Lipoprotein LptE is required for the assembly of LptD by the β-barrel assembly machine in the outer membrane of Escherichia coli

Abstract

Most Gram-negative bacteria contain lipopolysaccharide (LPS), a glucosamine-based phospholipid, in the outer leaflet of the outer membrane (OM). LPS is unique to the bacterial OM and, in most cases, essential for cell viability. Transport of LPS from its site of synthesis to the cell surface requires eight essential proteins, MsbA and LptABCDEFG. Although the key players have been identified, the mechanism of LPS transport and assembly is not clear. The stable LptD/E complex is present at the OM and functions in the final stages of LPS assembly. Here, we have identified the mutant allele lptE6, which causes a two–amino-acid deletion in the lipoprotein LptE that affects its interaction with LptD. Highly specific suppressor mutations were isolated not only in lptD but also in bamA, which encodes the central component of the β-barrel assembly machine. We show that lptE6 and both suppressor mutations affect the assembly of the LptD/E complex and suggest that the lipoprotein LptE interacts with LptD while this protein is being assembled by the β-barrel assembly machine.

The cell envelope of Gram-negative bacteria such as Escherichia coli consists of an inner membrane (IM), an aqueous peptidoglycan-containing periplasm, and an outer membrane (OM) (1). Unlike the IM, which is composed entirely of phospholipids, the OM is an asymmetric lipid bilayer with an inner leaflet composed of phospholipids and an outer leaflet composed of lipopolysaccharide (LPS) (2). Strong lateral interactions between neighboring LPS molecules make the OM more impermeable than a phospholipid bilayer to a wide range of hydrophobic dyes, hydrophobic antibiotics, and detergents (including bile salts), allowing Gram-negative bacteria to survive in environments such as the human intestine (1, 3).

The LPS biosynthetic pathway is well understood (4), but the process of LPS transport to the OM and assembly at the cell surface is just beginning to be characterized (5). LPS is synthesized at the inner leaflet of the IM (6). In the first step of LPS transport, the essential ABC transporter MsbA flips LPS from the cytoplasmic to the periplasmic face of the IM (7–9). Seven essential lipopolysaccharide transport (Lpt) proteins transport LPS from the IM to the cell surface (10–14). When any one of the Lpt proteins is depleted, de novo synthesized LPS can no longer reach the cell surface and abnormal membrane structures accumulate in the periplasm (10–13). There is at least one transport pathway component in each cellular compartment and their locations suggest order of function in LPS transport. LptB/C/F/G (formerly YhbG/YrbK/YjgP/YjgQ) form an ABC transporter at the IM that is believed to extract LPS from the periplasmic leaflet of the IM (11, 13, 15). The periplasmic protein LptA (formerly YhbN) binds LPS (16) and is thought to aid in the transport of the amphipathic LPS across the aqueous periplasm (10). LPS is transported to the OM in spheroplasts even though soluble periplasmic factors have been lost (17). In addition, it has recently been reported that all seven Lpt proteins form a transenvelope complex and are found in a light OM fraction containing both the IM and the OM (18). Therefore, LPS transport across the periplasm most likely takes place at membrane contact sites through a protein bridge. The LptD/E complex (formerly Imp/RlpB) at the OM presumably functions in the final stages of assembling LPS into the outer leaflet of the OM (12). LptE binds LPS in a specific manner and could be playing a role in receiving LPS from the periplasm (19). Although the individual roles of these OM proteins are yet to be established, some insights have begun to emerge through recent developments in the biochemical characterization of this two-protein complex.

LptD is an outer membrane protein (OMP) with a soluble N-terminal domain (residues 25–202) and an integral C-terminal β-barrel domain (residues 203–784) (19, 20). Two nonconsecutive intramolecular disulfide bonds covalently join the two domains (21, 22). LptE is a lipoprotein anchored to the OM via N-terminal lipid moieties (23). Both OMPs and lipoproteins pass through the Sec translocon after synthesis in the cytoplasm but their biogenesis pathways diverge at the periplasmic face of the IM. After signal sequence cleavage, OMPs are thought to be protected in their unfolded state by periplasmic chaperones until they are delivered to the BamABCDE complex for assembly into the OM (24). Indeed, biogenesis of LptD is dependent on both the periplasmic chaperone SurA (25) and the Bam complex (12). In the case of lipoproteins, after lipid modifications to the N-terminal cysteine residue and signal sequence cleavage, those destined for the OM, like LptE, are released from the IM by LolCDE and handed off to the periplasmic chaperone LolA, which delivers them to LolB for assembly into the OM (26). At the OM, LptD and LptE form an extremely stable 1:1 complex and the β-barrel domain of LptD is sufficient for this interaction with LptE (19). Not only does LptE stabilize LptD but also it is required for the proper oxidation of LptD, suggesting a crucial role for LptE in the folding of its complex partner (19, 22).

