Three-dimensional structure of tropism-switching Bordetella bacteriophage
Associated Data
Abstract
Bacteriophage BPP-1, which infects Bordetella species, can switch its specificity by mutations to the ligand-binding surface of its major tropism-determinant protein, Mtd. This targeted mutagenesis results from the activity of a phage-encoded diversity-generating retroelement. Purified Mtd binds its receptor with low affinity, yet BPP-1 binding and infection of Bordettella cells are efficient because of high-avidity binding between phage-associated Mtd and its receptor. Here, using an integrative approach of three-dimensional (3D) structural analyses of the entire phage by cryo-electron tomography and single-prticle cryo-electron microscopy, we provide direct localization of Mtd in the phage and the structural basis of the high-avidity binding of the BPP-1 phage. Our structure shows that each BPP-1 particle has a T = 7 icosahedral head and an unusual tail apparatus consisting of a short central tail “hub,” six short tail spikes, and six extended tail fibers. Subtomographic averaging of the tail fiber maps revealed a two-lobed globular structure at the distal end of each long tail fiber. Tomographic reconstructions of immuno-gold-labeled BPP-1 directly localized Mtd to these globular structures. Finally, our icosahedral reconstruction of the BPP-1 head at 7Å resolution reveals an HK97-like major capsid protein stabilized by a smaller cementing protein. Our structure represents a unique bacteriophage reconstruction with its tail fibers and ligand-binding domains shown in relation to its tail apparatus. The localization of Mtd at the distal ends of the six tail fibers explains the high avidity binding of Mtd molecules to cell surfaces for initiation of infection.
Bacteriophage BPP-1 initiates infection of Bordetella species (the bacteria causing whooping cough) by binding the cell-surface receptor pertactin (Prn). Prn is expressed only when induced by the BvgAS two-component virulence control system (1). In the absence of Prn expression, BPP-1 can switch its infectious specificity, or tropism, by altering the major tropism determinant (mtd) gene. This gene encodes the protein responsible for receptor binding (2). Mtd variants are generated at high frequency by a diversity-generating retroelement (DGR) (1–4). This DGR creates a mutant copy of an invariant DNA template repeat (TR) by transcription and reverse transcription. The TR-derived DNA copy replaces the variable repeat (VR) that encodes the C terminus of Mtd, thus diversifying the receptor-binding protein.
The Mtd protein of BPP-1 exhibits only weak affinity for Prn. Instead, bacteriophage infection relies on the interactions between several Mtd and Prn molecules. This multivalent binding enhances BPP-1 affinity compared to purified Mtd in a receptor binding pattern characterized by avidity (4). The structures of several Mtd variants have been solved by x-ray crystallography (3). These structures show that Mtd is a multidomain protein with a C-lectin-type fold at its C terminus and that Mtd assembles into globular trimeric structures (3). The VR-encoded regions form a discreet surface on one face of the pyramid-shaped trimer. Although each tropic variant displays different VR surface residues, their Mtd structures are otherwise identical. Despite detailed structural, biochemical, and genetic analysis of isolated Mtd, our understanding of the mechanistic basis of BPP-1 avidity remained limited due to the unavailability of the full 3D structure of the phage particle.
BPP-1 belongs to the short-tailed dsDNA phage family Podoviridae (2–5). Although the 3D structures of a number of phages in the Podoviridae family have been determined by cryo-electron microscopy (cryoEM) and single particle analysis (e.g., refs. 6–9), BPP-1 is novel in its possession of long, flexible tail fibers. The extended tail fibers of phages, including those of BPP-1, have been a particular challenge to resolve as their flexibility interferes with conventional structural approaches such as cryoEM single-particle analysis and x-ray crystallography.
