Genetic, Biochemical, and Molecular Characterization of the Polypeptide Transport-Associated Domain of Escherichia coli BamA

Abstract

The BamA protein of Escherichia coli plays a central role in the assembly of β-barrel outer membrane proteins (OMPs). The C-terminal domain of BamA folds into an integral outer membrane β-barrel, and the N terminus forms a periplasmic polypeptide transport-associated (POTRA) domain for OMP reception and assembly. We show here that BamA misfolding, caused by the deletion of the R44 residue from the α2 helix of the POTRA 1 domain (ΔR44), can be overcome by the insertion of alanine 2 residues upstream or downstream from the ΔR44 site. This highlights the importance of the side chain orientation of the α2 helix residues for normal POTRA 1 activity. The ΔR44-mediated POTRA folding defect and its correction by the insertion of alanine were further demonstrated by using a construct expressing just the soluble POTRA domain. Besides misfolding, the expression of BamA_ΔR44 from a low-copy-number plasmid confers a severe drug hypersensitivity phenotype. A spontaneous drug-resistant revertant of BamA_ΔR44 was found to carry an A18S substitution in the α1 helix of POTRA 1. In the BamA_{ΔR44, A18S} background, OMP biogenesis improved dramatically, and this correlated with improved BamA folding, BamA-SurA interactions, and LptD (lipopolysaccharide transporter) biogenesis. The presence of the A18S substitution in the wild-type BamA protein did not affect the activity of BamA. The discovery of the A18S substitution in the α1 helix of the POTRA 1 domain as a suppressor of the folding defect caused by ΔR44 underscores the importance of the helix 1 and 2 regions in BamA folding.

INTRODUCTION

The β-barrel outer membrane protein (OMP) assembly machine in Gram-negative bacteria, mitochondria, and chloroplasts features a common core protein that belongs to the highly conserved Omp85 protein family (8, 23, 27, 40, 43). In Escherichia coli, the β-barrel assembly machine (BAM) has five components: BamA (Omp85) and four outer membrane lipoproteins unique to bacteria, BamBCDE (22, 37, 46). BamA has two distinct domains: the transmembrane β-barrel and the soluble periplasmic polypeptide transport-associated (POTRA) domain (34). High-resolution structures of the POTRA domain have been solved (6, 7, 17, 18, 47). All POTRA domains have the same basic fold comprising a three-stranded β-sheet and a pair of helices arranged in the order N-β₁α₁α₂β₂β₃-C (17). The number of POTRA domains in Omp85 orthologs ranges from one to seven, with most Gram-negative bacteria having five POTRA domains. In E. coli, all five POTRA domains of BamA play a role in OMP assembly: the POTRA 1 domain is required for normal BamA assembly (1), and single deletion of POTRA 2 to 5 disrupts the interaction of BamA with one or more of the Bam lipoproteins BamBCDE (17).

The OMP assembly pathway begins with the translocation of nascent OMPs through the Sec machinery located in the inner membrane (for a recent review, see reference 13). Once in the periplasm, nascent OMPs interact with periplasmic chaperones, such as SurA. Recently, SurA has been shown to interact directly with the α2 helix of BamA POTRA 1 (1). No data exist showing interactions of BamA with other periplasmic chaperones, including Skp and DegP. Chaperone-bound OMPs are then thought to be offloaded to the BamA POTRA domain by a mechanism known as β-augmentation (17, 19), in which the β-strands of substrate OMPs align with the β-strands of POTRA in a sequence- and orientation-independent manner (29). These initial POTRA-OMP interactions culminate in the completion of OMP β-barrel assembly at the membrane interface. The β-barrel domain of BamA, whose structure and function are currently unknown, is thought to catalyze the final steps of assembly and insertion of new OMP β-barrels into the outer membrane.

The roles of the Bam lipoproteins in β-barrel OMP assembly are unclear at present. High-resolution structures of all four Bam lipoproteins of E. coli have also been solved (14, 16, 15, 18, 25, 35). These structures suggest that some of the lipoproteins may also interact with the β-barrel OMP substrates (14, 15, 35), although experimental data supporting this notion are lacking. Unlike BamBCE, BamD is present in almost all members of Proteobacteria and has been shown to be essential in E. coli (21, 26) and Neisseria meningitidis (42). The essentiality of BamD indicates that it plays a more critical role in OMP biogenesis than the other three Bam lipoproteins. Consistent with this notion, BamD depletion has the most dramatic effect on β-barrel OMPs (21), followed by the absence of BamB (4), while no significant differences in β-barrel OMP levels were observed in the absence of BamC (46; R. Tellez, Jr., and R. Misra, data not shown) or BamE (37, 39). While the loss of BamBCE individually has small to moderate effects on cell viability and OMP biogenesis, the pairwise removal of these three nonessential lipoproteins causes significant growth and OMP biogenesis defects (26, 37, 39). Recently it was shown that the inability of BamA to fold correctly is one of the reasons for the severe growth and OMP phenotypes of a ΔbamB ΔbamE double mutant (39).

When OMPs cannot assemble properly due to alterations in their primary sequence, a defective periplasmic folding environment, or a defective BAM complex, cellular envelope stress responses are triggered (1, 10). These responses, which are controlled primarily by the σ^E and CpxRA pathways (10, 28, 30), lower envelope stress by increasing the synthesis of folding/assembly factors (e.g., Bam proteins, SurA, and Skp), periplasmic proteases (e.g., DegP), and OMP synthesis-inhibitory small regulatory RNAs (e.g., MicA and RybB). The EnvZ/OmpR two-component system has also been implicated in reducing envelope stress caused by aberrant OMP assembly (9). In part, the EnvZ/OmpR pathway is activated by its modulator, MzrA, whose synthesis is controlled by both σ^E and CpxRA (9, 11).

In this paper, we sought to gain a better understanding of the role that BamA POTRA 1 plays in OMP assembly by investigating a BamA POTRA 1 mutant in which a residue in the α2 helix, R44, has been deleted (1). (BamA residue numbers are relative to the mature protein lacking the 20-residue N-terminal signal sequence.) Expression of the mutant BamA protein, BamA_ΔR44, from the chromosome or a low-copy-number plasmid confers a modest or severe OMP biogenesis defect, respectively (1, 10). Reversion and site-directed mutagenesis analyses revealed the molecular basis for the defects of BamA_ΔR44, as well as ways in which they can be rectified. The data revealed the importance of the correct side chain orientation of the α2 helix residues for normal POTRA 1 activity and the involvement of the α1 and α2 helices in POTRA 1 folding.

