Peptidoglycan maturation controls outer membrane protein assembly

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 environments1–5. The function of the outer membrane is dependent on outer membrane proteins (OMPs), which are inserted into the membrane by the β-barrel assembly machine6,7 (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.

The insertion of outer membrane β-barrel proteins into the membrane of Gram-negative bacteria is catalysed by the BAM complex 6,7 , comprising the β-barrel protein BamA and four accessory lipoproteins, BamB, BamC, BamD and BamE [10][11][12] . BamA is essential for viability, which, in conjunction with its surface exposure, makes it a promising target for Gram-negative-specific antibiotics [13][14][15][16] . Much is known of the mechanism by which BamA catalyses the folding of OMPs in vitro [17][18][19][20] . By contrast, little is known about BamA-mediated OMP biogenesis in the asymmetric outer membrane of live bacteria. Some studies suggest that new OMPs are inserted preferentially in the mid-cell region 8,21,22 , whereas others suggest that OMPs are inserted throughout the membrane 23,24 . These contrasting views are confounded by recent super-resolution microscopy studies in fixed cells that show BamA clustered throughout the outer membrane 25 .

BAM activity is linked to the cell cycle
To determine where OMPs emerge and whether their insertion in the outer membrane is regulated, we first determined the localization of BamA in live E. coli cells. We visualized BamA using epifluorescence microscopy and 3D structured illumination microscopy (SIM) following labelling with a specific, high-affinity monoclonal Fab antibody 13 (MAB2) that binds extracellular loop 6 in BamA with no impact on growth (Extended Data Fig. 1a-c). We found that BamA clusters into small (average diameter approximately 150 nm), uniformly distributed islands (8-10 per μm 2 ) on the cell surface (Extended Data Fig. 1d-g and Supplementary Video 1), in agreement with data from fixed cells 25 . In contrast to another recent study on BamA localization, in which extensive cell permeabilization of fixed E. coli cells was needed for detection of BamA 26 , we saw no enrichment of BamA at division sites. We next investigated where newly synthesized OMPs appear on the cell surface in relation to this distribution of BamA. We focused on two TonB-dependent transporters (TBDTs), the siderophore transporter FepA and the vitamin B 12 transporter BtuB. Both TBDTs were labelled with high-affinity, fluorescently labelled colicins that exploit these OMPs as receptors 8,27 . ColB was fused to mCherry or GFP to locate FepA and ColE9 was labelled with Alexa Fluor 488 (AF488) to locate BtuB. Both OMPs were expressed separately in E. coli from plasmids and induced with arabinose (Extended Data Fig. 2a-c). Time-course studies on FepA indicated that the OMP appeared on the bacterial surface approximately 3 min after induction (Extended Data Fig. 2d), and so all subsequent experiments included at least 3 min for induction. Within populations of cells that had been induced to express either FepA or BtuB, two clear zones of OMP biogenesis were observed; sites of cell division were predominant, yielding labelled septa and cells with unipolar labelling. We also observed OMPs on the long axis of cells (Fig. 1a,b and Supplementary Video 2). Little or no OMP biogenesis was observed at old poles (Extended Data Fig. 2e-i), in agreement with previous reports 8, 21 . We next co-labelled BamA and FepA following a brief period of induction of the latter. Co-labelling revealed a clear Article divergence between BamA localization and the sites of OMP insertion. Whereas BamA was distributed uniformly across the entire cell surface, including the poles, OMP biogenesis was confined to the long axis of cells and division sites (Fig. 1c,d and Extended Data Fig. 3a). We observed a similar biogenesis pattern for FepA expressed from its endogenous promoter (Extended Data Fig. 3b,c), demonstrating that the phenomenon is not exclusive to plasmid-based expression of OMP genes. Intense OMP biogenesis at (aberrant) cell division sites near cell poles were also seen in a minicell mutant (∆minB) (Extended Data Fig. 3d), showing that enhanced OMP biogenesis is not limited to the mid-cell position but localizes at active cell division sites. These data demonstrate conclusively that BamA is not equally active in the outer membrane and that its OMP insertion activity is regulated in a cell cycle dependent manner.

OMPs assemble at PG synthesis sites
Peptidoglycan (PG) is made of glycan chains connected by short peptides and forms a thin, net-like layer called sacculus underneath the outer membrane. The patterns of OMP biogenesis are reminiscent of those seen previously for PG in E. coli 28,29 . We therefore investigated the correlation between PG and OMP biogenesis through co-labelling experiments. PG biogenesis was imaged following incorporation of the fluorescent d-amino acid 7-hydroxycoumarincarbonylamino-d-alanine 30 (HADA) (Extended Data Fig. 4a-c and Supplementary Video 3) and OMP biogenesis was imaged by labelling induced FepA with ColB-GFP. Cells were incubated for 7 min with both HADA and arabinose, the latter to induce fepA expression. Cells were then fixed and surface-exposed FepA was labelled with ColB-GFP. Similar patterns of fluorescence labelling were observed, with particularly strong labelling for both PG and OMP at division sites ( Fig. 2a-c). Co-localization analysis of fluorophore pixel intensity across cells also revealed a strong correlation between PG and OMP fluorescence and this correlation was significantly greater than that for BamA and OMP fluorescence (Extended Data Fig. 4d,e), a conclusion that was confirmed by a triple-label experiment (Extended Data Fig. 4f and Supplementary Video 4). Closer inspection of new OMP and PG labelling at division septa revealed that they segregated into two populations; those with both PG and OMP labelling (group 2) and those with only PG labelling (group 1) (Fig. 2b). There were no cells with only OMP labelling at division sites. Fluorescence scans across each cell type and cell population demographs (Fig. 2b,c and Methods) demonstrated that PG biogenesis always appeared more advanced than OMP biogenesis. In addition, septal widths were narrower for group 2 cells compared to group 1 cells (Fig. 2b). Although the interpretation of these differences is complicated by the distinct, multi-step biosynthetic routes that result in label incorporation, they suggest that cell wall biogenesis precedes the emergence of OMPs at division sites. Dual-labelling experiments that included an additional (3 min) pre-induction period for fepA expression before the addition of HADA generated patterns of PG and OMP labelling similar to those without pre-induction (Extended Data Fig. 4e,g,h). Finally, we observed a similar correlation between OMP and PG biogenesis in Klebsiella pneumoniae (Extended Data Fig. 5a-g), including the emergence of PG before OMPs at the septum (Extended Data Fig. 5f). Spatial coordination between the two processes was also apparent in Pseudomonas aeruginosa (Extended Data  that it is conserved for all Gram-negative bacteria. We conclude that the patterns of OMP and PG biogenesis closely mirror each other in exponentially growing cells, suggesting extensive crosstalk between these layers of the bacterial cell envelope.

