Electrostatic interactions between the CTX phage minor coat protein and the bacterial host receptor TolA drive the pathogenic conversion of Vibrio cholerae

Vibrio cholerae is a natural inhabitant of aquatic environments and converts to a pathogen upon infection by a filamentous phage, CTXΦ, that transmits the cholera toxin-encoding genes. This toxigenic conversion of V. cholerae has evident implication in both genome plasticity and epidemic risk, but the early stages of the infection have not been thoroughly studied. CTXΦ transit across the bacterial periplasm requires binding between the minor coat protein named pIII and a bacterial inner-membrane receptor, TolA, which is part of the conserved Tol-Pal molecular motor. To gain insight into the TolA–pIII complex, we developed a bacterial two-hybrid approach, named Oxi-BTH, suited for studying the interactions between disulfide bond-folded proteins in the bacterial cytoplasm of an Escherichia coli reporter strain. We found that two of the four disulfide bonds of pIII are required for its interaction with TolA. By combining Oxi-BTH assays, NMR, and genetic studies, we also demonstrate that two intermolecular salt bridges between TolA and pIII provide the driving forces of the complex interaction. Moreover, we show that TolA residue Arg-325 involved in one of the two salt bridges is critical for proper functioning of the Tol-Pal system. Our results imply that to prevent host evasion, CTXΦ uses an infection strategy that targets a highly conserved protein of Gram-negative bacteria essential for the fitness of V. cholerae in its natural environment.

Phage/bacterium interaction is one of the driven forces for gene acquisition and bacterial host adaptation to their environment, and it has been frequently associated with increased virulence of the bacterial host. A striking example of this parasitism-dependent adaptation is Vibrio cholerae, a bacterial natural inhabitant of estuaries, and the causative agent of epidemic disease cholera. Although there are more than 200 O-antigen serogroups described, only two have been reported to cause the pandemic disease cholera, the O1 and O139 serotypes, due to the production of two essential virulence factors, the toxin coregulated pilus (TCP) 3 and the cholera toxin (CT). Interestingly, the genes ctxAB encoding the enterotoxin CT are not carried by the core genome of the bacterium but can be acquired after infection by a lysogenic bacteriophage known as CTX⌽ (1). Once infected, the bacterium produces CT and assembles new phage particles (carrying the ctxAB genes) that will be secreted in the environment, and it may convert nonpathogenic V. cholerae cells to pathogenicity.
Most of the knowledge on CTX⌽ infection have been extrapolated from the canonical model of Escherichia coli F-pilusspecific coliphages Ff (including f1⌽, fd⌽, and M13⌽). CTX⌽ and Ff⌽ both belong to the genus Inovirus, which are filamentous particles containing a circular single-stranded DNA genome. The genome of inoviruses includes about 10 genes and is generally organized in a conserved modular structure in which functionally related genes are grouped together (2,3). Ff⌽ and CTX⌽ binding and uptake into the host cell rely primarily on the minor coat protein pIII located at the distal tip of the phage and present at three to five copies. Although there is no sequence conservation between pIII Ff and pIII CTX , both proteins are composed of three distinct domains separated by two low-complexity regions that serve as linkers. Although the N-terminal (N1) and the central domains (N2) are exposed at the capsid surface, the C-terminal domain (N3) anchors the pIII protein to the phage particle through hydrophobic interactions (4 -6).
Filamentous phage infection of the bacterial host is seen as a sequential two-step process. First, phage recruitment occurs upon specific binding between the phage capsid pIII-N2 domain and a primary receptor exposed at the surface of the cell host, the conjugative F pilus for E. coli (3,7) and the TCP for V. cholerae (1,5). In E. coli, ATP-dependent retraction of the F pilus brings the phage in contact with the cell envelope, promoting its transport across the outer membrane (OM) by an unknown mechanism. Then, pIII must partially unfold to separate N1 and N2 domains (8). This event is crucial in the infec-tion process as it unmasks the pIII-N1 domain for subsequent binding to a second receptor, the TolA Ec protein located in the cell envelope (8 -12). In E. coli, it has been proposed that the pIII-N1 Ff /TolAIII Ec interaction triggers conformational modifications permitting the pIII-N3 domain to form a pore in the bacterial IM, allowing the subsequent phage DNA injection into the cell cytoplasm (13). The nature of the force driving the DNA out of the capsid remains unknown.
In V. cholerae, TCP retraction seems central to the phage infection process, as TCP production alone is not sufficient to allow CTX⌽ uptake (14). Although TCP parasitism facilitates the introduction of CTX⌽ into the host cell, subsequent phage binding to TolA Vc appears to be the limiting step of the infection process (5,6,15). Thus, Heilpern and Waldor (5) have shown that a chimeric fd phage displaying the pIII-N1 CTX domain fused to the pIII-N3 fd domain can successfully infect V. cholerae cells. This demonstrates that the pIII-N1 CTX domain displayed at the tip of the capsid is critical and sufficient to ensure host-specific recognition in a TCP-independent manner (5).
TolA is the central protein of the Tol-Pal cell envelope system, which is highly conserved in Gram-negative bacteria (16,17). In addition to TolA, the Tol-Pal complex is composed of two IM proteins, TolQ and TolR, of the OM lipoprotein Pal and of the periplasmic protein TolB. In several species, including E. coli and V. cholerae, two additional proteins complete the system, the periplasmic protein CpoB (previously YbgF) (18) and the cytoplasmic thioesterase YbgC (19). The Tol-Pal complex is suspected to function as a nanomachine, using the proton-motive force of the IM to generate movements and to transfer energy to OM proteins. Multiple interactions connecting the different components of the Tol-Pal system have been identified. TolA, TolQ, and TolR form a complex anchored in the IM. The TolA protein extends in the periplasm thanks to a long ␣2-helix (TolAII domain), whereas its globular C-terminal domain (TolAIII) interacts with TolB, and with Pal in the presence of proton-motive force (20 -22). Moreover, Pal interacts with TolB and with the peptidoglycan (23). Thus, the Tol-Pal system links the IM, the OM, and the peptidoglycan. The system is involved in maintaining OM integrity, conferring pleiotropic phenotypes when one of its genes is mutated: increased sensitivity to detergents, cell filamentation in low and high osmolarity media, and outer membrane hypervesiculation (6,16). In addition, the Tol system is involved in the late stage of cell division corresponding to the OM constriction (25) and has been found associated with the PBP1B-LpoB complex in E. coli (18). It is also required for proper localization of polar factor in Caulobacter crescentus (26) and of chemoreceptors in E. coli (27).
For both coliphage and CTX⌽, structural studies on isolated protein domains have provided new insights into the complex formed with TolAIII in the bacterial periplasm (Fig. 1). First, the structures demonstrate that although the CTX⌽ pIII-N1 and M13⌽ pIII-N1 domains have only 15% sequence identity, they are both dominantly composed of ␤-strands, and multiple disulfide bonds stabilize their structures. On the bacterial side, TolAIII Ec and TolAIII Vc are curved structures mixing ␣-helices and ␤-sheets. A high-resolution structure of E. coli TolAIII free in solution has been obtained by heteronuclear NMR (Protein Data Bank PDB code 1S62) (12), whereas the TolAIII in complex with coliphage pIII-N1 M13 (residues 11-86) has been obtained by X-ray crystallography (PDB code 1Tol). The structure shows that the pIII-N1 M13 domain binds the concave side of TolAIII Ec , forming a continuous interprotein ␤-sheet (8). In 2012, Ford et al. (15) determined the crystal structure of pIII-N1 CTX domain alone and in complex with the V. cholerae TolAIII domain (PDB code 4G7X). Surprisingly, the authors showed that pIII-N1 CTX binds on the convex face of TolAIII Vc , resulting in a continuous interprotein ␤-sheet (Fig. 1A). Thus, interaction between the two partners delineates a distinct interface compared with the coliphage model of infection (15).
Filamentous phages are not the only particles that parasitize the Tol-Pal system to penetrate E. coli cells. Colicins are bacterial toxins comprising various types of lethal activity targeting the IM, the RNA, or the peptidoglycan of its bacterial target. Tol-dependent colicins have been shown to interact with one or more of the Tol proteins during their translocation across the periplasm, showing some similarities with Tol-dependent filamentous phage uptake. In 2012, Li et al. (28) demonstrated that a colicin A peptide (residues 53-107) binds on the convex face of TolAIII Ec , forming an intermolecular antiparallel ␤-sheet.
It is puzzling to observe that despite the structural similarities between V. cholerae and E. coli TolAIII domains, the molecular binding interfaces with colicin A, pIII CTX , and pIII M13 differ, illustrating the versatile functioning of TolA as a periplasmic hub protein. In this study, our goal was to investigate the determinants allowing CTX⌽-specific host selection and periplasm transit in vivo thanks to a new oxidative bacterial two-hybrid approach combined to NMR and in vivo studies.

