Depletion of Apolipoprotein N-Acyltransferase Causes Mislocalization of Outer Membrane Lipoproteins in Escherichia coli*

Lipoproteins in Gram-negative Enterobacteriaceae carry three fatty acids on the N-terminal cysteine residue, two as a diacylglyceride and one through an N-linkage following signal peptide cleavage. Most lipoproteins are anchored in the outer membrane, facing the periplasm, but some lipoproteins remain in the plasma membrane, depending on the amino acid at position +2, immediately after the fatty-acylated cysteine. In vitro, the last step in lipoprotein maturation, N-acylation of apolipoproteins by the plasma membrane apolipoprotein N-acyltransferase (Lnt), is necessary for efficient recognition of outer membrane lipoproteins by the Lol system, which transports them from the plasma to the outer membrane (Fukuda, A., Matsuyama, S.-I., Hara, T., Nakayama, J., Nagasawa, H., and Tokuda, H. (2002) J. Biol. Chem. 277, 43512–43518). To study the role of Lnt in vivo, we constructed a conditional lnt mutant of Escherichia coli. The apo-form of peptidoglycan-anchored major lipoprotein (Lpp) and two other outer membrane lipoproteins accumulated in the plasma membrane when lnt expression was reduced. We also found that Lnt is an essential protein in E. coli and that the lethality is partially because of the retention of apoLpp in the plasma membrane. Topology mapping of Lnt with β-galactosidase and alkaline phosphatase fusions indicated the presence of six membrane-spanning segments. The lnt gene in a mutant of Salmonella enterica displaying thermosensitive Lnt activity (Gupta, S. D., Gan, K., Schmid, M. B., and Wu, H. C. (1993) J. Biol. Chem. 268, 16551–16556) was found to carry a mutation causing a single glutamate to lysine substitution at a highly conserved position in the last predicted periplasmic loop of the protein.

The Escherichia coli K12 genome encodes almost a hundred putative lipoproteins (1), a unique class of exported proteins, most of which are anchored in the inner leaflet of the outer membrane by their N-terminal fatty acids (2). The best characterized lipoprotein, Lpp, is present in two forms in the outer membrane. Approximately one-third of the Lpp molecules in the cell are cross-linked to the peptidoglycan via the C-terminal lysine residue (3), thereby contributing to outer membrane integrity (4). The remainder of the Lpp exists as a free (unbound) form. Lpp has been extensively used to study bacterial lipoprotein biogenesis and sorting to the outer membrane. Three fatty acids are bound to its N-terminal cysteine residue, two in a diacylglyceride that is linked via a thioether bond to the sulfhydryl group and one to the amine group that is liberated upon signal peptide cleavage (5,6). The fully matured lipoprotein is then transferred from the plasma membrane to the periplasmic lipoprotein chaperone LolA (7) by the ABC transporter LolCDE (8), and then delivered to the outer membrane docking protein LolB (9), whereupon it inserts into the inner lipid leaflet of the outer membrane. It seems reasonable to assume that most if not all E. coli outer membrane lipoproteins follow exactly the same route.
A relatively small number of lipoproteins remain in the E. coli plasma membrane. In the two characterized plasma membrane lipoproteins, the endogenous protein NlpA (10) and the Klebsiella oxytoca amylolytic enzyme pullulanase (11), retention in the plasma membrane requires the aspartate residue at position ϩ2, immediately after the fatty acylated cysteine. Furthermore, the introduction of an Asp ϩ2 residue into outer membrane lipoproteins or its presence in artificial lipoproteins (formed by fusing the signal peptide and first few amino acids of a lipoprotein to a reporter protein) causes their retention in the plasma membrane (12,13). Other residues except proline and aromatic amino acids cause lipoproteins to be routed to the outer membrane (12,13). The ability of Asp ϩ2 to cause efficient plasma membrane retention may be influenced both by amino acids in the adjacent sequence (14,15) and by the structure of the polypeptide of which it is part (16).
Asp ϩ2 lipoproteins differ from outer membrane lipoproteins in being unable to activate the LolCDE ATPase in proteoliposomes, suggesting that Asp ϩ2 functions as an Lol avoidance signal (8,17,18). Other details of the mechanism by which lipoproteins are retained in the plasma membrane are unclear. The unique physicochemical properties of aspartate seem to be important for this function (19), which makes it all the more surprising that the structurally unrelated aromatic amino acids and proline at position ϩ2 in an artificial lipoprotein can also function as efficient plasma membrane retention signals (13). Furthermore, it has not been established whether Asp ϩ2 functions as an Lol avoidance signal in other species of bacteria.
The plasma membrane enzyme that carries out the third and final step in lipoprotein processing and acylation (apolipoprotein N-acyltransferase or Lnt) was first identified in E. coli (20). A gene (lnt) encoding this enzyme was subsequently identified by the same group in Salmonella enterica sv. Typhimurium through studies of a temperature-sensitive mutant in which apoLpp (lacking the N-acyl group) accumulated at the nonpermissive temperature (21). The lnt gene E. coli (initially called cutE and referred to here as lnt Ec to reflect its known function and to distinguish it from the S. enterica gene, lnt Se ) was independently identified in studies of a copper-sensitive mutant (22).
