Peptidoglycan maturation controls outer membrane protein assembly

Abstract

Linkages between the outer membrane of Gram-negative bacteria and the peptidoglycan layer are crucial for the maintenance of cellular integrity and enable survival in challenging environments^1-5. The function of the outer membrane is dependent on outer membrane proteins (OMPs), which are inserted into the membrane by the β-barrel assembly machine^6,7 (BAM). Growing Escherichia coli cells segregate old OMPs towards the poles by a process known as binary partitioning, the basis of which is unknown⁸. Here we demonstrate that peptidoglycan underpins the spatiotemporal organization of OMPs. Mature, tetrapeptide-rich peptidoglycan binds to BAM components and suppresses OMP foldase activity. Nascent peptidoglycan, which is enriched in pentapeptides and concentrated at septa⁹, associates with BAM poorly and has little effect on its activity, leading to preferential insertion of OMPs at division sites. The synchronization of OMP biogenesis with cell wall growth results in the binary partitioning of OMPs as cells divide. Our study reveals that Gram-negative bacteria coordinate the assembly of two major cell envelope layers by rendering OMP biogenesis responsive to peptidoglycan maturation, a potential vulnerability that could be exploited in future antibiotic design.

Fig. 1. OMP biogenesis mirrors the cell cycle and not localization of BamA in the outer membrane of live E. coli cells.

a, Biogenesis patterns for FepA, stained with ColB–GFP, following a 5-min induction (0.4% arabinose). Shown are fluorescence heat maps of individual cells representing different cell cycle stages, including a recently divided cell (early), an elongating cell (middle) and a dividing cell (late). b, Demograph showing normalized fluorescence intensity across multiple cells following a 5-min induction of FepA biogenesis. Cells are aligned to show the more intense pole at the top (white asterisk). c, Co-labelling of BamA (with BamA antibody) and FepA (with ColB–mCherry) at different cell cycle stages following 7 min of FepA induction. d, Comparison of FepA biogenesis regions (7 min induction; mean (red line) ± s.d. shaded region) with the distribution of BamA-containing islands (bars) (see also Extended Data Fig. 1), in dividing and non-dividing E. coli cells. Norm., normalized. Scale bars, 1 μm. AU, arbitrary units.

Fig. 2. The cell wall has a pivotal role in OMP biogenesis.

a, Co-labelling of PG (with HADA) and the OMP FepA (with ColB–GFP). PG staining and FepA induction were carried out simultaneously for 7 min. b, Fluorescence intensity profiles (main graph) of PG (HADA) and OMP biogenesis (ColB–GFP) across the septum of dividing cells at two different stages of septum formation (group 1 and group 2) (top left). Top right, the width of individual cells at the division plane. Statistical significance was calculated using two-tailed Student’s unpaired t-test (****P = 0.0001). c, Demographs comparing the normalized fluorescence distribution of FepA and PG biogenesis in multiple cells from experiments performed as in a. d, Interaction of BAM proteins with purified PG analysed by SDS–PAGE and Coomassie blue staining (gel source data is presented in Supplementary Fig. 1). S, supernatant fraction; W, wash fraction; P, pellet fraction. The asterisk indicates the protein size marker lane. e, Scheme of poly-disaccharide-tetrapeptide chains (Tetra_n). GlcNAc, N-acetylglucosamine; m-DAP, meso-diaminopimelic acid; MurNAc, N-acetylmuramic acid. f, Interaction of BamA POTRA domain constructs with Tetra_n shown by MST. Data are mean ± s.d. (n = 3). Conc., concentration; F_norm, normalized fluorescence. g, Apparent (app.) K_d values (mean ± s.d.; n = 3) for the interaction of Bam proteins or sub-complexes with Tetra_n, measured by MST. ND, no interaction detected. h, Interaction of BamA and BamC with PG in E. coli MC1061 cells, detected with antibodies after in vivo cross-linking (blot source data are shown in Supplementary Fig. 1). Pal and CpoB were used as positive and negative controls, respectively. β-mercaptoethanol (β-me) releases cross-linked proteins from PG. l, Schematic depiction of the relative position of BAM (Protein Data Bank ID: 5AYW) in the cell envelope. The approximate width of PG (orange area), outer membrane (grey area) and outer membrane–PG distance were modelled on measurements from Matias et al. (2003). The position of each BAM subunit relative to the PG and how they change during the OMP folding cycle of BAM are not known.

