Structural changes in bacteriophage T7 upon receptor-induced genome ejection

Significance

Many double-stranded DNA (dsDNA) viruses, including the tailed dsDNA bacteriophages and herpesviruses, use molecular machines at a unique vertex of the capsid for genome delivery. Bacteriophage T7 provides a tractable model for the study of dsDNA viruses. Here, we report the structure of the mature bacteriophage T7, as well as the structures of the full and empty T7 particles in complex with their cell receptor lipopolysaccharide. The symmetry-mismatch machines at the vertex, including internal ejection proteins, have been resolved at near-atomic resolutions in our structures. The structures of the T7 particles at different stages reveal a series of structural rearrangements after viral genome delivery.

Abstract

Many tailed bacteriophages assemble ejection proteins and a portal–tail complex at a unique vertex of the capsid. The ejection proteins form a transenvelope channel extending the portal–tail channel for the delivery of genomic DNA in cell infection. Here, we report the structure of the mature bacteriophage T7, including the ejection proteins, as well as the structures of the full and empty T7 particles in complex with their cell receptor lipopolysaccharide. Our near–atomic-resolution reconstruction shows that the ejection proteins in the mature T7 assemble into a core, which comprises a fourfold gene product 16 (gp16) ring, an eightfold gp15 ring, and a putative eightfold gp14 ring. The gp15 and gp16 are mainly composed of helix bundles, and gp16 harbors a lytic transglycosylase domain for degrading the bacterial peptidoglycan layer. When interacting with the lipopolysaccharide, the T7 tail nozzle opens. Six copies of gp14 anchor to the tail nozzle, extending the nozzle across the lipopolysaccharide lipid bilayer. The structures of gp15 and gp16 in the mature T7 suggest that they should undergo remarkable conformational changes to form the transenvelope channel. Hydrophobic α-helices were observed in gp16 but not in gp15, suggesting that gp15 forms the channel in the hydrophilic periplasm and gp16 forms the channel in the cytoplasmic membrane.

Many double-stranded DNA (dsDNA) viruses, including tailed bacteriophages and herpesviruses, have a portal attached to a unique pentameric vertex of their icosahedral capsid shell (1–3). The portal is a dodecameric channel for viral DNA packaging and ejection. The tailed bacteriophages and herpesviruses encapsidate DNA in the capsid shell through the portal channel (4–10), and the last packaged DNA is held by tunnel loops (or β-hairpins for herpesviruses) in the portal (11–16). The last packaged DNA in most of the tailed bacteriophages and herpesvirus is the first to be ejected during the genome delivery (17). In tailed bacteriophages, the portal connects to a tail, which serves to recognize host cell receptors and deliver the genome into the cytoplasm (18). Gram-negative bacteriophage in Podoviridae initiate infection through a specific interaction of its receptor-binding protein with the receptor lipopolysaccharide (LPS) on the host cell surface. The phages in Podoviridae have a noncontractile tail that is too short to span the gram-negative bacteria envelope that comprises the outer membrane, the cytoplasmic membrane, and the peptidoglycan layer in the hydrophilic periplasm in between (19). After adsorption, a signal is transmitted for the release of internal ejection proteins to form a channel that extends the tail across the cell envelope and that allows for subsequent genome ejection into the infected cell (20–23). In many previous studies, structural analyses have been performed at resolutions of 9 to 40 Å on this highly coordinated dynamic infection process (21–26). These studies have provided insights on structural changes of phage particles that accompany the infection steps before and after the genome ejection. However, these studies did not resolve structures of the internal ejection proteins. Furthermore, the relative low resolutions cannot clarify the dynamic genome ejection process orchestrated by the ejection proteins, portal, and tail.

