Structure of a Brochothrix thermosphacta bacteriophage reveals cell wall adsorption mechanism in Gram-positive infecting siphophages

Abstract

Siphophages possess a long, flexible, and non-contractile tail that is responsible for host recognition, cell wall perforation, and genome delivery. Although the majority of structurally characterized siphophages target Gram-negative bacteria, those infecting Gram-positive bacteria remain elusive. Moreover, structural information concerning phage infection and genome release events in Gram-positive infecting siphophages is sparse. Here, we present a near-atomic resolution structure of Brochothrix thermosphacta bacteriophage NF5 determined by cryo-electron microscopy (cryo-EM). The structure comprises 11 proteins associated with the head, neck, tail tube, and baseplate, amounting to 643 polypeptides in total. Integration of cellular cryo-electron tomography (cryo-ET) showed the infection process of NF5, providing insights into the adsorption mechanism of siphophages targeting Gram-positive bacteria. Structural comparisons of baseplates from multiple siphophages targeting Gram-negative and Gram-positive bacteria reveal divergent compositional architectures and distinct assembly mechanisms. These disparities are evident in the domains across divergent baseplate proteins, suggesting evolutionary adaptations to host envelope architectures.

Mechanisms of siphophage infection in Gram-positive bacteria remain poorly understood. Here, the authors report the high-resolution structure of phage NF5 and visualize its cell wall adsorption and genome delivery process.

Introduction

Bacteriophages (or phages) are efficient candidates for a variety of biotechnological applications due to their host specificity and abundance of phage-encoded proteins^1,2. The majority of known phages belong to the order Caudovirales and possess a tail apparatus^3,4. Gene sequencing indicates that siphophages are the most abundant members of Caudovirales ⁵. Siphophage structure is characterized by a long flexible, non-contractile tail, attached via a neck (connector) composed of one or more proteins and a dodecameric portal to the genome-filled capsid⁵. The tail is vital for host cell recognition, cell wall perforation, and delivering of the viral genome into the host cytoplasm⁶.

Recent advancements in data collection and processing have enabled high-resolution characterization of several siphophage structures⁷, including R4C⁸, Lambda⁹, JBD30¹⁰, chi¹¹, T5¹², and DT57C¹³. Notably, all of these siphophages infect Gram-negative bacteria. Gram-positive bacteria possess complex surface structures to the human or animal hosts¹⁴. Gram-positive and -negative bacteria differ fundamentally in structure and composition^15,16, and the cell wall of the former possesses a thick peptidoglycan layer, whereas the latter is characterized by the outer membrane structure¹⁷. However, the adsorption process of Gram-positive infecting bacteriophages remains comparatively understudied relative to those infecting Gram-negative bacteria^16,18. To date, the only near-atomic resolution structure of a Gram-positive infecting siphophage is that of Mycobacterium smegmatis phage Bxb1¹⁹. Numerous medium- or high-resolution structures of isolated, monomeric proteins and recombinant tails of Gram-positive siphophages and tail-like machines, including Bacillus subtilis phage SPP1²⁰, Staphylococcus aureus phage 80α²¹, Lactococcal phage Tuc2009²², Listeria phage PSA²³, and tail-like bacteriocin monocin²⁴, have been determined by X-ray crystallography and cryo-EM. Additionally, low-resolution intact structures of Gram-positive siphophages, including Lactococcal phage 1358²⁵, p2²⁶, and TP901-1²⁷, have been reported. However, the structures of siphophages infecting Gram-positive bacteria at high-resolution remain elusive, and the infection process between siphophages and Gram-positive bacteria is poorly understood.

Brochothrix thermosphacta is a Gram-positive bacterium implicated in meat and meat product spoilage²⁸. Storage conditions for these foods favor the growth of B. thermosphacta at low temperature²⁹. B. thermosphacta is phylogenetically closely related to Listeria, belonging to the family Listeriaceae³⁰. Recently, extensive research has focused on the use of phages to control specific foodborne pathogens^31,32, with B. thermosphacta and Listeria phages emerging as potential agents for controlling food spoilage^33,34. Previous studies have classified 21 strains of B. thermosphacta phages using electron microscopy³⁵. However, to our knowledge, no structural characterization of these viruses exists. NF5 is a temperate phage infecting B. thermosphacta HER1188 and represents a group containing 14 B. thermosphacta phage strains³⁰.

Here, we present a near-atomic resolution structure of bacteriophage NF5 determined by cryo-EM. Our structure enables us to build the atomic models of NF5. The neck–tail complex is organized into a continuous assembly comprising the 12-fold adaptor (gp8), hexameric stopper (gp9), tail terminator (gp11), and tail tube (gp12) rings, followed by the threefold symmetric tape measure protein (gp15) and a baseplate. The baseplate consists of hexameric distal tail proteins (gp16), the threefold symmetric tail-associated lysin (gp17), and six trimers of the receptor binding protein (gp18) and side fiber protein (gp19). This complex connects the icosahedral capsid, built from the major capsid protein (MCP) gp7, to the 12-fold (gp3) portal. Using cellular cryo-ET after cryogenic focused ion beams (cryo-FIB) milling enables elucidation of the initial stages of NF5 infects of B. thermosphacta. Our structure supports previous hypotheses and provides structural insights into the molecular events involved in the adsorption of Gram-positive infecting siphophages. Furthermore, structural comparisons of baseplate proteins from multiple siphophages infecting Gram-negative and Gram-positive bacteria reveal distinct compositional and assembly mechanisms, with domain-level divergences in baseplate proteins suggesting evolutionary adaptations to host envelope architectures.

