Membrane integration of an essential β-barrel protein prerequires burial of an extracellular loop

Significance

Membrane proteins must be assembled to adopt a mature functional form. Integral membrane proteins with β-barrel structure are found in the outer membranes of Gram-negative bacteria, mitochondria, and chloroplasts. To understand the mechanism by which these proteins are assembled, intermediate states in the process must be accumulated and characterized. Here, we study assembly-defective mutants of the β-barrel protein BamA. Our findings reveal that a highly conserved, hydrophilic loop is buried within the β-barrel of BamA before membrane integration, suggesting that one barrier to assembly is the translocation of hydrophilic sequences across the membrane. The specific interactions we identify in this early assembly step could potentially be used for design of mechanism-based inhibitors with antibiotic activity.

Abstract

The Bam complex assembles β-barrel proteins into the outer membrane (OM) of Gram-negative bacteria. These proteins comprise cylindrical β-sheets with long extracellular loops and create pores to allow passage of nutrients and waste products across the membrane. Despite their functional importance, several questions remain about how these proteins are assembled into the OM after their synthesis in the cytoplasm and secretion across the inner membrane. To understand this process better, we studied the assembly of an essential β-barrel substrate for the Bam complex, BamA. By mutating conserved residues in the β-barrel domain of this protein, we generated three assembly-defective BamA substrates that stall early in the folding process in the periplasm. Two of the three defective substrates, which harbor mutations within β-strands, fail to associate productively with the Bam complex. The third substrate, which harbors mutations in a conserved extracellular loop, accumulates on BamD during assembly, but does not integrate efficiently into the membrane. The assembly of all three substrates can be restored by artificially tethering a region of the substrate, which ultimately becomes an extracellular loop, to the lumen of the forming β-barrel. These results imply that a critical step in the folding process involves the interaction of residues on the interior of the nascent β-barrel wall with residues in one of the extracellular loops. We conclude that a prerequisite for membrane integration of β-barrel proteins is burial of the extracellular loops within the forming β-barrel.

Membrane proteins with β-barrel structure can spontaneously assemble into their folded states in detergent solutions and artificial membrane bilayers (1, 2). Nevertheless, in living systems such as Gram-negative bacteria, chloroplasts, and mitochondria, these proteins are assembled into membranes by conserved machines (3–7). Clearly, there must be kinetic barriers to the rapid assembly of these proteins in cells that must be overcome by catalysis. Numerous in vitro biophysical studies on the uncatalyzed assembly pathway for β-barrel proteins into lipid bilayers have led to the conclusion that the slow step in assembly is membrane integration, and that the step involves concerted formation of tertiary structure (2, 8–10). Two limiting possibilities exist for the folding of membrane β-barrels in vivo. The assembly machines could somehow accelerate the uncatalyzed folding mechanism, or they could provide an alternative mechanism to accelerate assembly. It has been suggested that in vivo folding occurs through iterative insertion of β-strands or β-hairpins into the membrane through the assembly machine, which implies that the catalyzed pathway is different from the uncatalyzed pathway (11–13).

One approach to answer the question of how β-barrels are assembled in vivo is to characterize the machines that assemble them. A considerable amount of biochemical, genetic, and structural work has focused on the role of the components of these machines. These experiments have revealed which components perform essential chemistry (5, 14–16), how the components interact with each other (12, 13, 17), and what conformational states may exist during function (18–21). Largely, these studies have emphasized how the machine might alter the membrane structure or how its components might sequentially bind and thereby guide substrates into the membrane. However, much less has been done to understand how structure emerges within a substrate during assembly by these machines. Here, we have taken the approach of characterizing the nascent structure of a substrate as it interacts with the machine to identify the features of the substrate required to permit its proper membrane integration.

We have been studying β-barrel assembly in Escherichia coli, which is accelerated by the β-barrel assembly machine (Bam) (7, 22). The Bam complex comprises a central β-barrel protein, BamA, and four lipoproteins, BamBCDE (5, 7, 15, 23, 24). Two of the five components, BamA and BamD, are essential for viability and participate directly in β-barrel assembly (5, 7, 15). The Bam complex has broad substrate scope, in that it can accelerate folding of substrates with as few as 8 or as many as 26 β-strands, substrates with large periplasmic or extracellular domains, and even substrates that encapsulate other proteins (25–27). These diverse substrates may demand somewhat different assembly processes, but key aspects of the β-barrel assembly mechanism are likely shared among all substrates. In an effort to identify common mechanistic features, we have focused on the two essential substrates folded by the Bam complex: LptD, which facilitates lipopolysaccharide transport to the outer membrane (OM), and BamA itself. LptD is a 26-stranded β-barrel, which encapsulates the lipoprotein LptE within the lumen. Here, we study assembly of the smaller 16-stranded β-barrel BamA, which does not encapsulate another protein. Because BamA is both a substrate for the Bam complex and a component of it, it can be challenging to separate those features of its structure that are important for its folding from those that are important for its function once assembled—in fact, some features of the BamA protein likely have dual roles. The experiments we describe here have been designed to report on how the sequence and structure of the BamA substrate affect its own assembly.

