Crystal Structures of a CTXφ pIII Domain Unbound and in Complex with a Vibrio cholerae TolA Domain Reveal Novel Interaction Interfaces*

Background: CTXφ infection of Vibrio cholerae confers toxigenicity. Results: We report crystal structures of CTXφ-pIII domain N1 alone and bound to V. cholerae TolA domain C. Conclusion: CTXφ and coliphage pIII use distinct mechanisms to bind to TolA. Significance: These structures contribute to our understanding of a critical step in the evolution of pandemic V. cholerae with implications for emerging V. cholerae pathogens. Vibrio cholerae colonize the small intestine where they secrete cholera toxin, an ADP-ribosylating enzyme that is responsible for the voluminous diarrhea characteristic of cholera disease. The genes encoding cholera toxin are located on the genome of the filamentous bacteriophage, CTXφ, that integrates as a prophage into the V. cholerae chromosome. CTXφ infection of V. cholerae requires the toxin-coregulated pilus and the periplasmic protein TolA. This infection process parallels that of Escherichia coli infection by the Ff family of filamentous coliphage. Here we demonstrate a direct interaction between the N-terminal domain of the CTXφ minor coat protein pIII (pIII-N1) and the C-terminal domain of TolA (TolA-C) and present x-ray crystal structures of pIII-N1 alone and in complex with TolA-C. The structures of CTXφ pIII-N1 and V. cholerae TolA-C are similar to coliphage pIII-N1 and E. coli TolA-C, respectively, yet these proteins bind via a distinct interface that in E. coli TolA corresponds to a colicin binding site. Our data suggest that the TolA binding site on pIII-N1 of CTXφ is accessible in the native pIII protein. This contrasts with the Ff family phage, where the TolA binding site on pIII is blocked and requires a pilus-induced unfolding event to become exposed. We propose that CTXφ pIII accesses the periplasmic TolA through retraction of toxin-coregulated pilus, which brings the phage through the outer membrane pilus secretin channel. These data help to explain the process by which CTXφ converts a harmless marine microbe into a deadly human pathogen.

Vibrio cholerae colonize the small intestine where they secrete cholera toxin, an ADP-ribosylating enzyme that is responsible for the voluminous diarrhea characteristic of cholera disease. The genes encoding cholera toxin are located on the genome of the filamentous bacteriophage, CTX, that integrates as a prophage into the V. cholerae chromosome. CTX infection of V. cholerae requires the toxin-coregulated pilus and the periplasmic protein TolA. This infection process parallels that of Escherichia coli infection by the Ff family of filamentous coliphage. Here we demonstrate a direct interaction between the N-terminal domain of the CTX minor coat protein pIII (pIII-N1) and the C-terminal domain of TolA (TolA-C) and present x-ray crystal structures of pIII-N1 alone and in complex with TolA-C. The structures of CTX pIII-N1 and V. cholerae TolA-C are similar to coliphage pIII-N1 and E. coli TolA-C, respectively, yet these proteins bind via a distinct interface that in E. coli TolA corresponds to a colicin binding site. Our data suggest that the TolA binding site on pIII-N1 of CTX is accessible in the native pIII protein. This contrasts with the Ff family phage, where the TolA binding site on pIII is blocked and requires a pilus-induced unfolding event to become exposed. We propose that CTX pIII accesses the periplasmic TolA through retraction of toxin-coregulated pilus, which brings the phage through the outer membrane pilus secretin channel. These data help to explain the process by which CTX converts a harmless marine microbe into a deadly human pathogen.
Vibrio cholerae are Gram-negative, rod shaped bacteria that cause the gastrointestinal disease cholera. There are Ͼ200 known V. cholerae serogroups, yet only two, O1 and O139, cause pandemic disease. Pathogenic serogroups are distinguished from non-pathogenic strains by the acquisition of two mobile genetic elements: the Vibrio pathogenicity island (1) and the CTX element, which is a prophage of the filamentous bacteriophage CTX (2,3). The Vibrio pathogenicity island contains the tcp operon encoding the toxin-coregulated pilus (TCP), 4 which is necessary for V. cholerae colonization of the human intestine (4,5), and TcpF, a soluble colonization factor of unknown function (6,7). The CTX element contains repetitive sequences involved in site-specific recombination of the CTX genome and a core region encoding the phage assembly proteins and the cholera toxin subunits A and B. Expression of tcp and the cholera toxin genes ctxA and ctxB are regulated by the same transcriptional activator, ToxT (8,9).
CTX has a 7-kb single-stranded circular DNA genome that integrates into the V. cholerae chromosome as a prophage. Phage particles are produced from extrachromosomal DNA replicated as a plasmid (10). The CTX⌽ genome is similar in gene size and organization to Ff family filamentous phage that infect Escherichia coli (coliphage), including M13 and fd (3). The core region of the CTX prophage contains genes cep, orfU, ace, zot, ctxA, and ctxB. Although not homologous in sequence, the cep, orfU, ace, and zot genes are similar in size and synteny to the M13 genes VIII, III, VI, and I, which encode the structural proteins pVIII, pIII, pVI, and pI, respectively (3,11). In M13, the major coat protein pVIII forms the long, cylindrical phage coat that packages the phage genome. pIII, pVI, and pI are minor coat proteins located at the phage tips (12)(13)(14). Although pVI and pI are very small, pIII is a large protein, present in four-five copies at one end of the phage particle. pIII (also called g3p for gene 3 protein) mediates phage binding, uptake, and assembly (15)(16)(17)(18)(19).
pIII is well characterized in Ff phage, in part because of its application in phage display technology (20,21). fd and M13 pIII amino acid sequences are almost identical, with an 18amino acid signal peptide and a 406-amino acid mature protein organized into three distinct functional domains, N1, N2, and C (also called D1, D2, and D3 or CT), linked by glycine-rich segments of low structural complexity, LCR1 and LCR2 (19,22,23)). The hydrophobic C-terminal segment is required for insertion of the virion into the inner membrane and for excision after phage assembly (15,24). CTX pIII has very little sequence homology to the Ff pIII proteins but is predicted to have a 14-amino acid signal peptide and 3 domains separated by serine/proline-rich linkers (25), and CTX pIII-C contains a very hydrophobic segment that likely represents an inner membrane anchor.
