A new paradigm for outer membrane protein biogenesis in the Bacteroidota

Abstract

In Gram-negative bacteria, the outer membrane is the first line of defence against antimicrobial agents and immunological attacks¹. A key part of outer membrane biogenesis is the insertion of outer membrane proteins by the β-barrel-assembly machinery (BAM)^2-4. Here we report the cryo-electron microscopy structure of a BAM complex isolated from Flavobacterium johnsoniae, a member of the Bacteroidota, a phylum that includes key human commensals and major anaerobic pathogens. This BAM complex is extensively modified from the canonical Escherichia coli system and includes an extracellular canopy that overhangs the substrate folding site and a subunit that inserts into the BAM pore. The novel BamG and BamH subunits that are involved in forming the extracellular canopy are required for BAM function and are conserved across the Bacteroidota, suggesting that they form an essential extension to the canonical BAM core in this phylum. For BamH, isolation of a suppressor mutation enables the separation of its essential and non-essential functions. The need for a highly remodelled and enhanced BAM complex reflects the unusually complex membrane proteins found in the outer membrane of the Bacteroidota.

Fig. 1. Structure of the BAM_Fj complex.

a, Size-exclusion chromatography profile of the purified BAM_Fj preparation and Coomassie-stained SDS–PAGE gel of the indicated fractions. Bands were identified by peptide fingerprinting. Fraction A was used to determine the structure of the full BAM_Fj complex and fraction B was used for the structure of the BamAP complex. Similar data were obtained for three independent preparations. b, Cryo-electron microscopy (cryo-EM) volume for the BAM_Fj complex overlaid on the hybrid model shown in d. The volume is shown at a high contour level (coloured) and at a low contour level (semi-transparent). c,d, Comparison of the most similar E. coli BAM complex structure (darobactin 9-bound complex; PDB: 8ADI) (c) with the BAM_Fj complex (d). Structures are shown in cartoon representation with lipids and metal ions in space-filling atom representation coloured by element. For BAM_Fj, the poorly resolved BamA POTRA 1–3 domains and BamP C-terminal domain are modelled by placing AlphaFold structures in the electron microscopy density (lighter coloured domains). e–i, The BAM_Fj hybrid model (Supplementary Data 1) with protein components in space filling representation and lipids shown as atom spheres coloured by element. e, View in the same orientation as d, left. f, The N-acyl and S-diacylglyceryl groups attached to the N-terminal cysteine of BamH. g, The resolved portion of a lipopolysaccharide (LPS) molecule in the outer leaflet of the OM and two ordered phospholipid molecules on the inner leaflet of the OM. h, View from the periplasm with the periplasmic side of the complex cut away to the membrane midpoint. i, View from the exterior with the extracellular side of the complex cut away to the membrane midpoint. Source data

Fig. 2. Structural features of the BAM_Fj subunits.

a,b, Comparison of the F. johnsoniae (a) and E. coli (PDB: 8ADI) (b) BamA barrels. The strands closest to the viewer have been removed, revealing BamP within the F. johnsoniae barrel. The structure in a also highlights the hydrogen-bonding interaction between the side chain of Gln801 (substituted in the bamH suppressor mutant) and the main chain of Gly591 (both in ball and stick representation). c, Cartoon representation of BamP (orange). The C-terminal domain (pale yellow) is an AlphaFold model docked into the electron microscopy density. Portions of the BamA barrel (blue) are shown for orientation. d, Sequence conservation and intra-chain interactions of the inter-domain loop of BamP (cartoon with ball and stick side chains) within the BamA barrel (surface representation). Min, minimum; max, maximum. e,f, Superimposition of BamG (chainbows colouring) and E. coli FadL (grey; PDB: 3DWO). A proposed substrate-mimicking C₈E₉ detergent molecule in FadL is shown in grey spheres. In f, the front walls of the barrels, oriented as in e, left, are cut away and the N-terminal amino acid of FadL together with the equivalent sequence position residue in BamG are shown as spheres. g, View from outside the cell showing how the N-terminal region of BamH is bound by BamG. BamG is in surface representation with the N tail (residues 1–32) coloured blue. Partial structures of BamA and BamH are shown in cartoon representation with the N-terminal cysteine of BamH and its attached lipid groups shown as atomic spheres and coloured by atom. h, The BAM_Fj extracellular canopy viewed from BamA. Bound calcium ions and their coordinating side chains in BamM and glycosylation of BamH are shown in ball and stick representation. i, Surface conservation (left) and electrostatics (right, kcal (mol.e)^–1 at 298 K) of the extracellular canopy in the same orientation as in h.

