Function of Escherichia coli MsbA, an Essential ABC Family Transporter, in Lipid A and Phospholipid Biosynthesis*

The Escherichia coli msbA gene, first identified as a multicopy suppressor of htrB mutations, has been proposed to transport nascent core-lipid A molecules across the inner membrane (Polissi, A., and Georgopoulos, C. (1996) Mol. Microbiol. 20, 1221–1233). msbA is an essentialE. coli gene with high sequence similarity to mammalian Mdr proteins and certain types of bacterial ABC transporters.htrB is required for growth above 32 °C and encodes the lauroyltransferase that acts after Kdo addition during lipid A biosynthesis (Clementz, T., Bednarski, J., and Raetz, C. R. H. (1996) J. Biol. Chem. 271, 12095–12102). By using a quantitative new 32Pi labeling technique, we demonstrate that hexa-acylated species of lipid A predominate in the outer membranes of wild type E. coli labeled for several generations at 42 °C. In contrast, in htrB mutants shifted to 42 °C for 3 h, tetra-acylated lipid A species and glycerophospholipids accumulate in the inner membrane. Extra copies of the cloned msbA gene restore the ability ofhtrB mutants to grow at 42 °C, but they do not increase the extent of lipid A acylation. However, a significant fraction of the tetra-acylated lipid A species that accumulate in htrBmutants are transported to the outer membrane in the presence of extra copies of msbA. E. coli strains in which msbA synthesis is selectively shut off at 42 °C accumulate hexa-acylated lipid A and glycerophospholipids in their inner membranes. Our results support the view that MsbA plays a role in lipid A and possibly glycerophospholipid transport. The tetra-acylated lipid A precursors that accumulate in htrB mutants may not be transported as efficiently by MsbA as are penta- or hexa-acylated lipid A species.

The Escherichia coli msbA gene, first identified as a multicopy suppressor of htrB mutations, has been proposed to transport nascent core-lipid A molecules across the inner membrane (Polissi, A., and Georgopoulos, C. (1996) Mol. Microbiol. 20, 1221-1233). msbA is an essential E. coli gene with high sequence similarity to mammalian Mdr proteins and certain types of bacterial ABC transporters. htrB is required for growth above 32°C and encodes the lauroyltransferase that acts after Kdo addition during lipid A biosynthesis (Clementz, T., Bednarski, J., and Raetz, C. R. H. (1996) J. Biol. Chem. 271, 12095-12102). By using a quantitative new 32 P i labeling technique, we demonstrate that hexa-acylated species of lipid A predominate in the outer membranes of wild type E. coli labeled for several generations at 42°C. In contrast, in htrB mutants shifted to 42°C for 3 h, tetraacylated lipid A species and glycerophospholipids accumulate in the inner membrane. Extra copies of the cloned msbA gene restore the ability of htrB mutants to grow at 42°C, but they do not increase the extent of lipid A acylation. However, a significant fraction of the tetraacylated lipid A species that accumulate in htrB mutants are transported to the outer membrane in the presence of extra copies of msbA. E. coli strains in which msbA synthesis is selectively shut off at 42°C accumulate hexa-acylated lipid A and glycerophospholipids in their inner membranes. Our results support the view that MsbA plays a role in lipid A and possibly glycerophospholipid transport. The tetra-acylated lipid A precursors that accumulate in htrB mutants may not be transported as efficiently by MsbA as are penta-or hexaacylated lipid A species.
Many of the key enzymatic reactions involved in Escherichia coli LPS biosynthesis are known (2,4). Lipid A with its attached core (designated core-lipid A) and the O-antigen are synthesized as separate units (2,4). The important studies of Osborn and co-workers (9 -14) suggest that O-antigen is ligated to nascent core-lipid A on the periplasmic surface of the inner membrane (Fig. 1). The core-lipid A domain is made in association with the inner surface of the inner membrane (2,4). It must therefore be flipped across the inner membrane prior to O-antigen ligation. O-antigen attachment is not required for core-lipid A transit to the outer membrane, given that LPS consisting only of Kdo 2 -lipid A (Fig. 2) supports the growth of E. coli (2,4) and is translocated efficiently (7,10).
Elucidation of the molecular details of LPS transit from the cytoplasm to the outer membrane (7, 10) remains one of the most intriguing, unsolved problems of LPS biochemistry (Fig.  1). A possible clue to the mechanism of core-lipid A transport within the inner membrane has emerged from the recent discovery that HtrB functions as the lauroyltransferase that acts immediately after Kdo addition during lipid A biosynthesis ( Fig. 2) (15,16). Cells harboring htrB mutations fail to grow above 32°C (17), display unusual surface bulging (17), overproduce glycerophospholipids at 42°C (17,18), and accumulate under-acylated species of lipid A (19).
htrB insertion mutations were first isolated by Karow and Georgopoulos (17) during a search for new E. coli heat shock genes. Two multicopy suppressors of htrB mutations (designated msbA and msbB) were also reported (20,21). The msbB gene displays sequence homology to htrB (20,22) and encodes a distinct acyltransferase (16,19) (Fig. 2) that can substitute for htrB at 42°C, when overexpressed on high copy plasmids. The msbA gene shows no sequence relationship to htrB (21). Instead, msbA encodes an ABC family transporter with high amino acid sequence similarity to mammalian Mdr proteins (21). msbA is an essential gene (21), whereas msbB is not (20,23). When introduced on low copy plasmids, msbA complements most of the htrB mutant phenotypes, such as the morphological alterations, the overproduction of phospholipids, and the inability to grow above 32°C (21). Polissi and Georgopoulos (24) have recently shown that MsbA is an inner membrane, ATP-binding protein, as predicted from its sequence. When msbA and an essential downstream gene, orfE, are both inactivated because of an insertional mutation in msbA in a strain with a temperature-sensitive replication plasmid expressing both MsbA and OrfE, LPS accumulates in the inner membrane at 42°C (24). Similar findings have recently been reported (25) using the msbA(valA) and orfE(valB) genes of Francisella novicida. Moreover, LPS accumulates in the inner membranes of htrB-deficient mutants (24). Introduction of msbA and orfE on low copy plasmids partially restores translocation of LPS to the outer membrane at 42°C in htrB mutants (24). Taken together, these findings suggest that the HtrB-catalyzed transfer of laurate to lipid A ( Fig. 2) may be necessary for efficient core-lipid A transport and that MsbA and/or OrfE may be components of the transport machinery.
