The distinct trimeric structure of the immunodominant chlamydial antigen Major Outer Membrane Protein Authors: Yirui Guo, Megan L Shelby, Patrik D’haeseleer, Sukumar Pal, Anatoli Slepenkin, Beverly V Robinson, Zbyszek Otwinowski, Brent W Segelke, Matthew A Coleman, Dominika Borek, Luis M de la Maza Published: 2026 ABSTRACT AbstractThe chlamydial Major Outer Membrane Protein (MOMP) is a promising subunit vaccine candidate due to its abundance in the outer membrane (OM), the presence of four surface-exposed variable domains (VDs) containing neutralizing and serotyping epitopes, and confirmed T-cell epitopes located in its constant domains (CDs). However, recombinant and denatured forms of MOMP have failed to elicit immune responses comparable to native preparations, indicating that conformation is critical for immunogenicity. Here, we present two cryo-EM structures of native Chlamydia muridarum MOMP, isolated from infectious elementary bodies (EBs). EB MOMP forms a distinct trimer with a stem of three narrow, non-permeable β-barrels and an extracellular, folded antigenic cap that displays VDs. In complex with a neutralizing Fab fragment of a conformational antibody, the cap undergoes structural reorganization that shows how epitope presentation is modulated by immune engagement. These structures reveal the molecular basis of MOMP’s serovar specificity and provide a template for structure-based vaccine design.Subject terms: Cryoelectron microscopy, Molecular conformation, Vaccines Introduction Chlamydia trachomatis (Ct) is the most common bacterial sexually transmitted infection (STI) worldwide, with ~150 million people infected in 20231. Although Ct STIs are treatable with antibiotics, up to 80% of infected women and 50% of infected men are asymptomatic2–5, leading to undiagnosed cases and continued transmission. In women, acute and chronic Ct infections can cause serious complications, including infertility5–7. In developing nations, Ct ocular infections cause trachoma, preventable blindness8. Chlamydia pneumoniae is also a human pathogen that infects the respiratory tract9,10. Zoonotic infections by other chlamydial species11, such as Chlamydia psittaci, can be transmitted to humans and threaten animals, e.g. koalas are an endangered species due to Chlamydia pecorum genital infections12,13. The high prevalence and health and environmental burden of chlamydial infections drive vaccines development efforts14–18. Chlamydiae are Gram-negative obligate intracellular bacteria with a unique developmental cycle19–21. Their extracellular infectious forms, elementary bodies (EBs), are spore-like, stress-resistant, oxidized, and exhibit minimal metabolic activity19,22,23. EBs are endocytosed into host cells, where they enlarge into reticulate bodies (RBs) inside a cytoplasmic inclusion21,24. RBs are non-infectious, reduced, metabolically active, and replicate, with their asynchronous division producing both EBs and RBs until the inclusion ruptures the host cell or is extruded, releasing EBs to infect neighboring cells25. Unlike other Gram-negative bacteria, which have a peptidoglycan layer, EB cell envelope rigidity is conferred by cysteine-rich proteins crosslinked with disulfide bonds, forming the Chlamydial Outer Membrane Complex (COMC)23,26–29. The COMC is reduced during EB-to-RB differentiation, which is necessary for the growth of RB19,28. The main COMC component is the major outer membrane protein (MOMP), which accounts for ~60% of EB outer membrane mass30,31. EB MOMP is an outer membrane β‑barrel protein that is exported across the inner membrane by the Sec pathway and folded into the outer membrane by the BAM complex32,33. EB MOMP forms homotrimers, with each 42 kDa protomer comprising five conserved (CDs) and four variable sequence domains (VDs) that define the 15 major Ct serovars34,35 (Fig. 1a, Supplementary Fig. 1). MOMP has also been shown to act as a porin, diffusing sugars and ATP26,36,37, with an estimated pore size of ~1–2 nm based on liposome swelling assays26,37. Due to its abundance, role in host cell adhesion38, conserved T-cell epitopes in CDs, and the serovar/serogroup-specific protection induced by its extracellular VDs15,34,39–46, MOMP is the leading subunit vaccine candidate. Native trimeric MOMP induces stronger protection than denatured, peptide-based, or recombinant MOMP preparations, likely because such non-native forms lack proper folding14,16,17,41,47,48. Cryo-EM structure of the Cm EB MOMP trimer. a Schematic representation of the MOMP protomer sequence, with VDs and CDs labeled. The N-terminal stopper region is shown as diamond patterns. Transmembrane regions within the CDs are shown in orange. Periplasmic regions within the CDs are shown in gray. Extracellular regions within the CDs are shown in white. b Molecular density map of MOMP at a contour level of 0.2 (arbitrary units dependent on software-specific scaling). c–e The atomic model of the MOMP trimer, colored by chain (green, cyan, magenta), with side, periplasmic, and extracellular views and their dimensions labeled. f Detailed view of the MOMP protomer. The axis of the barrel is indicated by a gray dashed line. The extracellular domain includes ordered loops (circle 1), two β-strands exposed to the solvent (circle 2), two β-strands buried beneath the extracellular loops (circle 3), three β-strands at the trimerization interface (circle 4), and two β-strands extending above the membrane into the extracellular space at the trimer interface (circle 5). The transmembrane domain consists of a 10-stranded β-barrel with a shear number of 12 (circle 6). The cysteine-rich periplasmic region contains short β-turns and a partially disordered loop (circle 7). g, h Interior and exterior sides of the extracellular β-sheets at the trimer interface (circles 4 and 5 in f). Residues with side chains facing the barrel’s interior are mostly non-polar, while those facing the exterior are mostly polar. i, j Interior and exterior sides of the transmembrane β-sheets (circle 6 in f). Residues with side chains facing the barrel’s interior are mostly polar, while those facing the exterior are mostly non-polar. k Arrangement of VDs on the extracellular side (VD1: purple; VD2: red; VD3: yellow; VD4: green). One protomer is highlighted in tube representation, while the other two are shown as ribbon diagrams. Interactions between VDs are highlighted in boxes 1–5. l–n Surface electrostatic potential of MOMP and charge distribution inside the β-barrel, calculated at pH=7, showing that the interior is polar, with electronegative character in the transmembrane portion of the barrel interior. Zoomed in Mg2+ (white sphere) and its coordinating residue D311 and water molecules (pink spheres) are shown as an inset in (m). The importance of the native EB MOMP structure for vaccine design has driven 40 years of preclinical studies of Ct infections using primarily Chlamydia muridarum (Cm) as a model49,50. Cm causes genital infections in mice similar to Ct genital infections in humans51–53. However, progress toward an MOMP-based vaccine has been slowed by two obstacles. First, Chlamydia was genetically intractable until recently54,55, which prevented testing of MOMP functions. Second, recombinant expression systems have not produced MOMP in a properly folded, functional form, hindering efforts to study its structure, biochemical properties, and immunogenicity. Because of this, MOMP vaccine design has moved forward without structural guidance from well-controlled experimental systems. Here, we determined with cryogenic electron microscopy single particle analysis (cryo-EM SPA) structures of Cm MOMP isolated from EBs (Supplementary Fig. 2) alone and in complex with a conformational neutralizing monoclonal antibody Fab (mAb-18b) raised against Cm EB37. The binding of mAb-18b further validated that the isolated Cm MOMP retained its native structure in EB. This work reveals a distinct family of trimeric β-barrel adhesins and addresses longstanding questions about MOMP topology and its epitope structure and cellular presentation. Results Cm EB MOMP forms a distinct trimer exposing tightly folded antigenic cap The structure of Cm EB MOMP was reconstructed in C3 symmetry using cryo-EM SPA at 2.96 Å global resolution (Fig. 1b, Supplementary Fig. 3a–f, Supplementary Table 1). The density map reveals well-ordered side chains in both the β-barrel and extracellular domains (Fig. 1b, Supplementary Fig. 3d). The periplasmic region is less-ordered (Fig. 1b, Supplementary Fig. 3d). Lipid and detergent molecules are visible in the density, marking the EB MOMP transmembrane region (Fig. 1b, c, Supplementary Fig. 3d). The additional reconstruction in C1 symmetry and 3D classification of C3 symmetry expanded particles did not show significant differences compared with the reconstruction performed in C3 symmetry (Supplementary Figs. 4 and 5). The EB MOMP protomer is a 10-stranded anti-parallel β-barrel spanning the membrane (~40% of the polypeptide) with an extended extracellular structure (~50% of the polypeptide) and a periplasmic