Here we describe the isolation of highly specific suppressor mutations of a partial loss-of-function, 6-bp deletion mutation in lptE. Characterization of these suppressors provides insights into the structure and biogenesis of the LptD/E complex. These suppressors also provide a rational explanation for chemical conditionality, the genetic approach that was used to identify mutations in genes that specify components of the complex that assembles β-barrel proteins in the OM (27, 28).

Results

Isolation of Partial Loss-of-Function Mutations in lptE.

To understand the functional role of the OM lipoprotein LptE and probe its interaction with the β-barrel OMP LptD, we sought to isolate partial loss-of-function mutations in lptE. Random mutations were introduced into lptE by error-prone PCR. Mutagenized lptE was cloned into pBAD18, under an arabinose-inducible promoter, and the plasmids were introduced into a wild-type strain. To identify partial loss-of-function alleles carried on the plasmid, the wild-type lptE allele at the native chromosomal locus was replaced with a ΔlptE2::kan allele. At all times, expression from the plasmid was induced with 0.2% arabinose to facilitate identification of mutations that might cause severe defects in LptE function. The resulting lptE haploid strains were screened for increased sensitivity to the hydrophobic antibiotic rifampicin. Increased sensitivity to hydrophobic antibiotics is an indicator of a disrupted barrier function at the OM. One mutation obtained in this manner is lptE6. DNA sequence analysis of plptE6 revealed a 6-bp deletion that results in a YPISA (amino acids 116–120) to YRA change. We also isolated a transition mutation (CGT to TGT) that causes an Arg to Cys change in residue 157. Incidentally, unknown vector mutations lower lptE synthesis in this particular isolate, providing us with a useful control. The transition mutation is not solely responsible for the mutant phenotype and we refer to this mutant pBAD18lptE as plptE^low.

lptE6 and lptE^low Increase OM Permeability.

plptE⁺-carrying cells in which the chromosomal lptE gene is disrupted grow poorly in the absence of arabinose and plptE6-containing cells grow only in conditions where expression of the plasmid-borne gene is induced with arabinose. Because induction of expression of lptE⁺ from pBAD18 with 0.2% arabinose does not cause any phenotypic defects relative to the wild-type strain AM604 (Table 1, rows 1 and 2), all of the pBAD18-containing strains mentioned in this work were always grown with 0.2% arabinose. To quantify the OM permeability defects of the lptE mutants, we determined the minimal inhibitory concentration (MIC) of the hydrophobic antibiotics rifampicin and bacitracin for the mutant strains (Table 1, rows 3 and 4). Compared with the wild-type equivalent plptE⁺ strain, the plptE6 strain is 16-fold more sensitive to rifampicin and bacitracin. The plptE^low strain, on the other hand, is only two- to fourfold more sensitive than plptE⁺ to these antibiotics.

Table 1.

OM permeability defects

		MIC, mg·mL−1
	Strain	Rifampicin	Bacitracin
1	AM604	5	1,000
2	plptE	5	1,000
3	plptElow	2.5	250
4	plptE6	0.3125	62.5
5	plptE6; lptD7	1.25	250
6	plptE6; bamA8	2.5	250
7	plptE6; bamA6	0.078	31.25
8	plptE6; yfgL−	0.078	15.625
9	bamA6	2.5	500
10	yfgL−	0.625	250
11	lptD7	5	1,000
12	bamA8	5	1,000

lptE6 Affects the Interaction Between LptE and LptD.

The different kinds of LptE mutant proteins we expected to identify through the screen are those in which i) structural integrity of the protein is compromised, ii) interaction with LptD is disrupted, iii) interaction with LPS is disrupted, iv) interactions with other components of the Lpt pathway (LptABCFG) are affected, or v) interactions with unidentified pathway members are affected. Because mutants belonging exclusively to class i will be of limited utility for structure-function analyses, we first determined the effect of the two–amino-acid deletion in LptE6 on protein stability.