Escherichia coli bacteriophage T7, a member of the Podoviridae family, has been used as a model for understanding the DNA packaging and delivery mechanism that are common to tailed phages and related dsDNA viruses (10, 21, 27–33). T7 has an icosahedral capsid shell formed by gene product 10 (gp10). The 12-fold portal (gp8) shares a very similar topology with those in other phages and herpesviurses (14–16, 30, 34). The tail comprises a 12-fold adaptor protein gp11 assembly, a sixfold nozzle protein gp12 assembly, and six subunits of trimeric tail fiber gp17 (21, 30). These tail fibers are responsible for bacterial receptor recognition and adsorption (21, 33). On top of the portal within the capsid shell is a hollow cylinder-shaped core structure (10, 28) formed by the ejection proteins (core proteins) gp14, gp15, and gp16, which have been suggested to form a transenvelope channel for the genome delivery into the infected cell (20, 35, 36). The gp16 harbors lytic transglycosylase (LTase) activity, which allows for penetration into the bacterial peptidoglycan layer (37).

In this study, we present the structure of the mature bacteriophage T7 with internal core proteins at near-atomic resolution and the structures of the full and empty T7 particles in complex with their cell receptor at subnanometer and near-atomic resolutions, respectively. Our reconstruction reveals that the core in the mature T7 is formed by a fourfold gp16 ring, an eightfold gp15 ring, and a putative eightfold gp14 ring. The putative gp14 structures mediate the core–portal interaction. The gp15 and gp16 are mainly composed of helix bundles, and gp16 harbors a LTase domain. When the T7 phage interacts with the LPS, the tail nozzle opens. Six copies of gp14 anchor to the sixfold tail channel, extending the tail across the LPS lipid bilayer. A conformational change in the portal then triggers the genome ejection. Our structures reveal the structural changes of the phage genome-delivery molecular machines after the genome delivery.

Results

Structure of the Core, Portal–Tail Complex, and DNA in Mature T7 Phage.

We reconstructed the asymmetric structure of the whole mature phage at 7.0-Å resolution (Fig. 1A and SI Appendix, Fig. S1A) from 75,213 particle images by using cryoelectron microscopy (cryo-EM) and our symmetry-mismatch reconstruction method (38). Our local reconstruction shows that the core, portal, and tail can be resolved at a resolution of 5.8 Å together (SI Appendix, Fig. S1B), suggesting that the three complexes have relatively fixed interactions. Further local reconstructions improved the resolutions of the portal–tail complex and core to 3.4 and 3.7 Å, respectively (Fig. 1B and SI Appendix, Fig. S1 B–D). Our reconstruction shows that the core is an α-helix–rich tapered cylindrical structure of ∼150 Å in height sitting on the helical barrel domain of the portal (Fig. 1B and SI Appendix, Figs. S2 and S3). The bulky side chains in the core structure allowed us to build the atomic models of an eightfold gp15 ring and a fourfold gp16 ring unambiguously (Fig. 1C and SI Appendix, Fig. S3 A–C). The copy numbers of gp15 and gp16 in our structure are roughly consistent with the previous estimates of 9.1 ± 0.7 and 5.0 ± 0.4, respectively (39).

Fig. 1.

Structures of the core and portal–tail complex in the mature phage. (A) Surface and cut-open views of the reconstruction of mature T7. The color code applies to A, B, and K. (B) Cut-open view of the core, portal, and tail complexes. (C) Superposition of our atomic model on the density (mesh) extracted from box (gp16) in B. (D) Atomic model (Ribbion) of the core and portal complexes. The color code is identical to that in A, except that a subunit of gp15 in the octamer and a subunit of gp16 in the tetramer are colored by domains. The yellow rod in the middle is the density map of DNA. (E–H) Cross-section views of the column in D. (I and J) Structures of gp16 and gp15 monomers. (K) Surface view of the dsDNA around the portal.

Except for the missing 64 N-terminal residues and 36 C-terminal residues, the resolved gp15 structure comprises a sidewall and a platform domain (Fig. 1J and SI Appendix, Figs. S3E and S4A). Eight copies of gp15 assemble into an eightfold ring (Fig. 1 B–F and J). The sidewall domain, which presents a cascade of three helix bundles, forms the sidewall of the ring with an inner channel ∼90 Å in diameter (Fig. 1F). The C-terminal platform domain forms the top platform of the ring with an inner channel ∼30 Å in diameter (Fig. 1E). No hydrophobic α-helix was observed in the gp15 structure.

