Escherichia coli phage-shock protein A (PspA), a 25.3 kDa peripheral membrane protein, is induced under the membrane stress conditions and is assumed to help maintain membrane potential. Here, we report that purified PspA, existing as a large oligomer, is really able to suppress proton leakage of the membranes. This was demonstrated for membrane vesicles prepared from the PspA-lacking E. coli mutants, and for membrane vesicles damaged by ethanol and Triton X-100 prepared from the mutant and the wild-type cells. PspA also suppressed proton leakage of damaged liposomes made from E. coli total lipids. Furthermore, we found that PspA bound preferentially to liposomes containing phosphatidylserine and phosphatidylglycerol. All these effects were not observed for monomer PspA that was prepared by refolding of urea-denatured PspA. These results indicate that oligomers of PspA bind to membrane phospholipids and suppress proton leakage.
Despite of these in vivo studies, our understanding on the molecular functions of PspA has been still limited. PspA is a hydrophilic 25.3 kDa protein that does not contain any sequence characteristics of integral membrane proteins (Cserzo et al., 1997). Recent structural works revealed that PspA molecules assemble into a very large complex (∼1000 kDa) arranged as a ninefold rotationally symmetric ring (Hankamer et al., 2004). Its homologue, chloroplast and cyanobacterial Vipp1, also forms a very large complex (Aseeva et al., 2004; Liu et al., 2007). Interaction of PspA with integral membrane proteins PspB and PspC was reported (Adams et al., 2003), but it is not known how PspA interacts with the membrane. Here, we report an in vitro study on PspA function. We found that purified PspA in a large oligomeric state binds to liposomes containing phosphatidylserine (PS) and phosphatidylglycerol (PG) and repairs proton leakage of the damaged membranes.
Results and discussion
Synthesis of PspA is induced by high-level expression of ATP synthase
In the course of the study of proton-translocating ATP synthase (FoF1) from a thermophilic Bacillus strain PS3, we found that synthesis of a host E. coli protein with estimated molecular mass 23 kDa was induced when the thermophilic FoF1 was expressed in E. coli at high level (Fig. 1A, lane 3). This protein was identified as PspA from N-terminal amino acid sequencing of the band. Induction of PspA synthesis was weak when FoF1 was expressed at low level (lane 2). Interestingly, synthesis of PspA was only weakly induced by high-level expression of a mutant Fo(cE56Q)F1 (lane 4). Overexpression of FoF1 might accompany the increased proton leakage across membranes through proton-channel part Fo unassembled with F1. However, a mutant Fo(cE56Q)F1, lacking an essential carboxylic residue in the c-subunits of Fo, is unable to mediate proton leakage across membrane(Mitome et al., 2004). As a consequence, proton gradient across the membrane expressing this mutant FoF1 might be better maintained than in the case of the wild-type FoF1. Therefore, less induction of PspA by high-level expression of Fo(cE56Q)F1 indicates that proton leakage, rather than the high-level expression of membrane proteins itself, would be responsible for the induction of PspA synthesis.
To study the molecular function of PspA, we cloned pspA gene from E. coli, generated an expression vector and purified PspA from the cells harbouring the vectors. When the E. coli cells expressing PspA were disrupted and centrifuged, PspA protein was mostly recovered in the membrane fractions, whereas it was reported that PspA was approximately equally distributed between cytoplasm and inner membrane fraction (Brissette et al., 1990). This difference might be due to high and endogenous expression of PspA. PspA in the membrane fraction was solubilized with a detergent CHAPS. Once separated from membranes, PspA remained soluble in the absence of detergent and was purified to homogeneity as a soluble protein (Fig. 1A, lane 5). As reported (Hankamer et al., 2004), the purified PspA formed a very large complex (∼1000 k) as shown in gel filtration chromatography (Fig. 1B). Interestingly, when the purified PspA was denatured in 8 M urea and refolded by removal of urea with dialysis, thus-treated PspA did not assemble into a large oligomer, remaining to be a stable monomer as shown in gel filtration chromatography (Fig. 1B). The monomer PspA had the secondary structure as indicated from CD spectrum (Fig. 1C). Peak fractions of oligomer and monomer were used in the following experiments.
