Abstract
The outer membrane (OM) is essential for viability in Gram-negative bacteria, yet mechanisms to ensure its stability and homeostasis are not known. The trans-envelope Tol-Pal complex, whose physiological role has remained elusive, is important for OM stability, and coordinated OM invagination during cell division. Here, we establish the function of the Tol-Pal complex in OM lipid homeostasis in Escherichia coli. Cells lacking the complex exhibit defects in lipid asymmetry and accumulate excess phospholipids (PLs) in the OM. This imbalance in OM lipids is due to defective retrograde PL transport in the absence of a functional Tol-Pal complex. Thus, cells ensure the assembly of a stable OM by maintaining an excess flux of PLs to the OM only to return the surplus to the inner membrane via transport mediated by the Tol-Pal complex. Our findings also provide insights into the mechanism by which the Tol-Pal complex promotes OM invagination during division.
Lipid bilayers define cellular compartments, and thus life itself, yet our understanding of the assembly and maintenance of these structures are limited. In Gram-negative bacteria, the outer membrane (OM) is essential for growth, and allows the formation of an oxidizing periplasmic compartment beyond the cytoplasmic or inner membrane (IM)1. The OM is asymmetric, with lipopolysaccharides (LPS) and phospholipids (PLs) found in the outer and inner leaflets, respectively. This unique lipid asymmetry is required for the OM to function as an effective and selective permeability barrier against toxic substances, rendering Gram-negative bacteria intrinsically resistant to many antibiotics, and allowing survival under adverse conditions. The assembly pathways of various OM components, including LPS2, β-barrel OM proteins (OMPs)3, and lipoproteins4, have been well-characterized; however, processes by which PLs are assembled into the OM have not been discovered. Even though they are the most basic building blocks of any lipid bilayer, essentially nothing is known about how PLs are transported between the IM and the OM. Unlike other OM components, PL movement between the two membranes is bidirectional5, 6, 7. While anterograde (IM-to-OM) transport is essential for OM biogenesis, the role for retrograde (OM-to-IM) PL transport is unclear. How assembly of the various OM components are coordinated to ensure homeostasis and stability of the OM is also unknown.
The Tol-Pal complex is a trans-envelope system highly conserved in Gram-negative bacteria8, 9. It comprises five proteins organized in two sub-complexes, TolQRA in the IM and TolB-Pal at the OM. In Escherichia coli, these sub-complexes interact in a proton motive force (pmf)-dependent fashion, with TolQR transducing energy to control conformational changes in TolA and allowing it to reach across the periplasm to contact Pal10, 11, an OM lipoprotein that binds peptidoglycan12. TolA also interacts with periplasmic TolB13, whose function within the complex is not clear. The TolQRA sub-complex is analogous to the ExbBD-TonB system8, 14, 15, where energy-dependent conformational changes in TonB are exploited for the transport of metal-siderophores across the OM16. Unlike the ExbBD-TonB system, however, the physiological role of the Tol-Pal complex has not been elucidated, despite being discovered over four decades ago17, 18. The Tol-Pal complex has been shown to be important for OM invagination during cell division19, but mutations in the tol-pal genes also result in a variety of phenotypes, such as hypersensitivity to detergents and antibiotics, leakage of periplasmic proteins, and prolific shedding of OM vesicles, all indicative of an unstable OM8. In addition, removing the tol-pal genes causes envelope stress and up-regulation of the σE and Rcs phosphorelay responses20, 21. It has thus been suggested that the Tol-Pal complex may in fact be important for OM stability and biogenesis. Interestingly, the tol-pal genes are often found in the same operon as ybgC9, which encodes an acyl thioesterase shown to interact with PL biosynthetic enzymes in E. coli22. This association suggests that the Tol-Pal complex may play a role in PL metabolism and/or transport.
Cells lacking the Tol-Pal complex exhibit defects in OM lipid asymmetry
To elucidate the function of the Tol-Pal complex, we set out to characterize the molecular nature of OM defects observed in tol-pal mutants in E. coli. Defects in the assembly of OM components typically lead to perturbations in OM lipid asymmetry23, 24. This is characterized by the accumulation of PLs in the outer leaflet of the OM, which serve as substrates for PagP-mediated acylation of LPS (lipid A)25. To determine if tol-pal mutants exhibit defects in OM lipid asymmetry, we analyzed lipid A acylation in strains lacking any member of the Tol-Pal complex. We demonstrated that each of the mutants accumulate more hepta-acylated lipid A in the OM compared to wild-type (WT) cells (Fig. 1). This OM defect, and the resulting SDS/EDTA sensitivity in these tol-pal mutants, are all corrected in the complemented strains (Fig. 1, Extended Data Fig. 1). We also examined other strains with known OM permeability defects. We detected increased lipid A acylation in strains with either impaired OMP (bamB, bamD, ΔsurA) or LPS (lptD4213) biogenesis, as would be expected, but not in strains lacking covalent tethering between the cell wall and the OM (Δlpp) (Fig. 1). Even though the Δlpp mutant is known to exhibit pleiotropic phenotypes similar to the tol-pal mutants26, 27, it does not have perturbations in OM lipid asymmetry. In contrast to OMP or LPS assembly mutants, tol-pal strains produce WT levels of major OMPs and LPS in the OM (Extended Data Fig. 2). These results indicate that tol-pal mutations lead to accumulation of PLs in the outer leaflet of the OM independent of OMP and LPS biogenesis pathways.