Ff coliphage bind to E. coli using their minor coat protein, pIII, in a two-step process to initiate infection. First, the central N2 domain of pIII binds to the F pilus tip (26 -28), which spontaneously retracts (29 -31) to bring the pIII N-terminal domain, pIII-N1, into contact with the C-terminal domain of the periplasmic protein, TolA (32)(33)(34)(35). The TolA binding site on pIII-N1 is buried by pIII-N2 (36), and the interaction with the F pilus is required not only for bringing the phage to the bacterial surface but for inducing a conformational change in pIII that exposes the TolA binding site (18,22,23). The mechanisms by which the F pilus retracts and the bacteriophage gains entry into the periplasm to bind to TolA are not understood. The F pilus is required for efficient infection, but Ff phage can infect E. coli lacking the F pilus, albeit much lower levels than for Fϩ strains (18,27). In contrast, E. coli lacking TolA are resistant to phage infection (18,35).
The CTX phage utilize a similar mechanism to infect V. cholerae, as both TCP and TolA are required for maximal infectivity (3,11). Heilpern and Waldor (25) generated recombinant fd phage in which CTX pIII domains N1, N2, or N1N2 were fused to the N terminus of fd pIII or to deletion mutants lacking N1N2 and assessed the ability of these hybrid phage to infect V. cholerae. fd hybrid phage displaying CTX pIII-N1N2 fused to domain D3 of fd pIII showed high levels of infectivity for V. cholerae. When only CTX pIII-N1 was present, infectivity was reduced but measurable, but when only CTX pIII-N2 was present, infectivity was undetectable, demonstrating that whereas N2 is important for efficient phage uptake, domain N1 is critical.
Crystal structures are available for the N-terminal domains of both M13 and fd pIII, which are 99% identical in amino acid sequence (22,23). These structures reveal two discrete domains of similar, predominantly ␤-sheet fold joined by a linker and by a crossover of the C-terminal strand of N2 onto N1. N1 has a short N-terminal ␣-helix followed by a four-stranded ␤-barrel motif, and N2 is dominated by a twisted ␤-sheet. Not surprisingly, the glycine-rich LCR1 is disordered and is not resolved in the crystal structure. However, rather than acting as a linker between N1 and N2, this segment lies within N1 and is followed by an ordered loop and a ␤-strand, ␤5, that are part of the N1 domain. A short linker connects N1 to N2. The two domains also interact via non-covalent contacts between their ␤-sheet loops and by the C-terminal strand of N2, which extends across to N1 as a ␤-strand, ␤13, to form a 2-stranded ␤-sheet with ␤5 that lies across the N1 ␤-barrel.
A crystal structure was also determined of a fusion protein consisting of the M13 pIII-N1 and LCR1 fused to the C-terminal domain of TolA (36). The N1 fold is similar in both the pIII-N1/TolA-C fusion protein structure and the N1N2 structure but terminates just C-terminal to the disordered LCR1 in the fusion protein and, hence, lacks strand ␤5 and the remaining N2 domain. TolA-C interacts with N1 on a face of the ␤-barrel that is occupied by the ␤5␤13 sheet and part of the N2 domain in the N1N2 structure. Thus, for TolA to bind to pIII, pIII must partially unfold, separating N1 and N2 and removing the ␤5␤13 strand to expose the TolA binding site. The trigger that induces this conformational change is the interaction of N2 with the F pilus (18,22,23). This movement is proposed to involve isomerization of the Gln-212-Pro-213 bond immediately after ␤13 (37).
Commonalities and differences between the pIII proteins of CTX and the Ff phage prompted us to investigate the interaction between CTX and V. cholerae at a molecular level. Here we show by transmission electron microscopy (TEM) that phage binding to V. cholerae does not require TolA, but phage uptake does. We demonstrate a direct interaction between pIII-N1 and TolA-C and describe crystal structures of pIII-N1 alone and in complex with TolA-C. Our data reveal similarities in structure but surprising differences between CTX and the coliphage in their interactions with TolA, advancing our understanding of CTX infection of V. cholerae.

EXPERIMENTAL PROCEDURES
Bacterial Strains-Bacterial strains, plasmids, and primers are listed in Table 1. E. coli strains were grown in Luria-Bertani (LB) broth at 37°C with appropriate antibiotics. V. cholerae strains were grown in LB, pH 6.5, Sm at 30°C on a Ferris wheel rotator (TCP-inducing conditions). Antibiotics were used at a final concentration of 200 g/ml streptomycin (Sm), 50 g/ml kanamycin (Km), 100 g/ml ampicillin (Ap), and 12 g/ml tetracycline (Tc). Anti-TcpA antibodies were a gift from Ronald Taylor (Geisel School of Medicine).
CTX Transduction Assays-V. cholerae strains O395, DH3, and TCP2 used in the infection assays were grown overnight in TCP-inducing conditions. To prepare CTX, V. cholerae CL101 cells were grown under the same conditions but with Km in addition to Sm. CL101 cells produce CTX-Km⌽ in which the ctxA gene is replaced with the Km resistance marker (3). To produce fd⌬1-pIII CTX  phage, E. coli DH5␣pir(fd⌬1-pIII CTX (15-274)) cells were grown at 37°C overnight with Tc, 1.2 ug/ml. CL101, and E. coli DH5␣-pir cells were removed from the supernatant containing the phage by centrifugation and filtration on a 0.2-m pore filter. Transduction assays were performed by mixing 75 l of phage with 75 l of V. cholerae overnight culture and shaking for 20 min at room temperature. Serial dilutions of the infection mixture were plated on LB-Sm/Km agar plates to enumerate the transductants and on LB-Sm plates to enumerate the input bacteria. Cells were grown overnight at 37°C, and colony-forming units (cfu) were counted. The phage transduction frequency was calculated as the ratio of transductants to input V. cholerae cells. For the transfection inhibition assay, the filtered cell supernatant containing CTX or fd⌬1-pIII CTX (15-274) (50 l) was mixed with 3 l of His-TolA-C (100 M) or with buffer only and incubated at room temperature for 30 min. Seventy-five microliters of V. cholerae O395 overnight culture was added to 75 l of the phage mixture and incubated for another 20 min. The infection mixtures were serially diluted, plated on LB-Sm/Km and LB-Sm agar plates, grown overnight at 37°C, then counted for cfu.