Oligomer PspA suppresses proton leakage of the membrane vesicles
We prepared E. coli membrane vesicles from the ΔpspA strain (L2), L2's isogenic wild-type strain (L1), the ΔpspABC strain (J134), and J134's isogenic wild-type strain (K516), and analysed the effect of PspA on proton leakage across the membranes. The electrochemical potential gradient of protons was generated by respiratory proton pumps driven by succinate oxidation. It was confirmed that PspA, either oligomer or monomer, did not have significant effects (< ± 15%) on the activity of succinate oxidation. Upon addition of succinate, generation of inside-acidic ΔpH across the membrane started as shown by 9-amino-6-chloro-2-methoxyacridine (ACMA) fluorescence quenching, which reached the level where succinate-driven proton pumping was balanced with proton leakage across membranes. At the end of assays, ΔpH was abolished by addition of carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP). The maximum quenching, ∼25% of the initial fluorescent intensity, was observed when a saturating 10 mM succinate was given. Because the effect of PspA was masked when proton-pumping activity was strong, succinate concentrations that caused 5–10% quenching were chosen. The activity of succinate oxidation was very different from one strain to another, and different concentrations of succinate were used for each strain. The membrane vesicles from ΔpspA and ΔpspABC cells showed larger fluorescence quenching in the presence of oligomer PspA than in the absence of oligomer PspA (Fig. 2A and B). It appeared that without PspA, membranes are partly proton leaky and added oligomer PspA blocks this leak. We also examined membrane vesicles from the wild-type E. coli cells. The effect of oligomer PspA was not obvious on membrane vesicles from L1 strain (Fig. 2C). Oligomer PspA accelerated initial ACMA fluorescence quenching in the case of membrane vesicles from K561 strain even though the final level of quenching was not changed (Fig. 2D). We compared the amounts of endogenous PspA in the membrane vesicles of L1 and K561 strains by immunoblotting staining intensity, and found that the amount of endogenous PspA of K561 (0.09 mg mg−1 of membrane proteins) was nearly 10 times more than that of L1 (0.01 mg mg−1 of membrane proteins). This difference can explain the different response to the added PspA between membrane vesicles from the two strains. In all cases, the same amount of monomer PspA did not produce any effect. Similar results were obtained when ΔpH was generated by NADH oxidation (data not shown). It appeared that oligomer PspA, but not monomer PspA, suppressed the proton leakage of membranes of the vesicles.
Phage-shock protein A (PspA) suppresses proton leakage of damaged membrane vesicles
Addition of ethanol to the culture medium of E. coli induces synthesis of PspA (Model et al., 1997). Organic solvents tend to increase proton permeability of membranes, and indeed the membrane vesicles pretreated with 33% ethanol failed to generate succinate-induced ΔpH across membrane even after ethanol was diluted to 0.5% (Fig. 3A–D, none). We confirmed that activity of succinate oxidation itself was mostly retained after this ethanol pretreatment (107% for ΔpspA, 92% for ΔpspABC, 117% for L1 and 88% for K561) and was not changed by the presence of PspA, in either the monomer or oligomer state (See Fig. S1). When the ethanol-damaged membranes from ΔpspA and ΔpspABC strains were incubated with oligomer PspA, the succinate-induced ΔpH generation was observed in a dose-dependent manner, reaching the maximum ∼5% of quenching (Fig. 3A and B, traces 1–5). The plot of the extent of fluorescence quenching versus the amount of added oligomer PspA indicated that the ratios of amounts of oligomer PspA that caused a saturating effect relative to the amounts of membrane proteins and membrane phospholipids (w/w) were roughly estimated to be ∼0.3 and ∼0.7 respectively (protein : lipid ratio of E. coli membrane is ∼5:2; Burnell et al., 1980). Ethanol-damaged membrane vesicles from the wild-type strains L1 and K561 also exhibited the succinate-induced ΔpH generation only in the presence of oligomer PspA (Fig. 3C and D). Under the same assay conditions, membrane vesicles without ethanol pretreatment showed 8–10% quenching in the presence of oligomer PspA (see Fig. 2A–D). Monomer PspA did not show the effect in all experiments (Fig. 3A–D, mono-PspA). Therefore, it was clear that proton leakage of ethanol-damaged membranes was efficiently repaired by incubation with oligomer PspA.
