00126) and a C-terminal substrate-binding domain (PBP2_LTTR_aromatics_like; cd08414). The gene orf17162 was renamed tciR (transfer competence inducer regulator) in anticipation of the results described further below. Reverse transcription of P. putida UWC1 (ICEclc) RNA isolated from exponential phase-grown cells, followed by specific PCR amplification confirmed that the three genes are transcribed on the same mRNA, which ends downstream of tciR ( Figure S2). This implies that mfsR-marR-tciR form a single polycistronic unit.
ICEclc transfer frequencies in plate-mating assays with a gentamicin-resistant P. putida UWC1 as recipient were 2?10 3 -fold lower for UWC1 donors with ICEclc having an internal deletion in tciR compared to intact ICEclc ( Figure 2B). Complementation of the ICEclc-DtciR mutation with a single copy mini-Tn7 transposed fragment containing the tciR gene under the P mfsR -promoter (strain 4649, Figure 1C) restored transfer, even to much higher levels than wild-type ICEclc ( Figure 2B).
ICEclc transfer frequencies were 27-fold lower for UWC1 donors with ICEclc having an internal deletion in marR compared to intact ICEclc ( Figure 2B). Complementation of the ICEclc-DmarR mutation with a similar single copy mini-Tn7-marR insertion did not change transfer rates (strain 4804, Figure 2B). This suggests that the effect of the marR deletion on ICEclc transfer is rather due to polar disturbance of the expression of the downstream-located tciR.
In contrast, ICEclc elements with mfsR deletions [i.e., P. putida UWC1 (ICEclc-DmfsR) and UWC1 (ICEclc-DmfsR-D'marR)] transferred with 25-and 15-fold higher frequencies than wild type ICEclc, respectively (Figure 2A, B). Complementation of the ICEclc-DmfsR mutation with a single copy mini-transposed mfsR gene under control of its own promoter reduced ICEclc-DmfsR transfer frequencies by 10 4 -fold, also here much stronger than predicted from wild-type ICEclc itself ( Figure 2B). These results suggested that tciR is the actual regulator of ICEclc transfer, and further that mfsR is regulating expression of the mfsR-marR-tciR operon. Since MfsR is expected to be a repressor, its deletion would lead to higher expression of the downstream genes marR and tciR, which results in increased ICEclc transfer. The effect of the transposon insertions in mfsR (i.e., lower ICEclc transfer rates, Figure 2A) seems therefore due to a polar effect on marR-tciR expression.
tciR encodes a global activator of the genes in the ICEclc core region Next, we examined the effect of regulatory gene deletions on gene expression of ICEclc as a whole, using semi-tiling microarray analyses ( Figure 3). When P. putida UWC1 (ICEclc) wild-type cells are growing exponentially on 3-chlorobenzoate (3CBA), expression from the genes in the ICEclc core region (roughly the second half of ICEclc) plus the integrase intB13 is silent, whereas they are highly transcribed when cells are in stationary phase ( Figure 3A). Among others, the core region encodes genes implicated in ICEclc conjugative transfer [32,33]. P. putida with mutant ICEclc lacking either marR or tciR strongly diminished expression in the core region and of the integrase gene in stationary phase when compared to wild type ( Figure 3C, Figure S3). Lower core and integrase gene expression explains the lower ICEclc transfer rates from these mutants ( Figure 2B). In contrast, mfsR deletion resulted in much higher expression from the ICEclc core genes in exponentially growing cells ( Figure 3B), and even slightly higher expression in stationary phase than in wild-type ICEclc ( Figure S4), which explains the 10-to 100-fold ICEclc higher transfer rates ( Figure 2B). Expression of the mfsR-marR-tciR cluster itself was the same in the tciR and marR deletion mutants, and no different to the wild-type ( Figure 4A, C, E). In contrast, expression of the mfsR-marR-tciR cluster was higher in the mfsR deletion mutants than in wild-type, both in exponential and stationary phase cells ( Figure 4B, D). Since gene expression from ICEclc is similar in mutants lacking mfsR alone or mfsR plus the first 117 bp of marR ( Figure S4), we conclude that it is the LysR-type regulator encoded by tciR, which is the main activator for ICEclc core gene expression.
Microarray analysis also helped to understand the behaviour of the mfsR Km-insertion mutant ( Figure 3D). As for the tciR deletion mutant, expression of the ICEclc core region and of the integrase was dramatically lower than wild-type in stationary phase cells ( Figure 3D, Figure S4). On the other hand, both mfsR deletion and mfsR Km-insertion mutants showed increased expression of a group of genes on ICEclc coding for a putative efflux system ( Figure 3B, D). Detailed inspection of mfsR-marR-tciR expression in the Km-insertion mutant revealed that the first 160 bp of mfsR, upstream of the Km R -gene insertion were higher expressed than in wild-type cells ( Figure 4F). In contrast, the downstream genes marR and tciR were lower expressed compared to wild-type and to the mfsR deletion mutant ( Figure 4B, D, F). This confirmed, therefore, that insertion of the Km R -gene had caused a polar effect on expression of marR and tciR, which explains the strongly diminished expression of the ICEclc core genes in stationary phase in the mfsR Km-insertion mutant, and decreased ICEclc transfer.
