Dissection of an ABC transporter LolCDE function analyzed by photo-crosslinking

Abstract

The envelope of Escherichia coli contains approximately 100 different species of lipoproteins, most of which are localized to the inner leaflet of the outer membrane. The localization of lipoprotein (Lol) system, consisting of five Lol proteins, is responsible for the trafficking of lipoproteins to the outer membrane. LolCDE binds to lipoproteins destined for the outer membrane and transfers them to the periplasmic chaperone LolA. Although the cryo-EM structures of E. coli LolCDE have been reported, the mechanisms by which outer membrane lipoproteins are transferred to LolA remain elusive. In this study, we investigated the interaction between LolCDE and lipoproteins using site-specific photo-crosslinking. We introduced a photo-crosslinkable amino acid into different locations across the four helices which form the central lipoprotein-binding cavity, and identified domains that crosslink with peptidoglycan-associated lipoprotein (Pal) in vivo. Using one of the derivatives containing the photo-crosslinkable amino acid, we developed an in vitro system to analyze the binding of lipoproteins to LolCDE. Our results indicate that compound 2, a LolCDE inhibitor, does not inhibit the binding of lipoproteins to LolCDE, but rather promotes the dissociation of bound lipoproteins from LolCDE.

Graphical Abstract

Graphical Abstract.

Abbreviations

comp2

compound 2

DDM

n-Dodecyl-β-d-maltopyranoside

Pal

peptidoglycan-associated lipoprotein

pBpa

p-benzoyl-L-phenylalanine

TM

transmembrane segment

Vi

vanadate

Bacterial lipoproteins contain a lipid-modified N-terminal cysteine residue and are anchored to the cell membrane by these lipids (1–3). In Escherichia coli, there are approximately 100 lipoproteins, many of which are attached to the inner leaflet of the outer membrane facing the periplasm, while some are at least partially exposed to the cell surface. These lipoproteins play important roles in a variety of cellular processes, including biogenesis and maintenance of the outer membrane, stress sensing and signal transduction and transport of various types of molecules.

Lipoprotein precursors synthesized in the cytoplasm are translocated across the inner membrane and anchored to the outer leaflet of the inner membrane by signal peptides. They then undergo lipid modification and signal peptide cleavage. The transport of outer membrane lipoproteins from the inner to the outer membrane is carried out by the Lol system, which comprises five proteins: LolA, LolB and the three proteins forming the LolCDE complex (2,4). The mechanism of lipoprotein trafficking to the outer membrane via the Lol system has been studied in detail, particularly in E. coli, and a consensus pathway for this process has been established (4). The LolCDE complex, a member of the ABC transporter family, is composed of one molecule each of LolC and LolE, and two molecules of the ATPase subunit LolD. Mature lipoproteins destined for the outer membrane are released from the inner membrane by LolCDE and transferred to the periplasmic chaperone LolA. The water-soluble LolA–lipoprotein complex then crosses the periplasm to the outer membrane, where the lipoproteins are transferred to LolB. Finally, LolB embeds lipoproteins in the inner leaflet of the outer membrane.

Whether lipoproteins remain in the inner membrane or are transported to the outer membrane is determined by the amino acids near the signal peptide cleavage site (5,6). In E. coli, lipoproteins with Asp at amino acid position +2 from the N-terminus of the mature forms are not released by LolCDE and remain in the inner membrane, whereas those with other amino acids at the +2 position are released from the inner membrane by LolCDE and are subsequently transported to the outer membrane (7). However, it remains unclear how only fully modified (triacylated) lipoproteins with non-Asp at +2 are recognized as transport substrates by LolCDE.

In our previous studies, we successfully reproduced most of the fundamental processes of Lol-driven lipoprotein transport from the inner to the outer membrane in cell-free systems, enabling us to elucidate the mechanisms underlying this process. However, we were not able to demonstrate in vitro the first step of the pathway, namely the binding of outer membrane-destined lipoproteins to the LolCDE complex.

In this study, to analyze the interaction between LolCDE and lipoproteins, we established an experimental system to monitor the binding of lipoproteins to LolCDE using site-specific photo-crosslinking. In this approach, a non-canonical photoreactive amino acid is incorporated at a defined position in a protein by amber suppression (8,9). This photoreactive amino acid reacts with a nearby molecule upon UV irradiation to form a stable covalent bond, allowing the analysis of transient protein–protein interactions with high spatial resolution and sensitivity, both in vivo and in vitro. Here, we present biochemical evidence of a functional interaction between the Lol subunits and lipoproteins. Our data indicate that the reported cryo-EM structures of LolCDE are functionally important.