Detergents solubilize membrane lipids and break the ion-impermeable nature of membrane even at low concentrations. Oligomer PspA was also effective to repair the detergent-damaged membrane vesicles. The membrane vesicles pretreated with a low concentration (0.1%) of Triton X-100 failed to establish succinate-induced ΔpH even after dilution of Triton X-100 to 0.0015% (Fig. 3E–H, none). The activity of succinate oxidation was mostly retained after this Triton X-100 pretreatment (104% for ΔpspA, 98% for ΔpspABC, 118% for L1 and 91% for K561) and was not affected by the presence of PspA, in either the monomer or oligomer state. Similar to the cases of ethanol, Triton X-100-damaged membrane vesicles from each of four strains (ΔpspA, ΔpspABC, L1 and K561) were able to generate succinate-induced ΔpH when oligomer PspA was present (Fig. 3E–H, oligo-PspA). Monomer PspA did not show the effect in all experiments (Fig. 3E–H, mono-PspA). The similar effect of PspA was also observed when NADH was used as a substrate for respiratory proton pump (data not shown). These results suggest that PspA has the ability to suppress proton leakage of damaged membranes, and that PspB and PspC are not prerequisite for this function. This contention was further supported from the next experiments using reconstituted liposomes.
Phage-shock protein A (PspA) suppresses proton leakage of the damaged liposomes
The effect of PspA on the liposomes was examined. Liposomes were prepared from E. coli total lipid extracts. To incorporate a proton pump in the liposomes, we reconstituted FoF1 into liposome membrane. The addition of ATP to the FoF1-liposomes caused ∼40% quenching of ACMA fluorescence. As the membranes of FoF1-liposomes were tight enough against proton leakage, simple addition of oligomer PspA did not make any difference (data not shown). Therefore, we pretreated FoF1-liposomes with 33% ethanol or 0.1% Triton X-100 as describe above. ATP-hydrolysing activity of FoF1 was decreased to ∼40% by ethanol pretreatment and to ∼50% by Triton X-100 pretreatment, and PspA did not have effects on these ATP-hydrolysing activities. The damaged FoF1-liposomes failed to make ΔpH by addition of ATP (Fig. 4A and B, none). When oligomer PspA was present, significant ATP-induced ΔpH formation was observed (Fig. 4A and C, oligo-PspA). This result was expected only when FoF1 was active in proton-pumping and the membrane was tight enough to prevent rapid proton leak. Therefore, it was concluded that oligomer PspA repaired proton leakage of the damaged membrane to make proton-pumping activity of FoF1 visible. Consistent with specific binding of oligomer PspA to PG and PS as mentioned in the next paragraph, oligomer PspA did not have effects on damaged FoF1-liposomes whose lipid composition was 100% phosphatidylcholine (PC) (Fig. 4C and D). In any cases, monomer PspA did not have any effect (Fig. 4A and C, mono-PspA).