MATERIALS AND METHODS

Bacterial strains and media.

All bacterial strains were derived from MC4100 (3) and are listed in Table 1. Luria broth (LB) and Luria broth agar (LBA) were prepared as described previously (36). When appropriate, the growth medium was supplemented with l-arabinose (0.1% [wt/vol]), ampicillin (Ap; 25 μg/ml), chloramphenicol (Cm; 12.5 μg/ml), and kanamycin (Kan; 25 μg/ml).

Table 1.

Bacterial strains used in this study

Strain	Characteristicsa	Source or reference
MC4100	F− araD139 Δ(argF-lac)U139 rspL150 relA1 flbB5301 ptsF25 deoC1 thi-1 rbsR	3
RAM1292	MC4100 Δara714	45
RAM1431	MC4100 Δara714 ΔbamA::scar pBAD33-bamA (BamA-WT)	1
RAM1432	MC4100 Δara714 ΔbamA::scar pZS21-bamA6His (BamA-WT)	This study
RAM1433	MC4100 Δara714 ΔbamA::scar pZS21-bamA666His (BamAΔR44)	This study
RAM1438	MC4100 Δara714 ΔbamA::scar pZS21-bamA6His (BamAR44A)	This study
RAM1935	MC4100 Δara714 ΔbamA::scar pZS21-bamA6His (BamAΔR44A)	This study
RAM1936	MC4100 Δara714 ΔbamA::scar pZS21-bamA6His (BamAΔA45)	This study
RAM1937	MC4100 Δara714 ΔbamA::scar pZS21-bamA6His (BamA41A42)	This study
RAM1938	MC4100 Δara714 ΔbamA::scar pZS21-bamA6His (BamAΔR44, 41A42)	This study
RAM1939	MC4100 Δara714 ΔbamA::scar pZS21-bamA6His (BamA46A47)	This study
RAM1940	MC4100 Δara714 ΔbamA::scar pZS21-bamA6His (BamAΔR44, 46A47)	This study
RAM1941	RAM1432 ΔbamB::Cmr pBAD24-bamB6His (Apr)	This study
RAM1942	RAM1433 ΔbamB::Cmr pBAD24-bamB6His (Apr)	This study
RAM1943	RAM1438 ΔbamB::Cmr pBAD24-bamB6His (Apr)	This study
RAM1944	RAM1935 ΔbamB::Cmr pBAD24-bamB6His (Apr)	This study
RAM1945	RAM1936 ΔbamB::Cmr pBAD24-bamB6His (Apr)	This study
RAM1946	RAM1937 ΔbamB::Cmr pBAD24-bamB6His (Apr)	This study
RAM1947	RAM1938 ΔbamB::Cmr pBAD24-bamB6His (Apr)	This study
RAM1948	RAM1939 ΔbamB::Cmr pBAD24-bamB6His (Apr)	This study
RAM1949	RAM1940 ΔbamB::Cmr pBAD24-bamB6His (Apr)	This study
RAM1950	MC4100 Δara714 ΔbamA::scar pZS21-bamA666His (BamAΔR44, A18S)	This study
RAM1951	MC4100 Δara714 ΔbamA::scar pZS21-bamA666His (BamAA18S)	This study
RAM1952	MC4100 Δara714 ΔbamA::scar pZS21-bamA666His (BamAΔR44, A18T)	This study
RAM1953	RAM1950 ΔbamB::Cmr pBAD24-bamB6His	This study
RAM1954	RAM1951 ΔbamB::Cmr pBAD24-bamB6His	This study
RAM1955	RAM1952 ΔbamB::Cmr pBAD24-bamB6His	This study
RAM1956	RAM1292 ΔbamB::scar ΔbamA::Kmr pBAD24-bamB6His (Apr) pBAD33-bamA (BamA-WT) (Cmr)	This study
RAM1957	RAM1292 ΔbamB::scar ΔbamA::Kmr pBAD24-bamB6His (Apr) pBAD33-bamA66 (BamAΔR44) (Cmr)	This study
RAM1958	RAM1292 ΔbamB::scar ΔbamA::Kmr pBAD24-bamB6His (Apr) pBAD33-bamA (BamAA18S) (Cmr)	This study
RAM1959	RAM1292 ΔbamB::scar ΔbamA::Kmr pBAD24-bamB6His (Apr) pBAD33-bamA (BamAΔR44, A18S) (Cmr)	This study
RAM2015	RAM1292 recA::scar	This study
RAM2021	RAM2015/pBAD24-POTRA-WT (Apr)	This study
RAM2022	RAM2015/pBAD24-POTRAΔR44 (Apr)	This study
RAM2023	RAM2015/pBAD24-POTRAA18S (Apr)	This study
RAM2024	RAM2015/pBAD24-POTRAΔR44, A18S (Apr)	This study
RAM2025	RAM2015/pBAD24-POTRA41A42 (Apr)	This study
RAM2026	RAM2015/pBAD24-POTRAΔR44, 41A42 (Apr)	This study
RAM2027	MC4100 Δara714 ΔbamA::scar pZS21-bamA666His (BamAΔA45, A18S)	This study
RAM2041	MC4100 Δara714 rybB::lacZ recA::Cmr	This study
RAM2042	MC4100 Δara714 rybB::lacZ recA::Cmr pBAD24-POTRA-WT (Apr)	This study
RAM2043	MC4100 Δara714 rybB::lacZ recA::Cmr pBAD24-POTRAΔR44 (Apr)	This study

^aWT, wild type.

Antibiotic sensitivity assays.

Sensitivity to antibiotics was analyzed by placing either presoaked rifampin disks (5 μg/ml; Becton Dickinson) or blank paper disks soaked with vancomycin (75 μg/ml) on LBA plates overlaid with 4 ml of soft agar containing 100 μl of overnight-grown bacterial cultures. Plates were incubated for 16 h at 37°C, after which the diameters of inhibition zones were measured. All antibiotic sensitivity assays were performed in triplicate.

Mutant isolation.

Drug-resistant revertants of a strain expressing BamA_ΔR44 from pZS21 were isolated by plating overnight-grown cells on LBA medium containing rifampin and vancomycin, each at 1.25 μg/ml. Drug-resistant colonies arose at a relatively high frequency of 10⁻⁶ to 10⁻⁷. However, mutations responsible for the drug-resistant phenotype rarely moved with the pZS21-bamA_ΔR44 plasmid. Details on identifying isolates in which a plasmid-borne mutation was responsible for the drug-resistant phenotype are provided in Results.