Results

Overall structure of bacteriophage NF5

NF5 was purified from B. thermosphacta for cryo-EM data collection. We collected 19,681 cryo-EM movies of NF5 (Supplementary Figs. S1). Applying icosahedral and C5 symmetry, we determined the structures of NF5 icosahedral capsid and its fivefold region at resolutions of 3.56 Å and 3.39 Å, respectively (Supplementary Figs. S1–S3). We manually selected particles corresponding to the portal, neck, tail tube, and baseplate regions, resolving their structures to resolutions of 3.46, 3.36, 3.6, and 3.44 Å by imposing C12, C6, and C3 symmetry, respectively (Supplementary Figs. S1–S3). Based on the baseplate structure, coordinates were re-centered to regions below and surrounding the baseplate, improving resolution local structures for the tail-associated lysin and receptor binding protein regions at 3.45 Å and 3.76 Å by imposing C3 and C1 symmetry, respectively (Supplementary Figs. S1–S3). These high-resolution cryo-EM maps allowed us to build the atomic models for the majority of NF5 proteins, including the MCP gp7, portal protein gp3, adaptor protein gp8, stopper protein gp9, tail terminator protein gp11, tail tube protein (TTP) gp12, C-terminus (C-ter) of tape measure protein (TMP) gp15, distal tail protein (DTP) gp16, tail-associated lysin (Tal) gp17, receptor binding protein (RBP) gp18 and N-terminus (N-ter) of side fiber protein (SFP) gp19 (Fig. 1a, b). However, the N-terminus of TMP gp15, the C-terminus of SFP gp19, and the C-terminus of Tal gp17 were not resolved, likely due to their flexibility. Resolutions estimates used the Fourier shell correlation criterion with a cut-off of 0.143 according to the gold standard method³⁶. Data acquisition and reconstruction statistics are presented in Supplementary Table S1.

Fig. 1. Overall structure of B. thermosphacta bacteriophage NF5.

a Side (left) and cut-open (right) views of the intact structure of bacteriophage NF5. The inset cryo-micrograph shows a representative particle of NF5 (out of 64,549), with a scale bar indicating 50 nm. b Schematic diagram of the NF5 genome segment encoding structural genes. The color code of the open reading frames (arrows) corresponding to the structural proteins is applied to (a). MCP major capsid protein, TTP tail tube protein, TMP tape measure protein, DTP distal tail protein, Tal tail-associated lysin, RBP receptor binding protein, SFP side fiber protein.

The NF5 head is composed of 415 copies of the MCP gp7 (Fig. 1a, b). One pentamer of the icosahedral head is replaced by the portal complex and the neck, comprising the dodecameric portal protein gp3, the adaptor protein gp8, and the hexameric stopper protein gp9 (Fig. 1a, b). The NF5 tail, ~135 nm long, comprises the hexameric tail terminator protein gp11, 24 stacked discs of TTP gp12 hexamer, a threefold symmetrical TMP, and the baseplate (Fig. 1a, b). The core of the baseplate contains the hexamer of DTP gp16, the trimeric Tal gp17, and the C-ter of the TMP gp15 trimer, which fills the tail channel and extends to the Tal protein (Fig. 1a, b). The periphery of the baseplate comprises six trimers of RBP gp18 and the SFP gp19 (Fig. 1a, b).

Structure of NF5 capsid

The NF5 capsid exhibits a T = 7 icosahedral lattice with a longest diameter of 680 Å (Fig. 1a and Supplementary Fig. S4a). One asymmetric unit (ASU) of the NF5 capsid consists of a hexamer and one pentameric subunit (Supplementary Fig. S4b). Similar to the canonical HK97 MCP³⁷, NF5 gp7 contains four structural domains (Supplementary Fig. S4c): N-terminal arm (N-arm, residues 2–32), peripheral domain (P-domain, residues 33–44, 80–127, 223–258), extended loop (E-loop, residues 45–79), and axis domain (A-domain, residues 128–222, 259–269). Superimposing hexameric and pentameric monomers within an ASU shows a range of quasi-equivalent conformations, primarily in the N-arms and E-loops (Supplementary Fig. S4)^38,39. Furthermore, the A-domain shows distinct variations between the hexon and penton MCPs (Supplementary Fig. S4d), suggesting MCP adaptation to the diameter and curvature in different positions³⁹.

Adjacent MCPs in hexon and penton interact through complementary electrostatic interactions (Supplementary Fig. S5a), a conserved assembly mode among phages. Interactions between the N-arm, E-loop, and P-domain of the NF5 MCPs stabilize capsomer contacts at the threefold and pseudo-threefold axis (Supplementary Fig. S5b). Unlike the p-lock interactions in phage JBD30¹⁰ or the P-loop in phage R4C⁸, HK97³⁷, and phage-like particle RcGTA⁴⁰, salt bridges mediate interactions among the P-domains of three hexameric monomers on the threefold axis and between hexameric and pentameric monomers in NF5, while the N-arm and E-loop of the adjacent monomers also interact with these P-domains (Supplementary Fig. S5c). On the pseudo-twofold axis, similar to R4C⁸ and JBD30¹⁰, anti-parallel N-arms from two different hexameric or pentameric monomers interact with each other (Supplementary Fig. S5d, e).

Structures of the portal and the neck

The portal complex (12 × gp3) and neck occupy one of the twelve vertices of the icosahedral capsid (Fig. 1a). High-resolution cryo-EM maps resolve two components of the NF5 neck (Fig. 2a): the adaptor complex (12 × gp8), and the stopper complex (6 ×  gp9). We built the atomic model for NF5 portal protein gp3 lacking 3 N-terminal residues and 27 C-terminal residues. NF5 gp3 exhibits the canonical portal fold of Caudovirales phages^20,41,42, divided into four domains (Fig. 2b): wing domain (residues 4–222, 312–388), stem domain (residues 223–243, 296–311), clip domain (residues 244–295), and crown domain (residues 389–429). The outer surface of the wing domain contacts the capsid (Fig. 2a). The tunnel loop on the inner surface of the wing domain forms the narrowest channel of the portal, ~25 Å in diameter (Supplementary Fig. S6a). Similar structures are considered to stabilize the dsDNA and prevent leakage during maturation (e.g., phages T4⁴¹ and SPP1⁴³). The clip domain provides binding sites for the adaptor complex (Fig. 2a and Supplementary Fig. S6b).