Development of an oxidative bacterial two-hybrid approach dedicated to disulfide bond-folded protein analysis
To gain insights into the mechanism of CTX⌽ transit through the periplasm, we first analyzed the interaction that occurs between pIII-N1 CTX and TolAIII Vc , compared with the pIII-N1 M13 and TolAIII Ec interaction, using a bacterial twohybrid (BACTH) approach. This system relies on the reconstitution of the signaling cAMP transduction cascade in an endogenous adenylate cyclase-deficient strain (29). The TolAIII domains from V. cholerae and from E. coli were fused to the T18 domain in the pUT18 vector, whereas the pIII-N1 domains from M13 and CTX phages were fused to the T25 domain in the pKT25 vector. Constructs were introduced into the E. coli BTH101 strain, and co-transformants were tested on Mac-Conkey plates. As a control, we first observed that in this assay, the T18-TolAIII Ec construct gives a positive interaction signal with the colicin A N-terminal domain (ColA N ) as described previously (10,22,30), which validated our approach (Fig. 1B,  left panel). We also observed that a T18-TolAIII Vc construct is unable to bind T25-ColA N , attesting the specificity for partner recognition between the two bacterial species. Then, we tested Oxi-BTH study of the pIII CTX -TolA-binding determinants the TolAIII Vc and pIII-N1 CTX constructs together, but we did not detect interaction between these different domains. We obtained the same negative result when we tried to detect the TolAIII Ec /pIII-N1 M13 interaction (Fig. 1B, left panel). We hypothesized that this result could arise from improper folding of disulfide-bonded TolAIII, pIII-N1 M13 , or pIII-N1 CTX domains when expressed in the cell cytoplasm (Fig. 2, A  and B).
Thus, we envisioned that a bacterial two-hybrid assay in an oxidative environment would allow the proper folding of proteins with disulfide bonds. Several E. coli strains, such as Origami (Novagen) or SHuffle (New England Biolabs), have been engineered to optimize the purification of proteins with disulfide bonds and are commercially available. We chose the SHuffle T7 strain as a chassis because it is deleted for glutaredoxin reductase (gor) and thioredoxin reductase (trxB) genes, allowing disulfide bond formation in the cytoplasm. In addition, this strain expresses a cytoplasmic version of the disulfide bond isomerase DsbC, promoting correct disulfide bond formation and proper oxidative folding of proteins containing multiple cysteines (31). We transduced the cya mutation in the SHuffle strain to make the Oxi-BTH strain (for Oxidative Bacterial Two-Hybrid). The resulting strain was co-transformed with the pUT18 and pKT25 vectors expressing domains of interest. As shown in Fig. 1B (right panel), the Oxi-BTH strain allowed the detection of interaction between TolAIII Vc and pIII-N1 CTX , as well as between TolAIII Ec and pIII-N1 M13 . Thus, we concluded that the Oxi-BTH strain is a powerful tool to apply the BACTH system to the study of disulfidebonded proteins. Indeed, our data suggest that in both E. coli and V. cholerae, disulfide bond-dependent folding of the TolAIII domain and/or the phage minor capsid domain is required to allow binding of the two partners. Conversely, TolAIII Ec is able to interact with the colicin A N-terminal domain, which does not contain cysteines, in both reducing and oxidizing conditions (Fig. 1B). Finally, we did not observe cross-interaction between the two species (i.e. TolAIII Vc /pIII-N1 M13 or TolAIII Ec /pIII-N1 CTX ) despite strong structural conservation between E. coli and V. cholerae TolAIII domains.

