Deciphering Bacteriophage T5 Host Recognition Mechanism and Infection Trigger

ABSTRACT

Bacteriophages, viruses infecting bacteria, recognize their host with high specificity, binding to either saccharide motifs or proteins of the cell wall of their host. In the majority of bacteriophages, this host recognition is performed by receptor binding proteins (RBPs) located at the extremity of a tail. Interaction between the RBPs and the host is the trigger for bacteriophage infection, but the molecular details of the mechanisms are unknown for most bacteriophages. Here, we present the electron cryomicroscopy (cryo-EM) structure of bacteriophage T5 RBP_pb5 in complex with its Escherichia coli receptor, the iron ferrichrome transporter FhuA. Monomeric RBP_pb5 is located at the extremity of T5’s long flexible tail, and its irreversible binding to FhuA commits T5 to infection. Analysis of the structure of RBP_pb5 within the complex, comparison with its AlphaFold2-predicted structure, and its fit into a previously determined map of the T5 tail tip in complex with FhuA allow us to propose a mechanism of transmission of the RBP_pb5 receptor binding to the straight fiber, initiating the cascade of events that commits T5 to DNA ejection.

IMPORTANCE Tailed bacteriophages specifically recognize their bacterial host by interaction of their receptor binding protein(s) (RBPs) with saccharides and/or proteins located at the surface of their prey. This crucial interaction commits the virus to infection, but the molecular details of this mechanism are unknown for the majority of bacteriophages. We determined the structure of bacteriophage T5 RBP_pb5 in complex with its E. coli receptor, FhuA, by cryo-EM. This first structure of an RBP bound to its protein receptor allowed us to propose a mechanism of transmission of host recognition to the rest of the phage, ultimately opening the capsid and perforating the cell wall and, thus, allowing safe channeling of the DNA into the host cytoplasm.

INTRODUCTION

Bacteriophages, bacterial viruses also called phages, are the most abundant and diversified biological entity on Earth (1). As bacterial killers, but also as a means of genetic material exchange, they shape and drive microbial population evolution and diversity (2). The vast majority of phages indexed in databases are composed of a capsid protecting the DNA and a tail, which can be long and contractile (Myoviridae), long and flexible (Siphoviridae), or short (Podoviridae). The tail extremity serves to recognize the host surface and eventually perforate its cell wall and safely channel the DNA from the capsid to the bacterial cytoplasm (3, 4). The distal tip of the tail bears receptor binding proteins (RBPs) that reversibly or irreversibly bind saccharides and/or proteins. Depending on their morphology, RBPs can form fibers (often more than 1,000 residues long and often requiring chaperones to fold) or more globular RBPs, often called tail spike proteins (TSPs) (5). Fibers and TSPs of known structure share an elaborately interwoven, often trimeric state, rich in β-helices. TSPs are multidomain proteins, containing at least a phage binding domain and a receptor binding domain (6). The receptor binding domain can also bear enzymatic activity, digesting the saccharide receptors (5). Because they recognize the bacterial host surface with high specificity and because of their enzymatic activities, phage RBPs can be used to identify bacterial strains for diagnosis (7), to engineer phages with new host ranges (8), or in biotechnology (9). The stoichiometry of RBPs per virion can vary from 1 copy (e.g., RBP_pb5 in coli-siphophage T5) to 54 (e.g., in lacto-siphophage TP901-1) and maybe more. Phages can also combine RBPs with different binding specificities (10). Multiple binding is proposed to increase the affinity of saccharide binding by avidity. It also has been proposed to position the tail tube correctly for cell wall perforation (11) and has been shown to trigger infection (e.g., in T4 [12, 13] and T7 [14, 15]). All RBP structures determined to date are saccharide binding RBPs, except for the tip of T4’s long tail fiber, gp37 (16).

