Structural atlas of Pakpunavirus P7-1 reveals determinants of virion stability and genome ejection

Abstract

Bacteriophages of the Pakpunavirus genus exhibit a broad host range and potent bacteriolytic activity, making them promising candidates for clinical use. Here, we present a structural atlas of the therapeutic phage Pakpunavirus P7-1, a component of a phage cocktail targeting Pseudomonas aeruginosa that has undergone Phase 1/2 clinical trials. We determine the near-atomic structure of the extended virion and obtain a medium-resolution reconstruction of the contracted tail. Atomic models are presented for 20 structural proteins comprising the icosahedral capsid, neck, contractile tail, and baseplate. We identify six upward-pointing Short Tail Fibers that stabilize the extended sheath and six highly flexible Long Tail Fibers likely involved in host recognition. Ordered fragments of the Tape Measure Protein reveal six copies inside the tail tube, forming a 3-helix cork at the tail tip. Sheath contraction repositions the baseplate, projecting all twelve tail fibers outward, yet contraction alone is insufficient to trigger genome ejection.

Integrative structural analysis of the therapeutic phage P7-1 reveals determinants of virion stability, host adsorption, and genome ejection, guiding phage engineering

Introduction

Pseudomonas aeruginosa is a major cause of illness and death worldwide. The bacterium creates biofilms that are resistant to many antibiotics and common antimicrobials¹. This problem is exacerbated by the rapid emergence and spread of P. aeruginosa strains that have acquired antibiotic-resistant genes. P. aeruginosa infections are common in cystic fibrosis (CF), a multi-organ disease^2,3 marked by repeated bacterial infections in the lungs and ongoing inflammation. Early in life, the microbial community mainly includes Staphylococcus aureus, and P. aeruginosa becomes prevalent in later stages⁴. Strains from CF patients often mutate and develop adaptive responses to antibiotics, leading to multidrug-resistant strains⁵. Phage therapy for P. aeruginosa is increasingly seen as a promising treatment for CF-related infections^6,7. Therefore, there is growing interest in understanding known lytic P. aeruginosa phages and discovering new ones as antimicrobials against both antibiotic-sensitive and multidrug-resistant strains⁸. We previously shared a detailed structural analysis of Pseudomonas phages Pa193⁹ and Pa223¹⁰ from the Armata Pharmaceuticals collection. This study describes the complete architecture of the Pseudomonas phage P7-1, a Myoviridae phage from the Pakpunavirus genus, characterized by a long contractile tail¹¹. P7-1 was originally isolated from sewer samples in Southern California by Armata Pharmaceuticals. It has a 93.7 kb long double-stranded DNA (dsDNA) genome that includes 198 predicted ORFs (Supplementary Data 1) and 16 tRNA genes. Because of its stability and lytic activity, P7-1 was chosen for inclusion in a candidate anti-Pseudomonas phage cocktail, AP-PA02, which has undergone Phase 1/2 clinical trials for the treatment of patients with cystic fibrosis (SWARM-P.a., NCT04596319) or non-cystic fibrosis bronchiectasis (Tailwind NCT05616221)¹².

A growing number of new Pakpunaviruses have been isolated in recent years. The latest release of the International Committee on Taxonomy of Viruses (ICTV)¹³ (release v3, June 19, 2024) includes 35 members of the Pakpunavirus genus. However, over 60 Pakpunaviruses’ genomes are available in public databases and partially annotated. Typically, these phages have genomes ranging from 88 to 95 kb, about one-third larger than those of phages in the Pbunavirus genus, such as E217¹⁴ or Pa193⁹. The genome size of Pakpunaviruses is roughly half that of phage T4¹⁵, which has a genome of approximately 168.9 kbp, encodes about 300 gene products, and has been studied in much greater depth.

In this study, we use cryo-EM Single Particle Analysis (SPA), localized reconstruction¹⁶, proteomics, comparative genomics, and AlphaFold-guided structure prediction¹⁷ to generate a near-atomic atlas of P7-1 with the extended and contracted sheath. Our study identifies all the structural proteins of phage P7-1, including the multi-subunit baseplate complex. We identify features and conserved elements previously observed in other long-tailed bacteriophages, shedding light on the machinery and mechanisms of genome delivery in P. aeruginosa^18,19, and enhancing our understanding of Pakpunavirus biology.

RESULTS

Organization of Pseudomonas phage P7-1

We vitrified the P7-1 mature virion and collected a high-resolution dataset on a 300 kV Krios microscope (Table 1). Morphologically, the phage has a long contractile tail spanning more than 1200 Å and a small baseplate (Fig. 1a, Supplementary Fig. 1a), sharing some similarities with Pbunavirus phages E217¹⁴ and Pa193⁹. Surprisingly, the vast majority of vitrified virions (>99%) had an extended tail and retained the genome within the capsid, in stark contrast to E217¹⁴, which has a contracted tail in about 30% of particles after vitrification. To visualize the contracted tail, we subjected P7-1 to prolonged incubation at pH 4.5 (Table 1), which triggered tail contraction in approximately half of the particles on the grid (Fig. 1b, Supplementary Fig. 1b). However, unexpectedly, only a few particles with a contracted tail also ejected the genome, suggesting that tail contraction and genome ejection are not interdependent processes in this phage.

Table 1.

Cryo-EM data collection and refinement statistics

Data Collection Statistics		P7-1 Extended (pH 7.0)				P7-1 Contracted (pH 4.5)
Facility/Microscope		NCEF/Titan Krios				S2C2/Titan Krios
Camera		Gatan K3				Falcon 4
Magnification		X81,000				X81,000
Voltage (kV)		300				300
Electron exposure (e-/Å2)		50				50
Defocus range (um)		−1.2 to −2.4				−1.2 to −2.4
Pixel size (Å)		0.681 (physical: 1.362)				1.1
Total movies (frames/movie)		5714 (40)				10,329 (40)
Map and Model Refinement Statistics
PDB/Map entry codes		9ZBI/EMD-73989	9ZBS/EMD-73999	9ZD8/EMD-74057	9ZD7/EMD-74056	9ZDK/EMD-74065
Particles per reconstruction		12,148	12,148	10,400	10,400	7832
Map name		Capsid	Neck	Extended Baseplate	Extended Baseplate with STF	Contracted baseplate
Symmetry imposed		C5	C6	C3	C3	C6
Map resolution (Å)		3.3	3.6	3.6	4.2	4.4
FSC threshold		0.143	0.143	0.143	0.143	0.143
Models fit into density		Capsid, decoration proteins	Portal, HT-adaptor, HT-adaptor-2, sheath terminator	Tube-B/-C, sheath initiator, hub, spike, adaptor-1/2, Triplex, TMP	Tube-B/-C, sheath initiator, hub, spike, adaptor-1/2, Triplex, TMP, STF	Contracted Triplex, adaptor-2, sheath
Initial model (PDB code)		8FRS	8FVH	8FUV, 8EON	8FUV, 8EON AlphaFold3	8FVG
Model-to-Map Correlation Coefficient (mask)a		0.82	0.85	0.79	0.83	0.80
Model composition	No. of chains	9	36	63	81	30
	Non-H atoms	19,323	72,366	107,973	174,997	80,631
	Residues	2409	9006	13,938	22,895	10,527
R.M.S. deviations	Bond lengths (Å)	0.017	0.003	0.003	0.003	0.002
	Bond angles (°)	0.609	0.502	0.519	0.590	0.525
Validation	MolProbity score	1.78	1.72	2.40	2.47	2.61
	Clashscore	4.05	5.25	6.55	8.42	11.61
	Rotamer outliers (%)	5.30	1.50	7.24	8.26	11.14
Ramachandran plot	Favored (%)	97.83	95.67	94.31	95.44	93.14
	Allowed (%)	2.13	4.32	5.58	4.40	6.14
	Outliers (%)	0.04	0.01	0.11	0.17	0.72

^aCC is calculated inside a mask around the model.

