Antibiotic Resistance in Vibrio cholerae: Mechanistic Insights from IncC Plasmid-Mediated Dissemination of a Novel Family of Genomic Islands Inserted at trmE

The increasing association of the etiological agent of cholera, Vibrio cholerae serogroup O1 and O139, with multiple antibiotic resistance threatens to deprive health practitioners of this effective tool. Drug resistance in cholera results mainly from acquisition of mobile genetic elements. Genomic islands conferring multidrug resistance and mobilizable by IncC conjugative plasmids were reported to circulate in non-O1/non-O139 V. cholerae clinical strains isolated from the 2010 Haitian cholera outbreak. As these genomic islands can be transmitted to pandemic V. cholerae serogroups, their mechanism of transmission needed to be investigated. Our research revealed plasmid- and genomic island-encoded factors required for the resistance island excision, mobilization, and integration, as well as regulation of these functions. The discovery of related genomic islands carrying diverse phage resistance genes but lacking antibiotic resistance-conferring genes in a wide range of marine dwelling bacteria suggests that these elements are ancient and recently acquired drug resistance genes.

C holera is an acute diarrheal disease that leads to severe dehydration and often death in the absence of adequate treatment (1). The seventh cholera pandemic, which began in 1961, is caused by toxigenic strains of the Gammaproteobacteria Vibrio cholerae serogroup O1 biotype El Tor, or more sporadically its O139 variant (1). Since the late 1980s, antibiotic-resistant V. cholerae strains have emerged and spread globally (2). Development of drug resistance in seventh pandemic V. cholerae has been ascribed to mutation but it mostly involves acquisition of mobile genetic elements, including genomic island GI-15, integrative conjugative elements of the SXT/R391 family, and conjugative plasmids of incompatibility group C (IncC) (3)(4)(5).
IncC plasmids are large (Ͼ120-kbp) broad-host-range conjugative plasmids frequently associated with multidrug resistance in several species of globally distributed pathogenic enterobacteria, and in seventh pandemic V. cholerae strains from Africa, China, and Haiti (4,(6)(7)(8). Conjugative transfer of IncC plasmids is controlled by the FlhCD-like heteromeric transcriptional activator AcaCD that they encode (9). AcaCD activates 17 promoters conserved in IncC plasmids, including those driving expression of transfer genes and operons encoding type IV secretion system (T4SS) and conjugative pilus, relaxase TraI, putative type IV coupling protein (T4CP) TraD, and unknown function protein TraJ (9). TraI belongs to the MOB H1 family of relaxases, and together with the product of mobI C , is essential for initiation of conjugative transfer at the origin of transfer (oriT) (10,11). MobI C is responsible for recognition of the oriT locus of IncC plasmids that is located immediately upstream of mobI C (12). Unlike other transfer genes, mobI C seems to be expressed in an AcaCD-independent manner (9).
Furthermore, AcaCD also triggers excision of at least three types of genomic islands (GIs) shown to be mobilizable in trans by IncC plasmids. The first type, integrated at the 3= end of the gene of unknown function yicC, is exemplified by the 16.5-kbp MGIVmi1 of Vibrio mimicus (9). The two other types of GIs are inserted into the 3= end of trmE (also known as mnmE or thdF), a gene encoding the 5-carboxymethylaminomethyluridine-tRNA synthase GTPase subunit (13,14). One type of GI is illustrated by the 42.4-kbp Salmonella genomic island 1 (SGI1) that confers resistance to ampicillin, chloramphenicol, streptomycin/spectinomycin, sulfamethoxazole, and tetracycline (AC-SSuT) in Salmonella enterica (14,15). The other is illustrated by the 47.4-kbp MGIVchHai6 of V. cholerae HC-36A1 that confers not only the ACSSuT phenotype but also trimethoprim and possibly mercury resistance (Fig. 1A) (13). MGIVchHai6 was found in non-O1/non-O139 V. cholerae strains isolated from patients exhibiting symptoms of cholera at the onset of the 2010 cholera outbreak in Haiti. Despite integrating into the same site as SGI1, MGIVchHai6 encodes a distantly related integrase Int (67% identity) and recombination directionality factor (RDF) Xis (37% identity) (13). MGIVchHai6 also lacks most of the genes and sequences that enable the mobilization of SGI1 by IncC plasmids (13,16,17). On the basis of these structural differences, the mechanisms of mobilization of MGIVchHai6 and SGI1 by IncC plasmids are expected to differ considerably. In MGIVchHai6, xis is the last gene of a putative operon-like structure preceded by a putative AcaCD-controlled promoter (Fig. 1A). A second AcaCD binding site is located inside an open reading frame (ORF) that encodes a distant homolog of MobI C (27% identity over two fragments of 109 and 53 amino acid residues) (13). Besides AcaCD, IncC plasmid-and GI-encoded factors involved in MGIVchHai6 mobilization have not been characterized.