We observed previously that the effects of mutations that alter the LptD/E complex on protein stability are more pronounced in stationary phase (27). When we examined the stationary-phase levels of LptE in the mutants using SDS/PAGE and immunoblotting, we observed that plasmid-produced LptE6 is destabilized compared with plasmid-produced LptE⁺ (Fig. 1). The mutant protein levels are greater than those of LptE produced from the native chromosomal locus and these levels, in turn, are similar to those produced by plptE^low. The permeability defects of plptE6, however, are much greater than those caused by lowering lptE synthesis in the plptE^low strain. Therefore, the phenotypic defects caused by lptE6 cannot be solely attributed to lowered levels due to structural destabilization and must involve a functional aspect.

Fig. 1.

Levels of LptE and LptD. Whole-cell samples were obtained from AM604, AM604 ΔlptE (plptE), AM604 ΔlptE (plptE^low), AM604 ΔlptE (plptE6), AM604 ΔlptE lptD7 (plptE6), and AM604 ΔlptE bamA8 (plptE6) cells that were grown to stationary phase and subjected to SDS/PAGE and immunoblotting for LptE (Top), LptD (Middle), and DegP (Bottom). The asterisk indicates degradation products of DegP.

It is not possible to overproduce LptD unless LptE is also simultaneously overproduced (19). This result suggests that some mutations in lptE could affect LptD levels. To determine how the mutations in lptE impact LptD levels, we examined the stationary-phase levels of LptD and found that LptD is destabilized in the plptE6 mutant (Fig. 1, Middle, lanes 2 and 4). Because LptE6 is present at higher levels than LptE produced from the native chromosomal locus (Fig. 1, lanes 1 and 4), the loss of LptD stability in the plptE6 mutant cannot be due to a simple destabilization of the lipoprotein; instead, we posited that an interaction between LptE6 and LptD was affected, leading to a decrease in the structural stability of LptD.

To determine whether lptE6 affects an interaction between LptE and LptD, we constructed a strain in which we could test whether LptE6 has a lower affinity for LptD than does LptE⁺. LptE⁺ and LptE6 were C-terminally His-tagged by cloning the genes into the pET23/42 vector (12). To get comparable levels of LptE-His and LptE6-His, the start codon of lptE-his was changed from ATG to TTG (Fig. 2). When these plasmids are present in a wild-type strain, chromosomally produced LptE competes for LptD, lowering the amount of LptD that can copurify with the His-tagged lipoprotein. In samples from such wild-type lptE strains, LptD copurifies with LptE-His, but not with LptE6-His (Fig. 2). This result demonstrates that the lptE6 mutation directly affects the interaction between LptD and LptE.

Fig. 2.

The lptE6 mutation directly affects the interaction between LptE and LptD. Ni-NTA affinity purification is shown, using AM604 containing pET23/42 (lane1), pET23/42scT-lptE-his (lane 2), or pET23/42lptE6-his (lane 3). Samples before (whole-cell lysates) and after (Ni-NTA eluates) purification were subjected to immunoblotting using anti-LptD and anti-His antibodies. Positions of relevant molecular markers are indicated in kilodaltons.

lptE6 Impairs LptD Biogenesis.

LptD has two intramolecular disulfide bonds and mature, oxidized LptD (LptD_OX) migrates slower than reduced LptD (LptD_RED) during SDS/PAGE (21, 22). Unfolded LptD_RED runs at ∼83 kDa and LptD_OX runs at ∼120 kDa. When LptE is depleted, LptD_OX can no longer be detected even though LptD_RED is present at wild-type levels. Therefore, the interaction with LptE is required for the proper oxidation of LptD (22). Because LptE6 has lowered affinity for LptD, we wanted to determine if LptD biogenesis is impaired in the plptE6 mutant, using the oxidation state of LptD as a diagnostic tool.