Inserting the presumed mfsR promoter region upstream of a promoterless mcherry gene in single copy on the chromosome of P. putida UWC1 without ICEclc produced strong and homogenous mCherry expression among all cells ( Figure 5A, strain 3482). Inserting into this strain a single copy mfsR gene expressed from its own promoter abolished mCherry expression ( Figure 5A, strain 4302). Expression of mCherry from P mfsR in P. putida UWC1 (ICEclc) was very low, whereas disruption of mfsR on ICEclc again resulted in high mCherry expression ( Figure 5A, strain 3606). Complementation of this strain by a single copy mfsR gene under its own promoter caused repression of mCherry expression ( Figure 5A). All these data are consistent with the hypothesis that MfsR is repressing expression of itself and the downstream located marR and tciR genes. We then tested whether changed ICEclc transfer rates and core gene expression were in fact due to changes in the proportion of cells activating ICEclc. Hereto, a double promoter-reporter construct, carrying P int -gfp and P inR -echerry was inserted in single copy on the chromosomes of P. putida UWC1 (ICEclc), P. putida UWC1 (ICEclc-DmfsR), P. putida UWC1 (ICEclc-D'marR) and P. putida UWC1 (ICEclc-DtciR) ( Table 1). P int and P inR are the respective promoters for the integrase gene intB13 (integrated form) and the integrase activator gene inrR. Previous studies showed that both promoters are active only in a small subpopulation of cells, which are representative for transfer competent cells [26,29,30]. Consistent with previous data, the subpopulation of P. putida UWC1 (ICEclc) wild-type cells expressing P int and P inR in stationary phase suspended cultures represented a few percent ( Figure 5B). In contrast, deletion in mfsR resulted in 80-100% of cells expressing P int -and P inR -promoters ( Figure 5B). Expression of both promoters in P. putida UWC1 (ICEclc-DmfsR) occurred in early stationary phase whereas in wildtype cells their expression is maximal in late stationary phase ( Figure 5B). Conversely, P. putida UWC1 (ICEclc-DtciR) and P. putida UWC1 (ICEclc-DmarR) did not produce any detectable P intor P inR -expressing cells, neither in exponential nor in stationary phase ( Figure 5B). Considering a detection limit by microscopy of ,1 fluorescent cell among 1000-10,000 non fluorescent cells, the absence of detectable P int -or P inR -expressing cells in those mutants would be in accordance with absence of ICEclc core gene activation on microarrays ( Figure 3D) and lower transfer frequencies ( Figure 2B).

Mutants with mfsR deletion in ICEclc face a strong fitness cost
Given that P. putida UWC1 carrying ICEclc-DmfsR transferred at a much higher rate than wild-type ICEclc, and also expressed both P int -and P inR -promoters in almost all cells, we wondered why Figure 5. Effect of mutations in the mfsR region on the expression of the P mfsR -, P int -and P inR -promoters of ICEclc in P. putida UWC1. (A) mCherry expression from the mfsR promoter added in single copy to the chromosome of the indicated P. putida UWC1 strains (relevant genotypes and strain numbers specified below the graph). mCherry expression measured on individual cells (n = 1000) by epifluorescence microscopy in late exponential phase of cultures grown on 10 mM succinate and expressed as box plots (AU, arbitrary units at 20 ms exposure time). (B) Scatter plot of GFP and mcherry fluorescence in single cells of P. putida UWC1 (ICEclc) wild-type, DmfsR, DtciR or DmarR deletions, equipped with a single copy minitransposon containing the P int -egfp and P inR -mcherry fusions. Panels show expression of both markers at different growth phases, as indicated, with colors representing genotypes with independent mini-Tn5 insertions. Note as example the subpopulation of wild-type cells (dotted ellips) expressing both reporters, compared to the majority of cells in the mfsR deletion mutant but a complete absence of such subpopulation in the tciR and marR deletion mutants. doi:10.1371/journal.pgen.1004441.g005 such mutants did not become selected spontaneously. Both P. putida UWC1 (ICEclc) wild-type and (ICEclc-DmfsR) displayed statistically indistinguishable generation times during exponential growth on minimal medium with either 3CBA or succinate as carbon source (Table 2), although P. putida UWC1 (ICEclc-DmfsR) went through a longer lag phase ( Figure S5A). In contrast, the proportion of colony forming units (CFU) in samples taken from stationary phase cultures on 3CBA or succinate and plated on 3CBA solid medium was dramatically reduced for P. putida UWC1 (ICEclc-DmfsR) ( Figure 6). Whereas the number of colonies formed on 3CBA plates was the same as that on succinate for P. putida UWC1 (ICEclc), only 0.3-5.6% appeared on 3CBA