Asymmetric architecture and adaptation of Treponema flagella Authors: Jiaqi Wang, Kurni Kurniyati, Wangbiao Guo, Jack M. Botting, Hui Wu, Charles V. Sindelar, Chunhao Li, Jun Liu Published: 2026 ABSTRACT Spirochetes exhibit a distinctive corkscrew-like motility driven by periplasmic flagella that wrap around the cell body in a supercoiled configuration, yet the structural basis of this propulsion remains poorly understood. Here we combine cryo-electron microscopy, cryo-electron tomography, and genetic and biochemical analyses to determine the assembly and adaptation principles of the supercoiled flagellar filament in Treponema denticola, a major periodontal pathogen. Near-atomic structures reveal a glycosylated FlaB flagellin core encased by an asymmetric sheath. The major sheath protein FlaA forms the bulk of the sheath and mechanically couples to the core through defined interfaces required for efficient motility, whereas four minor sheath proteins (FlaA1, FlaA2, FlaAP1, and FlaAP2) assemble along the concave side of the filament to accommodate intrinsic curvature. Disruption of this asymmetric core–sheath organization compromises force transmission and impairs motility, establishing coordinated asymmetric assembly as a fundamental mechanism underlying spirochetal motility. Introduction Spirochetes are a distinct group of bacteria characterized by helical or flat-wave morphologies and a form of motility that enables efficient movement through highly viscous environments, including host tissues and mucus layers1–3. Several spirochetes are major human pathogens, including Treponema denticola (periodontitis), Treponema pallidum (syphilis), Borreliella burgdorferi (Lyme disease), Leptospira spp. (leptospirosis), and Brachyspira spp. (intestinal spirochetosis)4–6. Among these, the oral bacterium Treponema denticola is an obligatory anaerobic and motile pathogen implicated in both periodontal and endodontic infections and connects to several systemic illnesses including Alzheimer’s disease7–9. Motility, a central determinant of spirochetal pathogenicity, is driven by periplasmic flagella confined to the space between the outer membrane and peptidoglycan layer and wrapped around the cell body in a supercoiled form. Rotation of periplasmic flagella generates torsional forces that deform the cell body into helical or flat-wave conformations, producing a characteristic corkscrew-like motion that enables efficient migration through dense and heterogeneous environments1,2. In multiple species, genetic disruption of flagellar components abolishes infectivity, tissue penetration, and epithelial colonization, establishing motility as essential for host survival and transmission10–15. However, mechanisms underlying the spirochetal flagellar assembly and rotation have remained incompletely understood1–3,16. The periplasmic flagellum is architecturally distinct from the external flagellum in model organisms such as Escherichia coli and Salmonella enterica although they all share a conserved organization consisting of a basal body, a hook, and a filament16,17. In spirochetes, however, the filament is composed of a conserved core encased by a distinct proteinaceous sheath18. The core is formed by flagellins homologous to E. coli and S. enterica FliC19 which are exported through the flagellar type III secretion systems (fT3SS)18,20,21. In contrast, the proteinaceous sheath is largely composed of FlaA proteins that are unique to spirochetes, lack homology to flagellins, and are exported via the Sec-dependent pathway rather than the fT3SS18,22,23. Structural diversity among spirochetes is exemplified by Leptospira, whose flagellar filament contains an asymmetric, multi-protein sheath24. Disruption of genes involved in sheath formation not only impairs filament assembly but also compromises motility and infectivity13,25,26. Consistent with these observations, recent high-resolution structures of Leptospira flagellar filaments revealed pronounced asymmetry that is essential for proper filament assembly and motility27–29. In Brachyspira and Treponema species, the filament consists of a core formed by three flagellins (FlaB1, FlaB2, and FlaB3) surrounded by a proteinaceous sheath primarily composed of FlaA, while their structural details remain unknown22,30–34. Recent mass spectrometry analyses indicated that T. denticola FlaB proteins are extensively O-glycosylated on serine and threonine residues, and that this modification is essential for filament assembly and motility35. Despite these advances, sheathed flagella in spirochetes remain much less understood structurally and mechanistically than external flagella. Here, using cryo-electron microscopy (cryo-EM), we determine the near-atomic, asymmetric, multilayered architecture of the supercoiled flagellar filament in T. denticola. Our structure reveals that O-linked glycosylation of specific serine and threonine residues in FlaB flagellins stabilizes assembly of the filament core and promotes its interactions with the surrounding sheath. Structural analyses further show that the major sheath protein FlaA engages extensively with FlaB flagellins to form the bulk of the sheath, whereas four minor sheath proteins assemble selectively along the concave side of the supercoiled filament, accommodating the intrinsic curvature of the cell body. Finally, integrating cryo-electron tomography (cryo-ET) with genetic and biochemical analyses provides evidence that loss of any component of the asymmetric sheathed filament destabilizes flagellar assembly, disrupts force transmission, and severely impairs motility. Together, these findings establish a mechanistic framework for how spirochetes evolved an asymmetric periplasmic flagellar architecture to support their distinctive morphology and motility. Results Identification of flagellar proteins in T. denticola Previous studies have shown that T. denticola periplasmic flagellar filaments are composed of three flagellins, including FlaB1 (TDE1477), FlaB2 (TDE1004), FlaB3 (TDE1475), and a major sheath protein, FlaA (TDE1712)30,33–35. Whole-cell cryo-ET revealed that two to three periplasmic flagella originate from each cell pole, wrap around the cell body in a supercoiled configuration, and extend toward the cell center, where filaments from opposite poles overlap (Fig. 1a)34. Notably, periplasmic flagella exhibit pronounced length asymmetry: filaments anchored at the old cell pole are longer than those at the new cell pole (Fig. 1a)34. Furthermore, filament diameter varies along its length, with the proximal region significantly thicker than distal regions, suggesting that sheath formation is not fully synchronized with filament core assembly (Fig. 1a,b)34. Consistent with this interpretation, deletion of flaA markedly reduces filament diameter along the entire cell length and severely impairs motility, underscoring the essential role of the sheath in flagellar architecture and function34.Fig. 1Identification and characterization of flagellar proteins in T. denticola.a Schematic of T. denticola illustrating periplasmic flagella located between the inner and outer membranes, with variable filament lengths from the old and new cell poles. Notably, the sheath is colored in light green and the core is colored in orange. IM: Inner membrane; OM: Outer membrane. Representative cryo-EM images from cell tips of wild-type (WT) cells (b) and purified flagellar filaments (c). Green arrow: sheathed filaments; yellow arrow: non-sheathed filaments; pink arrow: thin protofilaments. Scale bars are 60 nm. d Identification of flagellar proteins of the T. denticola periplasmic flagella. SDS-PAGE analysis followed by Coomassie staining of purified periplasmic flagella from WT and ΔflaA strains. Bands indicated by red lines were excised and analyzed by nano LC-ESI/MS/MS for protein identification. The black-line-indicated bands represent periplasmic flagella-associated proteins that have been identified and studied. This experiment was performed one time. e–h Swimming motility of WT, ΔflgE, ΔflaA1, ΔflaA2, ΔflaAP1, and ΔflaAP2 cells in soft agar plates. ΔflgE, a non-motile mutant, was used as a control. g, Swimming ring diameters (n = 6 plates) are for wild-type (20.33 ± 0.35 mm), ΔflaA1 (7.00 ± 0.21 mm), ΔflaA2 (6.92 ± 0.38 mm), ΔflaAP2 (19.75 ± 0.45 mm), and ΔflgE (4.83 ± 0.21 mm). h, Swimming ring diameters (n = 5 plates) are for wild-type (22.50 ± 0.62 mm), ΔflaAP2 (13.00 ± 0.42 mm), and ΔflgE (5.40 ± 0.16 mm). The results shown here are representative of one experiment with at least five technical replicates. The experiment was repeated independently at least