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.  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 10 (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) 53 . The position of each BAM subunit relative to the PG and how they change during the OMP folding cycle of BAM are not known.

Article
cross-linked to PG isolated from DTSSP-treated cells (Fig. 2h), whereas BamB and BamE were not cross-linked to PG. As expected, the method identified the known PG-binding outer membrane lipoprotein Pal 35 , whereas the lipoprotein CpoB, which does not interact with PG 36 , was not detected. These results indicate that BamA and BamC interact with PG in E. coli cells. BAM proteins are likely to be close to the cell wall on the basis of dimensions of the complex inferred from structural studies (Fig. 2i). However, their relative positions are likely to vary owing to the highly dynamic nature of the OMP folding cycle 37 , which may explain why binding of PG to some BAM components is detected in vitro but not in vivo.

PG differentially affects BAM activity
Sites of PG biosynthesis are characterized by a transient enrichment in pentapeptides, which are present in PG precursors but not in mature PG owing to the action of dd-carboxypeptidases 9 . We therefore tested whether differences in PG composition might account for the cell cycle dependence of OMP biogenesis. Specifically, we tested whether differences in stem peptide composition between mature (tetrapeptide-rich) and nascent (pentapeptide-rich) PG affected the BAM-PG interaction. We purified the complete BAM complex (BamABCDE) and tested its binding to different PG sacculi preparations. We used MC1061 as a source for mature PG and the multiple dd-carboxypeptidase mutant CS703-1 38,39 as a source for pentapeptide-rich PG, mimicking nascent PG ( Fig. 3a and Extended Data Fig. 7a). The BAM complex pulled down with tetrapeptide-rich PG, whereas we observed almost no binding to pentapeptide-rich PG (Fig. 3b). Purified BAM subunits also showed a reduced interaction with pentapeptide-rich PG compared with tetrapeptide-rich PG (Extended Data Fig. 7b). We conclude that BAM proteins have greater affinity for tetrapeptide-rich PG than pentapeptide-rich PG.
We next tested whether the composition of the peptides in PG affected BAM activity in vitro using an established assay that monitors the liposome incorporation of the β-barrel protease OmpT. The protease activity of folded OmpT is detected as fluorescence following cleavage of a self-quenched peptide substrate 40-42 (Extended Data Fig. 7c). Addition of tetrapeptide-rich PG from MC1061 reduced BAM activity, whereas pentapeptide-rich PG from CS703-1 had little effect, consistent with the stronger binding of BAM proteins and the whole complex to PG (Fig. 3c-e). Control experiments indicated that the effects of PG on OMP folding activity were specific to BAM (Extended Data Fig. 7d-f). The effect of PG on BAM activity was dose-dependent, with a half-maximal effective concentration (EC 50 ) of 0.86 mg ml −1 for tetrapeptide-rich PG and more than 5 mg ml −1 for pentapeptide-rich PG (Fig. 3c-e). Finally, we demonstrated that Tetra n had little effect on BAM activity (Extended Data Fig. 7g), suggesting that cross-linked high-molecular weight forms of PG modulate BAM activity. It remains unclear why individual subunits of BAM vary in their ability to bind PG in vitro compared with the intact complex in vivo, which will require further mechanistic investigation. The variable nature of BAM binding to PG may have its origins in the large conformational changes undergone by the assembly during the OMP folding cycle 43 , which probably influences the positions of BAM components relative to the cell wall 37 (Fig. 2i).

PG modulates OMP patterning in vivo
Since mature and nascent PG have differential effects on the OMP folding activity of BAM in vitro, we hypothesized that cell wall composition is the driver of OMP patterning in the outer membrane of E. coli in vivo. We tested this hypothesis using two approaches: antibiotics that disrupt the PG and E. coli mutants with altered PG, in both cases monitoring PG incorporation using HADA labelling and OMP biogenesis Fluorescence (AU ×1,000)

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. using FepA labelling. Mecillinam, which inhibits the action of PBP2 within the elongasome causing cells to round up 44 , had a negligible effect on coordination between PG and OMP biogenesis as division sites were still formed (Extended Data Fig. 8a,b). Aztreonam, which inhibits PBP3-mediated PG cross-linking at the divisome 45 , inhibits formation of division septa. OMP biogenesis was observed across the filaments of aztreonam-treated cells except at the poles (Extended Data Fig. 8c,d and Supplementary Video 6), but this activity was lower than that at the septa of dividing cells, since removal of aztreonam allowed division sites to reform, accompanied by enhanced OMP biogenesis (Extended Data Fig. 8e-g).
In the second approach, we induced OMP (FepA) expression in a pentapeptide-rich mutant strain (CS703-1) and its tetrapeptide-rich parent (CS109). In both strains, FepA appeared on the cell surface shortly after arabinose induction (Extended Data Fig. 9a). However, whereas the distribution of FepA in the tetrapeptide-rich strain was similar to that observed in a wild-type (BW25113) strain, including mid-cell bias, the pentapeptide-rich strain displayed a defective biogenesis pattern with reduced mid-cell bias (Extended Data Fig. 9b-d). Furthermore, we found that OMP and PG biogenesis no longer mirrored one another as co-localization was significantly reduced (Fig. 4a,b and Extended Data Fig. 9e). Given the aberrant nature of OMP insertion in this and other PG mutant strains, we developed a live-cell assay by which we could determine the effectiveness of mid-cell OMP insertion even in highly disrupted cell envelopes. This assay exploited the fact that several OMP genes in E. coli, including fepA, are highly expressed in late-exponential and stationary phase but are suppressed in lag-phase and early exponential cells 46,47 . We used the displacement of FepA, expressed from its endogenous promoter, as cells revived from stationary phase as a proxy for mid-cell OMP insertion (Fig. 4c). We used this assay to assess the effect of PG mutant strains on their ability to drive polar displacement of FepA after first having determined that all strains expressed similar levels of BamA (Extended Data Fig. 9f,g and 10c). We also determined that they displayed uniform (or near uniform) distribution of FepA in stationary phase (Extended Data Fig. 10a). Following revival, FepA distribution was bipolar in the wild type but nearly 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.
uniform in the pentapeptide-rich mutant, indicative of defective OMP insertion at mid-cell in the cell wall mutant ( Fig. 4d and Extended Data Fig. 10b). Decreased polar displacement was also observed when only dacA (which encodes PBP5) was deleted ( Fig. 4d and Extended Data Fig. 10b). This strain (CS12-7 39 ) contains higher than normal levels of pentapeptide PG, but the increase is more moderate than in CS703-1 48 . Notably, restoration of FepA polar displacement was achieved upon plasmid expression of PBP5 in the CS703-1 background ( Fig. 4d and Extended Data Fig. 10b-e). This effect was retained in a derivative of CS703-1 lacking lpp (CS703-1Δlpp; Extended Data Fig. 10f-h), indicating that it is not the lack of outer membrane tethering to the PG that results in changes to OMP biogenesis patterns. Finally, we demonstrated that PBP5 expression in the CS703-1 background was sufficient to restore E. coli outer membrane stability when cells were challenged with the detergent sodium dodecyl sulfate (SDS) (Fig. 4e), demonstrating that the outer membrane instability associated with PG mutants is rescued when the cell wall defect is corrected. Cumulatively, our data show that it is the local density of pentapeptides that underpins the biogenesis patterns of OMPs and-consequently-the stability of the outer membrane in E. coli, since nascent pentapeptides do not dampen BAM OMP insertion activity as do matured tetrapeptides (Fig. 4f). Pentapeptides are absent in the PG of old poles but are abundant at division sites, whereas inhibitory tetrapeptides show the inverse distribution, explaining why little or no OMP biogenesis is seen in old polar regions of E. coli.