In this paper, we report the full 3D structure of BPP-1 phage by using an integrative approach combining cryoEM single particle analysis—in which 2D images of many “single” particles are averaged into a 3D reconstruction—and cryo-electron tomography (cryoET), in which each phage particle is reconstructed from a series of 2D images taken while the sample is progressively tilted in the electron beam. Single-particle cryoEM was used to obtain the structure of the icosahedral phage head and cryoET was employed to determine the structures of the tail hub and tail fibers. Mtd was localized to the distal globular domain of the tail fibers by using immunogold labeling and electron tomographic reconstruction. The 3D organization of phage structural components suggests a mechanism by which low-affinity Mtd supports high-avidity phage binding and DNA injection into the bacterium. The BPP-1 head structure, determined at 7-Å resolution by cryoEM single-particle icosahedral reconstruction, revealed that the BPP-1 major capsid protein (MCP) adopts a canonical HK97-like fold, highly conserved across DNA bacteriophages and human herpesviruses. Taken in its entirety, this structure of BPP-1 provides insights into the design of a previously undescribed short-tailed phage with a typical head, attached to a unique tail machinery with features for adaptive tropism-switching and high avidity binding to host receptors.
Results and Discussion
Structure of BPP-1 with Flexibly-Attached Tail Fibers Determined by cryoET.
BPP-1 phages are short-tailed Podoviridae with long tail fibers. These tail fibers are attached to the phage particle at a potentially flexible joint and their changing orientations relative to the phage capsid causes their density to disappear due to averaging during cryoEM single-particle reconstruction. Thus, we used cryoET to resolve the fiber structures by reconstructing each individual phage particle. We collected a total of five tilt series spanning an angular range of −70–70°, as shown in Movie S1 and exemplified by the five images in a tilt series in Fig. 1A. Typically, each of the reconstructed tomograms contains 40–60 individual phage particles in 3D (Movie S2). Despite noise, top-to-bottom section views of individual phage particles from these tomograms reveal densities corresponding to the phage head, tail hub, tail spikes (yellow arrow in Fig. 1B), and tail fibers (red arrow).
CryoET of the BPP-1 particles. (A) A representative tomography series with tilt range of −60° to 60°. (B) The top-to-bottom sectional views of a BPP-1 particle cut from the tomogram reconstructed from the tilt series shown in A. The thickness of each section is 7.86 Å, and the distance between two sections is 104 Å. Yellow arrow, tail spikes; red arrow, distal end of the tail fibers. (C) Averaged tail fiber. (D) Shaded surface representation of the composite BPP-1 map combining the averaged map of tail fibers (C) with that of the head and tail after averaging of 60 particles. The 11 pentons are colored in pink. The tail assembly sits on top of one vertex. (E) Side view of the BPP-1 structure showing the relative distances (in Å) between the tail, spikes, and the tail fibers.
During electron tomography, images in a tilt series are limited by the sample grid support to a useful tilt range typically from +70° and −70°, giving rise to the “missing-wedge” problem. As a result, densities in the 3D tomograms have anisotropic resolution with distorted structural features and are difficult to interpret. By averaging 60 BPP-1 particles extracted from different tomograms we eliminated most missing-wedge-related distortions and the averaged particle revealed isotropic surface protrusions on the head and a hexagonal tail (Movie S3). In addition, careful analysis of the angles between neighboring fibers in each phage particle showed a fluctuation of this angle from the expected 60° up to ±20°, indicating that the fiber is flexibly attached to the phage head (Fig. S1). Thus, a strategy of local alignment of individual fibers without the head and subsequent averaging of the aligned fibers was necessary to enhance the signal/noise ratio of the tail fiber density (Fig. 1 C and D). The diameter of the averaged fiber density varies along the length of the fiber and becomes the smallest/weakest in the region making connection to the tail spikes, suggesting this region is likely the hinge of the flexibly attached fiber (Fig. 1C).