Materials and Methods

Bacterial strains, plasmids and media

The amino acid p-benzoyl-l-phenylalanine (pBpa) was used as the photo-cross-linkable amino acid in this study. The bacterial strains and plasmids used in this study are listed in Supplementary Table S1. BL21(DE3) (10) was used as the host cell for in vivo photo-crosslinking and preparation of the peptidoglycan-associated lipoprotein (Pal) containing crude membrane extracts. MY201 (11) was used to prepare crude membrane extracts, which included pBpa-containing derivatives of LolCDE.

pBADCDE and its derivatives with amber mutations were used to express the wild-type and pBpa-containing derivatives of LolCDE, respectively. The expression of both constructs was governed by the araBAD promoter. The codes in parentheses indicate the site at which the mutation was introduced. For example, pBADC(V44am)DE contains an amber mutation in Val at position 44 of lolC. pSS4 was used to express the C-terminally His×6-tagged Pal (12). pSS4(DD) encodes a derivative of Pal in which the sorting signal, at the +2 and + 3 sites, are changed from SS to DD (13). pSS2(24am) was generated by introducing the following two alterations to the C-terminally His×6-tagged lolA in pSS2 (12): the terminating codon was changed from TAG to TAA, and Val at position 24 was changed to TAG (amber). pEVOL-pBpF was used to insert pBpa into amber codons (8). Mutations were introduced using the PrimeStar PCR mutagenesis kit, as described by the supplier (Takara Shuzo), and the oligonucleotides used are listed in Supplementary Table S2. Cells were cultured routinely at 30°C in LB medium (14).

In vivo photo-crosslinking

Cells carrying pEVOL-pBpF, pSS4, and one of the pBADCDE derivatives with an amber mutation at specific sites of lolC or lolE were grown to mid-log phase, and pBpa (1 mM at final concentration) and arabinose (0.02%) were then added to the culture to express lolCDE. Cells were grown to the late log phase, and IPTG (1 mM) was added for 30 min to induce Pal expression. The cells were then UV irradiated as previously described (15). Total cellular proteins were resolved by SDS-PAGE followed by immunoblotting.

Preparation of solubilized membrane and periplasmic fractions

MY201 cells carrying pBADC(V44am)DE were grown to mid-log phase, then pBpa and arabinose were added to a final concentration of 1 mM and 0.02%, respectively. After a further 3 h of culture, the cells were harvested, washed once with ice-cold 50 mM Tris–HCl (pH 7.5), and resuspended in the buffer. All subsequent procedures were performed on ice or at 4°C. The suspended cells were first disrupted by sonication. After removing undisrupted cells by centrifugation at 5000 × g for 10 min, the extract was centrifuged at 100,000 × g for 30 min. The resulting precipitate was suspended in 50 mM Tris–HCl (pH 7.5) containing 1% n-dodecyl-β-maltopyranoside (DDM) and 10% glycerol and kept on ice for 1 h to solubilize the membrane. The solubilized crude membrane extract was centrifuged again at 100,000 × g for 30 min, and the supernatant was recovered, aliquoted, and stored at −80°C. A crude membrane extract containing Pal was prepared in a similar manner. BL21(DE3) cells carrying pSS4 or pSS4(DD) were grown to the mid-log phase in LB medium, then IPTG was added to a final concentration of 1 mM to induce Pal expression. Cells were then cultured for 3 h. The solubilized crude membrane extract containing Pal was prepared in the same manner as described above. Crude membrane extracts containing Pal-liganded LolC(44pBpa)DE were prepared in the same manner as described above, from cells grown as in the in vivo photo-crosslinking experiments. The periplasmic fraction containing LolA(24pBpa) was prepared essentially as previously described (16). Protein concentrations were quantified with the BCA protein kit (Pierce), using BSA as a standard.

In vitro photo-crosslinking

One microliter (equivalent to 4 μg protein) each of the crude membrane extracts containing LolCDE and Pal were mixed in a buffer solution containing 50 mM Tris–HCl (pH 7.5) (20 μl of total volume) and then kept at 30°C for 15 min. The reaction mixtures were transferred to ice and irradiated with UV light. The reaction mixtures were then subjected to SDS-PAGE followed by immunoblotting. For UV irradiation, a UV LED spot irradiation device (MeCan Imaging Inc.) was used at a distance of 3 cm and maximum power for 1 min. Adenine nucleotides and MgSO₄ were added when needed at a concentration of 1 mM. LolCDE inhibitor compound 2 (comp2) was used at a final concentration of 20 μg/ml. To observe the transfer of lipoproteins to LolA, 1 μl of the periplasmic fraction containing LolA(24pBpa) was added.

Immunoblotting

For in vivo photo-crosslinking experiments, equivalent amounts of protein were loaded into each lane according to the OD₆₀₀ of the cultures. Rabbit anti-LolC, -LolE and -Pal antisera were described previously (17). The antiserum against LolE used in this study also reacts with Pal (15), and usually yields good results in the detection of LolC–Pal cross-link products. Therefore, in some experiments, anti-LolE was used to detect the LolC–Pal cross-linking products. Anti-pentaHis was obtained from Qiagen.