In this report, we established a model of mobilization of MGIVchHai6 by helper IncC plasmids and compared this model to SGI1 mobilization. Deletion mutants of the helper plasmid and MGI were used in mating assays to characterize the contribution of each element in MGIVchHai6 mobilization. The presence of AcaCD binding sites upstream of xis and inside the putative gene encoding a MobI C homolog suggests the presence of AcaCD-responsive promoters that were verified using lacZ reporter fusions. By analogy with IncC plasmids, we hypothesized that the MobI C homolog encoded by MGIVchHai6 recognizes a cognate oriT locus located upstream of its gene. The ability of IncC plasmids to mobilize chromosomal DNA through MGIVchHai6 was also tested and provided the directionality of transfer initiated at oriT. Finally, phylogenetic analyses based on the mobilization factor revealed the existence of a large class of potential IncC-mobilized GIs that are integrated at three different chromosomal sites in mostly marine dwelling species of Gammaproteobacteria. Our results show that MGIVchHai6like GIs share a mechanism of mobilization by helper IncC plasmids that differs from the one used by SGI1-like GIs.

RESULTS
int and xis are essential for excision and mobilization of MGIVchHai6. MGIVch-Hai6 carries a large cargo of antibiotic and mercury resistance genes. To make it more Green flags indicate the position and orientation of predicted AcaCD binding sites (13). The black flag indicates the position and orientation of the P int promoter. The insertion sites of In36A1 integron and Tn6310 transposon are shown. The gene numbers correspond to the last digits of the respective locus tags in GenBank accession no. AXDR01000001. ACSSuTmT, resistance to ampicillin, chloramphenicol, spectinomycin/streptomycin, sulfamethoxazole, trimethoprim, and tetracycline; Hg, mercury resistance. (B) Mobilization assays of MGI Kn or its Δint, Δxis, Δ86, or Δ85 mutants were carried out using E. coli GG56 (Nx) bearing pVCR94 Sp as the donor strain, and CAG18439 (Tc) as the recipient strain. Complementation assays were performed in the donor (D) or recipient (R) strain by expressing the missing gene from P BAD on pBAD-int or pBAD-xis. An "ϫ" indicates that the transfer frequency was below the detection limit (Ͻ10 Ϫ7 ). Bars represent the means Ϯ standard errors of the means (error bars) from three independent experiments. Statistical analyses were carried out on the logarithm of the values using a one-way analysis of variance (ANOVA), followed by Dunnett's multiple-comparison test with the wild-type (WT) MGI Kn as the control. Statistical significance is indicated as follows: ****, P Ͻ 0.0001; ns, not significant. (C) ␤-Galactosidase activities of P int , P 85 , and P 84 transcriptionally fused to lacZ. Colonies were grown on LB agar with or without arabinose to induce acaCD expression from pacaCD. (D) Induction levels of P int , P 85 , and P 84 in response to AcaCD. ␤-Galactosidase assays were carried out using the strains of panel C. Ratios between the enzymatic activities in Miller units for the arabinose-induced versus noninduced strains containing pacaCD are shown. The AcaCD-regulated promoter P traHs of SGI1 served as a positive control, and cells devoid of pOPlacZ served as a negative control. The bars represent the means Ϯ standard errors of the means (error bars) from two independent experiments. amenable for this study, we constructed a kanamycin (Kn)-resistant mutant of MGIVch-Hai6 that lacks In36A1, Tn6310, and res (Fig. 1A) To establish whether int is required for mobilization of MGIVchHai6, we carried out mobilization assays using pVCR94 Sp and a Δint mutant of MGI Kn . Deletion of int abolished MGI Kn mobilization (Fig. 1B). Complementation assays done by expressing int from the arabinose-inducible promoter P BAD on pBAD-int restored mobilization to the wild-type level only when pBAD-int was present in both the donor and recipient strains. Therefore, int is likely required for excision of MGI Kn in donor cells and for its integration into the chromosome of recipient cells.
The predicted (RDF) gene xis encodes a 114-amino-acid (aa)-residue protein containing a predicted helix-turn-helix (HTH_17, Pfam accession no. PF12728) domain (position 39 to 89). To assess the role of xis, we mobilized the Δxis mutant of MGI Kn using pVCR94 Sp . Transfer of this mutant was reduced ϳ200-fold compared to the wild type (Fig. 1B). Complementation using pBAD-xis in donor cells was sufficient to restore the mobilization of MGI Kn Δxis to the wild-type level, thereby confirming that Xis is required only in donor cells, likely to facilitate Int-mediated excision of the GI.
xis is located downstream of two open reading frames (ORFs), 85 and 86 (the gene numbers correspond to the last digits of the respective locus tags in GenBank accession no. AXDR01000001), of unknown function. The predicted translation products of 85 is a 65-aa-residue protein that, like xis, contains an HTH_17 domain (position 11 to 62). 86 encodes a predicted 558-aa-residue protein containing a domain of unknown function (DUF927, Pfam accession no. PF06048) in its N-terminal half. Mobilization assays using Δ85 and Δ86 mutants of MGI Kn and pVCR94 Sp revealed that neither 85 nor 86 is involved in the excision, mobilization, or integration of MGI Kn under laboratory conditions, as the frequency of mobilization remained unaffected by deletions (Fig. 1B).