BPP-1 particles have a capsid diameter of 680 Å (Fig. 1D). The head has partial icosahedral symmetry, characterized by 11 5-fold symmetry axes (colored pink in Fig. 1 D and E) and one unique portal vertex where the tail machinery is attached. The BPP-1 tail comprises one central cylinder-shaped tail hub, six surrounding tail spikes, and six long tail fibers, each with a globular distal end (Fig. 1D). The tail spikes are attached to the capsid shell without direct connections to the central tail hub. The six tail spikes exhibit a slight twisting pattern, with each tail spike tilting toward the central tail hub and clockwise (when viewed from the tail toward the head) toward its neighboring spike. One tail fiber originates from the head-proximal region of each tail spike. The distal end of each tail fiber terminates in a two-lobed structure (red arrow in Fig. 1B) attached by the upper region of one lobe to the rod-like section. In negative-stained 2D electron micrographs, each lobe of the globular structures appears as a separate density. This appearance may result from stain-induced distortions of the structures. Alternatively, slight flexibility might blur adjacent structures during averaging and alignment, creating the appearance of a single structure with two lobes. BPP-1 shows a number of distinctive features when compared to other short-tailed bacteriophages, such as P22 and epsilon15 (7, 8, 10) and to long-tailed phages. First, BPP-1 possesses long and flexibly attached tail fibers that are unusual in short-tailed phages and are not present in those for which high-resolution structural information is available. Second, although tail fibers are very common in long-tailed bacteriophages, they are typically attached to the central tail hubs, instead of to the tail spikes, as in BPP-1.
Localization of Mtd.
To determine the location of Mtd on the phage particle, we performed electron tomographic (ET) reconstruction of immunogold-labeled BPP-1 particles (Fig. 2). Purified BPP-1 phages were labeled with polyclonal antibodies raised against trimeric Mtd and then probed with secondary antibodies conjugated to 5-nm gold particles. EM images of negative-stained particles clearly show gold clusters localized exclusively to regions surrounding BPP-1 phage (Fig. 2 A and B). To directly visualize the specific sites of the gold particle attachment, we collected six ET tilt series of the labeled BPP-1 and performed 3D tomographic reconstructions. Immunogold labeling represents an appealing approach in electron tomography, where protein identification is difficult due to poor resolution. The immunogold particles conjugated to the secondary antibodies not only served as the specific tag for localizing Mtd protein, but also as fiducial markers for the tomography data collection and subsequent alignment. The 5-nm gold clusters are clearly visible under the magnification used (×31,000). Compared to markers that are nonspecific and mixed with the sample before cryo-freezing, the immuno-conjugated gold markers are relatively immobile during stage tilting, ameliorating a potential problem of traditional fiducial gold markers. Density slices from the 3D tomograms clearly show that the gold particles are exclusively localized at the distal end of the tail fibers, indicating the location of the Mtd in the globular lobes (Fig. 2C).
Localization of mtd to the distal ends of the tail fibers by 3D reconstruction of immunogold-labeled particles. (A and B) Images of immunogold-labeled BPP-1 particles. The phage particles are incubated with rat anti-mtd antibody, followed by incubation with anti-rat, 5-nm gold-conjugated secondary antibodies. The gold particles are shown specifically attached to the tail fibers with very clean background (average <1 gold per μm2). (C) Section from a 3D electron tomogram reconstructed from the particle shown in B, demonstrating the location of mtd. The arrow points to the two gold particles that are localized to the distal end densities of the same tail fiber. (D) Fitting of the crystal structure of the Mtd trimer (PDB ID code 1YU4) to the tail fiber distal end densities. Two Mtd trimers are modeled to the bottom portion of the distal end densities. Each trimer fits into one lobe with the variable region facing down.
Mtd exists as trimers in solution and its crystal structure [Protein Data Bank (PDB) ID code: 1YU4] shows extensive interactions among the three subunits (3), implying that Mtd may be trimeric in BPP-1 particles. In fact, Mtd trimers are similar in size and shape to the globular structures labeled during immunogold tomography (Fig. 2D). We fitted the Mtd trimer crystal structure into the bottom portion of each of the two lobes, with variable regions of Mtd oriented toward the bottom surface, to be accessible for receptor binding. In this model, Mtd is attached to the tail fibers at its N-terminal domain. Regions of Mtd encoded by VRs make no contact with the rest of the phage and, therefore, phages assembled from Mtd variants will be identical in their overall organization and structure. This fitting is consistent with the role of Mtd in BPP-1 tropism determination as discussed below.