Results

In vivo site-specific photo-crosslinking between LolCDE and pal

Recently, cryo-EM structures of E. coli LolCDE in detergent-solubilized or nanodisc-embedded states have been reported (18–20). Some of these structures were of LolCDE complexed with outer membrane lipoproteins and showed similar lipoprotein-binding modes, although the bound lipoproteins were different. In these structures, α-helices of the first and second transmembrane segments (TM1 and TM2, respectively) of both LolC and LolE extend into the periplasm, forming a cavity for lipoprotein binding (Fig. 1). In the initial step of lipoprotein transport, fully modified outer membrane lipoproteins could access the cavity laterally through the interface between LolC and LolE, and transiently form a lipoprotein–LolCDE complex. To investigate whether this is the case, we attempted to capture complex formation between LolCDE and lipoproteins, by introducing a photoreactive pBpa into the central cavity-forming helices of LolC and LolE.

Fig. 1.

Cryo-EM structure of LolCDE and the location of the photo-crosslinking residues. The cryo-EM structure of the Lpp- and AMP-PNP-bound LolCDE complex (PDB ID: 7ARJ) is shown as a ribbon representation. LolC and LolE are shown in blue and green, respectively. Two LolD monomers are colored orange and bright orange, respectively. The dotted lines indicate the estimated membrane boundaries. Lpp is shown in stick form colored magenta. Right panel is a close-up view of the gray box in the overall structure. The side chains of residues where photo-crosslinking with Pal was observed in Fig. 2 are shown as sticks. These residues are located near the N-linked lipid acyl chains of Lpp in the cryo-EM structure. In the stick representation, the atoms are colored as follows: O, red; N, blue; S, dark yellow; C, yellow (LolC); cyan (LolE); and magenta (Lpp).

We constructed a series of mutant lolCDE genes with a single amber mutation in the TM1s and TM2s of both LolC and LolE, as well as their periplasmic extensions. Specifically, mutations were introduced within amino acid residues 40–60 and 250–271 of LolC, 41–60 and 255–275 of LolE. Plasmids carrying these mutations were introduced into a strain carrying pEVOL-pBpF, which encodes the amber-suppressing tRNA and its cognate tRNA aminoacyl tRNA synthetase for the incorporation of pBpa at amber codons. The host strain also harbors pSS4, which encodes the outer membrane lipoprotein Pal. Pal expression was briefly induced using the T7 promoter prior to UV irradiation to increase the sensitivity of the detection of transiently formed LolCDE–substrate complexes.

Fig. 2 shows the results of the photo-crosslinking experiments with pBpa at LolC TM1 (panel A), LolC TM2 (panel B), LolE TM1 (panel C) and LolE TM2 (panel D). Each set of samples was electrophoresed in duplicates and detected using anti-LolC (upper rows) or anti-LolE serum (lower rows). The anti-LolE serum used in this study reacted with Pal, allowing detection of the LolC–Pal crosslink (15). For some LolC derivatives, results with anti-Pal antiserum are also shown (Supplementary Fig. S1). We observed prominent bands for several pBpa-containing derivatives with molecular weights corresponding to the sum of those of Pal and LolC or LolE. These were derivatives with pBpa at positions of V44 and F51of LolC and V43, G44, M48, F51, M267, I271, M272, L274 and A275 of LolE. Several other weakly reacting bands were detected in the long-exposed images in addition to these sites. In the cryo-EM structures, these amino acids were adjacent to the bound lipoproteins, particularly the acyl chain portions (Fig. 1). Residue V44 of LolC TM1 is located near one of the acyl chains of a diacylglycerol attached to the thiol group of the N-terminal cysteine. M267, I271, L274 and A275 of LolE TM2 are located on opposite sides of the same acyl chain. V43, G44, M48, F51 and M272 of LolE are located near the N-linked acyl group. In contrast, F51 in LolC is located near the N-terminal peptide portion of the bound lipoprotein. In terms of relative position to the membrane, most of these amino acids are located in the vicinity of the expected boundary between the outer leaflet of the inner membrane and the periplasm.

Fig. 2.

In vivo photo-crosslinking of Pal with LolC or LolE. BL21(DE3)/pEVOL-pBpF/pSS4 cells harboring pBADCDE (W.T.) or one of its derivatives carrying an amber mutation in lolC or lolE were subjected to pBpa-mediated photo-crosslinking experiments as described under Materials and Methods. The amino acid residues to which the amber mutation was introduced are shown above. Each set of samples was electrophoresed in duplicate and detected with anti-LolC (upper rows) or anti-LolE serum (lower rows). A triangle indicates putative intramolecular crosslink product of LolC. An asterisk denotes an unidentified band. Note that the anti-LolE antiserum used in this study also reacts with Pal, allowing detection of LolC–Pal crosslink. All samples were UV irradiated in these experiments. (E) Helical wheel projections of the photoreactive pBpa introduced area. TM1 of LolC (A40-L60), TM2 of LolC (R250-L271), TM1 of LolE (L41-L60) and TM2 of LolE (I255-A275) are shown. Sites referred to in the text where photo-crosslinking occurs by the introduced pBpa are shown in black.