Immunogold and Transmission Electron Microscopy-Sample grids were prepared by floating glow discharged carboncoated Formvar copper grids on transfection solutions (described above) for 10 min and then fixing with 2% paraformaldehyde in phosphate-buffered saline (PBS), pH 7.4 for 5 min. The grids were then washed twice on drops of PBS con-taining 0.15% glycine, blocked for 5 min with 1% bovine serum albumin (BSA) in PBS, and then incubated with polyclonal rabbit anti-TcpA antibodies (1:100 dilution) for 30 min. After 2 washes on PBS-glycine and 1 on PBS-BSA, the grids were incubated with 10-or 15-nm gold-conjugated protein A (diluted 1:70) for 20 min. After two washes on PBS, grids were fixed with 1% glutaraldehyde in PBS for 5 min, washed twice on PBSglycine, and then blocked for 5 min with PBS-BSA. This was followed by an incubation step on anti-M13 mouse monoclonal antibodies (anti-g8p, Progen Biotechnik GmbH, 1:100 dilution) for 30 min. After 2 washes on PBS-glycine and 1 on PBS-BSA, the grids were incubated with anti-mouse-coupled gold particles (5 nm) for 20 min. After 3 rinses on PBS and 1 on water, the grids were stained for 2 min with 2% uranyl acetate and viewed with a JEOL 1200EX TEM.
Expression of CTX pIII Domains-The gene fragment encoding the CTX pIII N-terminal domain and part of the signal peptide (pIII-N1, residues Ϫ5 to ϩ96) was PCR-amplified from orfU in CTX-Km DNA using primers pIII-N1-for and pIII-N1-rev (Table 1). A gene fragment encoding the CTX pIII-soluble portion lacking the C-terminal putative transmembrane segment (pIII-⌬TM, residues 1-355) was PCR-amplified using primers pIII-⌬TM-for and pIII-⌬TM-rev. In both PCRs the forward primers add an Nde1 site and the reverse primers add a BamH1 site to the PCR products. PCR products and the pET⅐15b expression vector (Novagen) were separately digested with Nde1 and BamH1 (Fermentas), and the digestion products were ligated for 3 h at 20°C. Insertion of genes between the Nde1 and BamH1 sites results in N-terminal addition of a hexahistidine tag and linker with the sequence GSSHHHHHHSSGLVPR2GSHM, where the down arrow indicates a thrombin cleavage site. Ligation products were transformed into E. coli DH5␣ and Rosetta-gami B (DE3) cells at 42°C using the CaCl 2 heat shock method (38). Cells were incubated overnight at 37°C on LB-Ap agar plates, and transformants were selected and grown overnight in LB-Ap broth. Plasmid was extracted from the DH5␣ strain and purified using the Qiaprep Plasmid Purification kit. Plasmids were sequenced using the T7 promoter and terminator primers ( Table 1) to verify that the insertions were correct. The pET-15b⅐pIII-⌬TM construct contains a mutation encoding a Ser65Thr substitution. A new plasmid encoding pIII-N1ϩ (residues Ϫ5 to ϩ134) was generated by introducing a stop codon at the nucleotide encoding pIII Asn-135 in pET-15b⅐pIII-⌬TM using primers pIII-N1ϩ-for and pIII-N1ϩ-rev (Table 1). Additional constructs were designed by inserting stop codons at other sites in pIII-⌬TM. These include pIII-N1N2 (residues Ϫ5 to 233), pIII-N1N2C (Ϫ5 to 330), and pIII-N1a (1 to 96). Of these constructs, only pIII-N1, pIII-N1ϩ, and pIII-N1a produced soluble protein.
Expression of the V. cholerae TolA C-terminal Domain-A gene fragment encoding residues 241-356 of TolA (TolA-C) was PCR-amplified from the genomic DNA of V. cholerae O395 using primers TolA-C-for and TolA-C-rev (Table 1). PCR products and pET15b plasmid were digested with NdeI and BamHI and ligated for 3 h at 20°C then transformed into E. coli BL21(DE3) cells as described above.
Expression and Purification of pIII and TolA Proteins-Bacterial cells transformed with the appropriate plasmid were grown overnight in 200 ml of LB broth containing 100 g/ml Ap at 37°C shaking. 10-ml aliquots of overnight culture were added to 1 liter of LB-Ap and grown at 37°C to an optical density at 600 nm (A 600 ) of ϳ0.6. Protein expression was induced by the addition of isopropyl ␤-D-thiogalactopyranoside to 1 mM and cells were grown overnight at 19°C with shaking. Cells were pelleted by centrifugation at 4000 ϫ g for 30 min, the supernatant was discarded, and the pellet was resuspended in lysis buffer (Bis-Tris pH 6.5, 100 mM NaCl, 10% glycerol, 0.1% Tween, 5 mM imidazole) with 10 g/ml lysozyme and a Complete Protease Inhibitor Mixture tablet (EDTA-free, Roche Applied Science). Cells were incubated in lysis buffer at room temperature for 1 h and then lysed by sonication. Cellular debris was removed by centrifugation at 40,000 ϫ g for 40 min, and the supernatant was filtered through a 0.45-m membrane. The filtered supernatant was loaded onto a gravity column containing Ni-NTA beads (Qiagen) for affinity purification using the N-terminal His tag. The column was washed using 3 column volumes of wash buffer (20 mM Bis-Tris, pH 6.5, 100 mM NaCl), and protein was eluted with 200 mM imidazole. Elution fractions containing the protein of interest were concentrated using a stirred cell concentrator (Amicon) with a 3000-Da molecular mass cut-off membrane and loaded onto a size exclusion column (HiPrep 26/60 Sephacryl S-100 HR, GE Health-care). Fractions containing purified protein were concentrated to 15-30 mg/ml, flash-frozen in liquid nitrogen, and stored at Ϫ80°C. In cases where the His tag was removed, the purified protein was incubated at 4°C ON with ϳ1 unit of thrombin per mg protein, after which the protein was loaded onto a Ni-NTA column to remove undigested protein and free His tag. The flow-through and wash fractions from the Ni-NTA column were concentrated in an Amicon stirred cell concentrator and loaded on to a size-exclusion column to remove the thrombin.