Conservation of the BPP-1 Head.
A single-particle cryoEM reconstruction approach was used to determine the shapes and molecular interactions between BPP-1 head proteins. By icosahedral averaging of ≈9,000 cryoEM particle images, we reconstructed a 7-Å resolution structure of the BPP-1 head. Each icosahedrally averaged head consists of 12 pentons and 360 capsid subunits on an icosahedral lattice with a triangulation number of 7 (T = 7) (Fig. 3B). This reconstruction allowed us to delineate most of the molecular boundaries and to discern individual subunits of the head. Two types of structural components can be identified: the MCP that makes up the shell of the BPP-1 capsid head and the cementing protein that joins neighboring MCPs. At 7-Å resolution, secondary structural elements such as α-helices and β-sheets are resolved in the subunits (Fig. 3C). A characteristic long helix (50 Å) is clearly resolved in the middle portion of the MCP subunit. The helix is flanked by an extended β-sheet that covers almost the entire length of the P (peripheral) domain. The fold of these secondary structural elements is reminiscent of a canonical fold first identified in the HK97 head protein gp5 (Fig. 3C) (11) and, subsequently, found in the major capsid proteins of many DNA viruses, including epsilon15 (6, 7), P22 (12), and herpesvirus capsid (13). A BLAST search for BPP-1 protein homologs identified that Bbp17 has 35% sequence identify to HK97 gp5. This sequence identity is exceptionally high considering the rather widely diverged cell attachment machinery between the two species. Indeed, Bbp17 has a molecular mass of 36.5 kDa, which is similar to that of HK97 gp5, and Bbp17 is the most abundant protein in purified phage preparations (data not shown), suggesting that it is likely to be the MCP of BPP-1. Each BPP-1 cementing protein (colored in red in Fig. 3D) appears as an elongated globular structure. Two cementing protein molecules situated at each 2-fold axis of the icosohedral MCP lattice form a pair of molecular clamps joining two underlying MCP molecules (Fig. 3 B and D). Extensive interactions are seen between the lower region of the cementing protein and the extended β-sheet in the MCP P domain (Fig. 4D). These interactions provide intercapsomer linkages that are likely to be critical for capsid stability, particularly when confronting pressure exerted by packaged DNA. Unlike HK97, which uses covalently crosslinked MCPs at the end of capsid maturation to stabilize the mature structure (11), BPP-1 uses the extra cementing proteins to achieve the same effect. A similar stabilizing mechanism has been seen in several other short-tailed bacteriophages, including epsilon15 (7) (6) and bacteriophage L (14), and in long-tailed bacteriophages λ (15) and T4 (16–18).
Subnanometer resolution reconstruction of the head revealed the fold of the MCP and the accessory stabilizing or cementing proteins. (A) A focal pair of cryoEM images of the BPP-1 particles embedded in vitreous ice. (Upper) The close-to-focus image that has high resolution information of the particles. (Lower) The far-from-focus image, which has high contrast for orientation determination. (B) Shaded surface view of the head. Three 6-fold axes and one 3-fold axis in the front are labeled. The dashed line box represents two MCP and two cementing protein from two neighboring hexons. (C) Fold of the MCP. One MCP monomer is extracted and fitted to the crystal structure of the HK97 MCP gp5. A long α-helix was identified at the P domain. Two short helices and some β-sheets are also visible and fit well with similar secondary structure elements present in the gp5 atomic structure. Domains indicated include axial (A) domain, peripheral (P) domain, and elongated (E) loop. (D) The interaction between MCP and cementing protein. The cementing protein is a elongated globular density that localized at the 2-fold axis as a dimer. The cementing proteins interact with the E domain and end of P domain of its adjacent MCPs to provide extra forces for the integrity of the capsid. In HK97, capsid stability is achieved through a completely different mechanism involving chemical bonds formed by the E loop. Blue, MCP; red: cementing protein.