Bands that reacted with both anti-LolC and -LolE antisera were observed for several derivatives (Fig. 2). These were derivatives carrying pBpa at V47, F51, A259 and M267 in LolC, and G44, A47, G50, F51, D264, I268 and R269 in LolE. Based on their molecular weights, these bands appeared to be cross-linked products of LolC and LolE. Since the cryo-EM structure suggests that close interactions between the TMs of LolC and LolE on opposite sides of the protein complex are unlikely to occur when the lipoproteins are bound to the central cavity, the crosslinks observed here might have captured the state in which the lipoproteins were not bound. However, the loop between TM3 and TM4 of LolC are also close to TM2 of LolE, and vice versa; therefore, the observed crosslinks may have occurred between them. Nevertheless, some derivatives formed cross-links with both lipoproteins and LolE or LolC. These results suggest that the cross-links observed here do not represent a particular state of LolCDE but rather capture several different conformations, including lipoprotein-bound states.

In vivo photo cross-linking between LolC(44pBpA)DE and pal

Next, we analyzed the cross-link formation in detail using one of the most prominent crosslink-forming derivatives LolC(44pBpa), with pBpa instead of V44 in LolC (Fig. 3). The formation of the photo-crosslinked product containing LolC(44pBpa) and Pal was dependent on the presence of the amber mutation in the lolC gene, pBpa in the growth medium, and UV irradiation (Fig. 3A), indicating that the crosslink observed here was formed through incorporation of pBpa at position 44 of lolC. Furthermore, the formation of this crosslink was dependent on the induction of Pal expression by IPTG (Fig. 3B). In addition, a faster migrating cross-linked product was observed irrespective of IPTG induction, especially in images with increased contrast. Since the formation of this cross-linked product was not observed in the lpp-deleted strain (Fig. 3B), it could be a cross-linked product between LolC(44pBpa) and Lpp, which is the most highly expressed outer membrane lipoprotein in E. coli(21). These results suggest that cross-link formation occurs not only with Pal, but also with outer membrane-destined lipoproteins.

Fig. 3.

Photo-crosslinking between LolC(44pBpa) and Pal. (A) BL21(DE3)/pEVOL-pBpF/pSS4 harboring pBADCDE or its derivative pBADC(V44am)DE, carrying an amber mutation at V44 of LolC, were grown in the presence or absence of pBpa (1 mM) and subjected to photo-crosslinking. The product of the chromosomally encoded lolC gene is barely detectable under these conditions. (B) Wild-type BL21(DE3) and its Δlpp derivative carrying pEVOL-pBpF, pSS4 and pBADC(V44am)DE plasmids were used for photo-crosslinking. Cells were grown to late log-phase and were induced or uninduced with IPTG (1 mM) for 30 min, then subjected to photo-crosslinking. (C) BL21(DE3)/pEVOL-pBpF/pSS4 harboring pBADC(V44am)DE or pBADC(V44am)DΔE were subjected to photo-crosslinking experiments. pBADC(V44am)DΔE lacks the entire lolE coding region. (D) BL21(DE3)/pEVOL-pBpF/pBADC(V44am)DE were transformed with pSS4 or pSS4(DD). Cells were subjected to photo-crosslinking experiments. Samples were electrophoresed in duplicate and detected with anti-LolC (left panel) and anti-His serum (right panel). Note that pSS4 and pSS4(DD) encode C-terminally His×6-tagged Pal, therefore the LolC–Pal cross-linked product can be detected with anti-His antibody.

For some LolC derivatives, including LolC(44pBpa), a band was observed that reacted with the LolC antiserum but not with the LolE antiserum, despite showing similar migration to the LolC–LolE cross-linked product (marked with an asterisk in Fig. 3). These products may represent mis-assembled complexes comprising only LolC and LolD (LolC₂D₂). The facts that these bands were observed only in the LolC derivatives (Fig. 2), and that deletion of the lolE gene from the mutant plasmids did not affect the formation of the bands, supports this notion (Fig. 3C). Furthermore, crosslinks between LolC(44pBpa) and Pal were not observed in lolE deletion mutants. This indicates that the cross-linked product containing LolC and Pal observed here was formed within the properly assembled LolC(44pBpa)D₂E complex, and not LolC(44pBpa)₂D₂, thus indicating that the LolE subunit is essential for lipoprotein binding.