Fig. 2. Structures of the NF5 portal, neck and tail tube.

a Overall view of the atomic models (ribbon) of the portal-neck and tail tube complex. The grey maps at the outside are the capsid. Only 3 rings of tail tube are shown, and the first and the other two rings are colored cyan, light blue, respectively. b–d Ribbon models of portal protein gp3 (b), adaptor protein gp8 (c), and stopper protein gp9 (d). All proteins are colored according to their domains. e, f Ribbon models of terminator protein gp11 (e) and tail tube protein gp12 (f) located in the middle layer of tail tube, shown in rainbow colors ranging from blue at the N-terminus to red at the C-terminus. g The atomic models of TTP gp12 superimposed among the proximal, middle, and distal rings showing conformational variations in both the C-arms and E-loops. h–j Zoomed-in views of the boxes in (a) showing the interactions between the C-arm of proximal gp12 and gp11 through π–π stacking and salt bridges (h), the interactions between the C-arm of middle gp12 and proximal gp12 (i), and the interactions between two adjacent rings of TTP (j), respectively.

The NF5 adaptor complex consists of 12 copies of gp8, with residues 2–115 modeled (Fig. 2a). Each gp8 monomer is composed of an α-helix bundle, a β-hairpin, and an extended C-arm (Fig. 2c), showing high structural conservation with siphophages SPP1, T5, and Lambda^9,12,20. The α-helix bundles stabilize the adaptor complex through electrostatic interactions (Supplementary Fig. S6c), while the β-hairpins from 12 subunits assemble into a β-barrel to reinforce the inter-subunit interactions (Supplementary Fig. S6d). The C-arm of gp8 interacts with two adjacent portal clip domains through β-sheet augmentation (Supplementary Fig. S6b). The bottoms of the α-helix bundle and β-hairpin provide binding sites for the stopper protein (Fig. 2a).

The adaptor is followed by the hexameric stopper protein gp9. According to the homologue designation (g7 from RcGTA and gp16 from SPP1), each gp9 monomer comprises a β-barrel core, a β2-β3 loop, and a β4-β5 loop (Fig. 2d). β-barrel cores from 6 copies of gp9 extend the neck lumen continuously from the adaptor to the stopper. Each stopper protein interacts with two neighboring adaptor proteins through salt bridges, mediating the structural symmetry mismatch between the 12-fold symmetric adaptor and the sixfold symmetric stopper (Supplementary Fig. S6e). Additionally, the C-terminal of gp9, together with two loops, forms a wedge-like surface for tail attachment (Fig. 2a). Consistent with observations in other siphophages^9,10,12, the portal complex and two neck components exhibit complementary electrostatic potential at their interfaces (Supplementary Fig. S7a).

Structure of the tail tube

The NF5 tail tube comprises a hexameric ring of tail terminator protein gp11 and 24 stacked rings of TTP gp12 (Figs. 1a and 2a). Each gp11 monomer consists of an antiparallel β-sheet core and two α-helices, with an O-shaped loop (O-loop, residues 23–37) and an extended loop (E-loop, residues 49–54) (Fig. 2e). Despite low sequence similarity, the structure is conserved among siphophages (e.g., gpU of Lambda⁴⁴, p142 of T5⁴⁵, and gp17 of SPP1²⁰) (Supplementary Fig. S8a, b). The α1, α2, O-loop of gp11 and the β2-β3 loop, β4-β5 loop, and C-ter of two adjacent gp9 mediate terminator-stopper interactions (Supplementary Fig. S9).

A tail tube ring is assembled by six copies of TTP gp12 wrapped around the lumen of the TMP gp15, with the twist and rise between two adjacent rings being approximately 41.9° +/− 0.1° and 41.7 +/− 0.1 Å, respectively (Figs. 1a and 2a). Each gp12 monomer contains two four-stranded β-sandwiches, flanked by an α-helix, with an extended loop (E-loop, residues 39–59) between β3 and β4 (Fig. 2f). The β-sandwich cores of six gp12 subunits form a highly negatively charged surface that may facilitate rapid DNA translocation (Supplementary Fig. S7b). HHpred analysis⁴⁶ shows that gp12 exhibits structural conservation with other siphophage TTPs, despite low sequence similarity (Supplementary Fig. S10). We resolved the structures of gp12 from the proximal, middle, and the distal rings of the tail tube. Superimposing gp12 monomers from different rings reveals conformational variations in both the C-arms and E-loops (Fig. 2g). Notably, the C-arm of the proximal gp12 shifts ~12 Å to contact α1 and α2 of the terminator through π–π stacking and salt bridges (Fig. 2h). In contrast, in other rings of gp12, the C-arms mediate inter-ring interactions (Fig. 2i). Such interactions are observed only in the Gram-positive infecting siphophage 80α²¹ and SPP1⁴⁷ (Supplementary Fig. S11 and Table S2). In Gram-negative infecting siphophage such as lambda, T5, YSD1, and JBD30^10,48–50, the C-ter of TTP includes an additional immunoglobulin (Ig)-like domain (Supplementary Fig. S11). Furthermore, the E-loop from the distal gp12 tilts ~6 Å compared to other TTPs, attributable to interactions with DTP gp16 in the baseplate (see below). Additionally, the interactions between the E-loop of the TTPs and adjacent TTP rings mediate ring-ring connection (Fig. 2j).

Structure of the core region in the NF5 baseplate

The baseplate complex is located at the distal end of the tail (Fig. 1a). As a high structurally diverse complex, the baseplate performs crucial roles in receptor recognition and host adsorption. The core region of NF5 baseplate is composed of a trimer of DTP gp16, a trimer of Tal gp17, and a trimer of TMP gp15 C-ter (Figs. 1a and 3a).