region (~10% of the polypeptide) (Fig. 1c–e, Supplementary Fig. 6). The β-barrel strands are angled ~41° to the axis (Fig. 1f, Supplementary Fig. 6), consistent with a shear number of 1256,57. In the transmembrane region of the β-barrel, hydrophobic residues face the lipid bilayer pointing outward, while polar residues point inward. Some β-barrel strands extend into the extracellular space, forming a collar-like structure where this inward-outward polarity pattern is reversed, characteristic of globular, soluble proteins (Fig. 1g–j). The extracellular region contains structured loops, which include the four variable domains VD1-VD4 (Fig. 1k, Supplementary Fig. 6), two antiparallel β-strands in solvent-exposed regions that flank VD1 (Fig. 1f, circle 2), two other short β-strands stabilized by VD3 and the extracellular β-barrel collar (Fig. 1f, circle 3 and Supplementary Fig. 6), and five β-strands at the trimer interface (Fig. 1f, circles 4 and 5) that extend from the β-barrel to form a β-sheet. VDs intra- and inter-protomer interactions stabilize the EB MOMP trimer. VD1 of one protomer (VD11) interacts with VD2 of the same protomer (VD21) and with VD4 of the neighboring protomer (VD42); VD21 interacts with VD31, VD41, and VD42; VD31 interacts with VD21 and VD41; and VD41 interacts with VD21, VD31, VD13, VD23, VD42 and VD43 (Fig. 1k, Supplementary Fig. 7a–i). These extracellular interactions are primarily mediated by hydrogen bonds (Supplementary Fig. 7a–d), with fewer non-polar interactions (Supplementary Fig. 7e, f). Each protomer contains a magnesium cation octahedrally coordinated by five water molecules and the Oδ1 atom of D311 (CD5) (Fig. 1k; Supplementary Fig. 7g). The coordinating waters form an H-bond network involving residues from VD13,12 and VD41,23. Alternative ions (Na+, K+, Ca2+, Zn2+) were evaluated during refinement. Discrimination relied primarily on coordination chemistry, as EM scattering factors and local resolution limit ion-specific density features. Na+, K+, Ca2+ require longer coordination distances that were incompatible with the model geometry. Zn²⁺ typically adopts tetrahedral coordination involving cysteine or histidine residues. The observed octahedral geometry, oxygen-based ligands, metal-ligand distances (~2.0–2.1 Å), and refinement behavior were all consistent with Mg2+ and incompatible with Zn2+. Mg2+ was also refined with atomic displacement parameters comparable to those of coordinating residues. Additionally, the Mg2+ assignment is consistent with an early observation in C. psittaci, showing that adding cations promoted chlamydial attachment, likely by counteracting repulsive forces between negatively charged surfaces of MOMP and host cells58. Two other ligands were modeled in the cap region: an N-acetylglucosamine molecule (NAG), a common component of chlamydial lipooligosaccharide (LOS), and lauric acid (DAO), one of the most abundant fatty acids known to be scavenged by Ct59 (Supplementary Fig. 7h, i). The transmembrane region of the EB MOMP trimer has fewer direct inter-protomer interactions than the extracellular regions (Supplementary Fig. 7j–n), with interactions primarily localized at the bottom of the β-barrel between strand 7 of one protomer and strand 10 of another. We observed elongated densities near the 3-fold axis that likely represent trapped lipids. However, the available information was insufficient to discriminate between isostructural fatty acid species; therefore, to avoid overinterpretation, these densities were left unmodeled. Non-protein densities away from axis could be modeled, taking into account compounds used in purification and the lipid composition of chlamydial and host membranes60. Host lipids are expected to bind because EB MOMP was isolated from Cm-infected HeLa-229 cells, and Chlamydiae are known to hijack host lipids for growth and development61. A sphingomyelin backbone was modeled between two protomers (Supplementary Fig. 7m, n). This molecule forms numerous hydrophobic interactions with the EB MOMP β-barrel, including the side chains of A100 (CD21), I333 (CD51), and A361 (CD52). It spans ~80% of the length of the EB MOMP β-barrel and buries a surface area of 307 Å2 with one protomer and 403 Å2 with another, thus likely playing an important role in stabilizing the trimeric assembly62. We observed additional molecular density near the modeled sphingolipid and on the β-barrel surface, suggesting possible LOS binding. However, local resolution was insufficient to confidently model LOS fragments. We observed no molecular density near any of the 16 asparagine residues in Cm MOMP that would be consistent with N-glycosylation by a large glycan containing 8–9 mannose residues, as previously reported in Ct MOMP63–65. Interactions between the EB MOMP protomers and ligands bury ~19,200 Å2, ~31% of the total surface area (~61,500 Å2), confirming the trimer as the functional unit of EB MOMP. Each protomer has ~40 periplasm-exposed amino acid residues, including residues 14–34 in the flexible loop, β turns in positions 107–109, 179–181, and 340–342, a fragment of a β-turn at N257 (Supplementary Fig. 7o, p). The periplasmic region includes five out of the eight cysteine residues discussed later in detail. Both the periplasmic and extracellular surfaces have a negative electrostatic potential (Fig. 1l–n). On the extracellular side, this negative potential is partially neutralized by Mg2+ (Fig. 1m). The transmembrane surface has a neutral potential with positively charged pockets, indicating putative binding sites for phosphate groups of phospholipids, LOS, or sulfate groups of proteoglycans that aid bacterial adhesion to host cells19,66,67. Inside the β-barrel, the transmembrane region is polar (Fig. 1l). The EB MOMP transmembrane β-barrel protomer displays canonical features of transmembrane β-barrel proteins68,69, including aromatic girdles that mark the membrane boundaries (Supplementary Fig. 6) and an alternating pattern of hydrophobic and hydrophilic residues that produces a polar interior and hydrophobic exterior70 (Fig. 1i, j). However, the trimeric architecture is distinct. Other experimentally determined 10-stranded transmembrane β-barrel structures, classified “OmpT-like”71, are monomeric multifunctional adhesins and include: OmpT, an outer membrane protease from E. coli, Pla, a plasminogen activator from Yersinia pestis, and OpcA, a proteoglycan-binding adhesin from Neisseria meningitidis. In these proteins, the extracellular segments extend above the membrane and confer activity and substrate specificity. EB MOMP similarly extends a significant portion of its β-barrel above the membrane, with the barrel occluded by extracellular loops, as observed for OpcA. A monomer of OpcA (PDB 2VDF72) aligns with an EB MOMP protomer with an RMSD of 3.6 Å over 175 Cα atoms. However, unlike monomeric OmpT-like β-barrels, EB MOMP is a trimer, with three β-barrels topped by a single globular extracellular cap formed by interacting VDs from all three protomers. This cap has its own hydrophobic core, burying ~10,000 Å2 out of the ~22,000 Å2 total cap surface (Supplementary Fig. 8), and presents the VDs as a continuous antigenic surface. Due to this organization, the extracellular region of the trimer extends ~50 Å from the membrane in EB MOMP, compared with ~25–35 Å in monomeric OmpT-like proteins. The formation of the folded cap requires additional sequence content and EB MOMP molecular mass of 42 kDa is significantly higher than the mass of typical OmpT-like proteins (~30–35 kDa). During revision of this manuscript, the structure of trimeric porin I from the Legionella Dot/Icm type IV secretion system (PDB 9NH1, [10.2210/pdb9NH1/pdb])73 was reported. This system uses trimers of 10-stranded β-barrels porins as architectural components. Although the 9NH1 trimer can be superposed onto the EB MOMP trimer with an RMSD of ~6.0 Å, there are significant differences in trimer shapes (Supplementary Fig. 9). Most importantly, 9NH1 lacks the extensive extracellular architecture of EB MOMP. Conserved cysteine residues in transmembrane region of EB MOMP Unlike most membrane β-barrel proteins, which lack cysteine residues74, MOMP contains eight cysteines (C26, C29, C33, C114, C179, C181, C204, and C329, Fig. 2a, b). Seven of these are highly conserved, whereas C204 is not (Fig. 2c, Supplementary Fig. 1, Supplementary Table 2). In the EB MOMP structure, C26, C29, C33, C179, and C181 are located on the periplasmic side; C204 is extracellular; and C114 and C329 reside in the transmembrane region (Fig. 2b–f, Supplementary Figs. 6 and 7p). Location of cysteine residues in the EB MOMP structure. a Positions of cysteine residues mapped on the trimer structure. Cysteine residues are shown as spheres to minimize visual occlusion. b Positions of cysteine residues mapped on the protomer structure. Cysteine side chains are shown as sticks and colored according to putative inter-trimer disulfide bond interactions discussed in this manuscript, with orange representing intramembrane interactions, and gray representing