Phage-shock protein A (PspA) binds to PS and PG
When PspA associates with membrane surface, it should have interactions with other membrane protein(s) and/or lipid head group(s). PspB and PspC would be themost possible candidate(s) for the former case (Adams et al., 2003). However, PspA was found in the membrane fraction no matter whether it was expressed in the ΔpspA or ΔpspABC strains of E. coli. Then, we examined whether PspA could interact directly with phospholipid head group. Taking into account lipid composition of E. coli membrane [75–80% phosphatudylethanolamine (PE), 15–20% PG, 1–5% CL, < 1% PS], we prepared liposomes that contained PC plus one of the above phospholipids. PC was not contained in E. coli membrane but used as a major component of liposomes because otherwise liposomes are not stabilized. Liposomes were incubated with purified PspA, applied to sucrose density gradient centrifugation for 15 h, and fractions were analysed by immunoblotting with anti-PspA antibody (Fig. 5A). A small amount of fluorescently labelled PC (0.5%) was contained in the liposomes for easy detection of the liposome fraction (Fig. 5B, top panel). Free PspA proteins, in both oligomer and monomer states, were recovered near the bottom of centrifuge tubes. Liposomes made from 100% PC did not bind oligomer PspA under the conditions tested. Binding of oligomer PspA was observed for liposomes containing either PG or PS, but not for those containing PE and CL (Fig. 5B). Interactions of monomer PspA with liposomes were also tested. Monomer PspA bound to liposomes with 100% PC. As all liposomes contained PC, binding of monomer PspA to PG and PS was not determined. We concluded that PspA preferentially binds to PS and PG and this ability is dependent on oligomer structure of PspA. It is interesting to note that the content of PG in E. coli membrane increased by treatment with heat or benzyl alcohol (Shigapova et al., 2005).
Possible mechanism of PspA function
We have shown here that oligomer PspA binds to PG and PS and can repair proton leakage of the damaged phospholipid membranes. In the damaged membranes, protons should diffuse out through distorted phospholipid bilayer. The exact mechanism of how oligomer PspA can suppress such proton leakage is yet unknown, but its specific bindings to PG and PS should contribute this function. It is likely that large oligomer of PspA (Hankamer et al., 2004) interacts with many PG and PS molecules in the membrane simultaneously and restrict their motion in the membrane, which alters the physical properties of the membrane, such as fluidity and stiffness. If it is really the case, suppression of proton leakage by oligomer PspA is caused as a result of change of physical properties of membrane. According to electron microscopic observations (Hankamer et al., 2004), the area of the one flat side of the ring-like assembly (presumably 36 mer) of PspA is assumed to be ∼300 nm2. If the ring complexes associate with the surface of membrane, rough estimation suggests that the observed amount of PspA to suppress proton leakage [PspA/phospholipid = ∼0.7(w/w) = ∼0.02 (mol mol−1) = ∼0.7(nm2 nm−2)] is almost enough to cover whole outer surface of phospholipid bilayers of the membrane vesicles. Estimated from phospholipid content of E. coli, this amount of PspA [PspA/phospholipid = ∼0.7(w/w)] is roughly equivalent to several per cents of total cellular proteins in E. coli cell that is well attainable when its synthesis is induced by the stress. Therefore, it is likely that the observed phenomena here are physiologically relevant.
Monomer PspA does not have any effect on the damaged membrane. A very recent paper reported that VIPP1, a homologue of PspA in plant plastids, assembles into a large oligomer by the assistance of molecular chaperone, Hsp70B and its cofactor (Liu et al., 2007). An attractive hypothesis to explain the effect of PspA in vivo is that, upon exposure to stress conditions, production of PspA and chaperones is induced, chaperones assist assembly of monomer PspA into oligomers, which then bind to the inside surface of the damaged cytoplasmic membrane to make the lattice-like scaffold, damaged membranes are stabilized and proton leakage is blocked. Further experiments are required to examine the validity of the hypothesis.