Four copies of the gp16 monomer are stacked on the top platform of the gp15 assembly (Fig. 1 D and G). The atomic model of gp16 contains 1,061 out of 1,318 amino acid residues (SI Appendix, Figs. S3F and S4B). The missing regions comprise the residues 1 to 10, 38 to 77, 157 to 235, 503 to 550, 746 to 761, 961 to 982, 1,048 to 1,054, 1,234 to 1,261, and 1,312 to 1,318. All these regions are exposed on the surface of the fourfold gp16 assembly, suggesting that they interact with DNA.

The gp16 structure comprises an N-terminal small nodule domain and a large nodule domain (Fig. 1I). The large nodule domain, which is mostly composed of α-helices separated by loops or the missing regions, can be divided into five-helix bundle subdomains (Fig. 2A). Among these α-helices, α-14, and α-39 (Fig. 2A and SI Appendix, Fig. S4B) are mostly hydrophobic and were predicted to be the transmembrane helices by the TMpred server (40). The remaining three predicted transmembrane helices (residues 322 to 346, 1,190 to 1,209, and 1,251 to 1,269) are a short helix, two loops, and an unstructured segment in our gp16 structure (SI Appendix, Fig. S4B). However, these three segments were predicted to be three long α-helices (SI Appendix, Fig. S4B) by the Psipred server (41), suggesting that they might become helices after gp16 is ejected.

Fig. 2.

Gp16 atomic model. (A) LTase domain (blue) and five subdomains I (yellow, 928 to 1,185), II (cyan, residues 273 to 330), III (orange, residues 670 to 792), IV (green, residues 551 to 643, 795 to 858, and 1,207 to 1,233), and V (red, residues 385 to 502, 866 to 908, and 1,277 to 1,311) in the gp16 model. The connecting region among these domain and subdomains are in gray. (B) Superposition of the LTase domain (blue) in gp16 on the counterpart (pink) in P. aeruginosa Slt (PDB ID: 6FCU).

According to a search of the Protein Data Bank (PDB) using the Dali server (42), the small nodule of gp16 (residues 11 to 142) resembles the superfamily of LTase in E. coli and other bacteria (43–45). Among these LTases, the small nodule most resembles the LTase of Pseudomonas aeruginosa (44) (Fig. 2B). The amino acid sequence identity between the small nodule domain of gp16 and the LTase domain of P. aeruginosa is 21%. We designated the small nodule domain as an LTase domain. Residues 38 to 77 were not resolved in the LTase structure because this region flexibly interacts with the genomic DNA in the capsid. However, the missing residues can be built by homologous modeling according to the homologous LTase structure (44). The gp16 LTase can be classified into family 1 of the LTase superfamily according to four consensus motifs (46) (SI Appendix, Fig. S5). Glu37 in gp16, which is the proton donor for initiating glycosidic bond scission, is strictly conserved among family 1 (43, 44). If this Glu37 is mutated, then the T7 takes about 20× as long to form the extended tail in the bacteria envelope (21, 47). The Ser46, Gln55, Tyr108, and Glu134 in gp16, which interact with the substrate in the other family 1 LTases (43, 44), are also conserved (Fig. 2). Each gp16 monomer spans three copies of the gp15 platforms. The gp16 large nodule interacts with two copies of the gp15 platform, and the gp16 LTase domain interacts with one copy of the two platforms and an adjacent third platform (SI Appendix, Fig. S3A).

The octameric gp15 assembly is stabilized on the portal through eight copies of four α-helix subunits surrounding the outside of the portal helical barrel domain (Fig. 1 B, D, and F and SI Appendix, Fig. S2). Although the side chains were not well resolved, we were able to build a 79-residue polyalanine backbone model (Fig. 1 D and F and SI Appendix, Fig. S3 C and D). We tentatively assigned these α-helices to gp14 according to their location because they could not be assigned to any of the missing regions of gp15, gp16, and gp8. The density of gp14 is slightly weaker compared with those of gp15, gp16, and gp8. Therefore, the imposition of four- or eightfold symmetry could have biased the density, leading to an incorrect potential chain trace. In addition, the gp14 copy number could constitute an artifact of the symmetry imposition. The conformation and location of gp14 were different from those in a previous structural study of the T7 procapsid structure at 20-Å resolution, in which 12 copies of putative gp14 in nodule conformation connect to the portal wing and the octameric gp15 assembly (28). These discrepancies of the gp14 structures may reflect the different states of the T7 phage (39).