Role of disulfide bonds in TolAIII Vc /pIII-N1 CTX interaction
Because the TolAIII Vc /pIII-N1 CTX interaction is only seen in our oxidative two-hybrid assay, we hypothesized that one or more disulfide bonds might be essential for bacterial and phage domain recognition. To test this hypothesis, each disulfide bond (one in TolAIII Vc and four in pIII-N1 CTX , Fig. 2, A and B) was sequentially abolished by introducing substitutions of individual cysteine to serine in the BACTH constructs. The resulting mutants were then tested in the Oxi-BTH assay. As shown in Fig. 2C, TolAIII Vc (C292S) construct was still able to interact with pIII-N1 CTX . This suggests that disulfide bond formation in TolAIII Vc is not required for CTX phage binding. We then targeted each of the four disulfide bonds present in pIII-N1 CTX . None of the individual mutations we performed was able to totally abolish binding to TolAIII Vc . Interestingly, mutation of the second or the third S-S bond (mutations C47S and C75S, respectively) of the phage pIII-N1 domain resulted in a faint interaction signal, suggesting a weaker binding affinity between TolAIII Vc and these two pIII-N1 CTX mutants compared with the wild-type construct. In agreement with this observation, a pIII-N1 CTX (C47S/C75S) double mutant was not able to interact with TolAIII Vc . These observations are unlikely to result from stability defects of the cysteine variants compared with the native pIII-N1 protein. Indeed, inserting the same mutations on the His-tagged pIII-N1 CTX domain expressed from the pIN vector resulted in equivalent expression of the different constructs (supplemental Fig. S1). Moreover, as the pIII-N1 CTX /TolAIII interaction is not seen in the regular BACTH assay, it is more likely that pIII-N1 CTX folding via its 2nd and 3rd disulfide bonds is critical for TolAIII Vc binding.

CTX⌽ central domain pIII-N2 does not block phage pIII-N1 accessibility to TolAIII Vc
It has been shown that Ff coliphages require an activation step to become able to infect the host cell. Indeed, in the native conformation of the minor capsid protein pIII Ff , the N2 domain is tightly associated with N1, which buries the phage TolAIIIbinding site at the domain interface. Phage activation is processed upon binding of N2 to the primary receptor, the F pilus, which initiates partial unfolding, prolyl cis-to-trans isomerization in the hinge between N1 and N2 and domain disassembly, thereby exposing its binding site for the ultimate receptor TolA (32). It has been proposed that the isomerization sets a molecular timer to maintain the binding-active state long enough for the phage to interact with TolA. Conversely, Craig and coworkers (15) have suggested that the TolA-binding site on pIII-

Oxi-BTH study of the pIII CTX -TolA-binding determinants
N1 CTX is permanently accessible and does not require initial pilus-induced conformational change. We wondered whether a fusion to the two-hybrid T25 domain would allow us to test the influence of pIII-N2 CTX on the accessibility of pIII-N1 CTX in an Oxi-BTH interaction assay. Indeed, our data show that the T25-pIII-N1N2 CTX construct is able to bind T18 -TolAIII Vc , whereas the T25-pIII-N2 CTX domain alone is not (Fig. 2C). This result demonstrates that pIII-N1 CTX exposes a TolAIII Vcbinding motif that is not masked by the pIII-N2 domain.

Specific recognition between the TolAIII Vc and pIII-N1 CTX domains relies on two salt bridges
We aimed to identify the residues showing a predominant role in the specificity of binding. Reversible protein/protein interactions involve low energy ionic bonds, hydrogen bonds, and van der Waals interactions. Based on the co-crystal structure of TolAIII Vc and pIII-N1 CTX (PDB code 4G7X (15)), we focused our analysis on the intermolecular salt bridges. Residues Lys-324, Arg-325, and Lys-347 on TolAIII Vc , respectively, interact with Glu-37, Asp-39, and Glu-92 on pIII-N1 CTX (Fig.  3A). To analyze the importance of these three salt bridges in the complex formation, we abolished them sequentially by introducing single residue substitutions on TolAIII Vc . We confirmed that the different TolAIII Vc variants showed production level and stability equivalent to the wild-type T18-TolAIII construct by immunodetection (Fig. 3B), and we tested their capacity to interact with T25-pIII-N1 in a Oxi-BTH assay (Fig. 3C). The TolAIII(K324E) and TolAIII(R325D) mutants were, respectively, partially and totally impaired in their ability to interact with pIII-N1 CTX , whereas the TolAIII(K347E) mutant retained its ability to bind pIII-N1 CTX . Conversely, we introduced the mirror mutations on T25-pIII-N1 CTX to generate the E37Q/D39N and the E92K variants. We demonstrated that T25-pIII-N1(E37Q/D39N) was unable to bind T18-TolAIII Vc , although T25-pIII-N1(E92K) was. Together, these data suggested that the two intermolecular salt bridges engaged between Lys-324 and Arg-325 on TolAIII Vc and Glu-37 and Asp-39 on pIII-N1 CTX are crucial for pIII-N1-TolAIII Vc complex stabilization.
To confirm the role of the TolAIII charged patch (Lys-324 -Arg-325) in pIII-N1 CTX binding, we swapped the charged residues engaged into salt bridges between the two partners. As predicted, the T18-TolAIII(K324E/R325D) double mutant was not able to interact with the wild-type T25-pIII-N1 CTX construct but regained interaction with the mutated variant presenting oppositely charged residues T25-pIII-N1(E37K/D39R) (Fig. 3C). Of note, the restored interaction signal appeared weaker than the one observed for wild-type constructs, possibly reflecting that additional local steric effects operate at the binding interface.
To further establish that our data result from a specificity of interaction rather than a folding defect of the protein variants, we developed an NMR approach. We used the pIN3-ompA2 vector (33) to overproduce 15 N-labeled wild-type pIII-N1 CTX , pIII-N1 CTX (E37Q/D39N), and pIII-N1 CTX (E92K) variants fused to a His C-terminal tag. The 1 H-15 N HSQC spectra of the proteins (Fig. 4, A, C, and E) shows 15 N/ 1 H correlations for the

