Cryo-EM structures of bacteriophage T4 portal-neck assembly intermediates reveal a viral genome retention mechanism

Abstract

Bacteriophage T4 has long served as an extraordinary model for tailed phages. During virion assembly, the viral DNA genome is tightly packed into the head, to which tail attaches via a portal-neck connector. Keeping this highly pressurized head leak-proof during these transactions is a challenge, yet the mechanisms remain poorly understood. Here we show that T4 seals its DNA-filled capsid using a double “genome-gate” mechanism. By reconstituting portal (gp20)–neck (gp13/gp14) assembly intermediates in vitro and determining their structures, we find that the gp14 hexamer forms a primary gate that closes the portal-neck opening. This gate is reinforced by a second gate formed by the host protein Hfq, which is hijacked by the virus as an accessory structural component. Hfq also stabilizes neck assembly and prevents its mis-assembly with portal. These studies define a viral genome retention mechanism in which a pre-assembled gp13/gp14/Hfq neck complex seals the pressurized, genome-filled capsid during virion maturation.

After genome packaging, tailed bacteriophages must seal the pressurized head to prevent DNA leakage. Here, the authors show that the T4 phage assembles a neck complex with closed double genome-gate by co-opting a host protein Hfq that plugs the head and retains the genome inside.

Introduction

The bacteriophage (phage) T4 is a contractile tailed phage that infects Escherichia coli bacterium. It is a member of Myoviridae family, which constitutes one the most abundant and widely distributed viruses on Earth¹. Historically, phage T4 has served as an important model to elucidate the fundamental principles of molecular biology, and as a prototypical virus to tease out the mechanisms of icosahedral virus assembly and infection^2,3. Indeed, many mechanistic parallels were discovered between the prokaryotic T4 phage and the eukaryotic herpes virus assembly and genome packaging^4,5. T4 has also emerged as a versatile platform to design next generation vaccines and gene therapies^6–9.

T4 first assembles a dodecameric, mushroom-shaped, portal protein (gp20) ring structure that acts as the initiator of head (capsid) assembly through interactions with the major capsid protein gp23 and the major scaffold protein gp22. A prolate icosahedral shell made of gp23 assembles around the scaffolding core made of gp22, with the dodecameric portal forming a distinct vertex. The rest of the eleven vertices are occupied by the pentameric gp24. Later, the portal serves as a docking site for assembling a pentameric DNA packaging motor, which powers translocation of the DNA genome into the capsid through a central channel^10–13. After encapsidating ~171-kb (“headful”) genome, packaging is terminated, and the motor is ejected, allowing the portal to interact with the neck proteins gp13 and gp14. Sequential assembly of neck, tail, and tail fibers then generates an infectious virion. During infection, the virion delivers its genome into the host cell through a tunnel formed by portal, neck, and tail^13–15.

The structural components of T4 have been extensively investigated. Numerous high-resolution structures have been determined by X-ray crystallography and/or cryo-electron microscopy (EM), including that of the head, the portal, the baseplate, and the tail^2,15–19. However, the structure of the neck is yet to be resolved. Neck is a critical component because, in addition to connecting the packaged head to the tail, it is thought to act as a plug (seal) to prevent leakage of the packaged DNA^20–24. The timing and integrity of this seal are crucial because the highly acidic genome compacted to near crystalline density (~500 mg/ml) creates tremendous pressure, ~25–35 atm^25–27 or 5-7 times the pressure in a champagne bottle. Packaging termination, motor ejection, and neck and tail attachment must be seamlessly coordinated while keeping the head leak-proof. Otherwise, the internal pressure could expel the DNA, resulting in abortive assembly. Thus, neck assembly is one of the most vulnerable steps in virion assembly^10,28,29. Though several neck structures have been determined^30–34, they are all in the mature phage, the finished product. The conformational dynamics of how a plug is created, and then unsealed, and the mechanics of how the genome is fully retained inside the head remain poorly understood in any phage or virus.

In this work, we developed an in vitro assembly system and successfully assembled a variety of neck (gp13/gp14) and portal-neck (gp20/gp13/gp14) complexes that represent the intermediates of virion assembly. We discovered that a Hfq of the E. coli host is associated with the in vitro-assembled neck complexes. Hfq, originally discovered as a host factor involved in phage Qβ replication, is an abundantly expressed hexameric protein that acts as a pleotropic regulator of gene expression in bacteria. Furthermore, we solved a series of high resolution cryo-EM structures, including gp13 (dodecamer)/gp14 (hexamer) complexes with and without Hfq (hexamer), to 2.79 Å and 3.37 Å resolution, respectively, and the gp20 portal dodecamer/gp13 dodecamer/gp14 hexamer/Hfq hexamer complex to 2.91 Å resolution. These structures reveal a phage T4 neck containing two closed genome-gates (“double genome-gate”); a primary gate formed by “stopper loops” of gp14 projected into the lumen of neck channel, which would stop the genome when descended into the neck channel from the pressurized head, and a secondary gate (plug) formed by the binding of an Hfq hexamer to gp14 hexamer at the bottom of the neck, which would reinforce the primary gp14 gate. Besides, Hfq stabilizes the gp13/gp14 complex and prevents its mis-assembly with the portal. Consistent with the structures, our biochemical and genetic studies show that Hfq functions as an accessory protein by transiently associating with the neck to maximize genome retention and then leaving the neck when the tail attaches. These studies lead to a detailed genome retention mechanism that securely contains the packaged genome in the pressurized head and ensures efficient infectious virion assembly. It also highlights how a phage, by co-opting an abundant host protein, evolves a fail-safe mechanism at a critical step in virion assembly, while incurring no genome or metabolic cost to itself.

Results

In vitro assembly of T4 portal-neck connector complexes

To investigate the structure and mechanism of the T4 neck-portal connector complex, we overexpressed gp20, gp13, and gp14 individually in E. coli, purified the proteins by Ni-affinity chromatography and size-exclusion chromatography (SEC) (Fig. 1a, d), and visualized the samples by transmission electron microscopy (TEM). Negative-staining EM showed that while gp20 formed dodecamers as expected (Fig. 1f), gp13 and gp14 existed as monomers (Fig. 1c). In contrast, when we mixed the E. coli lysates to allow the assembly to occur in vitro and then purified the proteins, we observed co-purification of a variety of oligomeric complexes.