three times. The results are expressed as the mean. Tracking analyses show a reduced velocity (i) and increased stops (j) of ΔflaAP1 cells compared to WT cells. The cell velocity (n = 100 cells) is 7.39 ± 0.29 mm for wild-type and 4.65 ± 0.19 mm for ΔflaAP1. The number of stops (n = 50 cells) is 1.80 ± 0.36 for wild-type and 5.90 ± 0.82 for ΔflaAP1. The results shown here are from three independent experiments. The results are expressed as the mean ± SEM. The data were analyzed using an unpaired t-test with Welch’s correction, with significance set at a two-tailed p < 0.01. Source data are provided as a Source Data file. Identification and characterization of flagellar proteins in T. denticola. a Schematic of T. denticola illustrating periplasmic flagella located between the inner and outer membranes, with variable filament lengths from the old and new cell poles. Notably, the sheath is colored in light green and the core is colored in orange. IM: Inner membrane; OM: Outer membrane. Representative cryo-EM images from cell tips of wild-type (WT) cells (b) and purified flagellar filaments (c). Green arrow: sheathed filaments; yellow arrow: non-sheathed filaments; pink arrow: thin protofilaments. Scale bars are 60 nm. d Identification of flagellar proteins of the T. denticola periplasmic flagella. SDS-PAGE analysis followed by Coomassie staining of purified periplasmic flagella from WT and ΔflaA strains. Bands indicated by red lines were excised and analyzed by nano LC-ESI/MS/MS for protein identification. The black-line-indicated bands represent periplasmic flagella-associated proteins that have been identified and studied. This experiment was performed one time. e–h Swimming motility of WT, ΔflgE, ΔflaA1, ΔflaA2, ΔflaAP1, and ΔflaAP2 cells in soft agar plates. ΔflgE, a non-motile mutant, was used as a control. g, Swimming ring diameters (n = 6 plates) are for wild-type (20.33 ± 0.35 mm), ΔflaA1 (7.00 ± 0.21 mm), ΔflaA2 (6.92 ± 0.38 mm), ΔflaAP2 (19.75 ± 0.45 mm), and ΔflgE (4.83 ± 0.21 mm). h, Swimming ring diameters (n = 5 plates) are for wild-type (22.50 ± 0.62 mm), ΔflaAP2 (13.00 ± 0.42 mm), and ΔflgE (5.40 ± 0.16 mm). The results shown here are representative of one experiment with at least five technical replicates. The experiment was repeated independently at least three times. The results are expressed as the mean. Tracking analyses show a reduced velocity (i) and increased stops (j) of ΔflaAP1 cells compared to WT cells. The cell velocity (n = 100 cells) is 7.39 ± 0.29 mm for wild-type and 4.65 ± 0.19 mm for ΔflaAP1. The number of stops (n = 50 cells) is 1.80 ± 0.36 for wild-type and 5.90 ± 0.82 for ΔflaAP1. The results shown here are from three independent experiments. The results are expressed as the mean ± SEM. The data were analyzed using an unpaired t-test with Welch’s correction, with significance set at a two-tailed p < 0.01. Source data are provided as a Source Data file. To further define the protein composition of the flagellar sheath, periplasmic flagella were purified from wild-type and ΔflaA (ΔTDE1712) strains. Purified flagellar samples were examined by cryo-EM (Fig. 1c) and analyzed by SDS-PAGE followed by Coomassie staining (Fig. 1d). Cryo-EM imaging revealed heterogeneous filament populations with distinct diameters, including fully sheathed filaments, non-sheathed filaments, and very thin protofilaments within the purified wild-type flagella (Fig. 1c). SDS-PAGE analysis showed several protein bands present in wild-type samples but absent in the ΔflaA mutant (Fig. 1d, red lines). These bands were excised and subjected to nanoLC-ESI-MS/MS analysis, which identified two FlaA homologs, TDE1408 (236 aa, 27.0 kDa) and TDE1409 (246 aa, 27.8 kDa), hereafter designated FlaA1 and FlaA2, respectively. In addition, two FlaA-associated proteins were identified: TDE2349 (234 aa, 26.0 kDa) and TDE1480 (139 aa, 16.0 kDa), designated FlaAP1 and FlaAP2, respectively (Fig. 1d and Supplementary Fig. 1). Source data are provided as a Source Data file. Construction and functional characterization of flagellar mutants To understand the functions of these identified proteins, deletion mutants of flaA1, flaA2, flaAP1, and flaAP2, were generated by allelic exchange and further confirmed by PCR analysis (Supplementary Fig. 2). Immunoblotting revealed that levels of FlaA and FlaB proteins were not significantly altered in the ΔflaA1 and ΔflaA2 strains. In contrast, ΔflaAP1 and ΔflaAP2 strains exhibited a modest reduction in FlaA levels accompanied by a slight increase in FlaB abundance (Supplementary Fig. 3 and Supplementary Tables 2,3). Functional consequences of these deletions were assessed using swimming plate assays. The mean diameters of swimming rings (n = 6 plates) were 7.00 ± 0.21 mm for ΔflaA1, 6.92 ± 0.21 mm for ΔflaA2, and 13.00 ± 0.42 mm for ΔflaAP2, all significantly smaller than those of the wild-type strain (20.33 ± 0.36 mm to 22.50 ± 0.62 mm) (Fig. 1e–h). We conclude that flaA1, flaA2, and flaAP2 play important roles in T. denticola motility (Fig. 1e–h). In contrast, the ΔflaAP1 mutant showed a modest reduction in mean swimming ring diameter (19.75 ± 0.45 mm) relative to wild type but also displayed spreading zones with indistinct boundaries on swimming plates (Fig. 1e, g and Supplementary Table 3). To better compare the motility phenotypes of wild-type and ΔflaAP1 cells, we deployed motion-tracking analysis to show that the ΔflaAP1 cells have reduced swimming velocity compared to wild-type cells in 1% methylcellulose (Fig. 1i; Supplementary Movie 1,2). In addition, the ΔflaAP1 cells have more stops compared to wild-type cells (Fig. 1j; Supplementary Movie 1,2). Together, these data indicate that FlaAP1 also plays a role in T. denticola motility. Near-atomic asymmetric structures of the supercoiled flagellar filament in T. denticola To define the molecular architecture of the supercoiled flagellar filament, we used single-particle cryo-EM to analyze filaments both in situ and in vitro (Supplementary Figs. 4–8). Asymmetric reconstruction and refinement in cryoSPARC36 yielded near-atomic structures of the supercoiled filament (Fig. 2a–d and Supplementary Fig. 4-8). The filament core consists of 11 flagellin protofilaments (C1-C11) with an overall diameter of 125 Å and a central lumen diameter of 24.5 Å (Fig. 2d). The core structure resembles a filament with a helical symmetry of a 4.80 Å rise and a 65.29° twist, as estimated by cryoSPARC36. However, each flagellin protofilament is slightly different from the adjacent ones to allow formation of the supercoiled filament (Fig. 2 and Supplementary Fig. 8). The filament sheath exhibits pronounced asymmetric features with an overall diameter of ~195 Å (Fig. 2a, d). Specifically, the sheath is mostly formed by eight protofilaments (S1-S8) of the major sheath protein FlaA and six protofilaments (AS1-AS6) of four minor sheath proteins (Fig. 2h). Consistent with flagellar proteins identified by nanoLC-ESI-MS/MS analysis (Fig. 1), the near-atomic map enabled de novo model building of the filament core protein FlaB1, the major sheath protein FlaA, and four minor sheath proteins (FlaA1, FlaA2, FlaAP1, and FlaAP2) (Fig. 2e–h and Supplementary Fig. 7a–d). The overall model reveals a molecularly heterogeneous flagellar architecture, providing a structural framework for understanding the specialized adaptations of the supercoiled periplasmic flagella and their contributions to the distinctive morphology and motility of spirochetes (Fig. 2i–m and Supplementary Fig. 9).Fig. 2Asymmetric architecture of the supercoiled flagellar filament in T. denticola.Cryo-EM structure of the sheathed flagellar filament viewed from the convex (a) and concave (b) sides. Green: FlaA; blue: FlaAP1; purple: FlaA2; pink: FlaA1; orange: FlaAP2; yellow: FlaB1; red: glycan. c Close-up view of the glycosylated filament core. Filament components are colored by protein identity; glycans are shown. d Perpendicular cross-section of the filament revealing a central lumen, a tubular filament core, and a surrounding protein sheath. The core is predominantly formed by FlaB1, whereas the sheath is composed of five proteins: FlaA, FlaA1, FlaA2, FlaAP1, and FlaAP2. Asymmetric model of the sheathed filament shown from the convex side (e), concave side (f), and the glycosylated filament core (g), with a perpendicular view of a single-layer filament model shown in (h). Light or dark green: FlaA; light or dark pink: FlaA1; light or dark purple: FlaA2; light or dark blue: FlaAP1; orange: FlaAP2; yellows: FlaB1. i–k Three different views of the architecture of the supercoiled periplasmic flagellar filament