Cells lacking the Tol-Pal complex have disrupted OM lipid homeostasis
We hypothesized that the loss of OM lipid asymmetry in tol-pal mutants is due to defects in PL transport across the cell envelope. To test this, we examined the steady-state distribution of PLs (specifically labelled with [3H]-glycerol) between the IM and the OM in WT and tol-pal strains. We established that tol-pal mutants have ~1.4-1.6-fold more PLs in their OMs (relative to the IMs) than the WT strain (Fig. 2a, Extended Data Fig. 4). To ascertain if this altered distribution of PLs between the two membranes was due to the accumulation of more PLs in the OMs of tol-pal mutants, we quantified the ratios of PLs to LPS (both lipids now labelled with [14C]-acetate) following OM isolation and differential extraction. tol-pal mutants contain ~1.5-2.5-fold more PLs (relative to LPS) in their OMs, when compared to the WT strain (Fig. 2b, Extended Data Fig. 5). Since tol-pal mutants produce WT LPS levels (Extended Data Fig. 2b), we conclude that strains lacking the Tol-Pal complex accumulate excess PLs in their OMs, a phenotype that can be corrected via genetic complementation (Fig. 2). Consistent with this idea, tol-pal mutants, unlike WT28, are able to survive the toxic effects of LPS overproduction (Extended Data Fig. 6), possibly due to a more optimal balance of PLs to LPS in their OMs. Importantly, having excess PLs makes the OM unstable, and can account for the permeability and vesiculation phenotypes observed in tol-pal mutants8, 27. Furthermore, cells lacking the Tol-Pal complex are on average shorter and wider than WT cells (when grown under conditions with no apparent division defects)19; this reflects an increase in surface area of the rod-shaped cells, possibly a result of increase in OM lipid content. As expected, we did not observe disruption of lipid homeostasis in the Δlpp mutant (Fig. 2). However, we observed higher PL content in the OMs of strains defective in OMP assembly. We reasoned that this increase may help to stabilize the OM by filling the voids created by the decrease in properly-assembled OMPs. Since strains lacking the Tol-Pal complex have proper OMP assembly (Extended Data Fig. 2a), the phenotype of excess PL build-up in the OM must be due to a different problem. Our results suggest that tol-pal mutations directly affect PL transport processes, and therefore OM lipid homeostasis.
Cells lacking the Tol-Pal complex are defective in retrograde PL transport
Unlike for other OM components, PL transport between the IM and the OM is bidirectional5, 6, 7. Therefore, a simple explanation for the accumulation of excess PLs in the OMs of cells lacking the Tol-Pal complex is that there are defects in retrograde PL transport. To evaluate this possibility, we used the turnover of OM PLs (specifically anionic lipids, including phosphatidylserine (PS), phosphatidylglycerol (PG), and cardiolipin (CL)) as readout for the transport of PLs back to the IM (Fig. 3a). As an intermediate during the biosynthesis of the major lipid phosphatidylethanolamine (PE), PS is converted to PE by the PS decarboxylase (PSD) at the IM, and typically exists only at trace levels29. PG and CL have relatively short lifetimes30, 31. While the pathways for CL turnover are not known, PG can be converted to PE via PS32. Since all known enzymes involved in possible pathways of converting PG to PS, and then to PE, are localized in the IM29, the turnover of OM anionic lipids require, and therefore report on, retrograde PL transport. Such an assay has previously been employed to demonstrate retrograde transport for PS7.