Fig. 1. Assembly of bacteriophage T4 neck-portal complexes and their functionality in vivo and in vitro.

a–f SEC profiles of and cryo-EM images of portal-neck complexes. a SEC elution profiles of gp13, gp14, and gp13/gp14 complex. b SDS-PAGE analysis of peak fractions. Hfq was co-purified with gp14. c Negative-stain EM images showing that gp13 and gp14 alone are monomeric, whereas gp13/gp14 complex assembles into an oligomeric structure. d SEC profiles of gp20, gp20/gp13/gp14, and gp20/gp13 complexes. e SDS-PAGE of the gp20/gp13/gp14 SEC peak fractions. f Negative-stain EM images showing that gp20/gp13/gp14 forms an oligomeric complex, whereas gp13/gp20 forms heterogeneous aggregates. g, h In vivo infection assay assessing the role of Hfq in T4 phage assembly: g Phage yield expressed as pfu produced per infected cell in hfq+ and hfq- E. coli infections. Data are presented as mean values +/- SD (n = 3). h Western blot detection of Hfq using Hfq-specific polyclonal antibodies. Hfq is absent in WT T4 phage or hfq + T4 phage. The hfq + T4 phage was constructed by inserting the hfq gene into T4 genome under the control of a strong late promoter of soc gene. i, j Binding of gp13/gp14/Hfq neck complex to packaged heads. i Cryo-EM of DNA-full heads purified by CsCl density gradient centrifugation. Blue arrows point to neck-like protrusions at the portal vertex, likely corresponding to bound gp13/gp14/Hfq complex. j Western blot analysis using His-tag–specific antibodies. Gp17, gp13, gp14, and Hfq contain a Histag either at N- or C-terminus of the recombinant protein. Lanes 2-4 represent 1:1 ratios of packaging motor (pentameric gp17) or gp13/gp14/Hfq complex to head particles. Lanes 5 correspond to CsCl-purified packaged head-gp13/gp14/Hfq complexes. Lanes 2-5 correspond to samples taken from the same experiment and lane 6 corresponds to the packaged head-gp13/gp14/Hfq complex from an independent experiment. Experiments in (b, e, j) were repeated three times with similar results. Molecular weight markers (M) are indicated.

Mixing gp13 and gp14 lysates resulted in the assembly of ring structures (Fig. 1a–c), whereas mixing of all three lysates of gp20, gp13, and gp14 resulted in even larger ring structures containing all three proteins (Fig. 1d–f). In contrast, mixing gp13 with gp20 did not produce ring structures but instead resulted in aggregates (Fig. 1f). These results demonstrate in vitro assembly of post-packaging neck connector structures and that the neck protein gp13 preassembles first with gp14 forming stable oligomeric rings before attaching to the dodecameric portal.

Involvement of a host factor, Hfq, in T4 neck assembly

Surprisingly, we observed an additional ~12 kDa protein band in the purified preparations containing gp14 (Fig. 1a, b). LC–MS/MS analysis identified this gp14-associated band as the Hfq protein of E. coli. Hfq is an abundant E. coli protein and originally discovered as a host factor for phage Qβ replication. It was later characterized as a pleiotropic regulator in E. coli and is commonly found in most bacterial genomes^35,36. Although Hfq primarily binds to RNA, it can also interact with a wide range of nucleic acids, including DNA, and regulates transcription, genome stability, and mRNA decay^37–39.

To assess the physiological importance of Hfq in T4 phage assembly, we first performed in vivo infection assays. The results showed that the phage yield of wild-type (WT) T4 phage on hfq-minus E. coli NM 22565⁴⁰ is ~2-fold lower than that from the isogenic hfq-plus E. coli NM22540 (Fig. 1g).

Second, we tested the ability of the in vitro-assembled gp13/gp14/Hfq complex to attach to the packaged T4 head by developing an in vitro packaging and neck assembly system using purified components. Empty heads were packaged with a ~ 6 kb linearized plasmid DNA by the packaging motor (gp17) in a defined packaging system and then incubated with the gp13/gp14/Hfq complex. The packaged heads were purified by CsCl density gradient centrifugation and analyzed by cryo-EM (Fig. 1i) and Western blotting (Fig. 1i, j). The images in Fig. 1i show that the heads were packaged with DNA to near headful capacity. Many of the full-heads show protrusions at the portal vertex, likely corresponding to the gp13/gp14/Hfq complex. The Western blotting data show that all the three proteins of the gp13/gp14/Hfq complex are now associated with the packaged heads (Fig. 1j, lanes 5 and 6). Furthermore, most of the packaging motor was dislodged from the head following packaging (compare gp17 band in lane 5 with control lane 2, Fig. 1j), while neck-like protrusions are now seen associated with the head at the portal vertex in the cryo-EM images (blue arrows in Fig. 1i).

Third, we observed that the ratio of Hfq to gp13 to gp14 in the packaged head-neck complexes varied, some experiments showing less than the stoichiometric amount of Hfq (compare lane 5 with control lane 3, Fig. 1j). Furthermore, Hfq is completely absent in the finished virion structure (Fig. 1h).

These results suggest that the Hfq hexamer is transiently incorporated into the neck complex that binds to the packaged head for enhanced genome retention and phage yield and then leaves the neck when tail attaches.

Cryo-EM structures of T4 neck complexes

Two types of neck complexes were observed in the cryo-EM images obtained from co-purified gp13 and gp14 fractions; the gp13/gp14 complex comprising ~34% of particles and gp13/gp14/Hfq complex comprising ~66% of particles (Supplementary Fig. 1). Cryo-EM maps were generated for both these complexes without imposing symmetry to 3.70 and 3.10 Å resolution, respectively (Supplementary Fig. 2). In the gp13/gp14 complex, gp13 forms a dodecameric ring and interacts with a hexameric gp14 ring creating a central channel. In the gp13/gp14/Hfq complex, the Hfq hexamer attaches to the gp14 hexamer, essentially covering the lower end of the channel (Supplementary Fig. 2c). Notably, the gp13 ring in the gp13/gp14 complex is incomplete, lacking some of the twelve subunits (Supplementary Fig. 2e). In contrast, the gp13 ring in the gp13/gp14/Hfq complex is complete (Supplementary Fig. 2c), suggesting that Hfq stabilizes the structure of the gp13/gp14 complex.

To further improve the resolution of the cryo-EM structures, C6 symmetry was applied and the structures of gp13/gp14 and gp13/gp14/Hfq complexes were refined to 3.37 Å and 2.79 Å resolution, respectively, allowing the building of atomic models (Fig. 2a, b; Supplementary Fig. 2d and f; Supplementary Table 1). The gp13/gp14/Hfq complex has an inverse bowler hat shape, with gp13 serving as the brim, gp14 as the body, and Hfq as the crown (Fig. 2b). The gp13 dodecamer forms a central tunnel with an inner diameter ranging from 56 Å at the top to 44 Å at the bottom (Fig. 2c). The gp13 subunit can be divided into five parts: domain I, a four-helix bundle (residues 3–51, 239–300); domain II, the “swing” domain (residues 52–192); the fibritin-binding domain III (residues 190–239) which is disordered in the current structure but resolved in the virion⁴¹; the gp14-binding adaptor loop (residues 266–279), and the portal-binding C-terminal arm (residues 301-308) (Fig. 2c, d).