formed by three core flagellins (FlaB1-FlaB3), five sheath proteins (FlaA, FlaA1, FlaA2, FlaAP1, and FlaAP2). l The models of the FlaB1 protofilament and FlaA protofilament at the convex side. The total length of the 7-subunit FlaA protofilament is 328 Å. m The models of the FlaB1 protofilament and FlaA2 protofilament at the concave side. The total length of the 7-subunit FlaA1 protofilament is 300 Å. The FlaA1 protofilament is slightly shorter than the FlaA protofilament. Asymmetric architecture of the supercoiled flagellar filament in T. denticola. Cryo-EM structure of the sheathed flagellar filament viewed from the convex (a) and concave (b) sides. Green: FlaA; blue: FlaAP1; purple: FlaA2; pink: FlaA1; orange: FlaAP2; yellow: FlaB1; red: glycan. c Close-up view of the glycosylated filament core. Filament components are colored by protein identity; glycans are shown. d Perpendicular cross-section of the filament revealing a central lumen, a tubular filament core, and a surrounding protein sheath. The core is predominantly formed by FlaB1, whereas the sheath is composed of five proteins: FlaA, FlaA1, FlaA2, FlaAP1, and FlaAP2. Asymmetric model of the sheathed filament shown from the convex side (e), concave side (f), and the glycosylated filament core (g), with a perpendicular view of a single-layer filament model shown in (h). Light or dark green: FlaA; light or dark pink: FlaA1; light or dark purple: FlaA2; light or dark blue: FlaAP1; orange: FlaAP2; yellows: FlaB1. i–k Three different views of the architecture of the supercoiled periplasmic flagellar filament formed by three core flagellins (FlaB1-FlaB3), five sheath proteins (FlaA, FlaA1, FlaA2, FlaAP1, and FlaAP2). l The models of the FlaB1 protofilament and FlaA protofilament at the convex side. The total length of the 7-subunit FlaA protofilament is 328 Å. m The models of the FlaB1 protofilament and FlaA2 protofilament at the concave side. The total length of the 7-subunit FlaA1 protofilament is 300 Å. The FlaA1 protofilament is slightly shorter than the FlaA protofilament. The core of the sheathed filament is mainly formed by glycosylated FlaB1 Assembly of the flagellar filament core in T. denticola is mediated by three FlaB flagellins that share highly similar sequences and overall structures but differ in their stoichiometries in the wild-type and in various flaB mutants34. For example, deletion of flaB3 increases the ratio of FlaB1 to FlaA from 0.71:1 to 1.12:1, while reducing the ratio of FlaB2 to FlaA from 0.61:1 to 0.36:134. These observations suggest that, while FlaB1 is the major flagellin protein, there may be functional redundancy among the three FlaB proteins. Consistent with this hypothesis, our cryo-EM analysis of purified sheathed flagella reveals that the core of the sheathed filament is composed of 11 FlaB1 protofilaments (Fig. 3). Each FlaB1 subunit consists of the conserved D0 and D1 domains (Fig. 3a). Moreover, in agreement with previous mass spectrometry analyses35, five serine or threonine residues (Ser115, Ser126, Thr137, Ser152, and Thr171) are modified by glycans corresponding to 2-methoxy-4,5,6-trihydroxy-hexanoyl-L-glycero-L-manno-nonulosonic acid (Fig. 3a–c and Supplementary Fig. 7e). These glycan modifications cluster around three surface-exposed loops in the D1 domain (loops 1–3), where they appear to stabilize inter-subunit interactions within the filament core (Fig. 3d, e). In addition, FlaB1 protofilaments assemble into a supercoiled filament through conserved longitudinal interactions mediated by residues D106, E92, R91, Q88, and E225, as well as lateral interactions involving residues R223, R244, and N277 (Fig. 3f, g).Fig. 3The flagellar filament core is composed of glycosylated FlaB proteins assembled into a conserved helical structure.a Atomic model of FlaB1 fitted into a single subunit of the cryo-EM map of the sheathed filament core at 2.87 Å resolution. Five extra densities corresponding to glycans are colored in red. b Five O-glycosylation sites (red) identified in the D1 domain of FlaB1, distributed around three flexible surface loops (magenta). c Refined models of glycans at residues S115, S126, T137, S152, and T171 fitted in the corresponding glycan densities. d Longitudinal FlaB1 dimer with glycans shown in red. The glycan attached to S152 appears involved in the interaction between two FlaB1 molecules. e Overall architecture of the glycosylated FlaB1 filament core. f Overview of the longitudinal interactions between two FlaB1 molecules along the 11-start protofilament (left panel). Detailed side-chain interactions (right panel). g Overview of the lateral interactions between two FlaB1 molecules along the 5-start protofilament (left panel). Detailed side-chain interactions (right panel). h Model of FlaB3 fitted into the single subunit map from the non-sheathed filament core at 2.87 Å resolution. Four extra densities corresponding to glycans are colored in red. i, Four O-glycosylation sites (red) identified in FlaB3 D1domain, distributed around three flexible surface loops (magenta). j Refined model of the FlaB3 glycans at residues S115, S126, T137, and T167 fitted in the corresponding glycan densities. k Longitudinal FlaB3 dimer with glycans shown in red. Notably, there is no glycan for S152 in FlaB3. The flagellar filament core is composed of glycosylated FlaB proteins assembled into a conserved helical structure. a Atomic model of FlaB1 fitted into a single subunit of the cryo-EM map of the sheathed filament core at 2.87 Å resolution. Five extra densities corresponding to glycans are colored in red. b Five O-glycosylation sites (red) identified in the D1 domain of FlaB1, distributed around three flexible surface loops (magenta). c Refined models of glycans at residues S115, S126, T137, S152, and T171 fitted in the corresponding glycan densities. d Longitudinal FlaB1 dimer with glycans shown in red. The glycan attached to S152 appears involved in the interaction between two FlaB1 molecules. e Overall architecture of the glycosylated FlaB1 filament core. f Overview of the longitudinal interactions between two FlaB1 molecules along the 11-start protofilament (left panel). Detailed side-chain interactions (right panel). g Overview of the lateral interactions between two FlaB1 molecules along the 5-start protofilament (left panel). Detailed side-chain interactions (right panel). h Model of FlaB3 fitted into the single subunit map from the non-sheathed filament core at 2.87 Å resolution. Four extra densities corresponding to glycans are colored in red. i, Four O-glycosylation sites (red) identified in FlaB3 D1domain, distributed around three flexible surface loops (magenta). j Refined model of the FlaB3 glycans at residues S115, S126, T137, and T167 fitted in the corresponding glycan densities. k Longitudinal FlaB3 dimer with glycans shown in red. Notably, there is no glycan for S152 in FlaB3. Given the essential role of glycosylation in flagellar filament assembly and motility in T. denticola35, these findings elucidate how specific glycan modifications and conserved inter-subunit interactions cooperate to drive filament core assembly and maintain the mechanical stability required to accommodate elastic deformation and supercoiling. Non-sheathed filaments in wild-type cells are formed by glycosylated FlaB3 Comparative analysis of in situ structures of sheathed and non-sheathed filaments in wild-type T. denticola cells reveals that the non-sheathed filament is predominantly composed of glycosylated FlaB3. Similar to FlaB1, FlaB3 contains the D0 and D1 domains, which mediate formation of the filament core through conserved inter-subunit interactions (Fig. 3h). Consistent with mass spectrometry analyses35, FlaB3 is glycosylated at four residues: Ser115, Ser126, Thr137, and Thr167 (Fig. 3i, j). Each glycan is clearly resolved in the cryo-EM maps (Fig. 3i–k). These glycans are positioned on surface-exposed regions of the filament core, suggesting a role in stabilizing inter-subunit contacts. The glycan located between flexible loop 2 and loop 3 at the interface of the FlaB1 longitudinal dimer likely contributes to filament stabilization (Fig. 3b, d). Its absence in the corresponding region of the FlaB3 longitudinal dimer may reduce the stability of this interface and the FlaB1 filament (Fig. 3i, k). Because non-sheathed filaments are consistently observed at the distal portion of flagella in wild-type T. denticola cells, these findings suggest that the three flagellins are sequentially assembled during flagellar biogenesis, with FlaB1 and FlaB2 incorporated at early stages and FlaB3 added later. The major sheath