Fig. 1. Pseudomonas phage P7-1 featuring an extended and contracted sheath.

Composite cryo-EM reconstructions of a P7-1 virion at neutral pH with an extended sheath and b virion after incubation at acidic pH that triggers sheath contraction.

We used approximately 12,000 particles to reconstruct the entire mature virion of P7-1 with an extended tail, using focused maps of the capsid, neck/tail, and extended baseplate (Table 1, Supplementary Figs. 2a–c). At the same time, approximately 7800 tail-contracted, head-filled virions were used to determine a medium-resolution (~4.2 Å) C6 reconstruction of the contracted tail (Table 1, Supplementary Fig. 2d). Surprisingly, these virions retained the genome in the capsid after contraction and revealed significant restructuring of the baseplate complex. Finally, using a small population of about 1300 particles that had ejected the genome onto the grid, we also obtained a near-atomic resolution reconstruction of the P7-1 empty capsid and neck (Supplementary Fig. 3a–c).

Using the five maps described above, we built atomic models for all structural proteins in P7-1. The capsid was visualized using a C5 localized reconstruction at a resolution approaching 3.3 Å (Supplementary Fig. 2a, e). The neck and tail were reconstructed using localized reconstruction, both from filled virions and empty particles, by selecting a class of particles with tails at one five-fold vertex and imposing C6 symmetry, which yielded a 3.6 Å map (Supplementary Fig. 2b, e, and Supplementary Fig. 3a). Finally, we reconstructed the baseplate from the extended tail by manually picking 10,400 baseplate particles and applying C6 and C3 symmetry to obtain reconstructions at 3.6 Å (Supplementary Fig. 2c, e). The contracted baseplate was modeled in a C6 4.2 Å density (Supplementary Fig. 2d, e).

All reconstructions exhibited strong density features, which we used to annotate 19 P7-1 Open Reading Frames (ORFs) and generate atomic models for 20 full-length gene products (gp), as illustrated in Fig. 2a–c. The structural atlas of P7-1 presented in this paper includes three capsid proteins (gp67, gp70, and gp71) (Fig. 2a), three neck proteins (gp72, gp74, and gp75), two tail proteins (gp76 and gp77) (Fig. 2b), and eleven baseplate proteins (gp78, gp79, gp83, gp84, gp85, gp86, gp87, gp88, gp89, gp90, and gp92) (Fig. 2c). The large Triplex component gp88 exists in two conformations (gp88-a and gp88-b). All atomic models built into cryo-EM maps were real-space-refined to achieve a model-to-map correlation coefficient (CC) between 0.79 and 0.85 (Table 1), except for the Long Tail Fibers (LTFs), which have weak density and were not refined (described below). Finally, we also modeled two small regions of the Tape Measure Protein (TMP) that are not shown in Fig. 2. All structural gene products annotated using cryo-EM were also identified and validated by liquid chromatography-mass spectrometry (LC-MS) of P7-1 virions (Supplementary Table 1), with one exception: gp68/gp69, which are likely located inside the capsid and are invisible by SPA.

Fig. 2. Structural atlas of phage P7-1 structural proteins.

Ribbon diagrams of 20 gene products were identified and built de novo in this study. The gene products are divided into three groups: a three capsid factors; b three neck factors and two tail factors; c eleven gene products that are part of the baseplate, encoded by ten genes. The Triplex large subunit factor gp88 exists in two conformations: conformation-1 (gp88-a) and conformation-2 (gp88-b). Fragments of the TMP (gp82) are not displayed in this figure.

The capsid and neck of Pseudomonas phage P7-1

P7-1 capsid is large, with a maximum diameter of approximately 800 Å. A 3.3 Å focused reconstruction of the P7-1 capsid’s five-fold vertex (Supplementary Fig. 2a) revealed two structural components: the major capsid protein gp71, which forms an icosahedron with a triangulation number T = 9, and the decoration protein gp70 (Figs. 1a, 2a, Table 1). Gp71 (344 aa) has a canonical HK97 fold, most similar to the major capsid protein of the flagellotropic phage YSD1²⁰, a Siphoviridae (E-value = 2.32e-15; RMSD =5.52 Å, 344 versus 354 residues in coverage). Gp70 folds into a homotrimer (Figs. 1a, 2a) that is topologically superimposable to phage YSD1 cementing protein²⁰ and overall very similar to the prototypical head decoration protein gpD from phage lambda²¹. One hundred eighty copies of the trimeric gp70 bind the capsid exterior at each local three-fold axis (Fig. 1a, b), generating a second capsid layer that decorates ~60% of the exterior surface of the P7-1 capsid. A similar pattern was observed in the Pseudomonas phages E217¹⁴ and Pa193⁹, consistent with gp70 functioning as a putative cementing protein.

We generated C12 and a C6 symmetric reconstructions of phage P7-1 unique vertex that revealed the architecture of the phage neck and the beginning of the extended tail (Supplementary Fig. 2b). The C12 map was used to identify and build de novo the portal protein gp67 and the head-to-tail (HT) adaptor gp72 (Figs. 2b, 3a, Table 1). These two proteins form a dodecameric complex at one of the 12 five-fold vertices of the capsid, replacing a penton. The portal interacts with the capsid protein (gp71) via a 12:10 symmetry-mismatched binding interface²². The P7-1 portal protein is relatively small, consisting of only 479 amino acids, compared to 722 residues in the podophage P22, for example^23,24. We built a nearly complete structure (res. 10–454) in a 3.6 Å map, while no density was visible for the first nine and last 25 amino acids. The P7-1 portal protein is most similar to the portal protein from bacteriophage G20C²⁵ (E-value = 1.67e-17), with 20.1% sequence identity and a comparable length (427 versus 479 residues). It features a classical portal protein fold consisting of crown, wing, and stem domains, but notably lacks a C-terminal barrel (Fig. 3a, b)²², possibly explaining its smaller size. Strong dsDNA density going around the portal wing perimeter is clearly visible in our reconstruction (Fig. 3a), as originally observed in the podophage P22^26,27 and later found in myophages²⁸. P7-1 HT-adaptor ring assembles co-axially under the portal, wrapping around the portal clip and stem regions²³. The HT-adaptor consists of an N-terminal 4-helix bundle core with an extended C-terminal arm that contains three α-helices wrapped around the portal clip and stem regions, resembling a bouquet wrap (Fig. 3a,b). HT-adaptor binding to the portal protein during genome packaging competes with the large terminase subunit (TerL), ending the packaging reaction²⁹. We also constructed a model of the portal:HT-adaptor complex using a 4.1 Å C6-averaged reconstruction derived from P7-1 empty particles. In both empty and DNA-filled virions, the portal remains at the same position of the 5-fold vertex and shows no signs of sinking after genome ejection (Supplementary Fig. 3a, b) or adopting a different position or organization relative to the capsid, as seen in podophages³⁰. The portal:HT-adaptor model showed a slight relaxation compared to the mature virion filled with DNA, but no other quaternary or tertiary structural changes were observed (RMSD of full versus empty HT-adaptor = 0.385 Å and 0.809 Å for the portal protein) (Supplementary Fig. 3c).