periplasmic interactions. c–f Molecular density maps showing the positions of cysteine residues and the absence of disulfide bond formation. Density maps for C26, C29, and C33 are shown in Supplementary Fig. 7p. Because MOMP is covalently incorporated into the disulfide-crosslinked COMC, reduction of intermolecular disulfide bonds is required to isolate MOMP trimers and to purify them from other COMC components. Reducing conditions must be maintained throughout isolation and purification, not only to purify MOMP but also to prevent reoxidation, aggregation, and loss of solubility. Although purification under reducing conditions removed disulfide bonds, our structures provide a framework for interpreting earlier biochemical observations of MOMP disulfide bonding, because disruption of a disulfide bond leaves the participating cysteines in spatial proximity, such that modest conformational rearrangement would permit re-formation of the bond under oxidizing conditions. Extracellular C204 is well ordered but too distant from any other cysteine, including C204 from neighboring protomers, to form disulfide bridges (Fig. 2a–c, Supplementary Table 3). It is also unlikely to participate in disulfide bonding with other MOMP trimers or COMC proteins. In contrast, under non-reducing conditions, the periplasmic cysteines are expected to engage predominantly in intermolecular disulfide bridges with other MOMP trimers and with additional COMC components19,26,27,29,75. Previously reported disulfide bonds involving C26-C33 and C26-C29, detected by mass spectrometry76,77, are not observed in our structure due to purification under reducing conditions (Supplementary Table 3). However, the periplasmic region (D14-D34) has sufficient conformational flexibility to support formation of disulfide bonds involving these residues within a MOMP monomer, between MOMP trimers, and with other COMC components. The formation of an intramolecular disulfide bond between C179 and C181 is unlikely, given their placement in a tight β-turn (Fig. 2f). Cysteine residues have not been previously observed in the transmembrane regions of experimentally determined β-barrel structures78, possibly due to their destabilizing effect in these segments79. However, C114 and C329 are exposed to the membrane and are evolutionarily conserved (Fig. 2a, b, d, e, Supplementary Fig. 1, Supplementary Table 2), suggesting an important role in MOMP function. Structure predictions for the chlamydial polymorphic outer membrane protein (Pmp) family similarly show outward-facing cysteines in their transmembrane β-barrels80 (Supplementary Figs. 10, 11, Supplementary Table 4). Like EB MOMP, several Pmps are components of COMC. Although cysteine positions are less conserved among Pmps than in EB MOMP, outward-facing cysteines occur in both β-barrel and periplasmic loops, and can form intermolecular disulfide bonds81,82. Our structure and Pmp models raise the possibility that COMC stability is mediated not only by periplasmic but also by intramembrane disulfide bonds. For example, conserved transmembrane cysteines in Cm Pmp A-C (C970 in PmpA, C1666 in PmpB and C1454 in PmpC) are at the same level in the membrane as conserved EB MOMP C114, making them candidates for transmembrane disulfide bridges. Earlier theoretical models placed MOMP cysteines in the extracellular space, complicating reconciliation with the known periplasmic disulfide network of the COMC23,28. The presented structure resolves this inconsistency and supports a model in which reversible, redox-regulated disulfide bonds in both periplasmic and transmembrane regions contribute to COMC assembly and remodeling during developmental transitions. The EB MOMP structure also provides a structural reference for interpreting EB outer membrane organization. The inner leaflet of the OM in EB is associated with a highly ordered, pseudo-crystalline hexagonal array28,64, which is thought to be composed primarily of MOMP, small cysteine-rich outer membrane protein OmcA, and the large cysteine-rich periplasmic protein OmcB. AlphaFold3 heteromultimer predictions of the MOMP trimer in complex with OmcA and OmcB consistently place OmcA and OmcB on the periplasmic side of MOMP and suggest the multiple intermolecular disulfide bonds between these components (Supplementary Figs. 12–14, Supplementary Data 1). OmcB is predicted to form elongated, star-shaped trimers, consistent with its proposed role as a crosslinked periplasmic protein scaffold that compensates for the absence of a peptidoglycan layer28. Its predicted immunoglobulin-like domains and remote homology to S-layer proteins further support this scaffolding function. OmcA is predicted to interact at the MOMP-OmcB interface and may tether the periplasmic protein network to the OM via its N-terminal lipid anchor28. In this model, MOMP and OmcB form the pseudo-hexagonal lattice, while OmcA occupies gaps and stabilizes the assembly through transmembrane interactions. Intra- and intermolecular disulfide bonds between MOMP and OmcB are proposed to provide periplasmic support for the array. Owing to the extended and flexible domain architecture of OmcB, precise lattice parameters cannot yet be defined. The EB MOMP structures reported here provide high-resolution templates for future cryo-electron tomography with subtomogram averaging studies, which will test this model and enable quantitative analysis of COMC and EB architecture. mAb-18b Fab binds a conformational epitope on the EB MOMP trimer mAb-18b is a mouse monoclonal neutralizing antibody (IgG2b/κ) raised against Cm EBs (Supplementary Table 5). It recognizes a conformational epitope of trimeric Cm EB MOMP but not Ct EB MOMP37. The complex of EB MOMP and mAb-18b Fab (MOMP/mAb-18b) was reconstructed at a global resolution of 3.36 Å (Fig. 3a, Supplementary Fig. 15a–g, Supplementary Table 1). The local resolution ranges from 3 to 3.5 Å for EB MOMP and 3.5 to 4 Å for the Fab variable domains. The Fab constant domains have lower resolution (>5 Å) due to flexibility at the elbow region (Fig. 3a–c, Supplementary Fig. 15e). Despite this, the reconstructed map was of sufficient quality to enable unambiguous discrimination between the variable heavy-chain (VH) and variable light-chain (VL) domains, despite their pseudo-symmetry, and to allow refinement of the atomic model against the reconstructed density. Reconstruction in C1 symmetry using the same set of particles was also performed to assess asymmetrical features, but no significant deviations from C3 symmetry were observed (Supplementary Fig. 16). Cryo-EM structure of the Cm EB MOMP/mAb-18b complex. a Molecular density map of the MOMP/mAb-18b complex (contour level 0.15) colored by chain (MOMP: green, cyan, magenta; mAb-18b Fab light chain: violet, light pink, tan; heavy chain: light blue, purple, blue). b Atomic model of the MOMP/mAb-18b complex, colored by chain (MOMP: green, cyan, magenta; mAb-18b Fab light chain: violet, light pink, tan; heavy chain: light blue, purple, blue). c Extracellular view of the MOMP/mAb-18b complex. For clarity, only one unit of the MOMP/mAb-18b complex is shown in color (MOMP: green; mAb-18b light chain: violet; mAb-18b heavy chain: light blue). d Schematic of MOMP/mAb-18b extracellular view showing one unit of mAb-18b Fab interacting with two MOMP protomers. e, f Structural changes in MOMP upon mAb-18b binding. For clarity, the main chain RMSD is mapped onto one MOMP protomer in tube representation, while the other two protomers are displayed as a gray ribbon. Significant changes are observed at VD1 (dashed box 1) and VD4 (dashed box 2). g Superposition of Fab-bound MOMP (green, cyan, and magenta) and unbound MOMP (gray). One protomer is highlighted in tube representation, with the other two protomers shown as a ribbon diagram. Dashed boxes 1 and 2 indicate the regions with the most substantial changes, as in (e). h Close-up view of VD1 in MOMP (gray) and in MOMP/mAb-18b complex (green), showing the largest change at residue Asp71, with a Cα distance of ~7.7 Å. i Close-up view of VD4 in MOMP (gray) and in MOMP/mAb-18b complex (green) showing the largest shift at residue T307, with a Cα distance of ~11.7 Å. Although the stoichiometry of the complex is 1:1, with three EB MOMP molecules binding three mAb-18b Fab molecules, each Fab interacts with the extracellular VDs of two EB MOMP molecules, inducing significant local structural changes (Fig. 3c–i, Supplementary Fig. 17a–d). The RMSD calculated between EB MOMP and the MOMP/mAb-18b structures is 2.0 Å. While the β-barrel segment of the trimer is unchanged upon Fab binding, VD1 (E61 to E82) undergoes dramatic rearrangements, with the largest main chain shift of ~8 Å at D71 (Fig. 3e–h, Supplementary Fig. 17b). VD4 (L284 to T310) is also rearranged, with large main chain shifts in the segment from S301 to I309, with T307 requiring an ~12 Å shift for optimal superposition (Fig. 3e–g, i, Supplementary Fig. 17b–d). The molecular density map for VD1 residues is well resolved (Supplementary Fig. 17e), while the density for two out of 27 VD4 residues in the