Derivatives of strain E. coli DH5α (Grant et al., 1990), containing different combinations of conjugative and/or mobilizable plasmids were used as donors in mating experiments (Table 1). A rifampicin-resistant derivative of E. coli MDS42 (Posfai et al., 2006) was used as the recipient strain. MOB_F11 conjugative plasmids were R388 (Datta and Hedges, 1972), pIE321 (Gotz et al., 1996), pMBUI4 (Brown et al., 2013), R7K (Coetzee et al., 1972), or R388ΔkfrA-osa (del Campo, 2016). Mobilizable plasmids used were the kanamycin-resistance derivatives of ColE1 (Van Rensburg and Hugo, 1969), and RSF1010 (Derbyshire et al., 1987), and the ampicillin-resistance derivative of CloDF13 (van Putten et al., 1987). A list of co-resident plasmids can be found in Table 1.

When appropriate, antibiotics (Apollo) were added at the following concentrations: ampicillin (Ap; 100 μg/ml), chloramphenicol (Cm; 25 μg/ml), kanamycin (Km; 50 μg/ml), nalidixic acid (Nx; 20 μg/ml), rifampicin (Rif; 50 μg/ml), tetracycline (Tc; 10 μg/ml), and trimethoprim (Tp; 20 μg/ml). Arabinose (Sigma-Aldrich) 0.001% (w/v) was used as transcription inducer. Bacterial cultures were set up in LB-broth and LB-agar (Pronadisa). M9-salts (Sigma-Aldrich) were used to resuspend bacteria after mating and to perform serial dilutions.

Coupling protein gene trwB from plasmid R388 (Datta and Hedges, 1972) was amplified by PCR using primers FKpnITrwB and RHindIIITrwB. Gene pifC from plasmid pOX38 (Chandler and Galas, 1983) was amplified using primers FKpnIPifC and RHindIIIPifC. Gene osa from plasmid R388 (Datta and Hedges, 1972) was amplified using primers FKpnIOsa and RHindIIIOsa (Supplementary Table 1). These primers introduced KpnI and HindIII sites at one and another end of the respective amplicon. Each amplicon, and the pBAD33 vector (Guzman et al., 1995), were digested with KpnI and HindIII endonucleases (Thermo-Fisher). Final constructs pBAD33::trwB, pBAD33::pifC, and pBAD33::osa were obtained after ligation, using T4 DNA ligase (Thermo-Fisher), and electroporation into E. coli DH5α competent cells.

Donor and recipient cultures in stationary phase were washed in LB-broth and mixed in a 1:1 donor-recipient ratio. Then, a 200 μl mix was centrifuged and resuspended in 15 μl LB-broth. 5 μl of this mixture were placed on top of 96-well microtiter plate wells containing 150 μl LB-agar and conjugation was allowed to proceed for 1 h at 37°C. Conjugation of a derepressed R27 derivative, drR27, was performed for 2 h at 25°C. In all cases, bacteria were then resuspended in 150 μl M9-salts and corresponding dilutions plated on selective media. Conjugation frequencies were estimated as the number of transconjugant cells per donor (T/D) and means were calculated using decimal logarithms of data.

Comparison of the means between two different conditions was carried out by using t-test tool from GraphPad Prism^® (v 5.0) biostatistics software (San Diego, CA).

Psi-blast searches, using proteins Osa_R388 (GenBank Acc. No. FAA00056.1), FiwA_RP4 (CAJ85704.1), FiwB components KlaA_RP4 (CAJ85667.1), KlaB_RP4 (CAJ85666.1), and KlaC_RP4 (CAJ85665.1), PifC protein RepC_F (NP_061420.1), FinC protein OrfD_CloDF13 (NP_052375.1), FinQ_R621a (YP_004823737.1), and FipA_pKM101 (AAC63100.1) as queries, were carried out against the 6,878 plasmid set available at NCBI database in May 12th, 2016 (ftp://ftp.ncbi.nlm.nih.gov/refseq/release/plasmid/). Hits (threshold e-value 0.001) were aligned with MUSCLE using default parameters (Edgar, 2004). HMMER 3.0 (Eddy, 2011) was used to build hidden Markov model (HMM) profiles from the alignments, which were in turn used in HMM searches against the plasmid database. For the plasmids encoding the retrieved hits, relaxase MOB families were defined as described in Garcillan-Barcia et al. (2009) and Guglielmini et al. (2011), and mating pair formation types (MPF) were defined as in Guglielmini et al. (2014) and (http://conjdb.web.pasteur.fr/).