To validate the involvement of int and xis in the excision step, we carried out PCR excision assays to detect the attP site on the plasmid-like form of excised MGI Kn , using E. coli GG56 (nalidixic acid) bearing MGI Kn or MGI Kn Δint. Spontaneous excision was undetectable, as shown by the absence of attP PCR product (see Fig. S1 in the supplemental material). Overexpression of int from P int did not trigger excision, whereas overexpression of xis resulted in excision of MGI Kn , but not of MGI Kn Δint. Thus, excision of MGIVchHai6 is induced by Xis and requires Int, in line with their proposed roles of RDF and integrase, respectively.
AcaCD-dependent activation of mobility genes of MGIVchHai6. AcaCD binding sites were previously detected upstream of 85 and inside 84 (13). To test whether AcaCD activates promoter sequences upstream of int and 85 or inside 84, P int , P 85 , and P 84 were introduced upstream of a single-copy, chromosomal promoterless lacZ gene cassette that is transcriptionally isolated by the terminator sequences rgnB and tl3 (9). The AcaCD-responsive promoter sequence P traHs of SGI1 (17) was used as a positive control. The ␤-galactosidase activity of each promoter was monitored in the presence and absence of ectopically expressed acaCD. P int yielded a weak yet constitutive ␤-galactosidase activity regardless of the presence of AcaCD (Fig. 1C), confirming that the promoter that drives expression of the integrase gene is not controlled by AcaCD. In contrast, P 85 , which likely drives expression of the RDF gene xis, did not appear to produce detectable ␤-galactosidase activity in the absence of AcaCD (Fig. 1C). When acaCD was expressed, P 85 activity increased 35-fold (Fig. 1D). Finally, P 84 exhibited a weak, constitutive expression similar to P int and weaker than P traHs in the absence of AcaCD (Fig. 1C). Like P traHs , P 84 was strongly induced (55-fold increase) upon expression of acaCD (Fig. 1D).
These results confirm that the two promoter sequences P 85 and P 84 containing the predicted AcaCD binding sites are activated by AcaCD. In contrast, P int drives low-level constitutive expression of the integrase gene.
MobI M is required for MGIVchHai6 mobilization. The 795-bp open reading frame 84 (locus VCHC36A1_0084 of V. cholerae HC-36A1) of MGIVchHai6 encodes a distant homolog of MobI C (27% identity). Since MobI C is a key factor for IncC plasmid transfer, we wondered whether 84 could play an important role in mobilization of MGIVchHai6. To test the hypothesis, we constructed MGI Kn Δ84 and carried out mobilization assays with pVCR94 Sp . MGI Kn transfer was strongly impaired by this deletion and reduced ϳ2,700-fold ( Fig. 2A). In this context, we observed a slight (12-fold), yet statistically significant, increase of pVCR94 Sp transfer. Complementation of the Δ84 mutation from pBAD-84 restored MGI Kn mobilization to the wild-type level, while reducing transfer of pVCR94 Sp ϳ40-fold, suggesting that 84 is essential for MGIVchHai6 mobilization. However, the position of the AcaCD-responsive P 84 promoter inside 84 suggested that the protein effective for mobilization is smaller than predicted. To test this hypothesis, we constructed complementation plasmids containing fragments of 84 starting at the following alternative start codons: ϩ31, GTG; ϩ106, GTG; ϩ118, ATG; ϩ304, ATG; and ϩ409, TTG (Fig. 2B). All plasmids but pBAD-84-304 and pBAD-84-409 restored mobilization of MGI Kn Δ84 by pVCR94 Sp to the wild-type level (Fig. 2C). Therefore, the ORF located downstream of P 84 and starting at ATG 118 in 84 is the likely gene that allows MGIVchHai6 mobilization. This gene, hereafter referred to as mobI M , produces a putative 225-aa-residue protein.
Localization and characterization of the origin of transfer (oriT). In IncC plasmids, oriT is located directly upstream of mobI C (10,12). By analogy, we predict that oriT of MGIVchHai6 is located at the corresponding position, i.e., upstream of mobI M . To confirm this hypothesis, we first cloned the intergenic region located between 85 and 84 (IG0) as well as 84 into the low-copy-number nonmobilizable plasmid pDC1. The resulting plasmid, pIG0-84, was mobilized by pVCR94 Sp at a frequency comparable to that of MGI Kn (Fig. 2D), thereby confirming the cloned fragment sufficient to support mobilization by the IncC plasmid even in the absence of MGIVchHai6. Site-directed mutagenesis G120C (ATG to ATC) and G306C (ATG to ATC) in 84 confirmed that ATG 118 is the start codon of mobI M since pIG0-84 carrying the mutation G120C was not mobilizable, whereas mutation G306C had no effect on transfer (Fig. 2D).