Protein methods.

OMPs from whole-cell protein samples were analyzed by Western blotting, and those from purified envelopes were analyzed by Coomassie blue staining after sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). To obtain whole-cell protein samples, cell pellets were resuspended in SDS sample buffer (5% glycerol, 5% β-mercaptoethanol, 1% SDS, and 62.5 mM Tris-HCl [pH 6.8]) and were heated at 95°C for 5 min. The heat-modifiability test to assess the folding status of BamA was carried out using purified whole-cell envelopes. Envelope samples were solubilized in SDS sample buffer. Prior to SDS-PAGE analysis, SDS buffer-solubilized envelope samples were either heated in a boiling water bath for 5 min or left unheated at room temperature. Cells were fractionated into periplasm, cytoplasm, and envelopes as described previously (11).

For Western blot analysis, proteins were transferred to polyvinylidene difluoride (PVDF) membranes (Immobilon-Millipore). Once the proteins had been transferred, the membrane blots were incubated with appropriate primary antibodies for 1.5 h, followed by incubation with goat anti-rabbit alkaline phosphatase-conjugated IgG secondary antibodies for 1 h. Finally, membrane blots were incubated with the Luminol substrate for 5 min, and protein bands were visualized using a chemiluminescence imager (Bio-Rad). Primary rabbit antibodies against the following proteins (at the dilutions given in parentheses) were used: AcrA (1:16,000), BamA (1:5,000), GroEL (1:50,000), LamB (1:10,000), LptD (1:5,000), MalE (1:10,000), OmpF, OmpC, and OmpA (1:16,000), SurA (1:5,000), TolC (1:5,000), and the N-terminal mature peptide BamA_A21-Q35 (1:2,000) (17). For envelope isolation, bacterial cells were lysed by the French press method (1). Whole-cell envelopes were isolated by centrifuging cell-free lysates for 1 h at 105,000 × g. Envelope pellets were resuspended in SDS sample buffer and were dissolved by boiling. To better resolve OmpC and OmpF bands, 4 M urea was added to the SDS-polyacrylamide running gel. OMPs were visualized after Coomassie blue staining.

Cross-linking and immunoprecipitation.

Cross-linking and immunoprecipitation were carried out as described previously (1, 44). The primary amine-reactive, homobifunctional cross-linker DSP (dithiobis[succinimidylpropionate]) and the primary amine- and sulfhydryl-reactive, heterobifunctional cross-linker SPDP (N-succinimidyl 3-[2-pyridyldithio]-propionate) (Pierce) were used for in vivo protein cross-linking. When full-length BamA protein was used, BugBuster protein extraction reagent (Novagen) was used for cell lysis, and immunoprecipitation was carried out using the IP50 protein G immunoprecipitation kit (Sigma-Aldrich). Immunoprecipitates were solubilized in SDS sample buffer containing β-mercaptoethanol at 95°C for 5 min. Proteins from solubilized immunoprecipitates were separated on 11% SDS-polyacrylamide gels and were visualized using the SilverQuest silver staining kit (Invitrogen) or Western blot analysis. BamB_6His from immunoprecipitates was detected using HisProbe-HRP (1:5,000) as described previously (44). When BamA constructs expressing just the POTRA domain were used for cross-linking, the His-tagged POTRA domain (POTRA_His) was purified from the periplasm by affinity chromatography. POTRA domains were visualized by Coomassie blue staining, while SurA and TolC were detected by Western blot analysis.

DNA methods.

Mutant bamA alleles used in this study were created using the QuikChange Lightning site-directed mutagenesis (SDM) kit (Agilent Technologies) according to the manufacturer's instructions. Mutations were created on the pZS21-bamA_6His, pBAD33-bamA, and pBAD24-POTRA_6His plasmid templates. Mutations in bamA were confirmed by DNA sequencing of the entire gene. Primers used for mutagenesis and sequencing are listed in Table 2.

Table 2.

Primers used for mutagenesis, sequencing, and cloning of bamA

Primer sequence (5′ to 3′)	Alteration(s) or primer nameb
Mutagenesis primersa
GAAGATATCAGTAATACCATTGCTCTGTTTGCTACCGGCAACTTTGAGG	ΔR44
GAAGATATCAGTAATACCATTGCCGCTCTGTTTGCTACCGG	R44A
GAAGATATCAGTAATACCATTGCCGCTCTGTTTGCTACCGG	ΔR44A
GATATCAGTAATACCATTCGCCTGTTTGCTACCGGCAACTTTGAGG	ΔA45
GAAGATATCAGTAATGCCACCATTCGCGCTCTGTTTGCTACCGGCAACTTTGAGG	41A42 insertion
GAAGATATCAGTAATGCCACCATTGCTCTGTTTGCTACCGGCAACTTTGAGG	ΔR44, 41A42 insertion
GAAGATATCAGTAATACCATTCGCGCTCTGGCCTTTGCTACCGGCAACTTTGAGG	46A47 insertion
GAAGATATCAGTAATACCATTGCTCTGGCCTTTGCTACCGGCAACTTTGAGG	ΔR44, 46A47 insertion
GAAGGCCTTCAGCGTGTCTCCGTTGGTGCGGCCCTCC	ΔR44, A18S
GAAGGCCTTCAGCGTGTCTCCGTTGGTGCGGCCCTCC	A18S
GAAGGCCTTCAGCGTGTCACCGTTGGTGCGGCCCTCC	A18T
GAAGGCCTTCAGCGTGTCACCGTTGGTGCGGCCCTCC	ΔR44, A18T
Sequencing primers
CGGGGTCACGACAGCTTTTAC	yaeT2rev
CCATTGCCAGCATTACTTTCTCCG	yaeT1fwd
CCAGGTCAGTCTGACGCCAG	yaeT3fwd
GTTATGGTACAGACGTGACGTTGG	yaeT5fwd
AGATGCCGTTCTACGAGAACTTC	yaeT6fwd
Cloning primers
AGGAGGAATTCACCATGGCGATGAAAAAGTTGC	bamAP1fwdEcoR1
TCGTTCTAGATTAATGGTGATGGTGATGGTGGGTGTTGCGCTCTTTTACC	bamAP56HrevXbaI

^aOnly forward mutagenesis primers are shown.