Discussion
Our data show that although BamA clusters are uniformly distributed in the outer membrane, their capacity for catalysing OMP insertion is not uniform, but is modulated by the maturation state of the underlying PG, an effect that is probably enhanced by the low mobility of OMPs in the outer membrane 49 . Maximum OMP insertion activity occurs at division sites, owing to the inability of nascent PG to inhibit BAM. By contrast, little or no OMP insertion occurs at cell poles, owing to the shutdown of OMP folding and insertion activity by mature PG. Intermediate levels of OMP insertion are observed on the long axis of cells, reflecting the mixed maturation state of PG as cells elongate (Fig. 4f). How different forms of PG affect BAM activity is not yet understood, but some elements of the crosstalk between these cell envelope layers can be inferred. Tetrapeptide-rich, non-cross-linked PG fragments bind to BAM proteins with micromolar affinity, but do not inhibit OMP foldase activity. Inhibition requires intact sacculi, suggesting that multiple sites within BAM have to be contacted simultaneously for inhibition to occur. This coordinate effect is reversed by the presence of a terminal d-alanine in nascent PG, blocking the ability of PG to associate with BAM. Tetrapeptide-rich PG could suppress BAM activity directly, for example by blocking access of unfolded OMP substrates to the lateral gate of BamA, or indirectly by restricting the conformational transitions of the BAM machinery that enable OMP folding and release into the outer membrane. Recent evidence points to the Sec translocon in the inner membrane, through which all OMPs are secreted, passing unfolded, chaperone-bound OMPs directly to BAM across the periplasm 50,51 . Thus, an additional element to PG-mediated control of BAM beyond that uncovered here could be the efficiency with which unfolded OMP substrates are transferred through the PG layer.
The modulation of OMP biogenesis by PG maturation has two important physiological consequences for the bacterium. First, it synchronizes insertion of OMPs with the growth of the PG. Such synchrony probably includes growth of the outer membrane itself, since the OMP LptD-which deposits lipopolysaccharides into the outer leaflet of the membrane-is a BamA substrate 52 . Second, biasing OMP insertion to septa or mid-cell regions enables the organism to turn over its OMPs simply by division (binary partitioning). As a result, the organism can rapidly alter its OMP composition in response to a changing environment without the need for active protein degradation 49 .

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-04834-7.

Strains, plasmids, oligonucleotides and antibodies used in this study
The bacterial strains used in this study are provided in Supplementary  Table 1. A list of plasmids and oligonucleotides appear in Supplementary Tables 2 and 3, respectively. Antibodies and engineered bacteriocins are listed in Supplementary Table 4.

Construction of pNGH206
For the construction of the pNGH206 plasmid, site-directed mutagenesis was used to introduce a solvent accessible cysteine (K469C) in the cytotoxic domain of a construct in which the N-terminal 62 amino acids of the colicin had been deleted (Δ2-61 ColE9).

Construction of pBAD33-dacA
The dacA gene was amplified by PCR with Q5 Polymerase (NEB), using chromosomal DNA from E. coli BW25113 as template and the oligonucleotides dacA_F and dacA_R (Supplementary Table 3). Plasmid pBAD33 was amplified by PCR with Q5 Polymerase using oligonucleotides pBAD33_FR and pBAD33_RF (Supplementary Table 3). Insert and vector were joined by ligation-independent cloning 54 . Positive clones were selected on LB agar + 25 μg ml −1 chloramphenicol and identified by colony PCR using GoTaq G2 Polymerase (Promega). The correct sequence of the cloned dacA gene was confirmed by double-strand sequencing using specific oligonucleotides pBAD33_seq_F1, dacA_ seq_R1, dacA_seq_F2 and pBAD33_seq_R2 (Supplementary Table 3). All oligonucleotides were obtained from Eurogentec.

