Characterization of the two-protein complex in Escherichia coli responsible for lipopolysaccharide assembly at the outer membrane

Abstract

Lipopolysaccharide (LPS) is the major glycolipid that is present in the outer membranes (OMs) of most Gram-negative bacteria. LPS molecules are assembled with divalent metal cations in the outer leaflet of the OM to form an impervious layer that prevents toxic compounds from entering the cell. For most Gram-negative bacteria, LPS is essential for growth. In Escherichia coli, eight essential proteins have been identified to function in the proper assembly of LPS following its biosynthesis. This assembly process involves release of LPS from the inner membrane (IM), transport across the periplasm, and insertion into the outer leaflet of the OM. Here, we describe the biochemical characterization of the two-protein complex consisting of LptD and LptE that is responsible for the assembly of LPS at the cell surface. We can overexpress and purify LptD and LptE as a stable complex in a 1∶1 stoichiometry. LptD contains a soluble N-terminal domain and a C-terminal transmembrane domain. LptE stabilizes LptD by interacting strongly with the C-terminal domain of LptD. We also demonstrate that LptE binds LPS specifically and may serve as a substrate recognition site at the OM.

The OM of Gram-negative bacteria is a unique asymmetric lipid bilayer in which the outer leaflet is composed almost entirely of LPS and the inner leaflet of phospholipids (PL) (Fig. 1) (1, 2). Building and maintaining this asymmetric bilayer is a challenge for the cell because the OM is located outside the cytoplasm, in an environment that lacks an obvious energy source such as ATP. LPS and PL are synthesized at the cytoplasmic face of the IM whereas lipoproteins and integral membrane proteins are synthesized in the cytoplasm; all these OM components must be transported across the IM and the aqueous periplasmic compartment to be assembled into the OM (3, 4). In the case of LPS assembly, the molecule must ultimately reach the outer leaflet of the OM (Fig. 1) (5, 6). LPS renders the OM a very effective permeability barrier against toxic compounds such as antibiotics and bile salts (7). Whereas many details of the biosynthesis of LPS inside the cell have been established, the subsequent steps of LPS assembly are uncharacterized.

Fig. 1.

LPS transport across the cell envelope (5, 6). The structure of LPS from E. coli K12, which lacks O-antigens, is shown on the Left. LPS is first synthesized as Kdo₂-lipid A or Re-LPS. Addition of inner and outer core sugars gives rise to Ra-LPS. As shown on the Right, after Ra-LPS is completed, ABC transporter MsbA flips the LPS molecule from the inner leaflet to the outer leaflet of the IM. At the periplasmic side of the IM, the O-antigen oligosaccharides can be ligated to Ra-LPS (not shown here). Another ABC transporter consisting of LptB, LptC, LptF, and LptG is thought to extract LPS from the IM by hydrolyzing ATP. LPS is then transported across the aqueous periplasm by LptA via an unknown mechanism. At the OM, LptD, and LptE form a complex that receives and assembles LPS into the surface of the cell. Gal, D-galactose; Glu, D-glucose; Hep, L-glycero-D-manno-heptose; Kdo, 3-deoxy-D-manno-oct-2-ulosonic acid; P, phosphate; PEtN, phosphoethanolamine.

LPS (Ra-LPS) is synthesized at the cytoplasmic side of the IM and then flipped across the bilayer to the periplasmic leaflet by an ATP-binding cassette (ABC) transporter known as MsbA (8) (Fig. 1). Current models propose that another ABC transporter consisting of LptB, LptC, LptF, and LptG (formerly YhbG, YrbK, YjgP, and YjgQ, respectively) extracts LPS from the IM, initiating the transport and assembly of LPS to its final destination (9–12). Because LPS is amphipathic, it is unlikely to diffuse across the periplasm unassisted. LptA (formerly YhbN), a periplasmic protein, is thought to mediate LPS transit across this aqueous compartment by an unknown mechanism (9, 10, 13–15). Once LPS arrives at the OM, a protein complex consisting of LptD and LptE (formerly Imp and RlpB, resp.) is required for LPS assembly at the cell surface establishing bilayer asymmetry (16). It is not known what individual roles LptD and LptE play in this final stage of LPS assembly and no biochemical characterization of this complex has been reported.