One feature of β-barrel substrates that is necessary for folding is a conserved C-terminal β-signal motif that must be recognized by the Bam complex in order for them to be recruited to the OM and assembled (28, 29). In bacteria, this motif mediates binding of unfolded substrates to BamD (30–33). The β-signal of E. coli BamA is slightly removed from the absolute C terminus of the protein; it spans residues 769–776, which are located in what ultimately becomes β-strand 14 of the folded structure (32). Amino acid substitutions at several positions within the β-signal were shown to decrease folding efficiency of BamA substrates in vivo. Most of these BamA variants were defective in binding to BamD in vitro as well; however, replacing a highly conserved glycine in the β-signal, G771, with alanine resulted in assembly defects without decreasing binding to BamD. Therefore, the β-signal must have additional functions beyond its binding to BamD. To define these requirements further, we began to investigate the role of G771 by mutating residues that interact with it in the reported crystal structures of BamA (13, 20, 21, 27, 34). We show here that BamA G771 participates in a network of tertiary interactions between the β-barrel lumen and an extracellular loop (L6); this network must be formed before the protein’s insertion into the OM. Our results suggest that hydrophilic extracellular loops must be shielded from the hydrophobic membrane during their translocation across the bilayer and that the Bam complex may facilitate β-barrel assembly by inducing unfolded substrates to form structures that internally bury such loops.

Results

A Change in the BamA β-Signal Creates an Assembly-Defective Substrate that Stalls in the Periplasm.

To characterize BamA(G771A) in vivo, we introduced plasmids that constitutively express either WT or mutant bamA into a bamA depletion strain in which the chromosomal copy of bamA is arabinose-inducible (7, 14). The strain expressing BamA(G771A) alone was viable (Fig. 1B), but had OM defects as indicated by hypersensitivity to rifampicin (Fig. 1C). Rifampicin does not easily penetrate the OM of Gram-negative cells, and therefore we use this antibiotic as a probe for OM permeability (35). To determine whether these OM defects resulted from impaired assembly of BamA(G771A), we expressed plasmid-encoded copies of His-tagged WT or mutant bamA in cells also containing WT bamA on the chromosome and compared levels of the His-tagged proteins. We were not able to detect BamA(G771A) in these cells (Fig. 1D), but we observed partial recovery of BamA(G771A) levels when DegP was genetically deleted (Fig. 1E). Because DegP is the principal protease that degrades misfolded OM proteins (OMPs) in the periplasm (36), these results imply that the assembly of BamA(G771A) is impaired at an early stage in the periplasm and, consequently, is degraded by DegP. It is worth noting, in contrast, that we have previously observed that DegP does not degrade an assembly-defective substrate that stalls at a late stage, LptD4213 (Fig. S1A). Furthermore, we did not observe discrete degradation fragments of BamA(G771A) at lower molecular weights by SDS/PAGE, suggesting that the substrate is not aberrantly degraded by DegP in its folded form (Fig. S1B).

Fig. 1.

Disrupting interactions between two conserved features of BamA impairs its assembly at an early stage. (A) Structure of the BamA β-barrel domain. The 16-stranded β-barrel BamA contains a weak point of β-barrel closure (β1 and β16; red strands) in addition to a long sixth extracellular loop (L6; blue). Position W776 of BamA resides at the end of β14 (orange) and was previously found to interact with BamD during its assembly as a substrate (32). Crystal structure images were generated from the E. coli BamACDE crystal structure [Protein Data Bank (PDB) ID code 5EKQ] (13). (B) BamA mutants are functional in vivo. The BamA depletion strain (JCM166) containing an arabinose-inducible WT bamA copy was transformed with plasmids that express WT or mutant bamA alleles (7, 14). (B, Left) In the presence of arabinose, all strains were viable. (B, Right) In the absence of arabinose (i.e., in the absence of chromosomal WT BamA expression), all strains were viable with exception of the vector control. (C) The OM of cells expressing BamA(G771A) are leaky to antibiotics. JCM166 cells expressing BamA(G771A) were plated in the presence or absence of 0.75 μg/mL rifampicin under chromosomal WT BamA depleting conditions. (D and E) BamA mutants are degraded by the periplasmic protease DegP. MC4100 (D) or MC4100 degP::cam (E) cells expressing WT or mutant bamA alleles were harvested and analyzed via SDS/PAGE followed by immunoblotting cell lysates. (F) BamA(G771A) is not membrane-integrated. Lysates from MC4100 degP::cam cells expressing WT BamA or BamA(G771A) were washed with 100 mM Na₂CO₃ and the resulting membranes were extracted and applied to SDS/PAGE followed by immunoblotting. (G) The conserved BamA L6 VRGF motif makes contacts to conserved residues on the β-barrel wall. In particular, the side chain of R661 (L6) interacts with F738 in β13. G771 resides beneath the phenylalanine side chain in close proximity to both F738 and R661.

Fig. S1.

The periplasmic protease DegP does not degrade the late-stage assembly defective substrate LptD4213, whereas it efficiently degrades BamA(G771A). (A) DegP does not affect LptD4213 levels. MC4100 cells expressing FLAG-tagged WT or mutant lptD alleles were grown to midlog phase, upon which time cells were harvested via centrifugation, resuspended in 1× SDS sample buffer and boiled. Samples were applied to SDS/PAGE followed by analysis via Western blotting. (B) Lower-molecular-weight degradation products were not observed from cells expressing BamA(G771A). MC4100 cells expressing His-tagged WT or mutant bamA alleles were grown to late-log phase and harvested via centrifugation. Sample analysis was conducted in the same manner as in A above.

