Introduction

Type IV pili (T4Ps) are surface filaments broadly distributed among bacteria and archaea, playing crucial roles in diverse biological processes, including surface motility, adhesion, biofilm formation, DNA uptake, protein secretion, immune evasion, cell signaling, and phage infection1,2,3,4,5,6,7,8,9,10,11. In Gram-negative bacteria, T4Ps are classified into three groups: type IVa pili (T4aP), type IVb pili (T4bP), and type IVc pili (T4cP)11,12. A distinguishing feature of T4bP is that its secretin component is a lipoprotein, in contrast to the non-lipoprotein secretins found in T4aP and T4cP. In enteropathogenic Escherichia coli (EPEC), a major pathogen responsible for infant diarrhea in developing countries13, the plasmid-encoded T4bP, known as the bundle-forming pilus (BFP), plays a crucial role in virulence. Specifically, BFP mediates bacterial auto-aggregation and localized adherence to epithelial cells—key processes for colonization and infection14,15. The essential function of BFP in EPEC virulence highlights the significance of T4bP in bacterial pathogenesis16.

The EPEC BFP system comprises 14 genes (BfpA to BfpL) organized in a single operon17, encoding the BFP biogenesis nanomachine responsible for pilus assembly and retraction (Supplementary Fig. 1). Prepilins are matured through cleavage of their N-terminal leader sequences by the prepilin peptidase BfpP18 (Supplementary Fig. 1b, c). The matured major pilin BfpA and other minor pilins associate with an inner membrane complex, where they polymerize to form the pilus structure. This process is energized by the cytoplasmic ATPase BfpD, while pilus retraction is driven by another cytoplasmic ATPase BfpF19,20 (Supplementary Fig. 1b, c). The T4bP secretin BfpB associates with BfpG to assemble a secretin channel complex facilitating pilus translocation across the outer membrane21,22 (Supplementary Fig. 1b, c). BfpG promotes BfpB multimerization through direct interaction, although this interaction is not required for BfpB proper localization to the outer membrane22. Interestingly, BfpB deficiency significantly reduces BfpG protein levels, whereas the absence of BfpG minimally affects BfpB levels23. The loss of either BfpB or BfpG results in the failure to form the BFP filament21.

Extensive research has elucidated secretin assembly mechanisms across various systems, including the type II secretion system (T2SS)24,25,26,27,28,29, type III secretion system (T3SS)30,31,32, type IV pilus system (T4PS)33,34,35, and the filamentous phage secretion system36. Secretins generally adopt a double-beta-barrel architecture in their secretin domains, with S domains located on the outer barrel surface24,30. The lipoprotein pilotin interacts with the S domain, facilitating the outer membrane targeting of secretins6,27. Unlike most secretins, BfpB exhibits a distinctive structural configuration as a lipoprotein, undergoing N-terminal cysteine acylation21, thereby eliminating the requirement for pilotin. High-resolution structures of other secretins reveal that the N0 domain is positioned at the base of the secretin channel24,30, distant from the outer membrane. This arrangement poses a structural puzzle for BfpB: the N-terminal acylated cysteine precedes the N0 domain in the primary sequence, yet how it spans the long distance from the channel base to the outer membrane to achieve proper localization remains unclear. Notably, resolved structures of secretins to date correspond exclusively to non-lipoprotein secretins24,25,26,27,28,29,30,31,32,33,34,35,36, leaving the assembly mechanism of the lipoprotein secretin BfpB largely unexplored.

To date, only four structures of fully assembled secretin complex have been resolved27,29,34,35. Three of these involve secretin-pilotin complexes: two from the T2SS, both exhibiting a 15:15 stoichiometry27,29, and one from the T4cP system, predominantly adopting a 14:14 assembly mode35. The fourth structure from the T4aP system features a secretin bound to its accessory protein in a 14:7 stoichiometry34. A recent cryo-electron tomography (cryo-ET) study provided a low-resolution view of the TcpC-TcpQ complex37, homologous to the BfpB-BfpG (hereafter referred to as BfpBG) complex. Despite these advances, the molecular mechanism governing BfpBG complex assembly remains poorly understood. In this work, we used single-particle cryo-electron microscopy (cryo-EM) to determine the structure of the EPEC BfpBG complex in the T4bP system. This structure uncovers distinctive structural characteristics of the secretin complex in the T4bP system. These findings establish the basis for understanding the assembly of the lipoprotein secretin BfpB and its cognate binding partner BfpG, offering deeper insights into the secretin complex assembly process.

Results

Structural determination and architecture of BfpBG complex

To investigate the structural organization of the BfpBG complex, a Strep-tag II was fused to the N-terminus of BfpG, positioned after the signal peptide, and the complex was co-expressed in the original EPEC strain. Size-exclusion chromatography (SEC) revealed a major peak at approximately 12.4 mL (Supplementary Fig. 2a), corresponding to a high-molecular-weight assembly. SDS-PAGE analysis of the peak fractions confirmed the presence of both BfpB and BfpG (Supplementary Fig. 2b). Purified BfpBG complex was further analyzed using single-particle cryo-EM to determine its structural features (Supplementary Fig. 2c–e).