most rearranged region suggests disorder (Supplementary Fig. 17d, f, g). The most probable conformation was selected for these two residues based on molecular density and the pattern of chemical interactions. The conformational changes in EB MOMP upon mAb-18b binding also alter the local environment that originally coordinated the Mg2+ cation (Fig. 3f, Supplementary Fig. 17g). The Oδ1 of D311 (CD51) rotates away from its coordinating position, while the side chain of I309 (VD41) moves into the space previously occupied by two water molecules coordinating Mg2+. Although alternative modes of Mg2+ binding cannot be entirely ruled out, the dominant conformation is incompatible with Mg2+ binding in its original position. In Ct, antibody neutralization efficiency decreases with increased Mg2+ concentration83. The structures presented here are consistent with this observation, as additional energetic costs would be incurred to induce conformational change of EB MOMP VDs to eject Mg2+ during antibody binding and neutralization. The interaction interface between the mAb-18b Fab and EB MOMP buries of ~1,175 Å2 surface area (Fig. 4a, b) and has a shape complementarity (Sc) of 0.6284, consistent with the expectations for an antibody-protein complex. Each Fab interacts with two MOMP protomers, engaging residues in the complementary-determining regions (CDRs) of both the light (L) and heavy (H) chains (Fig. 4a–j, Supplementary Table 5). The Fab primarily interacts with one protomer, which contributes ~90% of the buried surface area, while the second protomer contributes ~10% (Figs. 3c, d, 4c, d). CDR1 (L1) and CDR3 (L3) of the L chain interact with VD11 and VD21 (Fig. 4c–f) through four H-bonds. Although L2 does not directly contact MOMP, R60 forms a π−π interaction with Y104 (H3), stabilizing the Y104 side chain in an orientation that supports contact with the Cβ of A145 (VD21). Additionally, interactions with L1 reduce the solvent-accessible area of VD31. All three H-chain CDRs interact with EB MOMP (Fig. 4c, d, h–j). H1 interacts with VD21 and VD42, H2 interacts with VD11, VD21, VD41 and VD42, while H3 interacts with VD11 and VD21. These interactions involve ten H-bonds with the primary protomer and one H-bond with the secondary protomer. Binding is further stabilized by hydrophobic and van der Waals interactions across all interfaces. The binding affinity between MOMP and mAb-18b was estimated to be 7.6 pM (ΔG = −13.8 kcal·mol−1) based on the refined structure using PRODIGY webserver85. Neutralization with mAb-18b happens most likely through steric hindrance, with bound antibody shielding EB MOMP surfaces from binding to host cells. Cm EB MOMP and mAb-18b Fab binding interface. a MOMP residues involved in mAb-18b Fab binding are shown as buried surface area mapped onto one MOMP protomer. The other two MOMP protomers are displayed as ribbon diagrams. A surface representation of VDs in the same view is shown in the dashed box as a reference. b mAb-18b Fab residues involved in MOMP binding are shown as buried surface area mapped onto one Fab molecule (one light chain and one heavy chain). The other two Fab molecules are displayed as ribbon diagrams. Light chain residues (H36, S37, and A38) are marked with a black dashed line, while heavy chain residues are marked with a gray solid line. c Overall binding interface between MOMP and mAb-18b. Two MOMP protomers (green and cyan) and one mAb-18b Fab molecule (light chain: violet; heavy chain: light blue) are involved in binding. VDs of MOMP and CDRs of mAb-18b Fab are labeled (light chain CDRs: L1, L2, L3; heavy chain CDRs: H1, H2, H3) (Supplementary Table 5). d Overall binding interface between MOMP and mAb-18b. Epitopes and paratopes are shown as spheres. Red spheres represent MOMP residues in direct contact with mAb-18b (epitopes), while blue spheres are mAb-18b residues in direct contact with MOMP (paratopes). e–j Detailed view of interactions between MOMP and mAb-18b through VDs and CDRs (MOMP: green, cyan; Fab light chain: violet; Fab heavy chain: light blue). Main chain atoms are omitted for clarity. The recognition mode and conformational changes induced by antibody binding in the EB MOMP/mAb-18b complex will enable computational simulations to explore the dynamics of antigenic cap structural changes for Ct serovars. EB MOMP cannot function as a porin Some studies have shown that MOMP functions as a porin, diffusing sugars and ATP26,36,37, with an estimated pore size of ~1–2 nm, based on liposome swelling and planar bilayer assays26,37. However, in EBs, MOMP is not expected to function as a porin, as passive transport would likely compromise the stability of this largely metabolically inert developmental form26,37. Consistent with this expectation, the EB MOMP structures determined here are incompatible with the passive transport of the molecular species reported previously. The EB MOMP/mAb-18b structure is a complex with the Fab fragment of mAb-18b, an antibody raised against intact Cm EBs, and its formation indicates that the isolated protein retains its native EB conformation without major rearrangements that could alter pore size. The EB MOMP protomer is a 10-stranded, oval-shaped β-barrel, with dimensions of ~23 Å × ~19 Å (Fig. 5a). The first 12 N-terminal residues form a partially helical stopper segment that inserts into the barrel lumen from the periplasmic side, burying ~2100 Å2 of surface area against the β-barrel interior (Fig. 5b, c). This segment is stabilized by 15 H-bonds between the stopper and polar residues lining the barrel lumen, as well as by internal H-bonds within the stopper itself. Although the stopper segment does not fully occlude the barrel interior, Caver86 analysis revealed bottlenecks that prevent diffusion (Fig. 5d, e, Supplementary Fig. 18). No tunnels traverse the tightly folded extracellular cap. When the extracellular domain and the stopper segment are computationally removed, the remaining transmembrane β-barrel contains a tunnel with a minimal diameter of ~4.3 Å, which is sufficient for water passage (2.75 Å diameter) but incompatible with diffusion of sugars or ATP observed previously26,36,37. To accommodate the transfer of both water and larger solutes, EB MOMP would have to adopt a state that is topologically distinct from the current structure. Features of the EB MOMP β-barrel interior. a External dimensions of the β-barrel. b Arrangement of the N-terminal stopper segment (cyan) inside the barrel. c Interactions between the residues of the N-terminal stopper segment (cyan) and the side chains of residues lining the barrel interior (green) with selected hydrogen bonds labeled. d Tunnels inside of the MOMP protomer, calculated using Caver with a probe radius of 0.9 Å. Each tunnel is represented by a different color. e Radii of the β-barrel accessible volume mapped along each tunnel. The source data for this figure panel are provided in the Source Data file. The stopper likely serves a dual function: it occludes the barrel to prevent passive diffusion and stabilizes the β-barrel during folding, so that the folding process and transient conformational rearrangements during adhesion do not generate non-productive or permeable intermediates. Support for this folding-stabilization role comes from the presence of water molecules inside the barrel. Despite the modest global resolution of the EB MOMP reconstruction (2.96 Å), several ordered water molecules are resolved. One water molecule is located within the β-barrel lumen and is coordinated by the side chains of Q189 and Q249 and the backbone carbonyl of P2 from the stopper region (Fig. 5c). Because the β-barrel is impermeable, this water must have been trapped during folding. Current models of β-barrel biogenesis involve progressive β-sheet formation followed by barrel closure, so stabilization of the barrel interior by the stopper segment would suppress non-productive folding intermediates and favor correct assembly. If MOMP functions as a porin during other stages of the chlamydia developmental cycles, the necessary structural rearrangements likely occur during differentiation from EB to RB, coinciding with the reduction of disulfide bridges in COMC proteins and the increase in cell volume between EB and RB87. In RB MOMP, all cysteine residues are reduced23,26,28. However, the reduced form of EB MOMP structure reported here remains incompatible with passive transport, consistent with prior observations that reduction of disulfide bonds alone is insufficient to generate a porin-permissive MOMP conformation88. It remains to be determined whether a different structure of chlamydial MOMP exists, whether earlier experiments detected a contaminating porin, or whether MOMP misfolding in artificial membrane systems contributed to the reported transport activity. VDs, adhesion model, and subunit vaccine development Experiments using Cm in a mouse model have shown that the three-dimensional structure of EB MOMP is critical for inducing robust humoral and cellular immune responses37,39,41,47,48. EB MOMP is the dominant surface