We determined whether the change in the β-signal (G771A mutation) of the BamA substrate affected its ability to integrate into the membrane using a carbonate extraction procedure. This type of experiment is often used to distinguish between membrane-associated vs. membrane-integrated proteins (37). We have previously shown that this procedure also applies to membrane β-barrels (22). Using strains that lack DegP, we observed that BamA(G771A) was not retained in the membrane fraction, suggesting that this substrate is only loosely associated with the OM, rather than fully inserted (Fig. 1F). Because changing G771 to alanine makes the substrate more susceptible to proteases in the periplasm and reduces its association with the OM, we conclude that this change causes a folding defect early in the assembly process and that DegP sensitivity is a reasonable proxy for defects in proper membrane integration.

In the crystal structure of E. coli BamA, G771 resides in the third-to-last (14th) β-strand of the β-barrel (Fig. 1A). On the preceding (13th) β-strand is another highly conserved residue, F738 (Fig. S2). The phenylalanine side chain of this residue is positioned directly above G771 in the lumen of the β-barrel, and we wondered whether this arrangement of residues is functionally important (Fig. 1G) (13, 20, 21, 27, 34). Covariance of this glycine-aromatic or “mortise and tenon” motif in the context of autotransporters has been reported (38). We have recently suggested that the β-barrel folding mechanism begins by BamD binding the β-signal containing G771 (32). It seemed possible that unfavorable steric interactions between the side chains of G771A and F738 prevented subsequent steps in the folding process. Therefore, we expressed a plasmid-encoded BamA variant in which F738 was replaced with alanine, BamA(F738A), in the bamA depletion strain (Fig. 1B). BamA(F738A) exhibited similar phenotypes with respect to viability and DegP susceptibility as BamA(G771A) (Fig. 1 B, D, and E). Because mutation of F738 also results in assembly defects similar to BamA(G771A), we conclude that the interaction of G771 with F738 is important for assembly of BamA as substrate.

Fig. S2.

BamA loop-to-lumen contacts involve highly conserved residues. The conserved BamA L6 VRGF motif makes important contacts to residues E717, F738, and D740, which stabilize the loop within the BamA β-barrel lumen. G771 resides underneath the phenylalanine side chain of F738. Alignments were obtained by using Clustal-Omega.

Membrane Integration Requires Burial of L6 in the Lumen of the BamA β-Barrel.

In the final folded form of BamA, the phenylalanine side chain of F738 not only contacts G771 within the β-signal, but also contacts a conserved motif between β-strands 11 and 12 in extracellular loop 6 (L6), which is folded into the lumen of the β-barrel (Fig. 1A and Fig. S2) (13, 20, 21, 27, 34). This motif, consisting of the sequence VRGF, is the most highly conserved segment of BamA and is also found in the much broader Omp85 family of β-barrel proteins (39). It was previously shown that E. coli mutants containing two or three changes in the VRGF motif displayed severe defects in assembly of several different OMPs, including BamA. The authors concluded that these mutants impaired substrate BamA assembly, the activity of the mature Bam complex, or both (40).

Given the proximity of F738 to the VRGF motif, we wondered whether the mutations at F738 and G771 affected the long-range L6 VRGF contacts within the β-barrel lumen. To test this hypothesis, we made a double mutant encoding BamA(V660A/R661A), in which the first two residues of the VRGF motif were changed to alanine. When expressed in cells containing WT BamA, levels of BamA(V660A/R661A) were low, similar to BamA(G771A) and BamA(F738A) (Fig. 1D). Moreover, protein levels recovered when DegP was deleted (Fig. 1E). Therefore, changes in the L6 VRGF motif also result in early assembly defects of the substrate BamA that expose it to periplasmic degradation. Because changes to the conserved residues G771, F738, and V660/R661, which are close to one another in the crystal structure, all result in early assembly defects, we posit that they form a network of interactions important for early folding of BamA in the periplasm. If so, some amount of tertiary structure must be present at an early stage of assembly (see below).

BamA(V660A/R661A) Accumulates on BamD.

Because DegP is a general protease that recognizes misfolded β-barrel proteins in the periplasm, it was unclear whether all three assembly-defective BamA variants stall at the same early step during folding. Therefore, we reasoned that we could accumulate folding intermediates in a strain lacking DegP and assess whether cells expressing different BamA variants have different phenotypes. We assessed the rifampicin sensitivity of strains that lack degP but express the mutant bamA alleles in addition to a chromosomal WT bamA copy. We found that strains expressing BamA(G771A) and BamA(F738A) were resistant to rifampicin, but the strain expressing BamA(V660A/R661A) was sensitive to it (Fig. 2A). Therefore, expression of BamA(V660A/R661A) increases the leakiness of the OM, even though a WT copy of BamA is present. Expression of the mutant copy produces a dominant antibiotic-sensitivity phenotype, which might suggest that the mutant compromises the function of WT BamA.

Fig. 2.

BamA substrates containing VRGF motif mutations accumulate on BamD. (A) Expression of BamA(V660A/R661A) causes a dominant antibiotic sensitivity phenotype. MC4100 degP::cam cells expressing WT or mutant bamA alleles were plated in the presence or absence of 2.5 μg/mL rifampicin. (B) Overview of the in vivo photo-cross-linking assay used to probe interaction of substrate BamA with BamD. The unnatural amino acid pBPA was incorporated into position 776 of BamA, and cross-linking was induced with UV light. (C) An overview of the BamA constructs used in the cross-linking experiments. All constructs contain an N-terminal FLAG tag in addition to the W776pBPA mutation. Additional mutations were introduced as depicted in constructs 3–6. Construct 6 lacks the native cysteine residues (C690S and C700S) to ensure that only one disulfide bond can be formed. (D) BamA(W776pBPA/V660A/R661A) accumulates on BamD. BL21(DE3) cells expressing the indicated constructs were UV-irradiated or not, and analyzed via SDS/PAGE followed by immunoblotting cell lysates. [Note that the BL21(DE3) strain used in these experiments expresses higher and equal levels of the BamA substrates—unlike the MC4100 strain used in Fig. 1D. The higher expression allows observation of the cross-linked species.]

