Bacterial strain culturing and genetic manipulation

Bacterial strains used in this study are listed in Supplementary Table 6. E. coli strains were grown at 37 °C in lysogeny broth (20 g l−1, BioShop, LBL405) with 100 µg ml−1 ampicillin (BioShop, AMP201) and on 1.5% agar plates. For constructing B. thetaiotaomicron chromosomal mutants and protein production, strains were cultured in enriched brain–heart infusion (EBHI), containing 37 g l−1 brain–heart infusion powder (OXOIDLTD, CM1135B), 5 g l−1 yeast extract and 1 μg ml−1 haemin (Sigma-Aldrich, 51280), supplemented with 0.5 g l−1 cysteine (Melford, C50010) and 1 μg ml−1 vitamin K (Sigma-Aldrich, M5625). For growth curve and proteinase K experiments, defined minimal medium (MM) supplemented with 1 μg ml−1 haemin and 0.5% carbon source was used, unless otherwise indicated. Bacteria were grown under anaerobic conditions at 37 °C in a Don Whitley A35 workstation. B. thetaiotaomicron genetic deletions and mutations were created by allelic exchange using the pExchange-tdk vector67. Briefly, the constructed pExchange-tdk plasmids, containing the mutations and deletions plus ~700 bp flanks up- and downstream (DNA inserts were synthesized by Twist Bioscience), were transformed into S17 λ pir E. coli cells, to achieve conjugation with the B. thetaiotaomicron recipient strain. The conjugation plates were scraped, and B. thetaiotaomicron cells that underwent a single recombination event were selected for by plating on BHI–haemin agar plates containing gentamicin (200 μg ml−1, Melford, G38000) and erythromycin (25 μg ml−1, Duchefa, E0122). Eight colonies were restreaked on fresh BHI–haemin–gentamicin–erythromycin plates. Single colonies were cultured in BHI–haemin overnight and pooled. To select for the second recombination event, pooled cultures were plated on BHI–haematin agar plates containing 5-fluorodeoxyuridine (FUdR; 200 μg ml−1, Thermo Scientific, L16497.ME). After 48–72 h of growth, FUdR-resistant colonies were restreaked on fresh BHI–haematin–FUdR. From these, single colonies were cultured in BHI and genomic DNA was extracted and screened for the correct mutations using diagnostic PCR. This procedure was used to introduce the N-terminal His7-tag on BtBamA, and bamH and bamI deletions into the chromosome of B. thetaiotaomicron VPI-5482 tdk− and BamAhis, as well as a bamH deletion in the Bt1927-ON background.

P. gingivalis strains were grown anaerobically (90% nitrogen, 5% carbon dioxide and 5% hydrogen) at 37 °C in enriched tryptic soy broth (eTSB; 30 g l−1 tryptic soy broth (Sigma-Aldrich, T8907), 5 g l−1 yeast extract (BioShop, YEX401), 5 mg l−1 haemin (Sigma-Aldrich, H9039), 0.5 mg l−1 menadione (Sigma-Aldrich, M5625) and 0.25 g l−1l-cysteine (BioShop, CYS342)) or on eTSB blood agar plates with 1.5% agar (BioShop, AGR001) and 5% defibrinated sheep blood (Biomaxima SL0160). For mutant selection, appropriate antibiotics were added: erythromycin (Sigma-Aldrich, E6376) at 5 μg ml−1 or tetracycline (BioShop, TET701) at 1 μg ml−1. P. gingivalis mutants were generated through homologous recombination. For deletion mutants, 1-kb fragments flanking the bamH, bamI or bamK genes and chosen antibiotic resistance cassettes were amplified by PCR and cloned into pUC19 plasmid using restriction cloning. For expression of C-terminally-His-tagged BamH protein, master plasmid was first generated. A 1-kb fragment coding the C-terminal part of BamH and a 1-kb fragment downstream of it as well as an antibiotic resistance cassette were amplified and cloned into pUC19 plasmid using restriction cloning. The 7His-tag was then introduced into the master plasmid by PCR. Plasmids and primers used for mutant construction are listed in Supplementary Tables 7 and 8, respectively. Plasmids were electroporated into P. gingivalis competent cells, followed by plating the cells on eTSB blood agar with appropriate antibiotics and growth for 10 days. All generated plasmids were confirmed by PCR and Sanger sequencing. B. thetaiotaomicron and P. gingivalis clones were screened by PCR and verified by Sanger sequencing.

