In order to identify MOB_F12 plasmids, an initial search using a set of 26 known MOB_F12 relaxases (Table S1) taken from previous studies (Garcillán-Barcia et al., 2009; Alvarado et al., 2012) was carried out. MOB_F12 relaxases were defined as those having (D/E)NYY and D(L/F)TF amino acid motifs in the N-terminal relaxase domain of the protein (Garcillán-Barcia et al., 2009). These 26 relaxases were used as baits in protein BLAST searches of the NCBI plasmid database (6079 plasmids, 20th October 2015) using a threshold e-value of 1E-25. In this way, we retrieved a total of 256 plasmids containing relaxases that showed at least 40% sequence identity at the protein level to its closest database reference. This threshold was selected upon realizing that relaxases with low identity (down to 26%) were retrieved in an initial search. However, when we compared two of the most distant MOB_F12 plasmids, for instance, F and pAsa5, they showed an ID of 49%. We therefore established as selection criteria that IncF/MOB_F12 plasmids are those whose relaxase showed at least 40% ID with respect our homemade MOB_F12 relaxase DB.

To construct the presence/absence profile of the tra proteins of MOB_F12 plasmids, Psi-blast searches (Altschul et al., 1997) were performed against a protein database constructed from the annotations of the 256 MOB_F12 plasmids (Table S1) using the following conjugative proteins as queries: F plasmid proteins TraJ, TraA, TraL, TraE, TraK, TraB, TraP, TrbD, TrbG, TraV, TraR, TraC, TrbI, TraW, TraU, TrbC, TraN, TrbE, TraF, TrbA, ArtA, TraQ, TrbB, TrbJ, TrbF, TraH, TraG, TraS, TraT, TraD, TrbH, TraI (N-terminal 300 amino acids), and TraX as well as TraM, TraY, and FinO from plasmid R100. By default, hits below an e-value of 1E-3 were considered as positive hits. Selected protein hits were aligned using MUSCLE (Edgar, 2004). The resulting global alignments were used to reconstruct maximum-likelihood (ML) phylogenies using RAxML version 7.2.7 (Stamatakis, 2006). Twenty ML trees were executed using the JTTGAMMA model, and 100 bootstrap trees were inferred to obtain the confidence values for each node of the best ML tree.

To compute the co-occurrence matrix between tra genes, we used the presence/absence profile of all tra genes in our 256 plasmid dataset. Thus, for each tra gene, we defined a vector of 256 elements, with values 1 or 0. We then calculated the Hamming distance between each pair of vectors. Co-occurrence between two tra genes was expressed as the maximum possible distance (256) minus the Hamming distance between the pair of genes.

The 3D structures of TraJ_V, EntFR, and SphTR were predicted by homology modeling using the Phyre2 server (Kelley and Sternberg, 2009) Images of the resulting 3D models were generated using Pymol (DeLano Scientific, Palo Alto, CA, USA).

In order to identify IncF/MOB_F12 plasmids present in the databases, we used the conjugative relaxase gene as the lowest common denominator. This property of the relaxase to serve as classification guide was shown in previous works, and is widely used for plasmid typing (Garcillán-Barcia et al., 2009; Alvarado et al., 2012). In order to identify MOB_F12 relaxase-containing plasmids present in the NCBI plasmid database (6079 plasmids; October 20, 2015), we employed a set of 26 known MOB_F12 relaxases as baits (Table S1), as indicated in Materials and Methods. We retrieved a total of 256 plasmids containing relaxases that showed at least 40% sequence identity at the amino acid level to their closest database reference (see M&M). The resulting plasmid list is shown in Table S1. To reconstruct the phylogeny of MOB_F12 plasmids, we aligned the N-300 residues of the relaxase proteins as described in Materials and Methods and constructed a ML phylogenetic tree. We rooted the tree using the MOB_F11 relaxase TrwC_R388 as outgroup. The resulting tree is shown in Figure 1. The MOB_F12 phylogenetic tree includes plasmids isolated from α and γ-Proteobacteria, with 91% coming from species within Enterobacteria. The tree showed that 88% of the plasmids coming from enterobacterial species clustered in a monophyletic branch, well-supported by the bootstrap value (Figure 1, black vertical arrow). This branch included the “classical” IncF plasmids F, R1, R100, and pSLT. Enterobacterial plasmids not belonging to this branch included several from Enterobacter, which instead clustered in a second monophyletic branch (Figure 1, orange arrow). A third set of plasmids from Escherichia, Salmonella, and Klebsiella appeared in a third monophyletic branch (Figure 1, red arrow). As we will show later, these clusters contain plasmids harboring a typical MOB_F12 conjugation region, but showing different regulatory systems. Plasmids from α-Proteobacteria appeared in an ancestral, monophyletic group (Figure 1, green arrow).