Next, we examined whether the formation of the crosslinks was dependent on the sorting signal of Pal (Fig. 3D). Photo-crosslinking experiments were carried out using a derivative of Pal, Pal(DD), which has an efficient inner membrane retention signal ‘DD’ at the +2 and +3 positions (13). Samples were electrophoresed in duplicate; and one was detected with an antibody against LolC (Fig. 3D, left panel) while the other was detected with an antibody against pentaHis to detect Pal tagged with His×6 (right panel). Despite an equivalent amount of Pal(DD) being expressed, crosslink formation between Pal (DD) and LolC(44pBpa) was greatly reduced compared to that between Pal (W.T.) and LolC(44pBpa)DE. Similar experiments were performed with other pBpa-containing derivatives of LolC and LolE, including those that could possibly be cross-linked with acyl chains other than the acyl chain cross-linked with the pBpa introduced into LolC44 (Supplementary Fig. S2). The results showed that cross-linking with Pal(DD) was greatly reduced compared to that with wild-type Pal in most derivatives examined. Thus, we concluded that the crosslinking observed here captures the specific interaction of LolCDE with outer membrane-targeted lipoproteins. In one derivative, the pBpa substitution at LolE 271, the extent of cross-linking was also reduced, but to a lesser extent than in other derivatives. This may suggest that LolE at position 271 is important for the avoidance of inner membrane lipoproteins.

In addition to these bands, a LolC antiserum-reacting band that migrated slightly faster than LolC (44pBpa) was also observed in the UV-irradiated samples (marked with a triangle in (A) through (D) in Fig. 3). Similar fast-migrating bands were observed for several other derivatives (Fig. 2). These bands may represent intramolecularly cross-linked LolC, which might have more compact structures than those without cross-linking, and thus might have slightly higher electrophoretic mobilities. Since LolC TM2 is present in the vicinity of V44 in LolC (Fig. 1), the pBpa introduced at 44 could have formed a cross-link with TM2 in LolC in its Pal-unbound state.

In vitro detection of LolCDE–Pal complex formation using pBpa-containing LolCDE in detergent-solubilized crude membrane extracts

Next, we attempted to establish a highly sensitive experimental system to monitor the binding of lipoproteins to LolCDE in vitro using LolC(44pBpa)DE. We prepared a crude membrane extract from cells expressing the LolC(44pBpa)DE and solubilized it in 1% DDM. Similarly, the DDM-solubilized crude membrane extracts were prepared from cells expressing Pal. These two fractions were mixed, incubated, irradiated with UV light and analyzed by immunoblotting.

When only the LolC(44pBpa)DE-expressing solubilized crude membrane extract was UV irradiated, several cross-linked products were observed in addition to LolC(44pBpa), including LolC(44pBpa)–Lpp (Fig. 4A, upper panel, lane 8). This suggests that LolCDE is, at least partially, solubilized by DDM in a liganded form with substrate lipoproteins such as Lpp. Next we mixed the crude membrane extracts containing LolC(44pBpa)DE and Pal, incubated the mixture at 30°C, and then irradiated with UV light (Fig. 4A, lanes 5 and 6). The liganded LolCDE bands originally present in the crude membrane extracts containing LolC(44pBpa) were decreased, and the band representing the cross-linked products of between LolC(44pBpa) and Pal increased (compare lanes 6 and 8). The formation of crosslink between LolC(44pBpa) and Pal was dependent on the temperature and time of incubation, reaching a maximum around 25°C for 10 min (Fig. 4B, C). The amount of LolC(44pBpa)–Pal crosslinks did not change upon prolonged incubation until 40 min at 30°C (Fig. 4C). These results indicate that the binding and dissociation of lipoproteins to LolCDE occurred in this reaction system, and eventually, complex formation with externally added Pal became predominant at the final equilibrium state.

Fig. 4.

Binding of Pal to LolCDE in vitro with solubilized membrane extracts. (A) LolC(44pBpa)DE and Pal expressed crude membrane extracts were mixed, and UV irradiated as described under Materials and methods. Proteins were electrophoresed in duplicate, followed by immunodetection using anti-LolC (upper panel) or anti-Pal (lower panel) antisera. An asterisk denotes an unidentified band. We were unable to determine the nature of this band and did not investigate it further in this study. (B) Temperature dependence of LolCDE–Pal complex formation. LolC(44pBpa)DE and Pal expressed crude membrane extracts were mixed on ice and kept for 15 min at indicated temperatures. The mixtures were then transferred to ice and irradiated with UV light and analyzed as in (A). (C) Time dependence of LolCDE–Pal complex formation. LolC(44pBpa)DE and Pal containing crude membrane extracts were mixed on ice and then kept at 30°C for the indicated periods. The mixtures were then transferred onto ice, irradiated with UV light and analyzed as in (A). (D) Crude membrane extract containing LolC(44pBpa)DE were mixed with those expressing Pal or Pal(DD). Reaction mixtures were kept at 30°C for 15 min, then transferred to ice and irradiated with UV light. Samples were electrophoresed in duplicate and detected with anti-LolC (left panel) and anti-Pal serum (right panel). In (B), (C) and (D) all samples were UV irradiated. Pal×2 refers to the Pal dimer. Its formation appears to be dependent on UV irradiation, but the details are not known.