Expression and purification of antibodies and bacteriocins
Antibodies and engineered colicins used are listed in Supplementary  Table 4. Anti-BamA MAB2 Fabs: a construct suitable for periplasmic expression of Fab in E. coli and containing a sequence coding for Fab fragments of MAB2 was cloned, transformed into 34B8 E. coli cells and expressed at 30 °C under control of the phoA promoter in CRAP phosphate-limiting autoinduction medium (PMID: 12009210) supplemented with carbenicillin (50 μg ml −1 ). After 24 h, cells were collected and resuspended in PBS supplemented with one complete EDTA-free Protease Inhibitor Cocktail tablet (Roche) per 50 ml of lysis buffer, lysozyme (0.125 mg ml −1 ), and benzonase (0.01 mg ml −1 ). The prepared suspension was microfluidized at 15,000 psi and clarified at 50,000g for 30 min at 4 °C. The supernatant was then resolved on protein G Sepharose beads equilibrated with PBS, using 2 ml packed resin volume per original gram of cell paste. The column was washed extensively with PBS and Fabs were eluted under mildly acidic conditions (0.56% glacial acetic acid pH 3.6). Eluted Fabs were immediately dialysed overnight at 4 °C against buffer containing 500 mM NaCl, 10% glycerol and 100 mM Tris (pH 8.0). Fabs were further purified on an S75 16/60 gel filtration column (GE Healthcare) using PBS (pH 7.2) as the running buffer. MAB2 Fab fragments were labelled by using Alexa Fluor 488 Protein Labeling Kit (ThermoFisher Scientific) following the manufacturer's instructions. The fluorescently labelled Fabs were passed over HiPrep desalting column (GE Healthcare) to remove the excess dye. Peak fractions were collected and concentrated, and the degree of labelling was determined to be 1.42 dye molecules per Fab using liquid chromatography mass spectrometry (LC-MS). Colicin E9-AF488 expression and purification: the expression and purification of this protein has been previously described 8 . Here we used a modified construct with a single cysteine (Δ2-61 ColE9 K469C-Im9 His6 ). Cys469 in the C-terminal DNase domain of these ColE9 constructs was labelled with a threefold excess of Alexa Fluor 488-maleimide (Invitrogen), as previously described 8 . The labelling efficiency (typically 0.8 fluorophores per protein) was estimated spectrophotometrically (V550 spectrophotometer, Jasco). Colicin B-GFP or colicin B-mCherry expression and purification: the expression and purification of these proteins has been previously described 27 . Pyocin S5-AF488 expression and purification: the expression and purification of this protein has previously been described 55 . Pyocin S2-mCherry expression and purification: the expression and purification of this protein has previously been described 56 . Here we fused mCherry in a similar manner. Primers and plasmids used for the expression of this construct appear in Supplementary Tables 2 and 3. CloDF13-AF488 expression and purification: DNA encoding receptor binding domain of cloacin DF13 with a cysteine at its C terminus was cloned C-terminal to the gene for the colicin E9 immunity protein, Im9, in the pQE-2 vector (Qiagen), to give pNGH382. BL21 (DE3) cells transformed with pNGH382 were grown at 37 °C to an optical density at 600 nm (OD 600 ) of 0.8, upon which His6-Im9-CloDF13 301-460 Cys expression was induced through the addition of 1 mM IPTG. Cells were grown for a further 2 h at 37 °C, before being collected by centrifugation. Cells were resuspended in 20 mM Tris-HCl, pH 7.5, 8 mM imidazole, 0.5 M NaCl, 1 mM PMSF before being lysed by sonication. Cell lysate was clarified by centrifugation at 17,500g for 30 min at 4 °C, before passing the supernatant through a 0.45 μm filter and loading onto a 5 ml HisTrap HP column (Cytiva). Bound material was eluted from the column with a 4 to 500 mM imidazole gradient. Fractions containing His6-Im9-CloDF13 301-460 Cys were labelled with a 1.5-fold AF488-maleimide (Invitrogen) as previously described 8 . The labelling efficiency was estimated spectrophotometrically to be 98% (BioSpectrometer, Eppendorf).
Cells containing the appropriate plasmid were grown in LB medium supplemented with 100 μg ml −1 ampicillin for BamB, BamC, BamE and BamA P1,2, and 30 μg ml −1 kanamycin for P3,4 and P4,5, to an OD 600 of 0.4 and protein expression induced by the addition of 1 mM IPTG, at 18 °C overnight. Cultures were collected by centrifugation (6,000g, 15 min), resuspended in 50 mM sodium phosphate pH 7.5, 300 mM NaCl, 10 mM imidazole with EDTA-Free protease inhibitor tablets (Roche) and lysed using an Emulsiflex C3 cell disruptor (Avestin). The lysate was centrifuged at 75,000g for 45 min at 4 °C to pellet insoluble material. The supernatant was filtered through a 0.45 μM filter (Millipore), then purified via immobilized metal affinity chromatography using a 5 ml HisTrap HP column (GE Healthcare) in sodium phosphate buffer pH 7.5 followed by size-exclusion chromatography in 50 mM Sodium phosphate pH 7.5, 300 mM NaCl, using a Superdex 75 26/60 column (GE Healthcare). Fractions were assessed by SDS-PAGE, combined, concentrated using an Amicon Ultra 10 kDa MWCO centrifugal concentrator (Millipore) and stored at 4 °C for immediate use or frozen in liquid nitrogen and stored at −80 °C.

Expression and purification of BamCD and BamCDE
Plasmids expressing BamCD (pSK46) 40 and BamE with a C-terminal 6×His-tag (pBamE-His) 40 were transformed separately into BL21(DE3) cells (New England Biolabs). Cells were grown in LB broth (supplemented with 50 μg ml −1 streptomycin for BamCD and 100 μg ml −1 ampicillin for BamE), at 37 °C, to an OD 600 of 0.4 and expression induced by the addition of 0.5 mM IPTG, overnight, at 18 °C. Cells were collected separately by centrifugation (6,000g, 15 min), resuspended in 20 mM Tris pH 8.0, 150 mM NaCl with EDTA-Free protease inhibitor tablets (Roche) and lysed separately, using an Emulsiflex C3 cell disruptor (Avestin). The lysates were spun separately at 10,000g for 30 min at 4 °C and the supernatant centrifuged at 100,000g for 45 min to collect membranes. The membranes were solubilised (1 ml of buffer for every 40 mg of membrane) with 50 mM Tris pH 8.0, 150 mM NaCl and 1% n-dodecyl-β-d-maltoside (DDM) (Anatrace), combined and then rotated at 4 °C for 2 h. The solubilized membranes were centrifuged at 50,000g for 30 min; the supernatant was filtered through a 0.45 μm filter (Millipore), and then bound to equilibrated Ni-NTA agarose beads (Qiagen) overnight at 4 °C. The beads were washed with 3 column volumes of 50 mM Tris pH 8.0, 150 mM NaCl, 50 mM imidazole, 0.03% DDM and the protein was eluted with 2 column volumes of 50 mM Tris pH 8.0, 150 mM NaCl, 500 mM imidazole, 0.03% DDM. Fractions were assessed by SDS-PAGE and those containing BamCDE were pooled and further purified through size-exclusion chromatography using a Superdex 200 16/600 column (GE Healthcare) in 50 mM Tris pH 8.0, 150 mM NaCl and 0.03% DDM. Fractions were further assessed by SDS-PAGE, combined and stored at 4 °C for immediate use or frozen in liquid nitrogen and stored at −80 °C.
BamCD was co-purified from a modified version of plasmid pSK46 carrying a 6×His-tag at the C terminus of BamC following the same protocol, omitting the steps with plasmid pBamE-His.

Purification of SurA
The protocol was adapted from previous reports 40, 41 . SurA was overproduced in E. coli BL21(DE3) by growing cells in LB (10 g l −1 NaCl) containing 50 μg ml −1 kanamycin up to an OD 600 of ~1.0. The temperature was shifted to 16 °C and 0.1 mM (final concentration) IPTG was added and the cells incubated at 16 °C for ~16-18 h. Cells were collected (6,200g, 4 °C, 15 min), resuspended in buffer A (20 mM Tris/HCl, pH 8.0) and disrupted by sonication. The soluble fraction was incubated with 2 ml per l of culture volume of Ni-NTA agarose beads (Qiagen) and rotated overnight at 4 °C on a tube roller. Beads were washed in buffer B (20 mM Tris/HCl, 50 mM imidazole, pH 8.0) and the protein was eluted in buffer C (20 mM Tris/HCl, 500 mM imidazole, pH 8.0). Eluted fractions were dialysed against buffer D (20 mM Tris/HCl, 10% glycerol) overnight at 4 °C, then concentrated to ~5 ml and applied to a Superdex 75 (16/600) column (GE Healthcare), in filtered and degassed buffer D at 1 ml min −1 . Eluted fractions were analysed by SDS-PAGE to assess protein purity and yield. Fractions containing SurA were combined and concentrated to 250-300 μM in a Vivaspin Turbo 10 kDa centrifugal concentrator (Sartorius), and stored in aliquots at −80 °C.