LptD is an essential outer membrane protein (OMP) that is highly conserved in Gram-negative bacteria (17, 18). It is an ∼87 kDa protein that is predicted to assume a β-barrel conformation (18). Indeed, LptD is inserted into the OM by the β-barrel assembly machine (the Bam complex, formerly the YaeT complex) (16, 19–23). LptD was initially shown in Escherichia coli to be required for cell envelope biogenesis (18) and later demonstrated in Neisseria meningitidis to be specifically involved in LPS assembly (24). In N. meningitidis, LptD and LPS are not essential (24, 25). LPS molecules produced in the lptD knockout mutant are not assembled correctly on the cell surface as they are not accessible to LPS-modifying enzymes expressed in the OM or added to extracellular medium (24). These results implicated LptD in the translocation of LPS to the outer leaflet of the OM. In E. coli, LptD was shown to interact physically with the essential ~20 kDa lipoprotein LptE that is anchored to the OM via a lipid moiety attached to the N-terminal cysteine residue (16). Depletion of either LptD or LptE prevents newly-synthesized LPS from reaching the outer leaflet of the OM, suggesting that they are both required for LPS assembly at the cell surface.

We have overexpressed and purified LptD and LptE as a stable heterodimeric protein complex. We have defined a soluble N-terminal portion of LptD predicted to be periplasmic and a C-terminal transmembrane fragment likely composed of the β-barrel (3). We show that the C-terminal domain of LptD interacts strongly with LptE to form the β-barrel fold and demonstrate that LptE interacts specifically with LPS.

Results

LptD and LptE Form a Stable 1∶1 Complex in Vitro.

Both LptD and LptE have been characterized as low abundance proteins in E. coli (18, 26, 27). While it was not possible to overexpress LptD alone in E. coli in its native folded state, we were able to simultaneously overexpress and purify wild-type LptD together with a functional C-terminally His-tagged LptE (LptE-His) and obtain sufficient quantities of the LptD/E complex for biochemical characterization. Overexpressed proteins that are properly targeted were solubilized from fractions containing the OM. The protein complex was purified first on a Ni-NTA column and then by size-exclusion chromatography (SEC). Both proteins eluted together as one single peak. LptD and LptE-His purified in this manner are post-translationally modified correctly, as revealed by MALDI-MS (Fig. S1). LptD (∼87 kDa) has previously been shown to migrate as a higher-molecular-weight species due to the formation of disulfide bonds (16, 18). SDS-PAGE analysis of the purified LptD/E-His complex under non-reducing conditions showed two major bands at ∼100 kDa and ∼19 kDa when the sample was heated (Fig. 2A, Left, and + Heat). These bands correspond to LptD and LptE-His, resp., as judged by immunoblots (Fig. 2A, Middle, and Right). The observed migration for overexpressed and purified LptD is consistent with previous reports (16, 18).

Fig. 2.

LptD and LptE form a stable complex in vitro. (A) SDS-PAGE analysis of purified LptD/E-His complex under nonreducing conditions. Samples were heated or not for 10 min at 100 °C, as indicated. The protein bands were visualized either by staining with Coomassie-blue (Left) or immunoblot using antibodies against LptD (Middle) or a hexahistidine-tag (Right). (B) BN-PAGE analysis of purified LptD/E-His complex under nonreducing conditions. Positions of relevant molecular weight markers are indicated in kDa.

LptD and LptE-His co-migrate in a denaturing gel, suggesting that they form a very stable complex. In the absence of heating, SDS-PAGE analysis of the purified LptD/E-His complex (Fig. 2A, Left, and - Heat) showed a major band at ∼63 kDa. Immunoblots revealed that both LptD and LptE-His could be detected in this ∼63 kDa band (Fig. 2A, Middle, Right, and - Heat). The fast mobility of the complex (∼63 kDa without heating versus ∼100 kDa with heating) could be due to the formation of a β-barrel conformation in LptD (28, 29) or the binding of LptE-His or both.

The ability to overexpress and purify the LptD/E complex allows determination of stoichiometry. LptD and LptE-His migrate together as a stable complex just below the 140 kDa marker on non-denaturing blue-native (BN)-PAGE (Fig. 2B). When this band was excised and subjected to a second-dimension SDS-PAGE analysis, both LptD and LptE-His were present. Quantitative amino acid analysis of the respective bands obtained from the SDS-PAGE gel revealed that the stoichiometric ratio of LptD to LptE in the complex is 1∶1 (Fig. S2) (30). Taken together, these results indicate that the purified LptD/E-His complex is a heterodimer (∼107 kDa) containing one copy of LptD (∼87 kDa) and LptE (∼20 kDa).