Fig. 3. Structural and functional consequences of losing BamP.

a, Comparison of the structure of the BamAD complex from a BamP-deleted (ΔbamP) background and a BamAP complex from the wild-type (WT) background. The proposed phenylalanine molecule is shown in orange space-filling representation. b, Overlay of the structures shown in a aligned on the N-terminal 100 residues of the BamA barrel. c, Detail from b showing the enlargement and register shift of the sheet between the BamA barrel N and C terminal strands upon BamP removal and the incompatible binding modes of BamP and the putative phenylalanine (orange space-filling representation). Spheres show the Cα atom of Gly897 in each model. d, Cryo-EM volume for the BamAD complex from a BamP-deleted background reveals a partially occupied second barrel (silver). Inset shows the putative phenylalanine density. e, Superposition of the complex in d with an E. coli BamA–EspP complex (PDB: 8BO2; yellow) aligning on the blue BamA_Fj. The view is from the cell exterior but truncated in the periplasm for clarity. A second copy of BamA_Fj (silver) has been docked into the second barrel density in d and occupies the same position as the EspP substrate (yellow barrel, right). f, Removal of BamP homologues sensitizes F. johnsoniae to darobactin. The ΔporV Δplug background permeabilizes the OM by opening the T9SS translocon channel. g, BamP overexpression restores darobactin resistance to a strain lacking all BamP homologues. Strains contain plasmids overexpressing Twin-Strep-tagged BamP (p^TSBamP) or BamP4 (p^TSBamP4). f,g, Similar data were obtained for three biological repeats.

Fig. 4. Depletion analysis of the essential BAM_Fj subunits.

Strains are the wild-type and depletion strains for BamA (bamA^dep), BamG (bamG^dep) and BamH with either a strong (bamH^dep) or weak (bamH^low) inducible promoter. SkpA is a periplasmic protein to control for OM integrity; GroEL is a loading control. a–d, Strains were cultured in rich (Casitone yeast extract, CYE) medium. The aTC inducer of the target gene was removed at 0 h where indicated (−aTC) to initiate subunit depletion. Samples in b–d were taken at the indicated time points in a. a, Growth curves. Data are mean ± s.d. b, Immunoblots of whole-cell lysates. c, Representative transmission electron microscopy images showing OM defects in the depletion strains. Yellow arrows highlight budding OM vesicles; black arrows highlight OM blebbing and rupture. Scale bar, 500 nm. d, Depletion of BAM_Fj subunits for 6 h does not change the surface exposure of the SLP SusE as assessed by proteinase K accessibility. Triton X-100 permeabilizes the OM. Reactions were stopped immediately (t₀) or after 20 min (t₂₀) and analysed by immunoblotting. e, Comparative whole-membrane proteome analysis of depleted (−aTC) versus induced (+aTC) strains collected at the 6 h time point in a. Data points for OMPs and SLPs are coloured as indicated and the most highly expressed OM proteins are labelled. A significance threshold is drawn according to a two-tailed t-test with a false discovery rate (FDR) of 0.1 and a variance correction constant S₀ of 0.1. Data are averaged over three biological repeats. FC, fold change. f,g, Analysis of chronic BamH depletion in an induced strain (bamH^low + aTC) in which a weak promoter results in the incomplete restoration of wild-type BamH levels. f, Whole-cell immunoblots. Arrow indicates BamH; asterisk indicates a non-specific band. g, As e but comparing chronic BamH depletion (bamH^low strain + aTC) relative to the wild type. a–d,f, Similar data were obtained for three biological repeats. Source data

Fig. 5. Characterization of a bamH suppressor mutant.