Since we were able to use pBAD-84-118 to complement and mobilize the Δ84 mutant of MGI Kn that lacks the 117-bp segment containing the AcaCD-responsive promoter P 84 , the possibility that oriT could be located in this sequence or in mobI M was ruled out. Therefore, we cloned IG0 and the region extending upstream of mobI M (IG10) into pDC1 and carried out mobilization assays using pVCR94 Sp to provide the conjugative machinery. Donor cells also carried pT84 to provide MobI M . The empty vector and the same vector bearing IG0 in the absence of pT84 were both unable to transfer (Fig. 3A). In contrast, when pT84 was present in donor cells, the vector bearing IG10 or IG0 transferred at frequencies that were comparable to that of MGI Kn Δ84 complemented with pT84. This result indicated that oriT is located within IG0.
Insert reduction was then performed to find the minimal sequence of IG0 allowing mobilization by pVCR94 Sp . The smallest insert capable of acting as a suitable oriT, IG8, was 49 bp long and located immediately upstream of 84 (Fig. 3A). Although functional, IG8 provided only 1/10th of the mobilization activity of IG0 or larger inserts such as IG6 or IG13 that provided additional upstream or downstream sequences. The addition of either 43 bp upstream (IG6) or 30 bp downstream (IG13) led to transfer frequencies equivalent to that obtained with IG10, and addition of both fragments (IG12) did not enhance transfer further. Predicted folding of IG12 revealed three potential stem-loop structures within IG8 and on either side, highlighting the presence of repeated sequences potentially involved in relaxosome assembly (Fig. 3B).
Directionality of transfer initiated at oriT. To determine the direction of conjugative transfer initiated at oriT of MGIVchHai6, chromosomal markers located upstream and downstream of trmE, the integration site of MGIVchHai6, were tested for mobilization. Accordingly, MGI Cm or the excision-defective mutant MGI Cm Δxis were introduced together with pVCR94 Sp into Escherichia coli BW25113 derivatives carrying a kanamycin resistance (Kn r ) marker integrated at uhpC, yidX, purP, or gidB (Keio knockout collection) (18). These genes are located between 7 and 39.5 kb on either side of trmE (Fig. 3C). Mobilization assays failed to produce any transconjugants when the Kn r marker was inserted upstream of trmE. In contrast, transfer was easily detectable for Kn r insertions at purP and gidB that are located downstream of trmE, regardless of the  (Table 1). (D) Confirmation of ATG 118 as the genuine start codon of mobI M . Conjugation assays were performed using E. coli GG56 (Nx) containing the specified elements as donor strains and either CAG18439 (Tc) (A and C) or VB112 (Rf) (D) as the recipient strain. The bars represent the means Ϯ standard errors of the means from three independent experiments. Statistical analyses were carried out on the logarithm of the values using a one-way ANOVA followed by Dunnett's multiple-comparison test with the WT MGI Kn (A), pBAD-84 (C), or pIG0-84 (D) as the control. Statistical significance is indicated as follows: ****, P Ͻ 0.0001; ***, P Ͻ 0.001; **, P Ͻ 0.01; ns, not significant. ability of MGI Cm to excise from the chromosome (Fig. 3D). This result demonstrates that transfer of MGIVchHai6 initiated at oriT progresses downstream of mobI M and that the last genes translocated into the recipient cells are xis-86-85. Involvement of IncC DNA processing genes in MGIVchHai6 mobilization. To test whether DNA processing genes of IncC plasmids are involved in the mobilization of MGIVchHai6, we constructed nonpolar deletion mutants of mobI C , traI, traD, and traJ in pVCR94 Sp and carried out conjugation assays. Each individual deletion abolished transfer of pVCR94 Sp . Except for traI, all deletions could be complemented by ectopic expression of the corresponding gene (Fig. 4A). Complementation of traI could not be tested as attempts to clone an intact copy of this gene failed, suggesting that expression under P BAD , even in the absence of arabinose, is toxic.
As observed for pVCR94 Sp , deletion of traI, traD, and traJ abolished MGI Kn mobilization. Complementation of each mutation restored mobilization to the wild-type level (Fig. 4B). Therefore, these genes appear to be essential for processing and transfer of excised MGI Kn DNA. Interestingly, although MGIVchHai6 encodes its own MobI M factor, deletion of mobI C resulted in ϳ100-fold reduction of MGI Kn mobilization (Fig. 4B), whereas complementation with pBAD-mobI C restored mobilization to the wild-type level.