plates for UWC1 (ICEclc-DmfsR) ( Table 2). Also in absolute terms, the number of CFU/ml for UWC1 (ICEclc-DmfsR) cells taken from stationary phase cultures both on 3CBA and succinate was lower than that for UWC1 (ICEclc) ( Figure 6). Moreover, 8 of 10 tested colonies of UWC1 (ICEclc-DmfsR) grown on MM plates with succinate did no longer amplify the clcA gene of ICEclc (not shown), the remaining two still being able to grow on 3CBA. Furthermore, half or more of microcolonies formed from UWC1 (ICEclc-DmfsR) with a single copy P int -egfp insertion showed incidence of malformations and cell lysis, similar to what was reported previously for nutrient-reactivated tc cells [30] ( Figure 6B), but cells in the other microcolonies divided with generation times even slightly faster (1.4960.15 h) than those in microcolonies of P. putida UWC1 (ICEclc, 1.7960.09 h). This indicates that the mfsR deletion in ICEclc imposes a strong fitness cost on P. putida UWC1. Survival of UWC1 was restored to wildtype level when the ICEclc-DmfsR was complemented by the mini-Tn7 inserted mfsR gene ( Figure S6). In contrast, neither P. putida UWC1 (ICEclc) with marR or tciR deletion, nor the mini-Tn7 complemented strains of P. putida UWC1 (ICEclc-DtciR) and Shown are a regular stationary phase microcolony of P. putida with wild-type ICEclc and a transfer competent microcolony (tcm), occurring at 1-3% frequency as reported previously [30]. For comparison, massive lysis (white arrows) and cellular malformations formed in many microcolonies of P. putida UWC1 with the mfsR deletion. doi:10.1371/journal.pgen.1004441.g006  (ICEclc-DmarR) were impaired in survival compared to UWC1 (ICEclc) ( Figure S6).
tciR is a very widespread ortholog among ICE closely related to ICEclc Using bioinformatic queries, we retrieved orthologs to tciR and mfsR from sequenced bacterial genomes, and examined manually whether they occur in chromosomal regions qualifying as ICE (e.g., presence of an integrase gene nearby, see Materials and Methods). Interestingly, orthologs to the individual components of the mfsR operon are widespread but rarely occur in the same configuration (Figure 7). So far, the ICEclc mfsR-marR-tciR configuration is only found in the ICEclc variant of Burkholderia xenovorans LB400, whereas the Tn4371-element of Acidovorax sp. strain JS42 and Aeromonas hydropohila SSU (accession number AGWR01000022.1) both carry an mfsR homolog and a MarRtype regulator immediately downstream, but no tciR equivalent nearby. There is a tciR ortholog in Acidovorax sp. strain JS42, but not on the same chromosomal region as mfsR (Figure 7).
On the other hand, tciR seems much more widespread among ICEs, as homologs can be found in ICEclc-like elements GI1 and GI6 of Bordetella petrii DSM12804, in PAGI-2 of P. aeruginosa strain C, in diverse ICEs of X. campestris pv. vesicatoria str. 85-106, suspected ICEs in Herminiimonas arsenicoxydans, Cupriavidus metallidurans CH34, and Tolumonas auensis DSM 9187, among several dozens of others (Figure 7). Given that TciR is such a common regulator found in ICEs of the ICEclc family, it might be similarly implicated in their transfer control. This regulation is likely different in detail from ICEclc, given the absence of an mfsR and marR.

Discussion
ICEclc has two distinctive modes of existence: the integrated form, which is transmitted vertically, and the circular form, which can be horizontally transferred. Previous work in our laboratory has shown that the transition between these two states occurs in only a few percent of cells in a population under stationary phase conditions [26,30,34]. We have recently suggested to name cells in which the molecular decision occurs to activate the ICEclc horizontal transfer mode transfer competent (tc) cells [30]. Single cell time-lapse experiments indicated that ICEclc transfer -at least insofar as detectable by microscopy, only occurs from tc cells, which can be distinguished through simultaneous expression of fluorescent proteins from single copy transcriptional fusions to the P int and P inR promoters of ICEclc [30]. Activation of those two promoters is the likely outcome of a multi-step regulatory cascade that orchestrates expression of some fifty genes [32], but the key factors that determine the onset of this cascade and control the extent of bistability are still obscure. Previous work provided evidence for the role of the stationary phase sigma factor RpoS in activation of ICEclc promoters, and we could show that tc cells on average have higher levels of RpoS [34]. In the present study, we report the discovery of a cluster of three regulatory genes, two of which globally control ICEclc activation and transfer, and additionally maintain a cap on fitness loss induced by the ICE (Figure 8).