It was previously observed that levels of LptD4213, which carries a 23-amino-acid deletion and is severely destabilized, are higher in logarithmic-phase cultures compared with stationary-phase cultures (27). Therefore, to evaluate LptD assembly in the plptE6 mutant, we examined the levels of LptD_RED and LptD_OX in late logarithmic-phase cells. LptD_RED levels in the plptE6 mutant, like those of LptD4213, are higher during the logarithmic phase of growth (Fig. 3, Upper, lanes 1 and 2) but strikingly, most of the LptD present is not properly oxidized (Fig. 3, Lower, lanes 1 and 2). The misfolded protein must be present in some disulfide-bonded aggregate that either does not enter the gel or runs as a diffuse smear. This result demonstrates an LptE6-dependent LptD biogenesis defect.

Fig. 3.

The suppressors of lptE6 improve LptD assembly. Whole-cell samples from AM604 ΔlptE (plptE), AM604 ΔlptE (plptE6), AM604 ΔlptE lptD7 (plptE6), and AM604 ΔlptE bamA8 (plptE6) grown to an OD₆₀₀ of ∼1.0 were boiled in the presence (Upper) or the absence of β-ME (Lower) and Western blot analysis of LptD was carried out to detect reduced LptD (LptD_RED) or oxidized LptD (LptD_OX), respectively.

To determine whether the biogenesis defect is at the level of complex stability, complex formation, or both, we compared the stability of the LptD/E6 complex with that of the wild-type complex using our previously described methods (19). The wild-type LptD/E complex is very protease resistant and it does not dissociate upon SDS/PAGE provided the samples are not heated. In both types of experiments, the LptD/LptE6 complex behaves similarly (Fig. S1). These results suggest that the primary defect caused by LptE6 is in complex formation.

Mutations in lptD and bamA Suppress lptE6.

We hoped to genetically identify the regions of LptD that interact with LptE by finding conformational suppressor mutations of lptE6 in lptD that restore the barrier function of the OM. To select for suppressors, we used bacitracin instead of rifampicin to avoid the high number of antibiotic-resistant mutations we would isolate in rpoB (29). Specifically, we used bacitracin at a concentration normally lethal to the plptE6 mutant strain, but not to the plptE⁺ strain. Suppressors arose spontaneously at a frequency of 8 × 10⁻⁸.

A total of 11 independent suppressors were characterized. On the basis of colony morphology, these suppressors could be divided into two classes. The 3 suppressor strains in class I form small colonies, whereas the 8 suppressor strains in class II form normal-sized colonies. Genetic mapping showed that the class I suppressor mutations were linked to lptD, whereas the class II mutations were linked to bamA. bamA encodes a conserved integral membrane protein that is an essential component of the machinery involved in assembling β-barrel OMPs into the OM (30).

Allele Specificity of the lptE6 Suppressors.

DNA sequence analysis of the chromosomal lptD gene in the class I suppressors revealed that all three independent members of this class carry the same AT:GC transition mutation at codon 633, which results in a D633G substitution. This allele is called lptD7. Asp633 lies in the predicted β-barrel domain of LptD (19).

DNA sequence analysis showed that all eight independent suppressor strains in class II harbor mutations in bamA. Strikingly, all of them alter codon 159. Five suppressors have a P159L change and there is one suppressor each with a P159S, P159A, or P159Q change. Because all of the suppressors behave similarly, we went forward with the analysis of just the P159L mutant in this work and the allele is called bamA8. Pro159 is in the second polypeptide transport-associated (POTRA) domain of BamA. BamA has five periplasmic POTRA domains and they are thought to template β-strand formation during the folding of substrates (31, 32).

Both lptD7 and bamA8 improve the barrier function of the plptE6 mutant to similar extents, although not to wild-type levels, determined by a two- to fourfold increase in resistance to rifampicin and bacitracin (Table 1, rows 4–6). The fact that we repeatedly found suppression by mutating the same two codons suggests that the suppression is allele specific. Although we do not have a large collection of lptE mutants to combine with lptD7 and bamA8 and exhaustively test for allele specificity, we did check whether the suppressors of lptE6 could suppress lptD4213 and vice versa. bamA8 does not suppress the MacConkey sensitivity defect of lptD4213. Similarly, bamA6 (formerly yaeT6), which carries a two–amino-acid insertion in POTRA3 of BamA and suppresses lptD4213 (28), does not suppress the bacitracin sensitivity of lptE6. Null mutations in bamB can suppress lptD4213 under conditions of slow growth (27). However, a bamB null allele does not suppress permeability defects in the plptE6 mutant. In fact, both bamB⁻ and bamA6 aggravate the defects conferred by lptE6 (Table 1, rows 7–10). Therefore, the suppressors of lptE6 work by correcting a specific problem and not by ameliorating generalized OM defects.

lptE6 Suppressors Improve LptD Assembly.