Likewise, we confirmed that MGIVchHai6 relies exclusively on the T4SS encoded by IncC plasmids as deletion of traH C , traG C , or traN C abolished mobilization of MGI Kn (Fig. S2). Moreover, mobilization of MGI Kn by pVCR94 Sp into a recipient strain carrying In panels A and D, the bars represent the means Ϯ standard errors of the means from three independent experiments. "ϫ" indicates that the transfer frequency was below the detection limit (Ͻ10 Ϫ7 ).
MGIVchHai6 is the prototype of a large and diverse subfamily of GIs mobilizable by IncC plasmids. Given the importance of MobI-like factors, we searched the GenBank database for MobI homologs and extracted associated sequences to assess diversity of the GIs related to MGIVchHai6 (see Table S1 in the supplemental material). Phylogenetic analyses of MobI proteins revealed that MobI homologs encoded by GIs are distinct from those encoded by conjugative plasmids and cluster in two major groups (Fig. 5A). Group A contains GIs integrated at the 3= end of trmE or yicC, and at the 5= end of dusA. With the exception of the GI inserted at dusA, all GIs encoding group A MobI proteins exhibited the same structure: int-xis-86-85-ig-mobI, where ig likely contains the oriT, and xis-86-85 and mobI are divergent and preceded by AcaCD-like binding sites. In a few GIs, variable DNA is inserted between the convergent int and xis genes. MGIVchHai6 and GIPmi1 belong to group A. GIPmi1 of Proteus mirabilis resembles MGIVchHai6, and it contains a large multidrug resistance cluster inserted at the same position but lacks the mercury resistance transposon Tn6310 (19). Most trmEspecific group A GIs encode predicted type I restriction-modification (R/M) systems. In contrast, the yicC-specific group A GIs encode predicted type 1 BREX antiphage systems.
GIs coding for group B MobI proteins are integrated either at the 3= end of trmE or 3= end of yicC. In group B GIs, the xis-86-85 cluster is separated from mobI by a large region of variable DNA often encoding predicted type I or phosphorothioation R/M systems. Furthermore, while trmE-specific group B GIs conserved the overall structure of group A GIs, yicC-specific group B GIs lack 85 and have undergone an inversion of the xis-86-to-mobI region relative to the int gene. These GIs encode predicted type I and type II R/M systems. Surprisingly, analysis undertaken in this study led to discovery of MGIVchHai6s, a group B GI inserted at trmE in a tandem fashion directly downstream of MGIVchHai6 in V. cholerae HC-36A1. Unlike MGIVchHai6, MGIVchHai6s does not carry containing the specified elements as donor strains and CAG18439 (Tc) as the recipient strain. The bars represent the means Ϯ standard errors of the means from three independent experiments. "ϫ" indicates that the transfer frequency was below the detection limit (Ͻ10 Ϫ7 ). Statistical analyses were carried out on the logarithm of the values using a one-way ANOVA followed by Dunnett's multiple-comparison test with WT MGI Kn as the control. Statistical significance is indicated as follows: ****, P Ͻ 0.0001; ***, P Ͻ 0.001; ns, not significant. antibiotic or heavy metal resistance genes. The group B GI of P. mirabilis JN40 encoding AYM48754.1 is also integrated in a tandem fashion downstream of SGI1-PmJN40, an SGI1-like element that confers multidrug resistance (20). Thus, tandem integration of different GIs in trmE is not a rare occurrence.
Phylogenetic analysis of integrases encoded by respective GIs revealed similar clustering into groups A and B within trmE-and yicC-specific integrase clusters (Fig. 5B). Although the integrase encoded by SGI1 also mediates integration at trmE, it could not be ascribed to either group with confidence (Ͻ80 bootstraps), which is consistent with the considerable structural and genetic differences between SGI1-like and MGIVchHai6like GIs.

DISCUSSION
Bacterial conjugation typically results from concerted action of the cytoplasmic relaxosome responsible for DNA processing initiated at oriT, and the type IV secretion system (T4SS) that translocates the processed DNA across cell membranes into the recipient cell (21,22). The type IV coupling protein (T4CP), an inner membraneanchored protein, acts as a relaxosome docking station at the T4SS. In most cases, the relaxosome comprises a relaxase working together with auxiliary proteins that either help or are required for the relaxase activity. Most of the auxiliary factors hitherto described alter DNA topology by either locally bending DNA at the oriT or unwinding DNA through a helicase activity (23)(24)(25)(26)(27)(28)(29). Several oriT-binding factors, such as TraJ of RP4, TrwA of R388, and Int of Tn916, take part in specific recruitment of the relaxase at this locus (30)(31)(32)(33)(34). Other known auxiliary factors are involved in recognition of the T4CP or in relaxosome stabilization by protein-protein interactions (34)(35)(36).