A rod-like structure, which is assigned to the last packaged DNA in the T7 and other tailed phages (4–10), passes through the central channel of the core and portal and ends at the interface between the adaptor and nozzle. This DNA is held by tunnel loops (Fig. 1H and SI Appendix, Fig. S2A), which were suggested to function as clamps to retain packaged DNA in the portal of a Podoviridae phage Φ29 (11). The DNA end, which is located at the interface between the dodecameric ring of gp11 and hexameric nozzle of gp12, interacts with the platform domain of gp12 (we follow the domain nomenclature used in ref. 30, SI Appendix, Fig. S2B). The channel of the nozzle is closed (Fig. 1 A and B). Some coaxial ring–like structures circle around the portal near the portal–capsid interface (Fig. 1 A and K). The grooves in the ring-like structures indicate that these rings are dsDNA (Fig. 1K). The dsDNA helix is right-handed, and the pitch of the helix is 34 Å (Fig. 1K), consistent with the shape of B-form dsDNA. The corresponding ring structures were also observed around the portal in the herpesviruses and other phages, but the typical dsDNA grooves were not resolved (7, 9, 14, 15, 48–50).

Structural Changes of Full and Empty T7 Particles upon LPS Interaction.

The rough LPS from E. coli is the main T7 receptor that induces DNA ejection in vitro (25). To investigate the structural changes of the phage after DNA ejection, the purified mature T7 particles were incubated with the rough LPS for cryo-EM structural analysis. The cryo-EM images indicate that ∼85% of T7 particles ejected their genomes after 3 h of incubation with the LPS, whereas ∼15% particles appear to be full of DNA (SI Appendix, Fig. S6). The structure of the full particle in complex with the LPS at 15-Å resolution and the local structure of the portal–tail complex at 10-Å resolution were obtained using the asymmetric reconstruction and local reconstruction, respectively (SI Appendix, Fig. S1). The lower resolutions can be ascribed to limited number (7,428) of the full particles. The portal–tail complex adsorbed to the rough LPS lipid bilayer is shown in Fig. 3A. The distal half-fibers were not resolved due to their flexibility. The lipid bilayer of the rough LPS from E. coli (51), which is ∼80 Å in thickness (Fig. 3A), is similar to the outer membrane of the E. coli cell (21).

Fig. 3.

Structures of full T7 particle upon LPS interaction at a 10-Å resolution. The distal half-fibers were not resolved due to their flexibility. The color code is identical to that used in Fig. 1A. (A) Surface and cut-open views of the full T7 in complex with LPS. (B) A slab view of the portal showing that DNA is held by tunnel loops. (C) Six pairs of α-helices appear in the nozzle if the display threshold is increased.

Compared with that in the mature phage, the organization of two DNA layers adjacent to the inner capsid surface and the ring-shaped DNA around the portal appear to be unchanged, whereas the organization of the central DNA appears to be more diffusive than that in the mature phage (Fig. 3A). The absence of the core structure in the full particle suggests that the core becomes disordered. The portal helical barrel domain in the mature phage (SI Appendix, Fig. S2) also becomes disordered in the full particle in complex with the LPS. The rod-like DNA structure is held by the tunnel loops in the portal and ends at the interface between the adaptor and nozzle (Fig. 3 A and B), which are identical to those in the mature phage (Fig. 1A). The nozzle adopts an open conformation with an inner diameter of ∼30 Å, and there are six pairs of α-helices present in the nozzle (Fig. 3 B and C). Central slice view (SI Appendix, Fig. S7) of the full particle in complex with the LPS is similar to previously reported ∼40-Å–resolution structures of the T7 and a T7-like marine phage at the adsorbed stage, which have appeared to be full and to display an extension of the tail penetrating the cell envelope (21, 24).

The asymmetric reconstruction of the DNA-ejected empty particles in complex with the LPS at 8.2-Å resolution and the local reconstruction of the portal–tail complex at 4.3-Å resolution show the structural changes of the portal and tail after the LPS-induced genome ejection (Fig. 4 A and B and SI Appendix, Fig. S1 C and E). The overall structure of the empty particle is similar to that of the full particle, except for the absence of DNA in the empty particle (Fig. 4A). The distal half-fibers are detached from the capsid surface and interact with the LPS lipid bilayer (Fig. 4A and SI Appendix, Fig. S8 and Supplementary Text). The distal half-fibers were not well resolved due to their flexibility.