We examined whether this binding depended on the sorting signal of the lipoproteins. When a crude membrane extract expressing Pal(DD), with an inner membrane retention signal, was used, complex formation with LolCDE did not occur, as observed in the in vivo experiments (Fig. 4D). Taken together, we conclude that we could detect the first step of transport, the binding of lipoproteins to LolCDE, using detergent-solubilized LolCDE and lipoproteins in vitro.

In vitro transfer of Pal from LolCDE–Pal to LolA

Next, we examined whether Pal could be released from the LolCDE–Pal complex formed in vitro and transferred to LolA in an ATP-dependent manner (Fig. 5). We added LolA to the lipoprotein-binding reaction mixture and examined whether the LolA–Pal complex was formed. Formation of the LolA–Pal complex was assessed by photo-crosslinking between Pal and LolA(24pBpa), in which V24 was replaced by pBpa. The V24 residue is located near the entrance of the hydrophobic lipoprotein-binding cavity of LolA (16,22), and its pBpa-substituted derivative was one of the most efficiently crosslinked derivatives with Pal in our previous in vivo photo-crosslinking studies (16). In addition, LolA(24pBpa) could also cross-link to the periplasmic domain of LolC (16), presumably via the hook and pad regions (23).

Fig. 5.

In vitro transfer of Pal to LolA. Solubilized crude membrane extracts containing LolC(44pBpa)DE and Pal were mixed with LolA(24pBpa) on ice, then kept at 30°C for 10 min. ATP plus reactions contained 1 mM ATP and 1 mM MgSO₄. Photo-crosslinking was performed and detected with an antiserum against LolA. An asterisk indicates a putative tripartite complex of a lipoprotein, LolC(44pBpa) DE, and LolA(24pBpa). A triangle indicates a putative intramolecular cross-linking product of LolA. The components mixed for the reactions are shown above. All samples were UV irradiated.

When the solubilized crude membrane extract containing LolC(44pBpa)DE, Pal and LolA(24pBpa) was mixed in the absence of ATP, several crosslinked products were observed (Fig. 5, lane 7). The most prominent band appeared to be the product of a cross-link between LolC(44pBpa) and LolA(24pBpa), mediated by pBpa introduced at position 24 of LolA. The bands with slightly slower mobility than the LolC–LolA band appeared to be the product of a tripartite cross-link product between lipoprotein-liganded LolC(44pBpa) and LolA(24pBpa), in which pBpa at position 44 of LolC is linked to the lipoprotein and pBpa at position 24 of LolA is linked to LolC (marked with an asterisk in Fig. 5). The slowest migrating band among the putative tripartite complexes might be Pal–LolC–LolA, since this band reacted with the anti-Pal antisera (Supplementary Fig. S3) and was not detected in the reaction without the Pal fraction (compare lanes 5 and 7). The other high molecular weight bands observed could be tripartite complexes containing outer membrane lipoproteins, such as Lpp, present in the crude membrane extracts containing LolC(44pBpa).

When ATP was added to the reaction system, the putative tripartite complexes disappeared, and new cross-linked bands corresponding to LolA(24pBpa)–Pal and LolA(24pBpa)–Lpp were detected (Fig. 4, lane 14). In the reaction without the Pal-containing crude membrane extract, only LolA(24pBpa) crosslinked to Lpp was prominent (Fig. 5, lane 12). The LolA(24pBpa)–Pal cross-link product was not formed by simply mixing LolA(24pBpa) and the Pal-containing crude membrane extract, even in the presence of ATP (Fig. 5, lane 13), indicating that the formation of the Pal–LolA(24pBpa) complex was mediated by LolCDE.

These results support the notion that the LolCDE–Pal complex represents an intermediate in the transport process, and that the lipoproteins in the complexes are transfer-competent to LolA. Taken together, these results indicate that we could successfully reproduce the reaction of lipoprotein binding to LolCDE and transfer it to LolA using detergent-solubilized LolCDE with a photo-crosslinkable amino acid analog, even in the absence of a membranous structure.

When LolA coexisted with Pal (lanes 6, 7 and 13), a band was observed that migrated slightly faster than LolA (marked with a triangle). This might be due to the formation of an intramolecular cross-link in LolA. Pal might induce conformational changes in LolA without binding its lipid moieties to the central cavity of LolA, but details remain to be determined in future studies.

Effects of adenine nucleotides on the biding of Pal to LolCDE

Our previous experiments showed that the stability of lipoprotein–liganded LolCDE is affected by adenine nucleotides (24,25). Taking advantage of the solubilized membrane system, we investigated whether nucleotides affected the binding of lipoproteins to LolCDE in vitro (Fig. 6A). Solubilized crude membrane extracts containing LolC(44pBpa)DE were premixed with several adenine nucleotides in the presence (Fig. 6, lanes 6–10) or absence (Fig. 6, lanes 1–5) of vanadate (Vi). The Pal-containing fraction was then added to the binding reaction, irradiated with UV light and analyzed by immunoblotting. Vi is a phosphate analog that forms a complex with LolCDE together with ADP (25).

Fig. 6.