Phage T5, a member of the Siphoviridae infecting E. coli, is a model phage belonging to the T series introduced in the 1940s (17). It bears an icosahedral capsid that protects the 121-kb DNA and a 250-nm-long flexible tail (18). The host recognition apparatus, consisting of three side tail or L-shaped fibers and a central straight fiber, is located at the distal end of the tail. The central fiber is composed of a trimer of the C-terminal part of the baseplate hub protein (or Tal protein) pb3, followed by a trimer of pb4, at the tip of which is found a monomer of RBP_pb5, a 640-residue RBP (4, 19, 20). The side tail fibers, formed by pb1, reversibly bind lipopolysaccharide (LPS) polymannose O antigens O8 and O9 (21, 22), allowing the phage to “walk” at the surface of the bacterium until RBP_pb5 irreversibly binds to FhuA, an E. coli iron-ferrichrome transporter (23). Shigella and Salmonella strains also bear FhuA proteins, which share high identity with E. coli FhuA, and Salmonella enterica serovar Paratyphi at least can also be host to T5 (24). T5 side tail fibers can be dispensable under laboratory conditions, as the T5hd1 mutant that lacks these fibers only impacts host adsorption rates (19, 25), although their modification can also expand the host range (26). Inversely, a point mutation in the RBP_pb5 gene renders the presence of the side tail fibers indispensable. The RBP_pb5 gene was thus named oad: O antigen dependent (27). In vitro, mere interaction of T5 with purified FhuA commits T5 to infection and triggers DNA ejection (28), making T5 an excellent model to study phage-host interaction at the cellular and molecular levels (18, 29). Both RBP_pb5 and FhuA can be purified and can bind, in vitro, in a 1:1, irreversible complex (30, 31). Biophysical analysis of the individual FhuA and RBP_pb5 proteins compared to the complex showed that some secondary structure changes occur in RBP_pb5 upon binding to FhuA (31). Also, a low-resolution envelope of the proteins obtained by small-angle neutron scattering (SANS) did not show large conformational changes in either FhuA or RBP_pb5 upon interaction (32) but did show that RBP_pb5 is an 8-nm-long elongated protein. We recently solved T5’s tail tip structure before and after interaction with FhuA that was reconstituted into a nanodisc and built atomic models for all the tail tip proteins except RBP_pb5 and FhuA (29). These structures allowed us to describe the sequence of events undertaken by the T5 tail tip upon FhuA-RBP_pb5 binding, namely, straight fiber bending, opening of the tail tube, its anchoring to the membrane, and finally, formation of a transmembrane channel. Thus, unravelling the structure of RBP_pb5 before and after interaction with FhuA is the last piece of the puzzle to fully understand the infection trigger mechanism of phage T5 following FhuA binding. This structure is the first one of a nonsaccharide binding RBP bound to its receptor.

RESULTS AND DISCUSSION

Overall structure of the FhuA-RBP_pb5 complex.

The FhuA-RBP_pb5 complex being so stable and FhuA having been crystallized under different conditions and with different detergents (e.g., references 33 and 34), we tried to crystallize it but obtained only nonreproducible 8-Å anisotropically diffracting crystals (35). To improve the crystals, we produced nanobodies against the complex and used the lanthanide complex XO4 (nucleating and phasing agent) (36) to increase the chances of obtaining well-diffracting crystals. Under neither condition could reproducibly diffracting crystals be obtained. Two-angstrom diffracting two-dimensional (2-D) crystals were obtained and analyzed in electron diffraction and image analysis in collaboration with Mohamed Chami and Henning Stahlberg (Basel) (35). The crystals were, however, too thick, rendering structure determination impossible. Finally, taking advantage of the “resolution revolution” in electron cryomicroscopy (cryo-EM) (37), we investigated directly the FhuA-RBP_pb5 structure as a complex in solution by cryo-EM. Even though the complex is relatively small (150 kDa), we obtained a 2.6-Å-resolution three-dimensional (3-D) structure (Fig. 1A; Fig. S1 and Table S1 in the supplemental material).

FIG 1.