Fig. 3. Nascent (pentapeptide-rich) and mature (tetrapeptide-rich) PG differentially affect BAM activity.

a, Schematic representation of the PG structure in MC1061 and CS703-1, which are enriched in tetrapeptides (tetra-rich) and pentapeptides (penta-rich), respectively. b, Interaction of BamABCDE with tetrapeptide-rich and pentapeptide-rich sacculi in vitro after pull-down by SDS–PAGE and Coomassie blue staining (gel source data are shown in Supplementary Fig. 1). No PG, negative control without PG. c–e, Dose-dependent effects of tetrapeptide-rich (left) and pentapeptide-rich (right) PG on BAM-mediated OmpT assembly in vitro. Data are mean ± s.d. of three independent experiments.

Fig. 4. Cell wall composition affects the integrity and organization of the outer membrane.

a, Co-labelling of PG and the OMP FepA in a pentapeptide-rich strain (CS703-1). PG labelling (HADA) and FepA induction (0.4% arabinose) were initiated simultaneously for a total period of 7 min. Arrows indicate cells showing disparity between OMP and PG biogenesis. b, The fluorescence intensity of FepA versus HADA in tetrapeptide-rich (left) and pentapeptide-rich (right) strains following co-labelling as in a. The representative pixel-by-pixel cytofluorograms of single images show that the strong correlation of OMP and PG biogenesis is abrogated in the pentapeptide-rich strain. c, Polar displacement assay whereby mid-cell OMP biogenesis bias can be discerned from the movement of stationary phase FepA as cells revive. Representative images before and after resuspension in fresh medium (top) and an illustration of OMP movement during this period (bottom). d, Polar displacement of stationary phase FepA during revival. Representative images (top) and demographs of normalized fluorescence intensities across multiple cells (bottom) 45 min after resuspension in fresh M9 medium. The images and demographs show how differently OMPs segregate between tetrapeptide-rich and pentapeptide-rich strains. e, Effect of PG remodelling by plasmid-produced PBP5 on SDS sensitivity in CS703-1 and CS703-1Δlpp. Ectopic PBP5 production from pdacA partially restores outer membrane integrity. f, Model for spatial coupling of PG biosynthesis and BAM activity in the cell. BAM-mediated OMP insertion is dampened by mature PG, resulting in OMP biogenesis being largely absent at the old poles of cells and occurring predominantly at PG growth sites.

Extended Data Fig. 1. Surface labelling of BamA using a high-affinity monoclonal antibody.

(A) Growth curves of E. coli BW25113 cells grown in LB with or without the MAB2 Fabs used for live cell labelling showing the Fabs have no effect on bacterial survival. (B) A field of view showing DIC, epifluorescence and 3D-SIM images of BamA labelled using αBamA^AF488 Fabs. Scale bar, 1 μm. (C) BamA labelling by MAB2 is specific for E. coli BamA. BamA and BtuB, used as a control, labelling of the E. coli BW25113 (E. coli sequence) and a strain expressing a modified BamA β-barrel domain (E. coli/K. pneumoniae sequence). Shown are comparative images of cells labelled with αBamA^AF488 or ColE9^AF488. Scale bars, 1 μm. (D) Size distribution of BamA-containing islands demonstrating that the average island diameter is ~150 nm. (E) The density of BamA-containing islands on the surface of exponentially growing E. coli as measured by 3D-SIM (n = 30 cells). (F) Demograph of the fluorescence intensity across the long axis of multiple cells labelled with αBamA^AF488, emphasising the even distribution of BamA molecules in the OM. (G) Integrated localization data from multiple cells showing BamA islands are distributed through the E. coli OM. The analysis is based on the same cells shown in panel F and represents different stages of the cell cycle.

Extended Data Fig. 2. FepA and BtuB induction systems reveal the patterns of OMP biogenesis.

(A) Epifluorescence images of FepA in an inducible expression strain (GM07). Shown are images of cells after different induction regimes. FepA expression was induced with 0.4% arabinose and the cells stained with ColB-GFP. (B) Images of FepA labelling after 7 min induction with or without chloramphenicol pre-treatment demonstrating that labelling is dependent on new protein synthesis. (C) Epifluorescence images of BtuB stained with ColE9^AF488 under different induction regimes showing labelling is specific for BtuB. Expression was induced with 0.4% arabinose. (D) Timeline of FepA biogenesis in M9 media. Samples were taken and labelled with ColB-GFP at different time points after FepA induction. (E) The fluorescence distribution (± SD) of FepA labelling in dividing vs. non-dividing cells after 7 min induction (n = 25 cells for each condition). (F) A field of view showing grayscale and a corresponding fire heatmap of FepA labelling after 7 min induction. (G) Representative image of BtuB biogenesis after 5 min induction and staining with ColE9^AF488. Shown is a heatmap of a field of view. (H) 3D-SIM projections of BtuB biogenesis after 5 min induction. The images represent different cell cycle stages. (I) The fluorescence distribution (± SD) of BtuB labelling in dividing vs non-dividing cells after 5 min induction (n = 30). Scale bar, 1 μm in all microscopy images shown.