Comparative characterization of the recreated bamH^sup mutant (bamA^Q801K ΔbamH ΔbamH2) and wild-type strains. a, Growth on rich (CYE) medium in the absence of aTc. Data are mean ± s.d. b, Whole-cell immunoblots. SkpA is a periplasmic protein to control for OM integrity. GroEL is a cytoplasmic protein as loading control. BamA and BamG are detected via epitope tags. Asterisk indicates non-specific bands. c, Representative transmission electron microscopy images of the wild type and bamH^sup mutant. Yellow arrows highlight budding OM vesicles. Scale bar, 500 nm. d, Comparative whole-membrane proteome analysis of the bamH^sup strain relative to a BamH-induced strain (bamH^dep + aTC). Data points for OMPs and SLPs are coloured as indicated and the most highly expressed OM proteins are labelled. Proteins that show poor recovery in the bamH^sup strain in a post hoc ANOVA with BamH-induced and depleted strains are numbered as in Extended Data Fig. 8b. A significance threshold is drawn according to a two-tailed t-test with a FDR of 0.1 and a S₀ of 0.1. Data are averaged over three biological repeats. a–c, Similar data were obtained for three biological repeats. Cells were analysed (b,c) and membranes prepared (d) at the 6 h time point in a. e, Size comparison between BAM_Fj and a typical SusCD complex and OmpA. SusCD and OmpA are illustrated using homologous proteins of known structure from other organisms (labelled with their PDB accession numbers). Source data

Extended Data Fig. 1. Workflow for the cryoEM analysis of the F. johnsoniae BAM_FJ complex and map quality metrics.

a, Twin-Strep-tagged BamA complexes were purified by Streptactin affinity chromatography and size exclusion chromatography and the major (highest molecular size) peak was analyzed. See Fig. 1a for corresponding SDS-PAGE analysis of this material. Image processing workflow for the BamA complexes. b,c, Gold-standard Fourier Shell Correlation (FSC) curves used for global resolution estimation (b) and local resolution estimate (c) of consensus (left), extracellular (middle), or periplasmic (right) volumes from the BAM_FJ complex.

Extended Data Fig. 2. Further structural analysis of the BAM_FJ complex.

a, Chain ordering. The indicated subunit in each panel is rainbow-coloured from the N- (blue) to C-terminus (red). b, BamH (green) in cartoon representation with ligands (glycosylation and lipidation) as space fill representation. The protein is viewed from the direction indicated in (a) and overlaid with the closest structural homologue as judged by PDBeFold 2.58, the chondroitin sulfate-binding carbohydrate binding module of a chondroitinase (dark grey, PDB 8wab, RMSD 2.5 Å across 64 equivalent residues), which is defined as a DNRLRE domain-containing protein by UniProtKB. c, BamM (tan) in cartoon representation with bound metals shown as purple spheres and coordinating residues in ball-and-stick representation. The protein is viewed from the direction indicated in (a) and overlaid with the closest structural homologue as judged by PDBeFold 2.58, the peptidyl-prolyl isomerases (PPI) subunit (dark grey) from the Type 9 Secretion System translocon complex (PDB: 6h3i chain B, RMSD 0.75 Å across 74 equivalent residues). d, Glycosylation and lipidation of BamH shown in ball-and-stick representation within the cryoEM volume in the context of the full chain (left) and in closeup (right). The modelled glycosylation was assigned on the basis of the EM density informed by prior studies of O-glycosylation in Bacteroidota but without biochemical identification. e, Bound metals within BamM modelled as calcium ions (purple spheres) with coordinating residues shown in ball-and-stick representation. The model is displayed within the EM density (insets) or showing just the EM density for the metal ions (full structure). The metals were assigned as Ca ions based on their co-ordination chemistry (O-only ligation, variable co-ordination number and geometry, and appropriate bond lengths) and refining to thermal mobility (B) factors that matched those of the ligating protein atoms.

Extended Data Fig. 3. Genomic organisation of F. johnsoniae bam genes and biochemical evidence that BamP4 interacts with the BAM_FJ complex.

a, Genomic organisation of F. johnsoniae bam genes. porG and skpA at the bamA locus code for components of the Type 9 Secretion system. BamH2 would be unlikely to interact with BamM as it lacks the protruding β-hairpin that BamH uses for this purpose. BamP homologues have related folded domains but markedly diverge in the interdomain loop. b-d, BamP interacts with the BAM_FJ complex. Strains in which the native BamA protein was fused to a HA tag (HA-bamA allele) were transformed with plasmids overproducing N-terminally Twin-Strep-tagged BamP (p^TSBamP) or BamP4 (p^TSBamP4). wt, wild type. Similar data were obtained for three biological repeats. b, Immunoblots of whole cells showing overproduction of BamP or BamP4 relative to native BamP levels. The cytoplasmic protein GroEL was used as a loading control. c,d, Affinity purification of Twin-Strep-tagged BamP and BamP4 complexes. c, Coomassie-stained SDS–PAGE gel of the wash and elution fractions. BAM subunits were assigned by comparison with (d) and Fig. 1a. d, Immunoblotting analysis of the elution fractions with anti-HA (to identify BamA), anti-BamH, and anti-Twin-Strep (to identify BamP and BamP4) antibodies.