Fig. 3. Structure of the core region in the NF5 baseplate.

a Left and middle: side and cut-open views of the core region in the NF5 baseplate complex; right: side view of its atomic model. The core region of NF5 baseplate is composed of a trimer of DTP gp16, a trimer of Tal gp17, and a trimer of TMP gp15 C-ter. The color codes are identical to those used in Fig. 1a. This map is a composite map, with the cryo-EM maps of C6 TTP, C6 DTP, and C3 baseplate. TTP tail tube protein, DTP distal tail protein, Tal tail-associated lysin, TMP tape measure protein. b Ribbon model of the distal tail protein gp16 shown in rainbow colors. c, d Zoomed-in views of the red and black boxes in (a) showing the interactions between the DTP N-arm and TTP (c), and between the DTP N-arm and RBP (d) via salt bridges, respectively. e Ribbon model of gp17 colored according to its domains. f Slab view of the maps of the partial Tal and C-terminus of TMP. Top right: schematic model of the Tal plug. Bottom left: top-down view of the π–π stacking of the TMP monomers. Bottom right: zoomed-in view of the interactions between TMP and Tal.

DTP gp16 is constituted of two domains: an N-terminal domain (NTD, residues 1–145) with a β-sandwich core and a C-terminal domain (CTD, residues 146–255) with a galectin-like fold (Fig. 3b)^51,52. The NTD contains an N-terminal arm (N-arm, residues 1–23) and an extended loop (E-loop, residues 50–68) (Fig. 3b). The N-arm and β-sandwich of gp16 interact with the distal ring of TTP gp12 to attach the baseplate to the tail tube (Fig. 3c and Supplementary Fig. S12a), and the N-arm also provides binding sites for the RBP trimer (Fig. 3d). On the other side, the E-loop and CTD interact with Tal through electrostatic interactions (Supplementary Fig. S12b, c). NF5 gp16 is topologically identical to DTPs from Gram-positive infecting siphophage Bxb1, 80α, SPP1, TP901-1, and tail-like machine monocin (Supplementary Fig. S13 and Table S2)^{19,21,24,51,53}. However, in 80α DTP, additional domains (tail binding loop and RBP binding loop) connect to the tail tube and RBP, respectively²¹. In TP901-1⁵³ and Tuc2009²², these phages encode baseplate upper protein (BppU) to anchor the RBP. Thus, different assembly strategies of baseplate are employed despite structural similarity.

Cryo-EM maps of the baseplate and the Tal region allowed us to build an atomic model of residues 3–590 of 915-residue-long gp17 (Fig. 3e). Located at the distal end of the baseplate, gp17 is designated as the tail-associated lysin (Tal) in Gram-positive infecting siphophages^21,54, and its NTD contains hub domain (HD) I–VI, structurally conserved in baseplate hub proteins of siphophages and myophages (Fig. 3e, Supplementary Fig. S14a, and Table S2)⁶. An 8-stranded β-sheets formed by HD1 and HD3 extends the lumen to the end of the tail tip (Fig. 3a and e). Residues 350–390 form two α-helices that extend into the tail lumen, assembling as a trimeric plug to seal it (Fig. 3f). Residues 391–590 contain multiple antiparallel β-sheets forming a trimeric β-helix domain extending outward from baseplate (Fig. 3e and Supplementary Fig. S14b). This fold is common in fiber proteins and fiber-like proteins (Supplementary Fig. S14b)^55–57, suggesting that in NF5, the β-helix domain of Tal may be involved in host recognition and adhesion. Residues 591–915 of the gp17 CTD were not resolved due to flexibility. HHpred results reveal that residues 600–915 of gp17 share highly similarity (probability = 99.3%) with gp13 of phi29, a protein distal to the phi29 knob that functions as a peptidoglycan-degrading enzyme (PDBID: 3CSQ) (Supplementary Fig. S14c)⁵⁸. Thus, similar to gp13 in phi29, the CTD of NF5 gp17 may facilitate host cell penetration for DNA ejection⁵⁸.

In the tail lumen, additional density exists above the plug formed by gp17 α-helix bundle. This tripod-shaped density is constrained by the plug in the distal of tail lumen (Fig. 3f). Modeling demonstrated that this density belongs to residues 648–674 of the C-terminus of TMP. TMP is a multifunctional protein in siphophages and myophages that typically fills the tail lumen and is ejected during phage genome release, creating a channel for genome delivery^48,59. In mature NF5 particles, it is retained in the tail lumen by the Tal plug to prevent premature DNA release (Fig. 3f). The C-terminus of TMP stabilizes the trimer by π–π stacking and is anchored to the end of the tail tube lumen by electrostatic interaction with Tal (Fig. 3f).

Structure of the peripheral region in the NF5 baseplate

The peripheral region of the NF5 baseplate is composed of six trimers of RBP gp18, and six trimers of SFP gp19. The structure of it measures 30 nm in length, with a width of 39 nm and a lumen of 12.5 nm (Figs. 1a and 4a, f). We resolved residues 2–619 of the 626-residue-long gp18, which shows structural homology to receptor binding proteins in Gram-positive infecting sipho- and podophages^21,60 (Supplementary Table S2). Gp18 can be divided into three domains: stem domain, platform domain and tower domain (Fig. 4b). The stem domain forms three coiled coils connected by a flexible hinge undergoing a ~ 140° bend (Fig. 4b). The hinge angle differs significant from other phage, indicating its flexibility (Fig. 4c). This flexibility may enhance the function of phage-receptor binding, as proposed for S. aureus phage⁶¹. The Platform domain contains a classical five-bladed β-propeller fold with receptor binding sites⁶⁰. Despite structural homology, the receptor binding sites of NF5 differ from those of S. aureus phage, likely reflecting distinct host receptor requirements (Supplementary Fig. S15a). The tower domain is composed of two similar subdomains (tower1, tower2), each containing a four-stranded antiparallel β-sheet (Fig. 4b). The fold of each subdomain is homologous to putative major teichoic acid biosynthesis protein C and muramidases (Supplementary Fig. S15b), suggesting the tower domain may be involved in cell wall peptidoglycan degradation. Deleting the RBP in phage 80α abolished infectivity²¹. Thus, NF5 gp18 plays a vital role in host recognition and binding. Interactions of the NF5 gp18 stem domain with the DTP and the distal TTP anchor RBP to the NF5 baseplate (Fig. 3d and Supplementary Fig. S15c).