The three regulatory genes occur in a unique configuration of a TetR-type repressor (encoded by mfsR), followed by a MarR-type (orf17984) and a LysR-type regulator (tciR). Transcript and microarray analysis of the locus in wild-type and mutant ICEclc, plus analysis of reporter gene expression from the mfsR promoter in a variety of host backgrounds, showed that the three genes are expressed as a polycistronic unit and are under autoregulatory control by MfsR ( Figure 5A).
Precise gene deletions and complementations indicated that tciR is likely the main global regulator of ICEclc transfer activation. Deletion of tciR caused a 2?10 3 -fold lower frequency of ICEclc transfer compared to wild type ( Figure 2B), silenced expression of the ICEclc core region ( Figure 3B) and reduced the proportion of cells expressing P int and P inR in stationary phase ( Figure 5B). Complementation of the tciR deletion on ICEclc with a single copy mini-Tn7 inserted mfsR fs -marR fs -tciR fragment ( Figure 1C) fused to the mfsR promoter region restored the expected phenotype ( Figure 2B). TciR may act either directly as a regulator on a variety of individual ICEclc core promoters, or as a ''master'' regulator in a hierarchical activation cascade (Figure 8).
The role of marR is less clear and as yet unsolved. Deletion in marR resulted in essentially the same ICEclc transcriptome profile as deletion in tciR ( Figure 3D). It also resulted in a lower transfer frequency than wild-type but not as low as the deletion in tciR ( Figure 2B), and produced no detectable reporter gene expression from P int or P inR ( Figure 5). In contrast, complementation of the marR deletion on ICEclc by a single copy marR gene through mini-Tn7 delivery ( Figure 1C) did not restore ICEclc transfer MfsR autoregulates expression of itself and of the TciR activator, without which ICEclc transfer decreases by 2?10 3 -fold. TciR may activate specifically one or more promoters on ICEclc, such as the RpoSdependent P inR -promoter [29], or a promoter upstream of the parA-like gene [32]. Expression of the P inR -promoter occurs preferentially in cells having highest RpoS levels, and only 1-3% of cells in a population in stationary phase visibly express reporter gene from P inR [29]. InrR transmits bistable activation through an unknown process to the intB13 promoter [26], and possibly simultaneously to other promoters for the genes for the conjugative system in the ICEclc core region [32]. The exact mechanism of arisal of bistability is unclear as yet. For gene locations on ICEclc, see Figure 1A. doi:10.1371/journal.pgen.1004441.g008 ( Figure 2B). Furthermore, mutants with deletions in mfsR or mfsR plus the first 116 bp of marR ( Figure 1B), behaved quite similar in transfer frequency ( Figure 2) and showed similar ICEclc transcriptomes ( Figure S4). The marR gene therefore seems to have no clear role in ICEclc core gene expression.
The most surprising effect of deletions in mfsR was a complete deregulation of the ICEclc core gene expression. This became obvious from frequencies of ICEclc transfer being 10-100 fold higher than wild type, approximating 1 transfer per donor cell ( Figure 2B). The deregulation was also obvious in microarray data showing the ICEclc core region in the mfsR deletion mutants being already transcribed in exponential phase ( Figure 3B, Figure S3). Finally, 80-100% of individual cells in stationary phase expressed the reporter genes from P int and P inR in the mfsR deletion strain compared to 3-5% in wild-type ( Figure 5B). This can be explained by the fact that deletion of mfsR would abolish autorepression, which would lead to constant high expression of marR and tciR. This overinitiates ICEclc core gene expression, leads to more cells entering the tc state and to higher transfer rates. The balance of mfsR control appears to be extremely delicate, since even complementation with a single gene copy under control of the original promoter results in a stronger effect than the wild-type, both for tciR and mfsR ( Figure 2B). The delicate balance became also obvious from the polar effects of insertion of the Kmresistance gene within mfsR, leading to decreased transcription of marR and tciR, diminished core gene expression ( Figure 3D) and reduced transfer rates (Figure 2A).
The finding that deletion and complementation of mfsR or tciR drastically changes the proportion of cells activating ICEclc, could imply that the bistability seen in wild-type situation (i.e., 3-5% of cells in stationary phase becoming transfer competent) is a result of feedback at this locus. Cells activating ICEclc in the wild-type situation could arise as a consequence of ''sloppy'' control by MfsR, incidentally causing a few cells to escape its control and transcribing marR and tciR. We think this is an unlikely scenario, because mCherry expression from the mfsR promoter is homogenous among cells ( Figure 5A). Alternatively, there might be a chemical ligand that specifically binds to MfsR in a small subset of cells, upon which its repression is relieved in those cells. The resulting TciR would then be the necessary activator to trigger ICEclc core expression in cells with on average highest RpoS levels [29] (Figure 8). On the other hand, even though mfsR may be the first level of control, bistability may also originate at later checkpoints in the regulatory cascade, which depend on the presence of sufficient TciR.