Fig. 2. Cryo-EM structures of gp13/gp14 neck complexes.

a Cryo-EM structure of the gp13/gp14 complex shown in side (left) and top (middle) views, with gp13 in green and gp14 in orange. Right, ribbon representation of gp13/gp14 fitted into the density map. b Cryo-EM structures of the gp13/gp14/Hfq complex shown in side (left, cryo-EM map) and top views (middle, ribbon model) with gp13 in green, gp14 in orange, and Hfq in purple. Right, ribbon representation of gp13/gp14/Hfq fitted into the density map. c Domain organization in the gp13/gp14/Hfq complex. d Schematic representation of gp13 and gp14 secondary structures, with α-helices and β-strands labelled numerically. e Three glutamic acids (Glu284, 291, and 295) lining on α6 of gp13, forming a negatively charged surface for DNA passage. f Structural comparison of gp13/gp14 (colored in gray) and gp13/gp14/Hfq. The α6 helix of gp13 C-terminal has been used for the structural alignment.

The helix bundle and the swing domains of gp13 form the core and the peripheral ring, respectively. The helix bundle is made of four helices (α1, α2, α5, α6), with α6 forming the central channel. Three glutamic acids (Glu284, Glu291, and Glu295) on α6 face the channel, creating a negatively charged surface for smooth passage of DNA (Fig. 2e). The swing domain, so named because it is observed to swing upwards in the mature T4 virion (see below), is made of seven β strands and two α helices (β1-7, α3-4) (Fig. 2d). Other than these two main domains, gp13 contains two major loops located before and after α6, named as the adaptor loop and the C-terminal arm, respectively. The 14-aa adaptor loop inserts into gp14, forming extensive interactions (Fig. 3a–c). The C-terminal arms stretch upward, providing a binding site for gp20. The interaction between adjacent gp13 subunits is mainly mediated by hydrogen bonds between the swing and helix bundle domains (Fig. 3e).

Fig. 3. The networks of interactions between gp13, gp14 and Hfq.

a Ribbon diagram of the gp13/gp14/Hfq structure fitting into the cryo-EM density map. Four neighboring gp13 subunits are labeled sequentially as n, n + 1, n + 2, and n + 3 and colored in blue, lime, orange red and forest green; Two neighboring gp14 subunits are shown in violet and orange and two neighboring Hfq subunits are shown in cyan and red. b, c Close-up view of gp13 and gp14 interfaces. Residues involved in hydrophobic and electrostatic interactions are shown as sticks. d Close-up view of the gp14 and Hfq interface. Residues involved in salt bridge and hydrogen bonding interactions are shown as sticks. e Top view of the gp13 oligomeric ring, with each subunit in a different color. The boxed regions highlight the specific interaction interfaces between monomers. f Top view of the gp14 oligomeric ring, with each subunit in a different color. The boxed regions highlight the specific interaction interfaces between monomers. g Electrostatic surface representation of gp14 (bottom view) and Hfq (top view), illustrating positively charged (blue) and negatively charged (red) regions. h Ribbon diagram of the stopper loop region fitting into the cryo-EM density map of gp20/gp13/gp14/Hfq.

Gp14 consists of an N-terminal loop (residues 5-37), a long helix (residues 38-54), a long stopper loop (residues 89-112), and a core β-strands domain (residues 55-88, 113-177) (Fig. 2c, d). The long N-terminal loop and the N-terminal lateral helix (α1) of gp14 wrap around the adaptor loop of gp13, while six core β-strands (β1-6) form an anti-parallel β-barrel. The long loop between β2 and β3 protrudes into the center of the channel. Six such loops, referred to as “stopper loops”, form a gate-like structure that narrows the otherwise open channel (Fig. 3h), likely preventing the release of the tightly packed genome from the pressurized capsid. The C-terminal 80 amino acids are invisible, probably due to their flexibility. The loop between β5 and β6 interacts with the adjacent subunits mainly through hydrophobic interactions (I26, N28, Y36 and R54, Q52, Q126). Moreover, residues P157 and F158 form hydrophobic interactions with F140, P141and M147 of the adjacent subunit (Fig. 3f).

Gp13 and gp14 engage in intensive interactions, with one gp14 interacting with four gp13 subunits (n, n + 1, n + 2, n + 3) (Fig. 3a–c). In addition, the N-terminal loop of gp14 interacts with gp13 (n + 3) through both hydrophobic and hydrophilic interactions (Fig. 3c). The N-terminal helix α1 of gp14 spans the adaptor loops of three adjacent gp13 subunits (n, n + 1, n + 2), forming intensive hydrophobic interactions (M53, L46, V45, L42 from gp14 and I277, L271, I262, I24, I25 from gp13) (Fig. 3b). This explains why neither gp13 nor gp14 could form oligomers when expressed alone, but could do so when mixed.

Neck assembly and energy considerations

The above biochemical and structural data demonstrate that both gp13 and gp14 exist as monomers alone but when mixed together form a stable complex, a phenomenon also observed in the portal-neck assembly of P22 phage^42,43. The interface between gp13 dodecamer and gp14 hexamer is extensive with a buried area of 4181 Å, in contrast to the interface area between gp13 subunits (1914 Å) or gp14 subunits (1038 Å). Dissociation free energies (ΔG^diss) calculated by PDBePISA reveal an energetic landscape consistent with cooperative assembly (Supplementary Table 2). Among the gp13 dodecamer, gp14 hexamer, and gp13/gp14 complex, the isolated gp13 dodecamer and gp14 hexamer exhibit moderate thermodynamic stability (ΔG^diss = +46.2 and +45.3 kcal/M, respectively) but are characterized by large entropic penalties for dissociation (TΔS^diss = +158.6 and +66.7 kcal/M, respectively). The gp13/gp14 complex exhibits the lowest entropic penalty (TΔS^diss: +18.1 kcal/M) and highest ΔG^diss (+140.4 kcal/M), suggesting that complex formation is favorable. These are consistent with our experimental observations that gp13 and gp14 monomers do not spontaneously form stable oligomers individually, however mixing gp13 and gp14 monomers facilitates the formation of gp13/gp14 complex.

Thus, the T4 neck complex appears to assemble through a cooperative mechanism in which two metastable components converge to generate a hyperstable structure. This strategy ensures that a strong driving force for assembly is unleashed only upon correct formation of the functional unit, thereby preventing premature or erroneous interactions.