The isolation of a suppressor mutation that affects a POTRA domain of BamA strongly suggested that the LptD assembly defect conferred by lptE6 might be corrected by the suppressor mutation. Examination of LptD_OX levels in the suppressor strains during the late-logarithmic growth phase showed that there is more correctly oxidized LptD in the presence of either bamA8 or lptD7 (Fig. 3, Lower, lanes 2–4). Levels of LptD_RED are comparable in the plptE6 mutant and the suppressor-containing strains under these experimental conditions (Fig. 3, Upper, lanes 2–4). This result confirms that the difference in the levels of LptD_OX is a result of improved LptD assembly and not just a reflection of total protein levels.

When we examined the levels of LptD_RED in stationary phase, we found that LptD levels are higher in the suppressor strains (Fig. 1, Middle, lanes 4–6). lptD is regulated by σ^E, a transcription factor that responds to envelope stress (33). Because defects in OMP biogenesis and LPS biosynthesis can induce σ^E (34, 35), there is a possibility that the suppressor mutations in bamA and lptD result in increased LptD synthesis due to σ^E induction. However, the levels of a member of the σ^E regulon, the periplasmic protease DegP (36), are not increased in the suppressor-containing strains relative to the plptE6 mutant (Fig. 1, Bottom, lanes 4–6), ruling out further σ^E induction by the suppressor mutations. In fact, on the basis of the amounts of proteolyzed DegP (37), the suppressor mutations lower the σ^E stress response that is induced in the plptE6 mutant. The increased levels of LptD in the presence of the suppressor mutations in stationary phase, therefore, reflect an increase in LptD stability due to improved LptD/E biogenesis.

Phenotypic Characterization of lptE6 Suppressors.

lptD7 and bamA8 do not cause any permeability defects in an otherwise wild-type background (Table 1, rows 11 and 12). The bamA8 strain has wild-type levels of LptD and LptE, whereas the lptD7 strain appears to have slightly lowered levels of both LptD and LptE (Fig. 4). The reduced presence of the LptD/E complex at the OM in the lptD7 strain probably leads to induction of the σ^E stress response as seen by the up-regulation of DegP (Fig. 4). The D633G substitution in LptD, therefore, confers LptD/E complex instability in combination with LptE⁺ but has the opposite effect in combination with LptE6.

Fig. 4.

Effects of lptD7 and bamA8 on LptD and LptE levels. Western blot analysis is shown of LptD and LptE and DegP on whole-cell samples prepared from AM604, AM604 lptD7, and AM604 bamA8 after overnight growth. The asterisk indicates degradation products of DegP.

Discussion

The selection used in this study to isolate suppressors of lptE6 was quite demanding. The lptE6 mutation is a deletion; thus, reversion is not possible. In addition, there are no chromosomal mutations that would confer resistance to the antibiotic used, bacitracin, under the conditions used. Consequently, mutants that survive selection are rare and all of them contain bona fide suppressors. All of these suppressors fall into one of only two phenotypic classes, and each class carries alterations in one particular codon in one of two genes. Because these suppressor mutations have been isolated multiple times, we believe that the selection is saturated. Because of this striking specificity, we suggest that an understanding of these suppressors will provide clear insights into the nature of the defect caused by lptE6.

LptE is known to form a stable complex with the β-barrel protein LptD with a stoichiometry of 1:1 (19). As shown by Freinkman et al., LptE has an extensive interaction surface with LptD, suggesting that LptE is deeply embedded in the barrel (38). Although the residues altered in LptE6 do not directly contact LptD (38), LptE6 destabilizes LptD, suggesting that this mutation weakens an interaction between the two proteins. The suppressor allele lptD7, isolated three independent times, results in a single-amino-acid substitution in the β-barrel portion of LptD. The specificity of this suppressor mutation suggests that we have indeed found a classic interactive suppressor. The D633G mutation in lptD7 is either restoring a lost contact between LptE6 and LptD or relieving a constraint imposed by the YPISA to YRA change caused by lptE6. Because the lptE6 mutation does not destabilize the mature LptD/E complex, we suggest that these two mutations alter an interaction that is important for complex assembly. This hypothesis could also explain the phenotypes of lptD7 in an otherwise wild-type background. This mutation likely alters the assembly process with wild-type LptE.