Garrett et al. (26) have recently discovered that orfE (now designated lpxK) encodes the 4Ј-kinase of the lipid A pathway (53), a key enzyme that functions several steps before HtrB (Fig. 2). We have therefore re-evaluated the contribution of orfE/lpxK and msbA to the suppression of the htrB phenotype. We now show that the rescue of htrB mutants by overexpression of msbA is not accompanied by an increase in the extent of lipid A acylation at 42°C. By using a quantitative new 32 P i labeling assay, we demonstrate that htrB-deficient cells accumulate tetra-acylated lipid A molecular species and glycerophospholipids within their inner membranes at 42°C, whereas msbA-deficient strains accumulate hexa-acylated lipid A and glycerophospholipids in their inner membranes, even when extra copies of orfE/lpxK are provided in trans. Our data strengthen the view that MsbA is involved in the transport of newly synthesized core-lipid A molecules across the inner membrane (24,25) and suggest that MsbA may also transport glycerophospholipids.

EXPERIMENTAL PROCEDURES
Materials-32 P i and [1-14 C]dipalmitoyl-phosphatidylcholine were purchased from NEN Life Science Products. Glass-backed Silica Gel 60 thin layer chromatography plates (0.25 mm) were from Merck, Germany. Restriction enzymes EcoRI, EcoRV, HindIII, and SacII were purchased from New England Biolabs, and the Wizard Plus Miniprep DNA purification system was from Promega. Triton X-100 was obtained from Pierce and NADH from Sigma. Pyridine, methanol, and 88% formic acid were obtained from Mallinckrodt. Chloroform was purchased from EM Science. Chemicals of analytical grade were used throughout.
Bacterial Strains, Plasmids, and Growth Conditions-The bacterial strains and plasmids used in this study are described in Table I. Cells were generally grown at 30 or 42°C in LB broth, consisting of 10 g of NaCl, 5 g of yeast extract, and 10 g of tryptone per liter (27). Antibiotics were added when necessary at final concentrations of 50 g/ml for ampicillin, 12 g/ml for tetracycline, 10 g/ml for chloramphenicol, 50 g/ml for spectinomycin.
Recombinant DNA Techniques and Construction of pEH2-Plasmid DNAs were isolated by using the Wizard Plus Miniprep DNA purification system (Promega). Restriction endonucleases, T4 DNA ligase, and shrimp alkaline phosphatase were used according to the manufacturers' specifications. Other recombinant DNA techniques were performed as described by Ausubel et al. (28).
The pEH2 subclone containing the orfE/lpxK gene was made by digesting plasmid pGAP14 DNA with EcoRI and HindIII (24) and ligating the 1.9-kilobase pair fragment derived from pGAP14 with EcoRI and HindIII-digested pUC18 vector DNA. The ability of pEH2 to direct the overexpression (ϳ5-fold) of the lipid A 4Ј-kinase was confirmed by assaying extracts of cells (26) containing pEH2 compared with vector controls.
Construction of pKW2 Bearing the msbA Gene-Genomic E. coli DNA was isolated from strain W3110 according to Ausubel et al. (28). After the chromosomal DNA was isolated, the msbA gene was amplified by the polymerase chain reaction (PCR). The primer sequences used were as follows: the forward primer, 5Ј GCG CTC TAT CTC GCT CAA GGC TGG CGG ATT AGC AGC CTC AGG GAG C 3Ј and the reverse primer, 5Ј GCG CGC ATC GAT GGC CAT ACA ACC AGG AGA GTG GCA GCA ATA GCC GCC. The PCR reaction contained 100 ng of W3110 chromosomal DNA template, 0.2 g of each primer, 10 mM Tris-HCl, pH 9.2, 1.5 mM MgCl 2 , 75 mM KCl, 200 M each dNTP, and 1 unit of Taq polymerase (Stratagene). The mixture was subjected to 5 min of denaturation at 94°C, then 30 cycles of 94°C for 1 min, 50°C for 1 min, 72°C for 2 min, and ended with a 7-min extension at 72°C in a Perkin-Elmer GeneAmp PCR system 2400. The PCR product was then ligated into the TA cloning vector and transformed into the appropriate competent cells (Invitrogen Corp.) according to the manufacturer's instructions. Plasmid bearing colonies were selected by growth on LB agar plates containing 50 g/ml ampicillin and 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside at 37°C. Plasmid was isolated from positive (white) colonies, and the presence of the insert was confirmed by digestion with HindIII and EcoRV. Due to the reported sensitivity of the MLK53 cells to the presence of the msbA on high-copy vectors (21), we chose to move the insert into the low-copy vector, pACYC184 (New England Biolabs, Inc.). The TA plasmid bearing the msbA gene (pKW1) was digested with HindIII and EcoRV. The reaction was subjected to gel electrophoresis, and the insert fragment was purified using the GeneClean system (Bio 101). Simultaneously, the plasmid pACYC184 was digested with Hin-dIII and EcoRV, phosphatase-treated, and purified as described above for the insert. The digested pACYC184 and the insert were ligated. The ligation reactions were then transformed into SureCells (Stratagene), and plasmid containing cells were selected by growth on LB agar containing 30 g/ml chloramphenicol. The plasmid was isolated from chloramphenicol-resistant cells, and the presence of the insert was confirmed by digestion with HindIII and EcoRV. The pACYC184 containing msbA was named pKW2.