Expression of Selenomethionine-substituted pIII-N1ϩ-Selenomethionine (SeMet)-substituted His-pIII-N1ϩ was prepared by growing Rosetta-gami B (DE3) E. coli cells carrying pET-15b⅐pIII-N1ϩ in M9 minimal media containing SeMet and select amino acids (39). Cells were first grown overnight in LB-Ap at 37°C shaking. 10 ml of culture for every 1 liter of minimal media was centrifuged at 3500 ϫ g for 20 min. The supernatant was discarded, and the cells were resuspended in 1 liter of M9 minimal medium and grown at 37°C to an A 600 of ϳ0.4, at which point 60 mg/liter L-SeMet, 100 mg/liter lysine, threonine, and phenylalanine, and 50 mg/liter leucine, isoleucine, and valine were added. The culture was incubated for 15 min at 37°C with shaking before isopropyl ␤-D-thiogalactopyranoside was added to 1 mM to induce protein expression. Cells were grown overnight at 19°C with shaking. Cells were pelleted by centrifugation, resuspended in lysis buffer, and lysed by sonication, and protein was purified and stored as described above.
pIII-N1⅐TolA-C Binding Assays-His-pIII-N1 or His-TolA-C (1 mg) was added to 250 l of Ni-NTA beads. The beads were washed twice with 1 ml of wash buffer (20 mM Bis-Tris, pH 6.5, 50 mM NaCl, 5 mM imidazole). Next, 1 mg of the His tag-cleaved binding partner, TolA-C or pIII-N1, was applied to the Ni-NTA beads and washed twice. The His-tagged proteins were eluted from the column with 1 ml of elution buffer (20 mM Bis-Tris, pH 6.5, 50 mM NaCl, 200 mM imidazole). Elution fractions were analyzed by SDS-PAGE.
Purification of the pIII-N1ϩ⅐His-TolA-C Complex for Crystal Structure Determination-Purified His-TolA (30 mg) was loaded onto a Ni-NTA column (2 ml Ni-NTA beads), and the column was washed with 5 ml of purification buffer (20 mM Bis-Tris, pH 6.5, 50 mM NaCl). Next, ϳ27 mg of His tag-cleaved pIII-N1ϩ was loaded onto the column and washed with 5 ml of purification buffer. Proteins were eluted from the column with elution buffer and the fractions were concentrated in the Amicon stirred cell concentrator with a 10,000 molecular mass cut off membrane and then loaded onto the Sephacryl size exclusion column. The fractions containing the complex were concentrated to 27.1 mg/ml and stored as described.
Crystallization of pIII-N1ϩ and the pIII-N1⅐TolA-C Complex-All crystals were grown using the hanging-drop vapor diffusion method at 20°C. Initial crystals were grown by mixing 2 l of reservoir buffer from commercial crystallization screens with 2 l of protein in purification buffer (N-terminal His-tagged SeMet-pIII-N1ϩ alone) or protein buffer (20 mM Bis-Tris, pH 6.5, 20 mM NaCl for the pIII-N1ϩ⅐His-TolA-C complex). Promising conditions were further optimized by varying salt and precipitant concentrations and pH. Optimized conditions were then treated with additive and detergent screens (Hamp-ton Research) by mixing 2 l of protein with 1.6 l of concentration-adjusted reservoir buffer and 0.4 l of additive or detergent. SeMet-pIII-N1ϩ crystals grew from a 23.5 mg/ml protein solution in 20% PEG 3350, 8% Tacsimate, pH 5.0, and Fos-Choline-9. pIII-N1ϩ⅐His-TolA-C crystals grew from a 27.1 mg/ml solution in 25% PEG 6000, 100 mM MES, pH 6.0. Crystals appeared after 1-2 weeks and were flash-frozen in liquid nitrogen. Glycerol (25%) was included as a cryoprotectant for pIII-N1ϩ.
X-ray Diffraction Data Collection, Structure Determination, and Refinement-X-ray diffraction data for SeMet-substituted pIII-N1ϩ and pIII-N1ϩ⅐His-TolA-C crystals were collected remotely on the Stanford Synchrotron Lightsource beamline 9-2 at 100 K using Blu-Ice software (40). The initial test diffraction images were indexed, and x-ray fluorescence scans were performed on SeMet-substituted pIII-N1ϩ to select the appropriate wavelengths for multiple wavelength anomalous dispersion data collection. For pIII-N1ϩ, data sets were collected at wavelengths corresponding to the inflection point and high energy remote from a crystal that diffracted to ϳ4 Å resolution, and a second data set was collected at the peak wavelength for a crystal diffracting to 2.9 Å resolution ( Table 2). A native 1.44 Å resolution data set was collected for the pIII-N1ϩ⅐His-TolA-C complex. The dual wavelength pIII-N1ϩ data set was pro-cessed and scaled using iMosflm and SCALA (41,42). The single wavelength pIII-N1ϩ data set and the native pIII-N1ϩ⅐His-TolA-C data set were processed and scaled using the XDS suite (43) to resolutions 2.9 and 1.44 Å, respectively. Matthews coefficients (44) calculated using CCP4 (41) indicated three molecules per asymmetric unit for pIII-N1ϩ and a single heterodimer for pIII-N1ϩ⅐His-TolA-C. Initial phases for pIII-N1ϩ were determined by multiple wavelength anomalous dispersion phasing method using SOLVE (45), and density modification was performed with RESOLVE (46) yielding an interpretable electron density map. Backbone atoms were placed and model building was carried out in COOT (47). A few cycles of rigid body refinement followed by tight NCS-restrained refinement by REFMAC (48) brought the R cryst and R free values to 0.417 and 0.447, respectively. Iterative cycles of tight main chain and loose side chain restrained refinement and model building improved the model and statistics. After several cycles of TLS refinement (48), water oxygens were located using COOT and included in the refinement. The final R cryst and R free are 23.4 and 26.2%, respectively. The crystal data and refinement statistics are summarized in Table 2.
The pIII-N1ϩ⅐His-TolA-C structure was solved by the molecular replacement method. First, structure factors were determined for pIII-N1ϩ using PHASER (49) with our mono- meric pIII-N1ϩ structure as a search model. Once the orientation and position of pIII-N1 were determined, structure factors were calculated for V. cholerae TolA-C within the complex using the E. coli TolA-C structure in 1TOL as a model. The two TolA-C domains have 27% amino acid sequence identity. A few cycles of the "map improvements by atoms update and refinement" option of ARP/wARP (50) made a dramatic improvement to the map with clear density for main chain and side chain atoms for both proteins, with the exception of their N-terminal segments. The model was built manually in COOT, and several cycles of restrained refinement were carried out. Water oxygens were located in COOT, and after checking them manually, WATERTIDY in CCP4 (41) was used to move the water oxygen coordinates to the symmetry-related position nearest to the host protein molecule. Additional cycles of refinement and fitting brought the R cryst to 20.6% and the R free to 21.6%, with no subsequent improvement. Both pIII-N1ϩ and pIII-N1ϩHis-TolA-C structures were validated using PROCHECK and MolProbidity (51,52).