Together with SXT/R391 elements, IncC and IncA plasmids share a set of transfer genes encoding a T4SS of the MPF F family, a coupling protein (TraD), a relaxase of the MOB H1 family (TraI), and two essential factors thought to be part of the relaxosome (TraJ [DUF4400, PF14348] and MobI) (8,10,11,37). We have also shown here that TraJ is essential to transfer of IncC plasmids and MGIVchHai6 ( Fig. 4A and B). While results obtained with SXT and the IncHI1 plasmid R27 suggest a role as a relaxosome component (37)(38)(39), the exact function of TraJ remains unknown. MobI is required for conjugative transfer of IncC plasmids and SXT/R391 integrative and conjugative elements (ICEs) as well as GIs that mimic the oriT locus of the latter but lack a mobI gene (10,40,41). In contrast, deletion of mobI in the helper element does not abolish transfer of pCloDF13 and SGI1, which possess unique oriT loci and cognate mobilization proteins MobBC for pCloDF13 and MpsAB for SGI1 (12,16,40,42). MGIVchHai6 lacks the mobilization genes mpsAB and the oriT locus that are essential for SGI1 mobilization (13,16). Instead, we showed that MGIVchHai6 encodes MobI M , a distant homolog of the IncC plasmid-encoded MobI C that provides some independence from mobI C (Fig. 1A, 2A, and 4B). Deletion of mobI C in the IncC plasmid R16a was shown to enhance (45-fold increase) mobilization of SGI1-C, suggesting competition between SGI1-C and its helper plasmid for the conjugative machinery (12). In contrast, we found that the absence of mobI C impaired transfer of MGIVchHai6 (ϳ100-fold decrease) (Fig. 4B), suggesting that MobI C enhances initiation of transfer mediated by TraI and MobI M at oriT of MGIVch-Hai6. However, deletion of mobI M , while abolishing MGIVchHai6 mobilization, also enhanced transfer of the helper plasmid (12-fold increase) ( Fig. 2A), consistent with competition between the two elements.
We identified the oriT locus of MGIVchHai6 within the intergenic region upstream of  Table S1 in the supplemental material. mobI M (IG0) and found that a 49-bp region (IG8) was sufficient to promote mobilization of a nonmobilizable plasmid. While IG0 shares low nucleotide identity (45%) with the oriT of pVCR94 (10,12), the 49-bp IG8 shares 63% identity with this oriT, with the last 32 bp of IG8 sharing 84% with the corresponding region in pVCR94 oriT. Nevertheless, this level of conservation is notably limited compared to the similarity reported between oriT of SXT and the GIs it mobilizes (Ͼ63% over ϳ300 bp) (41). This lack of conservation likely accounts for the requirement for MobI M to achieve optimal transfer of MGIVchHai6. Together with the absence of a MobI homolog in SXT-mobilizable GIs, the specificity of MobI M and MobI C for their respective elements further supports the proposed role for MobI as an auxiliary relaxosome component involved in oriT recognition (40,43). SGI1 encodes three functional T4SS subunits, TraH S , TraG S , and TraN S , that displace the homologous subunits encoded by IncC plasmids despite strong amino acid sequence divergences (64,37, and 78% identity, respectively) (17). This alteration of the mating channel is crucial to enhance SGI1 mobilization (13,17). Incorporation of TraG S into the IncC mating pore allows SGI1 to circumvent entry exclusion exerted by an IncC plasmid in the recipient cells, allowing SGI1 to spread freely, even in a population of cells carrying IncC plasmids (17,44). In contrast, MGIVchHai6 lacks genes encoding T4SS subunits. Predictably, MGIVchHai6 was shown here to conform to IncC entry exclusion and thus is unable to transfer to a strain already containing an IncC plasmid (see Fig. S2 in the supplemental material).
In SGI1, AcaCD binding sites have been identified upstream of xis, S004-rep, traN S , traHG S , and S018 and corresponding promoters shown to respond to AcaCD activation (9,17,(45)(46)(47). In MGIVchHai6, putative AcaCD binding sites were predicted at the 5= end of 84 and upstream of an operon-like gene cluster containing xis, which encodes a putative protein sharing only 37% identity with the RDF Xis of SGI1 (13). In this study, Xis is indeed an RDF acting together with Int to catalyze excision under the control of AcaCD ( Fig. 1B and C and Fig. S1). Regulation of xis and mobI by AcaCD is consistent with the previously proposed model in which these GIs remain quiescent in the chromosome in the absence of an IncC plasmid (43).