BtBAM expression and purification

The B. thetaiotaomicron bamAhis strain was cultured overnight in EBHI and used to inoculate 10 l of rich medium (25 g l−1 brain–heart infusion powder, 3.5 g l−1 yeast extract, 1 µg ml−1 vitamin K, 0.5 g l−1l-cysteine) that was equilibrated at 37 °C overnight in an anaerobic chamber. For each 500-ml bottle of pre-warmed medium, 3 ml of overnight culture was used. The cultures were grown for 5–7 h until late exponential phase (optical density at 600 nm (OD600) ~1.8–2.0). Cultures were pelleted by centrifugation at 6,000 g for 30 min at 4 °C, and the pellets were stored at −20 °C.

Cell pellets were processed in 2-l batches to maximize yield. Cells were thawed, homogenized in Tris-buffered saline (TBS; 20 mM Tris–HCl pH 8.0, 300 mM NaCl), supplemented with DNase I (Roche, 04716728001) and lysed by a single pass through a cell disruptor at 22–23 kpsi. The total membrane fraction was isolated by ultracentrifugation at 200,000 g for 50 min at 4 °C. Membranes were homogenized and solubilized in 30 ml TBS with 0.75% DDM (Anatrace, D310S) and 0.75% decyl maltoside (DM, Anatrace, D322S) for 1 h at 4 °C with stirring. Insoluble material was pelleted by ultracentrifugation at 200,000 g for 30 min at 4 °C. The supernatant was loaded on a 1.5-ml chelating Sepharose (Cytiva, 17057502) gravity flow column charged with Ni2+ ions at 4 °C and equilibrated with TBS supplemented with 0.15% DDM. The column was washed with 30 column volume (CV) TBS supplemented with 30 mM imidazole and 0.15% DDM, and bound material was eluted with 4 CV TBS supplemented with 200 mM imidazole and 0.15% DDM. The eluate was concentrated using an Amicon Ultra filtration device column (100 kDa molecular weight cut-off, Sigma-Aldrich, UFC9100) and loaded on a Superdex 200 Increase 10/300 GL (Cytiva, 28990944) equilibrated in 10 mM HEPES–NaOH, 100 mM NaCl and 0.03% DDM pH 7.5. Peak fractions were analysed by SDS-PAGE and Coomassie staining for purity. Gel bands of interest were cut out and identified by peptide fingerprinting at the Metabolomics and Proteomics Laboratory, University of York. Material purified from five separate 2-l batches was combined and subjected to a final size-exclusion chromatography run as above. The BtBAM complex was concentrated, flash-frozen in liquid nitrogen and stored at −80 °C.

BtBAM cryo-EM structure determination

Purified BtBAM complexes at 8 mg ml−1 were applied to glow-discharged Quantifoil R1.2/1.3 holey carbon 200 mesh Cu grids (Oxford Instruments, 51-1625-0131), followed by blotting and plunge-freezing in liquid ethane using a Vitrobot Mark IV device at 4 °C and 100% humidity. Data were collected on a Titan Krios electron microscope operating at 300 kV accelerating voltage on a Falcon 4i direct electron detector with a Selectris energy filter (10 eV slit). Videos were recorded in electron-event representation format at a magnification of 165,000, corresponding to a sampling rate of 0.74 Å per pixel. The defocus was varied between −0.8 µm and −2.0 µm during data collection. The total dose for each video was ~35 e− Å−2. A total of 13,558 videos were collected in total. Data collection parameters are summarized in Supplementary Table 2.