Fig. 3. Structure of phage P7-1 neck assembly.

a Overview of the P7-1 neck complex. The dodecameric portal protein gp67 (cyan) binds to twelve copies of the HT-adaptor gp72 (light green), which then binds to the HT-adaptor-2 gp74 (light yellow) and the sheath terminator gp75 (light orange). b A ribbon diagram of a portal protein monomer bound to the HT-adaptor reveals the bouquet-wrapping interaction (red frame) between the C-terminal moiety of the HT-adaptor and the portal clip and stem regions. c A cutaway view of the P7-1 neck, as shown in (a), reveals a density that extends along the entire length of the channel. d The top view of the neck channel illustrates the interaction between the neck component and the density filling the channel (likely dsDNA) across three distinct regions: portal, HT-adaptor/HT-adaptor-2, and sheath terminator.

All P7-1 factors under the HT-adaptor are hexameric³¹ and were built in a C6 symmetry map (Supplementary Fig. 2b), specifically, the HT-adaptor-2 gp74 and sheath terminator gp75 that extend the portal: HT-adaptor channel by approximately 80 Å (Fig. 3c). Notably, the neck and tail channel is filled with a density, possibly dsDNA, which is C6 averaged in our reconstruction. Additionally, the internal diameter of the P7-1 channel varies from portal to sheath terminator. The portal channel and HT-adaptor inner cavities are narrow, with contacts less than 4 Å apart between basic residues extending into the channel (e.g., portal R296; HT-adaptor R64, etc.) and the inner density, which we hypothesize to be dsDNA (Fig. 3d). In contrast, the sheath terminator cavity at its widest point is substantially larger, measuring approximately 35 Å between juxtaposed residues. Interestingly, the walls lining the channel are markedly enriched in basic residues, especially in the HT‑adaptor‑2 and sheath terminator regions (Supplementary Fig.3d). The inner density in this portion of the channel appears bulkier than that observed in the portal channel (Fig.3d). This density may correspond to dsDNA smeared due to limited contacts with the channel lumen, or possibly to the N‑terminus of the TMP, which was also identified by LC‑MS analysis (Supplementary Table 2). Unfortunately, the current resolution and map quality in this region do not allow us to distinguish between these possibilities.

P7-1 tail and baseplate proteins

The P7-1 tail consists of a repetitive set of tail tube hexamers surrounded by sheath protein subunits arranged in a helical pattern. We reconstructed P7-1 with an extended tail approximately 1,100 Å long, comprising 204 copies of tail tube gp77 (res. 1-174) and 204 copies of tail sheath gp76 (res. 1-428)³² (Figs. 1a, 2a). The tail tube has 22.5% sequence identity to the Cyanophage Pam3 tube³³ (with comparable amino acid lengths of 174 versus 142 residues) and a similar 3D fold (RMSD of 2.1 Å). The extended tail consists of an internal component comprising 34 stacks of hexameric tube protein with a lumen approximately 37 Å wide. P7-1 sheath protein gp76 (res. 1–428) has a three-domain architecture similar to E217¹⁴, but it differs from the Pam3 sheath protein, which contains only two domains (Supplementary Fig. 4a, b). N- and C-terminal extensions connect sheath proteins to neighboring subunits, allowing the protein to assemble around the tube structure in a helical arrangement. The topology of the sheath assembly and the conformational changes that occur upon contraction are similar to those described for other contractile Myoviridae^9,14 and will not be discussed here.

The separate reconstruction of the P7-1 extended baseplate, determined at resolutions of 3.6–4.5 Å (Table 1, Supplementary Fig. 2c, e), allowed us to identify and build atomic models for 11 polypeptide chains (Fig. 2c), which are present in a total of 63 copies. The baseplate can be divided into three subcomplexes (Fig. 4a–c): the cap (gp78, gp79, gp83, gp85, and gp86), adaptor proteins (gp87 and gp84), and nut (gp88 and gp89), arranged into a metastable assembly ready for contraction³⁴. The baseplate cap (Fig. 4a) remains stationary during sheath contraction and can be thought of as an extension of the repetitive tail tube. It consists of two tail tube-like subunits (tail-tube-B gp78 and tail-tube-C gp79, which are 25.5% identical), linked by the sheath initiator (gp83) and hub (gp85) subunits to a trimeric tail spike (gp86) (Fig. 2c), which seals the tail channel. The tail tube variants gp78 and 79, and the sheath initiator form hexameric rings that are coaxial to the tail tube gp77. In P7-1, the tail tube variants are slightly shorter than the main tail tube subunit gp77, consisting of 166 and 159 residues, respectively, compared to 174 residues for gp77 (Supplementary Fig. 5a, b). This difference primarily reflects shortening of the gp77 C-terminal α-helix (res. 156–164), which engages the sheath proteins in the extended tail. In the baseplate, which lacks a sheath layer, the C-terminal helix of tail tube-C (res. 154–158) shifts upward to contact the corresponding helix in tail tube-B (res. 156–162). Unlike the hexameric tail tubes, P7-1 tail hub and tail spike are trimeric (Fig. 4a). The P7-1 tail spike features a classical triple-β-helix fold³⁵, which is also found in analogous contractile ejection systems like Pa193⁹, Pam3³³, R-type pyocin³⁶, and E217¹⁴. A characteristic feature of the P7-1 cap subcomplex is the sheath initiator (gp83), which makes extensive contacts with both the six-fold symmetric tail tube-C and the three-fold symmetric hub:spike assembly (Supplementary Fig. 6a). The P7-1 sheath initiator features an extension loop (aa 87–133) that extends upward and binds the tail tube-C. Each extension loop forms 20 hydrogen bonds and 2 salt bridges at the interface between two tail tube-C, totaling 120 hydrogen bonds and 12 salt bridges in the entire cap subcomplex (Supplementary Fig. 6a).