Using a strain expressing a temperature-sensitive (Ts) allele (psd2) of the gene encoding PSD33, we pulse-labelled PLs with [32P]-phosphate at the restrictive temperature (42°C), and monitored the turnover of individual PL species in the OM during a chase period at the permissive temperature (30°C). At 42°C, the psd2 strain accumulates substantial amounts of PS in both the IM and the OM (Fig. 3b, 0-min time point), as previously reported33. With the restoration of PSD activity at 30°C, we observed initial increase but eventual conversion of PS to PE in both membranes (Fig. 3b, after 45-min time point), indicating that OM PS is transported back to the IM, converted to PE, and subsequently re-equilibrated to the OM7. We also detected higher PG/CL content in the psd2 strain at 42°C, and saw rapid conversion of these lipids to PE in both membranes at 30°C (Fig. 3b), at rates comparable to what was previously reported (for PG)32. The fact that PS levels increase initially but decrease after 45 min into the chase is consistent with the idea that PS is an intermediate along the turnover pathway for PG32, as well as for CL. To confirm this observation, we also performed the chase at 42°C in the presence of a known PSD inhibitor34 (these conditions completely shut down PSD activity), and found quantitative conversion of PG/CL to PS in both membranes (Extended Data Fig. 7). We further showed that PG/CL-to-PE conversion is abolished in the presence of the pmf uncoupler carbonyl cyanide m-chlorophenyl hydrazone (CCCP)(Fig. 3c), demonstrating a requirement for cellular energy sources in this process32, and that conversion occurs in the IM. The observation of PG/CL turnover in the IM is thus expected. The fact that we also observed the conversion of OM PG/CL to PE points towards an intact retrograde PL transport pathway for these lipids in the otherwise WT cells. Notably, turnover of OM PG/CL appears to be slightly faster than that of IM PG/CL (Fig. 3b), suggesting that retrograde transport of these lipids may be coupled to the turnover process.
We performed the same pulse-chase experiments with psd2 cells lacking TolA. We detected PG/CL-to-PE conversion in the IM at rates comparable to WT (Fig. 3d, f; ~67% and ~71% PG/CL turnover at 2 h-chase in ΔtolA and WT IMs, respectively (Fig. 4a)), demonstrating that there are functional PG/CL turnover pathways in the ΔtolA mutant. In contrast, we observed substantial reduction of the turnover of OM PG/CL in these cells (Fig. 3d, f; ~53% PG/CL turnover at 2 h-chase in the ΔtolA OM, compared to ~79% for WT (Fig. 4a)), even though PS conversion to PE appears intact. These results indicate an apparent defect in the movement of PG and CL (but not PS) from the OM back to the IM, which is restored when complemented with functional tolAWT (Fig. 3e, f, Fig. 4a). ΔtolR mutant cells exhibit the same defect, and can similarly be rescued by complementation with functional tolRWT (Fig. 4a). In contrast, no rescue was observed when ΔtolR was complemented using a tolR allele with impaired ability to utilize the pmf (tolRD23R)14 (Fig. 4a, Extended Data Fig. 1); this indicates that Tol-Pal function is required for efficient PG/CL transport. We also examined PG/CL turnover in psd2 cells lacking BamB, which accumulate excess PLs in the OM due to defects in OMP assembly (Fig. 2). Neither IM nor OM PG/CL turnover is affected (Fig. 4a), highlighting the different basis for OM PL accumulation in this strain compared to the tol-pal mutants. Our assay does not report on the retrograde transport of major lipid PE, which is relatively stable30. However, since tol-pal mutants accumulate ~1.5-fold more PLs in the OM (Fig. 2) without gross changes in PL composition (compared to WT) (Extended Data Fig. 9), PE transport must also have been affected. We conclude that the Tol-Pal complex is required for the retrograde transport of bulk PLs in E. coli.
Removing the Tol-Pal complex does not completely abolish retrograde PG/CL transport, indicating that there are other systems involved in this process. The OmpC-Mla system is important for the maintenance of OM lipid asymmetry, and is proposed to do so via retrograde PL transport35, 36. We showed that cells lacking MlaC, the periplasmic lipid chaperone in the OmpC-Mla system, do not exhibit defects in OM PG/CL turnover (Fig. 4a). In addition, removing MlaC in cells lacking the Tol-Pal complex does not appear to exacerbate the defects in retrograde PL transport, given that overall turnover rates of IM and OM PG/CL are similarly reduced in the double mutant. These results indicate that the OmpC-Mla system may not contribute significantly to retrograde transport of bulk lipids when expressed at physiological levels, as has been previously suggested35. When MlaC and the IM MlaFEDB complex37 are overexpressed, however, we detected partial rescue of OM PG/CL turnover in the ΔtolA mutant (Fig. 4b). Further to validating the putative PL transport function of the OmpC-Mla system, this observation lends strong support to the notion that the Tol-Pal complex is indeed a major system for retrograde PL transport.
Discussion
Our work reveals that the Tol-Pal complex plays an important role in maintaining OM lipid homeostasis via retrograde PL transport. Removing the system causes accumulation of excess PLs (over LPS) in the OM (Fig. 2). While pathways for anterograde PL transport remain to be discovered, this result indicates that PL flux to the OM may be intrinsically higher than that of LPS. Evidently, the ability to transport high levels of PLs to the OM allows cells to compensate for the loss of OMPs due to defects in assembly (Fig. 2). Our data suggest that cells maintain an excess flux of PLs to the OM in order to offset changes in the unidirectional assembly pathways for other OM components, and then return the PL surplus to the IM via the Tol-Pal complex (and other redundant systems). Having bidirectional PL transport therefore provides a mechanism to regulate and ensure the formation of a stable OM.
