Visualizing Adsorption of Cyanophage P-SSP7 onto Marine Prochlorococcus

Marine cyanobacteria perform roughly a quarter of global carbon fixation, and cyanophages that infect them liberate some of this carbon during infection and cell lysis. Studies of the cyanobacterium Prochlorococcus MED4 and its associated cyanophage P-SSP7 have revealed complex gene expression dynamics once infection has begun, but the initial cyanophage-host interactions remain poorly understood. Here, we used single particle cryo-electron tomography (cryo-ET) to investigate cyanophage-host interactions in this model system, based on 170 cyanophage-to-host adsorption events. Subtomogram classification and averaging revealed three main conformations characterized by different angles between the phage tail and the cell surface. Namely, phage tails were (i) parallel to, (ii) ~45 degrees to, or (iii) perpendicular to the cell surface. Furthermore, different conformations of phage tail fibers correlated with the aforementioned orientations of the tails. We also observed density beyond the tail tip in vertically-oriented phages that had penetrated the cell wall, capturing the final stage of adsorption. Together, our data provide a quantitative characterization of the orientation of phages as they adsorb onto cells, and suggest that cyanophages that abut their cellular targets are only transiently in the “perpendicular” orientation required for successful infection.

icosahedral lattice. One of the twelve capsid vertices harbors a short (about 230 Å in length), non-contractile tail hub that is surrounded by six thin tail fibers. The capsid protein of P-SSP7 is structurally similar to that of other phages, including podoviruses [T7 25 , ε 15 26,27 , P22 28 , HK97 29 and Syn5 30 ] and myoviruses [T4 29 ] -all of which have the same protein fold 27,31 . In contrast, the structures of the tail hub and tail fibers are quite different among these phages, likely reflecting strong evolutionary pressure to optimize interactions with their hosts. In Salmonella phages P22 and ε 15, for example, six trimeric tailspikes (which differ from tail fibers in that they are more rigid) are bound in grooves of the tail hub; each has head binding, hinge and receptor binding domains, and the tips of the tailspikes point away from the capsid 32 . In phages T7 and P-SSP7, the tail fibers are highly flexible; however, their tips appear to bind to the capsid instead of pointing away from it 33 . These differences among phages suggest that their corresponding adsorption mechanisms onto their hosts may also differ. It has been suggested that tail fiber orientation may affect not only host recognition, but also the interactions between the tail fiber, the adaptor, and the portal, which are necessary to trigger internal core protein disassembly and DNA release 24 . Cryo-ET and subtomogram averaging have been increasingly used to study the interactions between phages and host cells in situ; for example, for Salmonella podovirus ε 15 34 , Escherichia coli podovirus T7 33 , and Synechococcus podovirus Syn5 35 . For ε 15, no structural rearrangement of the tail fibers was seen upon cell surface attachment, nor did the phages appear to change their orientation with respect to the cell membrane. In contrast, phage T7 tail fibers adopt various conformations during adsorption. Although the adsorption structure of biochemically purified Syn5 has been reported, a systematic study on whether its tail fibers change in conformation during phage adsorption onto the host bacterium is lacking. For Prochlorococcus podovirus P-SSP7, the focus of this study, our previous single particle cryo-EM study comparing DNA-containing (full) and DNA-released (empty) phages revealed extensive tail fiber rearrangements between these two states 24 . This suggests that infection of host cells by P-SSP7 might require such rearrangements. Here, we used cryo-ET to quantitatively characterize phage orientations with respect to the cell membrane and tail fiber conformations on a per-particle basis. Indeed, we observed tail fiber rearrangements in situ during P-SSP7 adsorption onto cyanobacteria.