Structure of the neck-Hfq complex

Notably, the E. coli Hfq hexamer fits precisely at the bottom of gp14. The Hfq hexamer engages with the gp14 hexamer primarily through hydrogen bonds, involving residues Q111, D112, E113, E134, K173, I175 of gp14 and K31, T49, Q33, S65, R19 of Hfq, resulting in a substantial buried interface of 3159.2 Ų. In addition, the negative charge at the bottom of gp14 and the positive charge at the top of Hfq are likely to facilitate their interaction (Fig. 3d, g). Remarkably, the disordered stopper loop residues (106-113) of gp14 become visible only upon Hfq binding, suggesting that Hfq binding stabilizes the stopper loop (Fig. 2f). Moreover, incorporation of Hfq hexamer constricts the central channel by approximately 2 Å when compared to the gp13/gp14 complex lacking Hfq (Fig. 2f). Thus, Hfq is likely functioning as a secondary genome gate to reinforce the primary gate to prevent DNA leakage (Fig. 2c).

Notably, the gp14 residues that mediate interactions with Hfq and the negatively charged interface are highly-conserved among T4-like phages (Supplementary Fig. 7). In contrast, these residues and the corresponding interface are missing in gp14 orthologs from phages such as SPP1 and lambda (Supplementary Table 4). This suggests that Hfq usage may represent a T4 lineage-specific evolutionary adaptation that co-opts an abundant host protein to install a fail-safe genome retention mechanism without incurring any genome or metabolic cost to phage itself.

Structure of the portal-neck connector complex

Cryo-EM structural analysis of the in vitro assembled gp20/gp13/gp14 complexes revealed that a majority (~90%) of these complexes also contained the Hfq hexamer. Therefore, the structure of this portal-neck connector complex that joins the head and the tail was resolved at 2.91 Å with C6 symmetry (Fig. 4a–c and Supplementary Figs. 3,4).

Fig. 4. Structural characterization of the gp20/gp13/gp14 complex with Hfq binding.

a Cryo-EM map of T4 gp20/gp13/gp14/Hfq complex, with gp20 in red, gp13 in green, gp14 in yellow, and Hfq in purple. b Ribbon diagram of the gp20/gp13/gp14/Hfq complex. A dsDNA model was modelled into the central tunnel. c Electrostatic surface representation of the gp20/gp13/gp14/Hfq complex. Unit on color key: kcal/(mol·e). d Interface between the C-terminal arm of gp13 and two neighboring subunits of gp20 highlighted residues involved in hydrophobic and electronic interactions. e Electrostatic surface representation showing interactions between gp20 subunits with gp13, emphasizing the distribution of charged regions. f Structural comparison of apo-gp20 (PDBID: 3JA7, shown in magenta) with gp20 in gp20/gp13/gp14/Hfq complex (shown in pink), showing the rotation of the α6 helix (4.7°) upon complex formation. g Structural comparison of gp13 between gp13/gp14/Hfq (light green) and gp20/gp13/gp14/Hfq (dark green) complexes, showing 40° rotation of C-terminal arm of gp13 upon gp20 binding. h Structural comparison of gp14/Hfq between gp13/gp14/Hfq and gp20/gp13/gp14/Hfq complexes, indicating no obvious conformational change.

The overall structure of the dodecameric gp20 in this complex is similar to that of the recombinant gp20, with a root mean square deviation (r.m.s.d.) of 1.4 Å across 411 (85%) aligned residues (Supplementary Fig. 5a). However, upon gp13 binding, the α6 helix of gp20 clip domain rotates 7° and shifts 4 Å to accommodate the binding of the gp13 (Fig. 4f). The C-terminal arm of gp13 rotates upward by 40° relative to that in gp13-gp14 complex (Fig. 4g), inserting itself between two clip domain α6 helices of adjacent gp20 subunits, and forming extensive hydrophobic interactions with residues on both the helices and with the β-strand of one of the clip domains (Fig. 4d). Additionally, the negatively charged residues at the C-terminus of gp13’s α2 and α4 helices (E47, D51, E295, D301) engage in electrostatic interactions with positively charged residues of gp20’s α6 helix (R295, K296, R339) (Fig. 4d, e). On the other hand, the conformation of the gp14 and Hfq complexes remained unchanged when the neck complex attaches to the portal (Fig. 4h; Supplementary Movie 1).

Hfq ensures correct assembly of portal-neck complex

We identified a class of particles among the portal-neck connector complexes in which gp14, instead of gp13, interacted with gp20 and is positioned between the gp20 portal dodecamer and the gp13 dodecamer (Fig. 5a; Supplementary Figs. 3 and 4). Notably, these gp20/gp14/gp13 complexes are devoid of Hfq and constituted ~14% of the portal-neck complexes (Supplementary Fig. 3). In this structure determined at 3.45 Å resolution, the C-terminal arm of gp14, which normally makes extensive interactions with gp15 subunits, inserts between the clip domains of adjacent gp20 subunits. Similar to the gp20/gp13 complex, gp20 and gp14 mainly interact through hydrophobic interactions and hydrogen bonds (K309, N291, N323, R311 of gp20 and D188, A110, Y187 of gp14) (Fig. 5b, c). Furthermore, residue K309 of gp20 forms a salt-bridge with D188 of gp14. Additionally, electrostatic interactions are also observed between the positively charged clip domain of gp20 (residues of R295, K296, H300 and H303) and the negatively charged C-terminus of gp14 (Fig. 5d).

Fig. 5. Cryo-EM structure of misassembled gp20/gp14/gp13 complex.

a Ribbon diagram showing the overall structure of the gp20/gp14/gp13 complex, with gp20 in pink, gp14 in yellow, and gp13 in green. b, c Close-up view of the interactions between gp20 (pink) and gp14 (yellow), showing key residues involved in the complex formation. Dashed lines indicate hydrogen bonds and salt bridges. d Electrostatic surface representations of the interactions between gp20 with gp14, showing the distribution of positive (blue) and negative (red) charges, showing the electrostatic interactions.

To further investigate this unexpected phenomenon, and if Hfq played a role in assembling the correct portal-neck complex (gp20/gp13/gp14), we produced recombinant Hfq protein and mixed it in excess first with gp14 extract and then with the gp13 extract, followed by the portal protein extract. SDS–PAGE analysis showed that, under excess Hfq, the portal–neck complex assembled into a homogeneous gp20/gp13/gp14/Hfq complex with an approximate 1:1 molar ratio of gp14 to Hfq, whereas limited Hfq resulted in a heterogeneous mixture containing both gp20/gp13/gp14/Hfq and gp20/gp14/gp13 complexes (~2:1 molar ratio of gp14 to Hfq) (Supplementary Table 3). Cryo-EM analysis confirmed that in the presence of excess Hfq, only correctly assembled gp20/gp13/gp14 complexes were formed and no mismatched gp20/gp14/gp13 complexes detected (Supplementary Fig. 6). Thus, Hfq, by binding to gp14, also inhibits its interaction with gp20, thereby facilitating the correct assembly of the portal-neck complexes.