The two-dimensional (2D) class averages revealed a ring-like structure comprising seventeen outwardly protruding densities encircling a central channel, indicating a heptadecameric assembly (Supplementary Fig. 2d). Subsequent three-dimensional (3D) reconstruction yielded a map at an overall resolution of 2.56 Å under C17 symmetry, with local resolutions varying across different regions (Supplementary Fig. 2e, 3a–c). To assess potential sample heterogeneity and possible symmetry mismatch, reconstructions were also performed with C1, C16, and C18 symmetry; however, only the C17 reconstruction yielded a map of sufficiently high resolution (Supplementary Fig. 2e). The density for BfpG was sufficiently resolved for de novo model building (Supplementary Fig. 3d). For BfpB, the map quality enabled de novo modeling of the N3 and secretin domains (Supplementary Fig. 3e, f, i). Although the N0 domain was less well-resolved, secondary structure features were still discernible, enabling the fitting of an AlphaFold3-predicted model into the unsharpened map (Supplementary Fig. 3g). Furthermore, the lipidation-BfpG binding domain (LG domain), located upstream of the N0 domain, displayed sufficient density for de novo modeling (Supplementary Fig. 3h, i).

The BfpBG complex exhibits a 17:17 stoichiometry, with 17 copies each of BfpG and BfpB (Fig. 1a, b). This assembly mode exhibits a higher symmetry compared to other secretins, which typically display C13–C15 symmetries24,25,26,27,28,29,30,31,32,33,34,35,36. The larger number of subunits results in an expanded channel diameter (~205 Å), with the lumen measuring ~110 Å due to a smaller inner lip (Fig. 1a, b). BfpB consists of four distinct domains: the LG domain, N0 domain, N3 domain, and secretin domain, arranged from the N-terminus to the C-terminus (Fig. 1c, d). Lacking the typical N1 and N2 domains, BfpB is shorter in overall height (~140 Å) (Fig. 1b). Like other secretins, BfpB adopts a closed state, with a central gate sealing the channel (Supplementary Fig. 4). Notably, BfpG localizes to the outer surface of the N3 ring (Fig. 1d, Supplementary Fig. 4)—a deviation from previously characterized secretin-binding partners, which typically associate outside the secretin channel27,29,34,35.

Fig. 1: Overall structure of the BfpBG complex.
figure 1

a, b Cryo-EM density map (left) and corresponding model (right) of the BfpBG complex shown in top (a) and side (b) views. The BfpB and BfpG are shown in cyan and purple, respectively. OM, outer membrane. c Schematic diagram of BfpB and BfpG domain organization. The BfpB sequence includes the signal peptide (gray), LG domain (green), N0 domain (blue), N3 domain (pink), and secretin domain (cyan), arranged from the N-terminus to the C-terminus. The BfpG sequence includes the signal peptide (gray) and N0-like domain (purple), arranged from the N-terminus to the C-terminus. d Structure of the BfpBG complex subunit. The structure is presented in two views, rotated 180° relative to each other.

Full size image

Secretin BfpB forms a distinctive third β-barrel within the N3 domain

The secretin domain of BfpB, like other secretins, forms two β-barrels: an inner and an outer barrel24,25,26,27,28,29,30,31,32,33,34,35,36 (Fig. 2a, Supplementary Fig. 5a). Each heptadecameric subunit contributes four β-strands to each barrel, resulting in a total of 68 β-strands per barrel (Supplementary Fig. 5a), which exceeds the number typically found in other secretins24,25,26,27,28,29,30,31,32,33,34,35,36. In addition to these two barrels, BfpB features a distinctive third barrel within the N3 domain, referred to as the N3 barrel (Fig. 2a, Supplementary Fig. 5a), which differentiates it from other secretins24,25,26,27,28,29,30,31,32,33,34,35,36. This N3 barrel is smaller than the inner and outer barrels, with each subunit contributing only two β-strands, yielding a total of 34 β-strands (Fig. 2a, Supplementary Fig. 5a). The β-sheets in the N3 barrel are tilted by approximately 55 degrees relative to those in the inner barrel, though they align more closely with the outer barrel (Supplementary Fig. 5b). Notably, all β-strands in the N3 barrel are antiparallel, a feature that enhances the stability of the barrel structure (Fig. 2a).

Fig. 2: Structural and functional analysis of the N3-barrel within the N3 domain.
figure 2

a Section view of the BfpB structure. The outer barrel, inner barrel, and N3 barrel are colored aquamarine, wheat, and light pink, respectively. The two adjacent N3 domains are highlighted in blue and yellow. b Close-up view of the N3 barrel within the N3 domain from (a). Hydrogen bonds between the main chains of the antiparallel β-strands are shown as violet dashed lines. c Bacteria auto-aggregation assay of the mutants in the N3 barrel. Phase contrast micrographs of E2348/69 ΔBfpBΔBfpF strain complemented with empty vector (EV), plasmids encoding wild-type BfpB (WT), BfpB Δ198-235, and BfpB mutants (T199P, K201P, and K230P) are shown. The red arrows highlight presentative bacterial auto-aggregation. Scale bar, 50 μm. Source data are provided as a Source Data file. d Western blot analysis of BfpB deletion and mutant variants shown in (c). BfpB was C-terminally Flag-tagged and detected with an anti-Flag antibody; GAPDH served as the loading control. Source data are provided as a Source Data file. The experiments in (c, d) were independently repeated twice with similar results.