MGIVchHai6-like elements have been detected in silico in environmental and clinical O1 and non-O1/non-O139 V. cholerae isolates from the Indian subcontinent and South America, and in Shewanella sp. from North America, although those seem to be devoid of antibiotic resistance genes (13). IDH᎑03944, a cotrimoxazole resistance-conferring GI related to MGIVchHai6, was recently reported in a 2011 isolate of V. cholerae O44 recovered from diarrheal patients in Kolkata, India (48). Comparative genomics revealed a much more diverse set of GIs related to MGIVchHai6 in the genome of marine dwelling species and integrated at the 3= end of trmE and yicC or at the 5= end of dusA (Fig. 5). The gene cargo of these GIs is predominantly associated with DNA modification (methylation, phosphorothioation) and restriction systems, as well as antiphage systems such as BREX (Fig. 5A). Only MGIVchHai6 and GIPmi1, together with IDH᎑03944 (48), contained integrons carrying antibiotic resistance genes, suggesting that these elements are an ancient, large reservoir of antiphage systems, recently hijacked as vectors for drug resistance genes. Most GIs of the MGIVchHai6 and MGIVchHai6s clades share a gene encoding a predicted RhuM-like virulence factor, usually immediately upstream or downstream of mobI M (Fig. 5). A mutant of rhuM located in S. enterica SPI-3 pathogenicity island is deficient for epithelial cell invasion, neutrophil transmigration, and killing of its Caenorhabditis elegans host (49). While these results suggest that MGIVchHai6-like elements may be involved in virulence modulation, the molecular function of RhuM is not known and its potency in the Vibrionaceae has not been assessed. One striking feature of MGIVchHai6-like elements is the syntenic conservation of an operon-like region typically including a predicted AcaCD binding site followed by genes encoding a putative transcriptional regulator containing a helix-turn-helix (HTH) domain, a DUF927 domain-containing protein, and the xis gene (Fig. 5A). While the DUF927 gene is ubiquitous, GIs of the MGIVmi1 clade lack the upstream transcriptional regulator gene and the AcaCD binding site is located directly upstream of the DUF927 gene. Such high conservation is even more surprising since 85 and 86 are dispensable for mobilization of MGIVchHai6 by its helper plasmid (Fig. 1B). Why such factors should be under the control of the IncC transfer activator AcaCD is puzzling. Interestingly, a similar region exists in GIs mobilized by ICEs of the SXT/R391 family (Fig. S3). In prototypical MGIVflInd1, a binding site of the transfer activator SetCD lies upstream of rdfM and cds8, which encode a predicted transcriptional regulator and a DUF927 domain-encoding protein. While the role of cds8 remains elusive, RdfM shares weak amino acid identity (25%) with the product of 85 and acts as an RDF, facilitating excision of MGIVflInd1 upon setCD expression (50). In addition to the aforementioned factors, MGIVflInd1 and several MGIVchHai6-like GIs share an integration site at the 3= end of yicC (Fig. 5B), an observation that prompted us to hypothesize that Xis and 85/RdfM act as alternative RDFs, each allowing excision from a specific integration site. To test this, a ΔrdfM mutant of MGIVflInd1 was complemented by overexpressing 85 in a strain also containing ICEVflInd1 to provide setCD. However, 85 failed to restore excision and an attP or attB junction was not detected (data not shown), suggesting that 85 has a different function or is too divergent to promote excision of MGIVflInd1.
SXT/R391 ICEs have been shown to promote conjugative transfer of chromosomal DNA located 3= of prfC, their integration site, and remotely 5= of yicC, the integration site of the GIs they mobilize (41,51). Thus, these elements are able to mobilize large stretches (Ն1Mbp) of chromosomal DNA locally and remotely in an Hfr-like manner, suggesting that SXT/R391 ICEs play an evolutionary role that extends beyond their own dissemination. Our results show that IncC plasmids can mobilize chromosomal DNA located downstream of trmE by mobilization of MGIs without their prior excision from the chromosome (Fig. 3). Given the presence of MGIVchHai6-like elements integrated at different chromosomal sites and across a wide range of Vibrionaceae and other species of Gammaproteobacteria, IncC plasmids and their subordinate GIs can be concluded to comprise a potent driving force in the gene flow circulating in many bacterial pathogens. In fact, this is superbly exemplified by the recently reported presence of an MGIVchHai6-like GI in V. cholerae Santiago-089, a non-O1/non-O139 clinical isolate harboring many virulence genes scattered throughout chromosome 1 (52). Not only is the GI itself poised to be mobilized by an incoming IncC plasmid-along with the GI-borne antibiotic and mercury resistance genes-but it is also plausible that it may, in fact, usher transfer of downstream elements GIVch-T3SS and VPI-2, thus simultaneously contributing to dissemination of virulence determinants.
Cholera continues to cause epidemics that include millions of cases worldwide (53). The geographical range of V. cholerae is expected to expand dramatically as climate change renders the marine environment increasingly hospitable to this pathogen (54,55). While the ability to promote epidemic outbreaks was traditionally regarded as an appanage of O1 and O139 strains (56), it is becoming increasingly clear that the actual picture is far more nuanced (52,(57)(58)(59)(60). In various species, cumulative acquisition of antibiotic resistance and/or virulence determinants through exchange of genomic islands has time and again allowed emergence of virulent strains, some of which lack canonical virulence hallmarks (61)(62)(63). IncC plasmids circulating in non-O1/non-O139 V. cholerae populations may prove to comprise the perfect trigger for emergence of unforeseen pandemics.