Fig. 4. Anatomy of phage P7-1 baseplate.

a A semitransparent map of the P7-1 baseplate, overlaid with ribbon diagrams of the baseplate cap, which includes one oligomer of the tail tube gp77 and three coaxial tube-like factors: gp78 (tail tube-B), gp79 (tail tube-C), gp83 (sheath initiator), tail hub gp85, and tail spike gp86. b The baseplate adaptor proteins, gp84 (adaptor-1) and gp87 (adaptor-2). c The baseplate nut complex consists of six copies of the triplex complex (gp88-a, gp88-b, and gp89). For (a–c), the maps are shown in a side view and rotated by 90 degrees. d Ribbon diagram of P7-1 triplex complex formed by gp88-a, gp88-b, and gp89.

The second component of the P7-1 baseplate consists of two adaptor subunits, gp84 (adaptor-1, 115 residues) and gp87 (adaptor-2, 123 residues) (Fig. 2c), which bind to the outward surface of the tail hub (gp85) and sheath initiator (gp83), respectively (Fig. 4b). The adaptor subunits decorate the cap complex without making direct contact with one another, and thus do not form an oligomer like all the other factors of the cap complex. Finally, the P7-1 baseplate is built by a nut-subcomplex (Fig. 4c) that slightly expands to a maximum diameter of ~290 Å at the tail spike distal to the capsid. The nut-subcomplex comprises six copies of the Triplex complex (Fig. 4d), attached like a nut to a bolt. Each Triplex complex consists of two copies of the large subunit gp88 in structurally distinct conformations, gp88-a and gp88-b, along with one copy of the small subunit gp89 (Fig. 2c). The three subunits combine to form an extended helical bundle that contributes to much of the association energy, holding the complex together. The nut-subcomplex is likely metastable in the extended tail conformation, ready for upward motion during sheath contraction¹⁴.

P7-1 contains two tail fiber proteins

The C3 reconstruction of the P7-1 baseplate revealed prominent upward-pointing density features, visible at a high contour, that spiral approximately 400 Å along the tail and contact the sheath lattice (Fig. 5a, left). We assigned this density to the Short Tail Fiber (STF) gp92 (Fig. 2c), which we modeled using AlphaFold3³⁷ as a slender trimer of 499-residue subunits. We docked the STF model into the density and real-space-refined it against a focused reconstruction of the baseplate to CC = 0.83 at 4.2 Å resolution (Table 1). STF measures approximately 410 Å in length and about 25 Å in thickness over most of its span (Fig. 5b). It features an N-terminal globular knob that is roughly 40 Å in diameter, as well as a C-terminal knob that is approximately 60 Å long (Fig. 5c). By lowering the map contour (Fig. 5a, right), we also noticed a second elongated density pointing upward along the tail sheath lattice, roughly parallel to STF. We attributed this density to the LTF gp90, which comprises 671 amino acids (Fig. 2c). Although the LTF density is much weaker than that of STF, an AlphaFold3 model of trimeric gp90 fits surprisingly well within the C3 cryo-EM map (Fig. 5a, right). Six copies of LTF were docked per baseplate, which, together with six STF, yield twelve tail fibers (Fig. 1a). Unlike STF, which refined well against the C3 density (CC = 0.83 at 4.2 Å, Table 1), the LTF model was not refined against the weak density to prevent overfitting. Comparing the two fibers reveals interesting differences. Both fibers are slightly bent, but LTF is longer (480 Å in length) and spirals as it meets the tail attachment site at the tail spike (Fig. 5c). It contains five knobs: three internal (I-knob-1, I-knob-2, and I-knob-3) and two at the opposite N- and C-terminal tips, each approximately 40 Å in diameter (Fig. 5c). Both STF and LTF attach to the Triplex complex small subunit, gp89 (Fig. 5d). This subunit specifically contains two fiber attachment loops (FAL), FAL1 (res. 92–112) and FAL2 (res. 215–236) (Fig. 5e), which act as flexible anchoring surfaces for STF and LTF, respectively.

Fig. 5. Structures of P7-1 tail fibers and their attachment to the baseplate.

a A semi-transparent C3 localized map of the P7-1 baseplate and tail, computed at 4.2 Å resolution, is contoured at 3.4σ (left) and 0.4σ (right). STF (cyan) is visible at both contours, while density features for LTF (salmon) are only detectable at the lower contour. b Cartoon schematic of the isolated STF and LTF assembled with a Triplex complex. c The solvent-surface representation of STF and LFT highlights the positions of the knobs. STF contains two knobs, while LTF has five knob-like motifs: two at the N- and C-terminal tips and three internal (annotated as I-knob-1/2/3). Magnified views of STF and LTF C-knobs reveal the differing bulkiness of these two motifs. d Magnified view of the interface between the baseplate Triplex complex (gp89, gp88-a, gp88-b) and the STF/LTF N-knobs. e A zoomed-in view of gp89 fiber attachment loops FAL1 (blue) and FAL2 (magenta) that bind to STF and LTF n-knobs, respectively.

The stronger electron density seen in the cryo-EM reconstruction for STF compared to LTF (Fig. 5a) suggests that STF primarily remains in a closed conformation, stabilizing the extended tail. In contrast, LTF is looser and likely accounts for the fibers extending outward from the baseplate in the cryo-micrographs (Supplementary Fig. 1a). We attribute the enhanced rigidity of STF to three factors. First, the STF N-termini obtain additional supporting stability from an extensive binding interface with the Triplex helical bundle (Supplementary Fig. 7a), which is not observed for the LTF (Supplementary Fig. 7b). Second, only the STF inserts into the sheath lattice, which likely explains the higher sequence conservation of STF versus LTF, as it must make extensive contacts throughout its 210 Å length (Supplementary Fig. 7c). Each sheath subunit makes two salt bridges, nine hydrogen bonds, and 198 nonbonded contacts with the STF, whose slender surface facilitates bonding with the sheath lattice. These interactions might be weaker for the LTF, where three bulky I-knobs (Fig. 5c) sterically hinder bonding with the sheath subunits. Finally, the STF C-knob is longer and bulkier than its counterpart in the LTF (Fig. 5c), allowing it to fit snugly within five layers of sheath molecules and anchor itself against the fifth layer (Supplementary Fig. 7c).

Evolution of P7-1 tail fibers

Phage P7-1 carries 12 tail fibers arranged around the tail spike, distal to the capsid (Fig. 1a). Although they appear similar, the two fiber types are evolutionarily, structurally, and likely functionally distinct. Comparative analysis of structural protein-coding sequences from 61 Pakpunavirus genomes (Supplementary Data 2) revealed that STFs are highly conserved, with 92.4–99.6% identity across the Pakpunavirus genus (Supplementary Fig. 8). In contrast, LTFs are more variable, with ~85% sequence identity overall (Supplementary Fig. 8). Domain-level analysis (Fig. 6a) revealed that the N-terminal region of LTF (res. 1–324), which attaches to the baseplate, is the most conserved (86.4–100%), while the middle and C-terminal regions are considerably more divergent.