BamA(V660A/R661A) is distinct from BamA(G771A) and BamA(F738A) because it has a functional β-signal; therefore, associated phenotypes may result from defects that occur subsequent to engagement with BamD within the Bam complex. If so, this substrate variant should have a longer lifetime of association with the Bam complex, which could be detected by using in vivo cross-linking. Previously, we showed that a C-terminal BamA fragment containing residues 715–810 containing the unnatural amino acid p-benzoylphenylalanine (pBPA) (41) at position 776 (Fig. 1A) cross-links to BamD in vivo (Fig. 2 B–D, construct 1) (32). We interpreted the observed cross-link to reflect binding of this fragment of the unfolded BamA substrate to BamD during assembly. To probe the interaction of full-length BamA substrates with BamD during assembly, we introduced pBPA into position 776 of the full-length proteins, expressed them in cells also containing WT BamA, and irradiated the cells with UV light (Fig. 2 B–D, constructs 2–5) (32). The extent of cross-linking was evaluated by Western blotting to determine whether protein intermediates accumulated on BamD during folding. We observed minimal cross-linking with BamA(W776pBPA) (construct 2), BamA(F738A/W776pBPA) (construct 4), or BamA(G771A/W776pBPA) (construct 5), but substantial cross-linking with BamA(V660A/R661A/W776pBPA) (construct 3; Fig. 2 C and D). Cross-linking cannot reflect interactions between folded BamA(W776pBPA) and BamD (as found in the assembled Bam complex) because the distance between these two features (20 Å) prevents them from interacting (Fig. S3) (13, 20, 21). Because cross-linking efficiency depends on the residence time of the bound substrate, this result implies that the β-signal mutants do not stably associate with BamD, whereas the L6 mutant accumulates on it. Therefore, the V660A/R661A mutation creates an assembly problem that differs from that created by the G771A and F738A mutations; the V660A/R661A mutant remains bound to BamD, but is still unable to fold efficiently.

Fig. S3.

BamA W776 can be used to detect substrate BamA–BamD interactions. The X-ray crystal structure of assembled BamAD reveals that residue W776 resides >20 Å away from BamD. Therefore, placement of pBPA at position 776 in full-length BamA allows selective reporting of substrate BamA–BamD interactions by photo-cross-linking rather than interactions within mature Bam complexes. Crystal structure images were generated from the E. coli BamACDE crystal structure (PDB ID code 5EKQ) (13).

We wondered whether we could further differentiate the mutants that alter the β-signal (G771A and F738A) from the mutant that alters the L6 VRGF motif (V660A/R661A). We hypothesized that if BamA(V660A/R661A) is stalled on BamD, as the cross-linking suggests, it may exhibit differences in cell-surface exposure compared with the variants that do not bind BamD productively. To test this hypothesis, we assessed cell surface susceptibility of BamA(V660A/R661A) and BamA(G771A) to externally added proteinase K. In intact cells, any protein that has surface-exposed regions is theoretically susceptible to cleavage by an externally added protease, whereas proteins that are not membrane-integrated should be protected (18, 19, 42). We treated cells expressing the mutant BamA proteins (and lacking DegP) with proteinase K and assayed total levels and cleavage products. BamA(G771A) was not susceptible to proteinase K degradation, consistent with an early folding defect that prevents engagement at BamD and subsequent membrane integration (Fig. S4). In contrast, BamA(V660A/R661A) demonstrated a susceptibility comparable to WT BamA. We conclude that BamA(V660A/R661A) is stalled on the Bam complex in a state that is partially exposed to proteases on both sides of the OM.

Fig. S4.

Proteinase K susceptibility reveals differences in cell surface exposure. MC4100 degP::cam cells constitutively expressing WT or mutant bamA alleles were grown to midlog phase and harvested via centrifugation. After washing, cells were treated with 0.5 mg/mL proteinase K for 30 min at 37 °C. Digestion was stopped via addition of PMSF, and samples were analyzed via SDS/PAGE, followed by immunoblotting cell lysates.

Tethering L6 to the β-Barrel Lumen Accelerates Membrane Integration.

Given that G771, F738, and the VRGF motif appear to form a network of interactions in the folded BamA structure, we thought it might be possible to rescue the assembly-defective BamA substrates by artificially restoring the interaction of L6 with the β-barrel wall in the lumen. Previously, a loop-to-lumen (LL) disulfide bond was used to tether an extracellular loop to the lumen of PhoE with no apparent effect on assembly; this disulfide bond was found to form in the periplasm before membrane integration (43). More recently, a double cysteine mutant in BamA (T659C and V713C; Fig. 3A) was shown to form a disulfide bond without compromising viability (11). We wondered whether this disulfide bond could rescue our assembly-defective bamA alleles by restoring L6-to-lumen contact in the periplasm. We replaced residues T659 and V713 with cysteine in WT BamA and in the three assembly-defective mutants, and we expressed all four of these LL tethered proteins in cells also containing WT BamA and DegP. Introduction of the LL disulfide bond increased protein levels of all three assembly-defective BamA mutants (Fig. 3B). This rescue of protein levels is specific to the tethering of L6 to the β-barrel wall, because disulfide cross-links between other regions of the BamA β-barrel did not improve assembly (Fig. S5). We conclude that the LL disulfide increases levels of the BamA substrates by preventing degradation in the periplasm by DegP.