Overall structure of isolated FhuA-RBP_pb5 and fit in the Tip-FhuA map. (A) Isosurface view of the FhuA-RBP_pb5 complex in two 90° related side views (left), and ribbon representation of the modeled proteins with FhuA in blue and RBP_pb5 in gold (right). The LPS density and model is labeled (LPS). (B) FhuA-RBP_pb5 map fitted into the SANS envelope of FhuA-RBP_pb5 (32). (C) Isosurface view at high contour level of the unmasked and unfiltered cryo-EM map of T5 Tip-FhuA (EMD-14799) (29) with the structure of FhuA-RBP_pb5 manually fitted into the empty density below the pb4 spike and anchoring into the nanodisc. Cyan, density corresponding to pb6; purple, p132; orange, pb9; yellow, pb3; green, pb4; red, pb2. Unattributed densities are in white. (D) Manual fit of the AlphaFold2-predicted RBP_pb5 pb4 binding subdomain into the density-extending pb4 spike, with the pb4 structure shown in ribbon from (EMD-14800, PDB code 7ZN4). Top left, section showing the two last β-strands of the pb4 spike trimer; bottom left, section of the density-extending pb4 spike. The positions of the sections are localized on the longitudinal thin section of the pb4 spike map and model to the right. The position of the latter section is indicated on the Tip-FhuA model by an asterisk in the inset (PDB code 7ZN2). (E) Overall position of the AlphaFold2-predicted RBP_pb5 structure in the Tip-FhuA map (same as in panel C) after manual fitting of the pb4 binding subdomain as in panel D.

The complex and RBP_pb5 within it have an oblong structure, the interaction surface occurring through the distal end of RBP_pb5, which penetrates into the FhuA barrel down to the FhuA plug. The overall FhuA-RBP_pb5 structure fits well in the SANS envelope (Fig. 1B) (32). It also fits well in the map of the T5 tail tip interacting with its receptor reconstituted into a nanodisc (Tip-FhuA) (Fig. 1C). In the FhuA-RBP_pb5 structure, both proteins are very well resolved, except for the FhuA His₆ tag located after residue 405 in loop 5 and residues 1 to 17 that contain the FhuA TonB box. These are disordered in all FhuA structures, except when FhuA is crystallized with the TonB C-terminal domain (23, 38). The lipopolysaccharide that copurifies with FhuA could also be fitted. In RBP_pb5, the first 37 N-terminal residues, as well as residues 333 to 352, 365 to 371, and 427 to 455, could not be traced in the cryo-EM map. These four sequences point to the RBP_pb5 apex, suggesting, from the fit in the Tip-FhuA map, that they form a phage binding subdomain in interaction with the pb4 spike C terminus. It seems that these loops are therefore unstructured without the pb4 spike partner.

Analysis of the FhuA-RBP_pb5 structure.

(i) RBP_pb5. As previously proposed (31), RBP_pb5 folds as a unique domain rich in β-structures. A large and long central 9-stranded β-sheet serves as the spine of the protein. It is surrounded by shorter β-sheets in the apical half and by random coil loops and small α-helices in the distal half of the protein (Fig. 2A). DALI is a software that searches the Protein Data Bank for structural homologues (39). A DALI analysis does not align the RBP_pb5 whole structure with any protein of known structure. However, a β-sandwich subdomain shows structural similarity with other proteins: a close analysis of the aligned DALI domains highlights three proteins with the same topology (Fig. 2B). These three matches are all from phage trimeric proteins, two from RBPs (S. aureus phage K gp144, PDB code 5M9F, DALI Z-score = 5.1, root mean square deviation [RMSD] 4.4 Å, number of equivalent residues (lali) 100, and phage T7 gp17, PDB code 4A0U, DALI Z-score = 3.2, RMSD 6.6 Å, lali 92) and one from a T4 baseplate protein (gp9, PDB code 1QEX, DALI Z-score = 4.9, RMSD 4.9 Å, lali 87, with, however, a slight topological difference on the last β-strand). In RBP_pb5, the β-strands are longer. This domain displays an original fold, which is nevertheless similar to the knob domain of other viral RBPs, that is believed to interact with the host receptor (40). This again illustrates the “Lego” strategy of phages and viruses, which use and exchange functional modules (41). The Sword software analyses a protein structure and proposes domain partitioning. It proposes that RBP_pb5 folds as a unique domain; as an alternative, the algorithm also proposes subdivision of RBP_pb5 into three subdomains, including the common RBP subdomain (Fig. 2C).

FIG 2.