Oxi-BTH study of the pIII CTX -TolA-binding determinants
NH group of each amino acid of the proteins. The 15 N and 1 H resonances are associated with the nuclear chemical environment; thus, the 1 H-15 N HSQC spectrum represents the fingerprint of a protein. In the case of the native and variant pIII-N1 CTX , the 1 H-15 N HSQC spectra are well resolved. This indicates that each residue has a particular environment in agreement with the X-ray structure of the protein (15). The fact that the spectra of the variant proteins are similar to that of the native protein demonstrates the conservation of the folding. Moreover, superimposition 1 H-15 N HSQC in the presence and the absence of TolAIII Vc showed chemical shift perturbations for pIII-N1 CTX (Fig. 4B) and pIII-N1 CTX (E92K) variants (Fig.  4D), whereas no perturbation was seen with the pIII-N1 CTX (E37Q/D39N) variant (Fig. 4F). These observations demonstrate that pIII-N1 CTX and pIII-N1 CTX (E92K) could interact with the TolAIII Vc domain, modifying the chemical environment of the nucleus located at the interface, whereas the pIII-N1 CTX (E37Q/D39N) variant could not, which confirmed the Oxi-BTH results.

In V. cholerae, periplasmic expression of the CTX pIII-N1 domain leads to Tol phenotypes
Oxi-BTH assay and NMR studies allowed us to identify key residues on the CTX⌽ minor capsid protein that were involved in TolAIII binding. To confirm these data in vivo, we used a periplasmic expression assay in V. cholerae. In E. coli, it has been previously shown that production of the N-terminal domain of colicin A (34) or of the pIII-N1 domain of coliphage f1 (10) into the periplasm of WT cells perturbs the Tol-Pal system by sequestering the TolA protein, and consequently results in tol phenotypes. This approach is particularly suitable to study the translocation step of phage infection, as it does not require production and assembly of mutated phage particles, their TCP-dependent reception, and their OM transport.
We used the pIN3-ompA2 vector to express the His-tagged pIII-N1 CTX wild-type sequence (or variants) fused to a N-terminal ompA signal sequence to allow transport of the produced

Oxi-BTH study of the pIII CTX -TolA-binding determinants
proteins into the periplasm via the sec pathway. Correct production of the resulting proteins in V. cholerae cells was confirmed by immunodetection, whereas transport of the produced domains into the periplasm was attested by sodium azide inhibition of the sec pathway (supplemental Fig. S2). Cells producing the WT pIII-N1 phage domain targeted to the periplasm presented a 5-fold decrease in susceptibility to phage infection (Fig. 5A) and high sensitivity to deoxycholate (DOC) (Fig. 5B) and to SDS (Fig. 5C), which are characteristic tol phenotypes.
This suggests that the exogenous pIII-N1 CTX domain interacts with the endogenous TolA Vc protein in the V. cholerae cell envelope, competing with the other Tol proteins of the system and preventing adequate functioning that ensures membrane integrity and permits phage uptake. We also observed that exogenous production of pIII-N1 M13 in V. cholerae periplasm does not result in tol phenotypes, which demonstrates further that the M13⌽ capsid protein pIII does not bind V. cholerae TolA protein. In accordance with the Oxi-BTH results,

Oxi-BTH study of the pIII CTX -TolA-binding determinants
periplasmic expression of the pIII-N1 CTX (E37Q/D39N) construct resulted in phage infection rate, DOC sensitivity, and SDS sensitivity similar to the control cells carrying an empty vector, whereas periplasmic expression of the pIII-N1 CTX (E92K) variant showed the same phenotypes as those obtained with the WT pIII-N1 CTX construct. In accordance with the Oxi-BTH results, these in vivo data confirm that the pIII-N1 CTX negatively charged residues Glu-37 and Asp-39 are required for pIII-N1 interaction with the native TolA protein.

Evidence that TolA Arg-325 residue is essential for proper functioning of the Tol-Pal system in V. cholerae
We decided to investigate whether the TolA charged patch (Lys-324 -Arg-325) required for phage binding is important for the Tol-Pal system function in V. cholerae. Although multiple attempts to delete the V. cholerae full-length tolA gene by double homologous recombination were unsuccessful, we found that a TolA⌬(41-421) mutant, deleted for its periplasmic domains but retaining its IM domain, is viable. The resulting strain presents characteristic tol phenotypes (6,35), including resistance to CTX⌽ infection, sensitivity to SDS and to DOC, and finally growth defect in a low osmolarity medium (tryptone broth, 66 mM NaCl) compared with LB medium (Fig. 6, A-E).
We first performed a complementation assay of the TolA⌬(41-421) mutant using a pBAD expression vector. TolA has been reported to be a low abundance protein (400 -800 copies in E. coli cells (36)). In our assay, basal level of TolA production from the pBAD-TolA vector was detected even in the absence of induction (Fig. 6A), and it was sufficient to almost fully complement (80 -100% rescue) the V. cholerae tolA mutant for phage infection (Fig. 6B), for resistance to SDS and DOC (Fig. 6, C and D), and for hypo-osmotic growth in tryptone broth (Fig. 6E). These results also confirmed that the deletion had no polar effect on the other genes of the Tol-Pal operon.
We then tested V. cholerae strains carrying TolA variants where residues Lys-324, Arg-325, or Lys-347 were mutated for oppositely charged residues. Although the K324E variant behaved similarly to the wild-type pBAD-TolA construct, rescue was no longer observed for the TolA(R325D) variant, despite adequate production of the mutant protein (Fig. 6, A-E). The TolA(K324E/R325D) double mutant showed a slightly stronger phenotype than the individual mutants for membrane integrity assays (Fig. 6, C and D). Finally, the TolA(K347E) variant complemented the mutant strain similarly to the WT TolA construct. Together, these findings support the conclusion that the positively charged residue Arg-325 plays a dominant role in TolA function in V. cholerae, whereas Lys-324 and Lys-347 do not.