It is not clear how the Tol-Pal complex may mediate retrograde PL transport. It is possible that this machine directly binds and transports lipids, although there are no obvious lipid binding motifs or cavities found in available structures of the periplasmic components38, 39. The Tol-Pal complex is related to the ExbBD-TonB14, 40, Agl-Glt41, and Mot14, 42 systems, each of which uses pmf-energized conformational changes to generate force for the uptake of metal-siderophores, for gliding motility, or to power flagella rotation, respectively. In addition, both the Tol-Pal and ExbBD-TonB complexes are hijacked by toxins (such as colicins) and bacteriophages to penetrate the OM43. It is therefore also possible that the Tol-Pal complex acts simply as a force generator to transport other PL-binding proteins across the periplasm, or perhaps bring the OM close enough to the IM for PL transfer to occur via hemifusion events. For the latter scenario, one can envision energized TolA pulling the OM inwards via its interaction with Pal, which is anchored to the inner leaflet of the OM12. While it remains controversial, the formation of such “zones of adhesion”, or membrane contact sites, has previously been proposed44, and in fact, was suggested to be a mechanism for retrograde transport of native and foreign lipids6.
That the Tol-Pal complex is involved in retrograde PL transport also has significant implications for Gram-negative bacterial cell division. As part of the divisome, this system is important for proper OM invagination during septum constriction19, 45, 46. How OM invagination occurs is unclear. Apart from physically tethering the IM and the OM, we propose that removal of PLs from the inner leaflet of the OM by the Tol-Pal complex serves to locally reduce the surface area of the inner leaflet relative to the outer leaflet47. According to the bilayer-couple model48, this may then induce the requisite negative curvature in the OM at the constriction site, thus promoting formation of the new cell poles.
Given the importance of the Tol-Pal complex in OM stability and bacterial cell division, it would be an attractive target for small molecule inhibition. This is especially so in some organisms, including the opportunistic human pathogen Pseudomonas aeruginosa, where the complex is essential for growth49, 50. The lack of understanding of the true role of the Tol-Pal complex, however, has impeded progress. We believe that our work in elucidating the physiological function of this complex will accelerate efforts in this direction, and contribute towards the development of new antibiotics in our ongoing fight against recalcitrant Gram-negative infections.
Author contributions
R.S. performed all experiments described in this work; X.E.J. performed experiments related to LpxC overexpression; R.S. and S.-S.C. analyzed and discussed data; R.S. and S.-S.C. wrote the paper.
Competing financial interests
The authors declare no conflict of interest.
Methods
Bacterial strains and growth conditions
All the strains used in this study are listed in Supplementary Table 1. Escherichia coli strain MC4100 [F- araD139 Δ(argF-lac) U169 rpsL150 relA1 flbB5301 ptsF25 deoC1 ptsF25 thi]51 was used as the wild-type (WT) strain for most of the experiments. To achieve accumulation of phosphatidylserine (PS) in cells, a temperature-sensitive phosphatidylserine decarboxylase mutant (psd2), which accumulates PS at the non-permissive temperature, was used33. NR754, an araD+ revertant of MC410024, was used as the WT strain for experiments involving overexpression of lpxC from the arabinose-inducible promoter (PBAD). ΔtolQ, ΔtolA and Δtol-pal deletions were constructed using recombineering52 and all other gene deletion strains were obtained from the Keio collection53. Whenever needed, the antibiotic resistance cassettes were flipped out as described52. Gene deletion cassettes were transduced into relevant genetic background strains via P1 transduction54. Luria-Bertani (LB) broth (1% tryptone and 0.5% yeast extract, supplemented with 1% NaCl) and agar were prepared as previously described54. Strains were grown in LB medium with shaking at 220 rpm at either 30°C, 37°C, or 42°C, as indicated. When appropriate, kanamycin (Kan; 25 µg ml−1), chloramphenicol (Cam; 30 µg ml−1) and ampicillin (Amp; 125 µg ml−1) were added.
Plasmid construction
All the plasmids used in this study are listed in Supplementary Table 2. Desired genes were amplified from MC4100 chromosomal DNA using the indicated primers (sequences in Supplementary Table 3). Amplified products were digested with indicated restriction enzymes (New England Biolabs), which were also used to digest the carrying vector. After ligation, recombinant plasmids were transformed into competent NovaBlue (Novagen) cells and selected on LB plates containing appropriate antibiotics. DNA sequencing (Axil Scientific, Singapore) was used to verify the sequence of the cloned gene.
To generate tolRD23R mutant construct, site-directed mutagenesis was conducted using relevant primers listed in Supplementary Table 3 with pET23/42tolR as the initial template. Briefly, the entire template was amplified by PCR and the resulting PCR product mixture digested with DpnI for > 1 h at 37°C. Competent NovaBlue cells were transformed with 1 μl of the digested PCR product and plated onto LB plates containing ampicillin. DNA sequencing (Axil Scientific, Singapore) was used to verify the introduction of the desired mutation.