RBP_pb5 structure analysis. (A) Top left, ribbon representation of RBP_pb5 structure in the FhuA-RBP_pb5 complex context in two 90° related side views; right, topology diagram. Both representations are rainbow colored: N-terminal end, blue; C-terminal end, red. Bottom left, blow-up of the apical region of RBP_pb5 with indication of the residues bounding the unresolved pb4 binding subdomain. (B) DALI alignment of RBP_pb5 (gold) with domains from phage trimeric proteins with PDB codes 5M9F (orange), 1QEX (red), and 4A0U (beige). (C) Subdomain partitioning proposed by the Sword software (https://www.dsimb.inserm.fr/sword/). The position of the aligned subdomain in panel B is boxed. (D) Top, best ranked AF2-RBP_pb5 ribbon structure colored by prediction confidence score (high pLDDT, blue; low pLDDT, dark magenta), alone (left) or aligned with the RBP_pb5 structure within the FhuA-RBP_pb5 complex (gold) (right). The pb4 binding subdomain, unresolved in our cryo-EM structure, is clearly identified at the apex of AF2-RBP_pb5. Bottom, pLDDT scores of the five AF2-RBP_pb5 predictions. (E) Cryo-EM RBP_pb5 ribbon representation colored by sequence similarity/divergence (46) (color key as shown) after ClustalO alignment of a BLAST search.

Structure prediction software programs have made tremendous progress over the last few years (42), AlphaFold2 in particular (43). We submitted the RBP_pb5 sequence to AlphaFold2 to investigate the structure of RBP_pb5 missing the phage binding subdomain and, also, to compare an RBP_pb5 predicted structure with our structure of RBP_pb5 in complex with FhuA. We obtained five different predicted models (AF2-RBP_pb5) (Fig. 2D; the PDB file of the best predicted model is given in the supplemental material). For all models, the AF2-RBP_pb5 structure can be divided into two halves: the distal half is predicted with a poor confidence (predicted local-distance difference test [pLDDT] < 50), while the apical half is predicted with high confidence (pLDDT > 70). This includes the N terminus and disordered loops of our RBP_pb5 structure, which are predicted to form a small β-helix, attached to the RBP_pb5 core by five linkers (residues 37 to 45, 327 to 336, 353 to 357, 426 to 432, and 456 to 463). This β-helix would perfectly extend the pb4 spike: indeed, it fits very well in the density present in the Tip-FhuA cryo-EM map extending the pb4 β-helix (Fig. 1D). However, the core domain of AF2-RBP_pb5 does not fit at all the rest of the low-resolution density available for RBP_pb5 (Fig. 1E). Indeed, our Tip-FhuA map rather suggests a different angle between the pb4 binding subdomain and the RBP_pb5 core domain, made possible by the flexibility given by the five linkers (see also below). This flexibility, visible on negative stain images (see Fig. 5 in reference 20) explains in part why cryo-EM image analysis of the T5 tip and Tip-FhuA did not generate a high-resolution map for RBP_pb5.

Considering the predicted structure of the AF2-RBP_pb5 core domain, the apical half aligns very well with our RBP_pb5 structure (RMSD ranging from 0.811 to 0.979 Å over 222 to 291 residues depending on the AF2 model). This is coherent with the good level of confidence calculated by AlphaFold2. On the other hand, alignment of the poorly predicted distal half, corresponding to the FhuA binding domain, is rather bad (overall RMSD ranging from 7.6 to 18.2 Å over all 572 residues for the five AF2-RBP_pb5 models). pLDDT values below 50 are indicated to be strong predictors of disorder (44), suggesting that the regions are unstructured under physiological conditions: they would only fold upon interaction with FhuA. Even though AlphaFold predictions are not always correct, disorder in the RBP_pb5 distal half is consistent with biophysical and biochemical data that allow the conclusion that isolated RBP_pb5 is less structured and sees an increase in its β-sheet content at the expense of other structures, as well as a rigidification of its tertiary structure, upon binding to FhuA (31). This also explains the strong thermal stabilization of RBP_pb5 upon binding to FhuA (melting temperature [T_m] = 43°C for RBP_pb5 alone and 89°C for FhuA-RBP_pb5) (31). We also submitted the FhuA-RBP_pb5 complex to AlphaFold-Multimer (45): the prediction exhibits a wrong orientation between FhuA and RBP_pb5 and a very small interacting surface. RBP_pb5 does, however, appear a bit more structured, in particular its large central sheet (Fig. S1F). We have proposed the FhuA-RBP_pb5 complex to CASP15 and are expecting the results by the end of 2022.