^bAlterations are shown for mutagenesis primers; primer names are given for sequencing and cloning primers.

RESULTS

Isolation of drug-resistant revertants of BamA_ΔR44.

The BamA_ΔR44 mutant, carrying a deletion of R44 from the BamA POTRA 1 α2 helix, was isolated among bacitracin-resistant revertants of a strain expressing the imp208 allele of the lipopolysaccharide transporter gene lptD (1, 33). LptD⁺ cells expressing BamA_ΔR44 from the chromosome show a modest OMP biogenesis defect (10) and are somewhat sensitive to drugs, such as vancomycin, that are too large to cross the outer membrane permeability barrier of wild-type E. coli cells (1). However, low constitutive expression of BamA_ΔR44 from pZS21 confers a severe biogenesis defect of LptD and other OMPs, as well as an acute drug sensitivity phenotype (1). The hypersensitivity to drugs provided us an opportunity to isolate drug-resistant revertants with the expectation that in some resistant isolates, the restored outer membrane permeability barrier would result from restored OMP biogenesis. Further characterization of these drug-resistant revertants is also expected to reveal the molecular basis of the OMP biogenesis defect caused by BamA_ΔR44 and the means by which cells can overcome this defect. In this study, we sought drug-resistant revertants of a strain expressing BamA_ΔR44 (pZS21; Km^r) and focused on isolates that carry reversion mutations within bamA.

Drug-resistant revertants were obtained on LBA medium containing vancomycin and rifampin, each at 1.25 μg/ml. Plasmids (pZS21-bamA_ΔR44) isolated from five independent pools of drug-resistant revertants were transformed into RAM1431 (ΔbamA pBAD33-bamA; Cm^r), in which BamA expression from a plasmid is under the control of the arabinose-inducible P_BAD promoter (1). In transformed cells, expression of BamA without arabinose was exclusively dependent on pZS21-bamA, thus allowing us to conveniently test the phenotype of pZS21-borne bamA alleles. Once the transformed isolate conferring drug resistance was identified, it was cured of the pBAD33-bamA plasmid by screening for Km^r Cm^s colonies. Only 1 out of 30 transformed colonies from five independent pools displayed drug resistance. After it was cured of pBAD33-bamA, this single isolate could still grow on the original antibiotic selection plate. DNA sequence analysis of the entire bamA gene from the pZS21 plasmid revealed a point mutation resulting in an A18S substitution in the POTRA 1 domain of BamA, in addition to the original ΔR44 alteration. A defect caused by ΔR44, which affects the α2 helix of POTRA 1, and its suppression by A18S, which is located at the start of the α1 helix of POTRA 1, suggest that the two POTRA 1 helices are functionally related (Fig. 1A).

Fig 1.

Locations of BamA POTRA 1 alterations and results of antibiotic disk sensitivity assays. (A) BamA POTRA 1 structure (Protein Data Bank identification code, 2QDF) with its α2 helix oriented in front. The image was drawn using PyMol. The locations of the two helices and three prominent β-strands, as well as those of Q15, R16, A18 and R44, are shown. The Q15 and R16 residues of POTRA 1 may interact with residues of POTRA 2 (17, 19). The BamA numbering is relative to the mature BamA protein lacking the 20-amino-acid N-terminal signal sequence. (B) Antibiotic sensitivities of cultures grown overnight at 30°C were measured. Zones of inhibition from three independent cultures were measured after 16 h of incubation at 30°C. WT, wild type.

As expected, cells expressing BamA_{ΔR44, A18S} display significantly lower sensitivities to vancomycin and rifampin than those expressing BamA_ΔR44 (Fig. 1B). Cells expressing BamA_A18S alone display a level of drug resistance similar to that of cells expressing wild-type BamA (Fig. 1B). We introduced an A18T substitution into the BamA_ΔR44 background via site-directed mutagenesis to determine the specificity of the A18S substitution. The results show that the A18S mutation is better than the A18T mutation at reversing the drug-sensitive phenotype of BamA_ΔR44 (Fig. 1B), thus revealing the stringent side chain specificity at residue 18 of the α1 helix of POTRA 1.

The A18S substitution reverses the OMP biogenesis defect of BamA_ΔR44 and improves its folding.

The biogenesis of β-barrel OMPs, including OmpA, OmpC, OmpF, and LamB, is significantly impaired in cells expressing BamA_ΔR44 (Fig. 2). At 30°C, the A18S and A18T substitutions were equally effective at reversing the OMP biogenesis defect of BamA_ΔR44, but at 37°C, the A18S mutation was better than the A18T mutation at restoring OMP biogenesis (Fig. 2). Note that the strain expressing BamA_ΔR44 does not grow in LB medium at 37°C, presumably due to an acute OMP biogenesis defect (1). Thus, the A18S substitution reverses both the drug sensitivity and OMP biogenesis defect phenotypes of BamA_ΔR44.

Fig 2.

Examination of OMP levels in various BamA backgrounds. OMPs were detected by Western blot analysis in whole-cell lysates prepared from overnight cultures grown in LB medium at 30°C or 37°C. Cells expressing BamA_ΔR44 could not grow in LB medium at 37°C. Each lane contains protein samples from equal numbers of cells (based on the optical density at 600 nm). Membrane blots were probed with antibodies specific to OmpA, OmpC, OmpF, LamB, and GroEL. GroEL served as a gel loading control. WT, wild type.

Reversal of the BamA_ΔR44-mediated drug sensitivity and OMP biogenesis phenotypes by the A18S substitution may be due to improvement of BamA and LptD biogenesis, which is severely impaired in the BamA_ΔR44 mutant (1). The OMP LptD functions as a lipopolysaccharide (LPS) transporter (2). The synthesis and structure of LPS greatly influence OMP biogenesis (20, 31) and the outer membrane permeability barrier (24). As expected, BamA_ΔR44 levels were significantly lower than those of wild-type BamA (Fig. 3A). The presence of A18S in wild-type BamA did not affect its level (Fig. 3A). However, when present in the BamA_ΔR44 background, A18S increased the level of BamA 2-fold over that of BamA_ΔR44 alone (Fig. 3A). Similarly, LptD levels were fully restored by A18S in the BamA_ΔR44 background (Fig. 3B). The improved biogenesis of LptD could explain the reversal of the drug sensitivity phenotype and OMP biogenesis defects of BamA_ΔR44 (Fig. 1 and 2).