The DNA sequence of the insert was confirmed by automated fluorescent sequencing at the Duke University DNA Analysis Facility, Durham, NC. The sequencing of pKW2 revealed that the plasmid contained the predicted junctions between pACYC184 (HindIII and EcoRV) and the expected TA sequence. However, the first 17 nucleotides of the forward primer, 5Ј GCG CTC TAT CTC GCT CAA GGC TGG CGG ATT AGC AGC CTC AGG GAG C 3Ј (shown in bold) were not present. The proposed function of MsbA as an inner membrane "flippase" for nascent core-lipid A molecules is indicated. Nothing is known about possible protein factors that might assist in the translocation of completed LPS and core-lipid A molecules across the periplasm or in their further assembly within the outer membrane. However, transport of core-lipid A molecules to the outer membrane can occur independently of ligation to O-antigen. Alternative transport models are discussed elsewhere (7, 10) but seem less likely, given the experimental evidence for periplasmic O-antigen intermediates (9,12). Recently, RfbX(Wzx) has been proposed as a possible O-antigen flippase (51). Quantitative in vitro assays for Oantigen polymerization and ligation have not been described (2). The accumulatiom of phospholipids in the inner membranes of msbA mutants suggests that MsbA might also transport phospholipids.
The absence of this portion of the primer sequence did not disrupt the coding sequence of msbA, since the primer was designed to a segment of DNA 92 nucleotides upstream of the start codon of msbA. The msbA coding sequence determined for pKW2 (data not shown) exactly matched the published msbA sequence (21).
Finally, the htrB mutant MLK53 was transformed with pKW2 at 30°C, selecting for tetracycline and chloramphenicol resistance. MLK53 is temperature-sensitive for growth at 44°C. When competent MLK53 cells were transformed at 30°C with pKW2, the temperature sensitivity was suppressed in all the transformants. In other experiments, MLK53 was transformed with pKW2 and subjected immediately to selection at 44°C. The frequency of transformants was the same as observed at 30°C.
Growth of 32 P i -Labeled Cells-Cells were grown and labeled using the methods described previously (19,29). In general, labeling of the glycerophospholipids and lipid A was achieved by dilution of overnight cultures with 125 ml of pre-warmed fresh LB broth containing appropriate antibiotics to yield A 550 ϭ 0.05. After addition of 4 Ci/ml 32 P i , non-temperature-sensitive and wild type cells were grown at 30 or 42°C (as indicated) until A 550 ϭ 0.6. All temperature-sensitive mutants were labeled for 3 h after a shift to 42°C. At the time of harvest, their A 550 was 0.4 -0.6.
Membrane Fractionation-Inner and outer membranes were separated by isopycnic sucrose gradient centrifugation following the procedure of Guy-Caffey et al. (30) with several modifications. Disruption of the lysozyme-and EDTA-treated spheroplasts was performed using a Branson sonifier (model 250) equipped with a microtip (31). Sonic disruption was performed while the sample was immersed in an icewater bath. One-minute intervals were allowed for cooling between each 30-s period of sonic oscillation (80 watts of power output). The whole particulate fraction was isolated on a sucrose cushion and washed once. Inner and outer membranes (recovered from 125 ml of cells and containing about 2 ϫ 10 6 cpm of 32 P) in 1.5 ml of 10 mM Tris acetate buffer, pH 7.8, with 0.5 mM EDTA were separated on a sucrose gradient prepared by layering 0.4 ml of 60% sucrose, 0.9 ml of 55% sucrose, 2.2 ml of 50% sucrose, 2.2 ml of 45% sucrose, 2.2 ml of 40% sucrose, 2.2 ml of 35% sucrose, and 0.4 ml of 30% sucrose (30). The gradient was buffered with 10 mM Tris acetate, pH 7.8, containing 0.5 mM EDTA. The separated membranes were collected in ϳ0.4-ml fractions by piercing the bottom of the centrifuge tube. Radioactivity in each fraction was measured with a scintillation counter.
Enzymatic Assays-Phospholipase A activity (an outer membrane marker) was determined as described (32) with minor modifications. The reaction mixture, prepared in a 0.6-ml plastic microcentrifuge tube, type of strains lacking htrB (21) and allows growth with only four acyl chains on lipid A. At 30°C, other acyltransferases, such as Ddg (37) (not shown), may substitute for HtrB. Finally, core sugars and other modifying groups are incorporated, including the additional phosphate residue present on a portion of the lipid A moieties of E. coli K-12, designated the trisphosphate species (bottom structure, dashed bond).

FIG. 2. Functions of orfE/lpxK, htrB, and msbB in the biosynthesis of Kdo 2 -lipid A in E. coli.