Full size image

The two β-strands in the N3 barrel of BfpB are connected by a β1 strand and an α1 helix (Fig. 2b, Supplementary Fig. 5c). In most secretins, this region is linked by a short linker24,25,26,27,28,29,33,34,35,36 (Supplementary Fig. 5c). However, in T3SS InvG, the corresponding region contains two shorter β-strands30,31,32 (Supplementary Fig. 5c). These β-strands in InvG do not form the significantly tilted structure observed in the N3 barrel of BfpB, nor do they establish main-chain hydrogen bonds with β-strands from adjacent subunits (Supplementary Fig. 5d). In contrast, the tilting of the β-strands in the N3 barrel of BfpB facilitates interactions, allowing portions of these strands to form antiparallel β-strands with adjacent subunits (Fig. 2b, Supplementary Fig. 5d). This structural variation, driven by the longer β-strands in BfpB, enables the formation of the N3 barrel (Fig. 2b, Supplementary Fig. 5d). Therefore, the elongated β-strands in BfpB are crucial for the assembly of this additional barrel.

To investigate the functional importance of the N3 barrel, we conducted a bacterial auto-aggregation assay, utilizing the characteristic feature of BFP. The assay was performed in a knockout background of the retraction ATPase BfpF, as the absence of this protein leads to enhanced bacterial auto-aggregation due to the failure of pilus retraction19 (Supplementary Fig. 5e). We generated a deletion of the Tyr198-Val235 sequence, linking the termini with a single glycine residue to provide adequate space for the connection (Fig. 2b). This deletion abolished the auto-aggregation phenotype (Fig. 2c), indicating that the integrity of the N3 barrel is essential for BFP function. To further explore the role of the N3 barrel, we focused on three key residues—T199, K201, and K230—that form critical inter- and intra-subunit main-chain hydrogen bonds (Fig. 2b). These residues were mutated to proline to disrupt β-strand formation. Mutations T199P, K201P, and K230P completely abolished the auto-aggregation phenotype (Fig. 2c), demonstrating that disruption of the hydrogen-bonding network in N3 barrel impairs BFP biogenesis. Notably, Western blot analysis revealed that these mutants maintained similar protein expression levels compared to the wild type (Fig. 2d). Given that the BfpB channel is larger than those of other secretins, the N3 barrel likely plays a crucial role in maintaining the structural integrity of the channel.

The LG domain bridges the N0 domain and the outer membrane

The overall architecture of BfpB resembles that of previously characterized secretins24,25,26,27,28,29,30,31,32,33,34,35,36, with the secretin domain positioned at the top, the N3 domain in the middle, and the N0 domain located at the base of the channel (Fig. 3a, Supplementary Fig. 4). As the N0 domain is located approximately 80 Å from the outer membrane (Fig. 3a), a structural linker is required to bridge this distance and enable membrane insertion of the N-terminal acylated cysteine, which precedes the N0 domain in the primary sequence. This linker corresponds to the LG domain, whose N-terminal head inserts into the inner leaflet of the outer membrane via the acylated cysteine, while its C-terminal tail connects to the N0 domain (Fig. 3a, Supplementary Fig. 6a). Such an arrangement ensures that BfpB, as a lipoprotein, is correctly localized to the outer membrane despite the acylated cysteine being positioned upstream of the N0 domain in the sequence.

Fig. 3: Structure of the LG domain.
figure 3

a Reconstructed unsharpened cryo-EM map of the BfpB embedded in a detergent micelle. The LG domain, N0 domain, secretin domain, and BfpG are depicted in green, blue, cyan, and purple, respectively. The distance from the N0 domain to the outer membrane is approximately 80 Å. b Relative positioning of the α1 helix of the LG domain and the outer barrel of adjacent BfpB subunits. c Close-up view of the outer membrane-targeting region of the LG domain. Unmodeled density that may correspond to the first three residues (Cys1, Ser2, and Gly3), as well as to one potential acyl chain of Cys1 (black asterisk), is highlighted with a red circle. The black dashed line indicates the approximate membrane boundary, with the upper part embedded in the membrane and the lower part located in the periplasmic space.