Fig. 4. Structure of the peripheral region in the NF5 baseplate.

a Left and middle top: side and top-down views of the peripheral region in the NF5 baseplate complex; right and middle bottom: side and top-down views of its atomic model. The color codes are identical to those used in Fig. 1a. SFP side fiber protein, RBP receptor binding protein. b Atomic model of an RBP trimer. Stem domain is composed of 10 nm and 5 nm segments. The tower, platform, and stem domains of one gp18 monomer are colored in red, green, and cyan, respectively, and the other two gp18 monomers are colored in grey. c Superimposition of NF5 RBP gp18 with the tail fiber (gp17) of P68 (PDBID: 6IAB) showing the flexibility of gp18 hinge. d Zoomed-in view of the box in (a) showing an SFP trimer is located at the top of an RBP trimer. e Top: atomic model of an SFP trimer. The N-terminus and α-helix domains of one gp19 monomer are colored in violet, and the other two gp19 monomers are colored in grey. Bottom: top-down view of the atomic model of six SFP trimers. f Left: The cryo-EM reconstruction of the baseplate at a resolution of 10 Å shows six rod structures spread around the baseplate, which are colored violet. The color of RBP is identical to Fig. 1a, and other proteins are colored grey. This structure is generated by low-pass filtering baseplate map to 10 Å. Right: predicted SFP atomic model fit well into the rod structure. g Predicted structure of SFP gp19 trimer after manual fitting, shown using different colors for its domains.

Each SFP trimer is composed of three gp19 subunits and located at the top of the RBP trimer (Fig. 4d). Due to the density beyond residue 162 was missing at high resolution, we resolved only residues 3–162 of the 639-residue gp19 (Fig. 4e). The low-resolution (10 Å) cryo-EM map of the baseplate revealed six discontinuous rod structures around the baseplate, into which the Alphafold3⁶² predicted structure of gp19 could be approximately fitted (Fig. 4f). Gp19 comprises an NTD, three tower domains, and a receptor binding domain, with the N-ter domain is connected to the tower domains by an α-helix domain (Fig. 4e and g). The N-ter contains a β-sandwich Ig-like domain, a fold also found in baseplate proteins p132 of Escherichia coli phage T5⁶³ and BppU of Lactococcus lactis phage TP901-1⁵³, serving to anchor the fiber to the baseplate (Supplementary Fig. S16a). At the C-terminus of SFP, the receptor binding domain is assembled from antiparallel β-hairpins (Fig. 4g). HHpred results reveal that it is homologous to the multiple Gram-positive infecting phage RBPs (98.5% probability to gp15 of PSA, 94.8% probability to ORF49 of TP901-1, 94.8% probability to RBP of p2) (Supplementary Fig. S16b and Table S2). Therefore, NF5 SFP may fulfill the combined functions of BppU and RBP, demonstrating structural diversity among Gram-positive infecting phages. The SFP trimers interact specifically with RBPs, mediated by the interactions of the two lower monomers near RBPs with the stem domain of RBP (Supplementary Fig. S16c). The SFPs together with the RBPs form a host recognition apparatus, anchored to the baseplate by the RBP stem (Fig. 4a and d and Supplementary Fig. S16c).

Cryo-electron tomograms reveal the initial stages of NF5 infection

To investigate the process of NF5 infecting host cells, we inoculated purified NF5 into B. thermosphacta NF4 cells. The cells were plunge-frozen, and cryo-FIB milling was used to generate lamellae with a thickness of ~250 nm. In the tomograms, we observed two distinct dense layers corresponding to the cytoplasmic membrane and the cell wall, with a lower density zone between them that may represent the cell periplasm (Fig. 5a). The combined thickness of the cell wall and periplasm is ~45 nm, which is consistent with previous studies on Listeria⁶⁴. Most phage particles were observed adsorbing to the host cells, with their baseplates attached to the cell wall (Fig. 5a and Supplementary Fig. S17).

Fig. 5. Cryo-electron tomograms of NF5 infection and proposed genome delivery mechanism.

a Tomographic reconstruction of B. thermosphacta cell infected by NF5. CW: cell wall; CM: cytoplasmic membrane. Triangular symbols of different colours in (a–e) denote different density level of the phage particles capsid. Green: full density, blue: weak density, and red: empty. The Scale bars in (a–e) are 20 nm. b Stage 1: initial adsorption of baseplate to the host cell wall. c Stage 2: insertion of the baseplate onto the cell wall, and the SFP undergoes a conformational change compared to its mature state. The purple lines indicate the possible density of SFPs. d Stage 3: the NF5 capsids show weak density, indicating process of genome ejection. Different sections denote distinct 2D tomographic slices viewed using 3dmod. e Stage 4: the NF5 capsid and tail lumen show empty density, indicating the end of genome ejection. f Schematic diagram summarizing the proposed mechanism of the NF5 host adsorption and genome delivery. The color codes of the NF5 proteins are identical to those in Fig. 1a. The red arrows indicate the possible movement directions of the corresponding elements. a–e display multiple tomographic slices, each selected to reveal specific structural features.