Five enzymes are required for the synthesis of the intermediate designated DS-1-P, the physiological substrate for the 4Ј-kinase (2). Structural genes encoding each of the indicated enzymes are in italics. The gene for the 4Ј-kinase has recently been identified as orfE (the downstream gene co-transcribed with msbA) (26). orfE is now designated lpxK. Following phosphorylation by the 4Ј-kinase, the Kdo transferase (encoded by kdtA) adds two Kdo residues to lipid IV A to form Kdo 2 -lipid IV A (41,52). The Kdo-dependent acyltransferases, HtrB and MsbB, then add laurate and myristate, respectively, to form Kdo 2 -lipid A (16,19). Kdo 2 -lipid A (with only five acyl chains) is sufficient to support the growth of E. coli. However, overexpression of msbA can suppress the temperature-sensitive pheno- This study contained 25 mM Tris-HCl, pH 8.0, 10 mM CaCl 2 , 0.1% Triton X-100, 5.5 M [1-14 C]dipalmitoyl-phosphatidylcholine (ϳ10,000 cpm/tube), and 2 l of membrane sample in a final volume of 10 l. After incubation at 37°C for 60 min, the reaction mixture was converted to a two-phase Bligh and Dyer mixture by addition of 22 l of chloroform/methanol (1:1, v/v). The lower phase was spotted onto a Silica Gel 60 high performance thin layer chromatography plate. The plate was developed in the solvent chloroform/methanol/water (65:25:4, v/v). The plate was dried and exposed overnight to a PhosphorImager screen. One unit of phospholipase A activity is defined as the amount of enzyme that generates 1 nmol of free fatty acid per min at 37°C. NADH oxidase activity (an inner membrane marker) was measured spectrophotometrically (33). One unit of NADH oxidase activity is defined as 1 mol of NADH oxidized per min at 25°C.

Rapid Detection of Lipid A Molecular Species Released by pH 4.5 Hydrolysis of 32 P i -Labeled Membrane
Fractions-A small amount of membrane fraction (usually several l containing about 2,000 cpm) was diluted to 10 l so that the final solution contained 12.5 mM sodium acetate, pH 4.5, and 1% SDS (34 -36). This solution was incubated at 100°C for 30 min in a 0.6-ml plastic microcentrifuge tube. The hydrolyzed material was converted to a two-phase Bligh and Dyer mixture by addition of 22 l of chloroform/methanol (1:1, v/v). The entire lower phase was applied to a Silica Gel 60 TLC plate. The plate was developed in the solvent of chloroform/pyridine/88% formic acid/water (50:50:16:5, v/v). The plate was dried and exposed overnight to a PhosphorImager screen to visualize both the glycerophospholipids (which are not degraded under these conditions) and lipid A (which is selectively released from the inner Kdo of the core without being dephosphorylated or deacylated).
Mass spectrometry of lipid A, purified after release from cells by pH 4.5 hydrolysis (36), was used to validate the identification of the radioactive lipids bands observed by thin layer chromatography. In wild type cells two forms of lipid A are recovered. Both are hexa-acylated, but a more hydrophilic form (about 30% of the total) contains an extra phosphate moiety, forming a diphosphate monoester (in place of the predominant monophosphate) at the 1-position of the proximal glucosamine residue (Fig. 2).

RESULTS
MsbA on pKW2 Suppresses the Temperature-sensitive Growth Phenotype of a Mutant Lacking htrB-Twelve colonies were isolated from a population of MLK53/pKW2 transformants selected on agar plates at 44°C, and plasmid DNA was isolated from these strains. All of the colonies that grew at 44°C contained pKW2, as determined by gel electrophoresis of the plasmid digested with EcoRV and HindIII. pACYC184 alone was unable to suppress the temperature-sensitive phenotype of MLK53.
As shown in Fig. 3, htrB mutant cells (MLK53) stop growing after about 4 h in shaking culture when shifted from 30 to 42°C. Transformation of strain MLK53 with pKW2 also restored growth at 42°C on liquid LB medium (albeit somewhat slower than wild type). These findings show that the orfE/lpxK gene downstream of msbA (21) is not needed for the suppression of the htrB mutant phenotype.
MsbA Does Not Increase the Extent of Lipid A Acylation in htrB-deficient Mutants at 42°C-Lipid A of wild type cells is acylated mostly with six fatty acids irrespective of growth temperature, as judged by 32 P i labeling and recovery of the lipid A 4Ј-monophosphates following strong acid hydrolysis (Fig. 4, lanes 2 and 3) (19). The msbB mutation (not shown in Fig. 4) blocks the last acylation of lipid A biosynthesis, resulting in the accumulation of the penta-acylated species (19,23). Five acyl chains on lipid A are sufficient to support growth (19,23). However, the htrB mutant MLK53 grown at 42°C generates mainly tetra-acylated lipid A moieties (Fig. 4, lane 5). When grown at 30°C, htrB mutants do synthesize some pentaand hexa-acylated lipid A moieties (19) (Fig. 4, lane 4), presumably by employing alternative acylation pathways, such as palmitoleate incorporation by Ddg (37) or palmitate transfer from glycerophospholipids (38). These additional acylation pathways do not seem to operate efficiently at 42°C. As shown in Fig. 4 (lanes 6 and 7), introduction of pKW2 into htrB deficient cells does not increase the extent of acylation of lipid A either at 30 or 42°C, even though the mutant cells are now able to grow at 42°C (Fig. 3). These findings demonstrate that  (19). Under these hydrolysis conditions both the Kdo and the 1-phosphate are released, giving rise to lipid A 4Ј-monophosphates (19,29). This procedure also causes 10 -15% O-deacylation. Approximately 1000 cpm of isolated lipid A 32 P-labeled 4Ј-monophosphates was applied to each lane of a Silica Gel 60 thin layer chromatography plate. The plate was developed in the solvent chloroform/pyridine/88% formic acid/ water (55:50:16:5, v/v). The plate was then dried and exposed overnight to a PhosphorImager screen. The positions of the hexa-acylated (Hexa-), penta-acylated (Penta-), and tetra-acylated (Tetra-) lipid A 32 P-labeled 4Ј-monophosphates, as well as the origin, are indicated. Lane 1 shows the position of a tetra-acylated lipid A 32 P-labeled 4Ј-monophosphate standard that was generated by acid hydrolysis of 4Ј-32 P-labeled lipid IV A . Even lanes are lipid A 32 P-labeled 4Ј-monophosphates from cells grown at 30°C and odd lanes from cells grown at 42°C. The strains are as follows: lanes 2 and 3, W3110; lanes 4 and 5, htrB mutant MLK53; lanes 6 and 7, MLK53/pKW2(msbA ϩ ).
