An Outer Membrane Enzyme Encoded by Salmonella typhimurium lpxR That Removes the 3′-Acyloxyacyl Moiety of Lipid A*

The Salmonella and related bacteria modify the structure of the lipid A portion of their lipopolysaccharide in response to environmental stimuli. Some lipid A modifications are required for virulence and resistance to cationic antimicrobial peptides. We now demonstrate that membranes of Salmonella typhimurium contain a novel hydrolase that removes the 3′-acyloxyacyl residue of lipid A in the presence of 5 mm Ca2+. We have identified the gene encoding the S. typhimurium lipid A 3′-O-deacylase, designated lpxR, by screening an ordered S. typhimurium genomic DNA library, harbored in Escherichia coli K-12, for expression of Ca2+-dependent 3′-O-deacylase activity in membranes. LpxR is synthesized with an N-terminal type I signal peptide and is localized to the outer membrane. Mass spectrometry was used to confirm the position of lipid A deacylation in vitro and the release of the intact 3′-acyloxyacyl group. Heterologous expression of lpxR in the E. coli K-12 W3110, which lacks lpxR, resulted in production of significant amounts of 3′-O-deacylated lipid A in growing cultures. Orthologues of LpxR are present in the genomes of E. coli 0157:H7, Yersinia enterocolitica, Helicobacter pylori, and Vibrio cholerae. The function of LpxR is unknown, but it could play a role in pathogenesis because it might modulate the cytokine response of an infected animal.

Lipopolysaccharide (LPS) 3 is the principal component of the outer leaflet of the outer membrane of Gram-negative bacteria. Recognition of LPS by the mammalian innate immune system results in the production of cell adhesion proteins in endothelial cells and of pro-inflammatory molecules such as tumor necrosis factor-␣ and interleukin-1␤ in monocytes (7,8). Lipid A, the hydrophobic anchor of LPS, produces most of these responses (9, 10) after its detection by Toll-like receptor 4 (TLR-4) (11)(12)(13). Lipid A of S. typhimurium and Escherichia coli is a ␤1Ј-6-linked disaccharide of glucosamine, phosphorylated at the 1 and 4Ј positions and acylated at the 2, 3, 2Ј, and 3Ј positions with R-3-hydroxymyristate ( Fig. 1A) (9,14,15). The hydroxyl groups of the R-3-hydroxymyristate residues, attached at positions 2Ј and 3Ј, are further acylated with laurate and myristate, respectively. Key features of the molecule shown to be important for TLR-4 activation include the phosphate groups and the secondary acyl chains (15)(16)(17). The biosynthetic enzymes that generate lipid A have been identified and their corresponding structural genes cloned, primarily from E. coli (15).
We now report a novel, Ca 2ϩ -dependent 3Ј-O-deacylase present in the membranes of S. typhimurium (Fig. 2). The structural gene (lpxR) encoding the 3Ј-O-deacylase was identified and expressed in E. coli K-12. LpxR, like PagP and PagL, is localized to the outer membrane. Expression of LpxR in the E. coli K-12 strain W3110 results in a significant production of 3Ј-Odeacylated lipid A species but does not affect growth. Ortho-logues of S. typhimurium LpxR are found in the genomes of several other Salmonella species, including both typhi and paratyphi, as well as several other Gram-negative pathogens. The function of LpxR is not known; however, partial 3Ј-Odeacylation of lipid A might be advantageous during infection, allowing S. typhimurium to evade the innate immune response by remodeling its lipid A.

EXPERIMENTAL PROCEDURES
Materials-32 P i and [␥-32 P]ATP were obtained from PerkinElmer Life Sciences. Silica gel 60 (0.25 mm) thin layer  (15). The numbering scheme is used for the NMR analysis. B, the phosphate groups and the acyl chains of S. typhimurium Kdo 2 -lipid A can be derivatized in a regulated fashion. The phosphate moieties of lipid A can be substituted with 4-amino-4-deoxy-L-arabinose by L-4-aminoarabinose transferase (ArnT) (18) and/or phosphoethanolamine by phosphoethanolamine transferase (EptA) (PmrC) (19,20). Formation of the 2-hydroxymyristate group requires an oxygen-dependent hydroxylase, designated LpxO (25). The incorporation of the acyloxyacyl-linked palmitate chain at the 2-position is catalyzed by PagP (26); removal of the ester-linked R-3-hydroxymyristoyl chain at the 3-position is catalyzed by PagL (27). chromatography (TLC) plates were from Merck. Chloroform, ammonium acetate, and sodium acetate were obtained from EM Science. Tryptone and yeast extract were from Difco. The bicinchoninic acid protein determination kit and Triton X-100 were from Pierce. All other chemicals were reagent grade and were purchased from either Sigma or Mallinckrodt.