Fig 3.

Determination of BamA and LptD levels in different genetic backgrounds. (A) BamA levels in whole-cell lysates prepared from cultures grown overnight in LB medium at 30°C were determined by Western blotting. Two identical membrane blots were probed with HisProbe-HRP and antibodies specific to AcrA; the latter served as a gel loading control. BamA levels were determined relative to AcrA levels and were then normalized to the wild-type (WT) value of 1. (B) LptD levels in either whole-cell lysates or purified envelopes from cultures grown overnight in LB medium at 30°C were determined by Western blotting. Note that anti-LptD antibodies also identify an unknown protein band that migrates just above LptD.

Earlier studies have shown that folded BamA from unheated samples migrates faster than its heat-denatured form on an SDS-polyacrylamide gel (17, 38, 39). Moreover, this heat modifiability characteristic of BamA has been attributed to the BamA β-barrel and not to the POTRA domain (38). We examined the heat modifiability of BamA in strains expressing wild-type BamA, BamA_ΔR44, BamA_A18S, or BamA_{ΔR44, A18S}. No denatured BamA could be detected in the unheated envelope samples of strains expressing wild-type BamA or BamA_A18S, indicating normal BamA folding (Fig. 4). However, for BamA_ΔR44, a significant amount of BamA from unheated envelope samples migrated at the denatured position, indicating a BamA folding defect (Fig. 4). The amount of the denatured form of BamA_ΔR44 was reduced substantially in the presence of A18S (Fig. 4). Because BamA has been proposed to assemble itself, and the POTRA 1 domain has been shown to play an important role in this process (1), the presence of unfolded BamA_ΔR44 in part reflects a self-assembly defect.

Fig 4.

Assessment of the heat modifiability of BamA in different backgrounds. Envelopes obtained from strains grown overnight in LB medium at 30°C were extracted with 0.1 M sodium carbonate at 4°C for 1 h. Sodium carbonate-insoluble membrane pellets were solubilized in an SDS buffer. Prior to SDS-PAGE analysis, SDS-solubilized samples were either left at room temperature (25°C) or heated (100°C). BamA was detected by Western blot analysis using BamA peptide-specific antibodies. BamA-D and BamA-N refer to the denatured and native forms of BamA, respectively. Note that the BamA peptide-specific antibodies are more reactive with BamA variants than with wild-type (WT) BamA.

The A18S substitution improves BamA_ΔR44-SurA interactions.

We have shown previously that BamA_ΔR44 is severely defective in its ability to interact with the major periplasmic chaperone SurA (1). We asked whether the presence of A18S in BamA_ΔR44 improves OMP biogenesis in part due to better BamA_ΔR44-SurA interactions. To evaluate this, we used the method described previously in which BamA was first copurified with His-tagged BamB, followed by detection of SurA that was cross-linked to BamA (1). Almost-identical amounts of BamB_His were purified from each bacterial culture, but the amount of copurified BamA differed owing to the presence of different amounts of BamA in different genetic backgrounds (Fig. 5A). We and others have shown previously that SurA copurifies with the BamB_His-BamA complex only when the proteins are chemically cross-linked in the cell culture prior to affinity purification (1, 37). The data showed that wild-type BamA and BamA_A18S were equally efficient at pulling down SurA (Fig. 5B). However, BamA_ΔR44 showed a 10-fold decrease in its ability to interact with SurA (Fig. 5B). The presence of A18S increased BamA_ΔR44-SurA interaction almost 2-fold (Fig. 5B). Apparently, this level of improvement, coupled with improvements in BamA and LptD biogenesis, is sufficient to restore normal OMP biogenesis, as judged by measurement of steady-state OMP levels (Fig. 2 and 3).

Fig 5.

DSP-mediated cross-linking of various BamA proteins and SurA. ΔbamA ΔbamB (pBAD24-BamB_His) strains expressing wild-type (WT) BamA and BamA variants from pBAD33 were grown in LB medium at 30°C and were cross-linked using DSP as described in Materials and Methods. Eluents from DSP-cross-linked cells (DSP +) and non-cross-linked cells (DSP −) were analyzed by SDS-PAGE, followed by detection of proteins by silver staining (A) or Western blotting (B). Six-His-tagged BamB and SurA were visualized using HisProbe-HRP and anti-SurA antibodies, respectively. SurA/BamA ratios were derived from BamA/BamB (A) and SurA/BamB (B) ratios. Relative SurA/BamA values normalized to the wild-type value of 1 are given in parentheses. Hc, immunoglobulin G heavy chain.

The defect of BamAΔR44 stems from its distorted POTRA 1 α2 helix.

Despite reversing the drug sensitivity and OMP biogenesis phenotypes of BamA_ΔR44, the presence of the A18S substitution in BamA_ΔR44 does not completely restore normal BamA levels. So we set out to determine why the deletion of a single residue, R44, in BamA causes severe biogenesis defects in BamA itself and other OMPs. The R44 residue in the wild-type BamA protein is located in the α2 helix of the POTRA 1 domain (Fig. 1A). Besides shortening the length of the α2 helix by one residue, the absence of R44 most likely also alters the α2 helix rotation and, consequently, the side chain orientation of the neighboring residues. Low BamA_ΔR44 levels suggest that deletion of R44 causes the protein to misfold/misassemble and be proteolyzed (Fig. 3 and 4). We found that the replacement of R44 with alanine in wild-type BamA or the insertion of alanine at position 44 in BamA_ΔR44 does not have a negative effect on BamA levels (Fig. 6A) or OMP biogenesis (Fig. 6C and D), showing that it is the loss of the residue at position 44 rather than the loss of a positively charged side chain at this location that causes the defects associated with BamA_ΔR44 (1).

Fig 6.

Analysis of BamA POTRA 1 α2 helix mutants by SDS-PAGE and Western blotting. The indicated proteins were detected in whole-cell lysates by Western blotting (A, B, and C) or in purified envelopes by Coomassie blue staining (D). Protein samples were prepared from cells grown overnight at 30°C in LB medium. The BamA proteins expressed in various strains are shown at the top. Each lane contains protein samples from equal amounts of cells, based on the optical density at 600 nm. BamA levels were determined relative to the levels of the gel loading control, AcrA, and then normalized to the wild-type (WT) value of 1. The BamA numbering corresponds to the mature BamA protein lacking the 20-amino-acid N-terminal signal sequence. For the locations of BamA α2 residues, see Fig. 1A.