protein FlaA is intimately coupled to the FlaB1 core Our cryo-EM map enabled construction of an atomic model of the major sheath protein FlaA, revealing how it assembles into the bulk of the filament sheath through extensive longitudinal and lateral interactions (Fig. 4a, b). Longitudinal contacts are mediated by a combination of electrostatic and hydrophobic interactions involving D293, D220, and D271 from one FlaA subunit and R104, R292, and R306 from an adjacent subunit (Fig. 4c). Lateral interactions are formed by salt bridges involving an extended peptide region, with key residues K109, K280, and R239 interacting with E345, E342, and E328 on neighboring subunits (Fig. 4d).Fig. 4The T. denticola flagellar sheath is predominantly composed of FlaA proteins.a Atomic model of FlaA fitted into a single FlaA subunit of the cryo-EM map. b Overview of the filament sheath formed by FlaA (light or dark green). Detailed views of longitudinal (c) and lateral (d) interactions between adjacent FlaA subunits. e Cut-away view of the sheathed filament model. One FlaA and one FlaB1 monomer are highlighted in brighter colors. f The FlaA-FlaB1 model showing interactions between FlaA and FlaB1. Detailed views of key interactions: Asn147 of FlaB1 interacting with Lys317 of FlaA (g), and Asp162 of FlaB1 interacting with Ser86 of FlaA (h). i, j Swimming plate assays showing that site-directed mutations in flaB1 significantly impair cell motility. Swimming ring diameters (n = 6 plates) for flaB1-N147A (11.50 ± 0.34 mm), flaB1-D162A (13.60 ± 0.50 mm), and flaB1-N147A/D162A (13.20 ± 0.39 mm) are smaller than those of wild-type cells (20.50 ± 0.54 mm). The results shown here are representative of one experiment with at least five technical replicates. The experiment was repeated independently at least two times. The results are expressed as the mean. Source data are provided as a Source Data file. Cryo-ET reconstructions reveals filament with variable diameters in ΔflaA (k), flaB1-D162A (l), and flaB-N147A mutants (m, n). Sheathed filaments (Thick) are labeled with green arrows, non-sheathed filaments are labeled with yellow arrows, and very thin protofilaments are labeled with purple arrows. Scale bars are 60 nm. The T. denticola flagellar sheath is predominantly composed of FlaA proteins. a Atomic model of FlaA fitted into a single FlaA subunit of the cryo-EM map. b Overview of the filament sheath formed by FlaA (light or dark green). Detailed views of longitudinal (c) and lateral (d) interactions between adjacent FlaA subunits. e Cut-away view of the sheathed filament model. One FlaA and one FlaB1 monomer are highlighted in brighter colors. f The FlaA-FlaB1 model showing interactions between FlaA and FlaB1. Detailed views of key interactions: Asn147 of FlaB1 interacting with Lys317 of FlaA (g), and Asp162 of FlaB1 interacting with Ser86 of FlaA (h). i, j Swimming plate assays showing that site-directed mutations in flaB1 significantly impair cell motility. Swimming ring diameters (n = 6 plates) for flaB1-N147A (11.50 ± 0.34 mm), flaB1-D162A (13.60 ± 0.50 mm), and flaB1-N147A/D162A (13.20 ± 0.39 mm) are smaller than those of wild-type cells (20.50 ± 0.54 mm). The results shown here are representative of one experiment with at least five technical replicates. The experiment was repeated independently at least two times. The results are expressed as the mean. Source data are provided as a Source Data file. Cryo-ET reconstructions reveals filament with variable diameters in ΔflaA (k), flaB1-D162A (l), and flaB-N147A mutants (m, n). Sheathed filaments (Thick) are labeled with green arrows, non-sheathed filaments are labeled with yellow arrows, and very thin protofilaments are labeled with purple arrows. Scale bars are 60 nm. Importantly, each FlaA subunit directly engages FlaB1 to form a repeating core-sheath composite unit that propagates along the filament core (Fig. 4e). At the FlaB1–FlaA interface, specific interactions were observed between N147 of FlaB1 and K317 of FlaA, as well as between D162 of FlaB1 and S86 of FlaA (Fig. 4f–h), indicating an intimate mechanical coupling between the FlaB1 core and the FlaA sheath. FlaB1 site-directed mutants destabilize the filament sheath and impair motility To test whether the two identified FlaB1–FlaA interactions are required for flagellar assembly and spirochetal motility, residues N147 and D162 of FlaB1 were individually or combinatorially substituted with alanine. Mutations were introduced by in-frame replacement of the wild-type flaB1 gene with the corresponding mutant alleles (Supplementary Fig. 2e, top). The resulting strains were validated by PCR (Supplementary Fig. 2e, bottom) and DNA sequencing. The three point mutants were designated flaB1-N147A, flaB1-D162A, and flaB1-N147A-D162A. Immunoblot analysis showed that levels of both FlaA and FlaB proteins were largely unchanged in the flaB1-N147A-D162A double mutant, slightly reduced in the flaB1-N147A mutant, and modestly increased in the flaB1-D162A mutant (Supplementary Fig. 2e and Supplementary Table 2,3). Despite these modest differences in protein abundance, all three mutants exhibited pronounced motility defects relative to the wild-type strain (Fig. 4i, j and Supplementary Table 3). Swimming plate assays revealed significantly reduced swimming ring diameters (n = 6 plates) for flaB1-N147A (11.50 ± 0.34 mm), flaB1-D162A (13.60 ± 0.50 mm), and flaB1-N147A-D162A (13.20 ± 0.39 mm), compared with wild type (20.50 ± 0.54 mm) (Fig. 4i, j). To further assess the structural consequences of these mutations, cryo-ET was used to compare flaB1-D162A and flaB1-N147A cells with the ΔflaA mutant. As expected, ΔflaA filaments were thin and non-sheathed (Fig. 4k)34. In contrast, filaments in both flaB1-D162A and flaB1-N147A cells were predominantly thick and sheathed (Fig. 4l–n). Notably, thin protofilaments were also observed in these mutants, indicating that although FlaA is able to assemble into a sheath, individual sheath protofilaments can partially detach from the filament core (Fig. 4k–n and Supplementary Table 3). Together, these findings suggest that mutations at FlaB1 residues N147 and D162 compromise motility, supporting their roles in core–sheath interactions and in maintaining the assembly and function of sheathed flagellar filaments. While not essential for sheathed filament assembly, these interactions are crucial for sustaining core–sheath stability, thereby reinforcing mechanical integrity and enabling efficient motility in T. denticola. FlaA1 and FlaA2 assemble along the concave side of the supercoiled filament The concave side of the supercoiled filament is associated with two minor sheath proteins: FlaA1 and FlaA2 (Fig. 5a). Both FlaA1 and FlaA2 are homologs of FlaA (Fig. 5b, c and Supplementary Fig. 1). Notably, two FlaA1 protofilaments assemble into a centrosymmetric architecture consisting of two opposing lobes arranged around the center as longitudinal dimers, flanked bilaterally by two FlaA2 protofilaments (Fig. 5a). Stabilization of this asymmetric extension is mediated by FlaA1–FlaA1 interactions involving residues E197, K203, and D221 from one subunit and R176, D227, and H72 from the adjacent subunit (Fig. 5d), while FlaA2 provides additional longitudinal reinforcement (Fig. 5e–g).Fig. 5FlaA1 and FlaA2 form the concave side of the supercoiled filament in T. denticola.a Surface view of the concave side of the supercoiled filament sheath, mainly composed of FlaA1 (light or dark pink) and FlaA2 (light or dark purple). Atomic models of FlaA2 (b) and FlaA1 (c) fitted into their corresponding cryo-EM maps of the filament sheath. d Detailed views of FlaA1 dimer interactions, showing longitudinal (top) and lateral (bottom) contacts. e Detailed view of longitudinal dimer interactions between FlaA2 subunits. Close-up views of key interactions between FlaA2 and FlaB1, including Glu146 of FlaB1 interacting with Tyr160 of FlaA2 (f) and Thr144 of FlaB1 interacting with Glu158 of FlaA2 (g). h No specific interaction between FlaB1 and FlaA1 is observed. i–k Cryo-ET reconstructions of the ΔflaA2 mutant reveal filaments with variable diameters, indicating the presence of sheathed filaments (green arrows), non-sheathed filaments (yellow arrows), and protofilaments (purple arrows). l, m Cryo-ET reconstructions of the ΔflaA1 mutant similarly show filaments with variable diameters, indicating the coexistence of sheathed filaments (green arrows), non-sheathed filaments (yellow arrows), and protofilaments (purple arrows). Scale bars are 60 nm. FlaA1 and FlaA2 form the concave side of the supercoiled filament in T. denticola. a Surface