Fig. 6. Evolutionary analysis of LTF in Pakpunaviruses.

a A diagram showing 7-1 LTF (gp90) conservation mapped onto its structure, divided into three regions. Each color represents an amino acid, with identical amino acids in phage 7-1 LTF shown in gray. Colored bars highlight discrepancies with the 7-1 LTF sequence. The N-terminal region (amino acids 1–324; green) is highly conserved across all phages. The middle segment (amino acids 325–540; orange) differs from the other three protein groups in phage 7-1. The C-terminal end (amino acids 541–671; red) is very similar between 7-1-like and PAK_P1-like phages but differs significantly between these groups and the EM-like and C11-like phages (Supplementary Table 2). b Phylogenetic tree of LTFs from 60 P7-1-like phages, showing four distinct clusters represented by P7-1 (yellow), PAK_P1 (cyan), EM (gray), and C11 (pink).

To evaluate whether LTF conservation reflects phage evolutionary subgroups, we performed a phylogenetic analysis of 60 publicly available Pakpunavirus genomes, excluding phage LCK69 due to an intron-split LTF gene (Supplementary Data 2). This analysis identified four phylogenetic groups (Fig. 6b; Supplementary Table 2). Twenty-nine phages, including P7-1 (Supplementary Table 3), cluster together with >95.4% sequence identity. A second group of 27 phages, represented by C11 (Supplementary Table 3), shares 91% identity among themselves but only 64.7–69.1% with other groups. C11-like and P7-1-like LTFs are highly conserved across residues 1–324 but diverge substantially thereafter (Fig. 6a; green vs. orange/red regions). A third group of three phages, represented by PAK_P1 (Supplementary Table 3), shows >93.3% identity among themselves, ~88% identity with P7-1-like LTFs, ~67% identity with C11-like LTFs, and 74% identity with EM. PAK_P1-like LTFs retain strong conservation in the N-terminal domain (res. 1–324; Fig. 6a, green) and in the C-terminal domain (>93% identity to P7-1-like LTFs; Fig. 6a, red), but differ markedly in the intervening region (res. 325–540; Fig. 6a, orange). Finally, phage EM represents a fourth (Supplementary Table 3), more divergent lineage, with its LTF distinct from all others except for limited conservation at the N-terminus (res. 1–80).

Consistent with these sequence-based patterns, preliminary data from Armata Pharmaceuticals’ proprietary phage collection suggest that LTF C-terminal domains contribute to host specificity. In particular, a P7-1-like C-terminus is associated with infection of P. aeruginosa strains expressing O3, O4, or O11 O-antigens, while a C11-like C-terminus correlates with infection of O2-expressing strains. These findings support a central role for the LTF C-terminal region in determining host range in Pakpunaviruses. The contribution, if any, of the variable middle region (orange in Fig. 6a) remains unclear.

Sheath contraction restructures the P7-1 baseplate

Prolonged incubation of P7-1 at acidic pH causes irreversible sheath contraction in approximately half of the phages on a cryo-grid (Supplementary Fig. 1b), which we used to determine a medium-resolution reconstruction of the contracted baseplate (Fig. 1b, Table 1). Even with its modest resolution, this C6 density proved helpful in modeling the contracted conformation of the P7-1 baseplate (Fig. 7a). Sheath contraction leads to the dissociation of the baseplate into two components, as seen in Twort-like Myoviridae

Fig. 7. Conformational changes during sheath contraction.

a The density map of the contracted baseplate, calculated at 4.2 Å, is displayed as a semi-transparent surface at a 2.5 σ contour level. Only six sheath proteins (purple) from the first layer and six copies of the Triplex complex are overlaid onto the density. b A magnified view shows a Triplex complex bonded to adaptor-2 (gp87) (light pink), positioned between two sheath subunits. c The adaptor-2 forms a 4-stranded β-sheet with N- and C-terminal ends from neighboring sheath subunits. d, e Ribbon diagram of the Triplex small subunit gp89 before and after contraction. FAL1 and FAL2 are colored blue and magenta, respectively, and they swing downwards after contraction.

phages³⁸. The baseplate cap subcomplex (comprising subunits gp78, gp79, gp83, gp85, and gp86) remains at the distal tail spike. The resolution of our reconstruction is too low to model these subunits. However, the fact that most contracted phages retain dsDNA in their heads (Supplementary Fig. 1b) strongly suggests that the tail spike continues to seal the tail after sheath contraction, and tail sheath contraction alone is not sufficient to induce genome release. In contrast, both the baseplate nut subcomplex (six copies of subunits gp88, gp89-a, and gp89-b) and six copies of the adaptor-2 (gp87) move upwards along the tail tube like a bolt around a nut (Fig. 7a). The adaptor-2 subunits, which decorate the outer surface of the sheath initiator complex in the extended baseplate (Fig. 4b), move upward to insert at the interface between sheath subunits, from the first layer of the contracted sheath. We modeled the N- and C-terminal moieties of the sheath from neighboring subunits that create a 4-stranded β-sheet with a β-hairpin from the adaptor-2 subunit, which effectively functions as a sheath terminator (Fig. 7c). A distinctive ~70 Å gap is visible between the first sheath layer and the contracted baseplate (Figs. 1b, 8a), corresponding to the height of the Triplex complex helical bundle that bonds adaptor-2 subunits (Fig. 7b). Presumably, the binding of the tail fibers, especially LTF, to their receptors pulls the FALs downward, enabling a conformational change in the triplex that releases it from the tail tube termination subcomplex and allows the sheath to contract, bringing the nut-bolt subcomplex with it (Fig. 7d). FAL1, the attachment site for STF, swings by 270°, while the more flexible LTF connecting loop, FAL2, swings by approximately 90°, projecting both short and long TFs downwards in the contracted state (Supplementary Fig. 9). Due to their flexibility and the lack of physical restraints in the contracted state, only about 200 amino acids of each fiber are visible in the reconstruction.

Fig. 8. P7-1 Tape Measure Protein.

a Cutaway view of P7-1 extended tail from a C3 density map calculated at a resolution of 4.5 Å. The extended tail and tail spike are colored gray and yellow, while the density for TMP is green. b Magnified view of the C3 density of the TMP 3-helix cork (green) bound to the N-terminal coiled-coil of the tail spike. c A ribbon diagram of the tail spike:TMP (res. 754–784) interaction is shown in side and top views. d A magnified view of the C3 density for the TMP 6-helix crown (green) located at the interface between the tail tube-B (pink) and tail tube-C (light green) hexamers of the cap subcomplex. e A ribbon diagram of the tail lumen, filled with six copies of the TMP (res. 689–706), is presented as a cutaway view. f A magnified view of the C3 density for the TMP 6-helix bundle filling the tail lumen. A six-helix polyalanine model, each about 70 residues long, was placed within the density to help visualize the six α-helices. g A ribbon diagram of the tail lumen, filled with six helical densities of TMP, is presented in a cutaway view.