Purification of MepM
An adapted protocol was used 33 . Soluble MepM carrying a C-terminal 6×His-Tag was overproduced in the cytoplasm of E. coli BL21 (DE3) from plasmid pMN86. Cells were grown in LB (10 g l −1 NaCl) containing 100 μg ml −1 ampicillin at 37 °C with shaking up to OD 600 ~0.6. The culture was shifted to 25 °C and supplemented with 50 μM IPTG after 30 min to induce protein overproduction. Induction was followed for 2 h. Cells were collected by centrifugation (6,200g, 4 °C, 15 min) and resuspended in buffer A (25 mM Tris/HCl, 300 mM NaCl, 10 mM MgCl 2 , 20 mM imidazole, 10% glycerol, pH 7.0). Cells were disrupted by sonication and the soluble fraction applied to a 5 ml HisTrap HP column at 1 ml min −1 . The column was washed with 5 column volumes of buffer A at 1 ml min −1 . MepM was eluted at 1 ml min −1 in buffer B (25 mM Tris/ HCl, 300 mM NaCl, 10 mM MgCl 2 , 400 mM imidazole, 10% glycerol, pH 7.0). Fractions containing MepM were pooled and dialysed overnight at 4 °C against buffer C (25 mM Tris/HCl, 300 mM NaCl, 10 mM MgCl 2 , 10% glycerol, pH 7.0), then concentrated to a volume of ~5 ml and applied to a Superdex 75 (16/600) column (GE Healthcare), in filtered and degassed buffer C at 1 ml min −1 . Eluted fractions were analysed by SDS-PAGE and fractions containing MepM were combined and stored in aliquots at −80 °C.

Isolation of PG sacculi
An adapted protocol was used 59 . Cells were grown in 4 l of LB (10 g l −1 NaCl) at 37 °C with orbital shaking (175 rpm), up to OD 600 ~0.5-0.6. Cultures were incubated on ice for 10 min to stop cell growth. Cells were collected (6,200g, 4 °C, 15 min) and resuspended in 40 ml of ice-cold Milli-Q water. The cell suspension was added drop-wise to 40 ml of boiling 8% SDS and boiled with vigorous stirring for 30 min. After cooling down to room temperature, sacculi were collected by ultracentrifugation (130,000g, 25 °C, 1 h) and washed in Milli-Q water. Ultracentrifugation and washing steps were repeated until samples were SDS-free 59 . Sacculi were resuspended in 9 ml of 10 mM Tris/HCl, 10 mM NaCl, pH 7.0, supplemented with 1.5 mg of α-amylase (Sigma-Aldrich) and incubated at 37 °C for 2 h. Samples were supplemented with 2 mg of Pronase E (Sigma-Aldrich) and incubated at 60 °C for 1 h. Reactions were stopped by adding 4% SDS (1:1 v/v) and boiling at 100 °C for 15 min, then samples washed until SDS-free as before. Purified sacculi were resuspended at ~10 mg ml −1 in 0.02% NaN 3 and stored at 4 °C. PG preparations were quantified by digestion with the muramidase cellosyl followed by the reduction of the resulting muropeptides by sodium borohydride and their separation by HPLC using detection at 205 nm. The total area of the muropeptides was compared with that of a standard sample with a known concentration that was estimated from UV absorbance, as described 59 . When comparing tetrapeptide-and pentapeptide-rich PG in pull-down assays and BAM activity assays, the amounts of the different PG preparations used was adjusted to the same UV absorbance of muropeptides released from the PG.

Preparation of disaccharide-tetrapeptide chains for MST experiments
Isolated sacculi from E. coli BW25113Δ6LDT 30 (~5.6 mg ml −1 ) were incubated with MepM (3 μM) in 25 mM Tris/HCl, 150 mM NaCl, 0.05% Triton X-100, pH 7.5. Negative control samples containing no sacculi (mock digests) for MST experiments were prepared in parallel. Samples were incubated for ~18 h at 37 °C with shaking and boiled at 100 °C for 10 min. The released soluble PG fragments (Tetra n ) were collected by centrifugation (17,000g, room temperature, 15 min) and dialysed against 50 mM sodium phosphate, 150 mM NaCl, pH 7.0 for ~24 h at room temperature in a 3.5 kDa dialysis membrane. Dialysed Tetra n was stored at −20 °C.

HPLC analysis of PG and Tetra n
For composition analysis and quantification, purified PG sacculi or Tetra n (~100 μg) were digested with cellosyl (0.5 μg ml −1 ) for 16-18 h at 37 °C in 20 mM sodium phosphate, pH 4.8 with shaking (1,000 rpm). Digestions were stopped by boiling at 100 °C for 10 min. Muropeptides were collected after centrifugation (15,000g, 15 min), reduced with sodium borohydride as described 59 and analysed by reversed-phase HPLC at 55 °C in a 90-or 180-min linear gradient from 50 mM sodium phosphate, pH 4.31 to 75 mM sodium phosphate, pH 4.95, 15% methanol, on an Agilent 1220 Infinity HPLC system (Agilent). Relative concentrations of muropeptides from different sacculi preparations were estimated by comparing the total peak area from the integration of the UV signal from HPLC chromatograms 59 .

PG pull-down assay
The protocol was adapted from previous reports 31,32 . PG sacculi (~1 mg) were incubated with purified protein (5 μM) in PG binding buffer (50 mM Tris/maleate, 50 mM NaCl, 10 mM MgCl 2 , pH 7.5) in a total volume of 100 μl. Samples were incubated on ice for 30 min, then pelleted by centrifugation (17,000g, room temperature, 10 min) and the supernatant was collected (supernatant fraction, S). PG pellets were washed by resuspending in 200 μl of PG binding buffer and pelleting again (17,000g, room temperature, 10 min) and the supernatant was recovered (wash fraction, W). PG-bound protein was released from sacculi by resuspending the PG pellet in 100 μl of 2% SDS and stirring for 1 h at room temperature. Samples were centrifuged (17,000g, room temperature, 10 min) and the supernatant was collected (pellet fraction, P). Fractions S, W and P were analysed by SDS-PAGE on 15% polyacrylamide gels and proteins visualised by Coomassie Blue staining. Protein retention in fraction P indicated binding of the protein to PG sacculi. For experiments performed with the full-length BamABCDE complex, PG binding buffer was supplemented with 0.05% Triton X-100. PG pull-down experiments with SurA were performed in 20 mM Tris/HCl, pH 6.5.