In order to determine the degree of association of MOB_F12 relaxases to the canonical F pilus, we determined the presence/absence of homologs of the F conjugation genes in our 256 plasmid dataset. For this purpose, we selected a total of 36 genes present in the conjugation region of plasmids F and R100, two well-studied IncF prototypes. As described in M&M, the presence of homologs of these 36 reference genes was determined by PSI-BLAST. Table S1 shows the accession numbers of each homolog identified for the 256 plasmids analyzed. Table S1 consists of a matrix in which rows (i) correspond to each MOB_F12 plasmid, while columns (j) indicate each tra gene. Thus, reference numbers in the i,jth position of the matrix correspond to the homolog to protein j present in plasmid i. We transformed this table into a binary matrix, such that each i,jth position was 1 if there was a homolog detectable by PSI-Blast, and 0 otherwise. This allowed us to determine the level of overall conservation of F conjugation genes among the 256 plasmids. Results showed that a TraD-like protein (the coupling protein (T4CP) of F-like plasmids) could be detected in 91% of the plasmids (Figure 2A). This intimate phylogenetic association between the relaxase and the coupling protein was shown previously to be a hallmark of mob genes (Fernández-López, et al., 2006; Garcillán-Barcia et al., 2011). The analysis of other plasmid groups showed, however, that the association between MOB and MPF genes is less stringent. For example, MOB_F11 relaxases are associated to N or W pili (variants of MPF_T) (Fernández-López, et al., 2006; Garcillán-Barcia et al., 2011). As shown in Figure 2A, the presence of MPF_F genes could be detected in more than 80% of the MOB_F12 plasmids in our dataset. Results also indicated that the conservation of MPF_F genes was not uniform. The most conserved MPF gene was traG, responsible for mating pair stabilization, which appeared in 88% of the plasmids. The least conserved gene was trbH, which could only be detected in 8% of the plasmids. Interestingly, genes that have been described as essential for F transfer appeared in more than 80% of the plasmids (Figure 2A, highlighted in red), while non-essential genes tend to appear at lower frequencies (Figure 2A, in black). An exception to this rule was traX, a gene that encodes an acetylase of pilin subunits. This gene was deemed non-essential for F plasmid conjugation (Maneewannakul et al., 1995), yet it was detected in 88% of the plasmids analyzed.

Since not all MPF_F genes showed the same degree of conservation, we wondered whether there were genes that showed preferential co-occurrence. To determine this, we built a co-occurrence matrix for the 36 MPF_F genes (Materials and Methods). Results (Figure 2B) showed that the highest co-occurrence values corresponded to the gene clusters traEBKL, trbCI, and traHUWNCF. These genes are essential components for the synthesis and function of F-pili, and thus their co-occurrence suggests the presence of functional F transfer systems. Interestingly, genes trbA, trbG, and artA also showed a high level of co-occurrence, despite their overall conservation is among the lowest overall (<25% identity). This indicates that this gene cluster is specific of a certain set of F-like plasmids. It was also noteworthy that regulatory genes (traM, traY, traJ, and finO) showed a lower degree of co-occurrence than structural genes, suggesting that alternative regulatory mechanisms could exist for MPF_F conjugation systems.

Since regulatory genes showed lower conservation than structural genes, we looked specifically at three key regulatory genes, namely traM, traJ, and finO. We did not include traY because, given its small size (around 225 bp), it is often not properly annotated. In order to distinguish between alternative regulatory schemes and major deletions that might have eliminated a substantial fraction of the transfer region, we also included a marker gene for the presence of the MPF apparatus. For this purpose, we used the essential ATPase traC. Using these genes as guidelines, we identified five major groups of F-like plasmids, which corresponded to major branches in the MOB_F12 phylogenetic tree.