Fig. 4.

Structures of DNA-ejected T7 particle upon LPS interaction. The color code is identical to that used in Fig. 1A. (A) Surface and cut-open views of the empty T7 in complex with LPS. (B) Cut-open view of the portal and tail complexes of the empty T7 (Left). Six pairs of α-helices of gp14 were resolved in the nozzle (Right). (C) Superposition of the model of gp14 α-helices on the density maps. (D) Schematic illustration of hydrophobic helices of gp14 in the outer membrane. The α1 to -3 are predicted helices in gp14 (SI Appendix, Fig. S4C), and the red region are the resolved structure in the nozzle as shown in panel (B). (E and F) Slab views of the portal–tail complexes of mature T7 and empty T7 in complex with LPS. The yellow rod indicates the density of DNA. Only two monomers of gp8 (purple), gp11 (green), gp12 (cyan), and gp14 (red) are shown. (G) Structural comparison of gp8 monomers in the genome-ejected T7 (green) and the mature T7 (purple).

The most distinct conformational change of the portal is that the longest α-helix (α-10) connecting the C terminus of the tunnel loop in the wing domain bends 30° upward at two glycines (Gly368 and Gly387) in the middle of the helix, resulting in an open conformation of the portal (Fig. 4 F and G and SI Appendix, Fig. S9). A similar bent conformation of the α-10 was observed in a crystal structure of the recombinant T7 portal (30). By contrast, in the structures of the mature phage and the full particle in complex with the LPS, the tunnel loop and α-10 interact with the DNA in the portal, which function as a clamp to retain the packaged DNA in the capsid (Figs. 1H, 3B, and 4E and SI Appendix, Fig. S2A).

We also observed other conformational changes in the portal of the empty T7 particle. The structures of the helical barrel and crown domains (13) become unstructured, and the channel is sealed by the unstructured helical barrel (Fig. 4A and SI Appendix, Fig. S9). The density map of the unstructured region is as strong as that of the remaining regions of the portal in the asymmetric reconstruction of the empty particles; however, it has not been resolved at higher resolution (Fig. 4B). In addition, minor shifts of α-helices and β-strands were also observed in the portal wing domain (Fig. 4E), whereas the stem and clip domains in the portal and the adaptor protein gp11 are structurally unchanged.

The overall movement of the six copies of gp12 monomers cause the opening of the nozzle channel (Fig. 4 A and B and SI Appendix, Fig. S10 A and B). Superposition of the gp12 monomer structures in the genome-ejected particle and in the mature phage shows minor structural element shifts in the nozzle tip domain (SI Appendix, Fig. S10C). Six pairs of antiparallel α-helices, which are anchored in the cleft between two adjacent gp12 monomers in the nozzle channel, were resolved (Fig. 4 B and C and SI Appendix, Fig. S10). We assigned the antiparallel α-helices to gp14 for the following reasons: 1) a previous biochemical result indicated that gp14 is localized to the bacterial outer membrane after being ejected into the cell (36) and 2) the PSIPRED server (41) predicted that gp14 had three α-helices, namely α-1, α-2, and α-3 (SI Appendix, Fig. S4C), and bulky side chains in α-1 and α-2 fit well with our density map (Fig. 4C). The α-3 (residues 153 to 176) and N-terminal region of α-1 (residues 1 to 17) were predicted to be transmembrane helices (40). These two transmembrane helices could form the channel in the lipid bilayer (Fig. 4D) on the basis of the location of the gp14 structure (Fig. 4B), although the channel was not well resolved in our structure of empty T7 in complex with LPS (Fig. 4A). However, because the copy number of gp14 in T7 was estimated to be 10.7 ± 2.4 (39), the locations and functions of the remaining copies of gp14 remain to be determined.