view of the concave side of the supercoiled filament sheath, mainly composed of FlaA1 (light or dark pink) and FlaA2 (light or dark purple). Atomic models of FlaA2 (b) and FlaA1 (c) fitted into their corresponding cryo-EM maps of the filament sheath. d Detailed views of FlaA1 dimer interactions, showing longitudinal (top) and lateral (bottom) contacts. e Detailed view of longitudinal dimer interactions between FlaA2 subunits. Close-up views of key interactions between FlaA2 and FlaB1, including Glu146 of FlaB1 interacting with Tyr160 of FlaA2 (f) and Thr144 of FlaB1 interacting with Glu158 of FlaA2 (g). h No specific interaction between FlaB1 and FlaA1 is observed. i–k Cryo-ET reconstructions of the ΔflaA2 mutant reveal filaments with variable diameters, indicating the presence of sheathed filaments (green arrows), non-sheathed filaments (yellow arrows), and protofilaments (purple arrows). l, m Cryo-ET reconstructions of the ΔflaA1 mutant similarly show filaments with variable diameters, indicating the coexistence of sheathed filaments (green arrows), non-sheathed filaments (yellow arrows), and protofilaments (purple arrows). Scale bars are 60 nm. In contrast to the strong FlaA–FlaB1 interactions, interactions between FlaA2 and FlaB1 are markedly weaker, and no significant hydrogen bonds or salt bridges were observed in the FlaA1–FlaB1 interfaces (Fig. 5h), indicating reduced core–sheath coupling along the concave side of the supercoiled sheathed filament. Cryo-ET analyses of ΔflaA1 and ΔflaA2 mutants reveal filaments with variable diameters, consistent with the ability of both mutants to assemble sheathed filaments but with reduced sheath stability. This instability results in the frequent presence of non-sheathed filaments and detached protofilaments. Importantly, both mutant strains exhibit pronounced motility defects (Fig. 1e–h), demonstrating that FlaA1 and FlaA2 are critical contributors to the mechanical integrity and structural stability of the supercoiled, sheathed flagellar filament. FlaA associated proteins FlaAP1 and FlaAP2 strengthen the supercoiled filament FlaAP1 and FlaAP2 are two sheath-associated proteins that localize to the concave side of the sheathed filament (Fig. 6a). FlaAP1 is a HEAT-repeat-containing protein composed of tandem repeats of two antiparallel α-helices connected by turns and arranged along a common axis, with adjacent repeats linked by flexible inter-repeat loops37. The FlaAP1s assemble into a longitudinal chain stabilized by extensive hydrophobic interactions and specific hydrogen bonds between N208 and K196 of one subunit and E53 and Y37 of the adjacent subunit (Fig. 6b, c). Occupation of this position by FlaAP1 induces a rearrangement of the terminal peptide of FlaA2 (Fig. 6f, g).Fig. 6FlaA-associated proteins FlaAP1 and FlaAP2 assemble along the concave side of the supercoiled filament.a Surface view of the concave side of the supercoiled filament sheath, minorly supported by FlaAP1 (light or dark blue) and FlaAP2 (orange). b Atomic model of FlaAP1 fitted into a single FlaAP1 subunit map. c Interfaces of two adjacent FlaAP1 molecules. Top, longitudinal contacts between adjacent subunits; bottom, hydrophobic interactions stabilizing the FlaAP1 dimer. d Atomic model of FlaAP2 fitted into a single FlaAP2 subunit map. e Interactions of FlaAP2 with three neighboring subunits, including FlaA2, longitudinal FlaA, and lateral FlaA. FlaAP1 binding induces a conformational rearrangement in the adjacent FlaA2. A flexible peptide segment (lighter color), which occupies a lateral position in the absence of FlaAP1 (g), shifts outward to accommodate FlaAP1 binding (f). h, i FlaAP2 binding alters the lateral FlaA conformation. A flexible peptide segment (lighter color) relocates from its original position (i) to enable FlaAP2 binding (h). Cryo-ET reconstructions reveal filaments with variable diameters in ΔflaAP2 (j, k) and ΔflaAP1 (l) mutant cells. Scale bars are 60 nm. m Model illustrating the assembly of minor sheath proteins along the concave side of the supercoiled filament while the major sheath protein forms the convex side of the sheath, enabling the filament to wrap tightly around the cell body. PG peptidoglycan, IM inner membrane. FlaA-associated proteins FlaAP1 and FlaAP2 assemble along the concave side of the supercoiled filament. a Surface view of the concave side of the supercoiled filament sheath, minorly supported by FlaAP1 (light or dark blue) and FlaAP2 (orange). b Atomic model of FlaAP1 fitted into a single FlaAP1 subunit map. c Interfaces of two adjacent FlaAP1 molecules. Top, longitudinal contacts between adjacent subunits; bottom, hydrophobic interactions stabilizing the FlaAP1 dimer. d Atomic model of FlaAP2 fitted into a single FlaAP2 subunit map. e Interactions of FlaAP2 with three neighboring subunits, including FlaA2, longitudinal FlaA, and lateral FlaA. FlaAP1 binding induces a conformational rearrangement in the adjacent FlaA2. A flexible peptide segment (lighter color), which occupies a lateral position in the absence of FlaAP1 (g), shifts outward to accommodate FlaAP1 binding (f). h, i FlaAP2 binding alters the lateral FlaA conformation. A flexible peptide segment (lighter color) relocates from its original position (i) to enable FlaAP2 binding (h). Cryo-ET reconstructions reveal filaments with variable diameters in ΔflaAP2 (j, k) and ΔflaAP1 (l) mutant cells. Scale bars are 60 nm. m Model illustrating the assembly of minor sheath proteins along the concave side of the supercoiled filament while the major sheath protein forms the convex side of the sheath, enabling the filament to wrap tightly around the cell body. PG peptidoglycan, IM inner membrane. FlaAP2 adopts a distinctive diamond-like fold and interacts simultaneously with three neighboring subunits – FlaA2, a longitudinal FlaA, and a lateral FlaA (Fig. 6d, e). Notably, interaction between FlaAP2 and the longitudinal FlaA induces a rearrangement of the terminal peptide of FlaA (Fig. 6h, i). In both cases, the corresponding terminal peptide densities of FlaA and FlaA2 are not resolved in the presence of FlaAP1 or FlaAP2, likely reflecting increased conformational flexibility. Cryo-ET analyses reveal that sheathed flagellar filaments are present in both ΔflaAP1 and ΔflaAP2 mutants; however, numerous thin protofilaments are frequently observed detached from the filament core, indicating reduced sheath stability in both mutants (Fig. 6j–l). Consistent with these structural defects, both mutant strains exhibit significantly impaired motility (Fig. 1e–j), implicating FlaAP1 and FlaAP2 as key contributors to the structural stability and mechanical performance of the supercoiled flagellar filament in T. denticola. Taken together, longitudinal cross-sectional views of the sheathed filament highlight extensive hydrophobic regions at interlayer interfaces, forming hydrophobic contacts that contribute to overall filament integrity (Supplementary Fig. 10). Representative hydrophobic interactions within FlaA longitudinal dimers and within FlaAP1 longitudinal dimers further underscore the importance of hydrophobic packing in maintaining filament architecture (Supplementary Fig. 10). Electrostatic surface analyses of the sheathed filament reveal distinct charge distributions across the structure (Supplementary Fig. 11). The outer surface of the sheath is predominantly negatively charged, whereas positively charged regions on the inner surface of the sheath are positioned adjacent to the negatively charged, glycosylated outer surface of the filament core (Supplementary Fig. 11). This complementary charge distribution is consistent with electrostatic contributions to core–sheath association and stabilization of the sheathed filament. Discussion To infect and disseminate within mammalian hosts, spirochetes have evolved a morphology and mode of motility adapted for propulsion through viscous media and dense tissue barriers1. Central to this adaptation are periplasmic flagella, which reside within the periplasmic space and are fundamentally distinct from the external flagella. Rather than extending into the extracellular milieu, spirochetal flagella wrap tightly around the cell body in an intrinsically supercoiled configuration, and their rotation drives coordinated deformation and propulsion of the entire cell, enabling efficient motility in environments where external flagella are ineffective16. Here, through an integrated structural, genetic, and biochemical approach, we define the asymmetric and multilayered architecture of the supercoiled flagellar filament in T. denticola. The filament core is composed of