Improved Mtd avidity during BPP-1 attachment to host cell. (A) A proposed model of attachment. (1) When a phage particle contacts a host bacteria, Mtd molecules on one of the six tail fiber ends first reaches the cell membrane receptor, retaining the particle near the cell surface. (2) Cooperative attachment. The initial binding of Mtd molecules on one tail fiber increases the probability of random collisions leading to the binding of other Mtd molecules on the same particle, thus providing a phage particle an increased strength of attachment to a cell surface. (3) Binding of multiple tail fibers to the cellular membrane receptors orients the phage particles toward the cell surface. Lateral movement of the Mtd-bound receptors may pull the particle toward the cell surface, providing possible mechanical support for the tail insertion. Because the height difference between the tail and the surrounding tail spikes is smaller than the average thickness of the periplasm, the tail spikes have to undergo a conformational change to alleviate the space hindrance. (4) Penetration of the tail through the periplasm to the inner membrane triggers the release of DNA into the cell. The attachment of tail spikes to the capsid shell suggest that conformational changes of the tail spikes during tail insertion might be the signal passed to the portal complex via the capsid shell to trigger DNA release. (B) Thin section EM image of BPP-1 infecting Bordetella bacteria, showing BPP-1 attachment spread uniformly across the bacterial surface.
Implications for Cooperative, Mtd-Mediated Phage Attachment and Entry.
The 3D structure of BPP-1 phage and the localization of Mtd to the distal end of flexibly attached tail fibers provide insight into several intriguing observations of BPP-1 infection (Fig. 4A). A distinctive feature of BPP is that Mtd variants have the potential to bind many different receptors. These variants share low individual binding affinities for their bacterial receptors, but enable efficient phage infection due to the high avidity of whole phage particles (19). In the initiation of infection, the receptor-binding surface of one Mtd trimer (comprising 3 VR-encoded regions) first recognizes a cellular receptor on the outer membrane. The distal, outward-facing orientation of the VR-encoded receptor binding surfaces suggests that this initial binding would align all five unbound Mtds with the bacterial outer membrane and, thereby, increase the probability of additional receptor-binding events. These secondary binding events would account for phage avidity. The positioning of Mtd trimers at up to 440Å radially from the central tail would allow the phage to search relatively wide swaths of cell surface for additional receptor molecules, which are likely arranged nonsymetrically on the surface, and may also allow new phage variants to use low-density cell surface molecules as partners during avidity-mediated attachment. The flexibility of the tail fibers is required for efficient scanning of the bacterial outer membrane; however, we observe that tail fiber flexibility is partially restrained, with receptor-binding surfaces always positioned on the “downward” face of each fiber, roughly parallel to the tail hub. Binding of multiple tail fibers to the host cell, along with this constrained flexibility, will position the tail hub perpendicular to the host cell surface and facilitate the interactions with the cell membrane that trigger DNA release.
For DNA injection to occur, the tip of the central tail hub must reach first the outer and then the inner membrane. Crystal structures show that Prn-binding does not trigger any significant conformational change in Mtd (19). As shown in Fig. 4A, when phage particles are at rest in solution the average angle between the vertical axis of the tail and the extended tail fibers is 66°. The tail fibers occlude the central tail, extending 75 Å vertically below the tail tip and potentially blocking outer membrane contact at most sites, although perhaps not at the curved surfaces of the bacterial poles. Exposure of the central tail hub would require a tail fiber-hub angle of at least 78°.
Indeed, images of phage-infecting bacteria (Fig. 4B) clearly show that phages attach not just at the curved poles, but along the full length of the bacteria. Intriguingly, purified BPP-1 Mtd trimers bind primarily to the cell poles, demonstrating that the receptor Prn—like many auto-exported proteins—is asymmetrically distributed after secretion through the bacterial outer membrane (20). The ability of BPP-1 phage to bind efficiently, even where free Mtd binds weakly, suggests the effectiveness of high-avidity binding. The slight enrichment for polar localization of phages may reflect a particularly strong Prn gradient, or a receptor-independent preference for polar localization similar to that observed for Escherichia coli phages (21).