The cryo-EM data processing workflow is shown in Supplementary Fig. 1. All data processing was done in cryoSPARC v4.5.3 (ref. 68). Videos were patch motion corrected, followed by patch contrast transfer function (CTF) correction. Particles were picked manually to generate 2D classes for template-based picking. In total, 2,266,553 particles were extracted in 224 × 224 pixel boxes (1.48 Å per pixel) and subjected to two rounds of 2D classification. Particles from 2D classes resembling protein density were pooled into a single particle stack containing 308,834 images, which was then used for ab initio reconstruction with 4 classes. 2D classes containing 57,300 noise and junk particles were used for ab initio reconstruction of 4 decoy classes. The 308,834-particle stack was subjected to heterogeneous refinement using the 8 ab initio classes as initial volumes. A total of 137,389 particles from the best two classes from the first round of heterogeneous refinement were subjected to 3D variability analysis32 and to another round of heterogeneous refinement against the same initial volumes as in the first round. This resulted in 2 classes with clear protein features: class 1 with 49,137 particles with no periplasmic density and class 2 with 55,224 particles and periplasmic density. Particles belonging to these two classes were independently used in non-uniform refinement69, yielding 3.41 Å (class 1) and 3.49 Å (class 2) reconstructions. Refined particles were re-extracted in 448 × 448 pixel boxes (0.74 Å per pixel) and subjected to another round of non-uniform refinement with enabled per-particle defocus and global CTF parameter (tilt and trefoil) refinement. A total of 48,843 particles from class 1 yielded a final reconstruction at 3.28 Å; 56,321 particles from class 2 yielded a final reconstruction at 3.46 Å. Global resolution was estimated using gold-standard Fourier shell correlation curves and the 0.143 criterion (Supplementary Fig. 1c). Particle viewing direction distribution plots and local resolution estimates are given in Supplementary Fig. 1d–f.

BtBamHIJK AF2 models were docked into the class 1 volume and subjected to cycles of manual adjustment in Coot70 and real space refinement in Phenix71. Mannose monosaccharides were placed in O-glycan densities to prevent the protein model from moving into the glycan density. Similarly, BtBamADG AF2 models were docked, manually adjusted and refined against the class 2 volume. The refined BtBamHIJK class 1 model was rigid-body-refined against the class 2 volume, which had much poorer density for the BtBamHIJK region than class 1 (Supplementary Fig. 1f), to obtain the complete BtBAM model. A composite map incorporating the best parts of the volume representing each class was generated using phenix.combine_focused_maps. Model refinement statistics are shown in Supplementary Table 2.

PgBAM cryo-EM structure determination

Purified PgBAM complexes at 4.9 mg ml−1 were applied to glow-discharged Quantifoil R2/1 holey carbon Cu grids (200 mesh) (EM Resolutions, QR21200Cu100), followed by blotting and plunge-freezing in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific) set to 95% humidity and 4 °C with the following blotting parameters: 2 s blot time, 0 s wait time, blot force −4, 0 s drain time and a total blot of 1. Data were collected using a Titan Krios G3i (Thermo Fisher Scientific; Solaris) electron microscope operating at 300 kV accelerating voltage on a Gatan K3 Summit direct electron detector with a Gatan Quantum energy filter. Videos were recorded in TIFF format at a magnification of ×105,000, corresponding to a sampling rate of 0.84 Å per pixel. The defocus was varied between −0.6 μm and −1.5 μm during data collection. The total dose for each video was ~41 e− Å−2. In total, 11,459 videos were collected. Data collection parameters are summarized in Supplementary Table 2.