Ordered regions of the TMP are visible within the cap subcomplex

The P7-1 tail tube is filled with density, likely consisting of the TMP (gp82), which is abundant in the MS analysis (Supplementary Table 1). By using a C3 map of the extended tail, we identified four stretches of ordered electron density spanning nearly 1000 Å of the tail lumen, named as (i)–(iv) in Fig. 8a. The first and most ordered region of density is located within the nut-subcomplex lumen at the interface with the tail spike, as well as slightly above it, although lacking continuity (Fig. 8a(i), b). The density was interpreted using ModelAngelo³⁹ and found to correspond to the C-terminus of P7-1 TMP, residues 754–784. As recently observed in phage Pa193⁹ and previously reported for Siphoviridae phage HK97⁴⁰, TMP C-terminal residues form a trimeric coiled-coil that sits like a cork onto the C-terminus of the tail spike, which is also arranged into a trimeric coiled-coil (Fig. 8c). We will refer to this part of TMP as the 3-helix cork. The cork is closely connected to the tail spike, which explains why the density is so well-ordered and traceable. The second TMP region visible in the C3 reconstruction consists of a helical hexamer (the TMP 6-helix crown) (Fig. 8a(ii), d), formed by three α-helical hairpins made up of TMP residues 689–706. This ring is located at the inner interface between tail tube-B and tail tube-C (Fig. 8d), marking the first nonsymmetric interface in the repetitive tail tube lumen. TMP 6-helix crown inserts into the acidic tail tube interface, primarily bonding through Arg 691 and limiting the tail tube lumen to about 15 Å (Fig. 8e).

Finally, extensive elongated density for TMP fills the entire tail tube, which in the C3 map appears to form two 6-helix bundles. The bottom bundle (labeled as (iii) in Fig. 8a) is more continuous. It can be tentatively traced with six poly-alanine α-helices of approximately 70 residues each (Fig. 8f). The top bundle (labeled as (iv) in Fig. 8a) is less continuous and tapers off toward the top of the tail lumen, where the thinner density suggests that TMP lost a helical fold.

Since the cryo-EM reconstruction does not directly reveal whether TMP is trimeric or hexameric inside the P7-1 tail, we examined whether the tail volume can hold six copies of the TMP. To answer this question, we used the Matthews coefficient probability, which has been used in protein crystallography for decades to estimate the approximate number of molecules in a crystallographic asymmetric unit⁴¹. The P7-1 tail lumen has a diameter of approximately 35.5 Å, calculated from juxtapposed arginine side chains lining the channel, and a length of roughly 1100 Å. This dimension can vary depending on where the DNA density ends and where the TMP begins, but we assume that TMP begins at the start of the sheath (Fig. 3b). Thus, the volume of this cylinder is V = h × r² × π = 1100 × 315.06 × 3.14 = 1,088,217.2 A³. We then approximated the P7-1 tail lumen as a triclinic unit cell (containing only one asymmetric unit) with dimensions a = 31.4 Å, b = 31.4 Å, and c = 1100 Å, yielding a unit cell volume of 1,084,556 ų, which is very similar to the volume of the cylinder described above. Assuming six copies of TMP (M.W. 85.9 kDa), the Matthews coefficient VM (V/M.W.) is 2.1 ų/Da, which is a plausible value for a hydrated unit cell since it falls within the typical range of 1.62 to 3.53 ų/Da. This corresponds to a unit cell with approximately 41.6% solvent content, similar to the tail tube filled with TMP observed experimentally by cryo-EM (Fig. 8g). Therefore, it is physically reasonable to have six copies of TMP inside the P7-1 tail lumen, supported by the presence of six identical helices in the TMP 6-helix crown (Fig. 8d). In contrast, three copies of TMP, as suggested by TMP 3-helix cork (Fig. 8b, c), lead to an unlikely Matthews coefficient of 4.2 ų/Da and an estimated 70.8% solvent content, which may be too high to observe ordered TMP density inside P7-1 tail through single-particle analysis.

Discussion

Pakpunaviruses are of significant biomedical interest due to their potent bactericidal activity and broad host specificity⁴². More than 60 Pakpunavirus genomes have been sequenced to date, and this number is expected to grow as large-scale sequencing efforts expand. However, the absence of structural and functional information has limited accurate annotation of their proteins. Here, using cryo-EM, localized reconstruction, proteomics, and bioinformatics, we annotated 19 structural genes of the Pseudomonas phage P7-1 and captured the virion in both extended, DNA-filled, and contracted, DNA-empty conformations. This integrated structural atlas reveals several unanticipated features of P7-1 biology.

First, P7-1 virions exhibit exceptional stability. More than 95% of vitrified virions retained extended tails 3 months after purification, in contrast to the Pbunavirus E217, which shows ~30% contracted, empty particles within days¹⁴. P7-1 was also resistant to alkalinization, which triggered contraction in E217, whereas prolonged acidification induced contraction in P7-1, although even under these conditions, most particles retained DNA. Fewer than 3% of contracted particles released their genome, confirming that contraction is necessary but not sufficient for genome ejection⁴³. Our cryo-EM reconstruction suggests that stability arises from several distinct determinants. The P7-1 sheath initiator forms extensive interactions with the tail tube-C and the spike:hub complex, effectively gluing the baseplate cap together and making this subcomplex a specialized extension of the repetitive tail tube. These interactions appear substantially stronger in P7-1 than in Pbunaviruses E217¹⁴ or Pa193⁹ (Supplementary Fig. 6a–c), which may explain why, even after sheath contraction, the tail remains sealed, and the genome is retained within the capsid. Additionally, our study shows negatively charged residues lining the neck channel, which may help stabilize interactions with DNA, as well as an ordered region of the TMP that makes extensive contacts with the hub:spike complex, supporting the existence of a stabilizing plug within the tail tube. Finally, P7-1 encodes two tail fibers per Triplex complex, yielding 12 fibers per baseplate, reminiscent of bacteriophage Sf14⁴⁴. Our structural analysis reveals that six STFs are embedded within the sheath lattice, physically stabilizing the Triplex complexes in an extended conformation and preventing premature contraction³⁴. In contrast, LTFs are more loosely positioned, consistent with a role as flexible probes for initial host recognition. Together, these dual fibers provide both mechanical stability and infection specificity.

Second, P7-1 also displays an unusual degree of fiber complexity. The presence of two sets of six fibers in P7-1, possibly arising from a duplication of the fiber attachment loop (FAL) in the Triplex small subunit gp89, is a feature absent in Pbunaviruses. In this respect, P7-1 resembles phage T4, which also carries both short and long fibers^45,46. Like T4, the larger genome of Pakpunaviruses (~30% larger than most Pbunaviruses) is associated with greater fiber complexity, although the baseplate remains comparable in size to Pbunaviruses and smaller than that of T4. Functional data indicate that replacing P7-1 LTF with heterologous LTFs from other Pakpunaviruses alters host range, supporting a role for LTFs in recognizing strain-specific O-antigens of P. aeruginosa¹⁴. STFs likely serve as secondary anchors that recognize a secondary receptor, possibly a component of the core LPS. Nonetheless, significant host-range overlap among phages with divergent LTFs suggests that multiple attachment mechanisms contribute to infection. Ongoing genetic work will clarify the relative roles of tail fibers and baseplate proteins in host specificity.