Effects of adenine nucleotides and comp2 on Pal binding to and dissociation from LolCDE. (A) Solubilized crude membrane extracts containing LolC(44pBpa)DE were mixed on ice with the nucleotides (1 mM) shown above and MgSO₄ (1 mM), in the presence and absence of Vi. After being kept on ice for 10 min, the solubilized membrane containing Pal was added and kept at 30°C for 10 min. The mixtures were then subjected to photo-crosslinking. (B) Crude membrane extracts containing LolC(44pBpa)DE, nucleotides, and comp2 (20 μg/ml) were mixed and placed on ice for 10 min, then the Pal containing fraction was added. After incubation at 30°C for 10 min, the mixtures were transferred onto ice for photo-crosslinking. (C) Solubilized crude membrane extracts containing Pal–pre-liganded LolC(44pBpa)DE, nucleotides, and comp2 were mixed on ice, allowed to stand for 10 min, and then subjected to photo-crosslinking. Experiments were performed four times, and a representative result is shown. (D) The band intensity of the LolC–Pal photo-crosslink product in each lane in the experiment in (C) was quantified with the band intensity of lane 2 as 100. The mean values of the band intensities in the four experiments are shown along with the standard deviations. LolC–Pal cross-link products were detected with LolC antiserum in (A) and (B). In (C), LolE antiserum, which cross-reacts with Pal, was used for detection of LolC–Pal. A triangle indicates a putative intramolecular cross-linking product. An asterisk denotes an unidentified band.

When LolC(44pBpa)DE was premixed with AMP-PNP, a non-hydrolyzable ATP analog, cross-linking products between LolCDE and Pal was almost undetectable, regardless of the presence of Vi (Fig. 6A, lanes 5 and 10). ADP inhibited Pal binding only in the presence of Vi (Fig. 6A, lanes 3 and 8). ATP alone showed slight inhibition (Fig. 6A, lane 4) and a marked inhibition in the presence of Vi (Fig. 6A, lane 9). We reasoned that ATP might be hydrolyzed by ATPase activity in the solubilized crude membrane extracts, and therefore showed an apparent inhibitory effect only in the presence of Vi. In addition, the presence of ADP, ATP or AMP-PNP increased the intensity of the band that migrates slightly faster than LolC (Fig. 6A, lanes 3–6 and 8–9), which could be an intramolecular crosslink product of LolC, suggesting that the binding of these nucleotides causes conformational changes in LolC. Together, these results indicate that ATP is not required for LolCDE to bind with Pal, and that in the solubilized crude membrane extract, LolCDE undergoes a conformational change by binding to nucleotides, preventing it from binding to lipoproteins. This is consistent with the finding that the cryo-EM structures of both AMP-PNP- and Vi-ADP- bound LolCDE showed closed nucleotide-binding domains and a closed central cavity, which did not allow lipoprotein binding (18–20).

Comp2 promotes dissociation of Pal from LolCDE

Recent high-throughput screenings of chemical libraries identified several compounds that inhibit lipoprotein modification and trafficking to the outer membrane (26–30). Among these, comp2, a pyridineimidazole compound, acts on LolCDE and inhibits LolA-dependent release of Lpp from spheroplasts (27). However, it is unclear how comp2 prevents LolCDE from extracting lipoproteins from the inner membrane and transferring them to LolA. To clarify this, we examined the effect of comp2 on the interaction between LolCDE and Pal using an in vitro photo-crosslinking system with detergent-solubilized crude membrane extracts containing LolC(44pBpa)DE (Fig. 6B and C).

We first, examined the effects of comp2 on the binding of LolCDE to Pal. When the crude membrane extract containing LolC(44pBpa)DE was premixed with comp2 and Pal was subsequently added for the binding reaction, the photo-crosslinking between LolCDE and Pal was not affected (Fig. 6B, lanes 1 and 2). A similar experiment was performed in the presence of different adenine nucleotides (Fig. 6B, lanes 3–5). The results showed that crosslink formation was significantly reduced only in the presence of ATP. These results suggested that comp2 either inhibits the lipoprotein-binding reactions to LolCDE or promotes the dissociation of Pal from the LolCDE–Pal complex only in the presence of ATP.

We have previously observed that ATP causes the dissociation of lipoproteins from lipoprotein-liganded LolCDE (24,25). We investigated whether comp2 could affect this process (Fig. 6C and6D). Solubilized crude membrane extracts expressing LolC(44pBpa)DE pre-liganded with Pal were prepared to determine whether Pal dissociation was affected by comp2. Similar to our previous observations, ATP or AMP-PNP alone slightly reduced the complex (lanes 2–4); comp2 also caused a slight reduction in the complex (compare lanes 2 and 5), whereas the simultaneous addition of comp2 and ATP greatly reduced the amount of the complex (lane 6). AMP-PNP showed no synergistic effect with Comp 2 (lane 7). These results suggest that comp2 promotes Pal dissociation from the LolCDE–Pal complex in an ATP hydrolysis-dependent manner rather than by inhibiting LolCDE–Pal binding.