Discussion

Our structures show that E. coli rough LPS interacts with the T7 phage and triggers the irreversible structural changes in the core and portal–tail complex. The structural analyses of the mature phage and the full and empty phage particles in complex with the LPS reveal the structural changes in the T7 phage from the mature state to the genome-ejected state. After stable adsorption, the tail nozzle opens upon interacting with the outer membrane and LPS (21). A signal must be transmitted from the nozzle for the ejection of the core proteins and genome. Although the adaptor protein and the adjacent portal stem and clip domains, which are located between the nozzle and core, remain structurally unchanged after the interaction with the LPS, the gp11 and the portal proteins may contribute to signal transmission through their temporary conformational changes. Alternatively, the signal may be transmitted through the DNA in the channel because the DNA interacts with the nozzle protein gp12 in the central channel and stacks onto the core in the mature phage.

The core proteins may be packaged in an energetic or metastable form, or internal forces associated with the packaged DNA could be used for their ejection (52). After receiving the signal, the helical barrel domain of the portal may become unstructured, allowing the passage of the gp14 molecules, which are located just around the domain in the mature T7. If the 12 tunnel loops that clamp on the DNA in the portal are partly unclamped, the partially unfolded gp14 molecules (with their helices retained) could pass through the portal channel without disturbing the portal clamp on the DNA.

The full particle in complex with the LPS could either be an intermediate or off-pathway particle. However, given the clear gp14 assignment from the structure of the empty particle, the structure of the full particle shows that the gp14 molecules extend the nozzle to the membrane, and the gp14 transmembrane helices probably form a conduit in the membrane. The transition of an internal protein into a well-defined, α-helical conduit upon host receptor interaction was also observed in other viruses. For example, calicivirus protein VP2 forms an α-helical portal in the capsid at a unique threefold axis following the receptor engagement (53), and internal protein H of the Φ-X174 bacteriophage forms an α-helical tube at a unique fivefold axis after host engagement (54). Similar to gp14, these conduits, which probably function as channels for viral genome delivery, contain a hydrophobic region or two transmembrane domains (53, 54). The hydrophobic tip of the calicivirus conduit probably penetrates into the endosomal membrane (53). Both ends of the Φ-X174 conduit contain transmembrane domains, which were suggested to anchor the assembled tubes into the inner and outer cell membranes (54). Similar types of transitions were also observed in some pore-forming toxins, which form a transmembrane helical barrel in the cytoplasmic membrane from a soluble state (55). Only six copies of gp14 were resolved in the empty T7 particle, whereas eight copies of gp14 were observed in the mature T7. The remaining two copies of gp14, which were not resolved in our structure, may play a role in extending the gp14 channel.

It is unlikely that gp15 and gp16 could exit without allowing the DNA to move out. These two core proteins were hypothesized to escort the leading end of the genome into the cytoplasm (36). The gp16 LTase domain is required for the penetration of the extended channel into the peptidoglycan layer. The gp16 molecules, which primarily comprises α-helices separated by loops, should be ejected in an unfolded conformation in order to pass through the portal channel, which is as narrow as 20 Å in diameter. The same is true for the gp15 molecules. The fact that no hydrophobic α-helix was found in the gp15 structure suggests that the middle part of the transenvelope channel in the hydrophilic periplasm, which is ∼210 Å wide (56), consists mainly of gp15. If the gp15 structure in the mature T7 phage shifts from a long and curved conformation to an unfolded, axially organized one, the eight copies of gp15 are able to form a channel spanning the periplasm. The transmembrane helices predicted in the gp16 structure suggest that gp16 could extend the gp15 channel in the hydrophobic cytoplasmic membrane.

The bend of the α-helix connecting the tunnel loop results in the opening of the portal in the empty particle. A similar open portal conformation was also observed in the DNA-ejected particle of the tailed SPP1 phage of the Siphoviridae family (57). In the DNA-ejected particle of the P68 phage, the corresponding α-helix and tunnel loop become flexible, resulting in the open conformation of the portal (5). Regarding herpesviruses, a channel valve (β-hairpin) was observed in the location corresponding to the tunnel loop in the phage portal, interacting with the DNA in the portal (14–16). Similar to the tailed phages, the last packaged DNA in the herpesvirus is ejected first during the genome delivery (17). These results suggest that the portal mechanisms for DNA ejection and retention are conserved in tailed phages and herpesviruses.