Sequence alignments reveal that homologues of Lnt are present in Gram-negative bacteria but apparently absent from most Gram-positive bacteria, including Bacillus subtilis and Staphylococcus aureus (23,24). This observation led to the proposal that lipoproteins produced by Gram-positive bacteria might be incompletely fatty-acylated and, since all of them are retained in the plasma membrane, that N-acylation might be a characteristic of Gram-negative bacterial outer membrane lipoproteins (13,23). In apparent agreement with this idea, LolCDE in proteoliposomes cannot promote the capture by LolA of the "apo-form" of another major E. coli outer membrane lipoprotein, Pal, suggesting that N-acylation is required for efficient recognition of nascent outer membrane lipoproteins by the Lol system (25). However, mass spectrometry demonstrated that plasma membrane-anchored Lpp with an Asp ϩ2 residue is fully N-acylated, indicating that this retention signal (or Lol avoidance signal) does not operate by preventing Nacylation (25). Furthermore, analysis of E. coli Lpp produced in B. subtilis (26) and of BlaZ ␤-lactamase in S. aureus (24) suggested that at least some polypeptides were N-acylated, indicating that these Gram-positive bacteria might possess an enzyme with a similar activity to Lnt.
In apparent contrast to the aforementioned retention of apo-Pal in proteoliposomes in the presence of LolA, studies with S. enterica carrying a temperature-sensitive (ts) mutation in lnt (referred to here as lnt ts ) revealed that apoLpp produced at the nonpermissive temperature (42°C) was covalently bound to the peptidoglycan and could not be extracted from the membranes with the detergent Sarkosyl, suggesting that it was localized to the outer membrane (21). This interpretation has been called into question (25), because it is known that Lpp that is artificially retained in the plasma membrane by introduction of Asp ϩ2 is also cross-linked to the peptidoglycan, leading to cell death (27).
This report documents the effects of Lnt Ec depletion in E. coli on lipoprotein localization and membrane organization. We constructed a conditional lnt Ec mutant in which the chromosomal lnt Ec gene is exclusively expressed from the tightly regulated arabinose-inducible araB promoter. This strategy avoids secondary effects introduced by the growth at high temperature that is necessary to inactivate the thermosensitive form of S. enterica Lnt (21). The question of whether apoLpp is retained in the plasma membrane in vivo was then addressed by using a variant of this protein that cannot be cross-linked to the peptidoglycan (27).

MATERIALS AND METHODS
Strains, Plasmids, and Construction of p ara -lnt Ec Cassette-Strains of E. coli and S. enterica sv. Typhimurium used in this study are listed in Table I. Strain PAP8504 was constructed by homologous recombination according to the method developed by Datsenko and Wanner (28).
The kan-rpoCter-p araB cassette was constructed by successive ligations between DNA fragments encoding the rpoCter gene, the kan gene, or the araB promoter. First, the terminator rpoCter was excised from pOM90 (29) using restriction enzymes EcoRI and KpnI and cloned in the plasmid pGP704 (30), at the same sites, giving pCHAP6560. Second, the gene encoding kanamycin phosphotransferase (kan) was PCR-amplified using primers 5Ј-kan and kan-3Ј (Table III) from pBGS18 (31). The fragment was cloned into EcoRI and NotI sites upstream from the rpoCter gene in pCHAP6560 to create pCHAP6561. Finally, the araB promoter was PCR-amplified using primers 5Ј-para and para-3Ј (Table  III) from pBAD33 (32), and the fragment was cloned into XbaI and SalI sites downstream rpoCter gene in pCHAP6561 to create pCHAP6563. E. coli strain BW25113 carrying pKD46 was electroporated with 50 ng of the kan-rpoCter-p araB fragment PCR-amplified from pCHAP6563 using primers 5Ј-ybe-k and p-lnt Ec -3Ј (Table III). These long primers include 45 nucleotides that hybridize with the ybeX or lnt Ec genes and 19 or 25 priming nucleotides. PAP8504 was obtained after selection of transformants on agar containing kanamycin (25 g/ml) and 0.2% arabinose and was then incubated at 37°C to eliminate the temperaturesensitive pKD46. The presence of the cassette in PAP8504 was verified by PCR amplification using primers 5Ј-ybe-k and p-lnt Ec -3Ј and by transduction into E. coli strain PAP105 using phage P1 (33). Transductants were selected on agar containing kanamycin (25 g/ml) and 0.2% arabinose.
Strain PAP8505 was obtained by transduction of lpp::Tn10 from an E. coli strain carrying this mutation (provided by S. Gupta) using phage P1 and selection for resistance to tetracycline (16 g/ml).
Plasmids used in this study are listed in Table II. pCHAP6571, encoding Lnt Ec , was constructed by ligating a DNA fragment, obtained by PCR amplification from DNA of strain MC4100 using primers 5Ј-cutE and cutE-3Ј (Table III), into the EcoRI and BamHI sites in pUC18. The same procedure was employed to construct pCHAP6573 and pC-HAP6574, encoding Lnt Se and Lnt Se (E435K), respectively, by using primers 5Ј-lnt and lnt-3Ј (Table III) and DNA from strains LT2 or SE5312. pCHAP6576 was obtained by site-directed mutagenesis of pCHAP6571 with the primer 5Ј-cutE435 (Table III). Oligonucleotidedirected mutagenesis was performed using a Quickchange site-directed mutagenesis kit (Stratagene).