The first and major MOB_F12 group includes a total of 200 plasmids (78% of the total), residing in the genera Escherichia, Salmonella, and Klebsiella. In the MOB_F12 tree of Figure 1, these relaxase genes are monophyletic. If we extrapolate from the relaxase to the whole TRA system, the tree structure implies that group A TRA_F system arose from a common ancestor that spread among these three bacterial genera. Plasmids of this group share a “classical” F plasmid conformation. Transfer genes are organized in a long, polycistronic operon, where gene synteny is preserved. Figure 3 shows the presence/absence table of tra genes in group A. It shows that we could detect the entire set of proteins deemed essential for F conjugation in a total of 150 plasmids. The essential genes, which were determined by transposon insertion analysis (Ippen-Ihler et al., 1972; Wu et al., 1987, 1988; Moore et al., 1990; Kathir and Ippen-Ihler, 1991; Maneewannakul et al., 1991, 1992; Maneewannakul and Ippen-Ihler, 1993), are shown in red in Figure 3. Remarkably, the regulatory components are also preserved among members of this group. TraM and TraJ appear in 100% of these plasmids. TraY homologs could be identified in 80% of the plasmids, but the real figure is probably higher, given the small size of the protein and lack of proper annotation. All MOB_F12 group A plasmids were finO+, the only exception being plasmid F itself. Since the finO phenotype in the F plasmid is due to an IS3 insertion, it is thus highly likely that the “original” F plasmid was also repressed.

Within group A, a total of 50 plasmids lacked some essential component of the transfer machinery. Within this group, we identified a set of small plasmids (about 15 kb long) present in Shigella sp. and E. coli O104:H4 Shiga-toxin containing species (Figure 3, bottom). These plasmids seem to have suffered massive deletion of the TRA region, with only traI, traX, and finO genes remaining. This should result in a non-transmissible plasmid, since these plasmids neither contain the essential genes for pilus formation, nor the coupling protein TraD. In some cases, the relaxase itself appears truncated. The presence of plasmids with this particular structure among Shigella and Shiga-toxin containing E. coli is puzzling. The presumptive inability of these plasmids for horizontal mobilization would point out to the vertical propagation of a single deletion event in the ancestral line shared by Shigella sp. and E. coli O104:H4. However, members of this group do not form a monophyletic branch in the relaxase tree (Figure 1), which would suggest repeated but independent deletion events. Further research is needed to clarify the evolutionary history of these plasmids, the functional advantage of these deletions, if any, and their relationship to the pathogenesis of Shiga-toxin containing enterobacteria.

A second group of F-like plasmids comprises a set of plasmids from Yersinia pestis (Figure 4A). Their relaxases appear as a monophyletic branch in the MOB_F12 tree (Figure 1), showing an ancestral relationship to plasmids from group A (Figure 1). Structurally, they are characterized by a bipartite operon structure (Figure 4B), with genes involved in relaxosome formation (traD, traI) transcribed divergently from genes involved in conjugative pilus formation. Group B preserves all F essential genes and also traP, traR, traR, trbI, traQ, trbB, and traX. They all contain finO, yet none contain homologs of traM, traJ, or traY. Moreover, we could not find any putative transcriptional regulator in the vicinity of their conjugation regions, opening the questions of (a) how this mating system is regulated and (b) whether these plasmids are self-transmissible, given the lack of relaxase-accessory proteins or a recognizable origin of transfer. Outside the conjugation region, group B plasmids show extensive homology to plasmid pMT. pMT plasmids are a fundamental component of Yersinia pathogenesis, carrying essential virulence determinants for flea colonization (Hu et al., 1998). Besides, it is known that pMT plasmids from all three Yersinia pestis biovars (Antiqua, Medievalis, and Orientalis) are not self-transmissible and contain no transfer genes. Indeed, all Yersinia plasmids with a MOB_F12 conjugation systems belong to isolates of Yersinia pestis pestoides, an atypical Y. pestis group, probably the closest to the ancestral lineage that gave rise to the pandemic biovars (Garcia et al., 2007). Incorporation of pMT to Y. pestis has been traditionally linked to horizontal gene transfer from other enterobacterial species (Hu et al., 1998; Lindler et al., 1998). According to the phylogeny shown in Figure 1, MOB_F12 group B pMT plasmids stemmed from group A plasmids. Specifically, group B plasmids are monophyletic with two group A plasmids from Klebsiella sp. (pKOX_NDM1 and pRJF866) that contain the entire repertoire of essential F genes. However, given the lack of relaxase-accessory proteins (traM and traY), it is unclear whether group B plasmids are self-transmissible.