In the tomographic reconstruction of unbound phage particles, the average height difference between the tail and the six surrounding tail spikes is 60 Å (Fig. 1E). In Gram-negative bacteria, the average cell membrane thickness is ≈100 Å. In the process of the tail insertion, the tail spikes—obstructed by the outer membrane—could prevent the tip of the central tail hub from reaching the inner membrane. Analysis of the tail spikes’ densities showed that the spikes are cylindrical structures that interact with head shell proteins, with no detectable interactions with the central tail (Fig. 1D). The tail spikes bend inward toward the tail, and show a slight clockwise propelling pattern from the bottom view (Fig. 1D). To expose an additional 40 Å of central tail during insertion, either the tail spikes or the central tail must undergo a conformational change. Based on the novel propelling pattern of the tail spikes in the unbound form, we hypothesize that after attachment the tail spikes bend a further 30 degrees inward toward the tail. This conformational change could be transmitted to the capsid and the portal complex, initiating DNA release into the cytoplasm as the exposed tail penetrates the inner membrane.
Compared to other phages for which detailed structural information is available, Mtd-bearing tail fibers are novel, adaptable structures for mediating bacteriophage attachment. Just as phage species use diverse structures to bind the bacterial surface, they rely on a range of mechanisms to deliver DNA from their capsid into the bacterial cytoplasm. BPP-1 belongs to the short-tailed Podoviridae family, whose members include T7, epsilon15, and P22. The T7 capsid contains a large protein core, which is ejected along with DNA, and enables penetration of the bacterial inner membrane. It has been hypothesized that core proteins lengthen the T7 tail, which otherwise would not reach the inner membrane. Epsilon15 has a core similar to T7, but P22 has a much less massive core. BPP-1 does not have a visible protein core, so its DNA delivery mechanism is likely to differ from those phages with a capsid core.
In summary, our BPP-1 structure reveals both ancient and recently evolved features of viral structural adaptation. The capsid shell into which DNA is packaged is composed of subunits containing a highly conserved capsid fold found in a number of other phage types. In contrast, the tail structure for host attachment and infection is distinctive, even from other viruses in the Podoviridae. These adaptations accommodate the ongoing evolution of the host recognition protein, Mtd, which can adapt to the transfomations of its ever-changing host’s surface. Mutations in the capsid protein are relatively uncommon and are likely to disrupt the conserved head structure. The tail apparatus of BPP-1 has evolved a novel structural solution that tolerates the dense mutagenesis created by the phage DGR and harnesses the resulting variation to increase the adaptability and, thus, the fitness and potential survival of its progeny phages. Conservation of head structure, assembly, and genome packaging, coupled with adaptation to optimize host infection, represents an emerging theme in viral evolution. The Bordetella phage system represents a particularly elegant and compact system to study the structural basis for host tropism, receptor-binding specificity and avidity and, possibly, the signaling mechanism that leads to DNA release from phage capsids.
Materials and Methods
Data Acquisition.
We obtained cryoET tilt series and cryoEM focal pair images of the BPP-1 phages in a 300-kV transmission cryo-electron microscope, the Polara G2 F30 from FEI (Hillsboro, OR), equipped with a 4k × 4k pixel TVIPS CCD camera. For cryoEM, 3 μL of purified sample was applied to 3.5/1.0-μm-hole type carbon-coated holey grids (Quantifoil Micro Tools GmbH, Jena, Germany) and quickly frozen in liquid-nitrogen-cooled liquid ethane, so that viral particles were suspended in a thin layer of vitreous ice spanning the holes of the supporting film. Focal-pair images for single-particle reconstructions were recorded with an electron dosage of ≈19–21 electrons/Å2 per micrograph and at a magnification of ×93,000, corresponding to an effective pixel size of 0.97 Å per pixel at the specimen level. The electron beam was under focused, with a 1.5-μm difference between the close-to-focus (≈1.0 μm under focus) and far-from-focus (≈2.5 μm under focus) images. For cryoET, 200 mesh grids were used so that the edge of the holes would not block the beam when the stage is tilted in the range of −70–70°. Computer-controlled tilting, sample tracking, and image recording were performed by using the EMMENU3.0 image acquisition software package (TVIPS, Germany), using a 2° incremental interval and a dosage of ≈1 electron/Å2. The accumulative dosage for each entire tilt series is ≈100 electrons/Å2. The magnification used is ×38,180, corresponding to a pixel size of 3.93 Å per pixel at the specimen level.