Conservation of the Lys-324 and Arg-325 residues in other TolA Vibrio species
The Tol-Pal system has a widespread distribution in Gramnegative bacteria. We first questioned TolA sequence conservation among 128 V. cholerae isolate genomes available on the NCBI database, and we found that the full-length protein is 100% identical (supplemental Fig. S3). We then broadened the study to the Vibrionaceae family, including 82 Vibrio species (supplemental Fig. S4) and 33 non-Vibrio species (six Aliivibrio, three Enterovibrio, five Grimontia, 13 Photobacterium, and six Salinovibrio (supplemental Fig. S5). PSIPRED analysis (37) of the TolAIII amino acid sequences suggests that the secondary structures are conserved among Vibrionaceae (data not shown). Alignments of TolAIII sequences from multiple Vibrio species indicates that Lys-324 is highly conserved (73/82

Oxi-BTH study of the pIII CTX -TolA-binding determinants
sequences) with only few variants where it is replaced by another positively charged residue, arginine (6/82). Three exceptions were found with Vibrio breoganii, Vibrio gazogenes, and Vibrio rhizosphaerae that carry an uncharged glutamine residue at position 324. Overall, the positive charge at position 324 appears to be conserved in other Vibrionaceaes tested (31/ 33), suggesting its importance in TolA function (supplemental Figs. S4 and S5). Conversely, the positively charged residue Arg-325 is less conserved in Vibrio species (51/82), frequently being replaced by the uncharged residue serine (25/82) or less frequently by alanine, threonine, or valine (6/82). Notably, the positively charged residue at position 325 was never observed in other Vibrionaceae species (0/33).
As we previously demonstrated that pIII-N1 CTX interaction with the V. cholerae TolA receptor is driven by two intermolecular salt bridges engaging TolA Lys-324 and Arg-325 residues, we expected orthologous TolAIII sequences carrying the charged patch (Lys-324 -Arg-325) to also be able to bind CTX pIII-N1. As shown in Fig. 7A, the TolAIII domain from Vibrio anguillarum, Vibrio tasmaniensis, and Vibrio alginolyticus carry the (KR) patch, whereas Vibrio harveyi has a (KS) motif and the Vibrionaceae Aliivibrio fischeri has a (KT) motif. The

Oxi-BTH study of the pIII CTX -TolA-binding determinants
TolAIII coding sequences from the different species were cloned into the pUT18 vector, and correct expression of the different constructs was assessed by immunodetection (Fig. 7B) before testing their interaction ability with T25-pIII-N1 CTX in the Oxi-BTH assay. As shown on Fig. 7C, T18-TolAIII from V. anguillarum, V. tasmaniensis, and V. alginolyticus showed positive interaction signal with T25-pIII-N1 CTX . In contrast, T18-TolAIII V. harveyi and T18-TolAIII V. fischeri , lacking the posi-tively charged patch (KR), were not able to bind pIII-N1 CTX . We analyzed the TolAIII sequences that were able to bind pIII-N1 to define a consensus sequence for partner recognition. We found that TolAIII tolerates multiple sequence variation in the ␤2-strand, likely because this region of the protein interacts with pIII-N1 ␤1-strand through backbone hydrogen bond interactions. Conversely, TolAIII ␣2-helix alignments allowed the definition of a consensus sequence (GD(S/T)R(L/V) CAA(T/A)KRA(V/I)AQ) surrounding the (KR) residues. We noticed that V. harveyi TolAIII carry the defined consensus sequence, apart from Arg-325. However, point mutation to generate T18-TolAIII V. harveyi S325R and reconstitute a positively charged patch (KR) was not sufficient to restore interaction with pIII-N1 CTX (Fig. 7C).

New in vivo protein-protein interaction assay dedicated to disulfide-oxidized proteins
For many proteins functioning in the periplasm, exposed at the cell surface, or secreted in the extracellular environment, stability and/or activity require the formation of disulfide bonds. For example, in V. cholerae, S-S bond proteins include the cholera toxin (38), the pilin subunits that assemble in different type IV pili (TCP, MshA, and PilA) operating into the various aspects of Vibrio ecology (39,40), the TolA protein in the Tol-Pal complex involved in envelope integrity and cell division (15), and sensors such as TcpP (41,42) and ToxR (43) operating signal transduction to regulate virulence pathways. Thus, identification and characterization of physiologically relevant interactions between these S-S bond proteins in vivo are crucial tasks for the understanding the molecular processes within the cell.
The original BACTH approach is an easy yet efficient technique that has become a common laboratory tool used to identify and dissect protein/protein interactions (29,44). As the interaction between candidates of interest is tested in the cell cytoplasm, studies were restricted so far to proteins in their reduced state. Recently, the development of new BACTH plasmids inserting a transmembrane segment downstream of the T25 and T18 fragments was shown to allow the detection of interactions occurring within the periplasmic space of the cell (45).
In this study, we extended the range of BACTH application to the study of disulfide-oxidized proteins directly in the reporter cell cytoplasm thanks to a new genetic background that we named Oxi-BTH. In this strain, correct disulfide reshuffling is catalyzed by a cytoplasmic version of the disulfide bond isomerase DsbC (31). The Oxi-BTH assay was validated by the visualization of interactions previously described between proteins with unique (TolAIII Ec and TolAIII Vc ) and multiple disulfide bonds (pIII-N1 M13 and pIII-N1 CTX ), which could not be observed in the regular BACTH assay. As a protein with eight cysteines, pIII-N1 CTX has a probability of less than 1% to form the correct four disulfide bonds, and our results also attest the efficiency of the cytoplasmic DsbC proofreading activity (Fig.  1). Finally, this tool was suitable to individually test the importance of each disulfide bond and of targeted residues in protein/

Oxi-BTH study of the pIII CTX -TolA-binding determinants
protein interactions (Figs. 2 and 3). Many BACTH pUT18 and pKT25 constructs have already been published, and genomewide libraries are available for E. coli (46) and Pseudomonas aeruginosa (47) allowing the constitution of a large collection of potential partners that can be directly tested in the Oxi-BTH assay without requiring new cloning steps. In this context, we believe that Oxi-BTH constitutes a robust and versatile tool to test the importance of oxidative folding in protein/protein interaction, to identify key amino acids involved, and to investigate the specificity of binding between two proteins of interest.