multiple glycosylated flagellins with distinct spatial distributions, encased within an asymmetric sheath in which the major sheath protein FlaA forms most of the surface, while four minor sheath proteins assemble preferentially along the concave side of the supercoil (Fig. 6m; Supplementary Movie 3). The asymmetric architecture of the T. denticola sheathed flagellar filament is broadly similar to that of Leptospira, although the latter is more complex due to the presence of additional sheath proteins involved in its assembly27–29. Importantly, genetic disruption of any sheath component or of the sheath–core interactions impairs T. denticola motility (Fig. 1e–j; Fig. 4i, j), demonstrating that this architectural asymmetry is not simply structural but functionally essential for spirochetal motility. The presence of extended non-sheathed distal regions supports a model in which flagellar core assembly precedes sheath formation. Flagellins are exported by the fT3SS18,20,21 and polymerize at the distal tip of the nascent filament to form a heterogeneous, supercoiled core, with distinct flagellin isotypes occupying spatially segregated regions along the filament axis. Similar organization has been recently observed in the sheathed flagella of Vibrio cholerae38 and Helicobacter pylori39, suggesting that differential flagellin incorporation is a common strategy during filament assembly. However, unlike the membranous sheaths of V. cholerae and H. pylori, which are derived from the outer membrane, the periplasmic flagellar core is encased by sheath proteins that are independently exported via the Sec secretion pathway18,22,23 and polymerize onto the pre-formed supercoiled core (Supplementary Movie 4). Glycosylation and flexible surface-exposed loops on the core facilitate intimate core–sheath interactions that likely enhance filament rigidity and torsional resistance. The asymmetric, multilayered architecture of the periplasmic flagellar filament has direct consequences for spirochetal hydrodynamics by generation of propulsive forces40,41. Such asymmetry stabilizes the supercoiled geometry during rotation and promotes efficient transmission of motor-generated torque. Rather than functioning as an isolated propeller, rotation of the periplasmic filament drives deformation and counter-rotation of the entire cell body (Supplementary Movie 4), providing a mechanistic basis for efficient spirochetal motility in highly viscous host environments. The mechanical performance of the periplasmic flagellar filament is tightly coupled to its genetic regulation. In T. denticola, four flagellar filament genes are partitioned between two transcriptional programs controlled by σ⁷⁰ and σ²⁸, providing temporal and quantitative control over the production of distinct filament components34,42. This layered regulation likely governs the sequence and stoichiometry of filament assembly, thereby fine-tuning filament length, thickness, stiffness, and core–sheath coupling. Because these morphological parameters are directly linked to distinct swimming mechanisms, as demonstrated in spirochetes such as Treponema, Leptospira, and Borrelia3,16, regulatory control over filament composition directly translates into control of flagellar assembly and biophysical properties and spirochetal swimming behavior. In sum, our studies reveal an asymmetric architecture of the supercoiled periplasmic flagellar filament that is required for tight wrapping around the spirochetal cell body, efficient torque transmission, and whole-cell deformation. We establish the periplasmic flagellar filament as a mechanically integrated and genetically tunable organelle optimized for rotation-driven propulsion, providing a structural and regulatory framework for understanding spirochetal motility in complex environments. Methods Bacterial strains, culture conditions, and oligonucleotide primers T. denticola ATCC 35405 (wild type, WT) was used in this study33. Bacterial cells were grown in either tryptone-yeast extract-gelatin-volatile fatty acids-serum (TYGVS) medium or Oral Bacterial Growth Medium (OBGM) at 37 °C in an anaerobic chamber in the presence of 90% nitrogen, 5% carbon dioxide, and 5% hydrogen43,44. T. denticola isogenic mutants were grown with appropriate antibiotic for selective pressure as needed: erythromycin (50 µg/ml) or gentamicin (20 µg/ml). Escherichia coli 5α strain (New England Biolabs, Ipswich, MA) was used for DNA cloning. The E. coli strains were cultivated in lysogeny broth (LB) supplemented with appropriate concentrations of antibiotics for selective pressure as needed: ampicillin (100 µg/ml). The oligonucleotide primers for PCR amplifications used in this study are listed in Supplementary Table 1. These primers were synthesized by IDT (Integrated DNA Technologies, Coralville, IA). Periplasmic flagella isolation Isolation of the periplasmic flagella was performed as previously described34. Briefly, 500 ml of the mid-logarithmic (log)-phase T. denticola cultures ( ~ 5 × 108 cells/ ml) were centrifuged at 5000 × g for 20 min at 4 oC. The cells were washed four times with phosphate-buffered saline (PBS, pH 7.4) and once with T1 buffer (0.15 M Tris-HCl buffer, pH 6.8). The final pellets were resuspended in 30 ml of T1 buffer. Three milliliters of 10% Triton X-100 were slowly added, mixed and incubated for 1 h at room temperature. And then, 3 ml of 200 µg/ml mutanolysin (Sigma-Aldrich, St. Louis, MO) was added slowly, followed by addition of 300 µl of T2 buffer (0.1 M Tris-HCl buffer, pH 6.8). The resultant mixture was first incubated for 2 h at room temperature and then at 4 oC overnight. After the incubation, 600 µl of 0.1 M MgSO4 was added, followed by addition of 600 µl of T2 buffer. The mixture was incubated for 5 min at room temperature and then centrifuged at 17,000 × g for 15 min at 4 oC. The supernatant containing periplasmic flagella was collected and 2 ml of 20% PEG 8000 (Alfa Aesar, London, UK) in 1 M NaCl was added and then incubated for 30 min on ice. The resultant sample was centrifuged at 27,000 × g for 30 min at 4 oC. The pellet containing periplasmic flagella was resuspended in alkaline solution (0.1 M KCl, 0.5 M sucrose, 0.1% Triton X-100, 50 mM sodium bicarbonate, pH 11) and incubated for 1 h on ice. The periplasmic flagella were finally collected by centrifugation at 80,000 × g for 45 min at 4 oC and washed once in 20 mM Tris-HCl buffer, pH 8. The flagellar pellet was resuspended in water and stored at 4 oC for further analyses. Electrophoresis and immunoblotting analyses Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting analyses were carried out as previously described45. For immunoblotting analysis, T. denticola cells were harvested in the mid-log phase ( ~ 5 × 108 cells/ml) Equal amounts of whole-cell lysates ( ~ 5 μg) were separated on SDS-PAGE. The resultant gels were subjected to Coomassie blue staining or immunoblotting analyses. The antibodies against T. pallidum FlaBs and T. denticola FlaA and DnaK are previously described35,46. The stoichiometry of each flagellar filament protein was analyzed using Image Lab software from Bio-Rad (Bio-Rad). Constructions of T. denticola flagellar mutants The diagrams in Supplementary Fig. 2 illustrate the deletion of the open reading frames (ORFs) of TDE1408, TDE1409, TDE2349, and TDE1480. TDE1408 is used as an example to describe how this vector (TDE1408::ermB) was constructed and used to delete the target gene. To construct TDE1408::ermB, the TDE1408 upstream flanking region (UR) and a previously described erythromycin B resistant cassette (ermB)47 were PCR amplified with primers P1/P2 and P3/P4, respectively, and then fused together with primers P1/P4 to generate UR-ermB. The downstream flanking region of TDE1408 (DR) was PCR amplified with primers P5/P6, and subsequently fused to UR-ermB fragment by PCR using primers P1/P6. The final fragment, UR-ermB-DR, was cloned into the pGEM-T easy vector (Promega, Madison, WI), generating TDE1408∷ermB (Supplementary Fig. 2a). The same method was used to construct TDE1409::ermB and TDE2349::ermB. To construct plasmid (TDE2349::ERM) to delete TDE2349, the following method was used. The TDE2349 upstream flanking region (UR) and the downstream flanking region (DR) were PCR amplified with primers P13/P14 and P17/P18, respectively, and then fused together with primers P13/P18 to generate UR-DR fragment. The UR-DR fragment was cloned into the pMD19 T-vector (Takara Bio USA, Inc, Mountain View, CA). The full-length erythromycin cassette was PCR amplified with primers P15/P16, generating Fragment ERM, which was subsequently cloned into the pGEM-T easy