Full size image

The LG domain represents a distinctive structural feature of the type IVb pilus system secretin. In contrast to typical secretins such as GspD, which contain a C-terminal S-domain that mediates outer membrane localization through interactions with the pilotin27, BfpB lacks this S-domain and instead employs an N-terminal LG domain to achieve outer membrane localization (Supplementary Fig. 6a, b). The LG domain adopts an extended conformation that spans both the N3 and secretin domains, containing two β-strands (β1 and β2; interacting with BfpG, as discussed in a later section) at its C-terminus and an α-helix (α1) at its N-terminus (Fig. 3a, Supplementary Fig. 6c). The α1 helix interacts with the outer surface of the outer barrel and is tilted approximately 30 degrees from the central axis of BfpB, enabling extensive interactions with two neighboring BfpB subunits (Fig. 3a, b, Supplementary Fig. 6c). The outer barrel interaction of α1 in the LG domain resembles that of α11 in the S-domain of conventional secretin GspD24,27 (Supplementary Fig. 6c, d). The short N-terminal tail, extending from α1 and composed of the first three residues (Cys1, Ser2, and Gly3), exhibits weak density and could not be reliably modeled in the high-resolution cryo-EM map (Fig. 3c), suggesting its flexibility. Nonetheless, the presence of detergent micelle density at this position (Fig. 3a) indicates that the acylated cysteine within this tail is embedded in the outer membrane. Collectively, these findings establish the LG domain as a previously unrecognized structural element of the lipoprotein-type secretin superfamily, bridging the N0 domain and the outer membrane.

The LG domain of BfpB forms zigzag-like interactions with BfpG

The BfpG interacts with the C-terminal region of the LG domain of BfpB, specifically between helix α1 and the N0 domain (Fig. 4a–d). This region is referred to as the BfpG binding region (GBR). Each BfpB subunit interacts with two adjacent BfpG subunits, and conversely, each BfpG subunit contacts two adjacent BfpB subunits. These interactions form a stable, continuous, alternating arrangement around the N3 ring, resulting in a zigzag-like assembly mode between the GBR and BfpG (Fig. 4a). These zigzag-like interactions promote tight binding between BfpG and BfpB, which explains the exceptionally high-resolution density observed for BfpG in the cryo-EM map (Supplementary Fig. 3d), a striking contrast to the typically lower-resolution densities seen for other secretin-binding proteins27,29,34,35. Furthermore, BfpG tightly binds the GBR, resembling a bucket hoop on the BfpB channel (Fig. 1a, b, Supplementary Fig. 4), which may explain why BfpG promotes BfpB multimerization22, especially considering the large size of the BfpB channel.

Fig. 4: Structural and functional analysis of the stable zigzag-like interactions between BfpG and the GBR of the LG domain.
figure 4

a Packing arrangement of neighboring BfpB subunits (yellow, green, and cyan) and BfpG subunits (purple and blue). The GBR of the LG domain is composed of two β-strands (β1 and β2) and a linker region (linkerβ12). b–d Close-up views highlighting the interaction details between the GBR of the BfpB and the two adjacent BfpG subunits. The interaction of β1 strand with right BfpG (b), the interaction of β2 strand with left BfpG (c), and the interaction of the linkerβ12 with two adjacent BfpG subunits (d) are highlighted in brown, blue and black box, respectively. Hydrogen bonds are shown as orange dashed lines. e Auto-aggregation assay assessing the role of specific residues mediating the interactions between BfpB and BfpG. Phase contrast micrographs of E2348/69 ΔBfpBΔBfpF strains complemented with empty vector (EV), wild-type (WT) BfpB, and various BfpB mutants (S41P, T43P, I46E, Y47A, I48E, S51P, and F53P) are shown. Scale bar, 50 μm. Source data are provided as a Source Data file. f Western blot analysis of C-terminally Flag-tagged BfpB mutants in (e), with GAPDH as the loading control. Source data are provided as a Source Data file. The experiments in (e, f) were independently repeated twice with similar results.

Full size image

The two β-strands (β1 and β2) in the GBR are pivotal for interactions with BfpG (Fig. 4a). The β1 strand of the GBR forms a parallel β-sheet with β5 of one BfpG subunit, while the β2 strand pairs antiparallel with β3 of an adjacent BfpG subunit, positioning β-strands from adjacent BfpB subunits on opposite sides of the BfpG subunit (Fig. 4b, c). In the first parallel β-sheet, the amide groups of Ser41 and Thr43 in the β1 strand form main-chain hydrogen bonds with the carbonyl groups of Ile97 and Phe99 in the β5 strand of BfpG, respectively (Fig. 4b). Similarly, the carbonyl groups of Met39, Ser41 and Thr43 in the β1 strand form main-chain hydrogen bonds with the amide groups of Ile97, Phe99, and Asn101 in the β5 strand of BfpG, respectively (Fig. 4b). Mutating Ser41 or Thr43 to proline disrupts these hydrogen bonds, leading to a significant reduction in auto-aggregation in EPEC (Fig. 4e). In the second antiparallel β-sheet, the amide groups of Ser51 and Phe53 in the β2 strand form main-chain hydrogen bonds with the carbonyl groups of Thr65 and Asn63 in the β3 strand of BfpG, respectively (Fig. 4c). Additionally, the carbonyl groups of Gly49, Ser51 and Phe53 in the β2 strand form hydrogen bonds with the amide groups of Asn67, Thr65 and Asn63 in the β3 strand of BfpG, respectively (Fig. 4c). Mutating Ser51 or Phe53 to proline abolishes these hydrogen bonds, completely preventing auto-aggregation in EPEC (Fig. 4e). Notably, Western blot analysis revealed that these mutations do not significantly affect protein expression levels, which remain comparable to the wild type (Fig. 4f). These results highlight the critical role of β-strand formation between BfpG and BfpB in the assembly of the secretin complex.