Refining the model of CTX⌽ uptake
Filamentous phage infection is a two-step process, requiring a primary receptor at the surface of the cell (the TCP pilus in the case of CTX⌽) and a secondary receptor, TolA, inside the cell envelope. In this study, our aim was to gain insight into the second step of the infection process by investigating in vivo the phage/TolA interaction. In contrast to E. coli, we did not manage to obtain the V. cholerae TolA clean deletion mutant by homologous recombination, although we were able to delete the periplasmic domains of the protein. It is noteworthy that the only other V. cholerae TolA Ϫ mutant published in the literature is a tolA::pGP704 disruption mutant (obtained by single crossover) that, according to the primer design, also resulted in a truncated TolA protein somewhere in the periplasmic domain (6). This suggests that, under laboratory conditions, TolA is essential in V. cholerae, as reported previously in E. coli O7:K1 (48), P. aeruginosa (49), and C. crescentus (26), but is dispensable in the E. coli K12 strain (35).
We observed that pIII-N1 CTX domain is sufficient for TolA binding and that pIII-N2 CTX does not impede the interaction between the two partners (Fig. 2). This result is in agreement with previous observations made by Ford et al. (15). Indeed, the authors previously showed that CTX⌽ incubation with an excess of purified TolAIII Vc domain before infection of V. cholerae O395 cells reduced the infection efficiency. However, in this study, binding between pIII-N1-N2 CTX and TolAIII Vc had not been formally tested because pIII-N1-N2 CTX could not be purified in a soluble form (15). Together, our data further demonstrate that the CTX⌽ naturally exposes a TolA-binding site on the N-terminal domain of pIII, which is reminiscent to IF1 and IKe coliphage infection strategy (50). It implies that CTX⌽ binding to the primary receptor TCP might be responsible for phage recruitment from the environment and possibly active pulling to the cell surface through retraction (14), but it is not required to unmask the pIII-N1 TolA-binding site.
The crystal structure of the complex (PDB code 4G7X) suggested that multiple interactions (ionic, hydrophobic, polar contacts, and van der Waals forces) contribute to the binding interface between TolAIII Vc and pIII-N1 CTX . The TolAIII Vc ␤2-strand folds around pIII-N1 CTX , resulting in a continuous intermolecular ␤-sheet that involves multiple hydrogen bonds between the backbone chains of TolAIII ␤2-strand and pIII-N1 ␤1-strand. TolAIII Vc residues Lys-324, Arg-325, and Lys-347 form three salt bridges with pIII-N1 Glu-37, Asp-39, and Glu-92 residues, respectively (Fig. 3). However, the contribution of each of the individual interactions in the TolAIII-pIII CTX complex formation cannot be assessed from the crystal structure. By combining in silico analysis and in vivo experimental data, we defined that the consensus sequence for CTX binding on TolAIII Vc is restricted to the ␣2-helix, whereas the ␤2-strand tolerates multiple residues variations (Fig. 7). Moreover, our work clearly demonstrates that the two salt bridges engaged between TolAIII Vc residues Lys-324 and Arg-325, and CTX phage pIII-N1 residues Glu-37 and Asp-39, respectively, provide the driving forces of the interaction, forming ion locks required for the complex to form, whereas the TolAIII Lys-347-pIII-N1 Glu-92 bond is dispensable. It is noteworthy that mutating the TolAIII Arg-325 residue alone had a stronger negative impact on phage binding than mutating the TolAIII Lys-324 residue (Fig. 3). We used the 2P2I database (http:// 2p2idb.cnrs-mrs.fr/2p2i_inspector.html) 4 (51,52) to gain insights into the protein-protein interface parameters. Results showed a total of 88 non-bonded contacts (distance cutoff 4 Å) between pIII-N1 CTX and TolAIII Vc (supplemental Fig. S4). Among these 88 contacts, TolAIII Arg-325 was engaged in 20 different contacts with pIII-N1 surface, whereas TolAIII Lys-324 was engaged in six (supplemental Fig. S6). These observations might explain the more predominant role of the TolAIII Arg-325 residue in pIII-N1 binding, compared with Lys-324. Overall, this is reminiscent of what was previously observed for M13 coliphage pIII-N1 binding on the concave side of TolAIII Ec , where the interaction makes an antiparallel continuous ␤-sheet stabilized by two salt bridges (TolAIII Ec Asp-210, Lys-212 interacting with pIII-N1 M13 Arg-29 and Asp-28, respectively (PDB code 1Tol (8)). It would be interesting to test the importance of each of these ionic interactions for coliphage-TolA Ec complex formation.
TolA is involved in the fitness of V. cholerae, in particular in low osmolarity conditions, such as those found when the bacterium reaches low salinity environments (fresh water), or in response to detergent components, such as the bile acid deoxycholate during its transit through the gastrointestinal track (Figs. 5 and 6) (6). We found that the TolA(R325D) variant was functionally unable to complement the V. cholerae tol mutant, whereas the TolA(K324E) mutant was able to do so (Fig. 6). We questioned the sequence conservation of TolAIII in other Vibrio species, but we did not observe strict correlation between conservation of the Lys-324 and Arg-325 residues and their importance in TolA function in the tested conditions. Interestingly, we noticed that in the 33 Vibrionaceae species studied, the TolAIII Arg-325 residue is absent (supplemental Fig. S3), whereas additional positive charges are found at more or less one pitch of the ␣2 helix: Arg-321 (occurrence 33/33) and Lys-329 (occurrence 24/33). Because TolA is a hub protein, and part of a multicomponent complex, the patch of positive residues on the ␣2 helix might be involved in partners binding, either with an already known partner of the TolAIII domain such as TolB or Pal (16,22,53) or an unknown partner that still has to be identified.
We demonstrated that the KR patch on TolA is crucial but not sufficient for CTX binding (Fig. 7). Indeed, study of the V. harveyi TolA(S325R) variant suggests that additional local steric effects operate at the binding interface, outside the ␣2 helix. Although salt bridges provide the forces to drive proteinprotein initial attraction, subsequent complex stabilization is usually the result of cooperative interactions that encompass multiple pairwise bonds (54). Our results allow a better understanding of the TolAIII/pIII-N1 interaction interface, with two salt bridges providing the driving forces of the interaction, whereas the formation of the intermolecular anti-continuous ␤-sheet provides stabilization interactions to the complex. By targeting important functional and/or conserved residues of TolA protein, CTX⌽ uses a parasite infection strategy that may help to prevent V. cholerae from escaping the infection.