Although all plasmids from the MOB_F12 group share a common mating apparatus (MPF_F), only some IncF plasmids contain the same replication and partition machineries. Thus, some IncF plasmids are able to co-reside together in the same cell, while others are not. Classical incompatibility testing identified several IncF subgroups, which were numbered from I to VII (de la Cruz et al., 1979). One of these IncF subgroups, IncFV, stood out because of its particular regulatory scheme. In our analysis, we identified a set of plasmids that display a typical IncFV arrangement (Figure 5). Plasmids from group C form a monophyletic branch in the MOB_F12 relaxase tree (Figure 1). They are characterized by a single promoter architecture. Conserved homologs include the same genes as in group A plasmids (Lu et al., 2002). The most conspicuous difference is that, although they encode a protein that is called TraJ, this protein actually shows no detectable homology to TraJ from Group A plasmids. Moreover, group C plasmids lack any recognizable homolog for FinO, yet experimental analysis showed that these plasmids are not de-repressed (Lu et al., 2002). To avoid confusion with the classical TraJ protein, hereafter we shall name this protein TraJ_V.

According to structural analysis of TraJ_V with Phyre2, this protein is predicted to contain a DNA/RNA-binding domain (DBD) formed by a 3-helical bundle fold in the 70 N-terminal amino acids. This type of DBD is found in transcriptional regulators such as LuxR/UhpA family. Most LuxR-type regulators act as transcriptional activators. They contain an HTH domain in the C-terminal part of the protein and an effector binding domain in the N-terminal domain. However, in TraJ_V the HTH domain is located in the N-terminal half of the protein. A comparison between the predicted 3D structure for TraJ_V N-terminal domain and the solved structure for F plasmid TraJ N-terminal domain (pdb 4KQD) (Lu et al., 2014) showed no structural homology (Figure 5C). In fact, TraJ is predicted to be similar to the canonical LuxR protein, with an N-terminal effector binding domain and a C-terminal DBD. Thus, TraJ_V retains the DBD of TraJ, but the different position of this DBD and the differences in the rest of the protein suggest that TraJ_V could play a different role than TraJ in the control of MOB_F12 group C plasmid conjugation.

A fourth group of F-like plasmids comprises members from another monophyletic branch in the MOB_F12 tree. This branch includes plasmids mainly coming from Enterobacter and the closely-related genus Pantoea, but also contains plasmids from other enterobacteria like Erwinia, Rahnella, and Kluyvera (Figure 6A). These plasmids maintain the classical group A genetic organization: a single operon including all tra genes, except traM and traJ regulators, (Figure 6B). Genes that were deemed essential for F transfer are also conserved among plasmids of group D, along with traV, trbI, traN, and trbB. Remarkably, group D plasmids are FinO and TraJ negative. However, the locus occupied by traJ in Group A contains a DNA-binding protein that is conserved within the group. We named this putative regulator EntFR.

According to the Phyre2 prediction, EntFR is a ribbon-helix-helix (RHH) DBD (Figure 6C). Although this DBD is not as common as the helix-turn-helix (HTH) domain, it is frequently found in accessory proteins that bind to the origin of conjugative transfer (TrwA in plasmid R388, TraJ in plasmid RP4). In F plasmids TraM and TraY proteins present this fold. Structurally, EntFR is thus more similar to these proteins than to TraJ. EntFR contains two RHH domains, a feature that is also shared by TraY. Based on this, it is possible that EntFR acts as the functional homolog of TraY in group D plasmids. However, given that there are known RHH containing proteins able to act as transcriptional activators (Schreiter and Drennan, 2007), the possibility of EntFR being the functional homolog of TraJ cannot be ruled out.