Plasmids encoding fatty-acylated variants of MalE (Table II) were derived from plasmids used in a previous study (13). Briefly, PCRs using appropriate oligonucleotide primers were used to change the second amino acid of the mature part of the protein referred to as CD-MalE in Ref. 13 to Ala, Phe, Pro, Trp, and Tyr using pCHAP511 as the template (34). A hexahistine tag was added to the end of each of the six lipoMalE constructs by replacing the 3Ј BglII-HindIII fragment at the 3Ј end of the malE gene by the corresponding segment of the malE::His6 gene carried by pMalE-His (provided by A. Davidson). The entire fragment was then subcloned as an EcoRI-HindIII fragment into pSU18 (35) or pBGS18 ϩ (31). Details of other plasmids listed in Table  II are given below.
For growth analysis and preparation of cell extracts, PAP8504 and PAP8505 were grown overnight in LB medium with 0.2% arabinose and washed in LB medium before being diluted 1:100 into LB medium with 0.2% arabinose or 0.4% glucose. Cells were grown at 37°C with agitation to A 600 0.8 and then re-diluted 1:10 of fresh medium. Immunoprecipitation of apoLpp-For radiolabeling, cells were grown exponentially in LB medium as above, washed, and resuspended in minimal medium supplemented with 0.4% glucose or 0.5% glycerol (when cells were grown in medium containing arabinose) for 15 min at 30°C. Proteins from 1 ml of each culture were labeled with 50 Ci of    Construction and Analysis of lnt Ec -lacZ and lnt Ec -phoA Gene Fusions-The phoA gene encoding alkaline phosphatase, PhoA, lacking the region coding for the signal peptide (from amino acid ϩ13) was PCR-amplified from plasmid pCHAP4020, 2 using primers 5Ј-phoA and phoA-3Ј (Table III), and cloned into PstI and HindIII sites in pBAD33 to obtain pCHAP6578. This plasmid was used to construct 9 plasmids, pCHAP6580 to pCHAP6588 (Table II), encoding Lnt Ec (30)-PhoA to Lnt Ec (512)-PhoA (numbers in parentheses correspond to the amino acid fusion site in Lnt Ec ). Each DNA fragment, obtained by PCR amplification with primers 5Ј-lnt Ec and lnt Ec 30 -3Ј (or lnt Ec 53-3Ј, lnt Ec 80 -3Ј, lnt Ec 117-3Ј, lnt Ec 154 -3Ј, lnt Ec 190 -3Ј, lnt Ec 218 -3Ј, lnt Ec 476 -3Ј, and lnt Ec 513-3Ј (Table III)), was inserted into the XbaI and PstI sites in pCHAP6578. The same procedure was employed to construct the 9 plasmids, pCHAP6589 to pCHAP6597, encoding Lnt Ec (30)-LacZ to Lnt Ec (512)-LacZ. First, the lacZ gene encoding ␤-galactosidase, LacZ (from amino acid ϩ9), was PCR-amplified from plasmid pRS552 (37) using primers 5Ј-lacZ and lacZ-3Ј (Table III) and cloned into the PstI and HindIII sites in pBAD33 to obtain pCHAP6577. An lnt Ec DNA fragment obtained by PCR amplification with primers 5Ј-lnt Ec and one of the nine 3Ј primers lnt Ec 30 -3Ј to lnt Ec 512-3Ј (Table III) was inserted into the XbaI and PstI sites in pCHAP6577.
Derivatives of strain KS272 producing the Lnt Ec -PhoA and Lnt Ec -LacZ chimeras or carrying pCHAP6577 or pCHAP6578 as controls were grown in LB medium at 30°C supplemented with chloramphenicol and arabinose. To measure alkaline phosphatase activity, cells were diluted in 1 ml of 50 mM Tris-HCl (pH 9.0) and then permeabilized by adding 50 l of 10% octylpolyoxyethylene. After incubation at 37°C for 5 min, the reaction was started by adding 100 l of para-nitrophenyl phosphate (10 mg/ml) and stopped with 500 l of 1 M NaOH. To measure ␤-galactosidase activity, the cells were diluted in Z-buffer (33) and permeabilized with octylpolyoxyethylene as above. The reaction was started by adding 200 l of orthonitrophenyl ␤-D-galactopyranoside (4 mg/ml) and stopped by adding 500 l of 1 M Na 2 CO 3 . Enzyme activities were calculated according to Miller (33) and are given in arbitrary units.
SDS-PAGE and Immunoblotting-Proteins solubilized in loading buffer were heated at 100°C for 5 min and separated by SDS-PAGE in gels containing 9, 10, or 12% acrylamide and, in some cases, 8 M urea to improve separation of apolipoproteins. Proteins were detected by staining with Coomassie Blue or after transfer onto nitrocellulose membranes and incubation with primary polyclonal antisera to Lpp, SecG (provided by W. Wickner), Pal (provided by E. Bouveret), NlpD (provided by S. Clarke), FhuA (provided by M. Bonhivers), ␤-galactosidase (Cappel), or alkaline phosphatase, and then by incubation with horseradish peroxidase-conjugated secondary antiserum (Amersham Biosciences). His-tagged proteins were immunodetected using INDIA HisProbe-horseradish peroxidase (Pierce). Bound horseradish peroxidase-labeled antibodies were detected by enhanced chemiluminescence.