RESULTS AND DISCUSSION
TolA Is Not Required for CTX Binding to V. cholerae but Is Required for Uptake-To investigate the role of V. cholerae TolA in CTX infection, phage were incubated with V. cholerae strain O395 and with an O395-derived ⌬tolA deletion mutant, DH3 (11), and visualized by immunogold labeling and TEM. V. cholerae produce relatively low levels of CTX that are difficult to detect by TEM. Thus, we used the engineered fd phage fd⌬1-pIII CTX , which expresses CTX pIII domains N1 and N2 in place of the native fd N1N2, causing a switch in specificity from E. coli to V. cholerae (11). Plasmidexpressed fd⌬1-pIII CTX  are produced in high levels from E. coli DH5␣-pir (53). fd⌬1-pIII CTX (15-274) phage were incubated with V. cholerae O395 and DH3 cells and applied to TEM grids. TCP were immunolabeled with 10-or 15-nm gold particles, and phage were labeled with 5-nm gold particles. Although the phage and TCP are both long thin filaments of comparable diameters (ϳ8 nm), they are readily distinguishable by TEM, as the pili are several microns in length and bundled, whereas the phage are typically shorter (200 -400 nm) single filaments (Fig. 1). Phage were observed in contact with TCP in both V. cholerae strains. In many cases the phage contact the pili via their tips (Fig. 1, A and B), but lateral interactions were also observed. In some cases the phage appear to interact directly with the V. cholerae surface (Fig. 1D). There were no apparent differences in the phage interactions with V. cholerae O395 (Fig. 1, A-C) and the tolA deletion strain DH3 (Fig. 1, D and E). These findings are consistent with CTX initially binding to V. cholerae via an interaction between pIII and TCP in a TolA-independent manner. However, TolA is required for the subsequent phage uptake step, as transduction frequencies of fd⌬1-pIII CTX  and CTX are negligible for the TolA-deficient strain (Fig. 1F and Ref. 11).
Interaction between CTX pIII-N1 and the C-terminal Domain of V. cholerae TolA-Because the N-terminal domain of coliphage pIII binds to the C-terminal domain of E. coli TolA (18, 32, 33, 35, 36), we sought to demonstrate a direct interaction between CTX pIII-N1 and V. cholerae TolA-C. Recom-binant pIII-N1 was prepared comprising residues 1-96 of the mature CTX pIII protein with an N-terminal hexahistidine (His) tag for purification using metal affinity chromatography ( Fig. 2A). Five residues of the signal peptide are also present in this construct, which was designed based on an earlier prediction of a signal peptidase cleavage site between the 14th and 15th residue of the pre-protein. However, current bioinformatics tools predict the cleavage site to lie between residues 19 and 20 ( Fig. 2A).
The C-terminal domain of V. cholerae TolA was delineated based on amino acid sequence comparison with E. coli TolA and secondary structure prediction using PsiPred (67). E. coli TolA is a 421-residue periplasmic protein with an N-terminal inner membrane anchor, a central ␣-helical segment with a Lys 1-2 -Ala 3-4 -(Glu/Asp) repeat sequence, and a C-terminal domain that binds to coliphage pIII-N1 as well as to colicins (18,33). V. cholerae TolA is a relatively uncharacterized 356-amino acid protein with a putative N-terminal transmembrane segment, a central Lys/Ala/Glu-rich region predicted to be an extended ␣-helix, and a C-terminal domain ( Fig. 2A). V. cholerae and E. coli TolA share 35% sequence identity in their N-terminal ϳ150 amino acids and 27% identity in their C-terminal domains. The C-terminal domain of TolA (TolA-C, residues 241-356) was expressed and purified from E. coli with an N-terminal His tag (His-TolA-C) and mixed with pIII-N1, which had its His tag removed by thrombin digestion. The pIII-N1⅐TolA-C complex was isolated by incubating the sample with Ni-NTA beads (Qiagen), washing the beads to remove unbound protein, and eluting bound protein using imidazole. Samples were analyzed by SDS-PAGE (Fig. 2B). pIII-N1 bound to the Ni-NTA-coated beads when His-TolA-C was present, but not in its absence, suggesting that these domains interact with each other. Similar results were obtained for reciprocal experiments in which His-pIII-N1 was mixed with non-tagged TolA-C (data not shown).
To further establish an interaction between pIII-N1 and TolA-C, the proteins were loaded onto a size exclusion chromatography column individually and after co-incubation. His-TolA-C, which has a calculated molecular mass of 14.8 kDa, elutes at ϳ15 kDa, and pIII-N1, with a calculated mass of 11.5, and elutes at ϳ12 kDa (Fig. 2C). When His-TolA-C is preincubated with pIII-N1, these peaks disappear, and a new large peak elutes from the column at a mass of ϳ28 kDa, corresponding to the pIII-N1:His-TolA-C complex, with a calculated mass of 26.3 kDa. The ability of pIII-N1 to remain bound to TolA-C through size exclusion chromatography demonstrates the stability of this interaction and indicated that this complex was a good candidate for x-ray structure determination.
Crystal Structure of CTX pIII-N1-We first solved the crystal structure of pIII-N1 alone. CTX pIII constructs of varying lengths were designed for expression as N-terminally Histagged proteins to identify crystallizable fragments or domains. E. coli expression strains, growth, and induction conditions were optimized for each construct. In addition to His-pIII-N1, described above, two other constructs produced soluble protein from E. coli Rosetta-gami B (DE3): pIII-N1ϩ encoding residues Ϫ5 to 134, which includes LCR1 plus 19 residues of N2 ( Figs. 2A and 3A), and pIII-N1a, encoding residues 1-96. Pro-teins were purified, crystallization conditions were screened, and diffraction-quality crystals were obtained for His-pIII-N1ϩ belonging to the P6 3 22 space group. Because pIII is not homologous in sequence to any protein of known structure, molecular replacement was not feasible for determining its structure. Thus, SeMet-substituted His-pIII-N1ϩ was prepared for structure determination using multiple wavelength anomalous dispersion methods. SeMet pIII-N1ϩ crystals were obtained in conditions similar to those of the native crystals, and a 2.9 Å resolution structure was solved by multiple wavelength anomalous dispersion phasing ( Table 2, Fig. 3, supplemental Fig.  S1A).