The C-Terminal Domain of LptD and the Core Structure of LptE Constitute a Trypsin-Resistant Fragment of the Protein Complex.

To characterize the structure of this complex, purified LptD/E-His was treated with trypsin and analyzed by BN-PAGE (Fig. 3A) and SEC (Fig. 3B). Interestingly, a major part of the trypsin-digested complex remained intact, migrating as one band in BN-PAGE and eluting as a single peak in SEC. SDS-PAGE analysis of the peak from SEC of the digested complex indicated that this trypsin-resistant fragment contains both LptD and LptE (Fig. 3C). The fact that both proteins display trypsin-resistance suggests that they are tightly folded.

Fig. 3.

The C-terminal domain of LptD and the core structure of LptE constitute a trypsin-resistant fragment of the protein complex. (A) BN-PAGE analysis of purified LptD/E-His complex before and after limited trypsin treatment under nonreducing conditions. (B) Size-exclusion chromatography of purified LptD/E-His complex before (Red) and after (Blue) limited trypsin treatment under nonreducing conditions. Number beside each peak represents elution volume. Difference in elution volumes corresponds to a mass shift of ∼35 kDa. (C) SDS-PAGE analysis of peak fractions from SEC in (B). Edman sequencing revealed that the LptD truncation is N-terminally cleaved at F203, which contains the predicted β-barrel domain of LptD. Bands labeled * and ** are further N-terminal truncation products from LptD (at N470) and LptE (at K 70), resp. Positions of relevant molecular weight markers are indicated in kDa.

The trypsin-resistant fragment of the LptD/E-His complex was separated into four major bands of ∼62 kDa, ∼30 kDa, ∼16 kDa, and ∼12 kDa by SDS-PAGE (Fig. 3C). N-terminal Edman sequencing revealed that the ∼62 kDa and ∼30 kDa bands are N-terminal truncations of LptD at residues F203 and N470, resp. LptD is predicted to contain an N-terminal periplasmic OstA domain (Pfam accession number PF03968, a.a. 6–195, numbering includes signal sequence (a.a. 1–24)) and a C-terminal β-barrel OstA_C domain (Pfam accession number PF04453, a.a. 308–697) (3, 31, 32). The F203 truncation, therefore, contains the predicted β-barrel domain. The two lower bands (∼16 kDa and ∼12 kDa) are truncations of LptE-His. Both bands can be immunoblotted by an α-LptE antibody but not by an α-His antibody (Fig. S3A and B). This suggests that both bands contain C-terminal truncations. Multiple attempts to sequence the N terminus of the ∼16 kDa band were unsuccessful, suggesting that the N-terminal lipidated cysteine residue of LptE is still intact. However, the ∼12 kDa band could be sequenced and was shown to be N-terminally truncated at K70.

We performed multiple sequence alignment of LptE from E. coli with its orthologs from N. meningitidis (33), Shewanella oneidensis (34), and Nitrosomonas europaea (35), whose structures have been solved (Fig. S4). These orthologs all contain a common folded core structure; E. coli LptE contains an additional 28 amino acids at the C terminus of this predicted core. Trypsin cleavage at either R163 or R168 in this C-terminal extension of E. coli LptE could give a truncation product that is consistent with the ∼16 kDa band. Therefore, we believe that the core structure of LptE is still preserved in the ∼16 kDa fragment.

The change in retention time for the trypsin-digested complex observed in SEC, relative to the untreated complex (∼107 kDa), corresponds to a loss of ∼35 kDa in mass (Fig. 3B). This mass shift is consistent with the loss of the N-terminal periplasmic domain of LptD (a.a. 25–202, ∼25 kDa) and the short C terminus of LptE-His (∼4 kDa). The resulting trypsin-resistant fragment, therefore, contains both the ∼62 kDa C-terminal domain of LptD and the ∼16 kDa core structure of LptE. Although smaller fragments of LptD and LptE (∼30 kDa and ∼12 kDa, resp.) were also generated, we did not observe additional peaks on SEC or bands on BN-PAGE. This suggests that tryptic peptides that resulted from any further digestion of LptD or LptE in the initial trypsin-resistant fragment must still remain bound to the complex. It appears that the secondary structure of the trypsin-resistant fragment of the LptD/E complex is very stable.