The cryo-EM data processing workflow is shown in Supplementary Fig. 3. All data processing was done in cryoSPARC v4.5.3. Videos were patch motion corrected, followed by patch CTF correction. Three rounds of 2D classification were performed to generate 2D classes for template-based picking. A total of 4,830,281 particles were extracted in 200 × 200 pixel boxes (1.68 Å per pixel) and subjected to two rounds of 2D classification. Particles from 2D classes resembling protein density were pooled into a single particle stack containing 349,717 images, which was then used for ab initio reconstruction with 4 classes. 2D classes containing 9,273 noise and junk particles were used for ab initio reconstruction of 3 decoy classes. Particles were re-extracted in 384 × 384 pixel boxes (0.84 Å per pixel). The 104,912-particle stack was subjected to heterogeneous refinement using the 5 ab initio classes as initial volumes. A total of 101,685 particles from the best 2 classes were combined and subjected to ab initio reconstruction (2 classes) with the class similarity parameter set to 0.0001. The class containing 86,584 particles was used in non-uniform refinement, yielding a 3.45-Å reconstruction. Refined particles were subjected to local CTF refinement and global CTF refinement and used again in non-uniform refinement, yielding a final reconstruction at 3.24 Å (PgBAM without periplasmic density). For PgBAM with periplasmic density, after the 2 initial rounds of 2D classification following template picking, the particles were subjected to a further 5 rounds of 2D classification (Supplementary Fig. 3). The resulting 13,929 particles were subjected to heterogeneous refinement with one protein and 2 decoy volumes. Protein class was used in non-uniform refinement, yielding a final reconstruction at 4.26 Å (PgBAM with periplasmic density). Global resolution was estimated using gold-standard Fourier shell correlation curves and the 0.143 criterion (Supplementary Fig. 3c). Particle viewing direction distribution plots and local resolution estimates are shown in Supplementary Fig. 3d–f. AF2 models of all subunits were docked into the final volume of PgBAM without periplasmic density and subjected to cycles of manual adjustment in Coot and real space refinement in Phenix. Mannose monosaccharides were placed in O-glycan densities to prevent the protein model from moving into the glycan density. Model refinement statistics are shown in Supplementary Table 2.

Growth curves

Overnight EBHI cultures of B. thetaiotaomicron were used to inoculate 5 ml of fresh EBHI (1:25), followed by incubation at 37 °C for 3–4 h in an anaerobic cabinet. The cells were collected by centrifugation for 3 min at 2,900 g, 20 °C, and resuspended in 1 ml pre-warmed MM. Glass tubes with EBHI or MM (including carbon source) and supplements indicated in the main text were inoculated with the resuspended cells at OD600 = 0.04, and the cultures were dispensed in 0.2-ml aliquots in sterile 96-well clear polystyrene cell culture plates (Corning Costar, 3599). Medium without any added cells was included as blank. Plates were incubated anaerobically at 37 °C for 30 min and sealed with a Breathe-Easy gas-permeable seal (Diversified Biotech, BEM-1), followed by growth monitoring for 48 h in a Biotek Epoch microplate reader housed inside an anaerobic cabinet. Blank well readings were subtracted from wells with added cells, and triplicate readings were averaged. The experiments were repeated at least two times on different days to ensure consistency.

P. gingivalis ΔbamH, ΔbamI and ΔbamK mutants were first grown in eTSB overnight. The cultures were resuspended in fresh eTSB or washed and resuspended in DMEM (Thermo Fisher Scientific, A1443001) supplemented with 1% BSA (BioShop, ALB001), 0.25 g l−1l-cysteine and either regular or decreased amounts of vitamin K (0.5 mg l−1 and 0.05 mg ml−1, respectively) and haemin (5 mg l−1 and 0.5 mg l−1, respectively). In each case, the OD600 was adjusted to 0.2 and the growth was monitored for 48 h through OD600 measurements at predetermined time intervals, as P. gingivalis does not grow in multi-well plates.

PgBAM expression and purification

The PgBAM complex was isolated from His-tagged BamH-expressing P. gingivalis according to a previously described protocol44. Briefly, cells from 5 l were collected by centrifugation at 6,500 g for 40 min and resuspended in TBS with 1 mM Nα-tosyl-l-lysine chloromethyl ketone (TLCK, Sigma-Aldrich, T8907), 1× cOmplete EDTA-free Protease Inhibitor Cocktail (Roche, 11873580001) and 20 μg ml−1 DNase I (Roche, 11284932001), followed by lysis with Constant Systems Cell Disruptor (Model TS2) at 23 kpsi. The total membrane fraction was separated from the soluble fraction through ultracentrifugation at 200,000 g for 1 h and then homogenized and solubilized in TBS with 1% n-dodecyl β-d-maltoside (DDM, Anatrace, D310) at 4 °C overnight. The extract was clarified by ultracentrifugation at 200,000 g for 50 min and loaded on 1.5 ml pre-equilibrated nickel-affinity resin (Chelating Sepharose, GE Healthcare, 17531802). The column was washed with 15 CV of TBS containing 0.2% DDM and 30 mM imidazole (BioShop, IMD508), and the bound protein was eluted with 2 CV of TBS containing 0.2% DDM and 250 mM imidazole. The complex was further purified by size-exclusion chromatography in 10 mM HEPES pH 8.0, 100 mM NaCl and 0.03% DDM using a Superdex 200 Increase 10/300 GL column (Sigma-Aldrich, GE28-9909-44). In pulldown experiments shown in Fig. 5g,h, RagB-His from wild-type and ΔbamH cells was purified in the same manner, except lauryldimethylamine-N-oxide (Anatrace, D360) was used instead of DDM.