Bacterial Strains-The bacterial strains used in this study are described in Table 1. Typically, bacteria were grown at 37°C in LB medium (28). When required for selection of plasmids, cells were grown in the presence of 100 g/ml ampicillin, 12 g/ml tetracycline, 25 g/ml chloramphenicol, or 30 g/ml kanamycin.
Molecular Biology Applications-Protocols for handling of DNA samples were those of Sambrook and Russell (29). Transformation-competent cells of E. coli were prepared by the method of Inoue et al. (30). When required, E. coli cells were prepared for electroporation by the method of Sambrook and Russell (29). Plasmids were isolated using the Qiagen Spin Prep kit. DNA fragments were isolated from agarose gels using the QIAquick gel extraction kit. Genomic DNA was isolated using the protocol for bacterial cultures in the Easy-DNA™ kit (Invitrogen). T4 DNA ligase (Invitrogen), restriction endonucleases (New England Biolabs), and shrimp alkaline phosphatase (U. S. Biochemical Corp.) were used according to the manufacturers' instructions. Double-stranded DNA sequencing was performed with an ABI Prism 377 instrument at the Duke University DNA Analysis Facility. Primers were purchased from MWG-Biotech.
Screening of a S. typhimurium Genomic DNA Library for the 3Ј-O-Deacylase-A bacterial artificial chromosome (BAC) library generated from genomic DNA of the S. typhimurium strain LT2 and harbored in an E. coli K-12 background was obtained from the Salmonella Genetic Stock Center (University of Calgary). Each of the 71 BAC clones harbors the pBeloBAC11 vector (31) with inserts of S. typhimurium genomic DNA. The BAC clones were grown in 5 ml of LB medium with 25 g/ml chloramphenicol to an A 600 ϳ 1.0. The membrane fraction of each BAC clone was prepared as previously described (32) and assayed for 3Ј-O-deacylase activity. Protein concentrations were determined by the bicinchoninic acid method using bovine serum albumin as the standard.
Construction of lpxR Expression Vectors-S. typhimurium lpxR, formerly STM1328, was cloned into pET23a (Novagen) behind the T7 promoter. The predicted coding region for lpxR was amplified by PCR from S. typhimurium strain LT2 genomic DNA using Pfu DNA polymerase (Stratagene), according to the manufacturer's instructions. The forward primer contained a clamp region, an NdeI site (underlined), and the lpxR-coding region with its start codon (bold). The reverse primer contained a clamp region, a BamHI site (underlined) and the coding region with its stop codon (bold). Sequences of the forward and reverse primers were 5Ј-GCGCGCCATATGAACAAATA-CAGCTATTGCGCAACG-3Ј and 5Ј-GCGCGCGGATCCT-CAGAAGAAGAAGGTGATGTCTCC-3Ј, respectively. The PCR product and the vector were both digested with NdeI and BamHI, ligated together, and transformed into XL-1 Blue cells (Stratagene) for propagation of the plasmid, designated pLpxR1. For some experiments, pLpxR1 was used directly for LpxR expression in E. coli NovaBlue(DE3) ( Table 1). Alternatively, the lpxR gene was transferred from pET23a to pWSK29 (33), a lac-inducible, low-copy expression vector. The XbaI/ BamHI-digested fragment, consisting of the lpxR gene as well as the pET23a-derived ribosome binding site, was ligated to the corresponding restriction sites of pWSK29. This plasmid, designated pLpxR2, was then transformed into competent E. coli W3110 or S. typhimurium LT2.
Heterologous Expression of Salmonella LpxR-E. coli Nova-Blue(DE3) harboring either the pET23a vector or pLpxR1 were grown at 37°C until the A 600 reached 0.4. Isopropyl-1-thio-␤-D-galactopyranoside at 1 mM was then added, and the cultures were shifted to 25°C for 4 h more. Washed membranes were prepared as previously described (32) and stored in 50 mM MES, pH 6.5, at Ϫ80°C. Cultures of W3110 harboring either pWSK29 or pLpxR2 were grown at 37°C with 1 mM isopropyl-1-thio-␤-D-galactopyranoside present throughout the growth. Cells were harvested when the A 600 of the cultures reached 1.0, and washed membranes were prepared as previously described (32). Membranes were also prepared from LB-grown cultures of S. typhimurium LT2 grown without CaCl 2 or in the presence of 5 mM CaCl 2 using the methods described above.