Extended Data Fig. 4. Phenotypic characterisation of strains with deletions in BAM_FJ subunits or BAM_FJ subunit homologues.

a, Results of attempts to delete Bam_FJ subunits and their homologues in different genetic backgrounds. Mutations and their combinations that were viable are indicated by green dots, while those that could not be constructed are indicated by red dots and are assumed to disrupt essential cell functions. b, Immunoblots of whole cells and isolated membranes of strains containing in-frame deletions of bamM or bamP. The cytoplasmic protein GroEL and OM protein OmpA were used as loading controls. *, non-specific band. Similar results were obtained from 3 biological repeats. c, Growth curves on rich CYE medium. Shown are the means ± 1 SD from three biological repeats. d, Growth curves on minimal medium containing either galactomannan or xyloglucan as carbon source. Shown are the means ± 1 SD from three biological repeats. e, Spreading (gliding) morphology of colonies on agar. Scale bar, 5 mm. Similar data were obtained for three biological repeats. f,g, OM integrity assays. Cells were grown on CYE agar with the indicated additions. wt, wild type; ΔbamP-P3, strain deleted for BamP, BamP2, and BamP3; ΔbamP-P4, strain deleted for all BamP homologues (ΔbamP-P3 ΔbamP4). The ΔporV Δplug background permeabilizes the OM through opening the T9SS translocon channel. Similar data were obtained for three biological repeats. h, BamP overproduction restores vancomycin resistance to a strain lacking all BamP homologues (strain ΔbamP-P4). Where indicated strains were transformed with plasmids overproducing N-terminally Twin-Strep-tagged BamP (p^TSBamP) or BamP4 (p^TSBamP4). wt, wild type. Similar data were obtained for three biological repeats.

Extended Data Fig. 5. Workflow for the cryoEM analysis of the BamA complex isolated from a ΔbamP background.

a, Size exclusion chromatography profile of BamA complexes purified from a BamP-deleted background together with a Coomassie-stained SDS–PAGE gel of the indicated peak fraction that was used for structure determination. BamA* indicates a proteolysis product of BamA. Similar results were obtained from 2 biological repeats. b, Image processing workflow for the ΔBamP complex. c, Gold-standard Fourier Shell Correlation (FSC) curves used for global resolution estimation. d, Local resolution estimate of the volume, displayed at two contour levels. e, Density for the unassigned second barrel taken from the focused 3.7 Å volume is shown with either a second copy of BamA or BamG docked. Two views are shown from the side (left) or the cell exterior (right). The shape and size of the volume is clearly more consistent with this being a second copy of BamA than BamG. Source data

Extended Data Fig. 6. Workflow for the cryoEM analysis of the BamAP complex.

The sample used was fraction B from Fig. 1a. a, Image processing workflow for the BamAP complexes. b, Gold-standard Fourier Shell Correlation (FSC) curves used for global resolution estimation. c, Local resolution estimate of the volume.

Extended Data Fig. 7. Depletion analysis of the essential F. johnsoniae BAM complex subunits.