Quasi-global appearance of transfer competence across all cells in the mfsR deletion mutant resulted in massive fitness loss (Figure 6), which became evident at two levels. First of all, timelapse observations indicated lysis and aberrant cell growth in more than 50% of microcolonies ( Figure 6B). This lysis and growth arrest are similar to what we previously described as being a side consequence of becoming transfer competent in wild-type cells [30], and is caused by the parA-shi gene products on ICEclc [30]. Secondly, there was a strong loss of the capacity to grow on 3CBA among cells sampled from stationary phase cultures of the mfsR mutant compared to wild-type ( Figure 6A, Figure S6), indicative for loss of ICEclc and counterselection against maintaining ICEclc-DmfsR. However, those cells that maintained ICEclc-DmfsR could still grow on 3CBA and showed indistinguishable exponential growth rate (Table 2). This paradox can be understood when modeling the number of tc cells in batch culture populations for ICEclc wild-type (probability of tc arisal, P tc , in stationary phase of 0.025) and for the ICEclc-DmfsR mutant (P tc = 0.5). This model ( Figure S5B) shows that whereas a large proportion of tc cells appear in ICEclc-DmfsR mutant cultures in stationary phase, these can only divide 2-3 times upon reinoculation into fresh medium before lysing. This causes an apparent prolongation of a lag phase visible as stagnant culture turbidity, but does not influence the overall predicted population exponential growth rate in batch culture ( Figure S5B).
As expected from the postulated role of TciR, its complementation in trans also leads to increased ICEclc transfer, but interestingly, only the mfsR deletion caused strongly decreased cell survival ( Figure S6). We therefore hypothesize that ICEclc activation may follow two separate processes: transfer and tc cell growth arrest [30], that may both be initiated at the mfsR locus. Deleting mfsR would then deregulate both processes, whereas expressing tciR in trans would only increase activation through the transfer branch (Figure 8).
The configuration of the mfsR-marR-tciR operon of ICEclc is unique, but tciR alone is a very common part of ICE similar to ICEclc (Figure 7). We therefore speculate that mfsR-marR are a more recent acquisition in ICEclc, which drastically changed the expression of the tciR gene. Unfortunately, expression of tciR analogs in other ICEs has not been studied and very little has been reported on the transferability of ICEs related to ICEclc. The exceptions being GI3 of B. petrii that transfers at extremely low frequencies (,10 -7 ) [35], and the P. aeruginosa PAGI-2 element for which transfer has not been detected at all [16]. In comparison, wild-type ICEclc transfers at rates of 10 -2 to 10 -3 per donor (Figure 2), suggesting that it was perhaps the acquisition of the mfsR regulatory control that led to expression of transfer activity in a larger proportion of cells in the population. As we show here, the downside of increasing the proportion of ICEclc tc cells is an increase of the proportion of cells displaying growth arrest through the shi-parA pathway [30]. Likely, the MfsR autoregulation evolved to a stage of permitting efficient transfer but avoiding too much fitness loss to the population. Even though the mechanistic details are different for ICEclc, double control layers are more common for various ICEs and typically involve a variety of regulators acting on each other and/or in response to specific chemical ligands [18,19,[23][24][25]. It will be highly interesting to further study the mechanistic details of the control systems that maintain very low ICE transfer rates, and to understand whether and how such control can evolve to allow hyperefficient transfer. Table 1 lists the strains used in this study. Escherichia coli DH5a (Gibco Life Technologies, Gaithersburg, Md.), E. coli DH5a lpir, E. coli BW20767/pRL27 were cultured at 37uC on Luria-Bertani (LB) medium [36]. Pseudomonas species were cultured at 30uC on LB or 21C minimal medium (MM) [37] complemented with one of the following carbon sources: 0.5, 5, or 10 mM 3-chlorobenzoate (3CBA), 15 mM succinate or 10 mM fructose. Antibiotics were supplemented to the growth medium to select for maintenance of genetic constructions at the following concentrations: kanamycin (Km) 25 mg/ml, chloramphenicol (Cm) 20 mg/ml, rifampicin (Rif) 50 mg/ml, nalidixic acid (Nal) 50 mg/ml, gentamicin (Gm) 20 mg/ ml, and ampicillin (Ap) 100 mg/ml.

Strain constructions and DNA techniques
DNA purification, PCR, restriction enzyme digestions, DNA ligations and electro-transformations were performed according to standard procedures [36]. Deletions in ICEclc genes were created by double recombination techniques as described elsewhere [29,38]. Nucleotide positions are given according to AJ617740 (ICEclc). Primers used for strain constructions are listed in table S1.