If distorted rotation of the α2 helix causes the defect of BamA_ΔR44, then insertion of a small neutral amino acid residue, such as alanine, upstream or downstream of ΔR44 should restore both the wild-type helix length and rotation and thus should restore the levels and activity of BamA. Also, deletion or insertion of a residue proximal to R44 in the wild-type sequence may reduce the levels and activity of BamA. To test these possibilities, five additional alterations were created in the wild-type or BamA_ΔR44 backbone, and their effects on BamA and OMP levels were examined.

Deletion of the residue A45 drastically reduces BamA levels (Fig. 6A) and confers an OMP biogenesis defect similar to that caused by BamA_ΔR44 (Fig. 6C and D). On the other hand, insertion of alanine between residues 41 and 42 or between residues 46 and 47 in a ΔR44 background increases BamA levels 30- or 6-fold, respectively, over the BamA_ΔR44 level (Fig. 6A) and improves OMP biogenesis (Fig. 6C and D). The presence of the same insertions in the wild-type BamA sequence decreases BamA levels >3-fold from the wild-type BamA level (Fig. 6A). These results are consistent with the assertion that the rotation of the α2 helix and the orientation of its side chains are critical for the folding/assembly and function of BamA.

In vivo characterization of the BamA POTRA domain.

The data presented above indicate that the distorted α2 helix of the POTRA 1 domain is the root cause of the defect in BamA_ΔR44. If this is true, ΔR44 will also cause a folding defect in a BamA POTRA construct devoid of the C-terminal β-barrel domain. To test this, plasmid clones expressing the wild-type and mutant BamA POTRA domains under the control of an arabinose-inducible promoter of pBAD24 were constructed. Like the full-length BamA protein, the recombinant BamA POTRA domain constructs contained a His tag at the N terminus (POTRA_His) to facilitate detection and affinity purification. Cell fractionation studies showed that the mature form of wild-type POTRA_His was localized exclusively in the periplasm (Fig. 7A). In contrast, the precursor form of wild-type POTRA_His, migrating just above the periplasmically localized POTRA_His protein, partitioned exclusively with the membrane fraction (Fig. 7A). Expression of wild-type POTRA_His did not significantly affect bacterial growth, indicating that POTRA_His does not impose ill effects on normal cellular activities, including that of the full-length, chromosomally expressed BamA protein.

Fig 7.

In vivo expression of the BamA POTRA domain and cross-linking with SurA. The BamA POTRA domain, devoid of the C-terminal β-barrel domain, was expressed from the pBAD24 plasmid replicon. (A) Cells expressing the POTRA domain (POTRA_His) were fractionated into whole-cell extracts (WCE), cytoplasm (Cyt), periplasm (Per), and membranes (Mem). Samples from each fraction were analyzed by SDS-PAGE, followed by Western blotting. POTRA, TolC, and SurA were detected by HisProbe-HRP, anti-TolC, and anti-SurA antibodies. TolC and SurA served as controls for the membrane and periplasmic fractions, respectively. The precursor form of POTRA is marked by an asterisk. (B and C) Bacterial cultures expressing either wild-type BamA POTRA_His (B) or its variant with the R44C substitution (C) were subjected to cross-linking with DSP and SPDP. (Top) The POTRA domain was affinity purified from the periplasmic fraction and was visualized by Coomassie blue staining after SDS-PAGE analysis. (Bottom) SurA was detected by Western blot analysis using anti-SurA antibodies.

We proceeded to test whether wild-type POTRA_His can pull down SurA in a manner similar to that of the full-length wild-type BamA protein (1). In addition, a POTRA_{R44C, His} protein was constructed by site-directed mutagenesis so that site-specific, cysteine-mediated cross-linking of POTRA_{R44C, His} with SurA could be performed. We have shown previously that a full-length BamA protein with the R44C substitution can be efficiently cross-linked to SurA using a sulfhydryl-reactive cross-linker (1). Unlike the full-length BamA protein, which contains two native cysteine residues in the β-barrel domain, POTRA_{R44C, His} contains only one cysteine residue introduced by site-directed mutagenesis. After cross-linking, POTRA_His was affinity purified from the periplasm, and peak POTRA_His fractions were tested for the presence of SurA by Western blot analysis. When the amine-reactive cross-linker DSP was used, both wild-type POTRA_His and POTRA_{R44C, His} pulled down SurA efficiently (Fig. 7B and C). In contrast, when the sulfhydryl (cysteine)-reactive cross-linker (SPDP) was used, only the POTRA_{R44C, His} protein was able to pull down SurA (Fig. 7B and C). These results showed that the recombinant wild-type POTRA_His, which localizes in the periplasm, has the ability to interact with SurA independently of the BamA β-barrel domain.

We investigated the effects of ΔR44 in the POTRA_His background. Unlike that of wild-type POTRA_His, expression of POTRA_{ΔR44, His} on a growth medium supplemented with arabinose adversely affected bacterial growth (Fig. 8A) and elevated σ^E-mediated envelope stress (Fig. 8B). These observations pointed to severe misfolding and/or mislocalization of POTRA_{ΔR44, His}. Cell fractionation analysis revealed that in contrast to the wild-type POTRA_His protein, a significant amount of the mature form of POTRA_{ΔR44, His} was present in the membrane fraction (Fig. 8C). It is likely that misfolding of mature POTRA_{ΔR44, His} leads to its aggregation and mislocalization. Because expression of POTRA_{ΔR44, His} severely inhibited cell growth and increased envelope stress due to mislocalization, we did not determine the ability of POTRA_{ΔR44, His} to interact with SurA. Nonetheless, the results of this analysis confirmed that ΔR44-mediated misfolding of the POTRA domain is one of the major reasons why full-length BamA_ΔR44 is severely defective in its own biogenesis and that of other OMPs (1).

Fig 8.

Effects of wild-type POTRA and its variants on cell growth and envelope stress. (A) Cells expressing wild-type (WT) POTRA or different mutant forms were streaked onto unsupplemented LBA plates or LBA plates supplemented with 0.2% arabinose to induce POTRA expression from pBAD24. Growth was recorded after 18 h of incubation at 37°C. (B) RybB::LacZ activities were measured for freshly grown bacterial cultures expressing wild-type POTRA or POTRA_ΔR44. Expression of POTRA from pBAD24 was either induced (+) or not induced (−) with arabinose (Ara). (C) Fractionation of cells expressing wild-type POTRA or POTRA_ΔR44 into soluble fractions (cytoplasm and periplasm) and insoluble membrane fractions. POTRA, TolC, and MalE were detected after Western blot analysis using HisProbe-HRP, anti-TolC, and anti-MalE antibodies. TolC and MalE are controls for the soluble and membrane fractions, respectively.