Results and Discussion
Experimental overview and initial screenings. Cryo-ET was applied to samples collected from timecourse experiments designed to capture different stages in the adsorption cycle of P-SSP7 onto Prochlorococcus. Briefly, Prochlorococcus cells in late exponential phase were concentrated to ~10 9 cells/ml by centrifugation and mixed with ~40-fold the number of phages. The samples were imaged from P-SSP7 adsorption through cell lysis, up to ~386 min post-infection (m.p.i., see Methods).
Preliminarily screening of ~90+ grids frozen at different post-infection time points revealed almost no intact cells after 7 h post-infection. Grids that were significantly bent, devoid of sample, or that yielded thick ice were discarded. As a result, 64 grids were suitable for data analysis (Supplementary Table). Among these, the 16 best ones were selected for tiltseries collection based on ice thickness, phage-cell concentration, and infection time diversity. In the end, 70 suitable tiltseries were reconstructed into tomograms, corresponding to time points between 11 and 386 m.p.i. Many biological features are visible in our tomograms, as shown in Fig. 1a and annotated with different colors in Fig. 1b (also, see Movies S1-3). Prominent features include phages full of DNA (Fig. 2a,d,g,h), empty phages (Fig. 2b), and cellular features such as the cell wall and plasma membrane, thylakoid,  Fig. 1). An analysis of our time-course of tomograms ( Fig. 3) showed that phage particles were not detectably attached to the cell surface until ~23 m.p.i. Furthermore, phages did not appear to release their genome into their host until ~86 m.p.i., when empty capsids (i.e., phages devoid of DNA) were first observed. Of note, cells began to lyse also at ~86 m.p.i. Our imaging-based observations are consistent with a prior characterization of this system using PCR-based approaches 16 , except that we observed lysis at even earlier infection time points. Classifying and quantifying cyanophage particle orientations during adsorption. To better understand the interactions between MED4 and P-SSP7, we further examined our 16 reconstructed tomograms. Briefly, 178 manually selected subtomograms of "full" P-SSP7 particles close to the cell were computationally extracted (see Methods). We classified them according to the angles of their tails with respect to the cell surface (Fig. 4a), and then computationally aligned and averaged the particles within each class (see Methods). The angle of interaction between phage and host cell was not uniform across particles. Instead, three main orientations were observed, corresponding to angles of ~0° ("parallel"; n = 70), ~45° ("leaning"; n = 44) and ~90° ("standing"; n = 26) (Fig. 4b) between the phage tail and the cell surface. The remaining full capsids (n = 30) were not close enough to interact with the cells, and were designated as "free". Of note, eight subtomograms were not included in the analysis due to orientation ambiguity. If the distribution of phage orientations is reflective of their prevalence in natural systems, we might infer the following: First, phages on the host surface appear to spend considerable time with their tail hubs parallel to the cell surface, perhaps "exploring" the cell surface using tail fibers to identify cellular receptors. For T4-like phages, this is a well-known phenomenon mediated by the reversible binding of long tail fibers and the irreversible binding of short tail fibers 36 . For T7-like phages, however, phage tails and tail fibers are less prominent 37 . The idea that phages "walk" along the cell surface has also been suggested for phage T7 38 , but has not been quantitatively demonstrated. Second, of the three major orientations observed, only the "standing" class represents an adsorbed phage poised for successful DNA injection, and this class comprised only 19% of observed full-capsid phages. This suggests that either few cyanophage-host contacts result in successful adsorption and genome injection, or that "infection competent" conformations (i.e., "standing" particles) are short-lived. Interestingly, tail fibers might interact asymmetrically with the cell surface during adsorption, as seen for T7 33 . For example, although we resolve clear density to account for all fibers at the base of the tail without imposing symmetry (Fig. 5a, Supplementary Fig. 2), the density of some fibers is stronger and resolved over a longer range than that of others. Indeed, strong pseudo-six-fold symmetry is evident in self-correlation plots (Fig. 5b, see Methods).