Fig. 3.

Efficient membrane integration requires burial of L6 within the lumen of BamA. (A) Mutation of residues T659 and V713 to cysteine enables formation of a LL disulfide bond that tethers the L6 VRGF motif to the β-barrel wall (11, 13). (B) The LL disulfide rescues assembly of BamA variants. MC4100 cells expressing WT or mutant bamA alleles were harvested, and cell lysates were analyzed via SDS/PAGE, followed by immunoblotting. The LL disulfide was introduced into bamA alleles lacking the native cysteine residues (C690S and C700S). (C and D) The LL disulfide promotes membrane integration of BamA(G771A) and BamA(V660A/R661A). Lysates from MC4100 degP::cam cells were washed and analyzed in the same manner as described in Fig. 1F. Lanes 1 and 2 of each gel in C are also shown in Fig. 1F. (E) BamA(V660A/R661A)-LL has a shorter residence time on BamD. UV Photo-cross-linking was performed in the same manner as described in Fig. 2 B–D.

Fig. S5.

The LL disulfide rescues assembly in contrast to disulfides that tether other regions of the BamA β-barrel. MC4100 or MC4100 degP::cam cells constitutively expressing WT or mutant bamA alleles were grown to midlog phase and analyzed via SDS/PAGE, followed by immunoblotting cell lysates. “Native” is defined as the WT cysteine residues present within BamA (C690/C700); “seam” is defined as the disulfide formed between the N and C termini of the β-barrel (G429C/T809C); and “exit pore” is defined as the disulfide formed between L6 and the N terminus of the β-barrel (E435C/S665C). All nonnative disulfides were introduced into bamA alleles lacking the native cysteine residues (C690S/C700S). Disulfide formation between the nonnative cysteine residues has been reported previously (11).

Finally, we wanted to establish whether the LL disulfide increases the amount of functional and membrane-integrated (i.e., assembled) protein. Both the BamA(G771A)-LL variant and BamA(V660A/R661A)-LL variant were resistant to membrane extraction by carbonate wash (Fig. 3 C and D) and were more susceptible to externally added proteinase K (Fig. S4). Increased levels and membrane integration of BamA(G771A)-LL also correlated with improved OM barrier function, because cells expressing this protein were no longer sensitive to rifampicin under depleting conditions in the bamA depletion strain (Fig. S6). Moreover, cross-linking of BamA(V660A/R661A)-LL to BamD was reduced, consistent with restored membrane integration and assembly (Fig. 3E). Collectively, these experiments suggest that a network of contacts between the VRGF motif and the β-hairpin containing F738 and G771 is required for efficient membrane integration.

Fig. S6.

The improved levels of BamA caused by introduction of the LL disulfide correlates with improved outer membrane barrier function. JCM166 cells (7, 14) constitutively expressing WT or mutant bamA alleles were plated in the presence or absence of 0.75 μg/mL rifampicin under chromosomal WT BamA-depleting conditions (absence of arabinose).

Discussion

We have identified three variants of BamA that reveal how the native protein assembles into the OM. Each of these substrates has early folding defects preventing their efficient membrane integration. We showed that assembly defects in two of the substrates, BamA(G771A) and BamA(F738A), which have changes in the β-signal proposed to be important for binding to BamD, were, in fact, due to failure to successfully engage BamD. Although a small amount of these BamA substrates can assemble into the membrane, most are degraded in the periplasm (Fig. 1). The other substrate, BamA(V660A/R661A), which contains amino acid substitutions in the VRGF motif, engaged BamD, but was slow to integrate into the OM (Fig. 3). This substrate displayed a dominant antibiotic susceptibility phenotype, evidently because it stalls on the Bam complex (Fig. 2). We suspect that the susceptibility of these substrates to degradation by DegP is also due to this accumulation, which results in prolonged exposure of nascent protein to periplasmic proteases.

We have also shown that it is possible to promote membrane integration of these slow-folding substrates by forming a disulfide bond that tethers L6 to the β-barrel lumen. This result provides a clue as to how membrane β-barrel proteins solve a crucial problem. These proteins typically contain large extracellular loops that must cross the hydrophobic membrane (26). Our studies show that a network of noncovalent interactions between the β-barrel lumen and the VRGF motif anchor L6 in the lumen so that it is not exposed to the hydrophobic membrane. Disrupting this constellation of interactions slows assembly even when BamD is engaged, but it is possible to restore assembly by covalently tethering the loop to a neighboring β-strand in the lumen. Therefore, these studies allow us to assign a role for the conserved L6 VRGF motif in promoting the assembly of BamA itself. The fact that the disulfide rescues all three BamA substrates, even though they stall at two different stages of assembly, suggests that the identified interactions may form in a single step.