Analysis of [32P]-labelled lipid A
Mild acid hydrolysis was used to isolate lipid A as previously described55 with some modifications. 5-ml cultures were grown in LB broth (inoculated from an overnight culture at 1:100 dilution) containing [32P]-disodium phosphate (final 1 µCi ml−1; Perkin Elmer product no. NEX011001MC) till mid-log phase (OD600 ∼0.5 - 0.7). One MC4100 WT culture labelled with [32P] was treated with EDTA (25 mM pH 8.0) for 10 min prior to harvesting. Cells were harvested at 4,700 × g for 10 min, washed twice with 1 ml PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, pH 7.4) and suspended in PBS (0.32 ml) again. Chloroform (0.4 ml) and methanol (0.8 ml) were added and the mixtures were incubated at room temperature for 20 min with slow shaking (60 rpm) to make the one-phase Bligh-Dyer mixture (chloroform:methanol: water = 1:2:0.8). Mixtures were then centrifuged at 21,000 × g for 30 min. Pellets obtained were washed once with fresh one-phase Bligh-Dyer system (1 ml) and centrifuged as above. Resulting pellets were suspended in 0.45 ml of sodium acetate (12.5 mM, pH 4.5) containing SDS (1 %) and heated at 100°C for 30 min. After cooling to room temperature, chloroform and methanol (0.5 ml each) were added to create a two-phase Bligh-Dyer mixture (chloroform:methanol:water = 2:2:1.8). The lower (organic) phase of each mixture was collected after phase partitioning via centrifugation at 21,000 × g for 30 min. This was washed once with upper phase (0.5 ml) of freshly prepared two-phase Bligh-Dyer mixture and centrifuged as above. Finally, all the collected lower phases containing [32P]-labelled lipid A were air-dried overnight. Dried radiolabelled lipid A samples were suspended in 50 μl of chloroform:methanol (2:1) and equal amounts (~1,000 cpm) of radioactivity were spotted on silica-gel coated TLC (Thin Layer Chromatography) plates (Merck). TLCs were developed in chambers pre-equilibrated overnight with solvent system chloroform:pyridine:98 % formic acid: water (50:50:14.6:5). TLC plates were air-dried overnight and later visualized by phosphor imaging (STORM, GE healthcare). The densitometric analysis of the spots obtained on the phosphor images of TLCs was carried out using ImageQuant TL analysis software (version 7.0, GE Healthcare). Average levels of hepta-acylated lipid A (expressed as a percentage of total lipid A in each sample) were obtained from three independent experiments.
Sucrose density gradient fractionation
Sucrose density gradient centrifugation was performed as previously described56 with some modifications. For each strain, a 10/50-ml culture (inoculated from an overnight culture at 1:100 dilution) was grown in LB broth until OD600 reached ~0.5 – 0.7. For radiolabeling, indicated radioisotopes were added from the start of inoculation. Cells were harvested by centrifugation at 4,700 × g for 10 min, suspended to wash once in 5 ml of cold Buffer A (Tris-HCl, 10 mM pH 8.0), and centrifuged as above. Cells were resuspended in 6 ml of Buffer B (Tris-HCl, 10 mM pH 8.0 containing 20% sucrose (w/w), 1 mM PMSF and 50 μg ml −1 DNase I), and lysed by a single passage through a high pressure French press (French Press G-M, Glen Mills) homogenizer at 8,000 psi. Under these conditions, lipid mixing between inner and outer membranes is minimal56. Unbroken cells were removed by centrifugation at 4,700 × g for 10 min. The cell lysate was collected, and 5.5 ml of cell lysate was layered on top of a two-step sucrose gradient consisting of 40% sucrose solution (5 ml) layered on top of 65% sucrose solution (1.5 ml) at the bottom of the tube. All sucrose (w/w) solutions were prepared in Buffer A. Samples were centrifuged at 39,000 rpm for 16 h in a Beckman SW41 rotor in an ultracentrifuge (Model XL-90, Beckman). 0.8-ml fractions (usually 15 fractions) were manually collected from the top of each tube.
Analysis of OMP and LPS levels in isolated OMs
OM fragments were isolated from 50 ml of cells following growth, cell lysis and application of sucrose density gradient fractionation, as described above. Instead of manual fractionation, OM fragments (~1 ml) were isolated from the 40%/65% sucrose solution interface by puncturing the side of the tube with a syringe. Buffer A (1 ml) was added to the OM fragments to lower the sucrose concentration and reduce viscosity. The OM fragments were then pelleted in a microcentrifuge at 21,000 x g for 30 min and then resuspended in 200 - 250 μl Buffer A. Protein concentrations of these OM preparations were determined using Bio-Rad DC protein assay. The same amount of OM (based on protein content) for each strain was analyzed by reducing SDS-PAGE and immunoblotted using antibodies directed against OmpC, OmpF, LamB, BamA, LptE and LPS.