Microscale thermophoresis
Purified proteins were labelled with Red-NHS (NanoTemper Technologies) in MST labelling buffer (50 mM sodium phosphate, 150 mM NaCl, 10% glycerol, pH 7.0) according to the manufacturer's protocol. The concentration of fluorescently labelled proteins and efficiency of labelling was determined spectrophotometrically. MST experiments were performed as follows: serial dilutions of Tetra n (from ~5.6 mg ml −1 to ~0.2 μg ml −1 , 16 samples in total) or mock digestions were prepared in MST buffer in a total volume of 10 μl, mixed to an equal volume of labelled protein at 100 nM in MST buffer supplemented with 0.1% reduced Triton X-100 (Sigma-Aldrich) in order to obtain a serial dilution of ligand from ~2.8 mg ml −1 to ~0.1 μg ml −1 (~90 μM to ~3 nM, estimating an average molecular weight of 30 kDa for Tetra n ), a final protein concentration of 50 nM, detergent concentration of 0.05% and reaction volume of 20 μl. Samples were incubated for 5 min on ice and 5 min at room temperature and loaded into standard-coated MST capillaries. Measurements were performed in a Monolith NT.115 (NanoTemper Technologies). LED power for each set of experiments was chosen in order to obtain initial fluorescence values between 200 and 2,500 counts for each individual protein during the capillary scan. Thermophoresis was analysed at the steady-state region of each thermogram. Curve fitting for K d measurements was performed by plotting the normalized fluorescence intensity (F norm ) at the steady state for each sample against ligand concentration according to a 1:1 binding model, as an average of three independent experiments. Results were analysed using the MO-Affinity Analysis software (NanoTemper Technologies).
Alternatively, for proteins that exhibited variations in initial fluorescence greater than ±10% of the average fluorescence along the serial dilution prior to the application of the temperature gradient, curve fitting was performed by directly plotting the initial fluorescence of samples against ligand concentration. K d was calculated assuming a 1:1 binding model from an average of three independent experiments. To confirm that initial fluorescence changes were ligand-dependent, SDS-denaturation tests were performed as follows: after preparing the serial dilution as for the main interaction experiments, the initial fluorescence of three capillary samples representative of the bound fraction (capillary 1, 2 and 3) and three capillary samples representative of the unbound fraction (capillary 14, 15 and 16) was first measured to determine the differences between bound and unbound states. Samples were then centrifuged at 15,000g for 10 min, and the supernatant mixed 1:1 with 2× SD-mix (40 mM DTT, 4% SDS) and boiled at 95 °C for 10 min. The initial fluorescence for the chosen samples was then measured again and compared to the initial fluorescence observed before the SDS-denaturation tests. The initial difference was confirmed to be ligand-dependent if initial differences in fluorescence between samples from the bound and unbound fractions disappeared by the SDS-treatment. SDS-denaturation tests (SDS-tests) were performed in triplicate and analysed using the MO-Affinity Analysis software (NanoTemper Technologies).

Reconstitution of BamABCDE into liposomes
The protocol was adapted from previous reports 40,41,60 . E. coli polar lipids (Avanti) were resuspended at 20 mg ml −1 in water, well dispersed by sonication and 200 μl were mixed with 1 ml of freshly purified BAM complex, and incubated on ice for 5 min. The mixture was diluted with 20 ml of 20 mM Tris/HCl, pH 8.0 and incubated on ice for 30 min. Proteoliposomes were pelleted by ultracentrifugation (135,000g, 4 °C, 30 min) and washed in 20 ml of 20 mM Tris/HCl pH 8.0, pelleted again and resuspended in 800 μl of 20 mM Tris/HCl pH 8.0. Efficiency of reconstitution was assessed by SDS-PAGE, analysing the supernatant, wash and pellet fractions. Proteoliposomes prepared by this method contain BAM complexes that are almost exclusively oriented outwards (that is, the periplasmic part of the complex is exposed on the liposome surface 60 ). Aliquots (20 μl) of proteoliposomes were snap-frozen in liquid nitrogen and stored at −80 °C.

In vitro BAM activity assay
Two 25 μl sub-reactions (A and B) were assembled in 20 mM Tris-HCl, pH 6.5 as follows: sub-reaction A contained SurA (140 μM) and OmpT (20 μM); sub-reaction B contained BAM proteoliposomes (2 μM), fluorogenic peptide (Peptide Synthetics) (2 mM). The two sub-reactions were assembled in half-area, black microplates (Corning) and incubated at 30 °C for 5 min, and mixed to start OmpT folding (final concentrations: 1 μM BAM proteoliposomes, 1 mM fluorogenic peptide, 70 μM SurA and 10 μM OmpT in a total volume of 50 μl). When required, PG sacculi or Tetra n prepared in 20 mM Tris-HCl, pH 6.5 were supplemented to sub-reaction B. Fluorescence emission (excitation at 330 nm, emission at 430 nm) upon cleavage of the fluorogenic peptide by folded OmpT was monitored at 30 °C for 1 h 20 min after in a FLUOStar Microplate Reader (BMG Labtech), with readings every 20 s and 5 s orbital shaking prior to each reading. Three independent replicates were analysed for each experiment. Activity rates for each replicate were analysed over the linear range in the fluorescence release curve, averaged and converted into percentage relative to control reactions containing no PG. EC 50 values for tetrapeptide-rich and pentapeptide-rich PG were estimated using the online MyCurveFit tool (https://mycurvefit.com/).

In vitro BAM activity assay with pre-folded OmpT
For experiments in which OmpT was folded prior to fluorescence measurements, reactions were assembled as described and incubated for 2 h 30 min at 30 °C with orbital shaking to allow BAM-mediated OmpT assembly. Samples were then mixed with 50 μl reactions containing either 2 mM fluorogenic peptide or 2 mM fluorogenic peptide and 5 mg/ml sacculi from E. coli MC1061 in 20 mM Tris-HCl, pH 6.5, and OmpT activity was monitored as described.

In vitro BAM activity assay with an excess of SurA
Experiments containing an excess of SurA were performed as follows. SurA (15 μM) was mixed with 1 mg of sacculi from E. coli MC1061 in 20 mM Tris-HCl, pH 6.5, in a total volume of 200 μl. Control samples contained no PG. Samples were incubated on ice for 30 min, then split in half. One half was added to OmpT folding reactions, assembled as described (BAM activity control reactions containing no PG and no extra SurA, or PG only, were included). The other half was used to monitor PG binding of SurA as described. OmpT activity was measured as above.