We propose the following model for assembly of BamA. Substrate BamA arrives at the OM in an unfolded state (state I; Fig. 4). The substrate proceeds to state II, in which interactions between F738 and G771 in the β-signal mediate binding to BamD and help organize the binding site for the VRGF motif of L6. Transition from state II to III is mediated by the interaction of L6 with the binding site created in the nascent β-barrel lumen. Burial of L6 within the lumen now allows rapid membrane integration. Membrane integration of substrate BamA requires the formation of conserved tertiary contacts between the hairpin containing the β-signal (β-strands 13 and 14) and the highly conserved L6. Substrates that are able to bind BamD, but cannot form critical tertiary contacts, such as BamA(G771A) (Fig. 1G), cannot be membrane-integrated and are therefore susceptible to periplasmic proteases. Introduction of the LL nonnative disulfide bypasses the requirement for a well-defined binding site for the VRGF motif to achieve state III.

Fig. 4.

Model of β-barrel membrane integration catalyzed by internal burial of hydrophilic extracellular loops. BamD initiates the assembly of substrate BamA by binding a sequence near the C terminus of the unfolded protein (i.e., the β-signal). This binding promotes the nucleation of a set of interactions that fold a conserved extracellular loop (L6) into the center of the forming β-barrel. Membrane integration proceeds efficiently after burying this loop, as the barrier to translocating a hydrophilic sequence across the membrane has been removed. We note that we do not mean to imply distinct steps involving intermediates I–IV, but, rather, a continuous folding process.

There are parallels between folding requirements for BamA and LptD, a large 26-stranded β-barrel that is assembled by Bam around a lipoprotein plug, LptE. Previously, we identified a key interaction between an extracellular loop of LptD with LptE that is required for release from the Bam complex (44). A separately folded protein plug (LptE) is required to scaffold assembly of LptD (25, 45) because its increased size makes folding a more challenging entropic problem; however, the requirement for contacts from a loop to the interior of the β-barrel appears to be conserved. Accordingly, we suggest that luminal scaffolds (LptE for LptD or the L6 VRGF motif for BamA) are required to interact with the interior surface of β-barrels as they are undergoing folding and membrane integration.

Burial of an extracellular loop against conserved residues in the β-barrel appears to be common in these OMPs. Crystal structures of PhoE, OmpF, and LamB each reveal similar placement of a phenylalanine side chain proximal to both a glycine on an adjacent strand and a large extracellular loop (46–48) (Fig. S7). This interaction was shown to form early in the periplasm during assembly of PhoE without adverse effects (43). It remains to be determined whether other β-barrel substrates use such highly specific contacts that facilitate burial of extracellular loops. Tucking extracellular loops within the nascent β-barrel not only provides a mechanism for translocation of these polar chains across the membrane to the cell surface, but also prevents entrapment of phospholipids within the lumen as it penetrates the membrane. The discovery of specific contacts required for membrane integration of substrate BamA also raises the possibility that these contacts could be targeted with an inhibitor that interferes with the formation of this critical state.

Fig. S7.

β-barrel substrates other than BamA contain extracellular loops that contact the β-barrel wall in the lumen. Crystal structures of PhoE (45), OmpF (47), and LamB (46) reveal the presence of an extracellular loop (blue) that closely contacts an interior phenylalanine or tyrosine residue (orange). In analogy to BamA, the side chains of these phenylalanine or tyrosine residues are positioned over a glycine (magenta) on the adjacent strand. PDB ID codes were as follows: 1PHO (PhoE), 4GCP (OmpF), and 1MAL (LamB).

Materials and Methods

Experimental Procedures, Bacterial Strains, and Plasmids.

Experimental procedures for in vivo photo-cross-linking, carbonate washing of membranes, proteinase K digestions, and plating are provided in SI Materials and Methods. The strains used in this study are described in Table S1. All plasmids are listed in Table S2.

Table S1.

Strains used

Bacterial strains	Genotype and relevant features	Source
JCM166	MC4100 arar/- Δ(λatt-lom)::bla PBADyaeT araC ΔyaeT (BamA depletion strain)	Refs. 7 and 14
MC4100	F−araD139 Δ(argF-lac)U169 rpsL150 relA1 flbB5301 deoC1 ptsF25 rbsR thi	Ref. 49
NR779	MC4100 degP::cam	Laboratory strain

Table S2.

Plasmids used

Name	Description	Source
pJW451	pZS21	This study
pSK476	pZS21/His-BamA	Ref. 7
pJW207	pZS21/His-BamA(G771A)	This study
pJW445	pZS21/His-BamA(F738A)	This study
pJW261	pZS21/His-BamA(V660A/R661A)	This study
pJW446	pET23a/FLAG-BamA(715-810/W776Amber)	This study
pJW443	pET23a/FLAG-BamA(W776Amber)	This study
pJW472	pET23a/FLAG-BamA(V660A/R661A/W776Amber)	This study
pJW473	pET23a/FLAG-BamA(F738A/W776Amber)	This study
pJW444	pET23a/FLAG-BamA(G771A/W776Amber)	This study
pJW234	pZS21/His-BamA(C690S/C700S/T659C/V713C)	This study
pJW223	pZS21/His-BamA(G771A/C690S/C700S)	This study
pJW237	pZS21/His-BamA(G771A/C690S/C700S/G429C/T809C)	This study
pJW236	pZS21/His-BamA(G771A/C690S/C700S/E435C/S665C)	This study
pJW238	pZS21/His-BamA(G771A/C690S/C700S/T659C/V713C)	This study
pJW463	pZS21/His-BamA(F738A/C690S/C700S/T659C/V713C)	This study
pJW453	pZS21/His-BamA(V660A/R661A/C690S/C700S/T659C/V713C)	This study
pJW484	pET23a/FLAG-BamA(V660A/R661A/C690S/C700S/T659C/V713C/W776Amber)	This study
pLptD	pET2342/LptD-FLAG	Ref. 44
pLptD4213	pET2342/LptD4213-FLAG	Ref. 44

Analysis of Cellular BamA Levels.