Implication for the CTX infection host specificity
In this work, we observed that the CTX phage protein pIII-N1 responsible for host selection (5) can bind the TolAIII receptor domain from at least three other Vibrio species: V. alginolyticus, V. anguillarum, and V. tasmaniensis (Fig. 7). However, the host range reported for CTX from environmental sampling studies is surprisingly narrow, even in species carrying a conserved TolA sequence. Thus, although all the epidemic V. cholerae strains belonging to the O1 or O139 serogroups carry the CTX prophage, it is rarely detected (2-5%) in non-O1/non-O139 V. cholerae environmental isolates genomes (55)(56)(57)(58)(59). Apart from V. cholerae, CTX⌽ has been reported to infrequently infect other Vibrio species, even for very close species such as Vibrio mimicus (60). It has been proposed that the limited distribution of the primary receptor TCP among Vibrio species plays a predominant role in the fate of the entire infection process (55). It is interesting to note that TCP-independent CTX⌽ infection has been reported to occur in defined conditions (5). Thus, whereas TolA is a conserved protein, variation in the sequence of the TolAIII domain might serve as a second level of CTX⌽ host-specificity checkpoint.
Because bacteriophages, like any other viruses, are obligate intracellular parasites, successful uptake across the bacterial cell envelope is an essential condition to complete their life cycle. However, current knowledge on host/phage interactions is based on a limited number of microbial models. This scientific problem outreaches the question of V. cholerae pathogenic conversion, as other filamentous phages (sharing similarities with CTX⌽) have been demonstrated to affect the virulence and fitness of a large range of bacterial hosts: meningococcal disease-associated (MDA) phage in invasive isolates of Neisseria meningitidis (61); Ypf⌽ in the plague bacillus Yersinia pestis (62); and CUS-1 in the high-virulence clone E. coli O18:K1:H7 (63). In this context, our work emphasizes the necessity to study CTX⌽ infection on its own and to overtake the Ff coliphage model of infection. It also sheds light on the mechanism underlying the initial filamentous phage-bacterial host binding and provides basic knowledge that might serve the understanding of transduction-dependent spreading of virulence factors in bacterial populations.

Bacterial strains and growth conditions
Relevant bacterial strains and plasmids used in this study are listed in supplemental Table S1. Bacteria were routinely cultivated in Luria-Bertani broth (LB, 407 mosM) at 37°C (E. coli) or 30°C (V. cholerae). When specified, MgCl 2 (2 mM) was added to the culture medium to promote OM integrity and cell growth of tol mutants. Tryptone broth (1% (w/v) tryptone, 66 mM NaCl, pH 8.5) was used as a low osmolarity medium (123 mosM (64)). When indicated, antibiotics were added to the medium at the following concentrations: streptomycin (100 g/ml); ampicillin (50 or 100 g/ml); kanamycin (50 g/ml); chloramphenicol (30 g/ml for E. coli and 1 g/ml for V. cholerae). V. cholerae tolA ⌬(41-421) in-frame deletion mutant was constructed as described previously using the primers listed in supplemental Table S2 and the suicide plasmid pWM91 (65,66). The absence of the TolA protein in the mutant strain was confirmed by Western blotting using polyclonal antibodies directed against E. coli TolAII-III protein (30) and cross-reacting with TolAIII Vc .

Plasmid construction
Polymerase chain reactions (PCR) were performed using Q5 High Fidelity DNA polymerase (New England Biolabs). Primer sets required to generate genetic constructs were synthesized by Sigma (supplemental Table S2). Enzymes (New England Biolabs) were used according to the manufacturer's instructions. Plasmids have been constructed either by standard restriction/ ligation protocol, by Sequence and Ligase Independent Cloning (SLIC) (67) as modified by Jeong et al. (68), or by restrictionfree cloning as described previously (69). Briefly for restrictionfree cloning, genes of interest were amplified with oligonucleotides introducing extensions annealing to the target vector. The double-stranded product of the first PCR has then been used as oligonucleotides for a second PCR using the target vector as template and Pfu Turbo polymerase (Stratagene, La Jolla, CA). For BACTH plasmid constructs, the V. cholerae El Tor N16961 genome was used as a template to PCR-amplify the pIII CTX -encoding gene (orfU, at loci vc1460). The tolA sequence was amplified from V. cholerae O395 (locus VC0395_A1430) or from E. coli W3110 (locus BL257_ RS03625) genomes. The pG3 vector (10) was used as a template to amplify pIII M13 . Amplified products were cloned into a pUT18c or a pKT25 expression vector (29) to generate fusions with the adenylate cyclase T18 and T25 domains. For construction of the pBAD-TolA rescue plasmid, the native sequence of V. cholerae O395 TolA, retaining the start and the stop codons, was amplified by PCR. The forward primer was designed to introduce a Shine-Dalgarno consensus sequence GAAGGAGATATACATACCC directly upstream of the start codon. The amplification product was then introduced into the pBAD18-Kan vector (70). For periplasmic expression, pIN-pIII-N1 CTX plasmid was constructed by PCR-amplifying the pIII-N1 CTX sequence (without start codon) with an upstream oligonucleotide encoding Strep-tag II (WSHPQFEK) and a downstream oligonucleotide encoding a C-terminal His 6 sequence. Digestion of the PCR product and the pIN-ompA2 Oxi-BTH study of the pIII CTX -TolA-binding determinants vector (33) with EcoRI and BamHI enzymes allowed subsequent ligation of the PCR product into the vector. The pIN-PIII-N1 M13 plasmid (previously named pG3) has been described previously (10).

Construction of the Oxi-BTH strain
To conduct bacterial two-hybrid experiments in an oxidative environment, the E. coli K12 SHuffle T7 strain (New England Biolabs) was engineered further. This initial strain is deleted for glutaredoxin reductase (gor) and thioredoxin reductase (trxB) genes, which allows disulfide bond formation in the cytoplasm. Moreover, cytoplasmic expression of the disulfide bond isomerase DsbC acts on proteins with multiple disulfide bonds, promoting correct disulfide bond formation and proper folding. In this genetic background, the cya°mutation was transduced using a P1 lysate of an E. coli K12 cya°strain. The resulting mutant strain, named Oxi-BTH, was unable to ferment sugars, and consequently it grew as white colonies when plated on MacConkey plates.