antigen of chlamydial species and, in Ct, is also the primary determinant of serovar specificity encoded in its surface-exposed VDs. Ct serovars A-K infect the epithelia, whereas LGV serovars invade deeper into the reticuloendothelial system. This difference in tissue tropism suggests that adhesion mechanisms vary among chlamydial serovars and species89. Although adhesion is modulated by both chlamydial and host factors and remains incompletely defined2,90,91, the correlation between serovar-specific tissue tropism and serospecificity encoded in MOMP VDs supports a central role for MOMP in host-cell adhesion. Our structures are consistent with an emerging multi-gate model of EB adhesion, involving at least two steps: an initial reversible attachment mediated by electrostatic interactions between EBs and glycosaminoglycans (GAGs), followed by irreversible binding to as-yet-unidentified species- or serovar-specific receptors and subsequent internalization92,93. Early studies proposed heparan sulfate-like GAGs as adhesion receptors, but results were inconsistent94–99. A loss-of-function screen in human haploid cells for Ct serovar L2 showed that sulfation levels, rather than the amount of heparan sulfate, are critical for EB attachment100. Proteolysis-based assays provided early functional evidence linking MOMP to adhesion. Trypsin treatment of purified EBs reduced attachment and infectivity in a serovar-dependent manner, and two-dimensional electrophoresis after proteolysis identified MOMP-derived peptides mapping to VD regions, implicating VD2 and VD4 in serovar B and VD4 in serovar L101,102. Although relatively crude by modern standards, these experiments directly linked surface accessibility of MOMP regions to infectivity and support a role for MOMP in the initial adhesion step. In the Cm EB MOMP structure, VD1, VD2, and VD4 are exposed on the extracellular antigenic head, with VD2 and VD4 forming the binding interface for the Fab of the neutralizing antibody mAb-18b. The trimeric antigenic head is negatively charged but contains positively charged patches and grooves, including a binding site occupied by N-acetylglucosamine (Figs. 1l, m and 6a, Supplementary Fig. 3h, i, and Supplementary Fig. 19), which could mediate interactions with sulfated proteoglycans on host cells95. Together with its high abundance on the EB surface, these features make MOMP a compelling candidate for the adhesin that mediates the reversible attachment step, although other proteins also contribute21,103. Additional factors, such as glycosylation or interactions with positively charged cations, may further modulate the reversible attachment. In our Cm EB MOMP structure, none of the asparagine residues is modified by high-mannose N-glycosylation. However, N-glycosylation has been reported for MOMP from some Ct serovars and in C. pneumoniae38,63,65,91, and shown to modulate, but not abolish, infectivity, indicating a modulatory rather than essential role. Similarly, we observe a bound Mg2+ cation in the Cm EB MOMP structure, but AlphaFold models of MOMP from other species suggest that the corresponding coordinating cavity may not be conserved (Supplementary Figs. 20, 21, Supplementary Data 2, 3). Structural analysis of species-specific conserved epitopes in VD4: Cm TTWNPTISG and Ct TTLNPTIAG. a CTH522 extVD4 mapped onto the Cm MOMP (± Fab) structures and its involvement in the N-acetylglucosamine (NAG) and lauric acid (DAO) binding sites. Gray arrows indicate residues marking the beginning (T292) and the end (G300) of the Cm TTWNPTISG motif. b CTH522 extVD4 mapped on the MOMP monomer. Gray arrows indicate residues marking the beginning (N257) and the end (N321) of the flanking region. c CTH522 extVD4 mapped onto the MOMP antigenic cap. MOMP structures with and without Fab are superimposed to highlight Fab-induced changes in the CTH522 extVD4 region. Residues indicating the beginning (T287) and the end (T310) of the large conformational changes induced by Fab are highlighted by dashed arrows. d Evolutionary conservation of VDs across all Chlamydia species, as determined by ConSurf154 mapped onto the trimer. e Conservation patterns on the antigenic cap, which mediates host cell adhesion. f Locations of the VDs on the antigenic cap in relation to the conservation pattern. The transition from reversible GAG- and possibly MOMP-mediated docking to irreversible commitment to internalization can be blocked by lowering the temperature to 4 °C in some species92. One mechanism supported by our structures is that this step involves conformational rearrangements at both the EB surface and the host membrane to enable high-avidity receptor engagement. A species-specific constant region in VD4 (TTLNPTIAG in Ct, corresponding to 292-TTWNPTISG-300 in Cm) has been proposed as the host cell binding site101 (Supplementary Fig. 1, Supplementary Table 6). This motif is conserved across Ct serovars50, induces neutralizing antibodies83,104, and is included in the MOMP-based vaccine candidate CTH522, which has completed two Phase I clinical trials46,105. The motif alone is not immunogenic and requires the flanking regions and repeats to elicit an immune response106,107. In both EB MOMP structures, this nonapeptide is not solvent-accessible but instead participates in trimerization and contributes to ligand-binding sites occupied by N-acetylglucosamine and lauric acid (Fig. 6a, Supplementary Fig. 19). In the EB MOMP structure, residues from this motif also coordinate water molecules around an Mg2+ ion (Fig. 6a, Supplementary Fig. 19b). Binding of mAb-18b induces a local conformational change in VD4 immediately downstream of the motif (Figs. 3e–g, 6a–c, Supplementary Fig. 19b). We therefore propose that, during EB adhesion, engagement of an appropriate host receptor induces a larger conformational rearrangement of the antigenic cap, exposing this otherwise buried motif so it can bind the receptor. Lowering the temperature would be expected to block such rearrangements. Although past antibody accessibility experiments suggest that this conserved motif can be targeted in some serovars, AlphaFold models indicate that it is generally solvent-inaccessible across species and serovars (Supplementary Fig. 20d, h). Antibody binding likely reflects stabilization of a transient “open” conformation rather than constitutive exposure. Differences in the energetic landscape of this transition, driven by local cap stability, may underlie species- and serovar-specific differences in adhesion, tissue tropism, and immune recognition. Shielding the species-specific nonapeptide by the antigenic cap likely provides an evolutionary advantage: mutants defective in adhesion are eliminated, while immune responses are diverted toward exposed regions of the VDs. The cap is formed by three copies of all VDs folded into a single domain, with the most divergent sequences exposed on the surface and at the ligand-binding sites (Fig. 6d). In this arrangement, even a single amino acid substitution is amplified by the trimeric organization of MOMP, leading to coordinated changes in three-dimensional structure, surface charge and complementarity, and conformational dynamics, which together influence adhesion specificity and immune recognition. Such functional impact of minor sequence variations has been observed for Ct A serotype, where infections with two strains differing by only two residues in the VDs produced distinct neutralizing antibody titers108. The role of MOMP as a reservoir of antigenic variability is also supported by the observation that the nonsynonymous mutation rate in cap-forming regions is ~3 times higher than in the rest of the protein109. AlphaFold models further show that minor sequence changes in VDs can dramatically alter surface charge and complementarity (Supplementary Fig. 20). mAb-18b binds to the EB MOMP VDs, where predicted and observed B-cell epitopes are located. However, resolving intracellular infections requires T-cell-mediated immunity, which is critical for protection against Ct-induced upper genital tract pathology42,110. To assess the relevance of our structure, we mapped data from the Immune Epitope Database (IEDB)111 for Ct serovar D, which has the most published epitope data, onto a Cm MOMP-based homology model of Ct serovar D MOMP. Mapping results show that B-cell epitopes are more frequent in the extracellular antigenic head, while T-cell epitopes are more prevalent in the conserved parts of EB MOMP, with VD3 harboring both B-cell and T-cell epitopes (Supplementary Fig. 22). Reconciling the serovar/serogroup specificity of Ct MOMP with the location of T-cell epitopes in the conserved β-strands presents an immunological paradox that requires further studies. The lack of information on chlamydial non-linear B- and T-cell protective epitopes limits our ability to generate synthetic antigens. The identification of the antigenic structure recognized by mAb-18b is the first step towards solving this problem. Discussion The EB MOMP structures show a distinct trimeric assembly of 10-stranded β-barrels, in