Pulldown proteomics sample preparation, mass spectrometry and analysis

Protein solutions were diluted 1:1 with aqueous 10% (v/v) SDS and 100 mM triethylammonium bicarbonate (TEAB, Sigma-Aldrich, T7408). Protein was reduced with 5.7 mM tris(2-carboxyethyl)phosphine (TCEP, Thermo Fisher Scientific, T2556) and heating to 55 °C for 15 min before alkylation with 22.7 mM methyl methanethiosulfonate (Sigma-Aldrich, 208795) at room temperature for 10 min. Protein was acidified with 6.5 μl of aqueous 27.5% (v/v) phosphoric acid then precipitated with dilution sevenfold into 100 mM TEAB 90% (v/v) methanol. Precipitated protein was captured on an S-trap (Protifi–C02-micro) and washed five times with 165 µl 100 mM TEAB 90% (v/v) methanol before digesting with the addition of 20 µl 0.1 µg µl−1 Promega Trypsin/Lys-C mix (V5071) in aqueous 50 mM TEAB and incubation at 47 °C on a hot plate for 2 h. Peptides were recovered from the S-trap by spinning at 4,000 g for 60 s. S-traps were washed with 40 µl aqueous 0.2% (v/v) formic acid and 40 µl 50% (v/v) acetonitrile:water and the washes were combined with the first peptide elution. Peptide solutions were dried in a vacuum concentrator then resuspended in 20 μl aqueous 0.1% (v/v) formic acid. Peptides were loaded onto EvoTip Pure tips for desalting and as a disposable trap column for nanoUPLC using an EvoSep One system. A pre-set EvoSep 100 SPD gradient (from Evosep One HyStar Driver 2.3.57.0) was used with an 8-cm EvoSep C18 Performance column (8 cm × 150 μm × 1.5 μm). The nanoUPLC system was interfaced to a timsTOF HT mass spectrometer (Bruker) with a CaptiveSpray ionization source (Source). Positive parallel accumulation serial fragmentation-data-dependent acquisition (PASEF-DDA), electrospay ionisation (ESI)-mass spectrometry (MS) and MS2 spectra were acquired using Compass HyStar software (version 6.2, Bruker). Instrument source settings were as follows: capillary voltage, 1,500 V; dry gas, 3 l min−1; and dry temperature, 180 °C. Spectra were acquired between m/z 100 and 1,700. Trapped ion mobility spectrometry settings were as follows: inverse reduced ion mobility rate (1/K0), 0.6–1.60 V s cm−2; ramp time, 100 ms; and ramp rate, 9.42 Hz. Data-dependant acquisition was performed with 10 PASEF ramps and a total cycle time of 1.17 s. An intensity threshold of 2,500 and a target intensity of 20,000 were set with active exclusion applied for 0.4 min after precursor selection. Collision energy was interpolated between 20 eV at 0.6 V s cm−2 and 59 eV at 1.6 V s cm−2.

Liquid chromatography (LC)-MS data, in Bruker.d format, were processed using DIA-NN (1.8.2.27) software and searched against an in silico predicted spectral library, derived from the B. thetaiotaomicron subset of UniProt (UP000001414, 2024/06/10) appended with common proteomic contaminants. Search criteria were set to maintain a false discovery rate of 1%. High-precision quant-UMS72 was used for extraction of quantitative values within DIA-NN. Peptide-centric output in.tsv format, was pivoted to protein-centric summaries using KNIME 5.1.2 and data filtered to require protein q-values < 0.01 and a minimum of two peptides per accepted protein. Calculation of log2 fold difference and differential abundance testing was performed using limma via FragPipe-Analyst73. Sample minimum imputation was applied and the Hochberg and Benjamini approach was used for multiple test correction. Volcano plots were made using the EnhancedVolcano74 package in R.