MsbA is not an alternative acyltransferase that can compensate for the absence of HtrB.

Rapid Detection of Lipid A Molecular Species in Separated Inner and Outer Membranes of Wild Type and Mutant
Strains-In order to compare the composition and subcellular localization of lipid A species in various mutant strains, inner and outer membranes were separated by isopycnic sucrose gradient centrifugation (30) from cells that had been grown in the presence of 32 P i , as described under "Experimental Procedures." Inner membranes were detected by assaying NADH oxidase, and outer membranes were located by determining the activity of phospholipase A in each fraction, as shown in Fig. 5 (upper panel) for the wild type strain W3110 grown at 42°C. Excellent separation was observed. The total amount of 32 P in each fraction was determined by liquid scintillation counting (Fig. 5, upper panel). Lipid A species were then released from the labeled membrane fragments by pH 4.5 hydrolysis (a milder procedure than the one used in the experiment of Fig. 4) that breaks the Kdo-lipid A linkage (Fig. 2) without causing 1-dephosphorylation or deacylation (34 -36). Membrane glycerophospholipids also are retained during pH 4.5 hydrolysis. As shown in Fig. 5 (lower panel), chloroform extraction of the pH 4.5 hydrolyzed membranes, followed by thin layer chromatography and PhosphorImager analysis, allowed simultaneous detection of the major glycerophospholipids (phosphatidylethanolamine and phosphatidylglycerol plus cardiolipin) and the two major hexa-acylated lipid A molecular species present in wild type cells. These two forms of hexa-acylated lipid A differ only by the presence of an additional phosphate residue. The more slowly migrating form corresponds to the trisphosphate species. The molecular weights of these lipid A species were confirmed by mass spectrometry (36). 2 Both the lipid A bisphosphate and the trisphosphate species (Fig. 5, BP and TP, respectively) were localized mainly in the outer membrane. The structures of the hexa-acylated lipid A bisphosphate and trisphosphate species are shown at the bottom of Fig. 2.
The glycerophospholipids (39) were identified by their migration with commercial standards. A higher percentage of phosphatidylethanolamine compared with phosphatidylglycerol plus cardiolipin was observed in the outer membrane compared with the inner membrane (Fig. 5), consistent with reported compositions (40). Similar results to those in Fig. 5 were obtained with other wild type strains irrespective of growth temperature (data not shown).

Localization of Penta-acylated Lipid A Species in the Outer Membranes of msbB Mutants Grown at 42°C-Mutant
MLK1067 (msbB Ϫ ), which is not temperature-sensitive, is defective in the enzyme that incorporates the sixth fatty acid present on lipid A, usually a myristate residue (Fig. 2) (19, 20). MLK1067 was labeled at 42°C for several generations. The membrane separation and the thin layer analysis of the hydrolyzed fractions are shown in Fig. 6. In MLK1067, penta-acylated lipid A species predominate as both the bis-and the trisphosphates (Fig. 6). Both forms of lipid A in MLK1067 are located mainly in the outer membranes, as judged by the phospholipase A profile (Fig. 6). Mass spectrometry of the lipid A isolated from msbB-deficient cells (23) reveals the absence of the myristate moiety normally present on the distal subunit of lipid A (Fig. 2).
Accumulation of Tetra-acylated Lipid A Species in the Inner Membranes of htrB Mutants Shifted to 42°C-MLK53 (htrB Ϫ ) cannot grow above 32.5°C (16,17). MLK53 was labeled with 32 P i either at 30 or at 42°C for 3 h. The membrane separations and the lipid profiles of the individual fractions are shown for cells grown at 30°C (Fig. 7) or shifted to 42°C (Fig. 8). In 30°C grown cells, the membranes were well resolved (Fig. 7). Most of the lipid A was present in the outer membranes, either as the bis-or trisphosphates of penta-and hexa-acylated species. As noted above, other enzymes, such as Ddg (37) or the one that transfers palmitate from glycerophospholipids to lipid A (38), may substitute for the absence of HtrB at 30°C, allowing for the generation of more highly acylated lipid A molecules. At 42°C, however, almost all the newly synthesized lipid A (and most of the glycerophospholipids) accumulate in the inner membrane (Fig. 8). The membranes are not as well resolved as those isolated from 30°C grown cells (Fig. 7), but they are nevertheless discernible (Fig. 8). There also appears to be a relative increase in the amount of glycerophospholipids (Fig. 8), consistent with the previous observation that the phospholipid to protein ratio doubles in htrB mutants at non-permissive temperatures (18) kind reaches pre-existing outer membranes at 42°C (as judged by the phospholipase A profile in Fig. 8). Predominantly tetraacylated lipid A bisphosphate (but very little trisphosphate) was found in association with the inner membranes, which are more heterogeneous in buoyant density than wild type. None of the accumulated tetra-acylated lipid A species could be extracted directly from the inner membranes of htrB mutants using chloroform/methanol mixtures without prior pH 4.5 hydrolysis (data not shown). These findings indicate that these abnormal lipid A species accumulating at 42°C probably contain a substantial core oligosaccharide, given that tetra-acylated precursors lacking a complete core, such as Kdo 2 -lipid IV A , are chloroform/methanol-soluble without prior pH 4.5 hydrolysis (41). Accumulation of underacylated lipid A-core molecules in the inner membrane may be toxic and account for the inability of htrB mutant bacteria to grow at 42°C.