For complementation of P. putida UWC1 (ICEclc-DmfsR, strain 4322) we first amplified the mfsR gene plus the 429 bp upstream region containing the mfsR promoter using PCR. This fragment was cloned into pGEM-T-easy and verified for correctness by DNA sequencing. The fragment containing the correct mfsR region was then recovered by restriction enzyme digestion with PstI and BamHI, and ligated into the mini-Tn7 vector pUC-miniTn7-Gm [39]. After transformation and verification in E. coli, the mini-Tn7 construct was introduced into P. putida UWC1 (ICEclc-DmfsR) by using the pUX-BF13 helper plasmid [40]. Clones resistant to Gm were selected and verified by PCR for correct insertion of the mfsR DNA in the attTn7 locus. To complement P. putida UWC1 (ICEclc-DtciR, strain 4321) and (ICEclc-DmarR, strain 4372) we amplified the complete mfsR-marR-tciR locus including the 429-bp upstream region. This fragment was cloned into pGEM-T-easy and again verified for correctness by DNA sequencing. The fragment was recovered by digestion with BamHI and StuI, and ligated with the mini-Tn7 vector. A frameshift was then introduced in the mfsR coding region by digestion at the unique NcoI-site, filling in using Klenow and religation. This will cause premature ending of the mfsR gene product (mfsR fs ). A second frameshift was subsequently introduced to inactivate the marR gene product, using the unique BsgI-site (marR fs ). After transformation and verification in E. coli, the construct was introduced in P. putida UWC1 (ICEclc-DtciR) as outlined above. This procedure was repeated to create a fragment with frameshifts in mfsR and in tciR (using the unique KpnI site), but maintaining an intact marR. This construct was introduced into P. putida UWC1 (ICEclc-DmarR). Gm-resistant clones were verified by PCR for the correct insertion at the attTn7-site, and for the presence of ICEclc.
A 656-bp region upstream of mfsR was amplified by PCR and fused to a promoterless mcherry gene. This fragment was introduced in single copy on the chromosome of P. putida UWC1, P. putida UWC1 (ICEclc) or P. putida UWC1 (ICEclc-DmfsR) using mini-Tn5 delivery. Three independent Km-resistant colonies were verified by PCR for the correct insertion and stored individually. P. putida UWC1 mini-Tn5-P mfsR -mcherry and P. putida UWC1 (ICEclc-DmfsR) mini-Tn5-P mfsR -mcherry were then further used as recipient to introduce the mini-Tn7-mfsR construct.

Random mutagenesis and screening
Random mini-transposon insertions in P. knackmussii B13 were generated by mobilization of the pRL27 suicide plasmid from E. coli BW20767 in a biparental mating. Hereto both strains were each cultured overnight in 3 ml LB, pelleted down, resuspended in 50 ml sterile saline solution (0.9% NaCl), mixed in a 1:1 (v/v) ratio and incubated on the surface of an LB agar plate for 24 hours at 30uC. The mixture was then resuspended with 1 ml saline solution, which was inoculated in 100 ml MM with 0.5 mM 3CBA plus Km to select for the mini-transposon insertion and Cm to counterselect against E. coli, and incubated at 30uC for 16 h with orbital shaking (180 rpm). An aliquot of 3 ml of this pool of enriched Km R B13 mutants was used en masse as donor in a subsequent mating procedure. Hereto, cells from the 3 ml suspension were pelleted by centrifugation, washed with 3 ml sterile saline and mixed with 3 ml of suspension of P. putida UWC1 recipient, that had been grown for 16 h on LB, was pelleted by centrifugation and resuspended in sterile saline. The mating mixture was again centrifuged, the cell pellet was resuspended in 50 ml sterile saline solution and spotted on the surface of a MM agar plate containing 0.5 mM 3CBA. The mixture was incubated for 72 hours at 30uC, after which the cells were washed from the plate with 1 ml sterile saline, which was further serially diluted and plated on MM agar plates with 5 mM 3CBA plus Km and Rif to select for transconjugants carrying mutant ICEclc. Individual colonies were purified, recultured in organized 96-well format and stored at -80uC after addition of and mixing with glycerol to 15% (v/v). Libraries were replicated and regrown in 100 ml LB plus Rif for 16 h in 96-well microtiter plates, mixed with 100 ml P. putida UWC1 Nal R recipient suspension, and incubated at 30uC for 48 h. Then 50 ml of each well was reinoculated into 170 ml of MM containing 5 mM 3CBA plus Km, Rif and Nal, and growth was measured by continuous OD-measurements in a multiplate reader (FluoStar Omega, BMG labtech). Absence of growth was taken as indication for absence of ICEclc transfer, in which case the donor culture was recovered for mapping of the transposon insertion.