Strains and plasmids
Escherichia coli strains L1 (wild-type), L2 [L1 ΔpspA::kan] (Brissette et al., 1991), K561 (wild-type) and J134 [K561 ΔpspABC::kan] (Weiner et al., 1991) were kindly provided by Drs P. Model, M. Russel and J. Dworkin. To make expression vectors for PspA, we amplified pspA gene from a genomic DNA of E. coli W3110 with polymerase chain reaction (PCR) by using primers containing restriction sites for EcoRI and PstI. The amplified pspA digested with EcoRI and PstI was ligated into the plasmid pTR19-ASDS (Suzuki et al., 2002) previously digested with both restriction enzymes. The obtained plasmid was named pTR19-pspA. The region amplified with PCR was verified by nucleotide sequencing.
Purification of PspA
Escherichia coli strain DK8 (Suzuki et al., 2000) harbouring pTR19-pspA was aerobically cultivated in 2× YT medium containing ampicillin (100 μg ml−1) at 37°C. Isopropyl-β-d-thiogalactoside (2 mM final concentration) was added at 6 h, and culture was continued for 2 h. The wet cells harvested from the culture were suspended in 7 vols of 10 mM HEPES/KOH buffer (pH 7.5) containing 5 mM MgCl2 and 10% glycerol, and disrupted twice by French press (1.2 bars). After removing cell debris, the membrane fraction was collected by a centrifugation (158 700 g, 1 h, 4°C). The obtained membrane fraction was washed and resuspended into the original wet-cell volume of the same buffer. This suspension was used as the (inverted) membrane vesicles in this study. To solubilize PspA from the membrane, the vesicles expressing PspA were diluted threefold with 10 mM HEPES/KOH buffer (pH 7.5) containing 100 mM NaCl and 1.2% CHAPS, and centrifuged (196 000 g, 20 min, 4°C). To the precipitated fraction, this solubilization procedure was repeated three times, and the combined supernatant was applied to a DEAE-TOYOPEARL column (TOSOH) previously equilibrated with 20 mM HEPES/KOH buffer (pH 7.5) containing 5 mM EDTA. PspA was eluted from the column with a linear gradient of KCl (0–2 M). The combined PspA fractions were concentrated to 10 mg ml−1 by a centrifugal concentrator and applied to a gel filtration column (Superose 6/20H, Pharmacia) equilibrated with 20 mM HEPES/KOH buffer (pH 7.5) containing 100 mM NaCl and 2 mM EDTA. The molecular weight standards of the chromatography were blue dxtran (2000 k), thyroglobulin (669 k), apoferritin (443 k), β-amylase (200 k), alcohol dehydrogenase (150 k), albumin (66 k) and carbonic anhydrase (29 k). The chromatography produced a major peak at the position of > 1000 kDa corresponding to the reported large complex of PspA (Hankamer et al., 2004). Monomer PspA was prepared with urea treatment. Purified PspA was dissolved in 10 mM HEPES/KOH buffer (pH 7.5) containing 150 mM KCl, 2 mM EDTA and 8 M urea at 4°C. The solution was dialysed against 1000-fold volume of 10 mM HEPES/KOH buffer (pH 7.5) containing 150 mM KCl and 2 mM EDTA at 4°C. The dialysis buffer was changed twice, and the dialysed solution was applied to a gel filtration column (Superose 6/20H, Pharmacia). PspA was eluted as a single peak at the position corresponding to monomer PspA. The peak fractions of oligomer and monomer PspA were collected individually, concentrated to 1 mg ml−1 in 10 mM HEPES/KOH buffer (pH 7.5) by a centrifugal concentrator and frozen with liquid N2. Circular dichroism spectra was obtained with JASCO-J820 CD spectrometer for oligomer PspA, PspA in 8 M urea and monomer PspA at protein concentration 10.6 μg ml−1. The membrane vesicles of E. coli L2 and J134 harbouring pTRc99A (Pharmacia) were prepared by the same procedures used for E. coli DK8 harbouring pTR19-pspA, except that isopropyl-β-d-thiogalactoside was not used. Protein concentrations were determined using a BCA protein assay kit (Pierce) with bovine serum albumin as a standard.