Conformational changes in portal-neck connector complexes

Comparative structural analysis showed that the structures of gp20, gp13, and gp14 in different in vitro assembled complexes are similar, with r.m.s.d. ranging from 0.68-1.47 (Supplementary Fig. 5). However, comparison of the in vitro assembled gp20/gp13/gp14/Hfq complex with the in situ gp20/gp13/gp14 atomic model of phage virion⁴¹ (Fig. 6) shows dramatic conformational changes in each of the neck components, which lead to a dynamic genome retention mechanism (Fig. 6a). Upon completion of headful packaging, a global conformational change occurs in the portal structure and the portal is pushed down by the pressure of the packaged genome exposing the “stem” and “wing” regions of portal for gp13 binding (Fig. 6b). After initial binding with the clip domain, as seen in the current structure, a major conformational transition occurs in the gp13 dodecamer. The swing domain flips upward by approximately 90° and its “anchor” region (aa 90-157) interacts with the now exposed stem and wing domains of the portal and the periphery domains of the major capsid protein gp23 (Fig. 6c). Furthermore, residues 193–238 which are not resolved in the current structure, model into domain III which binds to “fibritin”, a trimer of gpWac (whisker antigen control) (Fig. 6c). Twelve domains III of the gp13 dodecamer interacting with fibritins assemble twelve fibers around the neck, six as “whiskers” and the other six as “collar”. Later in virion assembly, these neck fibers interact with the long tail fibers (LTFs) in ‘up’ conformation in the pre-infection stage (Supplementary Movie 1).

Fig. 6. Structural comparison of in vitro and in situ portal-neck complexes.

a Left: in-vitro structure model of gp20/gp13/gp14/Hfq. Right: in-situ structure model of gp20/gp13/gp14 complex. b–d Structure comparison of in-vitro (left) and in-situ (right) gp20, gp13, gp14 proteins.

At the other end of the portal-neck connector, gp14 also undergoes a large conformational change. The C-terminal regions of gp14 (aa 178–256), which are disordered in the current structure, form new and extensive interactions with gp15 subunits (residues 203–245), while the β-barrel stopper loops rotate downward by ~90 ° to dock with the tail sheath terminator protein gp15 located at the tip of the tail (Fig. 6d). These conformational changes open the gp14 genome-gate, while also ejecting the Hfq hexamer.

Quantitative dissociation energy analysis using PDBePISA provides a mechanistic explanation for how the tail protein gp15 triggers Hfq displacement and allow stable docking of the tail. The gp14/Hfq complex exhibits both a low TΔS^diss and low ΔG^diss (+13.6 kcal/M), indicating that it is relatively unstable and can readily dissociate. By contrast, the ΔG^diss of the gp14/gp15 complex (+364.6 kcal/M) is approximately 25 times greater than that of gp14/Hfq, reflecting markedly higher stability. This large difference in stability indicates that, from a thermodynamic perspective, gp15 binding to gp14 is strongly favored.

This is consistent with Hfq’s role, which is restricted to the neck intermediate and dissociates upon tail attachment. Hfq binds gp14, forming a metastable complex. Tail attachment allows gp15 to displace Hfq, forming an ultra-stable gp14/gp15 interface that irreversibly locks the tail onto the neck. Together, these data demonstrate that gp15 provides the energetic driving force to disrupt the gp14/Hfq interaction, ensuring high-fidelity viral assembly (Supplementary Table 2).

These analyses reveal that the current structures represent intermediates trapped by in vitro assembly prior to tail attachment, thus preserving the structure of a transient genome plug formed to leakproof the head and retain the packaged genome while handing it over to the tail.

Discussion

One of the least understood mechanisms in large icosahedral virus assembly is how the highly pressurized viral genome remains securely contained within the capsid during the post-packaging transactions⁴⁴. After encapsidating a headful genome, packaging has to be terminated and the motor must be ejected to allow sequential assembly of the neck and tail structures in rapid succession. With ~25–35 atm pressure inside the packaged head^26,27, expulsion of even a few base pairs of DNA during these transactions could result in abortive assembly resulting in a non-infectious virus particle. Despite many phage and viral structures solved by cryo-EM, the dynamics of assembly and the conformational transitions that allow full retention of the genome at this critical juncture remain unknown.

Here, we report the generation of multiple in vitro-assembled portal-neck intermediates and determination of their structures to near atomic resolution. These structures reveal a closed, double genome-gate, composed of a primary gate formed by the essential neck protein gp14 and a secondary gate, attached to gp14, formed by the host component Hfq. Together, these gates create a tight seal at the portal vertex, preventing genome leakage from the highly pressurized packaged head.

Although a genome-gate has been implicated in other phages, such as SPP1^20,32,45,46 and lambda^47–49, no structure or neck assembly intermediate with a closed gate has yet been reported. Prior structural studies have been limited to mature virions or individual components of portal, adaptor, or stopper, therefore lacking high-resolution information on assembly intermediates, obscuring the dynamic assembly process. Furthermore, the double-gate structure uncovered in phage T4, through co-option of the host protein Hfq, represents a distinctive feature. Hfq, an abundant hexameric nucleic acid-binding protein, is ideally suited for this role^35,36. Notably, the negatively charged regions at the bottom surface of gp14 hexamer are highly conserved among T4-like phages but absent in orthologs from non-T4 phages such as lambda and SPP1, indicating a T4 lineage-specific evolutionary innovation (Supplementary Fig. 7, Supplementary Table 4). Nevertheless, Hfq is as an accessory factor rather than an essential structural component, recruited for enhancing neck assembly and genome retention through its transient association, as demonstrated by multiple and independent lines of evidence (discussed below).

Most important, our combined structural, genetic, and biochemical analyses reveal a conformation-driven viral genome retention mechanism in phage T4 (Fig. 7). Our in vitro assembly studies suggest that neither gp13 nor gp14 alone can oligomerize but together they do, forming a gp13 dodecamer/gp14 hexamer complex (Fig. 1). However, this complex is unstable losing some of the gp13 subunits in the dodecamer (Supplementary Fig. 2e). But when bound to Hfq, a stable gp13 dodecamer-gp14 hexamer-Hfq hexamer complex is assembled (Supplementary Fig. 2c) and the gp13’s unstructured C-terminal arms are stretched upwards and exposed on the surface, being ready to dock onto the portal vertex (Fig. 2a–c). Thus, while gp13 and gp14 homomeric assemblies are individually unstable, their association produces a robust complex that is further stabilized by Hfq. Notably, the cooperative mode of assembly, rather than the intrinsic stability of the individual components, underlies the fidelity and stability of the final structure. This strategy activates a strong assembly-driving force only upon correct formation of the functional unit, preventing premature or erroneous interactions.