Insertion mappings
DIG-labeled primers 070934 or 070935, annealing to one of the ends of the Km R insert but facing outward, were used (separately) in single-primer PCR with DNA from mutant UWC1 donors as templates. The reactions produced oligonucleotide probes with the 59-DIG label, the sequence of the end of the Km R gene and the adjacent sequence of the ICEclc insert position. Such products were used for rough localization of the insertion position by hybridizing to macroblot membranes (Eurogentec, UK), whose set of 55-mer oligonucleotides covers most of the ICEclc genes. Hybridization and detection of the DIG-marker were carried out according to the manufacturer's instructions (Roche Diagnostics GmbH, Mannheim, Germany). Once the insertion was roughly mapped on ICEclc, PCR-based sequencing was used to determine exact position of the Km R -gene insertion.

ICEclc transfer assays
The frequency of ICEclc transfer was determined in experimental conditions described previously [33]. P. putida UWC1 ICEclc wild-type or mutant derivatives were used as donors, whereas P. putida UWCGC (constitutively fluorescent, Gm R ) or P. putida UWC1 Km R were used as the recipient (Table 1). Briefly, donors and recipient were each cultivated on 5 mM 3CBA MM and 10 mM fructose MM, respectively, and combined on 0.5 mM 3CBA agar plates as a single concentrated pellet. After 48 hours incubation at 30uC, mating mixes were resuspended, diluted and plated on 5 mM 3CBA MM agar (counting of donor CFU) or 5 mM 3CBA Gm or Km agar (counting of transconjugant CFU). Transconjugants were checked by PCR and frequencies were expressed as the number of transconjugant CFU per donor CFU. Donor survival was used for the data shown in Figure S6.
Time-lapse microscopy P. putida UWC1 strains were precultured for 16 h at 30uC in LB medium, after which 100 ml were transferred to 20 ml fresh MM 4 mM 3CBA medium in presence of the appropriate antibiotics. This culture was incubated for 96 hours at 30uC and 200 rpm shaking, after which the cells were 100-fold diluted in MM without C-source and inoculated on agarose surfaces (gel patches) for timelapse microscopy [41]. Medium for gel patches consisted of 1% agarose dissolved by heating into MM with 0.1 mM 3CBA. Gel patches were created by pipetting 130 ml of the agarose-MM-3CBA solution kept at 55 uC on the surface of a circular cover glass (42 mm ø and 0.17-mm thick), placed in an autoclaved perfusion chamber (POC, H. Saur, Reutlingen, Germany), separated with a 0.5 mm thick silicon spacer ring and covering them with a second cover glass. After solidification of the agarose, the upper cover slip was removed and 6 ml of the diluted cell suspension was placed onto the agarose gel patch. As soon as the drops were dried on the surface, the patches were turned upside down and placed bacteria-facing-down on a new round cover glass [41]. A second silicon spacer ring was added to allow air circulation within the closed chamber and the glass sandwich was fixed into the metal cast POC chamber with a metal ring. Up to four patches could be placed simultaneously within a single glass sandwich in a POC chamber.
Microcolony development was followed directly on a Nikon Inverted Microscope Eclipse Ti-E, equipped with a Perfect Focus System (PFS), pE-100 CoolLED and a Plan Apo l 10061. 45 Oil objective (Nikon), installed in a controlled temperature room (22uC). Ten random regions of every patch were imaged automatically during 48 hours with intervals of 1 h, in Phase Contrast mode (10 ms exposure), eGFP (500 ms) and eCherry (500 ms). Images were recorded using Micro-Manager 1.4 (http:// www.micro-manager.org/) and fluorescence values were extracted using MetaMorph (Series 7.5, MDS, Analytical Technologies).

P. putida UWC1 ICEclc and mutant fitness tests
Triplicates of strains UWC1 (ICEclc) and UWC1 (ICEclc-DmfsR) were grown for 16 h in LB medium at 30uC. Both strains were then 500-fold diluted (starting OD 600 0.001) in MM with 5 mM 3CBA or 10 mM succinate. Upon reaching early stationary phase, strains were again diluted into fresh MM (starting OD 0.001) with the same carbon source, and growth was followed by frequent culture turbidity measurements (OD 600 ). 24 h after reaching stationary phase, each replicate culture was serially diluted in MM and plated onto MM agar plates with 5 mM 3CBA or with 10 mM succinate. The number of CFU/ml was scored and the ratio was calculated between the number of CFU/ml on MM agar with 5 mM 3CBA and the number of CFU/ml on MM with 10 mM succinate.
Ten randomly chosen colonies of UWC1 (ICEclc-DmfsR) cultivated in MM with succinate and grown on MM-succinate agar plates were retested for growth on MM agar with 5 mM 3CBA. The presence of ICEclc was determined by colony PCR on the same colonies by amplifying the clcA gene, which is carried by ICEclc and the gene product of which is essential for 3CBA metabolism.