Based on the baseplate position and the genome density, we classified these phage particles into several categories and speculate that these correspond to different stages of phage infection (stages 1–4) (Fig. 5b–f). In stage 1, the baseplate was observed in a tilted orientation contacting with the cell wall (Fig. 5b). We propose that this may according to the contact of peripheral SFPs and RBPs with the host receptors. This has also been hypothesized in previous studies²¹. In stage 2, the baseplate becomes nearly oriented perpendicularly to the cell wall and moves closer to it compared to the stage 1 (Fig. 5c). Particles in stages 1-2 retain the full-density genome (Fig. 5b, c). In stage 3, the phage genome within the capsid exhibit weak density, leading us to speculate that the phages are undergoing genome release (Fig. 5d). In stage 4, both the capsid and the tail lumen appear empty, indicating that the phage particles at this stage are consistent with the genome-released state (Fig. 5e). In stages 3 and 4, a potential channel-like structure is observed within the periplasm zone, with a width of ~3.6 nm (Fig. 5d, e). As previously studies proposed⁶⁵, the contact of the Tal C-terminus with the cell wall and its enzymatic activity may trigger Tal conformational changes. These structural changes facilitate TMP release from the tail tube, formation of a channel, and concomitant genome release into the cytoplasm⁶⁵. TMP has been proposed in previous studies to form a channel across the bacterial envelope, thereby preventing degradation of the phage genome by periplasmic endonucleases^59,66. Therefore, we propose that the channel-like structure observed in the tomograms may correspond to the TMP. Moreover, in contrast to the SFP conformation observed in mature particles (Fig. 4f), the SFPs in stages 2–4 appear to adopt an open state (Fig. 5c–e), implying that conformational rearrangements of the SFPs occur during infection (Fig. 5f).

Discussion

Here, we determined the near-atomic resolution structure of bacteriophage NF5 using single-particle cryo-EM. Our structure reveals a distinct baseplate architecture compared to Gram-negative infecting siphophages (Supplementary Table S2). Furthermore, by applying cryo-FIB and cellular cryo-ET to phage-infected cells, we characterized different stages of NF5 infection. Our findings provide advanced understanding of possible molecular events during adsorption in Gram-positive infecting siphophages, including initial host recognition and baseplate adsorption, as well as baseplate conformational changes and transmembrane channel formation, which facilitate DNA delivery.

The NF5 capsid lacks additional decoration protein and covalent cross-linked chain, which are common in the HK97 capsid³⁷. Capsid stability of NF5 relies solely MCP interactions. The NF5 capsid size is average among T = 7 capsids (diameter, 60–65 nm), similar to marine phage R4C⁸. However, NF5 encapsulates a smaller genome (36 kb) than phage T7 (40 kb), Lambda (48 kb), and P47-26 (83 kb)^67,68, all of which possess T = 7 capsids. Consequently, NF5 exhibits lower genome packaging density. Notably, phage with higher genome density typically possess decoration or cement proteins that stabilize the capsid^68,69. Thus, the reduced genome density and internal pressure in NF5 may be correlate with a simpler capsid assembly mode lacking covalent cross-links and decoration proteins.

The NF5 DTP connects the baseplate to the tail tube, while the RBP and SFP are anchored to the baseplate via RBP-DTP interactions. Thus, the DTP folding is pivotal for NF5 baseplate assembly. Comparing NF5 gp16 with other siphophage DTPs suggests: (1) DTPs in structural characterized Gram-positive infecting siphophages and phage tail-like machines typically contain a galectin-like domain, rarely observed in Gram-negative infecting siphophages (Supplementary Fig. S13). This domain interacts with the baseplate hub protein or Tal in Gram-positive phages, indicating different interactions in baseplate between Gram-positive and Gram-negative infecting siphophages. (2) DTPs of Many Gram-positive infecting siphophages and phage tail-like machines possess an extending N-arm toward the capsid (e.g., NF5, monocin, TP901-1, and Bxb1) or an additional domain (e.g., the RBP binding loop in 80α; Supplementary Fig. S13). Conversely, the DTPs of Gram-negative infecting siphophages lack both N-arms and additional domains (Supplementary Fig. S13). These domains mediate interactions with host adsorption apparatus (RBP, fiber, and BppU) in the Gram-positive infecting phages, further suggesting different architecture in host adsorption apparatus between siphophages infecting Gram-positive and Gram-negative bacteria. Based on these structural comparisons, we propose that the structural differences in bacterial cell walls uncover the divergence of key structural proteins between these two groups of siphophages. Moreover, these differences reveal the host specific adaptations of structural proteins in Gram-positive siphophages.

In Gram-positive infecting phage TP901-1⁵³, PSA²³, 80α²¹, and tail-like machine monocin²⁴, phage-encoded BppU (FtbN in monocin) and RBP (FtbP in monocin) form an oligomeric receptor-recognition device, anchored to DTP by BppU (FtbN). Comparisons with the NF5 SFP reveal that the N-terminal region and part of the α-helix of the NF5 SFP are highly similar to these BppUs (FtbN) (Supplementary Fig. S18). Moreover, the C-terminal region of the NF5 SFP is highly similar to these RBPs (FtbP) (Supplementary Fig. S18). Therefore, we hypothesize that NF5 SFP combines the functions of the BppU and the RBP found in the phages described above. This phenomenon has also been observed in Gram-negative infecting siphophage T5⁶³ and Lambda⁹, where proteins gpL, gpI, and the N-terminal of protein gpJ in Lambda correspond to the baseplate hub protein pb3 in T5⁹. This domain insertion/deletion and distribution in receptor recognition devices persists despite differences in tail components between Gram-positive and Gram-negative infecting phages, implying the potential evolutionary diversity phages and tail-like machines have developed to adapt to disparate host environments (Supplementary Fig. S19).

The tomograms provide a visual representation of the NF5 infection process of B. thermosphacta (Fig. 5a–f). The stage 1 is initial adsorption of RBP and SFP onto the host cell wall (Fig. 5b and f). According to 80α²¹ and our structure of stage 2–4, Tal requires perpendicular orientation to the cell wall for degradation activity. Thus, Tal needs also recognize and adsorb to the host cell wall. The β-helix domain of Tal, homologous to the CTD of other phage fibers, may mediate Tal localization to the cell wall (Figs. 3e and 5f and Supplementary Fig. S14b). As proposed for 80α²¹ and Bxb1¹⁹, once Tal degrade peptidoglycan, the baseplate needs to approach the cell membrane. Our structure shows that flexible hinges in RBP and SFP allow significant tilting, facilitating baseplate movement (Fig. 4b, g). The SFP angle in the stage 2–4 differs significantly from that in the natural state (Fig. 5c–e). Similar fiber conformational changes occur during mycobacteriophage Bxb1 infection¹⁹.