LptE and LptD are targeted to the OM by different pathways (Fig. 5B) and they must interact to form a stable, properly oxidized complex at the OM. LptD must be at least partially folded to bring the Cys residues in the N- and C-terminal domains into close proximity for disulfide bond formation and interaction with LptE is required for this process (22). Suppressor mutations of lptE6 that improve LptD biogenesis were also found in bamA, which encodes the essential β-barrel component of the machinery involved in assembling β-barrel OMPs such as LptD into the OM. Eight independent selections yielded Ser, Leu, Ala, or Gln substitutions at the same residue, Pro159. The varied nature of the substitutions suggests that the loss of the Pro residue is vital for the suppression.

Fig. 5.

Genetic interactions between lpt and bam. (A) Genetic interactions (dashed lines) between genes encoding members of the Lpt and Bam complexes exist. lptD4213 is suppressed by mutations in bamA and bamB (27, 28). lptE6 is suppressed by mutations in lptD and bamA. Physical interactions in the LptDE (12) and BamABCDE (30, 40) complexes are shown with solid lines. (B) LptE and LptD are targeted to the OM by different pathways. Lipoprotein LptE is targeted to LolB by the Lol pathway (26). OMP LptD is targeted to the BamABCDE complex (24). Defects in LptD assembly caused by the presence of LptE6 are suppressed by mutations in the gene encoding BamA, suggesting that LptD and LptE interact at the Bam complex during biogenesis of the LptDE complex.

The astonishingly specific genetic interaction identified here between lptE and bamA suggests that LptE interacts with the rudimentary barrel domain of LptD while it is being held in the Bam complex. The suppressor mutation in bamA8 is perhaps able to bias the folding of LptD such that the interaction between LptE6 and LptD is strengthened. LptE could participate in LptD folding by either acting as a scaffold or stabilizing the newly folded protein. In either case, completion of disulfide bond formation in LptD must take place at the Bam machine or after the LptD/E complex is released from the machine.

The lptD gene was first identified in a genetic selection for mutations that disrupt the OM permeability barrier. One mutation that answered this selection, lptD4213 (formerly imp4213), renders E. coli sensitive to a wide variety of antibiotics and detergents (39). As summarized in Fig. 5A, components of the Bam complex, which assembles OMPs (30, 40), were identified by mutations that suppress the permeability defects caused by lptD4213, using a genetic strategy termed chemical conditionality (27, 28, 41). Basically, it was shown that different toxic small molecules could be used to identify suppressors that correct the lptD4213 defects to varying degrees. Suppressor phenotypes correlated with the physical properties of the small molecules and the suppressor mutations identified the structural genes for BamA and BamB. Why mutations that alter the complex that assembles LPS at the cell surface could be suppressed by mutations that alter the complex that assembles OMPs was not clear. One possible explanation suggests that the activity of both machines must be balanced. Another possibility suggests that the activity of one machine may be regulated by directed interaction with the other machine.

Results presented here provide a more attractive explanation for chemical conditionality (Fig. 5B). We started out with a defect in LptE, the lipoprotein component of the LPS assembly machinery, and found a suppressor mutation that alters BamA, a part of the OMP assembly machinery. The permeability defects caused by the mutation in LptE are due to defects in the biogenesis of the β-barrel component of the LPS assembly machinery, LptD, and suppressor mutations in lptD and bamA both fix this assembly defect. In analogous fashion, we suggest that the lptD4213 mutation also causes a defect in the assembly of LptD and suppressors that alter components of the Bam complex address this problem. Different suppressors that correct the OM permeability defect to varying degrees do so by fixing the LptD/E assembly defect to varying degrees.

Our results argue that lptD4213, lptE6, and lptD7 all affect assembly of the LptD/E complex. Using a completely different, but complementary approach, Freinkman et al. (38) have identified another deletion in the β-barrel domain of LptD that affects assembly as well. We suggest that suppressors will be useful not only for identifying important genes but also for probing the mechanism of β-barrel assembly.

Materials and Methods

Bacterial Strains and Growth Conditions.