which extracellular loops from all three protomers fold into a compact, magnesium-binding antigenic cap containing the immunogenic VDs. The EB MOMP structural distinctness lies in the organization of the extracellular region: cap formation results in a “3 × β-barrels + 1 × cap” domain architecture rather than the “3 + 0” or “3 + 3n” architectures observed in other trimeric β-barrel assemblies. No experimentally determined or predicted structures identified in our searches for structurally similar models exhibit this feature. Searches performed using the extracellular cap alone, with the transmembrane β-barrels removed, likewise failed to identify structurally related proteins. Therefore, such a trimeric assembly can be considered the founding member of a distinct family of trimeric adhesins. The cap binds Mg2+ and contains several positively charged cavities, some occupied by ligands, indicating putative interaction sites for chlamydial and host molecules. A sphingolipid molecule intercalated between two EB MOMP protomers suggests that sphingolipid-mediated stabilization may contribute to Chlamydiae’s reliance on host sphingolipids. Although bacterial and mammalian sphingolipids differ structurally112, Ct cannot synthesize sphingomyelin, cholesterol, and phosphatidylcholine113 and therefore likely depends on host sphingomyelin to generate Chlamydia-specific sphingolipids. The EB MOMP structures differ significantly from earlier theoretical models proposing 16-stranded68,77,114,115 or 14-standed69 β-barrel topologies, indicating the need for a critical reassessment of experimental data on MOMP accumulated over the past four decades and interpreted in the context of those models. The EB MOMP structures are incompatible with the porin function proposed previously. The tightly folded antigenic cap and the N-terminal stopper segment bound in the lumen of each β-barrel preclude passive transport of solutes, including water. Even if the cap undergoes rearrangement during adhesion, the stopper region interacts extensively with the barrel lumen, preventing the transport. These interactions, together with a water molecule trapped inside the β-barrel, suggest that the stopper is inserted during folding. Removal of the stopper to allow passive transport would likely require de novo folding of MOMP. This implies either that MOMP adopts an alternative β-barrel topology in RBs through de novo folding or that other proteins serve as porins. A similar challenge of a β-barrel too narrow to support passive transfer, despite porin activity observed in planar lipid bilayers and liposome swelling assays26,36,37, has been described for other systems. For instance, Pseudomonas aeruginosa OprF116 forms a closed 8-stranded β-barrel that supports the OM, while a minor alternative conformation with a wider pore (~2 nm) has been proposed to explain transport of large substrates in liposome swelling assays. Models of this transition invoke refolding influenced by redox potential and multiple folding intermediates. Redox changes are central to EB ↔ RB developmental transitions. To enable cryo-EM SPA, EB MOMP was isolated under reducing conditions to dissociate it from the disulfide-crosslinked COMC and obtain a homogeneous, soluble sample. These conditions have not altered the overall fold or quaternary organization of EB MOMP, as demonstrated by the structure of EB MOMP in complex with the Fab fragment of the neutralizing antibody mAb-18b raised against Cm EBs, which is also incompatible with passive solute transport through the barrel. Fab binding indicates that purified MOMP retains its native EB conformation, supporting the conclusion that EB MOMP is not a porin. Although disulfide bonds could not be observed in samples purified under reducing conditions, cysteine placement and Sγ–Sγ distance analysis indicate that MOMP is unlikely to form intra-monomeric or intra-trimeric disulfides and instead participates primarily in intermolecular disulfide linkages between MOMP trimers and with other cysteine-rich COMC components. The presence of conserved cysteines in transmembrane regions, together with similar cysteine placement in predicted models of Pmp proteins80, raises the possibility that redox regulation extends to membrane-embedded cysteines, providing an additional control layer during the developmental cycle. This idea is consistent with prior research showing that EBs contain a highly disulfide-crosslinked COMC, whereas RBs are more reduced and gradually re-oxidize during RB → EB maturation. Perturbation of redox status alters developmental progression, and chlamydial redox regulation is independent of host cell redox state. Chemical reduction of EB disulfides does not recapitulate full EB → RB differentiation, consistent with reduction being an early enabling step rather than a complete trigger. Although MOMP remains the most abundant protein across the chlamydial developmental cycle, other COMC proteins undergo proteolytic processing and are not resynthesized until the RB → EB transition. These observations support a model in which controlled COMC reduction relaxes EB envelope rigidity while preserving stable MOMP trimers or even monomers so that they can later be reincorporated into the COMC. The EB → RB transition is therefore not equivalent to COMC dispersal or to a “COMC-embedded MOMP→free MOMP” transition; rather, COMC relaxation enables subsequent processes that commit the cell to the RB program. One critical constraint is nucleoid structure: EB → RB conversion requires decondensation of the tightly packed nucleoid to permit transcriptional reactivation and replication, which in turn requires increased metabolism and cell volume and therefore early COMC relaxation. In addition to its structural role in the COMC, EB MOMP exposes VDs in the antigenic cap that, in Ct also determine serospecificity. The cap is strongly electronegative, with positively charged patches on its sides, consistent with selective engagement of negatively charged GAGs at defined binding sites rather than through nonspecific association. Antibody-induced conformational changes in the mAb-18b complex suggest how the cap can undergo structural rearrangements upon encountering an appropriate trigger, potentially exposing hydrophobic surfaces that support high-avidity interactions with specific host receptors. The amplification of modest sequence differences by trimerization, together with the cap’s role as a unified antigen-presenting surface, provides a structural basis for serospecificity and antigenic variation (Supplementary Fig. 20). Together, these structures reconcile multiple experimental observations and support a multifunctional role for EB MOMP as both (i) a trimeric building block that is crosslinked within the COMC and (ii) an antigenic cap that contributes to early, and potentially later, steps of adhesion. Additional functions are likely given MOMP abundance throughout the developmental cycle. These EB MOMP structures also provide a foundation for structure-based vaccine development against Ct. They enable the design of constructs with improved expression, solubility, and stability, facilitating the production of recombinant MOMP in its trimeric form. Deep learning tools can further optimize sequences by eliminating non-essential cysteines and modifying trimer features to enhance cross-reactivity117,118. The immunogenic extracellular cap can be grafted onto alternative scaffolds to improve vaccine efficacy, with immune refocusing targeting conserved, surface-exposed epitopes while excluding non-protective ones. Previous attempts to graft MOMP VDs onto the Neisseria PorB (an OmpF-like porin unrelated to chlamydial PorB) achieved limited success119–121. Although PorB is trimeric, its β-barrels are too wide to support formation of the antigenic cap, indicating once again that unfolded VDs that do not assemble into the native antigenic cap are insufficient to reproduce the immune response to the native protein, even when presented on the trimeric scaffold. These chimeric antigens may present linear epitopes but cannot recapitulate the conformational epitope structure of trimeric EB MOMP. The limited responses elicited by these constructs, together with our structures, underscore the importance of preserving protective conformational B-cell epitopes. Weak immunogenicity may also reflect the absence of Chlamydia-specific T-cell epitopes, which are enriched in more conserved regions (Supplementary Fig. 22). The structural analysis presented here will guide vaccine design that preserves the relevant tertiary and quaternary epitope architectures. Methods Isolation and purification of native Cm EB MOMP The trimeric native Cm EB MOMP was extracted and purified as previously described37. Briefly, Cm-infected Hela-229 cell (ATCC; Manassas, VA; cat # CCL 2.1) monolayers were collected, centrifuged at 10,000 × g for 30 min at 4 °C (Beckman Coulter, Inc. J2-21 centrifuge; Brea, CA) and washed with PBS (pH 7.4). The pellet was resuspended in 0.02 M Tris (pH 7.4), 1.0 M NaCl, 0.012 M MgCl2, and 1 mM