Effects of Extra Copies of msbA on the Membrane Composition of htrB Mutants-The msbA gene was discovered as a multi-copy suppressor of htrB mutants (21). In order to understand how msbA rescues htrB mutants at 42°C, the htrB mutant MLK53 was transformed with the msbA bearing plas-mid (pKW2) to generate strain MLK53/pKW2. MLK53/pKW2 cells were then grown and labeled for 3 h with 32 P i at 42°C. The membrane separations and the radioactive lipid profiles are shown in Fig. 9. The lipid A species present in the MLK53/ pKW2 membranes are similar to those found in MLK53 grown at 42°C (Fig. 8), but a significant fraction of the tetra-acylated lipid A species are recovered with the outer membranes. The formation of a tetra-acylated trisphosphate is also observed in MLK53/pKW2 in conjunction with transport to the outer membrane. A small amount of residual penta-acylated lipid A is also observed, as in Fig. 4. These observations support the view that MsbA is a lipid A transporter rather than functioning as a distinct lipid A acyltransferase.

Accumulation of Lipid A Species in the Inner Membranes of htrB ϩ Cells When msbA and orfE Are Not Expressed-Strain
AP191 contains an insertion in the chromosomal msbA gene, which also inactivates the downstream orfE/lpxK gene (21). In addition, msbA ϩ and lpxK ϩ are present on a plasmid with a temperature-sensitive origin of replication (24). After several hours at 42°C, LPS accumulates in the inner membranes of such constructs, as judged by N-acetyl[ 3 H]glucosamine labeling (24). To confirm these findings with our 32 P labeling technique, membrane separations and lipid A analyses were performed on both AP191 (as described in Fig. 10) and on a matched strain, designated AP70, that bears a plasmid with a normal origin of replication (24) (data not shown). The membrane separation and the 32 P-lipid profiles of the fractions derived from cells of AP70 grown at 42°C (data not shown) were very similar to those of W3110 (Fig. 5). AP191 was first shifted to 42°C for 5 h to dilute out the msbA/orfE bearing plasmid, and then the cells were labeled with 32 P i for 3 h (Fig. 10). Under these conditions almost all of the newly made radioactive lipids were located in the inner membranes (Fig. 10). Two lipid A-related compounds, not seen in the inner membranes of wild type cells (Fig. 5), also accumulated in the inner membranes of AP191 at 42°C (Fig.  10). One of these could not be extracted with chloroform/methanol mixtures without prior pH 4.5 hydrolysis and was identified as hexa-acylated lipid A bisphosphate (Fig. 10). The other lipid that accumulated was slightly more hydrophobic but could be extracted directly with chloroform/methanol mixtures. As shown in Fig. 10, this material had the same apparent mobility as a disaccharide 1-phosphate (DS-1-P) standard, the intermediate of the lipid A pathway that is the substrate for the 4Ј-kinase (Fig. 2) (26,42,43). Since orfE/lpxK has recently been shown to encode the 4Ј-kinase (26), the accumulation of DS-1-32 P in the inner membranes of AP191 at 42°C can be explained by the loss of orfE/lpxK.

Absence of MsbA Alone Causes Lipid A Accumulation in
Inner Membranes-To exclude the possibility that the inner membrane build up of hexa-acylated lipid A bisphosphate observed in AP191 at 42°C (Fig. 10) might represent an indirect effect of DS-1-P accumulation on lipid A transport, instead of being due to the loss of MsbA per se, AP191 was transformed with the orfE bearing hybrid plasmid, pEH2. The latter is compatible with the temperature-sensitive hybrid plasmid (bearing msbA and orfE) present in AP191. Furthermore, pEH2 is maintained and can express the 4Ј-kinase irrespective of the temperature at which the cells are grown. AP191/pEH2 displays the same temperature-sensitive growth characteristics as does AP191 (data not shown). Cells of AP191/pEH2 were therefore labeled in the same way as described for AP191 in Fig. 10, and the membranes were separated and analyzed for lipid A molecular species (Fig. 11). As observed with AP191, most of the lipid radioactivity of AP191/pEH2 present after 3 h labeling at 42°C was located in the inner membranes, but now only the hexa-acylated lipid A bisphosphate accumulated (Fig.  11). DS-1-P was not detectable. The results indicate that the absence of MsbA alone can account for the accumulation of hexa-acylated lipid A species within the inner membrane and that the effects of orfE inactivation in AP191 (Fig. 10) are peripheral. As in htrB mutants shifted to 42°C, no lipid A trisphosphates are observed when hexa-acylated lipid A fails to reach the outer membrane in AP191/pEH2 (Fig. 11). Taken together, these findings lend additional support to the idea that MsbA may function as a transporter of newly synthesized lipid A-core molecules in E. coli (24,25). DISCUSSION The msbA gene was discovered because multiple copies of msbA suppress the temperature-sensitive growth of htrB mutants (Fig. 2) (21). The connection to the LPS system was made in 1995 with the recognition that htrB is the structural gene for the lauroyltransferase of the lipid A pathway (Fig. 2) (15, 16). The fact that MsbA is highly homologous to a large family of ABC transport proteins and is required for the growth of E. coli raised the intriguing possibility that MsbA might play a role in LPS secretion, a process about which little is known (Fig. 1) (21).