A fifth group of F-like plasmids includes plasmids from the genus Sphingomonas, the relaxases of which form a monophyletic branch in the MOB_F12 tree (Figure 1). These plasmids exhibit a conserved architecture that is different from the arrangement in the other MOB_F12 groups (Figure 7). Instead of a long, single operon, tra genes from group E plasmids are split in two convergent operons. The first contains the genes responsible for the formation of the conjugative pilus (MPF_F genes), while the second includes the genes involved in relaxosome formation (Figure 7B). This architecture is reminiscent of that of conjugative plasmids with VirB-like mating apparatus, such as the MOB_F11 (IncW and IncN) groups. However, although the architecture of group E is VirB-like, all the constituent genes are related to those of plasmid F, i.e., they belong to the MPF_F family. Genes deemed essential for F conjugation are preserved in group E plasmids, although traA, the gene coding for the conjugative pilin, is significantly different. Apart from these essential genes, group E plasmids also maintain clear homologs of trbI and traN. Regarding the regulatory components of the transfer machinery, group E plasmids lack homologs of TraM, TraJ, TraY, or FinO. The only putative regulator that can be identified by BLAST analysis is a small, conserved protein present immediately upstream traD. In plasmids with MPF_T (VirB-like) conjugation systems, this position is usually occupied by a protein coding for a ribbon helix-helix relaxase-accessory protein (e.g., TrwA in plasmid R388) (Moncalián and de la Cruz, 2004; Varsaki et al., 2009). Relaxase-accessory proteins participate in relaxosome assembly as well as in regulation of the expression of relaxosome components. In group E plasmids, this small protein can be thus considered as the hallmark of the group; it will be named SphTR (for Sphingomonas transfer regulator). Structural modeling of SphTR using Phyre2 showed that this protein belongs to to the ribbon helix helix (RHH) superfamily of DNA binding proteins. Thus, it is likely that SphTR fulfills the mobilization-accessory role in this group of plasmids.

Plasmids containing MOB_F11 relaxases, like IncN and IncW plasmids, show an operon structure similar to group E plasmids. However, while in IncN and IncW plasmids it is possible to identify the regulators responsible for the independent control of the two operons (Fernandez-Lopez et al., 2014), we were unable to find any other putative DNA binding proteins in the vicinity of Group E transfer genes. Thus, it is unclear whether these genes are controlled by other plasmid/host regulators, by SphTR, or are not transcriptionally regulated. In any case, Group E plasmids constitute a valuable divergent evolutionary line of MOB_F12 plasmids that, at least in genome organization and regulatory components, represents a bridge between MOB_F11 and MOB_F12.

A set of 14 plasmids remain unassigned in our group classification. They correspond to MOB_F12 plasmids that were found outside the Enterobacteriaceae. Members of this group include plasmids from Vibrio, Aeromonas, Legionella, Fluoribacter, and Piscirickettsia. Although many of them contain an entire set of tra genes (like for example pLELO-like plasmids from Legionella), their genetic organization and putative regulators (indicated in Table S2) do not seem to be shared by the different plasmids (nor with Group E plasmids from Sphingomonas). It is entirely possible that this is due to under-representation of these genera in the nucleotide databases, compared to clinically-relevant enterobacteria. Anyhow, these plasmids serve to demonstrate that the IncF conjugation system is not restricted to the enterobacteria and that the MOB_F12 plasmid clade can assimilate a number alternative of regulatory proteins.

IncF plasmids are frequent carriers of antibiotic resistance genes and virulence factors, and a common finding in clinically relevant enterobacteria. Clinical microbiologists differentiate IncF plasmids using a sequence-typing system that takes advantage of the allelic diversity that IncF plasmids present in their replication regions (Villa et al., 2010). Analyzing replicon variants, Villa et al. were able to differentiate several IncF groups, according to their replicon sequence type (RST) (Villa et al., 2010). A comparison between replicon typing and the analysis of conjugation regions, revealed that plasmids with an assignable IncF RST belonged to Group A and Group B IncF plasmids, two groups that are monophyletic in the relaxase phylogenetic tree (Figure 1, black arrow). Group A plasmids, which include all classical IncF plasmids, present different RST profiles (Villa et al., 2010), indicating that within this broad group there is substantial sequence variation. Plasmids from groups C, D and E, however, cannot be assigned a typical RST profile, indicating that IncF plasmids in these groups are likely to exhibit different replication mechanisms.

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.