Analysis of steady-state [3H]-glycerol-labelled PL distribution in IMs and OMs
To specifically label cellular PLs, 10-ml cells were grown at 37°C in LB broth (inoculated from an overnight culture at 1:100 dilution) containing [2-3H]-glycerol (final 1 µCi ml−1; Perkin Elmer product no. NET022L001MC) until OD600 reached ~0.5 - 0.7. Once the desired OD600 was achieved, cultures were immediately mixed with ice-cold Buffer A containing CCCP (50 µM) to stop the labeling of the cultures. Cells were pelleted, lysed, and fractionated on sucrose density gradients, as described above. 0.8-ml fractions were collected from each tube, as described above, and 300 μl from each fraction was mixed with 2 ml of Ultima Gold scintillation fluid (Perkin Elmer, Singapore). Radioactivity ([3H]-count) was measured on a scintillation counter (MicroBeta2®, Perkin-Elmer). Based on [3H]-profiles, IM and OM peaks were identified and peak areas determined after background subtraction (average count of first 5 fractions was taken as background). For each strain, relative [3H]-PL levels in the IM and OM were expressed as a percentage of the sum in both membranes (see Fig. 2a upper panel). The average percent [3H]-PL in the OM for each strain (obtained from three independent experiments) was then compared to that for the WT strain to calculate fold changes (see Fig. 2a lower panel).
Determination of PL/LPS ratios in [14C]-acetate labelled OMs (see Extended Data Fig. 5 for workflow and results)
To specifically label all cellular lipids (including LPS), 10-ml cells were grown at 37°C in LB broth (inoculated from an overnight culture at 1:100 dilution) containing [1-14C]-acetate (final 0.2 µCi ml−1; Perkin Elmer product no. NEC084A001MC) until OD600 reached ~0.5 – 0.7. At this OD, cultures were transferred immediately to ice-cold Buffer A (5 ml), pelleted, lysed, and fractionated on sucrose density gradients, as described above. 0.8-ml fractions were collected from each tube, as described above, and 50 μl from each fraction was mixed with 2 ml of Ultima Gold scintillation fluid (Perkin Elmer, Singapore). Based on [14C]-profiles, IM and OM peaks were identified. OM fractions were then pooled, and treated as outlined below to differentially extract PLs and LPS for relative quantification within each OM pool. For each strain, the whole experiment was conducted and the OM PL/LPS ratio obtained three times.
Each OM pool (0.32 ml) was mixed with chloroform (0.4 ml) and methanol (0.8 ml) to make a one-phase Bligh-Dyer mixture (chloroform:methanol:water = 1:2:0.8). The mixtures were vortexed for 2 min and later incubated at room temperature for 20 min with slow shaking at 60 rpm. After centrifugation at 21,000 × g for 30 min, the supernatants (S1) were collected. The resulting pellets (P1) were washed once with fresh 0.95 ml one-phase Bligh-Dyer solution and centrifuged as above. The insoluble pellets (P2) were air dried and used for LPS quantification (see below). The supernatants obtained in this step (S2) were combined with S1 to get the combined supernatants (S3), which contained radiolabelled PLs. To these, chloroform (0.65 ml) and methanol (0.65 ml) were added to convert them to two-phase Bligh-Dyer mixtures (chloroform:methanol:water = 2:2:1.8). After a brief vortexing step, the mixtures were centrifuged at 3000 × g for 10 min to separate the immiscible phases, and the lower organic phases were collected. These were washed once with equal volumes of water and centrifuged as above, and the lower organic phases (containing radiolabelled PLs) recollected and air dried. Finally, the dried PLs were dissolved in 50 µl of a mixture of chloroform:methanol (2:1). Equal volumes (20 µl) of PL solutions were mixed with 2 ml of Ultima Gold scintillation fluid (Perkin Elmer, Singapore). The [14C]-counts were measured using scintillation counting (MicroBeta2®, Perkin-Elmer) and taken as the levels of PLs isolated from the OMs.
To quantify LPS, the P2 pellets were suspended in 2X reducing SDS-PAGE loading buffer (40 µl) and boiled for 10 min. Equal volumes (15 µl) were loaded and subjected to SDS-PAGE (15% Tris.HCl). Gels were air-dried between porous films (Invitrogen) and exposed to the same phosphor screen along with standards (GE healthcare). To generate a standard curve for LPS quantification, the WT OM pellet sample was serially diluted two-fold and equal volumes of diluted samples were resolved on SDS-PAGE and dried as above. The densitometric analysis of bands (i.e. LPS from each OM) was carried out using ImageQuant TL analysis software (version 7.0, GE Healthcare). To allow proper comparison and quantification, the LPS gels from triplicate experiments were exposed on the same phosphor screen along with the standards (see Extended Data Fig. 5b)
For each strain, the arbitrary PL/LPS ratio in the OM was obtained by taking the levels of PLs (represented by [14C]-counts of PL fraction) divided by the LPS levels (represented by gel band density), averaged across three independent replicates (see Extended Data Fig. 5b, Fig. 2b upper panel). The average PL/LPS ratio in the OM for each strain was then compared to that for the WT strain to calculate fold changes (see Fig. 2b lower panel).