Statistical analysis of in vitro BAM activity experiments
Statistical significance was calculated using two-tailed Student's unpaired t-test. Differences were considered statistically significant for P < 0.05. Statistical significance was indicated as follows: NS, P > 0.05 (not significant); *P < 0.05; **P ≤ 0.01; ***P ≤ 0.001. Exact P values are indicated in the figure legends.

In vivo cross-linking of Bam proteins to PG
The protocol was adapted from a previous report 34 . In brief, E. coli MC1061 was grown in 50 ml of LB (5 g l −1 NaCl) at 37 °C by orbital shaking up to OD 600 ~0.5. Cells were pelleted (4,500g, room temperature, 10 min), the cell pellet washed with 50 ml of phosphate-buffered saline (PBS) three times and the OD 600 adjusted to 2.0 with PBS. 3,3′-dithiobis (sulfosuccinimidyl propionate) (DTSSP, ThermoFisher) was freshly dissolved in 5 mM sodium citrate, pH 5.0 and added to cells to a final concentration of 0.5 mM. Cells were incubated at room temperature for 10 min. The cross-linking reaction was quenched by adding Tris/ HCl, pH 8.0 to a final concentration of 50 mM, incubating at room temperature for 15 min. Whole cell samples were taken for Western Blot analysis by concentrating 300 μl of cells 3-fold and mixing 1:1 v/v with 2× SDS-PAGE buffer (200 mM Tris/HCl pH 6.8, 4% SDS, 20% glycerol, 0.2% bromophenol blue) with or without 10% β-mercaptoethanol.
The rest of the bacterial suspension was added drop-wise to an equal volume of boiling 8% SDS and boiled with vigorous stirring for 30 min to isolate PG sacculi. After cooling down to room temperature, sacculi were pelleted by ultracentrifugation (130,000g, room temperature, 1 h) and washed twice in 2% SDS, then resuspended in ~100 μl of 2% SDS. To analyse PG-bound proteins, sacculi suspensions were boiled in SDS-PAGE buffer with or without 10% β-mercaptoethanol at 100 °C for 10 min, briefly centrifuged, and supernatants loaded on 15% polyacrylamide gels, together with whole cell samples taken after cross-linking. Proteins were transferred to nitrocellulose membranes for Western blot analysis.
SDS sensitivity assay E. coli strains were grown from a single colony in LB + 25 μg ml −1 chloramphenicol at 37 °C by orbital shaking for ~16-18 h. The OD 600 was adjusted to 2.0 and cells were serially diluted to 10 −7 in growth medium, then plated with a pin replicator on LB agar + 25 μg ml −1 chloramphenicol + 0.2% arabinose, with or without 2% SDS. Plates were incubated at 37 °C and photographed after 24 h of incubation.

Cell preparation for live microscopy
Overnight LB (10 g l −1 tryptone, 10 g l −1 NaCl, 5 g l −1 yeast extract (pH 7.2)), supplemented M9-glucose medium (0.4% (w/v) d-glucose, 2 mM MgSO 4 , 0.1 mM CaCl 2 , 1 mg ml −1 NH 4 Cl, 0.05% (w/v) casamino acids) cultures were grown at 37 °C and diluted 1:100 into fresh medium with appropriate antibiotics. Cultures were grown at 37 °C, unless stated otherwise, to mid-log phase (OD 600 = 0.2-0.7) and cells were centrifuged at 7,000g for 1 min. For translation inhibition experiments, cells were treated with chloramphenicol (30 μg ml −1 ) 30 min before samples were taken. Agar pads were prepared by mixing supplemented M9-glucose medium or PBS with 1% agarose and pouring 150 μl into 1.5 × 1.6 gene frame (Thermo Scientific AB0577) attached to the slide. For pad formation, the gene frame was sealed by a coverslip until agarose solidified. Six microliters of cells were pipetted onto the agar pad, allowed to dry and sealed with a clean coverslip. For the induction of OMPs from a plasmid, 0.4% (w/v) arabinose was added directly into the growing culture 7 min before samples were taken, unless stated otherwise.
For live-cell labelling, an equivalent of 1ml of cells at OD 600 = 0.25 were pelleted by centrifugation (7,000g, 1 min) and the samples were resuspended in supplemented M9-glucose medium containing 200 nM fluorescently labelled MAB2. Labelling was carried out for 20 min at room temperature with mixing by rotary inversion in an opaque tube. Subsequently the cells were washed twice (M9-glucose) by pelleting (7,000g, 1 min) and finally resuspended in ~50 μl M9-glucose. For fixed cell labelling, cells were pelleted by centrifugation (7,000g, 1 min) and the samples were resuspended in 4% formaldehyde (in PBS) at 4 °C immediately after sampling. After 20 min, cells were pelleted by centrifugation (7,000g, 1 min) and resuspended in 100 μl of fresh supplemented PBS containing 200 nM fluorescently labelled Fabs or colicins. Labelling was carried out as with live cells. After labelling, cells were washed 3 times (in PBS) by pelleting (7,000g, 1 min) and finally resuspended in ~50 μl PBS. To improve binding of the MAB2 Fabs in co-labelling experiments, after the labelling step cells were washed once (in PBS), pelleted (7,000g, 1 min) and resuspended in 4% formaldehyde (in PBS) at 4 °C for further 20 min. Cells were then washed twice (PBS) and resuspended in ~50 μl PBS.
The fluorescent d-amino acid HADA (Tocris Bioscience) was used for cell wall labelling. The labelling was carried as described previously 61 with minor adjustments. The final concentration of HADA in the growing culture was 500 μM and the incubation time varied according to the aims of the experiment. When both PG and OMP labelling were included, the HADA labelling protocol was used first. After completing the last step of HADA labelling (fixation), samples were labelled using the relevant colicins as described above. For P. aeruginosa polar displacement experiments, HADA was added to the overnight culture and washed before resuspension in fresh medium.

Induction and suppression of chromosomal OMP expression
For E. coli, chromosomal expression of FepA was induced by the addition of 200 μM 2,2 Bipyridyl (Sigma-Aldrich) to LB medium during mid-log phase. For, K. pneumoniae, chromosomal expression of IutA was induced by growing U11 cells in LB overnight (stationary phase) and transferring them into fresh M9. For P. aeruginosa, suppression of FpvAI and FptA expression for polar displacement experiments was carried out by growing PAO1 cells in M9 overnight (stationary phase) and transferring them into fresh LB medium containing 200 μM FeCl 3 .