Whereas there are no RBP_pb5 homologues in the PDB, a BLAST search finds a large number of proteins (>100), with identities ranging from 29.0 to 96.3%. They are described as E. coli, Salmonella, Shigella, or Klebsiella phage hypothetical RBPs or RBPs. This is consistent with the fact that the T5-like genus is well represented in the databases. Clustal alignment and the ConSurf Server (46) allow visualization of the distribution of conserved versus variable residues on the RBP_pb5 structure (Fig. 2E). The dichotomy of the distribution is quite striking: the apical half of RBP_pb5 is largely conserved, `in particular for the unresolved N terminus and loops (not shown), while the distal half is much more variable. This trend—phage binding domain conserved/receptor binding domain much more variable—is quite general in RBPs, fibers, and TSPs, highlighting that the receptor binding domains are under heavy evolutionary pressure to diversify as they interact with a large variety of host targets, while the phage binding domain, bearing a structural function, is more constrained by interaction with its phage partner(s) (47).

(ii) FhuA. The structure of FhuA has been solved in its apo form, in complex with many ligands and with the C-terminal domain of TonB (23). The different structures are remarkably similar, with RMSDs between the 11 different structures below 1.93 Å over all 696 common residues. The main difference between the different structures is the N-terminal region: in FhuA apo structures, residues 19 to 30 are structured as an α-helix, whereas in structures binding TonB-dependent ligands, a conformation that is superimposable to the TonB-bound conformation is adopted (23). In agreement with the fact that T5 infection is TonB independent, the conformation of RBP_pb5-bound FhuA residues 19 to 30 is that of apo FhuA.

FhuA-RBP_pb5 interface.

RBP_pb5 interaction with FhuA occurs through the distal β-turns and loops of the central sheet and through random coil loops and small α-helices in the distal half of RBP_pb5, with a total of 63 residues involved, as calculated by PISA, a software that analyses interfaces from PDB files (Fig. 3A) (48). In FhuA, interactions are mediated by the apical tip of the plug, including residues in loops 4 to 11 and in β-strands 7 to 12, with a total of 64 residues involved (Fig. 3B). The surface of interaction is large, with a buried area of 2,034 Ų and a PISA calculated energy of −21.4 kcal/mol, and the interaction is stabilized by 13 hydrogen bonds and hydrophobic interactions, explaining the quasi-irreversibility of the interaction (31). Interestingly, a C₁₀DAO (n-decyl N,N-dimethylamine N-oxide) molecule, contained in the buffer, could be modeled at the interface between the FhuA plug and RBP_pb5 (Fig. 3C; Fig. S1E). It overlapped the ferrichrome hydrophobic binding pocket and interacted, on the FhuA side, with 13 residues—four from the plug, four from loops 3 and 4, and five from β-strands 7 to 9, with a hydrogen bond between Y116 and the carboxyl of the amine oxide head group. The C₁₀DAO binding pocket is further completed by four residues from loops of the tip of the RBP_pb5 β-sheet, further stabilizing the interaction between the two proteins (Fig. 3C).

FIG 3.

Surface of interaction between FhuA and RBP_pb5. (A) Two orientations of a rainbow-colored ribbon representation of RBP_pb5’s distal part, with residues involved in the interactions, as determined by PISA, shown in sticks. The C₁₀DAO molecule is shown in pink. (B) Left, topology diagram of FhuA barrel (from reference 33), on which are indicated the residues involved in the interaction (dotted-line and filled symbols indicate residues interacting exclusively with the C₁₀DAO molecule). Right, ribbon diagram of FhuA plug, with residues involved in the interactions and the C₁₀DAO molecule shown in sticks. (C) Zoomed image of the FhuA-RBP_pb5 interface. FhuA is in cyan and RBP_pb5 in gold. Both proteins are in ribbon form, with the residues involved in C₁₀DAO binding in sticks. The C₁₀DAO density is shown in pink. FhuA PADKGH is highlighted in blue. In RBP_pb5, G166 is highlighted in orange and FhuA-F566 in blue sticks.