There are three molecules in the asymmetric unit of the pIII-N1ϩ crystal, all with close structural similarity (supplemental Fig. S2A). Residues 1-98 were resolved for chains A and C and 1-99 for chain B. No electron density was apparent for the N-terminal His tag or the five signal peptide residues at positions Ϫ5 to Ϫ1. Additionally, most of LCR1 (residues 100 -115) and the 19 residues of N2 are not resolved in the map. Thus, we refer to this structure from here on in as pIII-N1. The core of pIII-N1 is an incomplete ␤-barrel with side protrusions and irregular loops at both the N-and C-terminal ends (Fig. 3, B and  C). The U-shaped N-terminal loop that lies at the bottom of the ␤-barrel is followed by its first ␤-strand, ␤1 (residues [13][14][15][16][17]. From ␤1 the polypeptide chain crosses the top of the ␤-barrel to form ␤2 (26 -29) on its opposite side. A type II turn connects ␤2 with ␤3 (32-50), a long strand that is interrupted by a ␤-hairpin (37)(38)(39)(40)(41)(42)(43)(44) that protrudes from the ␤-barrel. ␤3a and ␤3b curve around the ␤-barrel at almost 90 o to each other. ␤3b is connected to another long strand, ␤4 (53-62), by a type I hairpin turn. The ␤3␤4 hairpin extends away from the top of the ␤-barrel. Following ␤4, the polypeptide chain exits the ␤-barrel and  (Fig. 3B). Four disulfide bonds stabilize CTX pIII-N1: between Cys-5 on the N-terminal loop and Cys-32 at the N terminus of ␤3a; between Cys-47 and Cys-56, joining strains ␤3b and ␤4; between Cys-75 at the top of the ␤-barrel on the meandering ␤4␤5 loop and Cys-85 on the short ␤5 strand; between Cys-90 and Cys-96 in the C-terminal loop. The conformations of the N-and C-terminal loops have small variations among the three molecules in the asymmetric unit (supplemental Fig. S2A). Additionally, the extended ␤3␤4 hairpin bends away from the ␤-barrel in chain C (Fig. 3, B and C) but is straighter in chains A and B, with a difference of ϳ7 Å between the Glu-51 C␣s at the hairpin tip (supplemental Fig. S2A). This difference suggests flexibility in this region, which is constrained by crystal contacts in chains A and B.
The crystal structure of CTX pIII-N1 was compared with its corresponding domain in pIII of fd, M13, and IF1 phage. fd and M13 pIII proteins have 99% amino acid sequence identity, and their crystal structures (domains N1 and N2 or N1N2) are essentially identical (22,23). pIII of the related coliphage IF1 pIII shares the same domain organization as the Ff phage pIII proteins and binds to TolA-C via N1 and to the E. coli I pilus via N2, but its domains are more loosely associated, and no conformational change is required for TolA binding (54). IF1 pIII-N1 is 31% identical to fd pIII-N1, and their structures are highly similar. Molecule B of fd pIII-N1N2 (PDB code 2G3P) is shown in Fig. 3E. Although CTX and the coliphage pIII proteins have similar domain boundaries, size, and functions, a pairwise BLAST search (blast.ncbi.nlm.nih.gov) revealed no significance sequence identity. In comparing the CTX pIII-N1 structure with that of fd, structural similarities are also not immediately apparent. However, close inspection reveals similar topologies for the two domains (Fig. 3, B-G), and sequence homology is evident when the CTX and fd N1 amino acid sequences are aligned based on their secondary structures (Fig.  3A). Like CTX, fd pIII-N1 has a core ␤-barrel composed of strands ␤1-␤4 with two similarly-placed disulfide bonds (Fig. 3, A, C, and F), but the ␤-barrel is more compact and lacks a fifth ␤-strand (for ease of comparison, we have not labeled a short ␤-strand between ␤1 and ␤2 of fd pIII-N1, which is absent in CTX N1). fd pIII-N1 has a short N-terminal ␣-helix not present in CTX pIII-N1, and its ␤3 and ␤3␤4 hairpins are shorter. The fifth strand of the CTX pIII-N1 ␤-barrel is located on a 26-residue insertion between ␤4 and LCR1 that is not present in fd pIII (Fig. 3A). This insertion in CTX pIII-N1 follows ␤4 as a meandering loop around the outside and across the top of the ␤-barrel, then feeds into ␤5, which is followed by a disulfide-bonded C-terminal loop. The meandering loop is disulfide-bonded to ␤5. The LCR1 segment that follows is disordered in both the CTX and coliphage pIII proteins. In the coliphage pIII the LCR1 is referred to as the N1N2 linker (22,23), yet it appears to be an integral if disordered part of N1 distinct from the linker connecting the two domains (Fig. 3). It is not clear whether CTX pIII-N1 possesses a two-stranded ␤-sheet corresponding to ␤5 and ␤13 in fd pIII, but the structural similarities between the N1 domains suggest that their N2 domains have similar orientations.
Crystal Structure of CTX pIII-N1 in Complex with V. cholerae TolA-CTX pIII-N1 (residues Ϫ5-96) was shown to form a stable complex with His-tagged TolA-C (residues 241-356, Fig. 2, B and C) as did a second pIII-N1 construct, pIII-N1ϩ (residues Ϫ5-139). Both pIII-N1⅐TolA-C complexes were isolated by size exclusion chromatography, concentrated, and used in crystallization trials. Crystals were obtained for the pIII-N1ϩ⅐His-TolA-C complex, referred to hereafter as pIII-N1⅐TolA-C. A crystal belonging to the p2 1 2 1 2 1 space group diffracted to 1.44 Å resolution, with one molecule in the asymmetric unit and a solvent content of 27.5% (Table 2). Structure factors were determined for pIII-N1 by molecular replacement using our unbound pIII-N1ϩ structure as a search model. After proper positioning of pIII-N1 in the unit cell, TolA-C structure factors were determined using the E. coli TolA-C structure, solved in a complex with M13 pIII-N1 (1TOL), as a search model (supplemental Fig. S1B).
pIII-N1 residues 3-94 are well-resolved in the pIII-N1⅐His-TolA-C crystal structure (Fig. 4A). The pIII-N1 structure in the bound complex is essentially identical to that of the unbound protein, superimposing on chain C with a root mean square deviation of 0.608 Å between main chain atoms spanning residues 11 to 85 (supplemental Fig. S2B). The high curvature of the ␤3␤4 hairpin seen for chain C of the unbound pIII-N1 is also seen in the bound structure. As with the unbound structure, the 5 residues from the signal peptide, LCR1, and the 19 amino acids from N2 are not resolved in the pIII-N1⅐His-TolA-C structure.