Separation of Plasma and Outer Membranes by Sucrose Flotation Gradient-One hundred ml of cultures at an A 600 of 0.8 -1.0 were collected by centrifugation, and the pellet was resuspended in 10 ml of 25 mM HEPES (pH 7.4). Cells were disrupted by two passages through a French press (1,200 bar), and the lysate was supplemented with 10 g/ml each of DNase I and pancreatic RNase A and then centrifuged for Edman Degradation of Lipoproteins-To purify the histidine-tagged lipoMalE proteins for sequencing, bacteria from 100 ml of saturated LB broth culture containing 1 mM IPTG (to induce the expression of the malE gene, which is under p lacZ control) were disrupted in a French pressure cell, and the cell envelope was collected by centrifugation at 160,000 ϫ g for 60 min. Membrane proteins were dissolved in 2% 3-(N,N-dimethylmyristoylammonio)propane sulfonate (Fluka) in 50 mM Tris-HCl (pH 8.0) and then adsorbed onto Talon cobalt affinity resin (Clontech) from which the lipoMalE proteins were eluted with 200 mM imidazole (note that MalE cannot be purified on amylose resin in the presence of detergents). Purified proteins were precipitated with 10% trichloroacetic acid and then separated by SDS-PAGE on 10% acrylamide gels and electroblotted onto Immobilon-P SQ polyvinylidene difluoride membranes (Millipore) for automated Edman sequencing on an Applied Biosystems 477A gas phase Sequenator.

RESULTS
Lnt Ec Is Essential in E. coli-To study the consequences of Lnt Ec depletion in E. coli, a conditional mutant was constructed in which the chromosomal lnt Ec gene (originally called cutE (22), see Introduction) is expressed only from the arabinoseinducible AraC-dependent promoter p araB . The lnt Ec gene is located in an operon downstream from the ybeX gene, whose function is unknown. The 25 nucleotides between ybeX and lnt Ec were replaced by a cassette containing a selectable kanamycin resistance gene (kan), the rpoC transcription terminator (rpoCter; to prevent transcription read-though from p ybeX ), and p araB (Fig. 1; see "Materials and Methods") by homologous recombination. The resulting p araB -lnt Ec strain (PAP8504) consistently ceased growth and began to lyse after 8 generations of growth in LB medium without arabinose and with glucose (to repress p araB expression) and did not form colonies on minimal glucose or LB agar (data not shown). The integration of the cassette between ybeX and lnt Ec was first verified by PCR amplification using primers flanking the entire cassette and then by transduction into the E. coli strain PAP105 using phage P1 with selection for the kanamycin resistance cassette (see "Materials and Methods"). Donor and transductants exhibited the same arabinose dependence and were complemented by a cloned copy of lnt Ec expressed from p lacZ in a pUC18 derivative (pCHAP6571; see below), demonstrating that Lnt Ec 2 O. Francetic, unpublished observations.

FIG. 1. Construction of the p araBlnt Ec strain.
The kan-rpoCter-p araB cassette was generated by ligation and overlapping PCR amplification of the individual components followed by amplification of the complete cassette (see "Materials and Methods"), which was electroporated into strain BW25113 carrying the plasmid pKD46 (28). Transformants were selected on LB agar containing kanamycin and arabinose. Arrows below the genes indicate the orientation of the transcription units.
is essential for the viability of E. coli, as shown previously (21) in S. enterica. Although the p araB -lnt Ec was integrated into the chromosome, the amount of Lnt Ec produced is probably higher than the levels achieved with the wild-type strain, but this does not affect its use in the experiments described below.
Immunoblotting with antibodies against the major outer membrane lipoprotein Lpp revealed the gradual appearance of the apolipoprotein form (apoLpp) when p araB -lnt Ec was repressed by glucose in LB cultures ( Fig. 2A), confirming that the Lnt Ec activity declines under these conditions. ApoPal could not be detected by immunoblotting (data not shown; see "Discussion"). Immunoprecipitation of [ 35 S]methionine-labeled proteins with anti-Lpp revealed almost complete absence of Nacylation as the arabinose-deprived cultures approached the time when they lyse (Fig. 2B).
Lnt Ec Depletion Induces Juxtapositioning of Plasma and Outer Membranes-To analyze the localization of apolipoproteins, membranes from cells of strain PAP8504 grown in LB containing arabinose or glucose were separated by sucrose gradient centrifugation. Lpp and Pal in membranes from arabinose-grown cells were detected almost exclusively in dense fractions containing the outer membrane porins, well separated from less dense fractions containing SecG, an integral plasma membrane protein. In contrast, distinct outer and plasma membrane peaks were not detected when membranes from Lnt Ec -depleted cells (6.5 generations after repression of p araB -lnt Ec ) were examined (Fig. 3). In these gradients, Lpp and apoLpp were both detected exclusively in the middle of the gradient, as were SecG, Pal, the porins, and all other membrane proteins detected by Coomassie Blue staining (Fig. 3). These data indicate that the architecture of the cell envelope is drastically altered by Lnt Ec depletion, probably because the outer and plasma membranes become juxtaposed (see below) shortly before the onset of lysis.