We predicted the structure of NF5 Tal using AlphaFold3 (Supplementary Fig. S20)⁶². Comparing our Tal structure with the prediction shows that HD I–IV fit well, except for the α-helix, β-helix, and CTDs (Supplementary Fig. S20a). However, the segmented β-helix domain in the predicted structure fits well into our Tal structure (Supplementary Fig. S20b). Thus, the deflection of the α-helix bundle and CTDs in the predicted structure likely results from tilt in the linker between the α-helix bundle and HD III (Supplementary Fig. S20c). The difference may reflect Tal conformational change from the monomer solution state to the assembled state. We infer that the NF5 baseplate undergoes structural changes analogous to tail tip rearrangements in T5 and Lambda^9,63,70. Moreover, the trimeric α-helix bundle forms the Tal plug. DNA and TMP release require plug opening, facilitated by its flexibility. In Bxb1, the BHP adopts an open state upon infection¹⁹. In TP901-Tal⁶⁵, α-helix mutations facilitate Tal rearrangements leading to TMP exit, necessary for DNA release. These results support our structural analyses. Combined with our structure and previous models^21,65, NF5 Tal likely undergo conformational changes in the host cell wall analogous to those of T5⁶³ and lambda⁷⁰ at the host outer membrane (Fig. 5f). During infection, contact between NF5 Tal and the host cell wall may trigger a conformational change in Tal, thereby disrupting the plug–TMP interactions (Fig. 3f) and leading to Tal opening that facilitates the release of TMP and DNA (Fig. 5f).

In summary, our combined cryo-EM and cryo-ET analyses propose a possible mechanism for cell wall adsorption and genome delivery of Gram-positive infecting siphophages, providing a structural foundation for understanding the infection biology of Gram-positive bacteria-phage. The proposed model is reconstructed from a series of static cryo-ET snapshots; however, the near-atomic resolution cryo-EM analysis provides complementary structural information. We believe this model provides a plausible infection pathway of gram-positive infecting siphophages, and may offer potential value for the design and development of medically important phages.

Methods

Production and purification of NF5

The host bacterium B. thermosphacta NF4 HER1188 was cultured in tryptone soy broth (TSB) medium for 24 h at 25 °C. Subsequently, NF5 phages were incubated with the cells at 24 °C for 30 min. The mixture was then inoculated into 1 L of B. thermosphacta cells and amplified at 24 °C for 12 h. After amplification, the culture was shaken with 1% chloroform and centrifuged (9000 × g, 30 min, 4 °C) to obtain phage-containing supernatant. The supernatant was then incubated with 1 M NaCl and 10% PEG8000 at 4 °C overnight, followed by centrifugation at 11,000 × g for 90 min at 4 °C to precipitate phage particles. The precipitates were resuspended in SM buffer (50 mM Tris-HCl, 8 mM MgSO₄, and 100 mM NaCl, pH 7.5), and the phages were purified by CsCl density gradient ultracentrifugation (1.7, 1.5, 1.4, 1.3 g/ml; 135,000 × g, 4 °C, 2.5 h). Purified phages were dialyzed overnight against SM buffer and stored in ice water for cryo-sampling.

Cryo-EM sample preparation and data collection

Aliquots (3 μL) of purified NF5 phage were loaded onto glow-discharged Quantifoil R2/1 copper grids (20 s at 30 mA) with a 3-nm-thick layer of continuous carbon. Vitrification was performed using a Vitrobot Mark IV (Thermo Fisher Scientific) at 100% humidity and 8 °C, with a blot time of 4 s, blot force of 0.0, and a wait time of 0 s. Data were acquired on a Glacios 2 transmission electron microscope (Thermo Fisher Scientific) operated at 200 kV and equipped with a Falcon 4i direct electron detector. Images were recorded in the 32-frames movie mode at a nominal magnification 100,000 × g, corresponding to a pixel size of 1.2 Å. The accumulated dose of each movie was approximately 30 e⁻/Ų. Data were collected automatically using Thermo Fisher EPU software, yielding 19,681 movies.

Cryo-ET sample preparation and focused ion beam milling

B. thermosphacta NF4 cells for cryo-ET were grown in TSB medium at 25 °C, 230 RPM to OD₆₀₀ ≈ 0.5. Then, purified NF5 was added to the culture at a multiplicity of infection > 1000. The culture was incubated statically at 24 °C. Ten minutes post infection, cells were pelleted (7000 × g, 1 min) and resuspended in fresh TSB to OD₆₀₀ ≈ 5. The infected cells suspension (3.5 μL) was applied to glow-discharged (30 mA for 20 s) Quantifoil grids (R2/2, Cu, mesh 300). Using a Vitrobot Mark IV (Thermo Fisher Scientific), grids were blotted (blotting force 0, blotting time 2 s, wait time 0 s, 8 °C, 100% humidity), plunge-frozen in liquid ethane and stored in liquid nitrogen.

Grids of vitrified cells were subjected to cryo-focused ion beam (cryo-FIB) milling using an Aquilos 2 cryo-FIB/SEM system (Thermo Fisher Scientific). Briefly, to protect the cells from the damaging electron and gallium beams during the milling process, the vitrified samples were sputter-coated with metallic platinum for 15 s, followed by deposition of carbon-rich platinum for 10 s using a gas injection system, and then the sample was sputter-coated with metallic platinum for 15 s to prevent drifting during milling. Milling was performed with MAPS in a stepwise manner with an ion beam of 30 kV while reducing the current from 1 nA to 50 pA. Final polishing was performed with 30 pA current to achieve a target lamellae thickness of ~250 nm.