T5 is a historic phage and a prototype of the siphophages, for which host binding has been extensively studied in the past. The presence of an RBP binding to FhuA was proposed while studying the oad mutant, which had a reduced affinity to FhuA (19, 27). Later, Mondigler et al. identified the oad mutation: Gly166 is switched to a Trp (49). In our structure, Gly166 participates in the binding interface, making a C-H…Π interaction with FhuA Phe566 (Fig. 3C). A Gly→Trp mutation would thus indeed destabilize the protein-protein interaction interface. Mondigler et al. also investigated the ability of different RBP_pb5 large-deletion (over 40 residues) and insertion mutants to bind to FhuA (49). Apart from one (deletion of the 153 C-terminal residues), none retained binding abilities, which is explained by the tight fold of FhuA-bound RBP_pb5.

FhuA’s binding to RBP_pb5 has also been studied from the FhuA perspective: before the FhuA structure was known, Killmann et al. investigated T5 binding loops in FhuA by competitive peptide mapping (50). They identified a peptide—₃₃₄PADKGH₃₃₉—of FhuA loop 4 that induced DNA ejection when incubated with T5, in a temperature-dependent manner. However, we could not reproduce the results with a slightly longer peptide (APADKGHY), and, as seen in our structure, this sequence does not appear to be involved in RBP_pb5 binding (Fig. 3C). Single substitutions impairing T5 binding could not be isolated unless they perturbed FhuA’s overall structure (23), and deletion of loops 3 to 11, except for loop 8, did not prevent T5 binding (51). This can be explained by the large interacting surface between the two proteins, which could compensate single substitutions or loop deletions. There does not seem to be a structural rationale for loop 8 being more important for RBP_pb5 binding than other loops. Deletion of loop 8 also impairs T1 and ϕ80 binding and ferrichrome transport, suggesting that its deletion might prevent the proper folding of FhuA. A FhuA plugless mutant was also investigated for its ability to bind T5: an E. coli strain with deletion of endogenous FhuA and expressing FhuAΔplug was resistant to T5, thus suggesting that RBP_pb5 did not bind to FhuAΔplug (52). Given the surface of interaction provided by the FhuA barrel loops, this result suggests that the latter are either flexible or adopt a different conformation in FhuAΔplug than in wild-type (WT) FhuA, preventing RBP_pb5 binding. Indeed, the plug does not provide a larger surface of interaction than individual FhuA barrel loops. The role of the plug is, however, important in the overall stabilization of the interaction between FhuA and RBP_pb5, as previously suggested (31): whereas FhuA displays two T_ms, one for the plug (64°C) and one for the barrel (75°C), the RBP_pb5-FhuA complex displays a unique T_m that is shifted to 89°C.

T5 is described as infecting E. coli; however, some other strains also bear FhuA. Thus, we tested T5’s sensitivity on different E. coli, Shigella, and Salmonella strains, which were chosen because of their different levels of FhuA identity with the FhuA of the E. coli laboratory strain, ranging from 96% to 100% (Fig. S2). T5 sensitivity tests were performed in the presence of dipyridyl, an iron chelator that boosts FhuA expression. Surprisingly, only Salmonella bongori was sensitive to T5. The FhuA sequence alignment, however, shows some differences in the residues interacting with RBP_pb5, presumably reflecting the resilience of the FhuA-RBP_pb5 interaction. We checked for T5 irreversible adsorption on all strains, but only Salmonella bongori showed adsorption, comparable to that of E. coli. Interestingly, enteropathogenic E. coli (EPEC) and the Shigella strains were resistant to T5 despite there being no difference in the interacting residues (Fig. S2). Resistance to T5 may be explained by a different O antigen content of the LPS molecules, FhuA being blocked by a superinfection exclusion protein coded by a prophage in their genomes, or by the synthesis of a capsule, shielding FhuA from the phage.

Conclusion.