It is the convex face of TolA, mainly ␣2 and ␤2, that interacts with pIII-N1. pIII-N1 uses its open "hand" to grasp TolA. ␤-barrel strands ␤3b␤4␤1, which form the "palm," together with the ␤3 hairpin "thumb," the protruding ␤3␤4 hairpin, and the C-terminal loop "fingers" (Fig. 3B) clamp onto TolA-C, burying a total of 1489 Å 2 in surface area (Fig. 4). Two acidic side chains, Glu-37 and Asp-39, on the pIII ␤3 hairpin, form key salt bridges with adjacent lysines (Lys-324 and Arg-325, respectively) on ␣2 of TolA-C (Fig. 4B). The Arg-325 side chain on ␣2 of TolA inserts into the ␤3 hairpin and makes polar contacts with several residues, and the hydrocarbon portions of both Lys-324 and Arg-325 side chains form van der Waals interactions with Val-44 and Leu-59 of pIII-N1. Tyr-41 on the ␤3 hairpin makes hydrophobic contacts with Leu-282 on the 3 10 helix loop of TolA. Across the pIII-N1 ␤-barrel the ␤3␤4 hairpin hooks around TolA (Fig. 4A), forming mainly polar interactions with the extended loop on TolA that follows ␣2. Thus, the curvature of the ␤3␤4 hairpin, observed in this complex and in chain C of the unbound pIII-N1 structure (supplemental Fig.  S1), provides stereochemical complementarity for TolA-C binding. Main chain hydrogen bonds between ␤1 of pIII-N1 and ␤2 of TolA-C result in a continuous anti-parallel ␤-sheet comprising strands ␤1, ␤4, and ␤3b from pIII and ␤2, ␤1, and ␤3 of TolA-C (Fig. 4, A and C). The C-terminal loop of pIII-N1 also contacts TolA via a salt bridge between Glu-92 and Lys-347 on the short loop connecting ␣3 and ␤3 of TolA-C (Fig. 4C). The ␤3␤4 hairpin fingers protrudes toward the reader. Cysteines and secondary structural elements are labeled, and cysteines are shown in ball-and-stick representation with sulfur atoms in yellow. D, secondary structure representation of CTX pIII-N1 is shown. Disulfide bonds are shown as black bars. E, the crystal structure of fd phage pIII-N1N2 (2GP3) with N1 in pale cyan and N2 in yellow is shown in the same orientation as CTX pIII-N1 in C. The gray dashed line represents LCR1, which is unresolved in the crystal structure. F, fd pIII-N1 residues 2-67 from 2GP3 are shown. G, shown is a secondary structure representation of fd pIII-N1. ␤-Strands in fd pIII are numbered according to the corresponding strands in CTX pIII-N1. Thus, a short strand between ␤1 and ␤2 in fd pIII-N1 is not included.
Comparison of the CTX pIII-N1⅐V. cholerae TolA-C Complex with that of M13 pIII-N1⅐E. coli TolA-C-The crystal structure of the complex between CTX pIII-N1 and V. cholerae TolA-C was compared with that of a complex between M13 pIII-N1 and E. coli TolA-C (1TOL). The C-terminal domains of V. cholerae and E. coli TolA have 27% sequence identity (Fig. 4E), and their structures are very similar (1.81 Å root mean square deviation between C␣ carbons for V. cholerae TolA-C residues 263-356 and E. coli TolA-C residues 334 -421) (Fig. 4, A and D). The two TolA-C domains differ primarily in the loop regions; the conformations of the 3 10 -helix loops differ somewhat, the ␤1␤2 hairpin of V. cholerae TolA-C is lon-ger and more ordered than that of E. coli TolA-C, and the ␣2-␣3 loop of V. cholerae TolA-C is longer by two residues. As described above, N1 of the CTX and the Ff phage pIII proteins differ significantly in their ␤-sheet loops and in their N and C termini but have comparable topologies. Given the similarities between CTX and M13 pIII-N1 and between V. cholerae and E. coli TolA-C, it might be expected that the pIII-N1⅐TolA-C interactions are similar in both complexes, but this is not the case. Whereas ␤-strands of both pIII-N1 proteins form edge-on contacts with ␤-strands in TolA-C to form a continuous ␤-sheet, the pIII ␤-strands and the face of the pIII-N1 ␤-barrel involved in this interaction differ in the two complexes. Fur- thermore, CTX and M13 pIII-N1 bind to opposite sides of TolA-C (Fig. 4, A and D). Whereas CTX pIII-N1 binds to the convex face of TolA-C via its concave face, interacting with TolA-C all along the length of its ␤-barrel, M13 pIII-N1 binds to the concave face of TolA-C via the convex surface of its ␤-barrel composed of the ␤3␤4 hairpin and the C-terminal half of ␤1. The N-terminal end of M13 pIII-N1 ␤3 hydrogen-bonds with E. coli TolA-C ␤3 to form a continuous 6-stranded ␤-sheet between the two proteins, comprising strands ␤3, ␤4, and ␤1 of M13 pIII-N1 and ␤3, ␤1, and ␤2 of TolA-C. The total buried surface area for the M13 pIII-N1 complex is 1768 Å 2 , which is larger than that of the CTX pIII-N1⅐TolA-C complex (1489 Å 2 ), yet both interfaces are highly complementary.