Cryo-ET data acquisition and processing

The lamellae were loaded into a Glacios 2 transmission electron microscope (Thermo Fisher Scientific) operated at 200 kV with a Selectris energy filter and a Falcon 4i direct electron detector. Tilt series acquisition was performed using Tomo5 Software (Thermo Fisher Scientific) at a nominal magnification of 39,000 × g (pixel size, 3.18 Å) with a dose-symmetric tilt scheme. Data were collected ranging from −42° to +66° relative to the lamella pretilt at +12°, using an angular increment of 3°, and a dose of 2 e⁻/ Ų per tilt at a target defocus of 4.0 μm. Tilt series were automatically performed motion correction, alignment, and 3D reconstruction using Tomo Live Software (Thermo Fisher Scientific). Tomogram visualization was carried out using Amira Software (Thermo Fisher Scientific) and 3dmod Software⁷¹.

Cryo-EM data processing

Image processing and 3D reconstruction were performed using cryoSPARC v4.6.0 software⁷² (Supplementary Fig. S1). All 19,681 movies were corrected for beam-induced motion using Patch Motion Correction and the contrast transfer function (CTF) was determined using the Patch CTF Estimation.

Initial particle picking of NF5 head using Blob Picker on a subset of 300 micrographs. One round of reference-free 2D classification selected good icosahedral head particles. These classes generated templates for Template Picker automatic picking. Several rounds of 2D classification removed bad particles, yielding a total of 64,549 high-quality particles (box size 680 × 680 pixels). These particles were subjected to Ab-initio Reconstruction with icosahedral symmetry. The resulting model was refined using Homogeneous Refinement with icosahedral symmetry, yielding the DNA-filled capsid map at 3.47 Å global resolution.

Particles for the portal, neck, tail tube, and baseplate regions were first selected manually using the Manual Picker to generate initial templates. These templates were used in Template Picker automatic picking for each region. Picked particles underwent 2D classified to remove bad particles. Initial volume for neck, tail tube, and baseplate were generated by Ab initio Reconstruction and refined iteratively using Homogeneous Refinement, Local Refinement, and Local CTF Refinement as needed.

For the portal complex, 40,860 particles were selected, and a final 3.46 Å map was obtained with C12 symmetry. For the neck complex, particle box centers were realigned to the neck region using the Volume Alignment Tool. Particle stacks underwent 3D classification with a reference model with C6 symmetry. A total of 21,051 good particles were selected and refined via Homogeneous Refinement with C6 symmetry, yielding a final 3.36 Å map.

For the tail tube, 92,135 particles (box size 200 × 200 pixels) were selected. A final 3.65 Å map was obtained using Homogeneous Refinement with C6 symmetry.

For the baseplate, a total of 38,436 particles (box size 360 × 360 pixels) were selected. The initial volume was generated using a single-class Ab-initio Reconstruction without symmetry. This was refined under C3 symmetry using Homogenous Refinement, yielding a 3.44 Å map. Particle centers were realigned to the Tal region using Volume Alignment Tool, and 37,938 particles (box size 256 × 256 pixels) were re-extracted. A final Tal-focused 3.46 Å map was obtained using Homogenous Refinement with C3 symmetry.

To reconstruct the RBP, particle coordinates from the baseplate reconstruction were realigned to the RBP trimer center using the Volume Alignment Tool. C6 symmetry was applied for symmetry expansion, yielding 230,062 particles (box size 256 × 256 pixels), which underwent multiple rounds of 3D classified. A local mask surrounding the RBP trimer was generated in ChimeraX⁷³. Finally, 57,406 particles of consistent conformation were selected for a final mask-focused local refinement without symmetry, achieving a 3.66 Å reconstruction of the RBP trimer.

Model building and refinement

Based on the cryo-EM maps of NF5, initial atomic models of MCP gp7, portal protein gp3, adaptor protein gp8, stopper protein gp9, tail terminator protein gp11, TTP gp12, C-terminal of TMP gp15, DTP gp16, Tal gp17, RBP gp18, and N-terminator of SFP gp19 were built automatically using ModelAngelo v1.0.13 software⁷⁴. These models were manually adjusted and refined in Coot v0.9.4⁷⁵, followed by real-space refinement in PHENIX v1.19.2⁷⁶. Refinement and validation statistics are in Table S1. Due to the limited resolution for the SFP trimers, the gp19 atomic model was predicted using AlphaFold3⁶² and manually fitted into the corresponding low-resolution map using ChimeraX⁷³.

Reporting summary

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

Author contributions

H.L., F.Y., and H.X. designed the research; Y.P., H.P. produced the NF5 sample and prepared cryo-EM grids; Y.P., Junquan Zhou collected the cryo-EM data; Y.P., H.X. produced the cryo-ET sample and prepared cryo-ET grids; Y.P., H.X. collected the cryo-ET data; Y.P., H.P., Jing Zheng, Junquan Zhou, W.C., F.Y., H.X., and H.L. analyzed data; Y.P., F.Y., H.X., and H.L. wrote the manuscript, with further edits from all authors. Since Jing Zheng and Junquan Zhou share the same initials, their full names are used here.

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

The cryo-EM maps have been deposited in the Electron Microscopy Data Bank (EMDB) under accession codes EMD-63659 (capsid); EMD-63664 (portal complex); EMD-63666 (neck complex); EMD-63681 (tail tube); EMD-63716 (baseplate complex); EMD-63696 (receptor binding protein); and EMD-63689 (tail-associated lysin). Cryo-electron tomograms of NF5 infection have been deposited in the Electron Microscopy Data Bank (EMDB) under the accession codes EMD-63704, and EMD-63721. The atomic coordinates have been deposited in the Protein Data Bank (PDB) under the accession codes 9M66 (capsid); 9M6K (portal complex); 9M6X (neck complex); 9M7C (tail tube); 9M8O (baseplate complex); 9M7X (receptor binding protein); and 9M7N (tail-associated lysin). Accession codes of previously published structures used in this study are 6IAB; and 3CSQ.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68477-2.