Classifying and quantifying cyanophage tail fiber orientations during adsorption. Previous
cryo-EM studies showed that tail fiber conformation differs between biochemically purified full and empty P-SSP7 phage particles, suggesting that it might serve a specific function during adsorption 24 . Here, we examined the tails of 170 full phages by extracting them from phage subtomograms and classifying them according to their characteristic Eigen images (Supplementary Fig. 3a; see Methods). We discovered three major groups (Fig. 6, upplementary Fig. 3b): (i) "folded fibers" (n = 70, where fibers near the tail hub ran along the capsid surface and the tail hub showed a smooth cone shape; (ii) "extended fibers" (n = 48), where the proximal section of the fibers extended horizontally, the tail hub was slightly shorter than that with folded fibers, and the tip of the tail hub showed a round shape; and (iii) "intermediate fibers" (n = 52), likely encompassing an average of folded and extended fibers within individual phages. Taken together, tail orientation and phage adsorption data suggest that tail fiber conformation is likely to have a biological function. Specifically, for free phages that are proximal to but not interacting with the cell surface, only 17% of tail fibers were in the extended position. Similarly, only 16% of the "parallel'' phages showed extended fibers. In contrast, 30% of the "leaning'' and 73% of the "standing'' phages showed tail fibers in the extended conformation (Fig. 6b). Thus, a majority of "parallel'' and "leaning'' phages exhibited folded fibers, while the majority of "standing'' phages had extended fibers. The fiber-base conformations we found here are consistent with those found in our previous single particle analysis study 24 . Namely, the basal region of tail fibers was in a "folded" conformation in DNA-full phages. In our P-SSP7 data here the full length of the fibers is not visible because of their structural flexibility and the limited resolution of cryo-ET. Nonetheless we successfully resolved conformational differences at the fiber-base regions (i.e., where fibers are anchored to the tail hub) of phages at different stages of adsorption.
Our data, particularly our "standing" phage average where extended fibers are seen interacting with the cell surface, suggest that tail fibers are essential to phage adsorption onto their host. Indeed, temperature-sensitive mutants of gp17 in bacteriophage T3 yielded fiberless, non-infectious phages at high temperature 39 . Similarly, in bacteriophage T7, a gp7.3 mutant with compromised tail fibers yielded full particles that failed to adsorb onto host cells 40 . A fiberless P-SSP7 mutant has not been studied; however, tail fiber conformations seem to exert the same function in adsorption as those in other bacteriophages.
Our findings suggest a role for tail fiber extension in aiding the adsorption process. On the other hand, in the absence of the host, folded tail fibers are likely adhered to the phages for protection. Tail fiber conformational changes have also been observed for phage T7, where the number of tail fibers bound to the host cell surface gradually increases during adsorption 33 , but not for phage ε 15 34 . Here, our results suggest that conformational changes do occur in P-SSP7 tail fibers during adsorption. Formation of tail extension for DNA injection. Although cyanophage P-SSP7 is a T7-like phage that lacks a contractile sheath (as is the case for T4-like phages), comparative genomics suggests that the P-SSP7 genome encodes homologs to some T7 phage tail proteins. By analogy to T7, it was hypothesized that the P-SSP7 tail releases core proteins (gp14/15/16) just prior to DNA injection to digest the bacterial envelope and build a long, narrow, extensible channel from the tip of the tail to protect viral DNA during transport into the cell 13,16 . Our cryo-ET observations here are consistent with these genomics-inferred hypotheses. For example, in plasmolysed areas, we observed a tail tube extending from several standing phages, ~500 Å or longer and half the width of the tail, which pierced through the cell wall and cell membrane (Fig. 7). Such tubular features were also observed for ε 15, T7 and syn5 phages during infection of Salmonella, Escherichia coli and Synechococcus cells, respectively [33][34][35] . These observations are consistent with the hypothesis, derived from comparative genomics, that P-SSP7 has an extensible tail homologous to that of T7 33 . Towards a mechanistic model of cyanophage adsorption. Based on our per-particle observations, we propose the following conceptual model to describe the adsorption process of phage P-SSP7 onto host Prochlorococcus MED4 (Fig. 8): Free phages, which have not yet interacted with the bacterial cell wall, approach the cell with folded fibers and then adhere to the host cell surface with the tail hub almost parallel or at an angle to the cell surface: the "parallel" and "leaning" states being indicative of pre-infection. Once adhered, and after a presumed "walking" stage where tail fibers search for receptors on the cell surface, some tail fibers appear to attach more firmly and to extend horizontally, enabling the phage to "stand". These conformational changes in the tail fibers might then trigger changes in the tail-portal structure, allowing the tail to firmly attach to the cell surface Lower panel: Number of phage particles with a particular angle of the tail with respect to the cell surface, with vertical (perpendicular) orientation considered as 90°. 70 phages (50%) are classified as "parallel" (left), in which the tail hubs are almost parallel to the cell surface (θ ≈ 0°) and do not appear to directly interact with the cell wall (outlined in green). 44 phages (31%) are classified as "leaning" (middle), in which the tail is at an angle to the cell surface (θ ≈ 45°) and partly interacts with it (outlined in yellow). 26 phages (19%) are classified as "standing" (right), in which the tail is vertically connected to (θ ≈ 90°) and completely interacts with the cell surface (outlined in purple). * Horizontally extended tail fibers are visible in the subtomogram average of "standing" phages. OM denotes the outer membrane of the host cell. Scale bar is 20 nm. and release internal phage proteins to the cell membrane. Finally, this fully infection-competent "standing" state leads to the release of viral DNA into the cell. This multi-stage process may serve to ensure that the phage is in the proper orientation to productively deliver its DNA into the host. Furthermore, phages might be able to use their tail fibers to assess whether or not a cell is able to host a productive infection. Similar hypotheses have been suggested for soil mycophages (e.g., "waking proteins") 41 , and uncultured phages from the oceans 42 . Conventional (plastic-embedded, stained and/or sectioned) observations using electron microscopy showed that all the classic T1 to T7 phages adsorb onto 200 to 400 specific positions (adhesion zones) on the cell surface, presumably where the wall and protoplasmic membrane adhere to each other 43 . In contrast, we observed only a few cases per cell where an infecting phage localized to regions where the space between the membrane and the wall was narrowed ( Supplementary Fig. 4). Furthermore, no such adhesion zones were observed with phage T7 infecting E. coli when samples were prepared using the same cryo-ET protocol 33 . Together, these observations suggest that abundant "adhesion zones" in previous studies using conventional electron microscopy might correspond to artifacts stemming from chemical fixation.