Third, we identified an oligomerization mismatch in TMP. At the distal tail tip, TMP adopts three-fold symmetry, whereas within the lumen it exhibits six-fold symmetry, as reported for other Myoviruses^14,47. Six ordered helices (res. 689–706) were resolved, confirming that TMP exists in six copies within the extended tail. A theoretical calculation based on the inner tail volume further supports this stoichiometry, demonstrating that six copies of an 85.9 kDa TMP are chemically compatible with the volume of the P7-1 tail. The observed TMP structure visualized within the extended P7-1 tail likely represents a pre-ejection conformation, consistent with evidence that TMP refolds during genome delivery⁴⁸. TMP is multifunctional: in addition to determining tail length during assembly, it may facilitate genome ejection by forming a channel across the host membrane^19,49. Such roles are well established in Siphoviridae and supported by studies of λ^50,51, HK97⁴⁰, and T5⁵². Moreover, TMP can encode peptidoglycan hydrolase domains⁵³. In P7-1, we identified a putative Cpl-7 lysozyme-like domain (res.s 641–674), although its enzymatic activity has not yet been experimentally validated. Further structural and functional studies are needed to define TMP rearrangement after contraction and its precise role in genome ejection.

Finally, our study provides important insights into the therapeutic potential of P7‑1. We identified the STF embedded in the sheath layer, explaining the phage’s exceptional structural stability. P7‑1 is among the most stable biomedicines in the Armata collection of therapeutic phages. Defining the binding interface between the STF and the sheath layer opens opportunities for targeted mutagenesis or in vitro evolution to modulate this interaction. In parallel, structural and evolutionary analyses of LTF revealed that its C‑terminal region (amino acids 541–671, Fig. 6a), potentially involved in O‑antigen binding, is conserved only in a subset of P7‑1-like phages and shows substantial variation between phage groups. This region, therefore, represents a promising target for domain swapping, offering an opportunity to alter the phage’s host range. Finally, the observation of contracted virions that retain DNA inside the capsid suggests that a co-receptor is likely required for genome ejection in vivo. Such a receptor may engage the tail hub or tail spike, which was proposed to function like a puncturing device⁵⁴.

In summary, this work establishes P7-1 as the first Pakpunavirus phage with a structurally annotated virion from head to baseplate. By integrating cryo-EM with proteomics and comparative genomics, we provide insights into virion stability, fiber-mediated host recognition, and the potential role of TMP in genome delivery. This structural atlas lays the foundation for mapping functional mutations and provides principles for the rational engineering of therapeutic phages with enhanced stability, specificity, and efficacy.

Methods

Origin and characteristics of Pseudomonas phage P7-1

P7-1 was isolated from sewer samples collected in Southern California. Fermentation was performed in batch mode and monitored until lysis. The crude lysate was purified using chromatography-based methods to achieve the desired levels of purity. After final concentration, the titer of the phage solution used to produce grids was 1 × 1013 PFU/mL. P7-1 was sequenced on an Illumina MiSEQ and was assembled and annotated at Armata Pharmaceuticals using proprietary pipelines. The resulting genome is 93,745 bp long and contains 1143 base-long Direct Terminal Repeats. The coding sequences relevant to this work are published in Supplementary Data 1. Bioinformatics analyses were performed using the most up-to-date version of Geneious Prime 2025.2 (https://www.geneious.com). Genomes of publicly available Pakpunaviruses were downloaded from NCBI, and their accession numbers are listed in Supplementary Data 2. Homologs of structural proteins for all phages were identified from whole genome alignments and corrected as necessary to match P7-1’s coding sequences. Genomes that lacked annotations were first annotated using the same pipelines as those used for the assembly and annotation of P7-1.

Vitrification and data collection

3.0 µL of P7-1 virions, measured at 1 × 1013 PFU/mL, was applied to a 200-mesh copper Quantifoil R 2/1 holey carbon grid (EMS) previously glow-discharged for 60 s at 15 mA using an easiGlow (PELCO). The grid was blotted for 7.5 s at blot force 2 and vitrified immediately in liquid ethane using a Vitrobot Mark IV (Thermo Scientific). Cryo-grids were screened on a 200 kV Glacios (Thermo Scientific) equipped with a Falcon4 detector (Thermo Scientific) at Thomas Jefferson University. EPU (v3.15) software (Thermo Scientific) was used for data screening using the accurate positioning mode. For high-resolution data collection of P7-1 with an extended tail, micrographs were collected on a Titan Krios (Thermo Scientific) microscope operated at 300 kV and equipped with a K3 direct electron detector camera (Gatan) at the National Cryo-Electron Microscopy Facility (NCEF) at the Frederick National Laboratory, MD. The software Leginon was used for data collection⁵⁵. To induce P7-1 tail contraction, virions were incubated at pH 4.5 for 2 h. A dataset was collected at the Stanford-SLAC CryoEM Center (S2C2) using a 300 kV Krios equipped with a Falcon 4 direct detector, operated with the EPU (v3.15) software (Thermo Scientific).

Liquid chromatography/mass spectrometry (LC/MS/MS) analysis

Phage samples were treated with 12 mM sodium lauryl sarcosine, 0.5% sodium deoxycholate, and 50 mM triethyl ammonium bicarbonate (TEAB), heated to 95 °C for 10 min, and then sonicated for 10 min, followed by the addition 5 mM tris(2-carboxyethyl) phosphine and 10 mM chloroacetamide to entirely reduce and alkylate the proteins in the sample. The sample was then subjected to trypsin digestion overnight (1:100 w/w trypsin added two times). Following digestion, the sample was acidified, lyophilized, and then desalted before injection onto a laser-pulled nanobore C18 column with 1.8 μm beads. This was followed by ionization through a hybrid quadrupole-Orbitrap mass spectrometer.

The most abundant proteins were identified by searching the experimental data against a phage protein database, a Pseudomonas host protein database, and a common contaminant database using the MASCOT algorithm⁵⁶. The predicted proteome of phage P7-1, generated by in silico translation of predicted open reading frames, has been deposited in Figshare (10.6084/m9.figshare.31934328)⁵⁷.