Quantitative proteomics sample preparation, mass spectrometry, and data analysis

For B. thetaiotaomicron membrane proteomics, three independent 0.5 l cultures of wild type and ΔbamH strains were grown and processed up to the membrane isolation step as described above for BtBAM complex purification. For P. gingivalis whole-cell proteomics, three independent wild-type and ΔbamH P. gingivalis cultures were grown overnight in eTSB. Cells from 50 ml of culture were washed three times with ice-cold PBS. Aliquots corresponding to 50 ml culture of OD600 = 0.6 were centrifuged, and the pellets were stored at −80 °C until needed.

P. gingivalis and B. thetaiotaomicron cell and membrane pellets were lysed by addition of 50 mM TEAB pH 8.5 with 5% (w/v) SDS and sonication using a UP200St ultrasonic processor (Hielscher) at 90 W, 45 s pulse and 15 s rest three times. B. thetaiotaomicron lysis buffer included 1× cOmplete protease inhibitor cocktail (Roche, 11836170001), for P. gingivalis, 2× protease inhibitor cocktail and 100 mM N-ethylmaleimide (NEM, Sigma-Aldrich, 34115-M). Samples were then denatured with 5 mM TCEP at 60 °C for 15 min, alkylated with 30 mM iodoacetamide (B. thetaiotaomicron) or 10 mM NEM (P. gingivalis) at room temperature for 30 min in the dark, and acidified to a final concentration of 2.7% phosphoric acid. Samples were then diluted eightfold with 90% MeOH 10% TEAB (pH 7.2) and added to the S-trap micro columns. The manufacturer-provided protocol was then followed, with a total of five washes in 90% MeOH 10% TEAB (pH 7.2) and trypsin added at a ratio of 1:10 enzyme:protein and digestion performed for 18 h at 37 °C. Peptides were dried using a vacuum concentrator and stored at −80 °C, and, immediately before mass spectrometry, were resuspended in 0.1% formic acid.

LC was performed using an Evosep One system with a 15-cm Aurora Elite C18 column with integrated captivespray emitter (IonOpticks), at 50 °C. Buffer A was 0.1% formic acid in HPLC water; buffer B was 0.1% formic acid in acetonitrile. Immediately before LC–MS, peptides were resuspended in buffer A and a volume of peptides equivalent to 500 ng was loaded onto the LC system-specific C18 EvoTips, according to manufacturer instructions, and subjected to the predefined Whisper100 20 SPD protocol (in which the gradient is 0–35% buffer B, 100 nl min−1, for 58 min; 20 samples per day are permitted using this method). The Evosep One was used in line with a timsToF-HT mass spectrometer (Bruker), operated in diaPASEF mode. Mass and IM ranges were 300–1,200 m/z and 0.6–1.4 1/K0; diaPASEF was performed using variable width IM-m/z windows, as described previously75. TIMS ramp and accumulation times were 100 ms; total cycle time was ~1.8 s. Collision energy was applied in a linear fashion, where ion mobility = 0.6–1.6 1/K0 and collision energy = 20–59 eV.

Raw diaPASEF data files were searched using DIA-NN V 1.8.2 beta 27 (ref. 76), using its in silico-generated spectral library function, based on reference proteome FASTA files for B. thetaiotaomicron (UP000001414, downloaded from UniProt on 1 December 2021) or P. gingivalis (UP000008842, downloaded from UniProt on 13 February 2024) and a common contaminants list77. Trypsin specificity with a maximum of 2 variable modifications and 1 missed cleavage were permitted per peptide, cysteine carbamidomethylation (B. thetaiotaomicron) or NEM (UniMod:108, P. gingivalis) was set as fixed, and oxidation of methionine was set as a variable modification. Peptide length and m/z were 7–30 and 300–1,200; charge states 2–4 were included. Mass accuracy was fixed to 15 ppm for MS1 and MS2. Protein and peptide false discovery rates were both set to 1%. Match between runs and real-time normalization were used for B. thetaiotaomicron but not for P. gingivalis. All other settings were left as defaults. The protein group matrix outputs were processed (separately for each species) in R using Rstudio, with filtering to exclude contaminants, and include a minimum of 2 peptides per protein group, and proteins present in a minimum of 2 out of 3 replicates in any one condition. Statistical analysis was performed using Limma78, with statistical significance inferred by t-test with Benjamini–Hochberg correction for multiple comparisons. Volcano plots were made using the EnhancedVolcano74 package in R.