Preparation of FoF1-liposome
ATP synthase (FoF1) of thermophilic Bacillus strain PS3 with 10 histidines at N-terminus of its β subunits (Suzuki et al., 2002) was used in this study. FoF1 was reconstituted into liposomes by the freeze-thaw method (Kaim and Dimroth, 1998). E. coli total lipid extracts (44 mg, Avanti) was suspended in 1 ml of 10 mM HEPES/KOH buffer (pH 7.5) containing 5 mM MgCl2 and 10% glycerol, and subjected to sonication for 1 min on ice water. The mixture was frozen with liquid N2 and thawed at room temperature to make liposome suspension. Purified FoF1 (10 μl of 10 mg ml−1) was mixed with 200 μl of the liposome suspension, and the mixture was again subjected to the freeze-thaw procedures.
Analyses of ΔpH generation
To assess proton permeability of membranes, ΔpH generation across membranes was monitored with fluorescence at 480 nm (excitation at 410 nm) of a fluorescent indicator, ACMA (Molecular Probes) (Aggeler et al., 1995). All experiments were carried out at 40°C. E. coli membrane vesicles (5.5 mg protein ml−1) was mixed with 1.5 vols of PspA solution (1 mg ml−1), and the sample mixture (30 μl) was injected into 1.2 ml of the assay solution [10 mM HEPES/KOH buffer (pH 7.5), 5 mM MgCl2, 100 mM KCl, 0.3 μg ml−1 ACMA]. The reaction was initiated by adding succinate. The amount of added succinate was adjusted to the level that caused 5–10% ACMA fluorescence quenching, where the PspA effect was most pronounced. Succinate oxidation activities were very different among E. coli strains, and final concentrations of succinate were chosen to produce proton-pumping activity suitable to measure the PspA effect: 1 mM for vesicles from strain L2 and K561, 0.3 mM for those from L1, and 25 μM for those from J134. In the case when NADH was used as a substrate for respiratory proton pump, 7 μM (final concentration) was offered for vesicles from L2 and J134. After a couple of minutes, the assays were terminated by the addition of FCCP. For the analyses of damaged membrane vesicles, membrane vesicles (3.7 mg protein ml−1) pretreated in 33% of ethanol or in 0.1% Triton X-100 were mixed with equal volume of PspA solution (1 mg ml−1 or varied concentrations), and the mixture (36 μl) was injected into 1.2 ml of the assay solution. For the analyses of the effect of PspA on FoF1-liposomes, FoF1-liposome suspension (125 μg FoF1 ml−1) was mixed with 2 vols of PspA solution (1 mg ml−1), and the mixture (18 μl) was added to 1.2 ml of the assay solution. The ΔpH generation was initiated by adding 1 mM ATP. When indicated, FoF1-liposome suspension pretreated in 33% ethanol was mixed with 1.5 vols of PspA solution (1 mg ml−1), and the mixture (21 μl) was injected into 1.2 ml of the assay solution.
Assays of succinate oxidation activity
Succinate oxidation activity of membrane vesicles was estimated as described (Sakamoto et al., 1996) under the same conditions of the ΔpH assays. The concentrations of succinate were 1 mM (L2 and K561), 0.3 mM (L1) and 25 μM (J134). At first, succinate dehydrogenase activity of membrane vesicles was measured with the reduction of oxidized form of ubiquinone (Q1) that was monitored by decrease of absorbance at 278 nm. Re-oxidation of Q1 by terminal oxidase was blocked by KCN. The same sample mixtures that were used in ΔpH formation assay were added to 1.2 ml of 10 mM HEPES/KOH buffer, pH 7.5, containing 100 mM KCl, 5 mM MgCl2, 0.1% SML (Mitsubishi-Kagaku Foods, Japan), 100 μM oxidized form of Q1, and 5 mM KCN. The reaction was initiated by the addition of succinate. The average reducing rate in a time period from 10 to 80 s after initiation of the reaction was taken as activity. Next, the rate of reduction of oxidized form of Q1 was measured in the absence of KCN. Without KCN, reduced Q1 was oxidized by the terminal oxidase and the apparent reducing rate of Q1 was slowed down. The difference between rates in the presence and absence of KCN was taken as an index of succinate oxidation activity. The relative succinate oxidation rates of four strains thus measured were 100% (L2), 44% (J134), L1, 69% (L1) and 83% (K561). NADH oxidase activity (respiration driven by NADH) was estimated at 37°C by monitoring absorbance at 340 nm in 1.2 ml of 10 mM HEPES/KOH, pH 7.5, 100 mM KCl, 5 mM MgCl2 and 200 μM NADH. The reaction was initiated with the addition of the sample mixture, and the average dehydrogenase activity (5–20 s) after the initiation was taken as activity.