Bacterial two-hybrid assay in E. coli BTH101 and Oxi-BTH strains
The adenylate cyclase-based bacterial two-hybrid technique was used as published previously (29), with the following modifications. Pairs of proteins to be tested were fused to the isolated T18 and T25 catalytic domains of the Bordetella adenylate cyclase. After transformation of the two plasmids producing the fusion proteins into the reporter BTH101 or Oxi-BTH strains, plates were incubated at 37°C for 24 h. Three colonies for each transformation were inoculated into 600 l of LB medium supplemented with ampicillin, kanamycin, and IPTG (0.5 mM). After overnight growth at 30°C, 5 l of each culture were dropped onto MacConkey plates supplemented with ampicillin, kanamycin, and IPTG (0.5 mM). Plates were incubated for 16 -24 h (BTH101) and 2-3 days (Oxi-BTH) at 30°C. The experiments were done at least in triplicate, and a representative assay is shown.

SDS-PAGE and immunoblotting
Protein samples resuspended in 2ϫ loading buffer were subjected to SDS-PAGE. When specified, 2-mercaptoethanol (5% final) was added to the samples. For detection by immunostaining, proteins were transferred onto nitrocellulose membranes, and immunoblots were probed with primary antibodies and goat secondary antibodies coupled to alkaline phosphatase and developed in alkaline buffer in the presence of 5-bromo-4chloro-3-indolyl phosphate and nitro blue tetrazolium. The anti-TolAIII Ec polyclonal antibodies are from our laboratory collection, and the anti-5 His monoclonal antibody (Qiagen), the anti-CyaA monoclonal antibody (3D1, Santa Cruz Biotechnology), and alkaline phosphatase-conjugated goat anti-rabbit and anti-mouse antibodies (Millipore) have been purchased as indicated.

V. cholerae phenotypic analysis
Sensitivity test to SDS-V. cholerae cells harboring the empty plasmid as a control or the plasmid encoding the constructs of interest were grown in LB medium at 30°C until stationary phase, then back-diluted to initial A 595 nm ϭ 0.2 in LB supplemented or not with 0.125% SDS, and grown for 7 h at 30°C with agitation. The percentage of surviving cells was estimated from the turbidity ratio of the SDS-treated cells and the control samples. Experiments were performed in triplicates.
Sensitivity test to deoxycholate-Normalized serial dilutions of strains to be tested were spotted onto 1% DOC-supplemented LB plates. After overnight incubation at 37°C, survival was reported as the highest dilution of strain able to form colonies.
Growth in low osmolarity conditions-The different strains were grown in LB medium at 30°C until stationary phase, then back-diluted 100-fold in LB medium (osmolarity 407 mosM) and in tryptone broth (1% (w/v) tryptone, 66 mM NaCl, pH 8.5, 123 mosM (64)), and incubated for 16 h at 30°C. The percentage of growth was estimated from the turbidity ratio of the tested strains and the control sample. Experiments were performed in triplicates.

CTX-cm phage particle preparation
The dif1 Ϫ strain BS2 (71) harboring the chloramphenicolmarked CTX El Tor phage replicative form (pCTX-cm) was used as a donor strain to produce CTX phage particles. The donor strain was streaked onto LB-Cm (1 g/ml) plates and incubated at 37°C overnight. A single colony was used to inoculate a 2-ml LB ϩ Sm 100 culture and incubated at 37°C for 16 h. Cell were pelleted by centrifugation, and the supernatant containing phages was filter-sterilized using a 0.22-m syringe filter. Phage preparation was checked for sterility by plating on an LB plate.

Susceptibility to CTX phage infection assays
Cell susceptibility to phage infection was conducted as described previously (72). Briefly, for each recipient strain three independent clones were cultivated separately in TCP-inducing conditions (LB 1% tryptone, 0.5% yeast extract, 0.5% NaCl, pH 6.5, 30°C). After 20 h of growth, 75 l of cells were mixed with 75 l of CTX-cm phage suspension. The mixture was incubated during 30 min at room temperature without shaking, then 500 l of LB was added, and the cell suspension was incubated at 37°C with vigorous shaking for 45 min to allow cell recovery. Then dilutions of the cell suspension were plated on LB agar plates supplemented with Sm (100 g/ml) or with Sm (50 g/ml), and Cm (1 g/ml) to enumerate total cells and transductant cells, respectively. The frequency of infection was determined by dividing the number of transductants by the number of total recipient cells.

NMR spectroscopy
For NMR studies, a TolA(237-356) Vc (TolAIII Vc ) construct was included in a plasmid with the gene of OmpA signal sequence for protein secretion at the N terminus and His 6 tag Oxi-BTH study of the pIII CTX -TolA-binding determinants for the purification at the C terminus. For the native and variant pIII-N1 CTX , the OmpA signal sequence and His 6 tag construct were used, with an additional strep tag at the N terminus of the gene sequence. 15 N isotopic labeling of native and mutant pIII-N1 CTX was obtained from bacteria grown on M9 medium containing 1 g/liter 15 NH 4 Cl as sole source of nitrogen. The proteins were overexpressed in E. coli BL21 strain. Protein production was obtained after 2 h of IPTG induction at 37°C. Protein purification was obtained from periplasmic extract pulled on nickel-nitrilotriacetic acid-agarose and eluted with imidazole step gradient.
NMR spectra were recorded on a Bruker 600 MHz spectrometer equipped with a TCI cryoprobe at 300 K. 1 H-15 N HSQC spectra were processed with TopSpin software. For NMR experiments pIII-N1 CTX wild type and (E37Q/D39N) double mutant, the concentration was 0.16 mM in 50 mM NaPO 4 , 50 mM NaCl buffer at pH 6.9. In the case of the complexes, the final TolAIII Vc concentration was 0.32 mM.

In silico analysis
Search for V. cholerae O395 TolA orthologous sequences was performed with BlastP and restricted to Vibrionaceae species. Multisequence alignments were performed using Clustal Omega (73) and color-coded with JalView2 tool (24). The 2P2I database (51,52) was used to list all the non-bonded contacts at the complex interface (PDB code 4G7X) with a cutoff distance of 4 Å.