vector (Promega). The Fragment UR-DR and ERM were digested using NotI and then ligated, generating the TDE2349::ERM plasmid. The resulting constructs were transformed into T. denticola wild-type competent cells via electroporation and then plated on soft agar containing erythromycin, as previously described42. The deletions were confirmed by PCR (Supplementary Fig. 2, bottom panels). The resultant mutants were designated as ΔTDE1408, ΔTDE1409, ΔTDE2349, and ΔTDE1480, respectively. Source data are provided as a Source Data file. Construction of flaB1 site-directed mutants in T. denticola The vector (FlaB1::ermB) shown in Supplementary Fig. 2e was constructed to replace the wild-type flaB1 (TDE1477) with genes carrying site-directed mutations at amino acids N147A, D162A, or both N147A and D162A, using two-step PCR and DNA cloning as previously described42. Briefly, to construct this plasmid, the upstream flanking region of FlaB1 along with the wild-type flaB1 gene (UR), ermB, and the downstream flanking region of flaB1 (DR) were PCR amplified using primers P28/P29, P30/P31, and P32/P33, respectively. Then, UR and ermB were PCR ligated using P28/P31 to generate the UR-ermB fragment, which was subsequently PCR ligated to DR using primers P28/P33. The final PCR product, UR-ermB-DR, was cloned into the pJET1.2 vector (Thermo Fisher Scientific, San Jose, CA). Site-directed mutagenesis of flaB1 in this plasmid was performed using the Q5 Site-Directed Mutagenesis Kit (New England Biolabs) according to the manufacturer’s instructions. The primers used to replace N147 and N162 with Ala were P35/P36 and P37/P38, respectively. The resulting mutations were confirmed by DNA sequencing analysis. The resulting constructs were transformed into T. denticola wild-type competent cells via electroporation and then plated on soft agar containing erythromycin, as previously described42. Site-directed mutations in the resulting colonies were confirmed by PCR followed by DNA sequencing. Source data are provided as a Source Data file. Bacterial swimming plate assay and motion tracking analysis A swimming plate assay of T. denticola was performed as previously described34. Swimming plate assays were performed with plates containing 50% (v/v) TYGVS:50% (v/v) PBS with 0.35% (w/v) SeaPlaque agarose. Briefly, 10 ml of the late-log phase T. denticola cultures (109 cells/ ml) were harvested by centrifugation at 5000 × g for 10 min at room temperature. The cell pellets were resuspended in ~40 µL of spent culture supernatant. Each plate was inoculated with 4 µL of resuspended cell cultures and incubated anaerobically at 37 °C for 2–3 days to allow the cells to swim out. The plates were incubated anaerobically at 37 °C for 3–5 days to allow the cells to swim out. The diameters of the swimming rings were measured in millimeters. As a negative control, a previously constructed T. denticola non-motile mutant, ΔflgE, was included to determine the initial inoculum size48. The average diameters of each individual strains were calculated from three or six independent plates; and the results are represented as the mean of diameters ± standard error of the mean (SEM). For motion-tracking analysis, 100 μl of mid-log phase T. denticola cultures was first diluted (1:1) in OBGM medium and then 10 μl of diluted cultures was mixed with an equal volume of 2% methylcellulose with a viscosity of 4000 cp (MC4000). T. denticola cells were videotaped and edited using a computer-based bacterial tracking system alongside the software Velocity (Improvision Inc., Coventry, UK), as previously described34. For each bacterial strain, the cells were recorded for up to 30 s. Source data are provided as a Source Data file. In-gel trypsin digestion The excised bands were cut into ~2 mm cubes and subjected to in-gel digestion49,50. The excised gel pieces were washed consecutively with 200 µL deionized water, followed by 200 µL 50 mM ammonium bicarbonate in water/50% acetonitrile (ACN) and finally 200 µL 100% ACN. The dehydrated gel pieces were dried in a speed vacuum (SpeedVac SC110 Thermo Savant, Milford, MA) and reduced with 10 mM DTT (dithiothreitol) (w/v) in 100 mM ammonium bicarbonate in water for 1 h at 60 °C, then alkylated by adding 55 mM iodoacetamide in 100 mM ammonium bicarbonate (w/v) and incubation at room temperature, in the dark, for 45 min. Wash steps were repeated as described above. The gel pieces were dried in a speed vacuum and rehydrated with 100 µl trypsin (Promega Sequencing Grade) at 15 ng/µl (w/v) in 50 mM ammonium bicarbonate/10% ACN on ice for 20 min, topped with 150 µl 50 mM ammonium bicarbonate, 10% ACN (just until gel was covered), and incubated at 37 °C for 16 h. The digestion was stopped by addition of 10 µl 100% formic acid (FA) in water, incubated at room temperature for 5 min and the supernatant transferred to a clean 1.5 ml polypropylene low-bind microfuge tube. The gel pieces were then extracted twice by adding 150 μl of 50% ACN/5% FA and vortexing at 1800rpm for 30 min followed by sonication for 5 min, and once by adding 150 μl of 75% ACN/5% FA with incubation at room temperature for 5 min. The extraction supernatants were pooled with the first supernatant, dried in a speed vacuum. The final sample was reconstituted into 2% ACN/0.5% FA and filtered through a 0.22 µm cellulose acetate spin filter (Cornin75Costar Spin-X) prior to nanoLC-MS/MS analysis. Protein identification by nano LC/MS/MS analysis The analysis was carried out using an Orbitrap FusionTM TribridTM (Thermo Fisher Scientific) mass spectrometer equipped with a nanospray Flex Ion Source, and coupled with a Dionex UltiMate 3000 RSLCnano system (Thermo, Sunnyvale, CA)49. The peptide samples (10 μL) were injected onto a PepMap C-18 RP viper trapping column (5 µm, 100 µm i.d x 20 mm) at 20 µL/min flow rate for rapid sample loading and then separated on a PepMap C-18 RP nano column (2 µm, 75 µm x 25 cm) at 35 °C. The tryptic peptides were eluted in a 90-min gradient of 5% to 35% ACN in 0.1% formic acid at 300 nL/min, followed by an 8-min ramping to 90% ACN-0.1% FA and an 8-min hold at 90% ACN-0.1% FA. The column was re-equilibrated with 0.1% FA for 25 min prior to the next run. The Orbitrap Fusion was operated in positive ion mode with spray voltage set at 1.2 kV and source temperature at 275 °C. External calibration for FT, IT, and quadrupole mass analyzers were performed. In data-dependent acquisition (DDA) analysis, the instrument was operated using FT mass analyzer in MS scan to select precursor ions followed by 3 s “Top Speed” data-dependent CID ion trap MS/MS scans at 1.6 m/z quadrupole isolation for precursor peptides with multiple charged ions above a threshold ion count of 10,000 and normalized collision energy of 30%. MS survey scans at a resolving power of 120,000 (fwhm at m/z 200), for the mass range of m/z 300–1600. Dynamic exclusion parameters were set at 50 s of exclusion duration with ±10 ppm exclusion mass width. All data were acquired under Xcalibur 4.3 operation software (Thermo-Fisher Scientific). Source data are provided as a Source Data file. Data analysis The DDA raw files with MS and MS/MS were subjected to database searches using Proteome Discoverer (PD) 2.4 software (Thermo Fisher Scientific) with the Sequest HT algorithm. The PD 2.4 processing workflow containing an additional node of Minora Feature Detector for precursor ion-based quantification was used for protein identification and relative quantitation of identified peptides. The database search was conducted against T. denticola ATCC 35405 NCBI database that contains 3704 sequences. The peptide precursor tolerance was set to 10 ppm and fragment ion tolerance was set to 0.6 Da. Oxidation of M, deamidation of N and Q were specified as dynamic modifications of amino acid residues; protein N-terminal acetylation, M-loss and M-loss plus acetylation were set as variable modifications, carbamidomethyl C was specified as a static modification. Only high confidence peptides defined by Sequest HT with a 1% FDR by Percolator were considered for confident peptide identification. Source data are provided as a Source Data file. Cryo-EM sample preparation For in situ data, T. denticola cells were centrifuged at 2000 × g for ~5 min in 1.5 mL tubes, and the resulting pellet was gently rinsed with phosphate-buffered saline (PBS). The cell pellet was resuspended in PBS to a final optical density (OD600) of 1.0 for plunge-freezing preparation. For in vitro data, 4 µl aliquot of purified flagellar filaments was applied onto a freshly glow-discharged grid (200 mesh, R2/1, Quantifoil). The grids were blotted with filter