We asked whether the presence of the A18S substitution or the ₄₁A₄₂ insertion in the POTRA_{ΔR44, His} background would reverse the toxic effect of the mutant protein. Whereas the expression of POTRA_{ΔR44, His} bearing A18S continued to impose cell growth defects (Fig. 8A), the presence of the ₄₁A₄₂ insertion in the POTRA_{ΔR44, His} background restored normal cell growth (Fig. 8A). The observed effects of A18S and ₄₁A₄₂ in the POTRA_{ΔR44, His} background are consistent with their effects in the full-length BamA_ΔR44 background, where ₄₁A₄₂ increases BamA_ΔR44 levels, and hence folding, by 30-fold (Fig. 6A), compared to a 2-fold increase produced by A18S (Fig. 3A).

Functional defects of BamA mutants are revealed in the absence of BamB.

We asked whether some of the BamA mutants discussed above, in which the steady-state OMP levels are similar to those of the strain expressing wild-type BamA, would produce a synthetic phenotype in the absence of BamB and thus reveal a functional defect not apparent by measurement of steady-state OMP levels (Fig. 6D and E). The rationale behind this investigation was based on a previous observation showing that chromosomally expressed BamA_ΔR44 causes only a modest reduction in steady-state OMP levels (10) yet confers a synthetic lethal phenotype in the absence of BamB (1). A plasmid expressing bamB under the control of an arabinose-inducible promoter (pBAD24-bamB; Ap^r) was first transformed into strains expressing BamA, BamA_ΔR44, or their derivatives, from the pZS21 replicon. The chromosomal bamB gene was then replaced with a null ΔbamB::Cm^r allele by P1 transduction. Cm^r transductants, selected and purified on a selection medium containing appropriate antibiotics and arabinose, were tested for growth on a medium lacking arabinose. Only strains expressing wild-type BamA or, to a large extent, BamA_R44A or BamA_ΔR44A, grew normally without arabinose (and hence, without BamB) at all temperatures (Table 3). The remaining strains, expressing other BamA variants, displayed either synthetic lethal or conditional lethal (temperature-sensitive) growth defects (Table 3). These results further underscore the stringent side chain orientation requirement of the POTRA 1 α2 helix residues for BamA activity.

Table 3.

Growth of cells expressing BamA variants in the presence or absence of BamB

BamA protein varianta	Growthb
	30°C		37°C		40°C
	+ Ara	− Ara	+ Ara	− Ara	+ Ara	− Ara
Wild type	+++	+++	+++	+++	+++	+++
ΔR44	++	−/+	++	−/+	−/+	−
R44Ac	+++	++	+++	++	+++	++
ΔR44Ac	+++	++	+++	++	+++	++
ΔA45	+++	−/+	++	−/+	+/−	−/+
41A42 insertion	+++	+	+++	+/−	+++	−/+
ΔR44 with 41A42 insertion	+++	+	+++	+/−	+++	+/−
46A47 insertion	+++	+/−	+++	−/+	+++	−/+
ΔR44 with 46A47 insertion	+++	+	+++	+/−	+++	−/+
ΔR44 A18S	+++	+/−	+++	+/−	ND	ND
A18S	+++	+++	+++	+++	ND	ND

^aThe strain background is MC4100 Δara714 ΔbamA::scar ΔbamB::Cm^r pBAD24-bamB_6His pZS21-bamA. In the ΔR44A variant, R44 was first deleted, and an alanine residue was then inserted at its place.

^bGrowth was rated as follows: +++, large single colonies; ++, medium single colonies; +, small single colonies; +/−, growth in the first two or three streaked areas and formation of heterogeneous colonies; −/+, weak to no growth in the initial streaked area with no single colonies; −, no growth. At each temperature, growth was assessed on LBA plates containing ampicillin supplemented with 0.1% or no l-arabinose (Ara). Expression of BamB is under the control of the arabinose-inducible promoter of pBAD24. BamA is constitutively expressed from the pZS21 plasmid. ND, not determined.

^cThe R44A and ΔR44A BamA variants are identical, but the former was created from a template encoding wild-type BamA and the latter from a template encoding the ΔR44 variant in order to show that the phenotype of BamA_ΔR44 is due to that deletion alone.

DISCUSSION

Our data show that BamA_ΔR44 affects OMP biogenesis through both direct and indirect mechanisms. First and foremost, BamA_ΔR44 is severely impaired in its own folding and assembly, as evidenced by its altered heat modifiability and a dramatic reduction in the BamA protein level. A defect in the self-assembly of BamA_ΔR44 would compromise the activity of the BAM complex and cause a broad OMP biogenesis defect. The observation that a POTRA 1 mutation affects BamA assembly is consistent with our previous data showing that the deletion of POTRA 1 severely compromises the assembly of BamA (1). Although the A18S substitution in the α1 helix only slightly improves BamA_ΔR44 biogenesis compared to the insertion of alanine between residues 41 and 42 of the α2 helix, both alterations restore BamA_ΔR44 activity, which, in turn, improves general OMP assembly. The second direct mechanism entails defective BamA_ΔR44-SurA interactions, which compromise the assembly of all SurA-dependent OMPs, including OmpF, LamB, and LptD (1, 41). However, because BamA assembles independently of SurA (1, 41), a defective BamA_ΔR44-SurA interaction, and hence defective LptD (LPS) biogenesis, would not contribute to aberrant BamA_ΔR44 assembly. The A18S substitution increases BamA_ΔR44-SurA interactions 2-fold, thus possibly improving the SurA-dependent OMP assembly pathway. We think that several engineered alterations, which restored the wild-type helical turn within the α2 helix of POTRA 1, although not analyzed directly, improve OMP assembly by improving the folding of the mutant BamA protein and its interaction with SurA.