MC4100 (49) cells harboring pSK476, pJW207, pJW445, pJW261, pJW234, pJW238, pJW463, or pJW453 were grown at 37 °C in LB supplemented with kanamycin to an OD₆₀₀ of ∼1.0 over 8 h. The cells from a 1-mL sample normalized to an OD₆₀₀ of 0.8 were collected by centrifugation at 10,000 × g for 10 min. The resulting cell pellets were resuspended in 80 μL of 1× SDS-sample buffer and boiled for 8 min. The samples were applied to SDS/PAGE and analyzed via Western blotting. LptE was detected by immunoblotting with LptE antiserum and secondary anti-rabbit–HRP conjugates. His–BamA was detected by using anti–penta-His–HRP antibodies. The same procedure was followed to analyze protein levels in the MC4100 degP::cam strain, with the exception that chloramphenicol was also added to the growth medium.

SI Materials and Methods

Please see Materials and Methods for key experimental procedures.

Materials.

All cultures were grown in LB medium from Difco. Kanamycin (kan) and carbenicillin (carb) were used at a final concentration of 50 μg/mL, chloramphenicol (cam) was used at a final concentration of 30 μg/mL. TBS is defined as 20 mM Tris (pH 8) with 150 mM NaCl unless otherwise noted. FLAG-BamA was detected on Western blots with monoclonal anti-FLAG M2-peroxidase (HRP) mouse antibody from Sigma-Aldrich. The penta-His–HRP mouse antibody (Qiagen) was used to detect His-tagged proteins, and the secondary anti-rabbit–HRP conjugate (GE Healthcare) was used to detect the DegP, DsbA, SurA, BamD, and LptE primary antibodies. All HRP conjugates were visualized with the Lumigen PS-3 detection reagent (GE Healthcare). NovaBlue (Novagen) and BL21(DE3) (Novagen) strains were used for expression and cloning procedures. Proteinase K (Omega Bio-Tek) was dissolved in water and used directly in the protease digestion experiments. 2× SDS sample buffer refers to a mixture of 125 mM Tris (pH 6.8), 4% (wt/vol) SDS, 30% (vol/vol) glycerol, 0.005% bromophenol blue, and 5% (vol/vol) β-mercaptoethanol. The 1× SDS sample buffer refers to dilution of the above 2× SDS sample buffer 1:1 with TBS. Stock solutions of pBPA (Bachem) were prepared at a concentration of 0.8 M in 1 M sodium hydroxide. All SDS/PAGE analyses were performed by using 4–20% (vol/vol) gels at 200 V for 50 min.

Equipment.

Photo-cross-linking was performed by using a UVP Blak-Ray B-100AP high-intensity UV lamp with a 100 W spot bulb. A280 values were obtained by using a NanoDrop 2000 (Thermo Fisher Scientific) and OD600 values were obtained by using a Genesys 10-UV spectrophotometer (Thermo Fisher Scientific). Mechanical cell lysis was achieved by using an EmulsiFlex-C3 cell disrupter (Avestin) at a pressure of 10,000–15,000 psi. Ultracentrifugation was performed by using an Optima XE-90 ultracentrifuge (Beckman Coulter). Western blots were imaged by using an Azure Biosystems C400 imager by using the “low” sensitivity setting and the tray placed on the upper pedestal.

Photo-Cross-Linking.

Photo-cross-linking was performed in a manner similar to that described by Hagan et al. (32). BL21(DE3) cells harboring pSup-BpaRS-6TRN (41) and pJW446, pJW443, pJW444, pJW472, pJW473, or pJW484 were grown at 37 °C overnight in LB supplemented with carbenicillin, chloramphenicol, and 0.2% glucose. These cultures were used to inoculate 12 mL of LB supplemented with 0.8 mM pBPA, carbenicillin, chloramphenicol, and 0.2% glucose via 1:100 dilution. These cultures were grown at 37 °C until an OD₆₀₀ of 0.11–0.16 was reached. A 1440 μL sample of each culture was removed and incubated on ice for use as a nonirradiated control. In a 96-well clear-bottomed plate, eight aliquots of 200 μL of each strain were incubated on ice and irradiated from above (∼5 cm away) with UV light at 365 nm for 10 min. A total of 180 μL was then collected from each well (1,440 μL total for each strain), and all samples were centrifuged at 16,000 × g for 10 min. The resulting cell pellets were resuspended in OD₆₀₀ × 400 μL of 1× SDS sample loading buffer and boiled for 8 min. The samples were applied to SDS/PAGE and analyzed via Western blotting. BamD was detected by immunoblotting with BamD antiserum and secondary anti-rabbit–HRP conjugates. FLAG–BamA was detected by using anti-FLAG–HRP antibodies.

Protease Digestion Experiments.