While it seems reasonable that the β-barrel secondary structure predicted in LptD would be protease-resistant, we are intrigued by the observation that LptE is resistant to trypsin digestion in the presence of LptD. Trypsin treatment of LptE-His alone resulted in complete degradation (Fig. S3C), even though evidence suggests that this protein is properly folded (vide infra). Because orthologs of LptE from other species can stably fold in the absence of LptD (33–35), this indicates that the C-terminal domain of LptD protects LptE from trypsin digestion.

The C-Terminal Domain of LptD Scaffolds LptE but Is Not Sufficient for Viability.

We have shown that LptE interacts very strongly with the C-terminal domain of LptD in vitro. To determine whether this interaction is relevant in a cellular context, we performed affinity purification experiments using wild-type cells exogenously expressing His-tagged versions of the N-terminal domain (a.a. 25–203, N-LptD-His), C-terminal domain (a.a. 203–784, C-LptD-His), or full-length LptD (a.a. 25–784, LptD-His) (Fig. 4A) and detected the amount of LptE that co-purified with each construct by immunoblot using the α-LptE antibody. All the constructs contain the native LptD signal sequence (a.a. 1–24) to target them to the cell envelope. The level of N-LptD-His expressed under these growth conditions (37 °C) was much lower than that of C-LptD-His or LptD-His, suggesting that N-LptD may not be stable in the absence of C-LptD. Therefore, no conclusions can be drawn for this domain. In agreement with our in vitro data, however, C-LptD-His pulled down as much LptE as full-length LptD-His (Fig. 4A). These data suggest that N-LptD is not required for interaction with LptE.

Fig. 4.

The C-terminal domain of LptD alone scaffolds LptE at the OM but is not sufficient for viability. (A) Ni-NTA affinity purification using wild-type cells (MC4100, Lane 1) containing pET23/42lptD-His (Lane 2) or pET23/42C-lptD-His (Lane 3). Samples were submitted to immunoblots using α-LptE antibody. The α-LptE antibody cross-reacts with the hexahistidine tag and allows for the visualization of LptD-His species (SI Text). Positions of relevant molecular weight markers are indicated in kDa. (B) Growth curves of the LptD depletion strain NR1072 containing pBAD33 (Empty Vector), pBAD33lptD (wt LptD), or pBAD33C-lptD (C-LptD) grown in the absence of IPTG (inducer for expression of LptD from P_T7lacO-imp) as measured by optical density.

Both LptD and LptE are essential for viability in E. coli (16–18). Because C-LptD is sufficient to scaffold LptE, we asked if C-LptD alone is able to support growth in the absence of wild-type LptD. The effects of wild-type LptD depletion were examined by monitoring optical density as a function of cell growth in strains expressing either C-LptD or full-length LptD (Fig. 4B). Only full-length LptD is able to restore growth under these conditions. In fact, we were also unable to knock out the chromosomal copy of lptD while expressing both N-LptD and C-LptD in trans (Table S1). These results allow us to conclude that C-LptD alone is not sufficient for viability and that the N and C termini of LptD cannot function if separated.

LptE Is Functional Without Its N-Terminal Lipid Anchor.

LptE is anchored to the inner leaflet of the OM via an N-terminal lipid moiety. To understand the role of this lipid anchor in bringing the LptD/E complex together in vivo, we engineered a plasmid in which the coding sequencing of LptE, excluding its signal sequence and the N-terminal lipidated cysteine (a.a. 1–19), is placed after the sequence encoding the pelB leader peptide. This plasmid constitutively expresses a soluble version of LptE in the periplasm that is no longer lipidated at the N terminus.

We first examined whether this delipidated LptE-His (dLptE-His) is still functional by trying to knock out the chromosomal copy of lptE in the presence of the mutant (Fig. 5A). Remarkably, the lptE∷kan allele could be easily introduced and the resulting strain showed no growth defects. We then carried out affinity purification experiments using wild-type cells exogenously expressing dLptE-His to demonstrate that dLptE still interacts with LptD (Fig. 5B). We found that we could detect much less dLptE-His than LptE-His by immunoblot. We believe that this could be due to an artifact produced by the immunoblot procedure, as has been observed for OM lipoprotein LolB (36). Interestingly, LolB has also been reported to be functional without its lipid anchor (36, 37). Here, we demonstrate that the lipid anchor of LptE is not required for the interaction between LptE and LptD and, therefore, conclude that dLptE is functional.