a, Design of an anhydrotetracycline (aTC)-inducible system for the depletion of essential target genes in F. johnsoniae. The TetR repressor is constitutively expressed under the control of the F. johnsoniae ompA promoter (P_ompA) and the target gene is regulated by a designed TetR-repressed promoter (P_induc). In the presence of the inducer aTC repression of the target gene by TetR will be released. The genetic system is integrated into the F. johnsoniae chromosome at a neutral locus. b, Sequences of the designed inducible P_ompA-induc and P_{fjoh_0824-induc} promoters. tetO2 arrays are placed upstream and downstream of the conserved −33 and −7 RNA polymerase binding sites (boxed) of the selected promoters. c, Tight regulation of protein expression by the designed inducible systems. Strains expressing NanoLuc under the control of either the P_ompAindc promoter (XLFJ_1095) or the P_{fjoh_0824induc} promoter (XLFJ_1100) were grown to mid-exponential phase (OD₆₀₀ = 0.6) in the presence or absence of aTC and the luminescence signal measured. Error bars represent the mean ± 1 SD from three biological repeats. P values were determined with a two-sided paired Student’s t-test. RLU, relative luminescence units. d, Comparison of the expression levels of Bam subunits in the wild type strain (wt, XLFJ_1078) and corresponding depletion strains grown in the presence of the inducer aTc (bamA^dep, XLFJ_1129; bamG^dep, XLFJ_1115; bamH^dep, XLFJ_1140). Whole cell immunoblotting of cells grown to mid-exponential phase (OD₆₀₀ = 0.6). The blots for the depleted subunit are boxed in red. The BamH blot for the BamH depletion comparison is overexposed (OE) relative to the other BamH blots in order to detect the low levels of BamH in the depletion strain. BamA and BamG are detected via epitope tags. *, non-specific band. Similar results were obtained for 3 biological repeats. e, Phase contrast images of cells sampled at the indicated time points in the BAM_FJ subunit depletion experiments shown in Fig. 4a. Scale bar, 10 µm. Similar results were obtained for 3 biological repeats. f, The major F. johnsoniae SUS complex is composed of SusC (Fjoh_0403), SusD (Fjoh_0404), and SusE (Fjoh_0405). The native SUS complex was purified via a Twin-Strep tag on the N-terminus of SusC followed by size exclusion chromatography and analysed on a Coomassie-stained SDS–PAGE gel. Proteins were identified by peptide mass fingerprinting. Similar data were obtained for two biological repeats. g, Outer membrane vesicle (OMV) production does not increase upon BAM depletion. Immunoblotting of the OM protein SusC in whole cells or the OMV fraction at the 6 h time point in Fig. 4a. GroEL serves as loading control. Similar results were obtained for 3 biological repeats. h, An exogenously-expressed His-tagged variant of SusE (SusE^His) is incorporated into the native SusCDE complex. SusC-containing complexes were purified as described in f from cells expressing SusE^His from a plasmid. The purified material was separated by SDS-PAGE and characterized by Coomassie-staining (Left) and anti-His tag immunoblotting (Right). Similar data were obtained for two biological repeats. i, Exemplar Coomassie-stained SDS-PAGE gel of the whole membrane samples used for the comparative proteome analysis (Fig. 4e,g) of induced/non-induced (i.e. undepleted/depleted) BAM subunit depletion strains harvested at the 6 h time point in Fig. 4a. Proteins present in the two obviously depleting bands were assigned by peptide mass fingerprinting. The data are representative of the three repeats used for the proteomics analysis. Source data

Extended Data Fig. 8. OM proteomics data comparisons.

Heat maps of the indicated strains after hierarchical protein clustering of the entire datasets and post hoc ANOVA testing. Only proteins classified as OMPs or SLPs are displayed. Colours indicate HSD (honestly significant difference) values according to the intensity panel. a, Comparison of the datasets used in Fig. 4e and g. b, Comparison of the bamH^sup mutant dataset with the induced and non-induced bamH^dep datasets. The non-recovered proteins are numbered as in Fig. 5d and assigned to SusC, SusD, other SUS SLP, or TonB-dependent transporter (TBDT) protein families. TBDTs are 22-strand OMPs that are related to the SusC family. Source data

Extended Data Fig. 9. Isolation of BamA complexes from subunit depleted backgrounds and phenotypic characterisation of the bamH^sup strain.

a,b, Isolation of BamA complexes either (a) in the absence of BamM or (b) after 6 h of depletion of the essential BamG or BamH subunits. Size exclusion chromatography profile of Twin-Strep-tagged BamA complexes purified by Streptactin affinity chromatography (Left) and a Coomassie-stained SDS–PAGE gel of the indicated peak fractions (Right). BamA* indicates a proteolysis product of BamA. The identities of the BamA* and BamD + BamP bands were assigned by peptide fingerprinting. Similar results were obtained from 2 biological repeats. c-f, Characterization of the recreated bamH^sup mutant (bamA^Q801K ΔbamH ΔbamH2). wt, wild type. Similar results were obtained for three biological repeats. c, OM integrity assays. Cells were grown on CYE agar with the indicated additions. d, Surface exposure of the SLP SusE. Strains expressing a protease-sensitive His-tagged variant of SusE (SusE^His) were treated as indicated with Proteinase K and the detergent Triton X-100 (to permeabilise the OM). Reactions were stopped immediately (t₀) or after 20 min (t₂₀) and analysed by immunoblotting with His tag antibodies. The periplasmic protein SkpA serves as an OM integrity control. e, Spreading (gliding) morphology of colonies on agar. Scale bar, 5 mm. f, Purification of the native SusCDE complex via a Twin-Strep tag on the N-terminus of SusC followed by size exclusion chromatography. Analysed on a Coomassie-stained SDS–PAGE gel. Source data