Conclusions
Our cryo-ET experiments, including subtomogram classification and averaging, provide a first look in situ at the direct interactions between P-SSP7 and the marine cyanophage Prochlorococcus MED4. Our observations suggest that productive genome injection after the adsorption of P-SSP7 onto MED4 is either rare or transient. Furthermore, adsorption itself is shown to be a multi-stage process. Lastly, our data are consistent with the genomics-based hypothesis of the formation of an extensible tube preceding infection. While cryo-ET is the best method to visualize interactions between macromolecules in situ 44 , it remains challenging to establish a quantitative understanding of such nano-scale events given the relatively low-throughput of this imaging modality. Nevertheless, our analyses of 170 near-cell phage particles paint a picture of phage-host interactions that will help guide future experiments across diverse phage-host model systems.

Methods
Preparation of P-SSP7 phage. Phages were prepared as previously described 13 . P-SSP7 particles were propagated in Prochlorococcus MED4. The particles were then precipitated with polyethylene glycol 8000, purified on a cesium chloride step gradient (steps were ρ = 1.30, 1.40, 1.50, and 1.65), spun at 104,000× g for 2 h at 4 °C, and dialyzed against a buffer containing 100 mM Tris-HCl (pH 7.5), 100 mM MgSO 4 , and 30 mM NaCl. Just before use, the phages were ultracentrifuged at 104,000x g for 1.5 h at 4 °C, and the pellets were re-suspended in buffer to yield a concentration of about 10 12 particles/ml.  In late exponential phase, these cells were concentrated by centrifugation to ~10 9 cells/ml. For infection, 100 μ l of cells (~10 9 cells/ml) were mixed with 400 μ l of phage (~10 10 phage/ml), to achieve a multiplicity of infection (MOI) of ~40 virus particles per cell. Then, 5 μ l aliquots of the cell-phage mixture were taken at 2-3 min intervals from 4 to 128 min, and from 381 to 403 min post infection (see Supplementary Table). All aliquots were mixed with 1 μ l of 150 Å colloidal gold (Catalog No. 25-40, Electron Microscopy Sciences), applied to previously glow-discharged R3.5/1 copper Quantifoil grids (Quantifoil Micro Tools GmbH), and rapidly frozen in liquid ethane using a Leica plunger (EM-CPC, Leica Microsystems) after blotting with Whatman No. 1 filter paper. Grids of cell-phage mixture were frozen and stored in liquid nitrogen.  Cryo-electron tomography. Imaging was done using a JEM3200FSC electron microscope (JEOL Inc., Japan) equipped with a field emission electron source operated at 300 kV and an in-column energy filter (slit width: 20 eV). Condenser and objective apertures were set at 50 and 60 μ m respectively. The specimens were kept at − 170 °C with liquid nitrogen. Tilt series images were collected using SerialEM automated data acquisition software 45 in a range of ± 62° at 2° increments. The images were recorded on a 4k × 4k pixels slow scan CCD camera (Gatan Inc.) at a nominal magnification of × 20 K, which resulted in an imaging resolution of ~5.4 Å/pixel. The total electron dose on the specimen per tilt series was kept under 80 electrons/Å 2 to minimize radiation damage. The tilt series were aligned using gold fiducials, and tomograms were reconstructed using filtered back projection or SIRT in IMOD, as previously described 46 .

Infection of MED4 cells with P-SSP7 phages. Prochlorococcus
Subtomogram classification and averaging. In this phage adsorption study, 178 subtomograms of infecting phages attached to or near the cell surface were extracted from 16 tomograms of phage-infected cells. Subsequent post-tomographic image processing was performed using EMAN 47 with the following steps. First, subtomograms subjected to a 20 Å low-pass filter were aligned to the single-particle icosahedral 3D map of P-SSP7 (EMDB 1713), where a tight spherical mask was used to exclude the tail. Second, these subtomograms were further aligned with an appropriate mask including the tail, and the tails were manually oriented to all point in the same direction in Z. Aligned particles were classified according to the angle the tail made with the host cell surface, and averaged in Fourier space, weighted to reduce distortions caused by the missing wedge 48 . Eight subtomograms yielding ambiguous orientations were discarded. We also performed self-symmetry analysis 49 on our subtomogram averages and multivariate statistical analysis (MSA) of the tail structures using EMAN2 software 50 to classify the fiber structures surrounding the tail hub in each adsorption class. Tail subvolumes were extracted from full phage subtomograms, 6-fold symmetry was enforced, and tails were classified into three groups by MSA ( Supplementary Fig. 3) to evaluate the conformation of the basal part of the fibers relative to the capsid surface.