Besides the β-strand formation between GBR and BfpG, the linker region between β1 and β2 (termed linkerβ12) also plays a pivotal role in stabilizing the BfpBG complex (Fig. 4a). The linkerβ12 inserts into the interface between adjacent BfpG subunits, acting as a molecular glue that strengthens the sandwich-like interaction between them (Fig. 4a). Within the linkerβ12 region, Ile46 and Ile48 form hydrophobic interactions with the left BfpG subunit, while Tyr47 interacts with the right BfpG subunit through hydrophobic interactions and van der Waals forces (Fig. 4d). These hydrophobic interactions likely explain why BfpB deficiency leads to a significant reduction in BfpG protein levels23, as the absence of BfpB exposes hydrophobic residues of BfpG to the hydrophilic environment. Mutation of Ile46 to glutamic acid causes a slight decrease in auto-aggregation, while mutation of Ile48 to glutamic acid disrupts these hydrophobic interactions and completely abolishes auto-aggregation (Fig. 4e). Similarly, mutation of Tyr47 to alanine also disrupts the hydrophobic interaction, completely inhibiting auto-aggregation (Fig. 4e). Notably, these mutations do not substantially alter protein expression levels (Fig. 4f). Sequence alignment reveals that Tyr47 and Ile48 are conserved (Supplementary Fig. 7), further emphasizing the importance of these hydrophobic interactions in the linkerβ12 region. Thus, the zigzag-like interactions involving β-strand formation and sandwich-like interaction of the linkerβ12 region are essential for the assembly and stability of the secretin complex in the T4bP system.

BfpG serves as an atypical binding partner for the secretin

Like its homolog TcpQ37, BfpG shares structural similarities with the N0 domains of secretins in the T2SS, T3SS, and T4aPS, as well as with the C-terminal domain of Xanthomonas citri VirB7 in the type IV secretion system (T4SS)33,38,39,40(Supplementary Fig. 8a). Despite lacking limited sequence identity with the N0 domains of secretins (Supplementary Fig. 8c), BfpG adopts an N0 domain-like fold. To explore whether similar architectural features are conserved across other T4bP systems, we examined AlphaFold3-predicted homologs of BfpG in other well-characterized T4bPs. Our analysis revealed that these homologs exhibit structural characteristics similar to those of BfpG (Supplementary Fig. 8b). Consistently, multiple sequence alignments revealed strong sequence conservation, particularly within regions corresponding to the core structural elements (Supplementary Fig. 8d), suggesting that this architecture is conserved across T4bP systems.

Notably, the structure of BfpG diverges significantly from those of pilotins and secretin accessory proteins (Fig. 5a). Pilotins in the T2SS (e.g., AspS-type and PulS/OutS-type pilotins)27,29, T3SS (e.g., MxiM-type and InvH-type pilotins)41,42, and T4aP or T4cP (e.g., PilF/TadD-type pilotins)35,43 show considerable structural differences compared to BfpG (Fig. 5a). Similarly, TsaP, a secretin accessory protein in T4aP34, also shows structural variation compared to BfpG (Fig. 5a). These differences highlight that BfpG is a typical T4bP-type binding partner with a structure that sets it apart from other secretin-binding partners.

Fig. 5: Structural diversity of secretin-binding partners and their assembly with secretins.
figure 5

a Structural gallery of various pilotins and secretin accessory proteins. The PDB codes for the different molecules include: T2SS: 5ZDH (AspS), 6HCG (PulS); T3SS: 2JW1 (MxiM), 6XFK (InvH); T4PS: 2HO1 (PilF), 8ODN (TadD), 6VE2 (TsaP). b Distinct assembly patterns of secretin complexes from T2SS and T4PS. The structures of the following complexes are compared: ETEC GspD-AspS (5ZDH), K. pneumoniae PulD-PulS (6HCG), P. aeruginosa PilQ-TsaP (6VE2), EPEC BfpB-BfpG (this study), and P. aeruginosa RcpA-TadD (8ODN). AspS, PulS, TsaP, BfpG, and TadD are shown as surface (upper) or cartoon (lower) representations, colored in dark red, blue, orange, purple, and magenta, respectively. The monomers of secretins are depicted in red, with the LG domain of BfpB and the S domains of GspD, PulD, and RcpA shown in green (upper). The N-termini of secretin-binding partners are indicated, with AspS, PulS, and TadD oriented toward the inner leaflet of the outer membrane (lower). OM, outer membrane.