paper (WhatmanTM) for ~6 s and rapidly plunged into a liquid ethane-propane mixture using a GP2 plunger (Leica). During the plunge-freezing process, the GP2 environmental chamber was maintained at 25 °C and 90% relative humidity to optimize vitrification. In situ cryo-EM data collection and processing Cryo-EM specimen was imaged using a 300 kV Titan Krios electron microscope (Thermo Fisher Scientific) equipped with a field emission gun and a post-GIF K3 Summit direct electron detector (Gatan). Images were recorded in nanoprobe mode with the following settings: magnification at ×81,000, spot size at 5, illumination aperture at 1.22 µm, slit width at 20 eV, C2 aperture at 50 or 70 µm, and objective aperture at 100 µm. A customized multishot script was used to directly image the cell tips on the cryo-EM grids with SerialEM51, applying defocus values from −1.6 µm to −2.2 µm. The total electron dose was ~70 e-/Å2. The movies were collected in super-resolution mode. An F-crop factor of 1/2 was applied during Patch Motion Correction, resulting in an effective pixel size of 1.068 Å/pixel. A total of 6672 micrographs were collected from T. denticola wild-type cells. All micrographs were first motion-corrected52 and then subjected to Patch CTF Estimation53 in cryoSPARC36. Filament particles were automatically picked using the Filament Tracer54 in cryoSPARC with a filament diameter of 240 Å, a minimum inter-particle distance of 36 Å, and a minimum filament length of 240 Å. Selected particles were extracted with a box size of 720 pixels and Fourier-binned to 360 pixels. 2D classification was used to remove non-filament and low-quality particles, yielding a dataset of 379,837 particles. An initial filament model was generated by ab initio reconstruction, followed by heterogeneous and local refinements for 3D reconstruction. Helical symmetry parameters, including helical rise and twist, were determined using the Symmetry Search Utility in cryoSPARC36. The top-ranked solution, with a helical rise of 4.80 Å and a twist of 65.29°, was selected. Using these parameters, the final helical reconstruction of the sheathed flagellar filament in T. denticola reached a near-atomic resolution of 3.20 Å. Subsequently, 3D classification was performed using a sheath-focused mask. Three distinct classes were identified based on sheath features: symmetric sheath, asymmetric sheath, and non-sheathed filaments. Particles from each class were subjected to local refinement to further improve resolution. During refinement, the rotational search range was progressively reduced from 20° to 2°, the translational search range from 10 Å to 2 Å, and the initial low-pass filter from 12 Å to 3 Å. This process yielded three final maps corresponding to symmetric sheath, asymmetric sheath, and non-sheathed filaments, containing 216,090 (after particle expansion), 148,892, and 112,507 particles, with FSC (0.143) resolutions of 3.20 Å, 3.95 Å, and 3.72 Å, respectively. To investigate the spatial distribution of the sheath along filament curvature and within the cell body, well-aligned particles were re-extracted using a larger box size of 1280 pixels and Fourier-binned to 448 pixels. 2D classification was then performed, and classes exhibiting pronounced curvature and double-membrane features, corresponding to the inner membrane of the cell body, were selected for Homogeneous Reconstruction only. The resulting low-resolution 3D map was visualized in ChimeraX55, revealing that the asymmetric sheath preferentially localizes to the inner curvature of the filament and appears oriented toward the inner membrane. In parallel, selected particles were mapped back onto the original micrographs by Manual Picker to confirm their spatial distribution in the raw data. All cryo-EM maps were further processed using the deep learning-based refinement algorithm DeepEMhancer56 to enhance visualization. In vitro cryo-EM data collection and processing Cryo-EM specimen of purified flagellar filaments from a T. denticola ΔflaB3 mutant was imaged using a 300 kV Titan Krios electron microscope as described above. The total electron dose was ~50 e-/Å2. A total of 4319 micrographs were collected from purified flagellar filaments of the T. denticola ΔflaB3. All micrographs were first motion-corrected52 with an F-crop factor of 1/2 applied and then subjected to Patch CTF Estimation53 in cryoSPARC. Filament particles were automatically picked using the Filament Tracer54 in cryoSPARC with a filament diameter of 240 Å, a minimum inter-particle distance of 24 Å, and a minimum filament length of 480 Å. Selected particles were extracted with a box size of 360 pixels. 2D classification was used to remove non-filament and low-quality particles, yielding a dataset of 643,096 particles. Homogeneous and local refinement was used for 3D reconstruction, using a ~ 20 Å lowpass-filtered in situ cryo-EM map as the initial model. Subsequently, 3D classification was performed using a sheath-focused mask. Three distinct classes were identified based on sheath features: disrupted or low-quality sheath, asymmetric sheath, and non-sheathed filaments. 350,796 particles in disrupted or low-quality sheath class were excluded for the reconstruction. 159,002 particles from asymmetric sheath were further classified using sheath-focused mask by 3D classification until each class displayed significant asymmetric features. Maps from each class were aligned using the Align 3D Maps in cryoSPARC36, with particle alignments updated to place the asymmetric region at a consistent location. All the well-aligned particles were refined by homogeneous refinement and followed by local refinements for 3D reconstruction. 133,271 particles from non-sheathed class were subjected to helical refinement to further improve resolution. During local refinements for all the classes, the rotational search range was progressively reduced from 20° to 2°, and the translational search ranges from 10 Å to 2 Å. This process yielded two final maps corresponding to asymmetric sheath and non-sheathed filaments, containing 159,002, and 133,271 particles, with FSC (0.143) resolutions of 3.44 Å, and 3.67 Å, respectively. Cryo-ET data collection and analysis Frozen-hydrated specimens of T. denticola cells were imaged on a 200 kV Glacios transmission electron microscope (Thermo Fisher Scientific). Tilt series were acquired using SerialEM51 in conjunction with the FastTomo script57 and PaceTomo algorism58, employing a dose-symmetric acquisition scheme. Imaging was conducted at a defocus of −4.8 µm, with a cumulative electron dose of ~70 e-/Å2 distributed evenly across 33 tilts, ranging from −48o to +48o in 3o increments. Raw frames were corrected for beam-induced motion using MotionCorr252, followed by tilt alignment and stacking in IMOD59. Tomographic reconstructions were generated from 4× binned tilt series using Tomo3D60. In total, 71 tilt series were acquired for seven mutants (Supplementary Table 4). Model building The amino acid sequences of FlaB1, FlaB2, FlaB3, FlaA, FlaA1, FlaA2, FlaAP1, and FlaAP2 were retrieved from UniProt (Accession numbers, FlaB1: Q73MN1; FlaB2: Q73NZ6; FlaB3: Q73MN3; FlaA: Q73M00; FlaA1: Q73MU9; FlaA2: Q73MU8; FlaAP1: Q73K73; FlaAP2: Q73MM8) and used as input for AlphaFold361 to generate predicted structural models. The predicted models were fitted into the cryo-EM map using ChimeraX55, followed by interactive manual adjustment in Coot62 and real-space refinement in Phenix63. The final monomer models exhibited excellent stereochemistry, as evaluated by PHENIX. Phenix reported a map-to-model correlation coefficient of 0.70 for ∆flaB3 sheathed filament, 0.72 for ∆flaB3 non-sheathed filament, 0.68 for WT non-sheathed filament, which served as robust cross-validation metrics for evaluating map quality and the accuracy of C1 symmetry parameters. Furthermore, several residues extended beyond the cryo-EM density, indicating poor model-to-map fit and the local flexibility. Sequence alignments were performed using Clustal Omega64. ChimeraX55 was used for cryo-EM map segmentation, visualization, and molecular surface rendering. Interface and interaction analyses among residues were performed using PDBe PISA65. ChimeraX55 and Blender (www.blender.org) were used for making animations. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Movie 1 Supplementary Movie 2 Supplementary Movie 3 Supplementary Movie 4 Reporting Summary Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Movie 1 Supplementary Movie 2 Supplementary Movie 3 Supplementary Movie 4 Reporting Summary Transparent Peer Review file Source data Source Data Source Data