phenylmethylsulfonyl fluoride (PMSF) (Calbiochem, La Jolla, CA) by sonication using a cup horn with iced water (Virsonic 300 Ultrasonic cell disrupter; The VirTis Company, Inc; Gardiner, N.Y.). After adding 25 µg of DNase (Sigma Aldrich, St. Louis, MO), the suspension was incubated on ice for 2 h with constant mixing. Following centrifugation, the pellet was resuspended by sonication in 0.2 M sodium phosphate buffer (pH 5.5), containing 1 mM each of EDTA and PMSF, 100 mM DTT (Roche Applied Sciences; Indianapolis, IN), and 2% CHAPS (Anatrace; Maumee, OH), and rotated at 37 °C for 2 h. After ultracentrifugation at 100,000 × g for 1 h at 4 °C (Beckman Coulter Inc. L2 ultracentrifuge), the pellet was resuspended by sonication and extracted again with CHAPS-containing buffer for 1 h at 37 °C. After ultracentrifugation, the pellet was resuspended by sonication and extracted with 2% Anzergent Z3-14 (Anatrace; Maumee, OH) in the same phosphate buffer. Following incubation at 37 °C for 2 h, the sample was centrifuged at 100,000 × g for 1 h at 4 °C, and MOMP was recovered in the supernatant. To purify the MOMP trimer, a 1 × 20 cm hydroxyapatite (Bio-Gel ®HTP- hydroxyapatite; Bio-Rad Laboratories Inc., CA) column was used31. The column was equilibrated with 0.02 M phosphate buffer (pH 5.5) containing 0.1% Z3-14 and 1 mM each of EDTA, PMSF, and DTT. Elution was performed using a 0.02 M to 0.5 M linear gradient of sodium phosphate buffer containing 0.1% Z3-14 and 1 mM each of EDTA, PMSF, and DTT. The peak fraction was collected, concentrated by ultrafiltration (Amicon ®Ultra 15; Merck Millipore, Ltd.; Darmstadt, Germany), and the purity of the MOMP trimer was confirmed by SDS-PAGE stained with Coomassie blue. The concentration of endotoxins as assessed by the Limulus amoebocyte assay (Associates of Cape Cod, Inc. MA) was less than 0.05 EU/mg of protein. Production of Cm EB MOMP monoclonal antibodies and Fab fragments purification The monoclonal antibody mAb-18b was previously generated and characterized by enzyme-linked immunosorbent assay (ELISA) as binding MOMP and recognizing Cm EBs76. Subsequent studies demonstrated that mAb-18b is a conformational antibody that specifically recognizes the trimer of Cm MOMP37,44. The antibody was obtained by inoculating intranasally 4- to 6-week-old female BALB/c-H2d mice with 104 inclusion-forming units (IFU) of Cm as previously described37,83,122. Animals were housed in isolation cubicles at a constant temperature of 24 °C with a cycle of 12-h light and 12-h darkness and fed chow and water ad libitum. Six weeks later, the mice were injected intraperitoneally with 107 IFU of Cm. Spleen cells from these infected mice were fused with the mouse myeloma cell line S194/5XX0.Bu.1 (ATCC cat # TIB 20), using the method described by Kohler and Milstein123. Culture fluid from the hybridomas was screened for the presence of antibodies to Cm EB using a microimmunofluorescence assay. Positive hybridomas were selected and cloned by limiting dilution. To produce ascites fluid, 106 cloned hybridomas cells were injected intraperitoneally into 6- to 8-week-old female BALB/c mice that had been pre-treated with Pristane (Sigma Chemical Co.; St Louis, MO) two weeks prior. Ascites fluid was collected, centrifuged at 10,000 × g for 30 min at 4 °C, and frozen at −70 °C. To map the binding site of the monoclonal antibodies, 25-mer MOMP peptides were chemically synthesized (SynBioSci Corp.; Livermore, CA), and an ELISA was performed (Fig. 2 in Yen et al.76). Monoclonal antibodies that did not recognize linear peptides were then tested against both heated and non-heated Cm EBs using a fluorescently labeled anti-mouse antibody. This test aimed to identify antibodies that recognize discontinuous epitopes on EB MOMP trimers present in the non-heated samples. Fab fragments were generated using Immobilized Papain Agarose Resin (cross-linked 6% agarose; Pierce Biotechnology, Rockford, IL) following the manufacturer’s protocol. Briefly, antibodies purified using Affi-Gel® Protein A MAPS® II Kit (Bio-Rad Laboratories; Hercules, CA) were dialyzed against sample buffer (20 mM sodium phosphate, 10 mM EDTA, pH 7.0) and concentrated to 5 mg/ml in a total volume of 2 ml. Immediately prior to digestion, L-Cysteine-HCl was added to the sample to a final concentration of 20 mM, and the pH was adjusted to 7.0. The Immobilized Papain Agarose Resin (0.5 ml) equilibrated with digestion buffer (20 mM sodium phosphate, EDTA 10 mM, L-Cysteine HCl 20 mM, pH 7.0) was mixed with the antibody’s solution at a 1:2 v/v ratio. Digestion proceeded for about 12 h in a high-speed shaking water bath at 37 °C. After digestion, the papain agarose was separated by centrifugation and washed with 10 mM Tris HCl pH 7.5. Fab fragments were then separated from undigested IgG and Fc fragments using the Affi-Gel® Protein A MAPS® II Kit, according to the manufacturer's instructions. Cryo-EM grids preparation and data collection Purified samples were applied to Quantifoil R 1.2/1.3 300 mesh gold grids that had been glow-discharged for 90 s at 30 mA using a PELCO easiglow™ Glow Discharge Cleaning System to render their surface hydrophilic. The glow-discharged grids were then used to prepare vitrified samples using the Thermo Scientific Vitrobot Mark IV System. A 3 μL aliquot of sample was applied to the grid at 4 °C and 100% humidity, followed by blotting for 5.0 to 5.5 s with a blot force of either 18 or 19. MOMP alone was used at a concentration of 1 mg/ml. For the MOMP-Fab complex, 15 µL of purified MOMP (1 mg/mL) was mixed with 15 µL of purified Fab (1.4 mg/mL) and incubated at 4 °C for 72 h prior to grid preparation. The data were acquired with a 300 kV Titan Krios G2 microscope (Thermo Fisher) equipped with a K3 Summit direct electron camera (Gatan) operating in super-resolution mode at a nominal magnification of 81,000×, with a physical pixel size of 1.079 Å. Neither a phase plate nor the objective aperture was used. SerialEM (version 4.0) was used for automated data collection in beam-image shift mode, with nine images acquired per stage movement with a defocus range from −1.0 to −3.0 μm and compensation for aberrations induced by beam-image shift124. The slit width of the GIF Quantum Energy Filter was set to 25 eV. Movies were dose-fractionated into 100 frames with a total dose of ~80 e−/Å2. We collected 2,826 movies for Cm EB MOMP and 3,569 movies for MOMP/mAb-18b complex. Cryo-EM image processing and model building All movies were imported into CryoSPARC (versions 4.1-4.7), followed by patch motion correction using binning of 2 and patch CTF correction125. For the EB MOMP dataset, particles were initially picked using the blob picker and extracted with a box size of 360 pixels. After several rounds of 2D classification, 118,614 clean particles were used for ab initio reconstruction with three classes. Non-uniform refinement with C3 symmetry using 87,699 particles resulted in a 2.96 Å resolution density map (Supplementary Fig. 3)126. To validate the C3-symmetric reconstruction, particles were also reconstructed without imposed symmetry (C1) (Supplementary Fig. 4), and additional 3D classification with filter resolution of 4 Å was performed on C3 symmetry-expanded particles (Supplementary Fig. 5). Neither approach revealed asymmetry or distinct conformational features beyond those observed in the original C3 reconstruction, indicating that application of C3 symmetry did not obscure underlying structural heterogeneity. For the EB MOMP/mAb-18b complex, an initial batch of 25,575 particles was obtained using the blob picker across all micrographs, extracted with box size of 360 pixels, and cleaned through several rounds of 2D classification. Ab initio reconstruction with three classes, using all particles, yielded a single interpretable 3D model of the EB MOMP/mAb-18b complex. The particle pool was further enriched using the Topaz particle picker (version 0.2.5)127. The Topaz model was iteratively trained and optimized on a subset of micrographs, then applied across all micrographs to extract 265,993 particles128. After one round of 2D classification, 110,855 particles were retained for further processing. These particles were further filtered through heterogeneous refinement against the three original ab initio classes. A final set of 51,710 particles was used as an input to non-uniform refinement with C3 symmetry, producing a 3.36 Å resolution density map (Supplementary Fig. 15). No additional masking was applied beyond cryoSPARC’s default dynamic masking, and no sharpening was applied to the deposited maps. Local resolution maps were generated using the CryoSPARC Local Resolution Estimation job. Reconstruction without imposing symmetry (C1) was also performed to validate that no asymmetry or conformational features beyond those observed in C3 reconstruction were present (Supplementary Fig. 16). No 3D classification was performed due to the limit in particle numbers. The initial atomic models for the EB MOMP structure were generated by docking the AF2 models into the molecular map using MOLREP (version 11.9), as implemented in CCPEM (version 1.6.0)129–131. For the EB MOMP-Fab complex, the refined EB MOMP model was docked in the same way, together with IgG2b/κ Fab chains from PDB 4LQF132. The models were then iteratively rebuilt in Coot and refined using Phenix.real_space_refine (version 1.21) and Servalcat (version 0.4)133–140, including the analysis of difference maps generated by Servalcat. Model validation statistics were calculated with MolProbity141, Servalcat, and CCPEM tools (Supplementary Table 1). Structure analysis General inspection and visualization of the structural models were performed with Coot (version 0.9.8), Pymol (version 3.1) and UCSF ChimeraX (version 1.7)142–144. Surface electrostatics analysis was performed using the APBS (version 3.0) PyMOL plugin, with default settings corresponding to pH 7, and PDB2PQR (version 2.1)145 was used to add hydrogens and assign protonation states, atomic charges, and radii146. The DALI server147, the Vector Alignment Search Tool (VAST)148, and Foldseek149 were used to search for structurally similar proteins, while ECOD database71 was used to assess potential relationships within the OmpT-like family and in related families. The comparisons of three-dimensional structures were conducted with LSQKAB150 and Coot (version 0.9.8)133,139 from the CCP4 suite (version 8.0)150. Shape complementarity was analyzed with the SC (version available in CCP4 8.0)84, and buried surfaces were calculated and analyzed using the PISA server (https://www.ebi.ac.uk/pdbe/pisa/)151,152. Contact patterns were examined using CONTACT (version available in CCP4 8.0), AREAIMOL (version available in CCP4 8.0), and LIGPLOT2+ (version 2.3)153. The Caver (version 3.0) Pymol plugin86 was used to search for tunnels, channels, and cavities, with the search conducted using a minimum probe radius of 0.9 Å and the starting point near Asn5 inside the β-barrel. The evolutionary conservation was mapped onto the experimental models using Consurf154 and Consurf Colab notebook. The notebook can be accessed at: https://colab.research.google.com/drive/1PhDXX7k12oUsV6T_xkXC3Rm9R99e7tHz The Prodigy webserver at https://rascar.science.uu.nl/prodigy/85 was used for in silico prediction of ΔG and Kd for the MOMP-Fab complex. Epitope mapping B- and T-cell epitopes for Ct serovar D MOMP were downloaded from the Immune Epitope Database (IEDB)111. Epitopes with less than 100 % sequence identity to the Ct serovar D sequence were excluded from the analysis, resulting in 447 B-cell and 22 T-cell epitopes. For each amino acid position, the number of positive and negative B- and T-cell assays reported in the IEDB for epitopes containing that amino acid was counted, and an amino acid score was calculated as log(1+positive assays)/(1+negative assays). The serovar D trimer structure was modeled by AlphaFold-Multimer using the ColabFold (version 1.5.5) notebook provided by Sergey Ovchinnikov, utilizing the Cm EB MOMP cryo-EM structure as a template.The notebook can be accessed at: https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipyn Conformational sampling was done using the Colab implementation of AlphaFold2-MULTIMER using MMseqs2, with settings max_msa: 16:32, num_seeds: 8. Note that the AF2 predicted structure for MOMP as stored in the AlphaFold Protein Structure Database and available through Uniprot and PDB is flawed, as it includes the N-terminal signal peptide, which distorts the β-barrel structure. 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 Data 1 Supplementary Data 2 Supplementary Data 3 Reporting Summary Transparent Peer Review file Source data Source Data Acknowledgements We thank Tabitha Emde for preparing grids for cryo-EM data collection. We acknowledge Ida Theodor, who, by not boiling the purified native MOMP before running the SDS-PAGE, observed a band that eventually was identified as the MOMP trimer. We thank Drs. Katherine Spoth and Devrim Acehan from Hauptman-Woodward Institute for collecting initial cryo-EM datasets with the Glacios 200 kV instrument; these data were not included in this study. We also thank the Cryo-Electron Microscopy Facility (CEMF) at UT Southwestern Medical Center for maintaining electron microscopes. Funding: This project was funded in whole or in part with federal funds from the National Institute of Allergy and Infectious Diseases, National Institutes of Health, Department of Health and Human Services, under Contract 75N93022C00035 (DB), NIH grants U19AI144184-01 (MAC and LMdlM), R01AI032248 and R01AI030499 (LMdlM), and R35GM145365 (ZO). This work was performed in part under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. The Cryo-Electron Microscopy Facility (CEMF) at UT Southwestern Medical Center is supported by grant RP220582 from the Cancer Prevention & Research Institute of Texas (CPRIT). Author contributions YG collected and analyzed cryo-EM data, solved structures, built structural models, analyzed their features, interpreted their implications for chlamydial biology, and drafted and revised the manuscript, including figures preparation. MLS conceived the experimental approach, analyzed preliminary data, analyzed the structural models’ features, interpreted their implications for chlamydial biology, and drafted and revised the manuscript, including figures preparation. PDH performed computational structure prediction and analysis, mapped existing epitope data to the structure, and drafted and revised the manuscript, including figures preparation. SP conceptualized and initiated the project and generated mAb-18b antibody. AS isolated and purified the Fab for mAb-18b. BVR contributed to the analysis of epitope data, created the initial version of Supplementary Fig. 6, and revised the manuscript. ZO contributed to model building, structural and evolutionary analyses, and revised the manuscript. BWS helped draft and revise the manuscript. MAC provided input on data analysis and helped draft and revise the manuscript. DB collected cryo-EM data, built structural models, analyzed their features, interpreted their implications for chlamydial biology, and drafted and revised the manuscript, including figures preparation. LMdlM conceptualized and initiated the project, grew Cm, purified MOMP, analyzed the structural models’ features, interpreted their implications for chlamydial biology, and drafted and revised manuscript. All authors approved the final version of the manuscript. Peer review Peer review information Nature Communications thanks Jeremy Derrick and the other anonymous reviewers for their contribution to the peer review of this work. A peer review file is available. Data availability The cryo-EM maps generated in this study have been deposited in the Electron Microscopy Data Bank (EMDB) under accession codes EMD-70742 and EMD-70746. The corresponding atomic coordinates have been deposited in the Protein Data Bank (PDB) under accession codes 9OQ7 and 9OQD. Previously published structures used in this study are available in the Protein Data Bank under accession codes 2VDF, 9NH1 and 4LQF. The AlphaFold models (Cm MOMP/OmcA/OmcB, Cpn MOMP, Ct MOMP) generated in this study are provided in the Supplementary Data files. The epitope data used in this study are available in the Source Data file. Source data are provided with this paper. Competing interests YG, DB, and ZO are co-founders of Ligo Analytics, a company that develops software for cryogenic electron microscopy. YG serves as the CEO and is currently employed by Ligo Analytics. ZO is a co-founder of HKL Research, a company that develops and distributes software for X-ray crystallography. A patent application related to this work has been filed on the use of the presented structural models and their features (Applicants: The Regents of the University of California and The Board of Regents of The University of Texas System; Inventors: Luis de la Maza, Dominika Borek, Zbyszek Otwinowski, and Yirui Guo; Application No.: PCT/US2025/051361; Status: pending). All remaining authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Yirui Guo, Megan L. Shelby. Contributor Information Dominika Borek, Email: dominika.borek@utsouthwestern.edu. Luis M. de la Maza, Email: lmdelama@uci.edu Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-72763-4. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Reporting Summary Transparent Peer Review file Source Data Data Availability Statement The cryo-EM maps generated in this study have been deposited in the Electron Microscopy Data Bank (EMDB) under accession codes EMD-70742 and EMD-70746. The corresponding atomic coordinates have been deposited in the Protein Data Bank (PDB) under accession codes 9OQ7 and 9OQD. Previously published structures used in this study are available in the Protein Data Bank under accession codes 2VDF, 9NH1 and 4LQF. The AlphaFold models (Cm MOMP/OmcA/OmcB, Cpn MOMP, Ct MOMP) generated in this study are provided in the Supplementary Data files. The epitope data used in this study are available in the Source Data file. Source data are provided with this paper.