Full size image

The assembly and structure-function relationship between BfpG and secretin further distinguish it from other secretin-binding partners. In fully assembled secretin complexes to date, BfpG is the only partner located at the periphery of the N3 domain, whereas other partners typically associate with the periphery of the secretin domains27,29,34,35 (Fig. 5b). Although neither BfpG nor TsaP is a lipoprotein, they bind to distinct regions of the secretins: BfpG interacts with the GBR of the LG domain, while TsaP binds to the outer surface of the secretin domain34 (Figs. 4a, 5b). Notably, BfpG binding to the LG domain is essential for the assembly of the BfpB secretin and BFP biogenesis, whereas TsaP’s interaction with the secretin domain is dispensable for PilQ assembly and twitching motility34,44,45,46. In secretin-pilotin complexes, the C-terminal tails of the S domains engage with cavities in the pilotins, playing a critical role in secretin-pilotin assembly27,29,35 (Fig. 5b, Supplementary Fig. 9a). In most reported pilotin complex structures, these C-terminal tails adopt a helical conformation27,29,35,41,42,47, although in the RcpA-TadD complex, the tail is a pentapeptide35(Supplementary Fig. 9).

In contrast, the BfpBG complex exhibits a distinct interaction mechanism, involving two β-strands (β1 and β2) and a linker (linkerβ12) in the LG domain (Fig. 4a). Another key difference lies in outer-membrane targeting. In secretin-pilotin complexes, the N-terminal regions of pilotins containing acylated cysteine residues are oriented toward the inner leaflet of the outer membrane27,29,35(Fig. 5b), consistent with the targeting function of pilotins. By contrast, BfpG, as a T4bP-type partner, does not directly participate in membrane targeting but instead binds to the GBR through stable zigzag-like interactions near the N3 ring (Figs. 4a, 5b), facilitating the assembly of the lipoprotein secretin BfpB in the outer membrane. This distinct binding mode differentiates BfpG from pilotins and underscores its specialized role as a T4bP-type binding partner within the secretin superfamily.

Discussion

Despite the discovery of the EPEC BFP system in 199148, the precise assembly mechanism of this typical T4bP system has remained largely unresolved for over three decades. Here, we employ cryo-EM to determine the structure of the EPEC BfpBG secretin complex. This work represents the first resolved structure of a secretin complex in the T4bP system and provides the first structural insights into a lipoprotein-associated secretin. This secretin structure presents distinctive assembly characteristics, including higher C17 symmetry, N3 barrel and LG domain. The study also reveals how BfpG, an atypical T4bP-type binding partner, interacts with BfpB to cooperatively assemble a complete outer membrane secretin complex in the T4bP system.

Our findings demonstrate that BfpB adopts a C17 symmetry, which is larger than the C13–C14 symmetries previously reported for T4aP and T4cP secretins34,35. Several T4P structures have been resolved to date, including examples from T4aP, T4bP, and T4cP subtypes49,50,51,52,53,54,55. The diameters of these pili vary among subtypes, with T4aP measuring 45–70 Å49,50,51,52,53, T4bP-associated Tcp pili approximately 80 Å54, and T4cP-associated Tad pili approximately 42 Å or 50 Å54,55. Given that T4bP is thicker than both T4aP and T4cP, the higher symmetry of BfpB likely accommodates the passage of the wider T4bP filament. Notably, the BfpB structure resolved in this study represents a closed state, with its central gate sealed. While open conformations of T3SS secretins have been characterized31,40, providing insights into secretin gating mechanisms, comparable structural data for secretins in T4P-like systems (e.g., T2SS and T4PS) remain unavailable. The mechanism underlying the opening of the BfpB channel thus remains unclear and warrants further investigation.

The BfpB channel exhibits a distinctive structural feature among secretins, with three β-barrels, including an additional N3 barrel. Functional studies demonstrate that disruption of the N3 barrel, either through deletions or site-specific mutations, critically impairs BFP biogenesis, underscoring its pivotal role in BfpB-mediated function. In the T3SS secretin InvG, similar to BfpB, each subunit contributes two antiparallel β-strands in this region30. However, these strands in InvG are insufficient in length to form a β-barrel, instead adopting a β-hairpin configuration30. Mutation of conserved residues Arg198 and Asp199 of InvG, located at the hairpin tip and interacting with the underside of the inner β-barrel, significantly compromises or completely abolishes secretion30. Thus, these β-strands are essential for the function of both InvG and BfpB. Structural analyses of secretins from the T2SS, T4aPS and the filamentous phage secretion system reveal the absence of β-sheet formation in the corresponding region, further highlighting the structural and functional characteristics of T4bP secretins.