Extended Data Fig. 3. OMP biogenesis is septal biased and cell cycle dependent in native expression systems and division mutants.

(A) Co-labelling of BamA and FepA with αBamA^AF488 and ColB-mCherry, respectively, after 7 min induction of FepA biogenesis. Shown is an overlay of the BamA (green) and FepA (red) channels. (B) Timeline of FepA biogenesis following native chromosomal expression under iron limiting conditions. Samples were taken and labelled with ColB-GFP at different time points after the addition of 2,2 Bipyridyl to LB growth media. (C) Fluoresence intensity of cells shown in panel B. (D) Epifluorescence images of FepA in a ΔminB strain (PB114). Shown are images of cells before and after chromosomal induction with 2,2 Bipyridyl. Scale bars, 1 μm.

Extended Data Fig. 4. The spatiotemporal organization of cell wall biogenesis mirrors OMP biogenesis.

(A) Images of E. coli BW25113 cells after incubation with the PG stain HADA for 3 or 10 min. (B) Demographs of the normalised fluorescence intensity across the long axis of multiple cells after 3 or 10 min incubation with HADA. Cells are aligned so the more intense pole appears at the top. The charts highlight the cell cycle dependent patterns of PG biogenesis. (C) 3D-SIM images of single E. coli cells after incubation with HADA for 10 min. Left to right, cells representing different stages of the cell cycle. (D) The fluorescence intensity of FepA vs HADA (left) or BamA (right) following simultaneous co-labelling. The representative pixel by pixel cytofluorograms of single images show that OMP biogenesis correlates better with PG biogenesis than BAM localization. (E) Jitter plot showing the correlation between FepA biogenesis and either PG biogenesis or BamA distribution. Shown are Pearson coefficients of single images and the average value. Statistical significance was calculated using two-tailed Student’s unpaired t-test (P = 0.0001). (F) Co-labelling of BamA, PG and OMP biogenesis. HADA was added 3 min after FepA induction and the total induction time was 10 min. (G) Co-labelling of PG and OMP biogenesis incorporating a time delay. HADA was added 3 min after FepA induction (stained with ColB-GFP). The total induction time was 10 min. The white/red arrows indicate cells from groups 1 and 2, respectively (see Fig. 2b). (H) Demographs comparing the normalised fluorescence distribution of FepA and PG biogenesis in multiple cells. Cells were treated as in panel G. Scale bars, 1 μm.

Extended Data Fig. 5. PG and OMP biogenesis are also coordinated in other Gram-negative bacterial species.

(A) Timeline of IutA biogenesis in K. pneumoniae. IutA OMP expression was induced by transferring K. pneumoniae cells from overnight LB culture (0 min) into fresh M9. Samples were taken and labelled with CloDF13-AF488 at different time points after IutA induction. (B) Fluoresence intensity of cells shown in panel A. (C) Demograph showing the normalised fluorescence distribution of IutA in multiple cells. Cells were treated as in panel A. (D) Co-labelling of PG and OMP biogenesis 80 min after lutA induction as described in panel A. HADA was added 10 min before sample collection. (E) Pixel by pixel cytofluorogram of a single field of view after co-labelling as in panel D emphasising the correlation of PG and OMP fluorescence. (F) Comparison of the fluorescence intensity profiles of PG and OMP biogenesis in dividing K. pneumoniae cells. Shown are profiles of cells at two different stages of septum formation (Group 1/Group 2). Inset displays the width of individual cells at the designated division plane. n_early = 40 cells, n_late = 20 cells. Statistical significance was calculated using two-tailed Student’s unpaired t-test (P = 0.0001). In the following panels polar displacement of OMPs and PG was used as a measure of their spatially coordinated insertion (see Fig. 4c & Methods). (G) Time lapse images of PG and OMP polar displacement emphasising coordination of these layers during growth of K. pneumoniae. Cells from an overnight culture were labelled and grown on M9 agar pads. Images were taken at the indicated time points. (H) Time course images of PG and OMP polar displacement during P. aeruginosa growth. Cells from an overnight M9 culture were labelled with HADA and resuspended in fresh LB+FeCl₃ to suppress expression of the OMPs FpvAI and FptA. Samples were taken at the indicated time points and each OMP labelled with PyoS2-mCherry and PyoS5-AF488, respectively. (I) Pixel by pixel cytofluorogram of a single field of view after co-labelling as in panel H. Charts demonstrate the co-localisation of both OMPs with one another (top) and with the old PG (bottom). Scale bar, 1 μm in all microscopy images shown.