Cell envelope instability was also observed by phase contrast microscopy of Lnt Ec -depleted E. coli, which were oval and swollen (data not shown). A similar phenotype was reported for the S. enterica lnt ts mutant SE5312 at 42°C (21). Furthermore, our sucrose flotation gradient analysis of membranes from this mutant grown at 42°C revealed that plasma and outer membrane proteins were in the same fractions in the center of the gradient, whereas these two classes of proteins were clearly separated in membranes from the wild-type strain (LT2) or from the mutant grown at 30°C (data not shown). Thus, Lnt Ec depletion in E. coli has the same effects on envelope architecture as Lnt Se inactivation in S. enterica.
A similar association of plasma and outer membranes to that caused by Lnt Ec depletion or inactivation was shown previously (27) to occur in E. coli after production of Lpp carrying the Asp ϩ2 signal. It was proposed that this protein (LppDK) prevents the separation of the two membranes according to their density and causes lysis because it is both anchored in the plasma membrane and covalently linked to the peptidoglycan by its C-terminal lysine residue (27). Most interestingly, introduction of an lpp::Tn10 mutation into S. enterica strain SE5312 abolished temperature sensitivity caused by the lnt ts mutation (21). Therefore, a major factor contributing to the lethality caused by Lnt Ec depletion in the strain carrying the p araB -lnt Ec allele could be the proposed retention of peptidoglycan crosslinked apoLpp in the plasma membrane (25).
To test this idea, we analyzed 12 independent revertants of strain PAP8504 (p araB -lnt Ec ) selected on LB agar without arabinose in a search for extragenic suppressor mutations in lpp. The presence of the kan-rpoCter-p araB cassette in the revertants was verified by PCR using primers flanking the cassette, which was then transduced into strain BW25113 by P1 phage. Three of the mutants were resistant to P1 phage, probably due to changes in surface lipopolysaccharide composition, whereas a further six had acquired a mutation in p araB that rendered them AraC-independent. Two of the remaining three mutants were found to be devoid of Lpp when examined by SDS-PAGE and immunoblotting, whereas the third produced a form of Lpp that migrated aberrantly upon SDS-PAGE (slower migrating monomeric form and abundant dimeric form; not shown). Sequence analysis revealed that the lpp gene in this mutant encodes a protein with glycine to aspartate substitution at position 14 in the signal peptide. Most interestingly, the same mutation was previously identified in an E. coli K12 mutant producing unmodified and unprocessed Lpp that was not crosslinked to the peptidogylcan and, therefore, was phenotypically Lpp Ϫ (38). These results corroborate the idea that cross-linking of plasma membrane apoLpp to the peptidoglycan is partially responsible for the lysis resulting from Lnt Ec depletion.
Finally, an lpp::Tn10 derivative of the p araB -lnt Ec strain PAP8504 was constructed by P1 transduction (PAP8505). This strain, like the revertants (see above) grew on LB agar but did not produce colonies upon drastic repression by glucose. In LB liquid cultures, PAP8505, like the Lpp ϩ parent strain, lysed 8 generations after replacement of arabinose by glucose. When pJY111 carrying the wild-type allele of lpp (27) was introduced into strain PAP8505, the transformants lost their ability to grow on agar in the absence of arabinose. In contrast, introduction of pJY151 (27), which carries an lpp allele encoding a variant of Lpp (LppSR) in which the C-terminal lysine that is normally cross-linked to the peptidoglycan is replaced by arginine, still allowed growth on agar in the absence of arabinose (data not shown). These data confirm that the toxicity caused by failure to express p araB -lnt Ec can be partially relieved by preventing Lpp synthesis, export, or cross-linking to the peptidoglycan.
ApoLpp Is Localized in the Plasma and Outer Membranes-Although PAP8505 cells lacking Lpp are fragile and leak periplasmic proteins (4), their membranes can be separated by sucrose gradient centrifugation (not shown). Synthesis of LppSR did not affect this separation (Fig. 4A), allowing it to be used as a marker for localizing apolipoproteins after Lnt Ec depletion. LppSR was detected in the outer membrane, whereas the apoLppSR was detected mainly in the plasma membrane fractions (Fig. 4A), although some apoLppSR also appeared in fractions enriched in outer membrane proteins (Fig. 4B). As already noted, apoPal could not be detected under these conditions, and Pal was exclusively present in the outer membrane fraction of Lnt Ec -depleted cells (Fig. 4A).