Assay of ATPase activity
ATP-hydrolysing activity of FoF1-ATP synthase was determined at 40°C in 50 mM HEPES/KOH buffer, pH 7.5, containing 100 mM KCl, 5 mM MgCl2, 5 mM KCN and 1 mM ATP with ATP-regenerating system (Suzuki et al., 2000). The reaction was initiated by the addition of the same FoF1-liposome suspension used for ΔpH generation assays. Average hydrolysis rates (3–6 min) after initiation of the reactions were measured.
Binding of PspA to phospholipid
Soybean PC and synthetic phospholipids 1,2-dimyristoyl-sn-glycero-3-phosphoethanolamine, 1,2-dimyristoyl-sn-glycero-3-phospho-l-serine, 1,2-dimyristoyl-sn-glycero-3-[phospho-rac-(1-glycerol)] and 1,1′,2,2′-tetramyristoyl cardiolipin (AVANTI) were used for liposome formation. Five kinds of liposomes with different lipid compositions (100% PC, 25% PC + 75% PE, 90% PC + 10% CL, 90% PC + 10% PG, 95% PC + 5% PS) were prepared by the reverse-phase evaporation method (Szoka and Papahadjopoulos, 1978) with some modifications. PC (20 mg) was dissolved in 800 μl of chloroform in a 30 ml round-bottom flask, and the indicated amount of other phospholipids was added to the solution. A fluorescent lipid 2-(12-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino)dodecanoyl-1-hexadecanoyl-sn-glycero-3-phosphocholine (excitation 460 nm, emission 534 nm, Molecular Probe) was also added to all solutions (0.5% of the total lipid). Chloroform was removed under reduced pressure by a rotary evaporator at 37°C. The dried phospholipids were redissolved in 1.5 ml of diethylether, and 500 μl of 20 mM HEPES/KOH (pH 7.5) buffer was added. The resulting two-phase system was sonicated for 2 min on ice water, and the mixture became an opalescent dispersion. The organic solvent was removed by a rotary evaporator under the reduced pressure (150 mmHg) for 10 min at 25°C, and the lipids gradually formed a viscous gel. The pressure was adjusted to 200 mmHg and kept for more than 10 min until the formation of a viscous gel was completed. The evaporation was further continued at 400 mmHg for 30 min at 25°C to remove chloroform completely. A resultant turbid solution was subjected to ultrafiltration through a 0.45 μm filter, followed by a 0.2 μm filter. The filtrated liposomes were adjusted to be a final lipid concentration of 16 mg ml−1. One microlitre of PspA (1 mg ml−1) was added to 50 μl of each liposome solutions (16 mg ml−1) and carefully loaded on 0–30% sucrose gradient in 20 mM HEPES/KOH buffer (pH 7.5) in centrifugation tubes. The tubes were subjected to centrifugation (154 000 g, 15 h, 4°C) with a swing rotor, and the fractions were analysed by immunoblotting with antibody raised against PspA.
We thank Drs P. Model, M. Russel (Rockefeller University), and J. Dworkin (Harvard University) for their kind gifts of E. coli mutant strains. We thank our colleagues Y. Matsumoto and Y. H. Watanabe for helpful discussions.