Extended Data Fig. 6. MST experiments for BAM proteins in the presence of Tetra_n.

(A) Generation of Tetra_n by enzymatic digestion of sacculi from BW25113Δ6LDT. The DD-endopeptidase MepM cleaves the DD-cross-links, generating soluble Tetra_n chains of variable length. (B) HPLC analysis of sacculi from BW25113Δ6LDT (top) and of Tetra_n (bottom), after digestion with the muramidase cellosyl to produce the disaccharide peptide subunits (muropeptides). Schematic representations of the chemical structures of muropeptides are shown on the right side. (C) MST experiment for BamA P1,2 and Tetra_n. Fig. 2f shows that BamA P1,2 does not interact with Tetra_n. Here, the right graph shows the lack of MST response in the control sample (mock PG digest without Tetra_n). (D) MST experiment as in C but with BamA P3,4. BamA P3,4 interacts with Tetra_n (Fig. 2f). (E) MST experiment as in C but with BamA P4,5. BamA P4,5 interacts with Tetra_n (Fig. 2f). Further MST experiments were performed in the presence or absence of Tetra_n or mock PG digests for BamB (F), BamC (G), BamE (H), BamCD (I) and BamCDE (J). (K) Control experiments performed with fluorescent-labelled BSA or BamCD in the presence of Tetra_n, showing that proteins do not generally bind to Tetra_n in MST experiments. (L) Control experiments performed with free fluorescent Red-NHS dye and no protein in the presence of Tetra_n, showing that the fluorescent dye does not bind to Tetra_n. Values shown in panels C-L are mean ± SD of three independent experiments. For source data see Supplementary Table 6.

Extended Data Fig. 7. Controls for PG pull-down experiments and BAM activity assays performed with tetrapeptide-rich and pentapeptide-rich PG.

(A) HPLC muropeptide analysis of tetrapeptide-rich PG from MC1061 and pentapeptide-rich PG from CS703-1. Schematic representation of the main muropeptides are shown. (B) PG pull-down assays for BAM proteins performed in the presence of tetrapeptide-rich PG (MC1061) or pentapeptide-rich PG (CS703-1), analysed by SDS-PAGE and Coomassie Blue staining (for gel source data, see SI Fig. 1). S, supernatant fraction; W, wash fraction; P, pellet fraction. Marker lanes are indicated by asterisks. Representative results from two independent replicates are shown. (C) BAM-mediated OmpT assembly in vitro. Control experiments without BAM (empty liposomes), OmpT or SurA are included. Values are mean ± SD of three independent experiments. (D) Effect of tetrapeptide-rich PG on OmpT activity. Sacculi did not reduce the cleavage of the fluorogenic peptide when added to reactions containing already folded OmpT (n = 3, P = 0.69). (E) Interaction of SurA with tetrapeptide-rich (MC1061) and pentapeptide-rich PG (CS703-1) in vitro, analysed by SDS-PAGE and Coomassie Blue staining (for gel source data, see SI Fig. 1). S, supernatant fraction; W, wash fraction; P, pellet fraction. (F) BAM activity measured in the presence of a 15 µM excess of SurA and tetrapeptide-rich PG, showing that SurA binding to PG is not a limiting factor for BAM activity in vitro (n = 3; control vs sample containing excess SurA: P = 0.21; control vs sample containing PG: P = 0.001; control vs sample containing PG + excess SurA: P = 0.0002; sample containing PG vs sample containing PG + excess SurA: P = 0.89. For gel source data, see SI Fig. 1). (G) Effect of Tetra_n on BAM-mediated OmpT assembly in vitro (n = 3, P = 0.02). Statistical significance of data shown in panels D, F and G was calculated using two-tailed Student’s unpaired t-test. Values are mean ± SD.

Extended Data Fig. 8. The effect of cell wall targeting antibiotics on OMP biogenesis.