Other Outer Membrane Lipoproteins Also Accumulate in the Plasma Membrane When Lnt Ec Levels Are Decreased-To de- termine whether the localization of other outer membrane lipoproteins was affected by Lnt Ec depletion, strain PAP8505 was transformed with pCHAP1447, encoding a fatty acylated variant of the E. coli periplasmic maltose-binding protein MalE (13) with C-terminal hexahistidine extension. This protein (li-poCA-MalE) has an alanine residue at position ϩ2 and is normally localized to the outer membrane (13). However, when strain PAP8505(pCHAP1447) was grown in glucose (LntϪ in Fig. 5), a substantial amount of lipoCA-MalE accumulated in the plasma membrane. Sequence analysis of lipoCA-MalE purified by cobalt affinity chromatography from the plasma and outer membrane peaks of the glucose-grown cells revealed the presence of protein with a free N terminus, indicating the presence of apoLipoCA-MalE in both peaks. The same protein purified from cells with Lnt did not have a free N terminus. These data are consistent with the expected consequences of Lnt depletion and indicate that failure to N-acylate the Nterminal cysteine residue of a lipoprotein can be detected by N-terminal sequencing of the purified protein.
Substantial amounts of the endogenous outer membrane lipoprotein NlpD (39) also accumulated in the plasma membrane fraction of the glucose-grown cells, although Pal was again located exclusively in the outer membrane in both glu-cose and arabinose-grown cells (Fig. 5). Although it seems probable that the plasma membrane-associated NlpD was in the apo-form, it could not be separated from the mature (outer membrane-associated) form by SDS-PAGE.
Alternative Plasma Membrane Retention Signals Do Not Operate by Impeding Lnt Ec -Aromatic amino acids and proline at position ϩ2 in the artificial lipoprotein, lipoMalE, cause its retention in the plasma membrane (13). To test whether these amino acids, unlike Asp ϩ2 (25), function by impeding Lnt Ec , we determined whether the N terminus of five lipoMalE derivatives that accumulate in the plasma membrane (Asp ϩ2 , Phe ϩ2 , Pro ϩ2 , Trp ϩ2 , or Tyr ϩ2 ) (13) was blocked to Edman degradation. Like the lipoCA-MalE described above, the proteins were tagged with a C-terminal hexahistine that allowed them to be affinity-purified from detergent-solubilized envelope preparations. All five proteins were found to have a blocked N terminus, indicating the presence of an N-acyl group (see above) or, less likely, some other N-terminal modification. Thus, none of the known plasma membrane retention signals appears to operate by preventing N-acylation of the cysteine residue by Lnt Ec , as already demonstrated for Asp ϩ2 (25).
Characterization of the Mutation in lnt ts Gene of the S. enterica ts Mutant-Because the mutation in S. enterica strain SE5312 that causes the temperature-sensitive phenotype (21) was not characterized, the lnt ts gene from this mutant was PCR-amplified and sequenced. A single mutation causing a glutamate to lysine substitution at position 435 was found in comparison with the lnt Se gene amplified from S. enterica strain LT2. Most interestingly, Glu 435 is located in a highly conserved region of Lnt Se .
The same mutation was introduced in the E. coli K12 lnt Ec gene by directed mutagenesis, thereby producing the lnt Ec (E435K) allele. Trans-complementation tests were then performed in S. enterica strain SE5312 and in E. coli PAP8504 with lnt or lnt(E435K) from E. coli or S. enterica cloned in high copy plasmids. Wild-type alleles of lnt Ec (pCHAP6571) and lnt Se (pCHAP6573) complemented both the lnt ts mutation in strain SE5312 at 37 and 42°C and the p araB -lnt Ec mutation in strain PAP8504 grown in glucose (Table IV), indicating that the Lnt enzymes from these two species are sufficiently similar (26 conservative substitutions and 28 nonconservative substitutions out of 512 residues) to be functionally interchangeable.
The p araB -lnt Ec mutant PAP8504 failed to grow at either 37 or 42°C when Lnt Se (E435K) (pCHAP6574) was produced, confirming that the temperature sensitivity is caused by the E435K substitution in Lnt Se . Most interestingly, both Lnt Se (E435K) and Lnt Ec (E435K) (pCHAP6576) prevented growth of S. enterica wild-type strain LT2 at 42°C, suggesting that overproduction of either protein in LT2 is either toxic or is dominant negative over the activity of chromosome-encoded functional Lnt Se at this temperature (Table IV). However, Lnt Ec (E435K) and Lnt Se (E435K) were not toxic at 42°C in E. coli strain PAP8504 on LB agar containing arabinose, possibly because more Lnt Ec is produced when lnt Ec is expressed from p araB in E. coli than when the lnt Se is expressed from its own promoter in S. enterica. Moreover, in contrast to Lnt Se (E435K), Lnt Ec (E435K) restored arabinose-independent growth of E. coli PAP8504 even at 42°C. These data show that, even though Lnt Ec and Lnt Se proteins are functionally interchangeable, the E435K substitution resulted in a temperaturesensitive enzyme only in Lnt Se , implying that other amino acid differences between these two proteins determine whether or not this substitution has an effect on enzyme activity or stability.