Transfer of IncW plasmids was systematically analyzed in the presence of a representative set of conjugative plasmids in donor bacteria. Depending on antibiotic resistance, either plasmid R388 or pIE321 were tested as prototype IncW conjugative systems. R388 was substituted by pIE321, with a backbone genome 97% identical to R388 (Revilla et al., 2008), when the co-resident plasmid encoded trimethoprim resistance (as in the cases of R751 and pCTX-M3). Conjugation frequencies for each IncW conjugative system and co-resident plasmid were normalized to the mean value of the tested system in the absence of any co-resident plasmid. The results obtained are summarized in Table 1. As shown in the Table, R388 conjugation was significantly affected by the presence of four different plasmids. Specifically, it decreased by 4 logs in the presence of either IncFI plasmid pOX38::CmR (F derivative), IncI1 plasmid R64drd11, IncP-1α plasmid pRL443 (a kanamycin-sensitive derivative of plasmid RP4), or IncX2 plasmid R6K. Except for the IncN plasmid pKM101, which caused a 3-fold increase in R388 conjugation, the other co-resident plasmids tested (IncFII, IncP-1β, IncX1, IncL/M, and IncHI1) produced no significant change in R388 or pIE321 conjugation. While FI of R388 by IncP-1α plasmids was known to be caused by fiwA and fiwB genes (Fong and Stanisich, 1989) and R6K inhibited R388 transfer by a reported but unidentified mechanism (Olsen and Shipley, 1975), FI caused by IncFI and IncI1 plasmids had not been reported before.

In turn, IncW plasmids affected the transfer of co-resident IncP-1 plasmids pRL443 and R751. A 2-log reduction was observed on R751 conjugation and 1-log reduction on pRL443 transfer, paralleling the previously observed effect of R388 on transfer of IncP-1 plasmid RP1 (Olsen and Shipley, 1975).

The F plasmid was previously described to inhibit the fertility of plasmid RP4 through its gene pifC (Miller et al., 1985). Gene pifC is present also in the F reduced version pOX38. A 99% identical pifC homolog is present in R64drd11, an IncI1 plasmid that also affects RP4 conjugation (Datta et al., 1971). The fact that pifC-containing plasmids pOX38 and R64drd11 inhibited conjugation of RP4 and R388, positioned pifC as a potential responsible gene in all cases. To prove this fact, pifC from pOX38 was cloned in vector pBAD33, giving rise to plasmid pBAD33::pifC. Then, conjugation of donor cells containing either IncW, IncP-1, IncN, or IncF conjugative plasmids was tested in the presence or absence of pBAD33::pifC. As shown in Figure 1, conjugation of the IncW prototype plasmid R388 and the IncW-like plasmid pMBUI4 were significantly affected by pifC expression in donors. This result was comparable to that exerted over the IncP-1 plasmids and indicated that pifC is actually the gene responsible for IncW conjugation inhibition by IncFI and IncI1 plasmids. However, the impact on R388 fertility of pifC expressed from pBAD33::pifC was unexpectedly weaker than the impact of the complete F plasmid, suggesting that one or more additional F loci might be involved in inhibiting R388 transfer. In addition, PifC had no effect on the transfer frequency of the IncN plasmid pKM101, and pifC overexpression did not affect pOX38 conjugation.

The R388 origin of transfer, TrwA accessory protein, TrwB coupling protein, and TrwC relaxase comprise the R388 MOB genetic module, involved in conjugative DNA processing prior to plasmid transfer. Among these components, previous work pointed to TrwB as the target of various FI systems. This was the case of FipA of pKM101 and PifC of F, which inhibited RP4 conjugation by affecting TraG coupling protein (Santini and Stanisich, 1998), and Osa of pSa, which prevented T-DNA transfer to plant cells by limiting substrate binding to VirD4 coupling protein (Cascales et al., 2005).

To demonstrate that TrwB was also targeted by PifC, we analyzed transfer of three mobilizable plasmids with different requirements in the presence and absence of pifC. For mobilization, R388 provided the MPF genes to construct the mating channel that connects donors with recipients. Plasmids ColE1 and RSF1010 need R388 coupling protein as well as the MPF genes for mobilization, whereas CloDF13 only needs R388 MPF because it encodes its own coupling protein (Cabezon et al., 1994). The results of these experiments are shown in Figure 2. While CloDF13 mobilization was not affected by pifC, transfer of ColE1 and RSF1010 plasmids was inhibited when expressing pifC. The absence of effect in CloDF13 mobilization discarded R388 MPF components as potential targets, confirming our previous hypothesis. The inhibitory effect of pifC in ColE1 and RSF1010 mobilization suggests that R388 coupling protein is the most likely target of this FI system. The other, more unlikely, option is that oriT, relaxase, or accessory proteins of R388, ColE1, and RSF1010 were independently affected by pifC. However, previous experiments showing high mobilization frequencies of ColE1 by either F and R64 helper plasmids (Cabezon et al., 1997), both encoding pifC, indicate that ColE1 mobilization machinery is not affected by pifC. To reinforce this data, RSF1010 plasmid was mobilized by the pifC-unaffected conjugative plasmid pKM101. The absence of effect when employing pKM101 coupling protein indicates that R388 coupling protein was the actual target of pifC inhibition. As controls, we included IncP plasmids R751 and pRL443 as helpers of ColE1, RSF1010, and CloDF13 mobilization, which behaved similarly to R388.