All strains (Table S1), plasmids, growth conditions, and antibiotic sensitivity assays are described in SI Materials and Methods.

Construction of the ΔlptE2::kan Allele.

Primers ACM143 and GC28 (Table S2), with homologies to regions upstream and directly downstream of lptE, were used to amplify the kanamycin cassette from pKD4 (42) by PCR. The resulting PCR product was used to replace the entire lptE gene with the kanamycin cassette in a recombineering reaction (43) to obtain the ΔlptE2::kan allele. Insertion of the kanamycin cassette was verified by PCR.

PCR Mutagenesis of lptE.

Random mutations were generated by error-prone PCR using the GeneMorph kit (Stratagene). Primers ACM146 and ACM147 (Table S2) were used at 125 ng each with 77 ng pBAD18lptE (12) as the template in a 50-μL reaction. The PCR conditions were 95 °C for 4 min followed by 30 cycles of 95 °C for 30 s, 51 °C for 30 s, and 72 °C for 1 min followed by a final extension for 10 min at 72 °C. The product was digested with EcoRI and XbaI (New England Biolabs), ligated into similarly cut pBAD18 (44), and transformed into AM604. Transformants were selected on LB agar with ampicillin.

Identification of Mutant lptE Plasmid Constructs.

To replace the endogenous lptE in the transformants with ΔlptE2::kan, the transformants carrying the mutagenized lptE on pBAD18 (∼100 per ligation reaction) were pooled and treated as the recipient culture in a P1 transduction reaction in the presence of arabinose. Transductants (∼100 per reaction) were selected on LB agar with kanamycin and patched onto both LB agar with kanamycin and LB agar with rifampicin. Patches that grew in the presence of kanamycin but not rifampicin were purified further on kanamycin and their sensitivities to rifampicin, bacitracin, erythromycin, and novobiocin were measured using the BBL Sensi-Disk Antimicrobial Susceptibility Test Discs (BD) as previously described (27). Candidates that showed significant zones of inhibition around the 6-mm discs were selected for further analysis and the plasmids were sent to Genewiz for DNA sequencing.

Affinity Purification.

Affinity purification experiments were performed as previously described (19).

Suppressor Selection and Mapping.

Overnight cultures of GC190 were grown from single colonies. One hundred microliters (∼10⁸ cells) of each culture were spread on LB agar with bacitracin. Suppressors were allowed to develop spontaneously at 37 °C overnight. Suppressors were purified in the presence of bacitracin and grouped into two classes on the basis of colony size. To find out if the suppressor mutations were in lptD, linkage to carB::Tn10, a marker that is 40% linked to lptD, was calculated. To map the mutation in a representative class II suppressor strain, the strategy outlined in Ruiz et al. (28) was used, using a pool of MC4100 mutants containing random mini-Tn10 insertions to donate wild-type alleles (45). We found that a yafD::miniTn10 allele was 27% linked to the suppressor mutation. Linkage analysis with yafD::miniTn10 showed that all members of class II were most likely carrying a suppressor mutation in bamA. The lptD and bamA genes were sequenced to identify the suppressor mutations in the respective mutants. The suppressor strains were rebuilt to ensure that that the mutations in lptD and bamA were sufficient for suppression.

Immunoblot Analysis.

To analyze the levels of proteins in stationary phase, 250 μL of overnight cultures was pelleted at 13,200 rpm for 1.5 min and resuspended in a volume of SDS sample buffer equal to OD₆₀₀/24. To analyze the oxidation state of LptD, 1:200 dilutions of overnight cultures were grown to an OD₆₀₀ of 1. Either 1 mL or 500 μL of culture was pelleted and resuspended in a volume of SDS sample buffer equal to OD₆₀₀/12 in the presence or absence of 5% (vol/vol) β-ME, respectively. Samples were boiled for 10 min before being subjected to SDS/PAGE. Rabbit polyclonal antisera against LptD (1:7,000 dilution) (21), LptE (1:40,000 dilution) (19), and DegP (1:50,000 dilution) (46) and donkey ECL horseradish peroxidase conjugate anti-rabbit IgG (GE Life Sciences) (1:10,000 dilution) were used for immunoblotting. To visualize the bands, the ECL antibody detection kit (Amersham Pharmacia Biotech) and HyBlot CL film (Denville Scientific) were used. Suitable controls are shown in Fig. S2.