B. thetaiotaomicron proteinase K shaving assay

Overnight EBHI cultures were used to inoculate fresh minimal medium supplemented with 0.5% fructose, or 0.5% maltose for the control susAhis strain, at OD600 ~0.1. After a 4-h incubation at 37 °C in an anaerobic chamber, 33 OD units of each strain were collected by centrifugation for 5 min at 4,000 g. Cell pellets were resuspended in 1 ml PK buffer (50 mM Tris–HCl pH 8.0, 50 mM NaCl, 1 mM MgCl2), split into four 250-μl aliquots and pelleted again. Two aliquots were resuspended in 180 μl PK buffer and supplemented with 20 μl PK buffer or 20 mg ml−1 fresh proteinase K solution (Sigma-Aldrich, P2308). The other two aliquots were resuspended in 180 μl PK buffer containing 1% Triton X-100 (v/v) (Fisher BioReagents, BP151-100) and supplemented with 20 μl PK buffer or 20 mg ml−1 fresh proteinase K solution. Proteinase K digestion was carried out at 37 °C for 2 h under aerobic conditions. Samples containing Triton X-100 were supplemented with 5 mM phenylmethylsulfonyl fluoride (PMSF) and incubated at 20 °C for 15 min. Samples without detergent were pelleted, washed twice with PK buffer containing 2 mM PMSF and lysed in 200 μl PK buffer with 1% Triton X-100 and 2 mM PMSF for 10 min at 20 °C. Samples were mixed with SDS loading buffer, boiled immediately for 10 min and separated on a 12% FastCast gel (Bio-Rad). Proteins were transferred from the gel to a 0.2-μm polyvinylidene fluoride membrane using a Bio-Rad Trans-Blot Turbo system. The membrane was stained with Ponceau S to confirm successful transfer, blocked for 1 h with 3% BSA solution in PBST (phosphate buffered saline with 0.1% Tween-20) and incubated for 2 h with anti-His-HRP conjugate monoclonal antibody (Roche, 11965085001) diluted 1:500 in blocking solution. The membrane was washed three times with PBST and HRP was detected using chemiluminescence. The membrane was then re-probed with StrepTactin-HRP (Bio-Rad) in blocking solution (1:10,000) for 2 h, and washed and developed as above.

P. gingivalis cell fractionation

P. gingivalis strains were grown overnight in eTSB. All cultures were adjusted to the same OD600 and centrifuged at 7,500 g for 40 min. The supernatant was further ultracentrifuged at 200,000 g for 1 h to separate OMVs from media. OMV-free media were concentrated 20×. OMVs were resuspended in PBS with 1 mM TLCK and sonicated. Cells were washed and resuspended in PBS with 1 mM TLCK, followed by sonication and centrifugation at 7,500 g for 15 min. Supernatant samples were collected and processed as for the soluble fraction. Pellets, corresponding to the insoluble fraction, were washed and resuspended in PBS with 1 mM TLCK.

SDS-PAGE sample preparation and western blotting analysis

Samples of P. gingivalis OMVs, OMV-free media, and soluble and insoluble fractions were mixed with NuPAGE LDS Sample Buffer (Thermo Fisher Scientific, NP0007) and boiled at 95 °C for 5 min. Dithiothreitol (BioSHop, DTT001) was added to 50 mM, and the samples were boiled again at 95 °C for 5 min. Samples were separated in NuPAGE Bis-Tris Mini Protein Gels, 4–12% (Thermo Fisher Scientific, NP0323BOX) and either stained with Coomassie Brilliant Blue G250 (BioShop, CBB555) or electrotransferred onto nitrocellulose membranes (Sigma-Aldrich, GE10600001) in 25 mM Tris, 192 mM glycine and 20% methanol at 100 V for 60 min. To visualize total protein, membranes were stained with 0.1% Ponceau S (BioShop, PON001) in 1% acetic acid and rinsed with distilled water. Membranes were blocked with 5% skim milk (SM, BioShop, SKI400) in TBS with 0.1% Tween-20 (TTBS) at 4 °C overnight followed by 1 h incubation with anti-HmuY antibodies (1 µg ml−1 in TTBS) and 50 min incubation with secondary HRP-conjugated anti-rabbit antibodies (Sigma-Aldrich, A0545) diluted 1:20,000 at room temperature. Signal detection was carried out with Pierce ECL western blotting substrate (Thermo Fisher Scientific, 32109).