Indirect mechanisms by which BamA_ΔR44 likely influences OMP biogenesis involve LPS. We have shown here that the biogenesis of LptD, which codes for the essential LPS transporter (2), is severely compromised in the BamA_ΔR44 background. The low LptD level in the BamA_ΔR44 background is not due to reduced lptD transcription, since lptD mRNA levels increase >2-fold in the mutant BamA background (1). The elevated LptD level observed in the A18S background is expected to restore LPS biogenesis and the ensuing steps of OMP assembly where LPS acts. The second indirect mechanism by which BamA_ΔR44 affects OMP assembly is the activation of the σ^E-mediated envelope stress pathway (1). This will reduce the expression of several OMPs, including OmpF and LamB (1), presumably by activating the expression of small regulatory RNAs, such as RybB and MicA (12). Additionally, elevated expression of the DegP protease in a BamA_ΔR44 background would degrade the pool of misfolded or unassembled OMPs (1). The multiple positive effects of A18S on BamA_ΔR44 folding and assembly, BamA-SurA interactions, and LPS biogenesis would improve OMP biogenesis and reduce the pool of misfolded or unassembled OMPs. This, in turn, would reduce envelope stress, thus obviating the inhibitory effects on OMP synthesis.

The BamA POTRA domain structure has been solved by X-ray crystallography (6, 17, 47) and nuclear magnetic resonance (NMR) (7, 18). The structures obtained by these approaches are very similar to each other; differences are due primarily to the conformational flexibilities between the POTRA domains. In vitro studies have provided evidence for interactions between the β-strands of OMPs and the three β-strands of the POTRA domain (19), presumably via β-augmentation, as proposed by Kim et al. (17). Through cysteine-specific in vivo cross-linking, we have shown that the α2 helix of POTRA 1 interacts with SurA (1). It is possible that SurA, bound to an unassembled nascent OMP, interacts with the α2 helix of POTRA 1 and facilitates interactions between the β-strands of OMPs and the POTRA domain.

The data presented in this study stress the importance of the side chain orientation of the α2 helix residues in BamA and OMP biogenesis. The correct side chain orientation of the α2 helix is not only needed to receive SurA; it may also be necessary, directly or indirectly, for the reception of incoming nascent BamA molecules and the completion of BamA folding/assembly. The latter possibility is based on the observation that BamA mutants with a residue deleted or inserted in the α2 helix of POTRA 1 manifest a severe defect in BamA biogenesis and folding, which is thought to be independent of SurA (1, 41).

It is not known what role, if any, the α1 helix plays in OMP/BamA biogenesis. Our data showing that a defect located in the α2 helix can be partly reversed by the A18S substitution in the α1 helix indicate that this region is also important for POTRA 1 folding. The replacement of a small, nonpolar alanine residue with a small polar serine residue suggests that A18S initiates a novel interaction when R44 is absent from the α2 helix. Moreover, the fact that A18S can reverse the effect of BamA_ΔR44 on OMP biogenesis slightly more than A18T can suggests a high degree of specificity of this novel interaction. It should be noted that A18S can also reverse the drug sensitivity phenotype of BamA_ΔA45 (data not shown), indicating that the positive action of A18S is not strictly site specific. We speculate that the loss of normal helical turn caused by ΔR44 or ΔA45 destabilizes the α2 helix, which broadly impacts the overall folding and assembly of BamA and causes a dramatic reduction in the BamA-SurA interaction. The A18S substitution lowers the global folding/assembly defect of BamA and moderately improves its interaction with SurA. The BamA crystallographic data suggest possible interactions between the POTRA 1 and 2 domains (17, 19). Q15 and R16 are among several residues that could participate in these interdomain interactions. The close proximity of A18S to Q15 and R16 suggests that it could also contribute to interdomain interactions and that such interactions might be particularly beneficial in a ΔR44 or ΔA45 background. The folding defect caused by ΔR44 can be recreated in a construct expressing just the POTRA domain of BamA. Interestingly, the insertion of alanine between residues 41 and 42, but not the A18S substitution, overcomes the toxic effects of POTRA_ΔR44 overproduction.

The growth phenotype of ΔR44 affecting the α2 helix of POTRA 1 is significantly exacerbated in the absence of BamB (1). This is also the case with a strain expressing BamA_ΔR44 bearing the suppressor alteration A18S, which reverses the OMP biogenesis defect of BamA_ΔR44 in a BamB⁺ background. Therefore, while A18S satisfies the drug resistance selection criterion and reverses the OMP biogenesis defect of BamA_ΔR44, it does not fully restore BamA structure and/or function. Similarly, alanine insertions in the α2 helix of POTRA 1, which partially restore BamA_ΔR44/OMP biogenesis, cannot support normal growth without BamB. These results emphasize the importance of BamB, whose dispensability in the wild-type background suggests that it plays a relatively minor role in the biogenesis of E. coli OMPs. In a recent study, we have shown that BamB and BamE play overlapping roles in BamA folding (39). It is possible that when the folding of BamA is compromised due to alterations in its POTRA 1 domain, the absence of BamB alone further exacerbates the folding defects of the mutant BamA protein, thus causing the observed synthetic lethal phenotype. Alternatively, the bamA missense and bamB-null mutations disrupt independent OMP biogenesis steps, and together they result in the synthetic lethal phenotype.

We do not fully understand the reason for the drug sensitivity phenotype of BamA_ΔR44. The fact that large hydrophilic (vancomycin; molecular weight, 1,485) and hydrophobic (rifampin; molecular weight, 823) antibiotics can enter E. coli cells expressing BamA_ΔR44 reflects a significant breach in the outer membrane permeability barrier. Our data suggest that the breach could result from defective LPS and OMP biogenesis. A void created by lower LPS and OMP levels in the outer membrane may be filled by phospholipids, thus creating patches of a phospholipid bilayer through which inhibitors can cross the outer membrane (24). The fact that the A18S variant was isolated among drug-resistant revertants meant that the outer membrane permeability barrier of BamA_{ΔR44, A18S} must be restored close to that of the wild type. The positive correlation between improved outer membrane permeability and improved biogenesis of OMP, and possibly of LPS, shows that isolation of drug-resistant revertants from a strain expressing BamA_ΔR44 is a viable genetic strategy for gaining deeper insight into the functioning of the BAM complex. It is noteworthy that bamB-null mutations were discovered among drug-resistant revertants of strains expressing the missense imp alleles of lptD (5). In addition, similar selections led to the isolation of missense bamA alleles (1, 32). Although at present it is not clear how mutations in bamB and bamA overcome the outer membrane permeability defects of the imp alleles, these mutant bamB and bamA alleles have provided an opportunity to better understand the interplay between the OMP and LPS biogenesis machineries and their roles in establishing an effective outer membrane permeability barrier.