Bioinformatic screening for ICE related to ICEclc
Homologues to tciR of ICEclc were detected by BLASTN to the nr/nt database at E-value ,1?10 -15 . The corresponding whole or draft genome sequences were retrieved and compaired by aligning to ICEclc (Accession number AJ617740.2) using Megablast. Detected regions were manually recovered and searched for the tciR homologue and an intB13 homologue within a 1-100 kb window. If annotated, the presence of a gene for tRNA-Gly nearby the intB13 homologue was scored. Regions covering all criteria (i.e., homology to ICEclc core region, tciR homologue and presence of integrase gene) were retained as containing putative ICE. Selected regions were further individually pair-wise compared by using the Artemis Comparison Tool within the WebACT service [42]. Figure S1 Outline of the random mutagenesis and subsequent selection procedure. Original ICEclc-host P. knackmussii B13 is randomly mutagenized by miniTn5-mediated insertions of kanamycin resistance inserts (Km R ). Mutant B13 are selected by culturing cells on minimal medium (MM) with Km and 3chlorocatechol (3CBA) as sole carbon and energy source. The pool of B13 mutants is cultured in batch and mixed with recipient strain P. putida UWC1 (resistant to rifampicin, Rif R ). The mixture is incubated in mating conditions for 72 hours and plated on MM with 3CBA, Km and Rif, to select for transconjugants. Individual colonies of transconjugants were restriked and organized into a mutant library in 96-well plates. Each mutant is used as donor in a new 96-well mating with recipient P. putida UWC1 resistant to Rif and nalidixic acid (Nal R ). Individual mating mixtures are grown on MM agar with 3CBA, Km, Rif, Nal in order to select for transconjugants. In absence of transconjugant growth, the donor of that particular mating was recovered from the library and had its insertion position mapped. Dots indicate the 2 log-fold change of hybridization signal per microarray probe in the comparison, plotted at their distance along the ICEclc sequence (X-axis; in kb). A scheme of ICEclc is redrawn at the bottom of each section, with regions of interest as grey boxes (+ or -indicate the DNA strand on which the region is encoded). Grey bars in the background indicate the two-fold cutoff level. (TIF) Figure S5 Observed and expected population growth in cultures of P. putida with wild-type ICEclc and the DmfsR mutant. (A) Measured turbidities in triplicate batch cultures growing on 3CBA as sole carbon and energy source. Note the increased lag time of the P. putida DmfsR mutant. (B) Modeled population growth of all cells (brown) and tc cells (red) in two subsequent batch cultures growing on 3CBA; the second being inoculated with 1:1000 volume from the first culture in stationary phase. Scenarios show predicted behavior for wild-type ICEclc (with a probability P tc of 0.025 of tc cell appearance in stationary phase) and for DmfsR (with P tc = 0.5). Note how a tc population appears in stationary phase, which is transferred to a new culture, but rapidly dies as a result of activation of the par-shi system [30]. Note further how exponential growth rates remain the same for both wild-type ICEclc and the DmfsR mutant. Modeling based on Monod kinetics, using the following parameters: total C = 4.25 mg C/ml, dry weight of a single cell = 2?10 -12 g, yield = 0.3 g/g, K s = 0.02 mg/mL, generation time of non-tc cells = 1 h, generation time of tc-cells = 2 h, death rate of tc cells = 0.4 h -1 . (TIF) Figure S6 Population sizes of P. putida UWC1 carrying wild-type (2737, 2738) or mutant ICEclc, measured as colony forming units (CFU) on 5 mM 3-chlorobenzoate agar plates, per ml of resuspended culture spotted and incubated for 48 h on 0.5 mM 3-chlorobenzoate containing agar medium. (A) P. putida UWC1 strains: 4165, ICEclc with a deletion in the genes for the suspected efflux system (as unrelated control); 4321, tciR deletion; 4322, mfsR deletion; 4372, marR deletion; versus P. putida UWC1 (ICEclc), 2738. (B) P. putida UWC1 strains: 2737, ICEclc wild-type; 4321, tciR deletion; 4322, mfsR deletion; 4372, marR deletion; 4646, mfsR deletion but complemented in trans by a single copy mini-Tn7 inserted mfsR gene under its own promoter; 4649, tciR deletion but complemented in trans by a single copy mini-Tn7 inserted fragment with the (mfsR fs )-(marR fs )-tciR genes under the mfsR promoter. 4804, marR deletion but complemented in trans by a single copy mini-Tn7 inserted fragment with the (mfsR fs )-marR-(tciR fs ) genes under the mfsR promoter. (A) and (B), Independently carried out experiments on different occasions and by different scientists. Letters indicate statistically indistinguishable groups identified from ANOVA tests on biological replicates, followed by Tukey's post hoc testing. P-values indicate the significance for the overall group difference (a versus b), or in one specific case, between the samples connected by the line. (TIF)