Plasma Membrane Topology of Lnt Ec -Bioinformatic analyses of lnt genes from Gram-negative bacteria predicted the presence of 8 segments of sufficient hydrophobicity to adopt a transmembrane topology ((Gln 10 -Ala 27 ) I , (Trp 34 -Asn 50 ) II , (Ala 57 -Val 75 ) III , (Pro 88 -Leu 112 ) IV , (Trp 121 -Leu 138 ) V , (Leu 163 -Leu 187 ) VI , (Leu 195 -Ile 211 ) VII , and (Trp 489 -Leu 507 ) VIII ; amino acid positions according to Lnt Ec ). This predicted topology of Lnt Ec was tested by constructing a series of lnt Ec -phoA and lnt Ec -lacZ gene fusions. The lacZ and phoA genes (encoding ␤-galactosidase LacZ and alkaline phosphatase PhoA, respectively) lacking their 5Ј-translation start signals (and, in the case of phoA, lacking a part of the region coding for the signal peptide) were PCR-amplified and fused to nine selected sites in regions of lnt Ec encoding loops between the predicted transmembrane segments, in plasmids under the p araB promoter. Each fusion was expressed in E. coli under arabinose induction, and the activity of the reporter proteins was determined in permeabilized cells. Immunoblotting using anti-PhoA (alkaline phosphatase) and anti-LacZ (␤-galactosidase) antibodies revealed that the chimeras were stable and that their estimated sizes were as predicted from knowledge of the fusion site (Fig.  6). Hybrid proteins with junction sites after Pro 53 , Pro 117 , Pro 154 , Arg 190 , and Lys 512 of Lnt Ec showed relatively high or moderate LacZ and low PhoA activities, indicating that these sites are exposed to the cytoplasm (PhoA is only active when exported to the periplasm, whereas LacZ is active only when retained in the cytoplasm (40)). When the reporters were placed after Ala 80 , Pro 218 , and Pro 476 , LacZ activities were at least 10 times lower, and PhoA activities at least 22 times higher than when the junction sites were at positions 53, 117, 154, 190, or 512 (Fig. 6), indicating that these sites are exposed on the periplasmic side of the plasma membrane. We note, however, that the specific ␤-galactosidase activity of LacZ fusions at positions 53, 117, 154, and 190 was considerably lower than at position 512, although it was reproducibly higher than at positions 80, 219, and 476 (Fig. 6). Most surprisingly, fusions constructed after Pro 30 of Lnt Ec gave relatively high PhoA and LacZ activities. The N-terminal hydrophobic segment of Lnt Ec presumably acts as a signal sequence to drive PhoA export, consistent with the presence of three arginine residues on its N-terminal (cytoplasmic) side (41), but cannot promote LacZ export, as reported previously for similar signal peptide-LacZ constructs (42). Similarly high enzyme activities were reported for other LacZ chimeras in which the fusion site was located after the first transmembrane segment of different E. coli polytopic plasma membrane proteins (43)(44)(45).
We conclude that Lnt Ec probably possesses six rather than the predicted eight transmembrane segments interconnected by hydrophilic loops, with the N and C termini located in the cytoplasm (Fig. 7)  itive-inside rule (41). According to this prediction, residue Glu 435 is located in the large periplasmic loop between transmembrane segments 5 and 6 ( Fig. 7). DISCUSSION The data reported here provide in vivo evidence that Lnt Ecmediated N-acylation of major outer membrane lipoprotein Lpp and probably also of outer membrane lipoprotein NlpD and the artificial lipoprotein lipoCA-MalE is required for their efficient release from the plasma membrane. These data corroborate previous in vitro studies showing that LolA cannot promote apoPal release from proteoliposomes by LolCDE (25). In view of the substantial structural differences between alternative lipoprotein plasma membrane retention signals (Phe ϩ2 , Pro ϩ2 , Trp ϩ2 , and Tyr ϩ2 ) (13) and the canonical plasma membrane retention signal or Lol avoidance signal (Asp ϩ2 ), it was speculated that one or all of them might operate by preventing N-acylation of normally outer membrane lipoproteins. We demonstrated that this is not the case by showing that lipoMalE variants possessing these signals have blocked N termini by Edman degradation, suggesting the presence of an N-acyl group. The way these amino acids prevent Lol-mediated lipoprotein transport to the outer membrane remains to be clarified.
Even though apoLppSR was clearly visualized after Lnt Ec depletion, the apo-form of another abundant outer membrane lipoprotein, Pal, could not be detected (data not shown), possibly because, like the apo-forms of lipoCA-MalE and NlpD, it cannot be separated from the mature form by SDS-PAGE. Furthermore, Pal was reproducibly found exclusively in the outer membrane in Lnt Ec -depleted cells, whereas apoLpp and two other outer membrane lipoproteins, NlpD and lipoCA-MalE accumulated in the plasma membrane (Figs. 4 and 5). There are four possible explanations for these observations.
(i) ApoPal has a higher affinity than apoLpp for Lnt Ec . Therefore, trace amounts of Lnt Ec remaining after long periods of p araB -lnt Ec repression are sufficient for N-acylation of apoPal.  (ii) ApoPal is transported to the outer membrane by the Lol system, although this would be in contradiction to the observation that LolCDE cannot release apoPal from proteoliposomes in the presence of LolA (25).
(iii) ApoPal is retained in the plasma membrane but is degraded.