Fig. 7. Mechanism of genome retention in bacteriophage T4.

After headful genome packaging, a global conformational change (a) in the dodecameric portal structure ejects the packaging motor and exposes binding sites for neck assembly. A pre-assembled gp13 dodecamer (adaptor)-gp14 hexamer (stopper I)-Hfq hexamer (stopper II) neck complex with a double genome-gate (b) docks on the portal clip domains through gp13’s C-terminal arms. This induces a global conformational change in gp13 adapter locking-in the neck-portal complex (c), sealing off the packaged genome-containing head with a double genome-gate, and attaching fibritin fibers (d).

Since the gp13 binding sites are not yet available during packaging, and since the packaging motor is still attached to the portal¹³, premature docking is prevented and the neck complex awaits packaging to complete. Once the head is filled and packaging terminated, the portal switches conformation, ejecting the packaging motor and is pushed down under pressure exposing the gp13 binding sites (Fig. 7a)⁴¹. The pre-assembled neck complex then docks onto the portal, first through insertion of the gp13 C-terminal arms into the portal clip domains, as observed in the current portal-neck intermediate formed in vitro when the portal clip domains are fully exposed (Fig. 5d). This then triggers a conformational change causing the gp13’s swing domain to flip upwards by 90° and embrace the stem and wing domains of the portal (symmetry-matching interactions) while another segment interacts with the “P” (periphery) domains of the major capsid protein gp23 (symmetry-mismatched interactions), as observed in the in situ portal-neck structure (Fig. 6c). Additionally, other previously unstructured regions of gp13 remodel into binding domains that capture the gpWac fibritin trimers. These interactions lead to the twelve fibers decorating the neck, six of them forming the propeller-shaped collar and the other six facing down forming the whiskers⁵⁰. Thus, a stable and intricately woven head-portal-neck structure is formed.

The viral genome thus would be safely sequestered inside this structure, with the last-packaged DNA stopped by the gp14 gate reinforced by the Hfq gate (Fig. 7c, d). In the next step of virion assembly, the pre-assembled tail attaches to the neck. As the gp15 hexamer residing at the tip of the tail approaches the neck, it triggers the genome gate to open and the Hfq hexamer is expelled (Fig. 6). The C-terminal negatively charged region of gp14 then forms extensive electrostatic interactions (30 salt-bridges) with the positively charged gp15 exposed at the surface of the tail tip locking-in the neck-tail junction. Additional conformational changes would then follow and the genome descends through the neck-tail channel and positioned in the innermost tunnel of the virion⁴¹.

Hfq’s role in the above mechanism is consistent with that of an accessory factor enhancing the efficiency of neck assembly and genome retention. This is evident from multiple sets of data. First, several independently solved structures show that Hfq forms a secondary gate by attaching to the bottom of the primary gp14 genome gate. While gp14 gate would be the one to contact the genome and stop it, Hfq attachment reinforces the primary gate. In principle, the gp14 genome gate might itself be sufficient (mutants of gp14 are lethal), but having a second gate would make it fail-safe. Second, Hfq’s association with the neck is transient because it leaves the neck when tail attaches. This is demonstrated both in vivo, since Hfq is not found in the finished virion structure (Fig. 1h), and in vitro, where we observed that some of the bound Hfq leaves the neck (even when the tail is absent), while the gp13/gp14 complex remains tightly associated with the packaged head (Fig. 1j). Third, energy calculations show that the gp14/Hfq complex exhibits both a low TΔS^diss and low ΔG^diss, indicating that this assembly is unstable. It would dissociate easily when gp15 hexamer at the tip of the tail interacts with gp14. The ΔG^diss of the gp14/gp15 complex is ~25 times greater than that of gp14/Hfq, reflecting markedly higher stability (Supplementary Table 2).

Our work provides a detailed and complete mechanism for neck assembly and viral genome retention in icosahedral phages and viruses. Particularly eye-opening are the dynamic conformation-driven transitions, illustrated as above, within every component of the portal-neck connector complex as the virus undergoes post-packaging assembly and morphogenesis, processes previously thought to involve only simple binding steps. Clearly, the structures of the finished phages alone^{20,29,45,47–51} could not reveal such conformational transitions^{20,30,47,50–54}. Given the structural and functional parallels across other large icosahedral phages and viruses, this basic mechanism is likely preserved in other phages and viruses, which might also present a new target for antiviral discovery.

Additionally, our studies uncover a strategy T4 had evolved that transiently hijacks host Hfq to streamline neck assembly and ensure efficient genome retention. Although not essential, such an engagement of an accessory host component enhances T4 phage production, thereby increasing its fitness to survive in a competitive phage universe. In fact, curiously, phage T4 appears to have evolved multiple such auxiliary functions, each providing a modest (~2-fold) enhancement in phage yield, as was observed with Hfq. Examples include MotB, gp5.4, gp45.2, and others^55,56. Referred to as “nonessential” genes, such accessory functions collectively constitute a significant portion of the T4 genome³. What is intriguing here is the co-option of a host protein to optimize genome retention and virion assembly with no genome or metabolic cost to phage itself.

Methods

Protein expression and purification

The full-length gp13 gene (NC_000866.4), gp14 (NC_000866.4) and Hfq (NC_000913.3) were synthesized (GeneScript) and inserted into the pET-29a, pET28b and pET-28a vectors with N-terminal His tag, N-terminal Strep II tag (Hfq), respectively. Plasmids information and primers Oligonucleotide sequences are provided in Supplementary Data. The protein was expressed in E. coli BL21(DE3) grown at 37 °C, induced by 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at an optical density (OD) of 0.6–0.8 at 16 °C. After 4 h, the cells were harvested by centrifugation at 5000 rpm (6240 g) with a JLA-8.1 rotor and resuspended in lysis buffer (20 mM Tris, pH 8.0, 200 mM NaCl, 20 mM imidazole, 1/500 Supernuclease S (Tiandirenhe), 1/10,000 Protease Inhibitor Cocktail (Roche). The mixture was then centrifuged at 40,000 g for 20 min at 4 °C and removed fragments by 0.22 µm filter (Millipore USA). The sample was loaded into Ni Smart Beads 6FF (Tiandirenhe) equilibrated in buffer A (20 mM Tris, pH 8.0, 500 mM NaCl), and eluted with buffer A containing 500 mM imidazole. The protein was then purified by Superose 6 Increase 10/300 GL (Cytiva) with buffer B (20 mM Tris, pH 8.0, 500 mM NaCl, and 1 mM TCEP). The elution volume indicated that gp13 and gp14 were monomeric in solution. For further studies, we mixed gp13 and gp14 lysates together from the very beginning in bacterial sonication and centrifugation, and then we used the same purification protocol described above. N-His tagged gp20 (residues 74-524) (NC_000866.4) and untagged gp22 were cloned into the pET-Duel vector for co-expression and purified as previously described¹. Finally, we mixed gp13/gp14 neck complex and portal protein gp20 together and incubated for 30 min, and then the mixture was concentrated with 50,000 Da Ultra tube (Millipore) into 0.5 ml for size-exclusion chromatography (Superose 6 Increase 10/300 GL, Cytiva).