Fig. 5.

Delipidated LptE is functional. (A) Growth curves of the MC4100 ara⁺ΔlptE ∷ kan strains containing pET23/42lptE-His (wt lipidated LptE-His), or pET23/42dlptE-His (delipidated LptE-His) as measured by optical density. (B) Ni-NTA affinity purification using wild-type cells (MC4100, Lane 1) containing pET23/42lptE-His (Lane 2) or pET23/42dlptE-His (Lane 3). Samples were submitted to immunoblots using α-LptD and α-His antibodies. Positions of relevant molecular weight markers are indicated in kDa.

LptE Binds LPS Specifically.

LptD and LptE are involved in the translocation of LPS across the OM and one or both of these proteins could recognize LPS molecules. The C-terminal domain of LptD is a predicted transmembrane β-barrel in the OM and is expected to associate with LPS (38). N-LptD and LptE are soluble domains of the LptD/E complex so we wondered whether either of them could be involved in binding LPS. To address this question, we overexpressed and purified N-LptD-His and dLptE-His and tested for LPS binding. Here, we were able to optimize conditions to allow high level overexpression of periplasmic N-LptD-His in E. coli BL21(λDE3) (Materials and Methods). Purified His-tagged protein was incubated with LPS and enriched on a Ni-NTA column. LPS molecules that remained bound to the His-tagged protein after extensive washing were visualized by silver staining after SDS-PAGE. LPS binding to dLptE-His was observed but no LPS binding to N-LptD-His could be detected in our assay (Fig. 6A). We cannot draw any conclusions about LPS interaction with N-LptD, in part because we do not know if the purified protein is properly folded. As a control, we showed that delipidated His-tagged LolB (dLolB-His), another protein known to bind lipids (displayed at the N-termini of OM lipoproteins), does not bind LPS (37). Furthermore, we demonstrated that dLptE-His binds different forms of LPS including Ra-LPS, Re-LPS and lipid A (Fig. S5). This suggests that LptE binds LPS by recognizing mainly the hydrophobic lipid A portion.

Fig. 6.

LptE interacts with LPS in a specific manner. (A) LPS binding assay using purified delipidated LolB (dLolB-His, Lanes 3 and 4), delipidated LptE (dLptE-His, Lane 5 and 6) or N-terminal periplasmic domain of LptD (N-LptD-His, Lanes 7 and I). The respective proteins were incubated with (Even Lane) or without (Odd Lane) E. coli K235 LPS and enriched on Ni-NTA columns. Eluates were analyzed by SDS-PAGE followed by Coomassie Blue (for protein) or Silver Staining (for LPS). Lane 1: pure LPS; Lane 2: no protein control. (B) Saturation binding curves. Purified dLptE-His was incubated with Ra-LPS (Top) or Re-LPS (Bottom) at increasing concentrations. The amounts of LPS bound to dLptE-His enriched on Ni-NTA columns were determined by SDS-PAGE/silver staining followed by densitometry. Quantification was performed using ratio standard curves (47). (C) Ligand titration experiment. Purified dLptE-His was incubated with a fixed amount of Re-LPS and increasing concentrations of Ra-LPS (Left) or cardiolipin (Right). The amounts of Re-LPS bound to dLptE-His enriched on Ni-NTA columns were determined by SDS-PAGE/silver staining. The amounts of bound Ra-LPS (where applicable) stayed constant as saturating amounts were used.

To establish that LPS binding to dLptE-His is specific, we examined whether binding is saturable and competitive. The amount of LPS bound to dLptE-His enriched on Ni-NTA increases steadily with increasing LPS concentrations but reaches a constant amount at higher LPS concentrations (Fig. 6B). Re-LPS binding to LptE can be competed using a different form of LPS (Ra-LPS, Fig. 6C). Furthermore, [³²P]-labeled Re-LPS bound to dLptE-His can be titrated away with saturating amounts of non-labeled Re-LPS (Fig. S6). In contrast, cardiolipin, a phospholipid composed of similar chemical functionalities, is unable to compete effectively with Re-LPS for binding to dLptE-His (Fig. 6C). These important controls establish the structural basis of LPS binding by LptE; the core carbohydrate portion of LPS is not critical for recognition, and LptE discriminates against structurally distinct E. coli phospholipids. Taken together, these results indicate that purified dLptE-His is properly folded and that LptE is a specific recognition site for LPS at the OM.