Protease digestion experiments were conducted in a similar manner to that described by Rigel et al. (18, 19). MC4100 degP::cam cells harboring pSK476, pJW207, pJW261, pJW453, or pJW482 were grown at 37 °C overnight in LB supplemented with kanamycin and chloramphenicol. These cultures were used to inoculate 5 mL of LB with the same additives via 1:100 dilutions. These cultures were grown at 37 °C until an OD₆₀₀ of 0.6–0.7 was reached. A total of 1 mL × 2 of culture was normalized to OD₆₀₀ = 0.6, and these cell suspensions were centrifuged at 10,000 × g for 10 min. Cells were washed with 20 mM Tris (pH 7.5)/100 mM NaCl and resuspended in 0.5 mL of the same buffer. Proteinase K was added to produce a final concentration of 0.5 mg/mL, and the samples were shaken at 37 °C for 30 min. All samples were treated with 25 μL of 100 mM PMSF in isopropanol and incubated on ice for 10 min. After inactivation of proteinase K with PMSF, samples were centrifuged at 10,000 × g for 10 min. The resulting cell pellets were resuspended in 1× SDS sample buffer and boiled for 8 min. The samples were applied to SDS/PAGE and analyzed via Western blotting. DsbA was detected by immunoblotting with DsbA antiserum and secondary anti-rabbit–HRP conjugates. His–BamA was detected by using anti–penta-His–HRP antibodies.

Analysis of Cellular LptD Levels.

Analysis of LptD levels was performed in a manner similar to that described by Lee et al. (44). MC4100 cells harboring pLptD- or pLptD4213-FLAG were grown at 37 °C in LB supplemented with carbenicillin to midlog phase. The cells from a 1-mL sample normalized to an OD₆₀₀ of 0.4 were collected by centrifugation at 10,000 × g for 10 min. The resulting cell pellets were resuspended in 80 μL of 1× SDS-sample buffer and boiled for 8 min. The same procedure was followed to analyze protein levels in the MC4100 degP::cam strain, with the exception that chloramphenicol was also added to the growth medium. All samples were applied to SDS/PAGE and analyzed via Western blotting. LptE and DegP were detected by immunoblotting with LptE and DegP antiserum, respectively, followed by secondary anti-rabbit–HRP conjugates. FLAG-LptD was detected by using anti-FLAG–HRP antibodies.

Plating of Strains Under Depleting and Nondepleting Conditions.

Plating of the BamA depletion strain was performed in a manner similar to that described by Wu et al. (7). JCM166 (7, 14) cells harboring pJW451, pSK476, pJW207, pJW445, or pJW261 were grown in LB supplemented with kanamycin and 0.05% arabinose at 37 °C until midlog phase (OD₆₀₀ = 0.5–0.6) was reached. These cultures were used to prepare LB (no arabinose added) supplemented with kanamycin to a normalized OD₆₀₀ of 0.1. Serial 10-fold dilutions of the normalized cultures were performed to prepare the indicated dilution series. The dilution series (5 μL) were plated on agar plates containing kanamycin, carbenicillin, and the indicated additive and incubated at 37 °C for 18–20 h. In cases where rifampicin was added, a final concentration of 0.75 μg/mL was used.

Plating of MC4100 degP::cam Strains.

MC4100 degP::cam cells harboring pJW451, pSK476, pJW261, pJW445, or pJW207 were grown in LB supplemented with kanamycin and chloramphenicol at 37 °C until early log phase (OD₆₀₀ between 0.12 and 0.20) was reached. These cultures were used to prepare LB supplemented with kanamycin and chloramphenicol with a normalized OD₆₀₀ of 0.1. Serial 10-fold dilutions of the normalized cultures were performed to prepare the indicated dilution series. The dilution series (5 μL) were plated on agar plates containing kanamycin, chloramphenicol, and the indicated additive and incubated at 37 °C for 18–20 h. In cases where rifampicin was added, a final concentration of 2.5 μg/mL was used.

Membrane Extraction with Sodium Carbonate.

Membrane extraction was performed in a manner similar to that described by Molloy (37). MC4100 degP::cam cells harboring pSK476 or pJW207 were grown at 37 °C overnight in LB supplemented with kanamycin and chloramphenicol. These cultures were used to inoculate 750 mL of LB with the same additives via 1:100 dilutions. These cultures were grown at 37 °C until an OD₆₀₀ of 1.1–1.2 was reached. The cells were harvested via centrifugation at 5,000 × g for 12 min and resuspended in 25 mL of resuspension solution [50 mM Tris (pH 8) and 10 μg/mL DNase]. Cells were lysed via two passes through the cell disrupter, and the resulting lysate was centrifuged at 5,000 × g for 10 min. The resulting supernatant was used to prepare normalized cell lysate suspensions to an A280 value of 30 (assayed by using NanoDrop). In 10-mL ultracentrifuge tubes, 8 mL of 100 mM Na₂CO₃ (not pH-adjusted) was cooled on ice and treated with 0.8 mL of normalized lysate. The solutions were rocked at 4 °C for 2 h, upon which time all samples were centrifuged at 100,000 × g for 30 min. The resulting supernatant was removed, and the pellets were treated with 0.5 mL of extraction solution [TBS and 2% (vol/vol) Triton X-100] and rocked for 1.5 h at room temperature. The extracts were transferred to epitubes, mixed 1:1 with 2× SDS sample buffer, and boiled for 8 min. The samples were applied to SDS/PAGE and analyzed via Western blotting. His–BamA was detected by using anti–penta-His–HRP antibodies.

Strain and Plasmid Construction.

The strains used in this study are described in Table S1. All plasmids are listed in Table S2.