Identification of proteins in the neck complexes by LC/CL-MS analysis

The gp14 and Hfq bands (the number of samples analyzed is n = 1, respectively) were excised from the gel and digested with Trypsin at 37 °C for 4 h. The concentrated peptides were analyzed by Orbitrap Fusion^TM Lumos^TM Tribrid^TM Mass Spectrometer (Thermo Fisher Scientific). Original data was analyzed by Thermo Proteome Discoverer 2.5.0.400 followed by searching for specific protein sequences in data bank.

Analysis of gp14/Hfq ratio based on the SDS-PAGE

The protein bands in the SDS-PAGE were calculated with imageJ software. The SDS-PAGE images were converted to grayscale images. For a more intuitive analysis, we use the inverted grayscale map for numerical calculation. The brightness of the bands indicates the mass of gp14 and Hfq proteins and is finally represented by the Mean value output by the software. The final binding mass and molecular weight were used to calculate the molar ratio of gp14 and Hfq bands on SDS-PAGE.

Phage yield in Hfq-minus E. coli

E. coli strains NM22565 (Hfq-plus) and NM22540 Dhfq::cat-sacB (Hfq-minus) were grown to mid-logarithmic phase (~2 ×10⁸ CFU/mL) in Luria-Bertani (LB) medium at 37 °C with shaking at 250 rpm (the strains were provided by Dr. Susan Gottesman from National Institutes of Health). The cells were infected with WT T4 phage at a multiplicity of infection (MOI) of 0.1 and incubated at 37 °C for 5 min. Immediately afterward, serial 10²-, 10⁴-, and 10⁶-fold dilutions were prepared in LB medium. The remaining infection mixture (non-diluted) was treated with chloroform to lyse the cells and quantify unbound phage. At 10, 20, 30, 40, and 60 min post-infection, aliquots were taken from the 10⁶-fold dilution and subjected to plaque assays by the soft agar overlay method using E. coli B40. Burst size was calculated as plaque-forming units (PFU) produced per infected cell.

Construction of Hfq⁺ T4 Phage

The Hfq⁺ T4 phage was generated using phage CRISPR-mediated gene insertion, as described previously⁵⁷. Briefly, E. coli B40 cells were co-transformed with a donor plasmid containing E. coli hfq gene flanked by T4 genomic sequences in the SegF region and a CRISPR-Cas12a-SegF spacer plasmid. The cells were infected with T4-Soc-del phage. Recombinant phages were selected by counter-selection in E. coli B40 expressing the Cas12a-SegF spacer. PCR analysis confirmed successful insertion of the hfq gene under the control of Soc promoter. To verify Hfq expression by the constructed hfq⁺ T4 phage, E. coli NM22565 (hfq⁻) cells were grown to mid-log phase in LB medium at 37 °C ( ~ 10⁸ CFU/mL) and infected with the Hfq⁺ phage at an MOI of 5. After 5 min of incubation at 37 °C, the culture was super-infected with an additional dose of hfq⁺ phage (MOI 5) to prevent lysis. At 30 and 60 min post-infection, ~10⁸ CFU of bacteria were collected for SDS-PAGE and Western blotting using Hfq polyclonal antibodies (see below). The data show that Hfq is well-expressed from hfq⁺ phage during phage infection.

Hfq western blotting

E. coli strains [B40, NM22565 (Hfq-plus) and NM22540 Dhfq::cat-sacB (Hfq-minus)] were grown to mid-logarithmic phase in LB medium at 37 °C with shaking at 250 rpm. Bacterial cells (approximately 10⁸ CFU) were harvested by centrifugation at 5000 x g for 5 min, resuspended in 200 µL 1X SDS-PAGE loading buffer (50 mM Tris-HCl pH 6.8, 2% SDS, 10% glycerol, 1% β-mercaptoethanol, 0.02% bromophenol blue), and denatured by boiling for 10 min. The samples were then electrophoresed on a 4-20% gradient Mini-PROTEAN TGX precast gel (Bio-Rad) and transferred to nitrocellulose membranes using a Trans-Blot Turbo transfer system (Bio-Rad). Membranes were blocked with 5% BSA in PBS-T (PBS containing 0.05% Tween-20, pH 7.4) for 1 h at room temperature with gentle shaking, then incubated overnight at 4 °C with anti-Hfq rabbit polyclonal antibodies (provided by Dr. Susan Gottesman from National Institutes of Health) at 1:1000 dilution in 5% BSA/PBS-T. After five 5-minute washes with PBS-T, membranes were incubated with HRP-conjugated goat anti-rabbit secondary antibody (1:10,000, Abcam) for 1 h at room temperature. Following five additional PBS-T washes, stained protein bands were visualized using an enhanced chemiluminescence substrate (BioRad) and imaged using a Bio-Rad Gel Doc XR+ imaging system with exposure times optimized for each blot.

Binding of gp13/gp14/Hfq complex to in vitro packaged heads

The empty heads were purified from 10am13am-9DE mutant phage infected cells as described previously⁹. The heads were packaged with the linearized ~6 kb AAV-GFP plasmid DNA in a defined packaging system containing the purified packaging motor protein (gp17) and ATP according to the protocol described earlier⁵⁸. After packaging for 5 min, the gp13/gp14/Hfq complex purified as above was added to the packaging reaction mixtures at a molar 3:1 ratio of neck complexes to packaging motors. After 40 min incubation, the packaged heads were separated from empty heads and unbound complexes by CsCl density gradient centrifugation. The DNA-full heads that migrated to near the bottom of the tube were extracted by puncturing the side of the tube and dialyzed first against 50 mM Tris-HCl pH 7.4 buffer containing 5 mM MgCl₂ and 200 mM NaCl for 1 h and then for another 1 h against the same buffer containing 100 mM NaCl. The packaged head-neck complexes were centrifuged at 36,000 x g for 1 h and resuspended in 50 mM Tris-HCl pH 7.4 buffer containing 5 mM MgCl₂ and 50 mM NaCl. The samples were then analyzed by cryo-EM and Western blotting using anti-His tag monoclonal antibody (Invitrogen). The motor protein gp17 and each of the three proteins of the gp13/gp14/Hfq complex contain Histag fused to either N- or C-terminus of the recombinant protein.