With the aim of improving R388 transfer efficiency in the presence of plasmid pOX38, we overexpressed TrwB coupling protein in donor cells. As observed in Figure 3, increased expression of trwB partially relieves inhibition, the frequency of transfer increasing 6-fold when compared to the frequency in the presence of basal trwB. As a control, in the absence of pOX38, trwB overexpression did not produce a significant change in R388 conjugation. In addition, the empty expression vector pBAD33 did not cause any impact in pOX38 inhibitory effect. Overall, the significant alleviation of R388 transfer by trwB overexpression indicates that the activity of the coupling protein is affected by PifC.

As shown in Table 1, IncW plasmids prevented conjugation of IncP plasmids pRL443 and R751. Although a similar effect was previously found (Olsen and Shipley, 1975), the responsible gene was not identified. Most IncW plasmids encode osa, a FI function affecting transfer of A. tumefaciens T-DNA to plant cells (Close and Kado, 1991). Plasmids of the IncW family that lack this gene, such as R7K (Coetzee et al., 1972) or the R388 derivative R388ΔkfrA-osa (del Campo, 2016), when used as co-residents with pRL443 and R751, resulted in conjugation frequencies as high as in donors containing just the IncP plasmids (Figure 4A). This result indicates that Osa is the FI function affecting IncP plasmids. To confirm this hypothesis, we cloned osa of R388 in the pBAD33 vector and expressed it in donor cells. As observed in Figure 4B, conjugation of IncP plasmids was inhibited when osa was expressed in donor cells. In this case, as expected, the FI effect is stronger when osa is expressed from pBAD33::osa instead of R388 plasmid. On the other hand, osa did not affect transfer of the IncN plasmid pKM101. It is worth to mention that conjugation of the osa⁺ IncW plasmid R388 was slightly affected by osa overexpression in donor cells. This inhibition could be related to the fact that osa and fiwA are homologs and thus, osa overexpression could mimic FiwA activity to some extent.

In order to check if the molecular target of Osa was the IncP coupling protein, we analyzed the mobilization of plasmids with and without their own coupling protein (CloDF13 and ColE1/RSF1010, respectively), using affected and unaffected helper plasmids (pRL443/R751 and pKM101, respectively) in the presence of osa. The results obtained are shown in Figure 4C. While mobilization of ColE1 and RSF1010 by IncP plasmids pRL443 and R751 was inhibited by osa, CloDF13 transfer was resistant to the inhibitory effect. These results suggested that TraG, the coupling protein of IncP plasmids, was the molecular target of Osa. The absence of effect when RSF1010 was mobilized by the unaffected plasmid pKM101 discarded other MOB components as potential targets, confirming our hypothesis.

From the FI genes named and reported in the literature, the gene sequence is available for just a few, namely those encoding factors Osa, FiwA, FiwB (KlaA, KlaB, and KlaC proteins), PifC, FinC, FinQ, and FipA. To assess their abundance and distribution in different plasmid families, we searched for homologs of these FI proteins in NCBI plasmid database. At least one protein with homology to one of the protein families was found for 129 out of the 6,878 plasmids contained in the database (Supplementary Table 2). Although FI was studied mainly in plasmids from enterobacteria and agrobacteria, other bacterial classes are represented in the list. The abundance of each protein is shown in Figure 5. KlaC and FipA were the most abundant. Profile-profile comparisons showed that Osa and FiwA were homologs, as well as KlaA and KlaB. The remaining protein families showed unrelated HMM profiles.

The FiwB-mediated FI phenotype exhibited by IncP-1α on IncW plasmid transfer requires three proteins, KlaA, KlaB, and KlaC (Goncharoff et al., 1991). Only 7 out of the 52 klaC-encoding plasmids also encoded klaA and klaB genes. In turn, nine plasmids lacking klaC but encoding other fiwB genes were found (three contained KlaA and KlaB, while six KlaA or KlaB). Although FiwB was described and studied in an IncP-1α/MOB_P/MPF_T plasmid, it is absent from other IncP-1 groups and is more widely found in different replicons containing a MOB_H/MPF_F transfer system (6 out of the 7 klaABC-encoding plasmids).

Each FI family comprises homologs retrieved from both, highly related backbones and plasmids from different replication, MOB and MPF groups. The exception was the FinC family, which comprises only two members encoded in highly-similar mobilizable backbones. Within each protein family, the closest homologs were carried by highly-identical plasmid backbones (i.e., members of the same Inc group). Plasmids of the same replication group tend to encode the same type of FI factor (e.g., FipA in IncN plasmids). Nevertheless, plasmid membership to an Inc group does not necessarily anticipate the presence of a common FI factor in all members. Conserved backbones can encode different factors (e.g., some IncI1 plasmids encode pifC, while others encode finQ; fiwA is present in the α subgroup of IncP-1 plasmids, but is absent in the rest of subdivisions).

Plasmids tend to encode a single FI factor. The exception, pKPC-LKEc, is a cointegrate that contains three replication initiators, two MOB regions and a MPF system. With only five exceptions, the retrieved plasmids contained either a MOB relaxase and/or a MPF system. Besides, only two plasmids smaller than 30 kb, CloDF13 (9.9 kb) and ColEST258 (13.6 kb), were found to encode a FI factor (FinC). Thus, the proteins analyzed here seem to be associated with plasmids transmissible by conjugation, and especially with conjugative plasmids.

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