Total internal reflection fluorescence microscopy acquisition
Live cells were imaged using an Oxford NanoImager (ONI) super-resolution microscope equipped with four laser lines (405, 473, 561 and 640 nm) and ×100 oil-immersion objective (Olympus 1.49 NA). Fluorescence images were acquired by scanning a 50 μm × 80 μm area with a 473 nm laser for AF488-and GFP-labelled proteins (laser power 1.4-2.3 mW) or 561 nm for mCherry-labelled proteins (laser power 2.1-3.4 mW). The laser was set at 50° incidence angle (200 ms exposition), resulting in a 512 × 1,024 pixel image. Images were recorded by NimOS software associated with the ONI instrument. Each image was acquired as a 20-frame stack for brightfield and fluorescence channels, respectively. For analysis, images were stacked into composite images using average intensity as a projection type in ImageJ (version 1.52p). To ensure non-uniform fluorescence of labelled OMPs was not the result of proximity to the coverslip, equivalent images were taken in epifluorescence and by 3D-SIM microscopy where such potential bias was absent.

Epifluorescence and SIM imaging
Cells were imaged using Deltavision OMX V3 Blaze microscopy system (GE Healthcare) equipped with four laser lines (405, 488, 593 and 633 nm), pco.edge 5.5 sCMOS cameras (PCO), a standard or a green/ red drawer filter set and a ×60 oil-immersion objective (Olympus 1.42 NA). Three-dimensional-SIM three-colour images were taken using Deltavision OMX-SR microscopy system (GE Healthcare) equipped with four laser lines (405, 488, 568 and 640 nm), pco.edge 4.4 sCMOS cameras (PCO) and a ×60 oil-immersion objective (Olympus PlanApo 1.42 NA). For both conventional and SIM imaging 1.512 index refraction immersion oil was used for AF488-and GFP-labelled proteins and for mCherry/GFP/HADA three-colour imaging. For mCherry-labelled proteins or AF488-mCherry dual-colour imaging 1.514 index refraction immersion oil was used. Conventional fluorescence images were acquired by imaging a 42 μm × 42 μm area with the 488 nm laser (5.7 mW, 500 ms exposure) resulting in a 512 × 512 pixel image. For SIM acquisition, a similar area was imaged using the 488 nm laser (2.7 mW, 200 ms exposure). Image stacks of 1-1.5 μm thickness were taken with 0.125 μm z-steps and 15 images (three angles and five phases per angle) per z-section and a 3D structured illumination with stripe separation of 213 nm and 238 nm at 488 nm and 594 nm, respectively. The SIMcheck plugin (ImageJ) was used to assess the data quality of SIM images. Image stacks were reconstructed using Deltavision softWoRx 7.2.0 software with a Wiener filter of 0.003 using wavelength specific experimentally determined OTF functions. Average intensity and 3D projections of 3D-SIM images were generated using ImageJ (V1.52p).
For the acquisition of multi-channel images, a DIC image was taken first followed by an imaging sequence which minimized any possible overlap between channels. Fluorophores with higher excitationemission spectra were imaged first and the fluorescent signal was bleached prior to the acquisition of the following channel. Alignment of dual-colour images was carried out using TetraSpeck Microspheres, 0.1 μm (ThermoFisher scientific) and the channel aligner tool (ImageJ V1.52p).

Image and data analysis
Two-dimensional SIM images of BamA labelled cells were binarized and regions of interest (ROIs) were generated. Non-distinct islands were manually excluded. The size of each island was calculated based on its Ferret's diameter (ImageJ V1.52p). For measurement of septal cell widths, the DIC and epifluorescence images were overlaid and the HADA channel used to determine the location of the developing septum along the long axis of the cell. Measuring the width of the cell at the selected region was done using the DIC channel (ImageJ V1.52p).
For the detection of islands and creating integrated localization maps an Unsharp mask filter (Radius 2 px, Mask weight 0.5) was applied to the raw images and BamA islands identified using the Find Maxima process (Prominence > 600) (ImageJ) and ROI were created. The fluorescence intensity of each ROI was measured using the raw images and background fluorescence was subtracted. Integrated localization maps of BamA islands were created using the ImageJ plugin MicrobeJ (v5.13m) 62 . For the detection of BamA islands with MicrobeJ, Unsharp Mask filter was applied, as described above. Cells and maxima points detection was carried out using the MicrobeJ plugin, In some cases the auto segmentation was manually corrected to exclude improperly detected or clustered cells.
Measurement of individual fluorescent intensity profiles and calculation of the normalized fluorescence distribution of OMPs were carried out as follows: For the early experiments (Fig. 1, and the related supplementary figures) fluorescence distribution across the long axis of the outer membrane was determined by the plot profile function (ImageJ V1.52). After measuring the raw values, they were normalized to 0−1 scale for comparison between cells. To enable the integration of fluorescence intensity distribution from cells of different lengths, the long axis of each cell was normalized to 0−100. In later experiments (Figs. 2-4 and the related extended data figures), the measurement of the fluorescence intensity profile and normalization of the long axis between 0-100 was automated using the MicrobeJ plugin (v5.13m). After the integration of profiles from all cells, the value at the poles was set to 1 and the remaining values normalized accordingly. Normalization was done using Excel and the data was plotted in GraphPad Prism 8 software.
Demographs presenting the fluorescence distribution of OMPs in a large population of cells were created using the MicrobeJ plugin (v5.13m). Each demograph included cells from at least three different fields of view. For all experiments, cells were sorted by length in order to highlight the different cell cycle stages. For induced OMP biogenesis experiments the pole with the higher fluorescence intensity was aligned to the top in order to highlight the unipolar distribution pattern.
For co-localization measurements the two compared channels were overlaid (following channel alignment) using the 'merge channels' tool (ImageJ V1.52) to confirm that cells are properly aligned. Next, the fluorescence image from the 405nm channel was thresholded and a ROI including only cell containing regions was created. The Coloc 2 tool was used to calculate the Pearson correlation coefficient of the ROI between the different fluorescence channels. Cytofluorogram plotting was done using the imageJ plugin JACoP 63 .
The following experiments are representative of four independent biological replicates: Extended Data Figs. 1b, 2f, 3b, 4a and 9b.
Significant P values were validated using a two-tailed Student's t-test or a non-parametric Mann-Whitney test.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability
The data supporting the findings of this study are available within the paper and its Supplementary Information files. All the images displayed in this study, raw MST data and raw BAM activity data are available as source data files accompanying this manuscript. Raw uncropped gel images appear in Supplementary Fig. 1 and full MST controls appear in Supplementary Fig. 2. Materials and reagents are available upon request. 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.