Cryo-EM sample preparation and data collection

All samples were evaluated by negative stain before preparing cryo-grids. 5 µL purified samples were applied to the glow-discharged copper grids (Electron Microscopy China) for 1 min then blotted with filter paper. The grids were then negatively stained with 2% (w/v) uranyl acetate, blotted and air-dried. Images were recorded with Talos L120C TEM (Thermo Fisher Scientific) equipped with a CCD camera at a nominal magnification of 73,000× or 92,000×, defocus between −0.5 µm and −3.5 µm.

All cryo-grids were prepared with the same condition. Here, we took gp20/gp13/gp14 as an example and more details are shown in the Supplemental Information. Quantifoil holey carbon grids (R20/20 Ni-Ti-Au300) were glow-discharged under the atmosphere of argon and oxygen mixture. 3 µL sample containing 0.05 % β-OG was loaded on the grids, and then blotted and vitrified using a Vitrobot Mark IV (Thermo Fisher Scientific). All cryo-EM data were collected at the Center for Cryo-Electron Microscopy (Fudan University) with Titan Krios TEM (Thermo Fisher Scientific) operated at 300 kV, equipped with BioQuantum energy filter (Gatan).

Cryo-EM images processing

For gp13/gp14/Hfq complex, a total data set of 12,169 images were collected with the defocus range of −1.2 μm to −2.2 μm, in super-resolution mode (magnification 81,000×) using K3 Summit direct electron detection device with a physical pixel size of 1.064 Å/pixel. The total exposure time was 3s with 40 frames giving an accumulated dose of 58.3 e^-/Ų. Automated data acquisition was performed with SerialEM software through the beam-image-shift method⁵⁹. More details are listed in Supplementary Table 1. All super resolution images were binned 2, dose weighted, and motion corrected using MotionCor2⁶⁰ and subsequent CTF (contrast transfer function) of micrographs was estimated using Gctf⁶¹. Bad images were excluded upon ice condition, defocus range and estimated resolution. Remaining 11,557 good images were imported into cryoSPARC⁶² for further patched CTF-estimating, blob-picking and 2D classification. Good 2D classes were selected as the template for template picking. From the 2D classes, good particles from blob-picking and template-picking were merged and deduplicated. The whole particle stacks were separated into subsets to accelerate processing. 3D classification in Relion of the first subset of 1,873,444 particles shows two major conformations: gp13/gp14/Hfq and gp13/gp14, these two conformations were selected and merged and after another round of 3D classification, followed by 3D auto-refinement with C6 symmetry, CTF Refinement and Bayesian Polish, a map of 2.79 Å gp13/gp14/Hfq was obtained from 381,536 particles. The whole dataset was used for iterative 3D classification to yield a 3.37 Å map of gp13/gp14 complex.

The gp20/gp13/gp14/Hfq dataset was collected at magnification 130,000x using K2 Summit direct electron detection device with a physical pixel size of 1.046 Å/pixel, total exposure time 8s, 36 frames, at an accumulated dose of 53 e⁻/Å. A total of 4252 images were collected and 4136 good ones were selected. Particles were Laplacian and templated autopicked in Relion, and 297,605 particles were used for iterative 3D classification. The best resolved classes were selected and yielded a 2.91 Å map of gp20/gp13/gp14/Hfq complex.

The reported resolutions are all based on the gold-standard Fourier shell correlation (FSC) 0.143 criterion. All the visualization and evaluation of 3D density maps were performed with UCSF Chimera⁶³. The above procedures of data processing are summarized in (Supplementary Figs. 1, 3 and 6). These sharpened maps were generated by DeepEMhancer⁶⁴.

Model building and refinement

Model building of gp13/14 was performed de novo in COOT⁶⁵, while Hfq was fitted and refined using previous crystal model (PDB:1HK9). After structure refinement in PHENIX⁶⁶, the model was fit into the gp20/gp13/gp14/Hfq map. Gp20 was built from the crystal model (PDB:3JA7) and manually adjusted in COOT. Statistics associated with data collection, 3D reconstruction and model refinement can be found in Supplementary Table 1.

Interface and dissociation energy analysis using PDBePISA

To analyze the protein–protein interfaces and assess the thermodynamic stability of the complexes, we utilized the PDBePISA (Protein Interfaces, Surfaces and Assemblies) web server (http://www.ebi.ac.uk/pdbe/pisa/). PISA was employed to calculate the buried surface area, dissociation free energy (ΔG^diss), and entropic contribution (TΔS^diss) for each complex. These parameters provide insights into the stability and assembly energetics of the gp13 adjacent subunitss, gp14 adjacent subunits, gp13/gp14, gp14/Hfq, and gp14/gp15 complexes. Hydrogen bonds and salt bridges were identified based on geometric criteria (distance ≤ 4.0 Å), and only interactions meeting these thresholds were included in the final analysis.

Reporting summary

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

Author contributions

L.S., Z.C., and V.B.R. initiated, conceived, and supervised the project. L.H. and Q.M. performed the cryo-EM structural study with assistance from X.J., X.D.W., Z.L., J.Z., and X.R.W. performed the in vivo phage infection assays and constructed mutant phages. N.U. performed in vitro packaging and head-neck complex assembly experiments. L.H., L.S., and V.B.R. co-wrote and edited the manuscript. Q.F. and A.F. edited the manuscript. V.B.R. provided overall direction and coordination for this and the accompanying project.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Coordinates and electron density maps associated with data reported in this manuscript have been deposited in the Electron Microscopy Data Bank (EMDB). The C6 symmetric reconstructions of the gp13/gp14 have been deposited with the accession codes EMD-63387. The C6 symmetric reconstructions of the gp13/gp14/Hfq have been deposited with the accession codes EMD-63388. The C6 symmetric reconstructions of the gp20/gp13/gp14/Hfq have been deposited with the accession codes EMD-63389. The C6 symmetric reconstructions of the gp20/gp14/gp13 have been deposited with the accession codes EMD-63390. The asymmetric atomic structure of the neck (gp13/gp14) has been deposited in the Protein Data Bank (PDB) with the accession code 9LU4. The asymmetric atomic structure of the neck (gp13/gp14/Hfq) has been deposited in the Protein Data Bank (PDB) with the accession code 9LU5. The asymmetric atomic structure of the portal-neck (gp20/gp13/gp14/Hfq) has been deposited in the Protein Data Bank (PDB) with the accession code 9LU6. The asymmetric atomic structure of the mismatched portal-neck (gp20/gp14/gp13) has been deposited in the Protein Data Bank (PDB) with the accession code 9LU7. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-69107-7.