Phosphatidylglycerol/Cardiolipin turnover assay (pulse-chase and single time-point (2-h analysis)
PG/CL turnover pulse-chase experiments were performed using the psd2 background, which accumulated PS and PG/CL during growth at restrictive temperature. For each strain, cells were grown in 70 ml LB broth (inoculated from an overnight culture at 1:100 dilution) at the permissive temperature (30°C) until OD600 reached ~0.15 - 0.2. The culture was then shifted for 4 h at the restrictive temperature (42°C) and labelled with [32P]-disodium phosphate (final 1 µCi ml−1) during the last 30 min at the restrictive temperature (42°C). After labeling, cells were harvested by centrifugation at 4,700 × g for 10 min, washed once with cold LB broth (10 ml) and centrifuged again at 4,700 × g for 10 min. Cells were then resuspended in fresh LB broth (70 ml) and the chase was started in the presence of non-radioactive disodium phosphate (1000-fold molar excess) at either the permissive temperature, with or without addition of carbonyl cyanide m-chlorophenyl hydrazone (CCCP; 50 µM), or at the restrictive temperature in the presence of hydroxylamine (HA; 10 mM). At the start (0 min) and different times (15, 30, 45, 90 and 120 min) during the chase, a portion of the culture (either 15 ml or 10 ml) was collected and mixed immediately with equal volume of ice-cold Buffer A containing CCCP (50 µM) and hydroxylamine (10 mM). Cells were harvested by centrifugation at 4,700 × g for 10 min and then resuspended in 6 ml of Buffer B containing CCCP (50 µM) and hydroxylamine (10 mM). Cells were lysed, and fractionated on sucrose density gradients, as described above. 0.8-ml fractions were collected from each tube, as described above. Fractions 7-9 and 12-14 contained the IM and OM fractions, respectively. To extract PLs from the IM and OM pools (2.4 ml), methanol (6 ml) and chloroform (3 ml) were added to make one-phase Bligh-Dyer mixtures. These were incubated at room temperature for 60 min with intermittent vortexing. Chloroform (3 ml) and sterile water (3 ml) were then added to generate two-phase Bligh-Dyer mixtures. After brief vortexing, the lower organic phases were separated from the top aqueous phases by centrifugation at 3,000 × g for 10 min. These were washed once with equal volumes of water and centrifuged as above, and the lower organic phases (containing radiolabelled PLs) recollected and air dried. Finally, the dried PLs were dissolved in 40 µl of a mixture of chloroform:methanol (2:1) and spotted onto silica-gel coated TLC plates (Merck). Equal amounts (in cpm) of radioactivity were spotted for each sample. TLCs were developed in pre-equilibrated chambers containing solvent system chloroform:methanol:water (65:25:4). TLC plates were dried, and visualized by phosphor imaging (STORM, GE healthcare). Densitometric analysis of the PL spots on the phosphor image of TLCs was conducted using the ImageQuant TL analysis software (version 7.0, GE Healthcare). The levels of each major PL species were expressed as a percentage of all detected PL species (essentially the whole lane), and plotted against time (see Figs. 3b-e, Extended Data Fig. 7).
For single time-point analysis, 30-ml cultures were grown and labelled with [32P]-disodium phosphate (final 1 µCi ml−1) at the restrictive temperature. For strains harboring plasmids used for overexpressing OmpC-Mla components, arabinose (0.2 %) was added during growth at the permissive as well as restrictive temperatures. After washing and resuspension in fresh LB broth (30 ml), the chase was started in the presence of non-radioactive disodium phosphate (1000-fold molar excess) at the permissive temperature. At start (0 h) and 2 h during the chase, a portion of the culture (15 and 10 ml) was collected and processed similarly as pulse chase analysis described above. The levels of PG/CL in the membranes at each time point were expressed as a percentage of the sum of PE, PS and PG/CL. For each strain, IM and OM PG/CL turnover were expressed as the difference between percentage PG/CL levels at 0-h and 2-h time points divided by that at 0-h. Average PG/CL turnover values were obtained from three independent experiments conducted (see Fig. 4, Extended Data Fig. 8).