3D Tomography Reconstruction.
To process the cryoET tilt series, we used Protomo, a program package for marker-free alignment (22). Briefly, a tilt series is coarsely aligned in a graphical user interface. The preliminary aligned image stack was further refined with a sequential alignment schema. When first started, the 0° tilting image was used as the reference to align the two nearest neighbors in the tilt series; then the reference is updated by combining all images already aligned; and geometry fitting further refines geometric parameters, including the direction of the tilt axis, an off-set angle to the goniometer readings, and the out-of-plane angle between the specimen and the stage. Approximately 10 rounds of refinement were carried out for each tilt series until no significant improvement in the alignment parameters were detectable. The final 3D map was then computed by weighted back projection algorithm (23).
3D Alignment and Averaging of cryoET Maps.
To overcome the missing wedge problem and improve overall quality, we aligned and averaged multiple particles. Sixty BPP-1 particles are extracted from the 3D tomogram and manually aligned to a common model. The tail assembly was used as a feature for the initial rough alignment based on visual inspection and manual adjustment. In subsequent 3D cross-correlation, restricted vertical and unrestricted in-plane rotation and local translation are allowed to refine the alignment. The sixty particles are averaged after alignment to the model to generate a final map of the capsid, the tail, and the tail spikes. For the structure of the tail fibers, local 3D averaging was performed to improve the signal-to-noise ratio of the 3D structure. First, three extracted tail fibers are picked out and roughly aligned to each other manually. Then, the three tail fibers are averaged to create an initial 3D average model to be used as a template for further iterative computational alignment. In the iterative local alignment, one tail fiber is aligned to the current model by 3D cross-correlation. Then, the aligned fiber is added to the current best model with appropriate weighting to generate a new model for the next iteration of alignment. We continued this process until 20 tail fibers were aligned and no further improvement of the structure can be seen by adding more extracted tail fibers. The tail fibers are merged with the averaged phage head to make a composite 3D map. The 3D maps were visualized by using Amira (Mercury Computer Systems, Chelmsford, MA).
Single-Particle Reconstruction.
For single-particle reconstructions with icosahedral symmetry, we used the IMIRS software package (24, 25). First, ≈12,000 particle images were boxed out from the far-from-focus micrographs of 1,406 focal pairs, and their center and orientation parameters were refined to ≈15 Å resolution by using a combination of cross-common-lines method and project matching. The orientation parameters were then applied to the corresponding particle images from the close-to-focus micrographs for iterative orientation refinement. Before merging of particles for 3D reconstruction, the Fourier transform values of individual images were corrected for the contrast transfer function as described (26). A decay factor of 70 Å2 was used to enhance the resolution structure features of the final map. Approximately 9,000 particles were chosen for the final reconstructed map. The effective resolution of the final map is evaluated to be 7 Å based on the criterion of the 0.5 Fourier shell correlation coefficient between two independent reconstructions The 3D map was first segmented by using Amira and displayed by using UCSF Chimera (27).
Acknowledgments
We thank Ivo Atanasov, Xuekui Yu, and Peng Ge for assistance in single-particle cryoEM imaging and Ms. Xiaorui Zhang for help in graphics. This research was supported in part by National Institutes of Health Grants R01GM071940; (to Z.H.Z.), R01AI071204;, and R01AI072504 (to J.F.M.).
Footnotes
The authors declare no conflict of interest.
Data deposition: The cryoET and cryoEM density maps have been deposited in the EBI under accession codes EMD-1619 and EMD-1620, respectively.
This article contains supporting information online at www.pnas.org/cgi/content/full/0915008107/DCSupplemental.