We present here the first structure of an RBP, RBP_pb5, in complex with its protein receptor, FhuA. Analysis of this structure compared to that predicted for RBP_pb5 and fitted into the T5 tail tip map after interaction with FhuA (Tip-FhuA) (29) allows us to propose a mechanism that would transmit FhuA binding to pb4 through RBP_pb5 (Fig. 4): the apical subdomain of AF2-RBP_pb5 shows a small structured β-helix that fits well in our Tip-FhuA map, in the densities extending the pb4 β-helix spike. This RBP_pb5 subdomain would serve as an adaptor between the trimeric pb4 and the monomeric RBP_pb5. The presence of linkers between this subdomain and the RBP_pb5 core confers flexibility between these two parts of the protein. The AF2-RBP_pb5 predicted structure and biophysical analysis of RBP_pb5 (31) suggest that the distal half of RBP_pb5 is unstructured when not bound to FhuA. Upon binding to FhuA, RBP_pb5’s distal half becomes structured, rigidifying the protein. In our Tip-FhuA map, the RBP_pb5 core and the pb4 binding subdomain lie at an angle of ~45°. We suggest that rigidification of the RBP_pb5 distal half is transmitted to the pb4 binding subdomain, where it imposes a ~45° kink. This would constrain the RBP_pb5 pb4 binding subdomain, which in turn would constrain the pb4 β-helix spike and induce the different twist in the pb4 β-helix. This last twist could then initiate the conformational change cascade that results in the tail tube opening and the tube anchoring to the membrane in register with a channel that goes through the outer membrane (29). Definite confirmation of this hypothesis will require determination of the structure of pb4-bound RBP_pb5 before and after interaction with FhuA.

FIG 4.

Proposed mechanism of signal transduction of FhuA’s binding to the pb4 spike through RBP_pb5. Left, within the T5 tail, the RBP_pb5 distal half is unstructured and the RBP_pb5 core is flexible with respect to the pb4 binding subdomain. Upon binding to FhuA, the RBP_pb5 distal half gets structured (1). This would induce rigidification of the whole protein, in particular of the linkers between the core of RBP_pb5 and the pb4 binding subdomain, which are stabilized at an angle of ~45° (2). This in turn would induce a modification of the twist of the pb4 β-helix (3), which induces the opening of the tail tube, its anchoring to the membrane, the expulsion of the tape measure protein, and the formation of a channel across the host outer membrane (29).

MATERIALS AND METHODS

Bacterial and phage strains.

The heat-stable deletion mutant T5st0 was used for sensitivity and adsorption tests. E. coli strain F, a fast-adsorbing strain for T5, was used for the production of T5st0 and for control of free T5 in adsorption tests. Shigella flexneri (CIP 106171), Shigella sonnei (CIP 106204), Salmonella enterica serovar Typhimurium (CIP 104474), Salmonella enterica serovar Indica (CIP 102501T), Salmonella bongori (CIP 82.33T), and enteropathogenic E. coli (CIP 52.170) were used for T5 sensitivity and adsorption tests. E. coli strains AW740 and BL21(DE3) were used for FhuA and RBP_pb5 expression, respectively.

FhuA and RBP_pb5 production and purification.

FhuA was purified from E. coli strain AW740 transformed with a plasmid encoding the complete fhuA gene, including its signal sequence, and in which a His-tag coding sequence has been inserted in extracellular loop L5. A His tag in this position was shown not to impair RBP_pb5 binding (30). RBP_pb5 was purified from E. coli strain BL21(DE3) carrying the oad gene encoding RBP_pb5 fused to a His tag-coding sequence 3′ from oad in a pET-28 vector (30). Cells were grown in LB medium at 37°C in the presence of the iron-chelating agent dipyridine (100 μM) for AW740 cells and at 20°C without induction for BL21(DE3) cells. FhuA and RBP_pb5 purifications were carried out as described previously (31). The FhuA-RBP_pb5 complex was formed by adding equimolar amounts of the two proteins, which results in 100% complex formation as described in reference 30. Detergent exchange on the FhuA-RBP_pb5 complex was performed by a 10-times dilution in deionized water and centrifugation at 100,000 × g for 45 min at 4°C. The pellet was rinsed with water and resuspended in 20 mM Tris, 1.6% C₁₀DAO at a protein concentration of 4.3 mg/mL.

Cryo-EM sample preparation.

Typically, 3.5 μL of the FhuA-RBP_pb5 complex was deposited on a freshly glow-discharged (25 mA, 30 s) Cu/Rh 400-mesh Quantifoil R 2/1 EM grid and flash-frozen in nitrogen-cooled liquid ethane using a ThermoFisher Mark IV Vitrobot device (100% humidity, 20°C, 2 s blotting time, blot force 1). Preliminary screening of freezing conditions was performed on an FEI F20 electron microscope.