In the present work, we have made the following new observations. (a) Tetra-acylated lipid A species accumulate in the inner membranes of htrB mutants at 42°C, but htrB mutants grown at 30°C are able to make penta-and hexa-acylated lipid A molecules, presumably using alternative acyltransferases. (b) msbA does not increase the extent of lipid A acylation in htrB mutants. (c) Extra copies of msbA introduced into htrB mutants stimulate the transport of a large fraction of the tetra-acylated lipid A species to the outer membrane at 42°C. (d) Hexa-acylated lipid A species accumulate in the inner membranes of cells depleted of MsbA, even when LpxK depletion is prevented by the presence of an additional lpxK bearing plasmid. (e) Glycerophospholipids accumulate together with lipid A in the inner membranes of both htrB and msbA mutant cells at 42°C, suggesting that MsbA might also play a role in glycerophospholipid transport. (f) When lipid A fails to reach the outer membrane, the trisphosphate form of lipid A is not observed.
LPS had recently been shown to accumulate in the inner membranes of mutants with disrupted htrB and msbA genes, as judged by gel electrophoresis, pulse labeling with N-acetyl[ 3 H]glucosamine (24), or antibody binding (25). However, no unambiguous way to quantify the amount of lipid A and to determine the extent of its acylation in isolated membrane vesicles had been described. To address this problem, we devised a rapid, new procedure for detecting 32 P-labeled lipid A residues in small biological samples. Our assay is based on the observation that the Kdo-lipid A linkage (Fig.  2) is cleaved selectively by pH 4.5 hydrolysis at 100°C in the presence of SDS without measurable loss of ester-linked fatty acids or of the anomeric phosphate (34)(35)(36). The hydrolysis procedure can be applied directly to radioactive cells or to isolated membrane fractions from a sucrose gradient without first extracting LPS with phenol-containing solvents. The lipid A released in this manner is recovered by Bligh-Dyer extraction together with the glycerophospholipids, which also remain intact throughout the procedure. The lipid A and the glycerophospholipids can then be separated and quantified simultaneously by TLC analysis (Fig. 5, lower panel). The identification of the hexa-acylated lipid A bis-and trisphosphate spots (Fig. 5, BP-hexa and TP-hexa, respectively) was validated by mass spectrometry (36), in conjunction with 1 H, 31 P, and 13 C NMR spectroscopy. 3 In several wild type E. coli strains grown at 30 or 42°C (not shown), the lipid A is largely hexa-acylated and is mainly recovered in the outer membrane fractions (Fig. 5). The small amount of lipid A found in the inner membranes (Fig. 5) may represent newly synthesized core-lipid A, given that the cells were labeled with 32 P i for several generations.
Previous genetic studies showed that both msbA and orfE/ lpxK are essential E. coli genes (21). Consistent with previous observations (21), the experiment of Fig. 3 shows that the phenotypic suppression of the htrB mutation can be attributed solely to msbA. Interestingly, the growth rate of the suppressed mutant is slower than wild type. Slow growth may be explained by the finding that msbA does not correct the lipid A acylation deficiency associated with the htrB mutation (Fig. 4). The tetra-acylated lipid A precursors that accumulate at 42°C in htrB mutants may not be rapidly transported, given their accumulation in the inner membrane (Fig. 8). The tetra-acylated lipid A species that accumulate in the inner membranes of htrB mutants at 42°C probably contain a complete core, as pH 4.5 hydrolysis is required to recover these species by Bligh-Dyer extraction (data not shown). Overexpression of msbA results in the translocation of a significant fraction of the tetra-acylated lipid A species to the outer membrane (Fig. 9). These findings support the view that MsbA acts by stimulating lipid A transport and that it is not likely to be an acyltransferase.
The htrB mutant MLK53 grown at 30°C synthesizes large amounts of penta-and hexa-acylated lipid A moieties (19) (Figs. 4 and 7), which are efficiently translocated to the outer membrane. The origin of these acylated species is not entirely clear, given that the htrB gene in MLK53 is disrupted by an insertion (17). Some of this acylation may be catalyzed by Ddg (a cold-shock induced palmitoleoyltransferase with homology to HtrB that appears to acylate the same OH group on Kdo 2 -lipid IV A normally employed by HtrB) (37). Ddg-catalyzed acylation could be followed by the action of MsbB (Fig. 2) to generate a hexa-acylated lipid A moiety. Alternatively, the enzyme that transfers palmitate residues from glycerophospholipids to lipid A precursors (38) (the structural gene of which is unknown) may be activated in htrB mutants. Mass spectrometry of lipid A isolated from MLK53 grown at 30°C (data not shown) suggests that palmitoleate and myristate attachment account for most of the observed penta-and hexa-acylated lipid A species (Figs. 4 and 7). Mass spectrometry of lipid A isolated from Salmonella typhimurium strains lacking htrB likewise suggests the activation of compensatory lipid A acylation reactions when such mutants are grown at 30°C (44). Mutants lacking both htrB and these other acyltransferases may therefore be unable to grow at any temperature in the absence of a covering plasmid carrying msbA.
Five acyl chains on lipid A seem to be sufficient to support the growth of E. coli cells containing only a single chromosomal copy of msbA (19,23). The growth rate of strains defective in msbB (Fig. 2) is nearly normal at 30 and 42°C, despite the fact that their lipid A consists almost entirely of penta-acylated species. The latter, like the penta-acylated lipid A moieties seen in htrB mutants grown at 30°C (Fig. 7), are translocated efficiently to the outer membrane (Fig. 6). Taken together, these findings indicate that one acyloxyacyl group on lipid A may be needed for proper recognition by the secretion machinery.