(A) Epifluorescence images of FepA in an inducible expression strain (GM07) treated with mecillinam. Shown are images of cells before and after induction with 0.4% arabinose. (B) Co-labelling of PG and OMP biogenesis in mecillinam treated cells. HADA and 0.4% arabinose were added simultaneously for 7 min. Shown are epifluoresence images (left) and pixel by pixel cytofluorogram of a single image (right). (C) Co-labelling of BamA and OMP biogenesis in aztreonam treated cells. Aztreonam was added 30 min before induction with 0.4% arabinose for an additional 7 min. (D) Co-labelling of PG and OMP biogenesis in aztreonam treated cells. HADA and 0.4% arabinose were added simultaneously for 7 min. Shown are epifluoresence images (left) and pixel by pixel cytofluorogram of a single image (right). (E) OMP biogenesis following the removal of aztreonam. Aztreonam was washed from the media at T = 0 min. OMP biogenesis was induced with 0.4% arabinose for 7 min prior to each time point. (F) Co-labelling of PG and OMP biogenesis 45 min after the removal of aztreonam from the media. HADA and 0.4% arabinose were added simultaneously for 7 min. (G) Comparison of the fluorescence intensity profiles of PG and OMP biogenesis across emerging septa as shown in panel F. Shown are profiles of cells at two different stages of septum formation (Group 1/Group 2). Inset displays the width of individual cells at the designated division plane. The chart demonstrates that PG biogenesis precedes the emergence of OMPs at division sites upon aztreonam removal. n_early = 50 cells, n_late = 30 cells. Statistical significance was calculated using two-tailed Student’s unpaired t-test (P = 0.0001). Scale bars, 1 μm.

Extended Data Fig. 9. Increased levels of pentapeptide-rich PG diminishes midcell bias of OMP biogenesis.

(A) Epifluorescence images of FepA, labelled with ColB-GFP in the indicated strains with or without induction of FepA expression. (B) Epifluorescence and 3D-SIM images of newly synthesised FepA in the indicated strains after 7 min induction (0.4% arabinose). Red arrows indicate cells exhibiting atypical biogenesis patterns compared to the CS109 parent strain. (C) Distribution of FepA localisation along normalised cell lengths in the strains from panel B (± SD, 3 biological replicates). (D) Normalised fluorescence intensity across the long-axis of multiple cells. Shown are demographs of FepA labelling 7 min after induction of the indicated strains. These charts and the images in panel B demonstrate the incoherent and reduced midcell bias for OMP biogenesis in the pentapeptide-rich strain. (E) Co-labelling of BamA, PG and OMP biogenesis in a penta-rich strain (CS703). HADA was added 3 min after FepA induction and the total induction time was 10 min. Arrows indicate septa where no increased OMP biogenesis was observed. (F) Epifluorescence (top) and 3D-SIM (bottom) images of the indicated strains labelled with αBamA^AF488 Fabs. The images highlight that BamA organization is similar despite the changes in PG composition. (G) Comparison of BamA distribution in the strains from panel F. Shown are integrated localization maps (top) and charts showing the distribution of islands along the long axis of the cells (bottom). The charts show that BamA is uniformly distributed in all strains. Scale bars, 1 μm.

Extended Data Fig. 10. Complementation experiments by ectopic PBP5 production from pdacA.

(A) Native FepA localisation before resuspension in fresh media (stationary phase). Shown are representative images of the indicated strains (top), demographs of the normalised fluorescence intensity across the long axis of multiple cells (middle) and distribution of native FepA before vs after revival (bottom). The charts and images demonstrate that FepA distribution is almost uniform at stationary phase in all strains but binary partitioning is diminished in penta-rich strains. (B) Distribution of FepA localisation along normalised cell lengths in the strains shown in Fig 4d. (C) Ectopic PBP5 production from pdacA in CS703-1 detected by Western Blot and specific antibodies (for blot source data, see SI Fig. 1). The marker lane is indicated by an asterisk. (D) HPLC analysis of cellosyl-digested PG sacculi shows that the CS703-1 cells expressing PBP5 restore a tetrapeptide-rich PG as is present in the CS109 strain. (E) Effect of PBP5-remodelled PG isolated from CS703-1 (with or without expression of PBP5) on BAM-mediated OmpT assembly in vitro (n = 3; control vs PG from CS109: P = 0.001; control vs PG from CS703-1: P = 0.34; control vs PG from CS703-1 pdacA: P = 0.003). Statistical significance was calculated using two-tailed Student’s unpaired t-test. Values are mean ± SD. (F) Western Blot analysis confirming the absence of Lpp in CS703-1Δlpp (for blot source data, see SI Fig. 1). The marker lane is indicated by an asterisk. (G) Polar migration of pre-existing FepA during revival from stationary phase in the Δlpp background. Shown are representative images (top) and demographs of the normalised fluorescence intensity across multiple cells (bottom) 45 min after resuspension in fresh M9 media. (H) Distribution of FepA localisation along normalised cell lengths in the strains from A. The charts demonstrate that dacA expression restores FepA polar migration in the pentapeptide-rich strain in Δlpp background.