Two different data sets were acquired on the same grid. For both data sets, 60-frame movies with a total dose of 60 e⁻/Ų were acquired on a ThermoFisher Scientific Titan Krios G3 transmission electron microscope (European Synchrotron Radiation Facility, Grenoble, France) (53) operated at 300 kV and equipped with a Gatan Quantum LS/967 energy filter (slit width of 20 eV used) coupled to a Gatan K2 summit direct electron detector. Automated data collection was performed using ThermoFisher Scientific EPU software. A nominal magnification of ×130,000 was used, resulting in a calibrated pixel size at the specimen level of 1.052 Å/pixel. For the first data set, 777 movies were acquired with a phase plate, close to focus (between −0.5 and −1.0 μm), while for the second data set, 8,752 movies were acquired without a phase plate and with a defocus ranging between −1.2 and −2.6 μm.

EM image processing.

For both data sets, the processing was done with Relion (54). Motion correction using 5-by-5 patches was done with Motioncor2 (55), while contrast transfer function (CTF) estimation was done with Gctf (56).

For Data Set S1, an initial set of particles was selected with the Laplacian of Gaussian (LoG) picker in Relion. After 2-D classification with 2-times-binned particles, nicely defined 2-D classes were used as references for a template-based picking. Following another 2-D classification, an initial model was obtained in Relion, which was then used to get a first 3-D reconstruction. 3-D classification (without a mask) was performed to select the best particles, which were then used to compute a final 3-D reconstruction at 6.7 Å (at a Fourier shell correlation [FSC] of 0.143) into which the atomic model of FhuA (PDB code 1QFG) could be unambiguously fitted (not shown).

For Data Set S2, 2-D projections of the 3-D model obtained from Data Set S1 were used as a template to automatically pick all micrographs. In all, 2,191,586 particles were obtained, binned two times and further subjected to two rounds of 2-D classification to remove obvious outliers. A first 3-D classification was done without any mask, and the selected particles were re-extracted without binning. Following a 3-D refinement step, two successive 3-D classifications with a mask following the contour of the complex were performed without any image alignment, leading to a homogenous subset of 109,350 particles. Then, particle polishing and two rounds of CTF refinements (magnification anisotropy, beam tilt, and per-particle defocus) were performed before a last 3-D refinement yielded the final 3-D reconstruction at 2.6-Å average resolution (at an FSC of 0.143). The final map was sharpened using DeepEMhancer (57).

Model building.

The RBP_pb5 protein model was built de novo in a FhuA-RBP_pb5 cryo-EM map using Coot (version 0.9.2) (58). FhuA was adapted from an FhuA structure solved by X-ray crystallography (PDB code 2GRX). The two individual proteins and the complex were then refined using the PHENIX (version 1.18.2-3874) Real Space Refine tool (58). Structure validation was done using the MolProbity online tool.

AlphaFold structure prediction.

AlphaFold prediction of RBP_pb5 was performed on the CollabFold server with AlphaFold2 version 1.3 (59). AlphaFold prediction of FhuA-RBP_pb5 was performed using AlphaFoldMultimer (45).

Phage sensitivity tests.

Shigella, Salmonella, and E. coli strains were grown in LB medium at 37°C, 180 rpm until reaching an optical density at 600 nm (OD₆₀₀) of 0.5. Amounts of 1.2 mL of LB soft agar were inoculated with 300 μL of culture and spread on an LB petri dish. Serial dilutions of phage T5, at 0 to 10⁹ PFU/mL, were directly spotted on the soft agar. The titer was estimated by counting lysis plaques in the spotted area after overnight incubation at 37°C.

Irreversible phage adsorption tests.

To 5 mL of each strain grown in LB medium at an OD₆₀₀ of 0.2, 0.01 M CaCl₂, 0.04 M MgCl₂, and T5 at a multiplicity of infection of 0.5 were added, and the mixtures incubated at 37°C. Amounts of 200 μL of infected cells were collected at different times, vigorously vortexed for 10 s, and centrifuged for 10 min at 15,000 × g. The supernatants containing free phage (unabsorbed and reversibly bound particles) were titrated on E. coli strain F using the double-layer agar method described above.

Data availability.

Cryo-EM density maps of FhuA-RBP_pb5 and the associated atomic coordinates have been deposited in the Electron Microscopy Data Bank (EMDB) and in the Protein Data Bank (PDB) under the following accession numbers: PDB code 8B14 and EMD-15802.