An important distinction between the two complex structures is that pIII-N1 and TolA-C in the M13⅐E. coli system were synthesized as a single fusion protein with an 18-residue linker. This unnatural covalent linkage could potentially force an interaction between these proteins that is not biologically relevant. However, the crystal structure of a complex between phage IF1 pIII-N1 and E. coli TolA-C, obtained from proteins that were expressed separately (54), is highly similar to that of the M13 pIII-N1⅐TolA-C complex. Thus, it appears that CTX pIII and the coliphage indeed bind to distinct sites on TolA.
The function of TolA in V. cholerae and E. coli is not known with certainty. TolA forms an inner membrane complex with TolQ, TolR, and TolB, encoded by the tolQRAB gene cluster conserved in Gram-negative bacteria (55) and is involved in outer membrane integrity, energy transduction, and cell division (56 -58). E. coli and V. cholerae tolQRA mutants are viable but are sensitive to detergents and antibiotics, suggesting a role for these proteins in maintaining outer membrane integrity (11,59,60). E. coli tolQRAB mutants are resistant to filamentous phage infection as well as to colicins, bacteriocidal molecules that are produced by some E. coli strains and act upon other E. coli strains (61,62). Like the coliphage, colicins utilize a twostep mechanism for uptake, first binding to outer membrane receptors on E. coli, then interacting with TolA upon transfer into the periplasm. Very recently a crystal structure of E. coli TolA-C was solved in complex with a polypeptide from the N-terminal domain of colicin A, TA 53-107 (PDB code 3QDR (63)). Interestingly, TA  binds to the side of TolA opposite to that used by coliphage pIII-N1, which corresponds to the CTX pIII-N1 binding site on V. cholerae TolA (Fig. 5). The main interacting segment in TA 53-107 is ␤-strand ␤5, which associates with TolA-C primarily via edge-on backbone hydrogen bonds. Antiparallel strands ␤4 and ␤5 of TA 53-107 form a continuous ␤-sheet with the three-stranded TolA-C ␤-sheet (Fig. 5A) and correspond to ␤1 and ␤4 in CTX pIII-N1 (Fig. 5B). TA  ␤5 also makes van der Waals contacts with TolA ␣2. The total buried surface area for this interaction (1303 Å 2 ) is similar to that of the pIII-TolA interaction (1489 Å 2 ). Several of the TolA residues involved in these interactions are in equivalent positions, but they are not conserved between E. coli and V. cholerae, and thus, colicin A is unlikely to bind to V. cholerae TolA.
Accessibility of the TolA Binding Site on Native CTX pIII-Because the TolA binding site on CTX pIII-N1 lies outside of the region that interacts with domain N2 in the fd and M13 coliphage, we wondered if it might be accessible to TolA bind-ing in its native form. In Fig. 6A the CTX pIII-N1⅐TolA-C complex and the fd pIII-N1N2 structure are presented with both N1 domains in approximately the same orientation. This shows that if N2 is located in the same position relative to N1 in CTX pIII as it is in fd pIII, it would not block TolA binding. To test the accessibility of the TolA binding site on CTX pIII-N1, phage were incubated with an excess of TolA-C before infection of V. cholerae O395 cells to see if TolA-C binding reduced infection levels. The transduction efficiency was reduced 3-fold for CTX and 5-fold for fd⌬1-pIII CTX (15-274) when pretreated with TolA-C (Fig. 6B), suggesting that TolA-C can bind to pIII-N1 in its native conformation. Thus, although TCP binding is important for bringing CTX to the V. cholerae surface, it may not be required to unfold pIII and expose the TolA binding site. This relationship is similar to that seen for the IF1 and IKe coliphage, where the N1 and N2 domains are not tightly apposed but instead are arranged like beads-on-a-string, and thus their TolA binding sites are accessible in the native pIII (54,64). The TolA binding domains of CTX, M13, fd, IF1, and IKe phage are all structurally similar regardless of whether their respective pilus binding domains are tightly associated or not. It is possible that CTX pIII also adopts a beads-on-a-string arrangement for N1 and N2 rather than the close packing seen in fd and M13 phage. We were unable to express CTX N1N2 in a soluble form to examine this possibility. However, the unique location of the TolA binding site on CTX pIII-N1 and the ability of free TolA-C to reduce phage infectivity together suggest that the TolA binding site is accessible in the native protein.
Despite differences between CTX and the coliphage with respect to their mode of binding to TolA, strong parallels between the two systems support a common two-step infection process whereby phage bind first to the bacterial pili via pIII-N2 and pilus retraction draws the phage tip into the periplasm and presents N1 to TolA (Fig. 7). Unlike the E. coli F pilus, TCP has not been shown to retract and lacks a "retraction" ATPase present in retractile Type IV pili. However, the TCP role in TcpF secretion presumably entails a piston-like motion involving cycles of pilus extension and retraction to extrude TcpF across the outer membrane. Such a mechanism has been proposed for the closely related type II secretion system (65), which also lacks a retraction ATPase. TCP retraction may occur randomly and spontaneously in the absence of a retraction ATPase (Fig. 7A), with CTX being drawn in serendipitously to the benefit of both the phage and V. cholerae but to the detriment of the human host.
Conclusions-We demonstrate here direct interactions between CTX and V. cholerae TCP and between pIII-N1 and TolA and present x-ray crystal structures of pIII-N1 alone and in complex with TolA-C. Despite the structural similarities between the CTX and coliphage pIII-N1 domains and between V. cholerae and E. coli TolA-C, the molecular interfaces are entirely different between the two pIII-N1⅐TolA-C complexes. The unique interaction between CTX pIII-N1 and V. cholerae TolA-C may abrogate the need for a pilus-induced conformational change in pIII to prime it for binding to TolA. CTX infection may involve spontaneous retraction of TCP to transport the phage tip through the pilus secretin channel. An emerging understanding of the Type IV pilus machinery and that of the related type II secretion system provide a structural framework to investigate the CTX infection process, with implications for understanding coliphage uptake in E. coli. These highly specific and efficient processes could potentially be exploited as delivery systems for antibiotics.  In the case of the Ff phage, this interaction induces a conformational change that exposes the TolA binding site of N1. B, pili retract, drawing the phage tip through the pilus secretin channel into the periplasm where the N1 domain of pIII interacts with the C-terminal domain of TolA. A different face of TolA is accessible to pIII-N1 in E. coli and V. cholerae, perhaps due to its orientation to the outer membrane secretin, which may also dictate colicin binding to TolA-C in E. coli (A).