Discussion

LolCDE is an ABC transporter classified as type VII due to its high amino acid sequence similarity to MacB, which constitutes part of a tripartite drug efflux pump (31–35). Recently, three groups have reported cryo-EM structures of E. coli LolCDE in detergent-solubilized or nanodisc-embedded states (18–20). These structures include the lipoprotein- and nucleotide-bound states of LolCDE, in addition to those of LolCDE alone and provide a significant advances in our understanding of the mechanism by which LolCDE functions. However, the mechanism by which LolCDE releases the outer membrane lipoproteins remains elusive. For example, it is still unclear how LolCDE recognizes only triacylated lipoproteins as substrates rather than diacylated lipoproteins (36,37), as is the mechanisms by which lipoproteins with an inner membrane retention signal (+2Asp) are avoided by LolCDE (7).

To address these issues, in this study, we established a sensitive method for monitoring the binding of outer membrane lipoproteins to LolCDE, the first step in lipoprotein transport, by systematically introducing a photo-crosslinkable amino acid into the lipoprotein-binding sites of LolCDE. We prepared a series of derivatives in which pBpa was introduced at approximately 80 different sites of the α-helices that form the central substrate-binding cavity of LolC or LolE, and examined whether they crosslinked with Pal in living cells. Approximately, one-sixth of the derivatives formed crosslinked with the outer membrane lipoprotein Pal. The pBpa-replaced sites that formed cross-links with Pal generally corresponded to the vicinity of the acyl groups and the N-terminal peptide portion of the lipoproteins in the structures of the LolCDE–lipoprotein complex, as revealed by cryo-EM. This suggests that our cross-linking experiments captured the same state of the LolCDE–lipoprotein complex as that analyzed by cryo-EM. These crosslinks, along with those between LolC and LolE, could be useful probes for detecting the conformational changes that occur in the LolCDE complex during the reaction cycle, both in vivo and in vitro.

Using this crosslink formation method, we developed an in vitro experimental system for lipoprotein binding to LolCDE. We showed that Pal binds to LolCDE when detergent-solubilized crude membrane extracts containing LolCDE and Pal, prepared separately in vitro, were mixed. Furthermore, bound Pal was transferred to LolA in an ATP-dependent manner. Thus, using this system, it is possible to study the binding of lipoproteins to LolCDE and their transfer to LolA. We have been unable to demonstrate the specific binding of lipoproteins to LolCDE by mixing lipoproteins and LolCDE in vitro, but stabilization of the complexes by crosslinking made this possible. This method uses solubilized crude membrane extracts and does not require highly purified components or reconstitution of the components into liposomes or nanodiscs. Therefore, they are suitable for studying the effects of a wide variety of mutations on the function of LolCDE and LolA. However, there are also limitations in the use of unpurified crude fractions. In our previous studies, we noted the importance of lipids as a mechanism by which LolCDE avoids inner membrane lipoproteins (38,39). However, because membrane lipids remain in this reaction, it is difficult to directly verify the role of lipids in sorting the inner and outer membrane lipoproteins. A similar experimental system using purified LolCDE containing pBpa and lipoproteins may provide a more direct approach for testing the role of membrane lipids in the avoidance of lipoproteins with an inner membrane retention signal by LolCDE.

Some outer membrane lipoproteins play essential roles in the biogenesis of the outer membrane (40–42), which is indispensable for cell survival. In addition, mislocalization of the major outer membrane lipoprotein, Lpp, to the inner membrane is extremely detrimental to cells (43). Thus, cells defective in any of the steps of lipoprotein maturation and trafficking to the outer membrane lose their viability, and factors involved in these processes have been considered promising targets for antimicrobial agents (44,45). A small molecule, comp2, that inhibits LolCDE function, has been previously identified (27). In this study, we found that comp2 does not inhibit the binding of outer membrane lipoproteins to LolCDE but rather promoted their dissociation from LolCDE in an ATP hydrolysis-dependent manner. Comp2 may affect the opening and closing of the lateral gates for lipoproteins between LolC and LolE, such that the acyl chains of a lipoprotein do not move towards LolA. Instead, they laterally escape from the binding cavity upon binding and hydrolysis of ATP, thereby aborting the cycle. Based on the structure of the LolCDE–lipoprotein complex and the results of biochemical analyses, it has been proposed that, of the two possible lateral entrances between LolC and LolE leading to the central cavity, lipoproteins enter from only one side via the LolC TM1 and LolE TM2 interface (18,19). Curiously, mutations that confer resistance to comp2 map predominantly to near the entrance on the other side of the complex, namely the periplasmic extensions of LolC TM2, and in the periplasmic loop between TM3 and TM4 of LolE (25). Further studies, especially concerning the structure of the LolCDE–lipoprotein–comp2 ternary complex, are required to elucidate the mechanism of action of comp2.