(iv) ApoPal associates with the peptidoglycan or with outer membrane components such as OmpA (46), and its association with the plasma membrane is consequently disrupted when the cells are broken in the French press. Cross-linking of the plasma membrane apo-form of Lpp to the peptidoglycan could explain the cofractionation of the plasma and outer membrane pools from envelopes of Lnt Ec -depleted cells in sucrose gradients, because the peptidoglycan would be linked to both the outer membrane (via mature Lpp) and to the plasma membrane (via apoLpp). An identical phenomenon occurs when proLpp (uncleaved and unacylated) accumulates in the plasma membrane because of the absence of phosphatidylglycerol caused by a mutation in the pgsA gene (47) or when Lpp (LppDK) is directed to the plasma membrane by a Asp ϩ2 (27). In all cases, rupture of the bacteria in a French press would result in the formation of outer and plasma membrane vesicles that are both linked to the peptidoglycan, causing them to float to an interme-diate density in the sucrose gradients.
The disorganization of the cell envelope that results from the mislocalization of peptidoglycan-bound apoLpp (Fig. 3) is presumably a major cause but not the only cause of lysis that ensues Lnt Ec depletion. Membrane disorganization might also explain why some apoLppSR, which is not linked to the peptidoglycan, was observed in fractions containing outer membrane proteins when Lnt Ec was depleted (Fig. 4B). Bulk mixing of plasma and outer membrane can be ruled out as a possible explanation because the integral plasma membrane protein SecG did not appear in increased amounts in the dense (outer) membrane fractions under these conditions (Fig. 4). One possible explanation for this observation could be the formation of heterotrimers of Lpp containing a mixture of mature and apo-forms (48), although we cannot rule out that apoLpp is transported, albeit inefficiently, to the outer membrane by the Lol system. ApoLipoCA-MalE was also found in outer membrane fractions when Lnt was depleted, but this observation is more difficult to interpret because the assay used (detection of a free N terminus that allows Edman degradation) is not quantitative. Therefore, only trace amounts of apoLipoCA-MalE, comparable with those of SecG, might be present on the outer membrane fraction because of mixing of membrane components.
The inhibition of growth caused by severe glucose repression of p araB -lnt Ec in E. coli was not overcome by elimination of Lpp, whereas lpp::Tn10 restored growth of the S. enterica lnt ts mutant at 42°C (21), probably because the lnt ts mutation does not completely inactivate Lnt Se at 42°C. Furthermore, the fact that glucose repression of p araB -lnt Ec in E. coli is lethal even in the absence of Lpp suggests that lnt expression is completely blocked under these conditions, leading to the inactivation of essential outer membrane lipoproteins (LolB, for example (49)) by delocalization (in the apo-form) to the plasma membrane (49,50).
Mutations in the lnt (cutE/actA) genes have been reported to confer copper sensitivity (22,51). This is an interesting observation, especially in view of the presence of a putative copperbinding site, HXXMXXM (amino acids 425-431; HFQMARM in E. coli Lnt (22)), in the large, well conserved periplasmic loop of Lnt Ec between transmembrane segments 5 and 6. Copper binding by Lnt has not been investigated, however, and only the first amino acid of this putative motif (His) is highly conserved in Lnt from different bacterial species (as are the glutamine and arginine that precede and follow it, respectively), whereas the two methionines are often replaced by nonconservative amino acids in predicted Lnt homologues from bacteria distantly related to E. coli (data not shown). Thus, the effects of lnt mutations on copper sensitivity are probably due to incorrect localization or inactivation of one or more outer membrane lipoproteins (as FIG. 6. Detection by immunoblotting and enzymatic activities of Lnt Ec -PhoA and Lnt Ec -LacZ hybrid proteins in E. coli. Crude extracts from KS272 producing Lnt Ec -PhoA and Lnt Ec -LacZ hybrid proteins or carrying plasmids without lnt Ec -derived inserts (control) were analyzed by SDS-PAGE and by immunoblotting with anti-␤-galactosidase and anti-alkaline phosphatase antibodies. Cells were grown in LB medium containing arabinose to induce expression of the gene fusions. Alkaline phosphatase and ␤-galactosidase activities were measured in permeabilized cells from the same cultures. Enzyme activities (in arbitrary units) are the average of three independent assays. Maximum variations were Ϯ14.6% for alkaline phosphatase and Ϯ23.8% for ␤-galactosidase. apolipoproteins) involved in copper homeostasis (e.g. NlpE (52)), rather than to direct copper-Lnt interactions.
The coding sequence of the mutant lnt (cutE) gene amplified from the copper-sensitive E. coli mutant reported by Rogers et al. (22) was found to be devoid of mutations (data not shown). This suggests that the mutation in this strain is not in lnt itself but might be in the upstream ybeX promoter and, consequently, might diminish lnt expression. On the other hand, the transposon insertion in the R. meliloti lnt (actA) gene that causes copper and acid sensitivity would be expected to inactivate the gene, implying that Lnt function is not essential in Rhizobium (51). Most interestingly, the amino acid substitution that inactivates the S. enterica Lnt at 42°C (E435K) (52) affects one of several totally conserved amino acids in Lnt homologues and might define part of the catalytic site of the enzyme. The mechanisms by which Lnt hydrolyzes phospholipids and transfers the resulting fatty acids onto the N-terminal cysteine residue of lipoproteins should be investigated.