Cryo-EM SPA

All micrographs of phage P7-1 were motion-corrected with MotionCorr2⁵⁸. RELION’s implementation of motion correction was applied to the micrographs with options of dose-weighted averaged micrographs and the sum of non-dose-weighted power spectra every 4 e^-/Ų. Contrast Transfer Function (CTF) was estimated using CTFFIND4⁵⁹. After initial reference picking and 2D classification, particles were subjected to a reference-free low-resolution reconstruction without imposing symmetry. The particles were then 3D classified into four classes, with I4 symmetry imposed. Among the four classes, the best one was selected and underwent 3D auto-refinement to finely align the particles. The particles were then expanded according to I4 symmetry using RELION’s relion_particle_symmetry_expand function to obtain 60 times the initial particles. A cylindrical mask (r = 200 Å) was generated using SCIPION 3.0⁶⁰ and then resampled onto a reference map covering the five-fold vertex in Chimera⁶¹. The cylindrical mask was then used for non-sampling 3D classification (as implemented in RELION 3.1.2^62,63) without imposing symmetry to search for the tail. Locally aligned particles were then combined, and duplicate particles were removed. The initial localized reference map was reconstructed directly from one of the classes using RELION’s ab initio 3D Initial Model. Selected 3D classes were auto-refined using C5 symmetry, followed by five-fold particle expansion. The expanded particles underwent a third 3D classification, and the map was symmetrized using C12 and C6 symmetries, which yielded the best densities for the portal/head-to-tail and HT-adaptor-2/sheath terminator/tail tube/sheath proteins, respectively.

To reconstruct the baseplate, we manually picked 10,400 particles containing the distal tip of the P7-1 tail and reconstructed it with C3 symmetry, as C6 symmetry would result in a loss of density features. For the contracted sheath, P7-1 with contracted tails were picked from 7832 movies collected on a 300 kV Krios/Falcon 4 and processed as detailed above. After extensive particle picking and 2D classification, 10,329 contracted tail particles were reconstructed in C6 symmetry. All steps of SPA, including 2D Classification, 3D classification, 3D refinement, CTF refinement, particle polishing, post-processing, and local resolution calculation, were carried out using RELION 3.1.2^62,63. The final densities were sharpened using phenix.autosharpen⁶⁴. RELION_postprocess^62,63 was used to estimate local resolution. All cryo-EM data collection statistics are in Table 1. CryoSPARC v5.0.2⁶⁵ was used to reconstruct the P7-1 empty particle. Both empty and full particles were picked, yielding 27,000 full particles and 1300 empty particles. Particle sets were merged for the icosahedral refinement of the capsid. The same strategy for separating the phage tail using a cylindrical mask on symmetry-expanded particles, as described earlier in this section, was adapted for cryoSPARC. After resolving the C6 tail, empty particles were separated out and reconstructed, followed by local refinement of just the empty particles, leading to a 4.1 Å C6-averaged map (deposited as EMD-75604).

De novo model building, oligomer generation, and refinement

All de novo atomic models presented in this paper were built using Coot⁶⁶ and ChimeraX⁶⁷. We used four different maps for model building: (i) a 3.3 Å C5-averaged localized reconstruction¹⁶ of the mature capsid (Supplementary Fig. 2a) yielded models of the capsid (gp71) and the decoration protein (gp70); (ii) 3.6 Å C12- and C6-averaged localized reconstructions of the extended neck and tail (Supplementary Fig. 2b) were used to build the dodecameric portal (gp67) and HT-adaptor (gp72), as well as hexameric HT-adaptor-2 (gp74), sheath terminator (gp75), tail sheath (gp76), and tail tube (gp77); (iii) a 3.6 Å C3-averaged reconstruction of the extended baseplate (Supplementary Fig. 2c) yielded models of the tube-B (gp78), tube-C (gp79), sheath initiator (gp83), adaptor-1 (gp84), hub (gp85), tail spike (gp86), adaptor-2 (gp87), Triplex subunits (gp88-a/gp88-b/gp89). ModelAngelo³⁹ was used to de novo build TMP (gp82) residues 689–706 and 754–784. The same C3 reconstruction of the extended baseplate and tail bottom, but calculated at 4.2 Å resolution, was used to dock AlphaFold3³⁷ models of STF and LTF. Both fibers were fit into the density using the “fit-into-map” command in ChimeraX⁶⁷. (iv) A 4.2 Å C6-averaged reconstruction of the contracted baseplate (Supplementary Fig. 2 d) was used to dock all the factors observed in the contracted tail and baseplate subunits. All atomic models, except for LTF, were refined against the different densities described above through multiple rounds of rigid-body, real-space, and B-factor refinement with phenix.real_space_refinement⁶⁸. Figure 2 shows all final models generated in this study (except for TMP fragments), and Table 1 presents the relative map and model statistics.

Structural analysis

All ribbon and surface representations were generated using ChimeraX⁶⁷. Drawings of electron density maps and local resolution maps were generated using ChimeraX⁶⁷. Structural neighbors and flexible regions were identified using Foldseek⁶⁹. RMSD between superimposed PDBs was calculated using SuperPose Version 1.0 (superpose.wishartlab.com)⁷⁰. Binding interfaces were analyzed using PISA⁷¹ and PDBsum⁷² to determine bonding interactions, interatomic distances, and bond types. Structure prediction was done using AlphaFold¹⁷, and the hexameric TMP was generated using AlphaFold3³⁷.

Statistics and reproducibility

The cryo-EM data were collected from several grids. Micrographs with poor ice were excluded from the reconstruction pool by visual inspection. Data collection, processing, and refinement statistics are summarized in Table 1. All atomic models were validated using MolProbity⁷³. All RMSDs in the Cα position between superimposed structures were calculated using Coot⁶⁶.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

F.L., N.F.B., C.-F.D.H., R.Y., and G.C. performed all steps of the cryo-EM data collection and analysis, deposition of atomic coordinates, and maps. R.K.L. helped with biochemical methods. G.C., P.K., D.B., and S.L. supervised the entire project. F.L., S.L., and G.C. wrote the paper. Z.K., R.G., A.S., S.B., L.S., and S.L. isolated, amplified, and purified P7-1 for cryo-EM analysis. A.S. analyzed the LC/MS MS data in conjunction with the service provider. R.G. and S.L. sequenced and analyzed the P7-1 genome. All authors contributed to the writing and editing of the manuscript.

Peer review

Peer review information

Communications Biology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editors: Dr. Nilanjan Banerjee. A peer review file is available.

Data availability

The atomic coordinates have been deposited in the Protein Data Bank under accession codes 9ZBI, 9ZBS, 9ZD8, 9ZD7, 9ZDK, and 11BP. The corresponding cryo-EM maps have been deposited in the Electron Microscopy Data Bank under accession codes EMD-73989, EMD-73999, EMD-74057, EMD-74056, EMD-74065, and EMD-75604. All data supporting the findings of this study are available within the Article and its Supplementary Information. The sequences of predicted structural proteins of phage P7-1 are provided in Supplementary Data 1. A list of Pakpunavirus genomes used in this study is provided in Supplementary Data 2. The predicted proteome of phage P7-1, generated by in silico translation of predicted open reading frames, has been deposited in Figshare (10.6084/m9.figshare.31934328). All other data are available from the corresponding author upon reasonable request.

Competing interests

Z.K., S.B., R.G., A.S., L.S., P.K., D.B., and S.L. are employees of Armata Pharmaceuticals Inc., a company involved in the development of bacteriophage therapies. The remaining authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-026-10148-5.