Hexa-acylated lipid A species accumulate in the inner membranes of strains depleted of MsbA, which cannot grow because MsbA is essential (Figs. 10 and 11). In these experiments, the depletion of MsbA is achieved by shifting a strain containing a chromosomal insertion mutation in msbA and a temperaturesensitive covering plasmid bearing msbA ϩ to 42°C. MsbA depletion takes 3-5 h in the available constructs. Growth inhibition is very gradual (Fig. 12). The cells nevertheless take up reasonable amounts of 32 P i from the medium during the 3-h labeling period after MsbA is diluted out, and they incorporate it into all major membrane lipids (as judged by the total counts recovered in Figs. 10 or 11 versus Fig. 5). Efficient labeling under these conditions argues that the effects of MsbA depletion are not indirect. The lack of the msbA covering plasmid during the period of 32 P i labeling in the experiments of Figs. 10 and 11 was confirmed by plating the cells in the presence of 50 g/ml spectinomycin and showing that the number of colonies was reduced by 4 orders of magnitude compared with what was expected from the optical density of the culture.
A problem with the published MsbA depletion experiments, as well as the experiment in Fig. 10, is that msbA is an upstream gene in an operon (24,25). The essential downstream gene (orfE/lpxK), now known to encode the 4Ј-kinase of lipid A biosynthesis (Fig. 2) (26), is inactivated together with msbA. Expression of extra lpxK genes on a separate plasmid in MsbAdepleted cells does not prevent lipid A accumulation in the inner membranes or gradual growth arrest (Figs. 11 and 12). Thus, lipid A accumulation in the inner membrane can be attributed solely to MsbA depletion (Fig. 11). The only difference is that the 4Ј-kinase substrate (DS-1-P) no longer accumulates when LpxK depletion is prevented (Fig. 11 versus Fig.  10). The identification of DS-1-P in Fig. 10 was confirmed by the finding that pH 4.5 hydrolysis was not required for efficient Bligh-Dyer extraction (data not shown). Recently, we have demonstrated that selective lpxK inactivation also results in growth arrest and DS-1-P accumulation (53). The effects of selective LpxK depletion on lipid A transport have not yet been examined. Therefore, we cannot exclude the possibility that MsbA and LpxK are both components of the putative core-lipid A transporter (Fig. 1).
The biosynthetic origin of the trisphosphate species of lipid A (45) (Figs. 5, 6, 7 and 9) has not been established. The trisphos-3 Z. Zhou and C. R. H. Raetz, manuscript in preparation.
FIG. 12. Temperature-sensitive growth of strains depleted of msbA. Strains were grown as described in Fig. 3, except that the antibiotic spectinomycin was left out of the LB broth when the temperature was shifted to 42°C. It took about 5 h to dilute out the covering plasmid that contains both msbA and orfE/lpxK (24). phate is not observed when lipid A transport is blocked (Figs. 8,  10, and 11). This interesting finding suggests that the trisphosphate is generated on the periplasmic surface of the inner membrane or in the outer membrane. Since ATP would not be available in the outer parts of the envelope, other donors of high energy phosphate moieties must be considered, such as bactoprenol-pyrophosphate, a product of O-antigen polymerization ( Fig. 1) (4), or diacylglycerol-pyrophosphate, a newly discovered, minor lipid (46). Attempts to generate trisphosphate in vitro using ATP and Kdo 2 -lipid IV A have been unsuccessful. 4 Our assays (Figs. 8, 10, and 11) indicate that glycerophospholipids accumulate together with lipid A in the inner membrane when lipid A transport is blocked in htrB-or msbAdeficient strains. One provocative explanation is that MsbA transports both lipid A and glycerophospholipids. This idea is consistent with several of the following observations. (a) The msbA gene (21) is a member of a large subset of ABC transporters of largely unknown function found in all eubacteria, archaebacteria, and eucaryotic cells. Multiple copies of msbAlike genes (coding for proteins of ϳ590 amino acids) are present in all genomes, as judged by BLAST searching (47,48). Since lipid A is found only in Gram-negative bacteria (2, 3), the wide distribution of msbA genes in all organisms suggests a more general function for MsbA. (b) The high sequence homology of E. coli msbA to the two homologous domains of mammalian Mdr proteins is especially striking (about 10 Ϫ84 and 10 Ϫ72 using the Gapped-Blast algorithm) (48). This score is comparable to the similarity of E. coli msbA to msbA genes found in other Gram-negative bacteria (48). (c) The function of Mdr-2 in the mouse has recently been determined by construction of a strain in which the mdr-2 gene is knocked out (49,50). This mouse displays a 10-fold reduction in the secretion of phosphatidylcholine into its bile, consistent with the high level of expression of Mdr-2 in the cells lining the biliary tree (49,50). This finding shows that Mdr proteins are capable of transporting specific glycerophospholipids in a physiologically relevant setting (49,50). The multiplicity of msbA-like genes in nature may therefore reflect the requirement for protein-mediated transport (i.e. flip-flop) of diverse lipid molecules across biological membranes.
To prove unequivocally that MsbA functions as a general membrane lipid transporter in E. coli, additional mutants defective in msbA need to be isolated, and more informative biochemical assays must be devised. For instance, a mutant allele of msbA that results in the complete loss of MsbA activity within minutes after a shift to 42°C would be very helpful, since such a strain should accumulate newly made lipid A and glycerophospholipids within its inner membranes shortly after a temperature shift. The time course of inactivation of other secretory events and of macromolecular synthesis could then be assessed and should help establish what effects are primary and secondary. Lastly, MsbA needs to be purified to homogeneity and reconstituted into model membrane vesicles in order to demonstrate conclusively its ability to catalyze ATP-dependent lipid A and glycerophospholipid flip-flop.