OM permeability assay
OM sensitivity against SDS/EDTA was judged by colony-forming unit (cfu) analyses on LB agar plates containing indicated concentrations of SDS/EDTA. Briefly, 5-ml cultures were grown (inoculated with overnight cultures at 1:100 dilution) in LB broth at 37°C until OD600 reached ~1.0. Cells were normalized according to OD600, first diluted to OD600 = 0.1 (~108 cells), and then serial diluted in LB with seven 10-fold dilutions using 96-well microtiter plates (Corning). Two microliters of the diluted cultures were manually spotted onto the plates and incubated overnight at 37°C.
LpxC overexpression (growth curves and viability assay)
For each strain, a 10-ml culture was inoculated in LB broth supplemented with arabinose (0.2 %) from the overnight culture to make the initial OD600 of 0.05. Cells were grown at 37°C and the OD600 of the cultures was measured hourly. At the start of growth (0 h) and at 4 and 7 h during growth, 100 µl of cells were collected and then serial diluted in LB/cam with six 10-fold dilutions using 96-well microtiter plates (Corning). Five microliters of the non-diluted and diluted cultures were manually spotted on LB/cam agar plates (no arabinose). Plates were incubated overnight at 37°C.
IM (NADH activity) and OM marker (LPS) analysis during sucrose gradient fractionation
The inner membrane enzyme, NADH oxidase, was used as a marker for the IM; its activity was measured as previously described56. Briefly, 30 µl of each fraction from the sucrose density gradient was diluted 4-fold with 20 mM Tris.HCl, pH 8.0 in a 96-well format and 120 µl of 100 mM Tris.HCl, pH 8.0 containing 0.64 mM NADH (Sigma) and 0.4 mM dithiothreitol (DTT, Sigma) was added. Changes in fluorescence over time due to changes in NADH (λex = 340 nm, λem = 465 nm) concentration was monitored using a plate reader (Perkin Elmer). The activity of NADH oxidase in pooled IM and OM fractions relative to the sum of these fractions was determined.
LPS was used as a marker for the OM and detected using LPS dot blots. OM fractions were pooled together and 2 µl of the fractions were spotted on nitrocellulose membranes (Bio-Rad). Spotted membranes were allowed to dry at room temperature for 1 h and then the membranes were probed with antibodies against LPS.
SDS-PAGE and immunoblotting
All samples subjected to SDS-PAGE were mixed with 2X Laemmli reducing buffer and boiled for 10 min at 100°C. Equal volumes of the samples were loaded onto the gels. Unless otherwise stated, SDS-PAGE was performed according to Laemmli using the 12% or 15% Tris.HCl gels57. Immunoblotting was performed by transferring protein bands from the gels onto polyvinylidene fluoride (PVDF) membranes (Immun-Blot® 0.2μm, Bio-Rad) using the semi-dry electroblotting system (Trans-Blot® TurboTM Transfer System, Bio-Rad). Membranes were blocked using 1X casein blocking buffer (Sigma). Mouse monoclonal α-OmpC antibody was a gift from Swaine Chen and used at a dilution of 1:5,00058. Rabbit α-LptE (from Daniel Kahne)56 and α-OmpF antisera (Rajeev Misra)59 were used at 1:5,000 dilutions. Rabbit α-BamA antisera (from Daniel Kahne) was used at 1:40,000 dilution. Mouse monoclonal α-LPS antibody (against LPS-core) was purchased from Hycult biotechnology and used at 1:5,000 dilutions. Rabbit polyclonal α-LamB antibodies was purchased from Bioss (USA) and used at 1:1,000 dilution. α-mouse IgG secondary antibody conjugated to HRP (from sheep) and α-rabbit IgG secondary antibody conjugated to HRP (from donkey) were purchased from GE Healthcare and used at 1:5,000 dilutions. Luminata Forte Western HRP Substrate (Merck Milipore) was used to develop the membranes and chemiluminescent signals were visualized by G:BOX Chemi XT 4 (Genesys version1.3.4.0, Syngene).
Acknowledgments
We thank Zhi-Soon Chong for constructing the ΔmlaC allele, and Manjula Reddy (Center for Cellular and Molecular Biology) for providing the lpxC1272 mutant and the pyciM plasmid. We thank Chee-Geng Chia for performing preliminary experiments. We are grateful to Swaine Chen (NUS), Rajeev Misra (Arizona State U), and Daniel Kahne (Harvard U) for their generous gifts of α-OmpC, α-OmpF, and α-LptE and α-BamA antibodies, respectively. Finally, we thank William F. Burkholder (Institute of Molecular and Cell Biology) and Jean-Francois Collet (U catholique de Louvain) for critical comments and suggestions on the manuscript. This work was supported by the National University of Singapore Start-up funding, the Singapore Ministry of Education Academic Research Fund Tier 1 grant, and the Singapore Ministry of Health National Medical Research Council under its Cooperative Basic Research Grant (NMRC/CBRG/0072/2014) (to S.-S.C.).