Discussion

We have characterized the LptD/E protein complex that is responsible for LPS assembly at the OM of Escherichia coli (16). LptD can be overexpressed with LptE. The two proteins form a stable heterodimeric complex. Because LptD cannot be overexpressed alone, LptE must play a structural role in stabilizing the folded state, or facilitate the folding and assembly of LptD. LptD contains an N-terminal soluble domain (a.a. 25–202) that is likely periplasmic and a C-terminal transmembrane domain predicted to be a β-barrel (a.a. 203–784). We show that LptD interacts very strongly with LptE through its C-terminal domain. This strong interaction may explain why the N-terminal lipid anchor of LptE is not required for function in vivo. We also demonstrate that purified LptE binds LPS specifically, suggesting that it is properly folded.

Purified LptE alone is susceptible to proteolytic degradation but resistant when in complex with LptD. That LptD protects LptE from proteolytic degradation suggests that LptE may also be important for the β-barrel fold in the protein complex. LptE may bind within the β-barrel of LptD. The reported structures of LptE orthologs show that the folded state of these proteins is composed of a four-stranded curved β-sheet with the concave face capped by two amphipathic α-helices (33–35). LptE and the C-terminal domain of LptD could form the β-barrel conformation of the complex together such that LptE provides part of the necessary β-strands to complete the barrel structure. Alternatively, and perhaps more likely, LptE may form a plug that sits at the base of, or is somewhat buried within, the lumen of the β-barrel formed by the C-terminal domain of LptD. Similar proteolytic protection has been observed for plug domains of the P pili assembly usher PapC (39, 40) and the OM iron-chelate transporter FhuA (41), although, in these cases, the respective plug domain is part of the same polypeptide as the β-barrel.

LptE does not simply play a structural role in LPS assembly at the OM. The fact that LptE binds LPS suggests that LptE may be the LPS recognition site at the OM in vivo, receiving LPS from the periplasm. Interestingly, the plug domains of FhuA (38, 42) and other iron-chelate transporters like FepA (43) and FecA (44, 45) are also known to contribute to the binding of their respective ligands at the extracellular mouth of the β-barrel. In addition, the plug domain of FepA has been shown to bind its ligand even in the absence of the β-barrel domain (46). It is believed that binding of ligand is followed by conformational changes in the transporter that eventually allow the energized transport of the iron-chelate across the lumen of the β-barrel and into the periplasm. A similar sequence of events may take place during LPS transport across the OM. If the plug-and-barrel model is correct, recognition of LPS by LptE could trigger a conformational change in the protein that could be propagated to LptD to allow LPS translocation.

Undoubtedly, there are many steps involved in LPS transport across the periplasm and assembly into the outer leaflet of the OM. In this paper, we have described the structure of the protein complex responsible for LPS assembly at the OM and proposed a function for LptE involved in LPS binding. LptE may receive LPS from the periplasm whereas the transmembrane domain of LptD acts to assemble LPS correctly in the OM. There is no mechanistic information on how LptD might mediate the process of LPS insertion into the outer leaflet. Indeed, there is no evidence that LPS ever even resides at the inner leaflet of the OM. We know that both the N- and C-terminal domains of LptD play distinct roles; the C-terminal domain binds LptE but alone cannot support growth. How the N-terminal domain of LptD participates in the process of LPS assembly may provide the key to ultimately understand how this two-protein OM complex is coordinated with the five other essential proteins that together are responsible for proper targeting and assembly of LPS.

Materials and Methods

Bacterial Strains and Plasmids.

Strains are described in SI Text. Plasmids are listed in Table S2.

Overexpression and Purification of LptD/E-His Complex.