BfpB exhibits a non-classical secretin architecture, comprising an LG domain, N0 domain, N3 domain, and secretin domain. Unlike classical secretins, BfpB lacks the S domain, likely due to its evolutionary adaptation as a lipoprotein. The LG domain’s α1 helix binds to the outer surface of the secretin channel, with its N-terminal acylated cysteine anchoring into the outer membrane. At the C-terminus of the LG domain resides the BfpG-binding region, which mediates stable zigzag-like interactions with BfpG. This BfpB-BfpG interaction is critical for the BFP biogenesis, as disruption of this interaction impairs BFP assembly. The LG domain, therefore, functions as a structural bridge between the N0 domain and the outer membrane. Notably, BfpG adopts an N0 domain-like fold, typical for T4bP-type binding partners. Structural studies further reveal that 17 BfpG molecules bind tightly to the outer surface of the N3 ring of BfpB, stabilizing its oligomeric state. The distinct assembly mechanism and structure-function relationship set BfpG apart from other secretin-binding partners.

In summary, this study provides significant insights into the structural features of the BfpB secretin and clarifies the distinctive assembly mechanism of the secretin complex within the T4bP system, thereby advancing our understanding of the functional diversity within the secretin superfamily.

Methods

Plasmid construction

The DNA sequence of BfpG-BfpB was amplified from the EPEC E2348/69 strain and subsequently cloned into the pBAD30 vector using the ClonExpress Ultra One-Step Cloning Kit V3 (Vazyme Biotech Co., Ltd). For BfpBG complex purification, Strep tag was introduced after the N-terminal signal peptide of BfpG (UniProtKB accession number B7UTD2), yielding the pBAD30-StrepBfpG-BfpB construct. Similarly, BfpB (UniProtKB accession number Q47068) with a C-terminal Flag tag was cloned into the pBAD30 vector to generate the pBAD30-BfpBFlag construct for complementation assays. All BfpB mutants and truncations were generated using the same cloning strategy.

Protein expression and purification

The plasmid pBAD30-StrepBfpG-BfpB was electroporated into the EPEC E2348/69 strain, and the transformed cells were cultured in lysogeny broth (LB) at 37 °C with shaking. Protein expression was induced by adding 0.2% L-arabinose when the culture reached an OD600 of 1.2. The culture was then incubated at 20 °C for 14 h. The cells were harvested by centrifugation at 5000 g for 30 min at 4 °C and resuspended in a buffer containing 50 mM HEPES at pH 7.4, 150 mM NaCl, and a protease inhibitor cocktail. The resuspended cells were disrupted using a homogenizer (AH-NANO, ATS) at 1000 bar for 4 cycles. The lysate was centrifuged at 17,420 g for 10 min to remove cellular debris. To isolate membrane fractions, the lysate was subjected to ultracentrifugation at 164,700 g for 30 min. The membrane pellet was solubilized in a buffer containing 50 mM HEPES at pH 7.4, 150 mM NaCl, and 1% n-dodecyl-β-D-maltopyranoside (DDM, Anatrace) for 1.5 h at 4 °C. Insoluble materials were removed by centrifugation at 27,220 g for 30 min. The detergent-solubilized supernatant was then loaded into Strep-Tactin resin (IBA) and washed with a buffer containing 50 mM HEPES at pH 7.4, 150 mM NaCl, and 0.05% DDM. Protein elution was performed with a buffer containing 50 mM HEPES at pH 7.4, 150 mM NaCl, 0.05% DDM, and 2.5 mM desthiobiotin. The eluted proteins were concentrated and further purified by size-exclusion chromatography using a Superose 6 Increase 10/300 GL column (Cytiva). Peak fractions were analyzed by 15% SDS-PAGE. The purified proteins were concentrated and stored at −80 °C for future use.

Cryo-EM sample preparation and data collection

3.5 µL of a 2.15 mg/mL protein sample was applied to a glow-discharged holey carbon copper grid (R1.2/1.3, 400 mesh, Quantifoil) with a thin continuous carbon film. The grid was incubated for 30 s, followed by blotting for 4 s at 8 °C and 100% humidity. The grids were then plunge-frozen in liquid ethane cooled by liquid nitrogen using a Vitrobot Mark IV and subsequently loaded into a Titan Krios G3 electron microscope operated at 300 kV. All movies were recorded with a K3 direct electron detector at a nominal magnification of 81,000x (pixel size: 0.88 Å/pixel) using EPU software. The movies were collected in automated super-resolution mode and dose-fractionated into 32 frames, with a total exposure time of 1.27 s and a total dose of 56 e-/Ų. A defocus range of −1.7 to −2.1 µm was applied to all movies.

Cryo-EM data processing

Beam-induced motion in super-resolution movie stacks was corrected using MotionCor256. Following this, image processing and cryo-EM map reconstruction were performed using cryoSPARC 4.0.357. Contrast transfer function (CTF) parameters were estimated using Patch CTF Estimation, and hundreds of micrographs were pre-processed to generate suitable 2D templates and an initial 3D reference. A total of 571,165 particles were picked from 6762 micrographs using a Topaz model previously trained. After removing undesirable particles through iterative rounds of heterogeneous refinement with C17 symmetry, 419,489 particles were retained; however, the density of the N0 domains of the BfpB subunit in the reconstructed map remained weak, requiring further analysis. Following several rounds of masked 3D classification, 75,445 particles were selected for subsequent refinement. The final map was refined using non-uniform refinement, yielding an average resolution of 2.56 Å.