Transmission electron microscopy with sectioned B. thetaiotaomicron cells

Cells were cultured to stationary phase (OD600 = 1.5–1.8), collected by centrifugation (5,000 × g, 5 min) and washed twice with 1 mM HEPES pH 7.0, 55 mM sucrose (BioShop, SUC700). Samples were fixed in 1.5% (v/v) glutaraldehyde (Sigma-Aldrich, G6257), 0.5% OsO4 (Sigma-Aldrich, 208868), 0.15% ruthenium red (Sigma-Aldrich, 557450) and 55 mM sucrose in 1 mM HEPES pH 7.0. Next, samples were dehydrated in an ethanol series (50%, 70%, 90%, 96%, 100%, 2 × 15 min each) and in propylene oxide (2 × 10 min) before embedding in a mixture of epoxy resins (Poly/bed 812, MNA, DDSA) (Polysciences, 08791). The following day, the preparations were transferred to DMP-30 resin (Polysciences, 00553) and polymerized in embedding forms at 60 °C. Embedded samples were sectioned with an EM UC7 microtome (Leica). Sections were contrasted on the grids using uranyl acetate and lead citrate. A Tecnai Osiris electron microscope (FEI) operating at an accelerating voltage of 200 kV was used to image the stained sections. Gatan Microscopy Suite (GMS) software v3.4.3 was used to take images on a Rio16 camera (Gatan).

Bioinformatics

To determine the taxonomic distribution of newly discovered BAM components, protein sequences were first subjected to SSDB Motif Search in the Kyoto Encyclopedia of Genes and Genomes database. Then, the conserved motif was used as a query in the Enzyme Function Initiative–Enzyme Similarity Tool79. Default parameters and the UniProt database were used for searches.

Homologues to BtBamG (Bt4367) were identified through complementary BlastP searches performed at the NCBI Blast server against the non-redundant protein sequences (nr)80. BtBamG and selected BamG homologues were used as queries in BlastP unrestricted and taxonomically restricted searches to identify a representative set of BamG homologues across sequence and taxonomic diversity. A selection of BamG homologues were combined to generate protein alignments, optimizing taxonomic and sequence diversity representation for phylogenetic inferences. SeaView v5.0.5 (ref. 81) was used to infer, manipulate and visualize the alignments to generate figures and initial phylogenies. Clustal Omega was used to generate the protein alignments82, and Gblocks83 was used to trim the alignments to select the sites with the strongest hypothesis of homology and to remove excessively divergent regions before phylogenetic inference. The maximum likelihood framework for protein phylogenetics implemented in IQ-TREE was used for tree inferences84,85. Support values for branches were estimated using ultrafast bootstrapping86. The best-fitting single-matrix homogenous substitution model for the alignment was inferred with the automatic model selection (function ‘Auto’), which was LG + G4 + I + F (Optimal pinv = 0.045, alpha = 1.879) for the Akaike information criterion (AIC) and corrected AIC and the Bayesian information criterion.

To investigate the distribution of BamG homologues across Bacteroidota genomes, the annotated proteins from complete genomes from the RefSeq genome database at the NCBI87 were downloaded (1,448 genomes on 17 September 2024). A local BLASTP 2.12.0+ search88,89 was performed using BtBamG as query (B. thetaiotaomicron VPI-5482, accession AAO79472.1 putative outer membrane protein, length = 431 residues). Blast hits with E-value ≤ 0.05 were recorded and are listed in Supplementary Table 2. tBlastn searches using the NCBI Blast portal were performed against individual genomes without strong BlastP hits (Supplementary Discussion).

Reporting summary

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