Materials and Methods

Mature T7 Sample Preparation and LPS Incubation Assay.

The mature T7 phage were prepared as described previously (13). Briefly, the T7 phage were multiplied before being precipitated with 1M NaCl and 10% polyethylene glycol8000 and purified on CsCl cushions (1.56 g/mL and 1.26 g/mL) through ultracentrifugation (197,000 g, 12 h at 10 °C). The bands were dialyzed in buffer A (50 mM Tris ⋅HCl, 10 mM MgCl₂, and 50 mM NaCl, pH 7.4) overnight. Negative staining electron microscopy showed that they were empty and full particles (mature), respectively. The mature T7 particles were collected for the LPS incubation assay.

Purified T7 reacted with rough LPS (250 µg/mL) from E. coli Serotype EH100 (Ra; Hycult Biotech, HC4046) in 37 °C water incubation for 3 h. Subsequently, the T7-LPS mixture was quickly placed in an ice water bath to stop the reaction at the end of each time point. Negative staining electron microscopy showed that almost no T7 released genome when the T7-LPS mixture was incubated in an ice water bath, even after 19 h. The mixture was further purified on 1.26 g/mL CsCl at 197,000 g for 4 h at 4 °C to remove the phages that were not bound to LPS. Finally, the T7-LPS complex band was collected and dialyzed in TNM buffer at 4 °C overnight. Subsequently, an aliquot of 3.5 μL sample was applied to a holey grid and blotted for 4 s in a chamber at 100% humidity using an FEI Vitrobot Mark IV.

Cryo-EM Imaging and Image Processing.

The mature T7, full, and empty particles in complex with LPS were imaged with an FEI Technai Arctica 200-kV electron microscope equipped with a Falcon II camera at a nominal magnification of 78,000×, corresponding to a pixel size of 1.27 Å, respectively. The electron dose of ∼25 e⁻/A² was fractionated into 30 movie frames, which were aligned and averaged to a single image (58). The astigmatism and defocus value of each image were determined by a program that we designed. The viral particles were boxed automatically using the software ETHAN (59).

The icosahedral reconstructions of the T7 mature phage were performed using our programs (60) based on the common-line algorithm (61, 62). The correct orientation of the tailed phage for each particle image was determined using our symmetry-mismatch reconstruction method (38): 1) For each particle image, we projected an initial model of the tailed phage to generate 60 projection images according to the 60 equivalent icosahedral orientations of the particle image. We searched the 60 projection images for the projection image that best matches with the particle image and then assigned the corresponding orientation to the particle image. 2) We reconstructed the tailed phage structure using the particle images according to the newly assigned orientations. We then obtained an improved model of the tailed phage. 3) We iterated steps 1 and 2 until the orientations of all particle images were mostly stabilized and the portal–tail complex structure could not be improved further.

The orientation and center parameters of each portal–tail complex in the particle image were further refined using our local refinement and reconstruction method. We used the portal–tail complex structure segmented from the whole phage structure as the model to refine the orientation and center for the portal–tail complex region in each particle image. The reconstruction and refinement were performed iteratively to improve the resolution until the orientations and centers of the portal–tail complex regions in all images were stabilized and the portal–tail complex structure could not be improved.

Following the same image processing protocol, we reconstructed the structures of the full and empty particles of the LPS-treated phages, the portal–tail complexes of the full and empty particles of the LPS-treated phages, and the core in the mature phage.

Atomic Model Building and Refinement.

Our models of the gp8, gp17, gp11, gp12, gp14, gp15, and gp16 of the empty particle in complex with the LPS and mature phage were manually built according to our cryo-EM density map using the COOT software (63). The models were refined using real-space refinement, implemented in Phenix (64). The refinement and validation statistics are shown in SI Appendix, Table S1.

Data Availability

The electron density maps and atomic coordinates have been deposited in the Electron Microscopy Data Bank (https://www.ebi.ac.uk/emdb/) and Protein Data Bank (https://www.rcsb.org) under accession codes EMD-31315–EMD-31322, 7EYB, 7EY9, 7EY8, 7EY7, and 7EY6. All other study data are included in the article and/or SI Appendix.