To further assess the validity of the C17 symmetry, we performed Ab-initio reconstructions with C1, C16, C17, and C18 symmetry on 571,165 particles extracted with Topaz. To minimize model bias, particle subsets corresponding to well-defined classes were refined by non-uniform refinement under either C1 or the symmetry applied in the preceding step. Furthermore, Ab-initio maps generated under different symmetries were used as references for subsequent heterogeneous refinement in C1 symmetry, confirming that only the C17 symmetry reconstruction yielded a high-resolution map. The overall resolution was determined using the gold-standard Fourier shell correlation (FSC) cutoff of 0.143, while local resolution ranges were assessed in cryoSPARC.

Model building and refinement

The BfpB and BfpG models, predicted by AlphaFold358, were fitted into the cryo-EM map using Chimera59 and manually adjusted in Coot60 based on the density. Bulky residues (Phe, Tyr, Trp, and Arg) and unique sequence patterns were used as markers to validate sequence assignments during model building. Several iterations of real-space refinement were performed using Phenix61. The geometries of all models were evaluated using MolProbity62. Cryo-EM data collection, refinement, and validation statistics are provided in Supplementary Table 1.

Gene knockout

A one-step inactivation procedure63 was used to delete genes from the EPEC E2348/69 strain. Briefly, a kanamycin resistance cassette was amplified from plasmid pKD4 using primers with approximately 60–70 bp extensions homologous to the regions flanking the BfpB gene. The PCR product was electroporated into the EPEC strain carrying the pKD46 plasmid, which had been induced with L-arabinose to express the recombinase. Kanamycin-resistant clones were screened by colony PCR and verified by DNA sequencing. The kanamycin cassette was then excised using the pCP20 plasmid, generating the BfpB knockout strain (E2348/69 ΔBfpB). This approach was also applied to generate the E2348/69 ΔBfpF and E2348/69 ΔBfpBΔBfpF strains. Strains and primers are provided in Supplementary Tables 2, 3.

Bacteria auto-aggregation assay

The pBAD30-BfpBFlag construct and various mutants were transformed into the E2348/69 ΔBfpBΔBfpF strain. The bacterial strains were cultured overnight in LB medium containing 100 µg/mL ampicillin at 37 °C. The OD600 was measured and normalized across all samples. Each sample was then diluted 1:50 into 10 mL of Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F-12, Gibco), supplemented with 100 µg/mL ampicillin and 0.2% L-arabinose for induction. The cultures were incubated at 37 °C for an additional 8 h. To assess bacterial auto-aggregation, 20 µL of each culture was placed on a glass slide, and the cells were observed under the EVOSTM M5000 microscope (Thermo Fisher Scientific). Images were captured and saved for subsequent analysis. Strains and primers are provided in Supplementary Tables 2, 3.

Western blot

Bacterial cells from the auto-aggregation assay were collected by centrifugation at 12,000 g for 5 min at 4 °C. Pellets were washed twice with pre-chilled PBS, resuspended in SDS buffer, and boiled at 100 °C for 30 min. Proteins were separated by 10% SDS-PAGE and transferred to Immobilon-P membranes at 200 mA for 1.5 h on ice. The membranes were blocked overnight at 4 °C in 5% skim milk in TBST buffer (25 mM Tris-HCl at pH 7.5, 150 mM NaCl, 0.1% Tween 20). Membranes were incubated for 1 h at room temperature with anti-Flag mouse monoclonal antibody (Abmart, M20008, 1:5000 dilution) to detect BfpB expression, and anti-GAPDH rabbit monoclonal antibody (Abways, AB0037, 1:10,000 dilution) as a loading control. After washing, membranes were incubated with HRP-conjugated goat anti-mouse (Abmart, M21001, 1:5000 dilution) or goat anti-rabbit (Abmart, M21002, 1:5000 dilution) secondary antibodies for 1 h at room temperature, respectively. Membranes were detected using the enhanced chemiluminescence (ECL) system and imaged with the ChemiDoc MP System (Bio-Rad).

Structural homology search

Structure-based homology analysis was conducted using the DALI server64 (http://ekhidna2.biocenter.helsinki.fi/dali). The experimentally determined structure of BfpG was submitted as the query, and the resulting hits were evaluated based on Z-score, RMSD, and alignment length. Structures with low resolution or incomplete models were excluded. Representative structurally similar proteins identified include N0 domains of GspD (PDB: 4JTM), InvG (PDB: 6DV3) and PilQ (PDB: 6W6M), and the C-terminal domain of VirB7 (PDB: 6GYB) (Supplementary Data 1). To identify sequence-based homologs of BfpG, multiple sequence alignment was performed using previously reported homologous proteins (CofC, LngC, TcpQ). The structures of these homologs were predicted using AlphaFold3 and subsequently compared with BfpG by structural superposition and visualization in Chimera.

Reporting summary

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