The LptD/E-His complex was overexpressed and purified from outer membranes of BL21(λDE3) harboring pET23/42lptD and pCDFlptE-His. A 10-mL culture was grown from a single colony in LB broth supplemented with 100 μg/mL carbenicillin, 50 μg/mL streptomycin, and 0.2% glucose at 26 °C until OD₆₀₀ ∼ 0.6. This culture was then used to inoculate a 1.5-l culture that was grown at 26 °C until OD₆₀₀ ∼ 0.6. At this time, 0.1 mM IPTG was added and the culture was grown at 26 °C for another 20 h. Cells were pelleted by centrifugation at 5,000  × g for 20 min and then resuspended in 60 ml TBS (20 mM Tris.HCl, 150 mM NaCl, pH 8.0) containing 1 mM phenylmethylsulfonyl fluoride (PMSF, Sigma) and 50 μg/mL DNase I (Sigma). The resuspended cells were lysed by a single passage through a French Press (Thermo Electron) at 16,000 psi. The cell lysate was centrifuged at 5,000  × g for 10 min to remove unbroken cells, and the supernatant was centrifuged at 100,000  × g for 30 min in an ultracentrifuge (Model XL-90). The pellet was extracted with 30 ml TBS/0.5% N-lauroylsarcosine (sodium salt) at 4 °C for 1 h and recentrifuged in the ultracentrifuge as above. The resulting pellet was resuspended in 30 mL of the same buffer and recentrifuged as above. The washed pellet was finally extracted with 18 mL TBS-B (20 mM Tris.HCl, 300 mM NaCl, 20 mM imidazole, pH 8.0)/1% Anzergent 3–14 (Anatrace) containing 100 μg/mL lysozyme (Sigma) at 4 °C for 2–3 h and recentrifuged as above. 1 mL of Ni-NTA resin (Qiagen) pre-equilibrated with TBS-B was added to the final supernatant and incubated at 4 °C for 1 h, with rocking. The mixture was loaded onto a column and allowed to drain by gravity. The filtrate was collected and reloaded onto the column and drained. The column was washed with 2 × 10 mL TBS-B/0.02% Anzergent 3–14 followed by 10 × 2 mL TBS-B/1% n-octylglucoside (OG) (Anatrace) and then eluted with 4 mL of TBS/1% OG containing 200 mM imidazole. The eluate was concentrated in an ultrafiltration device (Amicon Ultra, Millipore, 10 kDa cut-off) to ∼100 μL and further purified by SEC on a prepacked Superdex 200 column (GE Healthcare), using TBS/1% OG as the eluent. 8 μg of protein complex was analyzed by SDS-PAGE and 15 μg was analyzed by BN-PAGE.

LPS Binding Assay.

In a typical binding assay, 15 μg of purified dLptE-His (or molar equivalence of N-LptD-His or dLolB-His) was incubated with 10 μg of smooth LPS, Ra-LPS, Re-LPS or lipid A in 500 μL TBS/1% IP buffer (Sigma). LPS solutions were prepared according to protocol outlined in SI Text. The mixture was incubated at 4 °C for 30 min on a rotary shaker and passed four times through a column containing 100 μL Ni-NTA resin pre-equilibrated with TBS-B. The column was washed with 2 × 1 mL TBS-B/0.05% Triton X-100 (VWR), followed by 2 × 1 mL TBS-B, and then eluted with 300 μL TBS containing 200 mM imidazole. The samples were analyzed by SDS-PAGE. Proteins and LPS were visualized by Coomassie Blue and silver staining (47), resp.

For saturation binding, increasing amounts of Ra-LPS (0, 2.25, 4.5, 9, 22.5, 45, 90, or 180 μg) or Re-LPS (0, 2.5, 5.0, 10, 25, 50, 100, or 200 μg) were used in the above assay. Samples from the same saturation binding experiment were loaded and analyzed on the same SDS-PAGE gel. Proteins and LPS were both visualized by silver staining. The amounts of protein and LPS in each lane were determined by densitometry using ChemiImager (AlphaInnotech) and quantified accurately by using ratio standard curves (48). The relative amount of LPS to protein (in arbitrary units) at each concentration of LPS was plotted. To calculate molar concentrations, the molecular weights of Ra-LPS and Re-LPS were assumed to be 4145 and 2237 Da, resp.

For ligand titration experiments, Re-LPS (10 μg) was incubated with dLptE-His (15 μg) in the presence of increasing amounts of Ra-LPS (0, 93, 186, or 461 μg) or cardiolipin (0, 29, 57, or 143 μg). The binding assays were performed as above, except that the Ni-NTA resin was washed with 4 × 1 mL TBS-B/0.05% Triton X-100. Amounts of Re-LPS bound to dLptE-His were visualized by SDS-PAGE followed by silver staining.

Other Methods.

Other methods are provided in SI Text.