MATERIALS AND METHODS
Bacterial strains and media. The bacterial strains and plasmids used in this study are described in Table 1. The strains were routinely grown in lysogeny broth (LB) (EMD) at 37°C in an orbital shaker and stored at Ϫ80°C in LB broth with 15% (vol/vol) glycerol. The following antibiotics and concentrations were employed: ampicillin (Ap), 50 g/ml; chloramphenicol (Cm), 20 g/ml; kanamycin (Kn), 10 g/ml for single-copy integrants of pOPlacZ-derived constructs, 50 g/ml otherwise; nalidixic acid (Nx), 40 g/ml; rifampin (Rf), 50 g/ml: spectinomycin (Sp), 50 g/ml; streptomycin (Sm), 200 g/ml; tetracycline (Tc), 12 g/ml. Conjugation assays were performed as previously described (17). However, donors and recipients were selected according to their sole chromosomal markers. When required, mating experiments were performed using LB plates with 0.02% arabinose to induce expression of pBAD30-derived  Table 2. PCR conditions were as follows: (i) 3 min at 94°C; (ii) 30 cycles, with 1 cycle consisting of 30 s at 94°C, 30 s at the appropriate annealing temperature, and 1 min/kb at 68°C; and (iii) 5 min at 68°C. When required, the resulting products were purified using the EZ-10 Spin Column PCR Products purification kit (Bio Basic) following the manufacturer's instructions. E. coli strains were transformed by electroporation as described previously (64) in a Bio-Rad Gene Pulser Xcell device set at 25 F, 200 V, and 1.8 kV using 1-mm gap electroporation cuvettes.
Plasmid and strain construction. The plasmids and primers used in this study are listed in Tables  1 and 2, respectively. Detailed description of plasmid and strain construction is provided in Text S1 in the supplemental material.
Detection of MGIVchHai6 excision. Excision of MGIVchHai6 was detected by PCR on genomic DNA of the appropriate strains using primers listed in Table 2. The attR and attP sites were respectively amplified using primer pairs 43_attPF/Ec104D.rev and 43_attPF/Hai6_attPR. ␤-Galactosidase assays. Qualitative assays were performed by depositing 10-l aliquots of overnight cultures with appropriate antibiotics on solid agar supplemented with 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-gal) with or without 0.02% arabinose. The plates were observed after an overnight incubation at 37°C.
Quantitative assays were performed with 2-nitrophenyl-␤-D-galactopyranoside (ONPG) according to a protocol adapted from Miller (65). After an overnight incubation at 37°C with appropriate antibiotics, cultures were diluted 1:100 in 50 ml LB broth supplemented with 50 g/ml ampicillin and grown until an optical density at 600 nm (OD 600 ) of 0.2 was reached. Two series of 1/10 dilutions were then prepared in total volumes of 5 ml LB broth supplemented with 50 g/ml ampicillin with or without 0.2% arabinose and incubated for 2 h at 37°C.
Phylogenetic analyses. The primary sequence of homologs of MobI proteins encoded by MGIVch-Hai6 and MGIVmi1 were obtained using the NCBI blastp algorithm (66) against the nr/nt database restricted to Gammaproteobacteria (taxid: 1236). Primary sequences sharing less than 45% identity and under 85% minimum coverage were filtered out of subsequent analyses. Distant MobI homologs from SXT (GenBank accession no. EET25017.1), pVCR94 (GenBank WP_001447712.1), pAhD4-1 (GenBank ALZ82609.1), pAsa4 (GenBank ABO92354.1), and pAQU1 (GenBank WP_014386842.1) were introduced manually in the data set as an outgroup. MobI homologs were first clustered with CD-HIT (67) to the best cluster that met the 0.95 identity cutoff prior to alignment. The predicted primary sequences of Int homologs were recovered from a representative sample of GIs encoding MobI homologs. Int primary sequences of SXT and SGI1 (GenBank accession no. AF261825.2) were introduced manually. Phylogenetic analyses were computed using amino acid alignments generated by MUSCLE (68). Prior to phylogenetic analysis, poorly aligned regions were discarded using trimAl v1.3 software with the automated heuristic approach (69). Evolutionary analyses were performed within MEGA X (v 10.0.5) (70) using the maximum likelihood method (PhyML) (71) and either the JTT (MobI) or LG (Int) matrix-based models (72,73). The initial tree(s) for the heuristic search was obtained automatically by applying neighbor-joining and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model and then selecting the topology with superior log likelihood value. Red recombination thermo-inducible encoding plasmid (ts, Ap) 81 pMS1 Red recombination thermo-inducible encoding plasmid (ts, Gm) 10 pKD3 cat (Cm) template for one-step chromosomal gene inactivation 75 pKD4 aph (Kn) template for one-step chromosomal gene inactivation 75 pCP20 Flp recombinase thermo-inducible encoding plasmid (ts, Ap Cm) 82 pCP20-Gm Gm r derivative of pCP20 (ts, Gm Cm) 83 Genomic