Assay of the 3Ј-O-Deacylase-The 3Ј-O-deacylase activity of LpxR was assayed under optimized conditions in a 15-l reaction mixture containing 50 mM MES, pH 6.5, 1% Triton X-100, 5 mM CaCl 2 , and either 10 M Kdo 2 -4Ј-32 P-labeled lipid A, Kdo 2 -4Ј-32 P-labeled lipid IV A , or 4Ј-32 P-labeled lipid IV A (each at 50,000 cpm/nmol) as the substrate. Washed membranes were employed as the enzyme source. Reactions were incubated at 30°C for the indicated times. Reactions were stopped by spotting 4-l portions of the mixtures onto a Silica Gel 60 TLC plate, and the plate was dried under a cool air stream for 20 min. Reaction products were separated using the solvent chloroform, pyridine, 88% formic acid, water (50:50:16:5, v/v), whereas reactions containing Kdo 2 -4Ј-32 P-labeled lipid IV A were analyzed in the solvent chloroform, pyridine, 88% formic acid, water (30:70:16:10, v/v). The reactions that were carried out with the hexa-acylated substrate, Kdo 2 -4Ј-32 P-labeled lipid A, were analyzed in the solvent chloroform, methanol, water, acetic acid (25:15:4:4, v/v). After chromatography, the plates were dried, and radioactive substrates were detected using a GE Healthcare PhosphorImager (STORM 840) equipped with ImageQuant software. The enzyme units (nmol/min) were calculated by determining the percentage of the substrate converted to product.
Purification of the Salmonella 3Ј-O-Deacylase in Vitro Reaction Product-A large scale 3Ј-O-deacylase reaction mixture was prepared by assembling the following components: 50 mM MES, pH 6.5, 1% Triton X-100, 5 mM CaCl 2 , 100 M Kdo 2 -lipid A, and either 1 mg/ml S. typhimurium strain LT2 membranes or 0.1 mg/ml NovaBlue(DE3)/pLpxR1 membranes, as described below. The reaction was initiated by the addition of the membranes and incubated at 30°C for 12 h. The reaction mixture was then converted into a two-phase Bligh/Dyer (43) system consisting of chloroform, methanol, and 0.1 M HCl (2:2: 1.8, v/v). The phases were separated by centrifugation at 4000 ϫ g for 8 min. The organic phase was removed, and the resulting aqueous phase was extracted a second time by the addition of pre-equilibrated acidic organic phase. The organic phases were pooled and neutralized by the addition of a few drops of pyridine. The sample was dried by rotary evaporation, and lipids were fractionated using a DEAE cellulose anion-exchange column as previously described (44). Fractions containing the desired lipid species were pooled and converted to an acidic two-phase Bligh/Dyer mixture by the addition of appropriate amounts of chloroform and HCl. The organic phase was dried under a stream of N 2 and stored at Ϫ80°C.
Separation of Inner and Outer Membranes-Membranes isolated from either W3110/pWSK29 or W3110/pLpxR2 were separated by isopycnic sucrose gradient centrifugation using previously described methods (45,46). Each gradient fraction (0.5 ml) was assayed for NADH oxidase (inner membrane marker) and phospholipase A (outer membrane marker) (47). The amount of protein in each fraction was determined using the bicinchoninic acid assay (48). Each fraction was also assayed for 3Ј-O-deacylase activity using the standard conditions described above.
Protein Microsequencing of LpxR-Inner and outer membranes of W3110/pWSK29 and W3110/pLpxR2 were analyzed by SDS-PAGE on a Bio-Rad Protean II XI apparatus with a 12% polyacrylamide, 1.5 mm thick gel (49). A piece of the gel containing LpxR was excised. The protein was blotted onto an Immobilon-P polyvinylidene fluoride membrane (Millipore) equilibrated in 10 mM CAPS, pH 11, containing 10% methanol at 15 V for 30 min using a Bio-Rad Semi-Dry apparatus. The blot was stained with 0.1% Ponceau S in 1% acetic acid for 1 min followed by destaining with 1% acetic acid for 10 min. The band corresponding to LpxR was excised, rinsed three times with distilled water, and analyzed by high sensitivity protein microsequencing at the Harvard Microchemistry and Proteomics Facility (Cambridge, MA).
Large Scale Isolation of Lipid A from LpxR-expressing Cultures-Cultures of E. coli strain W3110 harboring either pWSK29 or pLpxR2 were grown in 1 liter of LB medium as described above. The cultures were harvested by centrifugation at 5000 ϫ g for 20 min when the A 600 reached 1.0. The cells were washed once with 40 ml of phosphate-buffered saline, pH 7.4. The final cell pellet was resuspended in 40 ml of phosphatebuffered saline, pH 7.4. Lipid A was released from the LPS, purified as described previously (38,39), and stored frozen at Ϫ80°C. Lipid A was also prepared from CaCl 2 (10 mM)-treated cultures of S. typhimurium strain LT2 harboring either pWSK29 or pLpxR2, as just described.
Mass Spectrometry of Kdo 2 -Lipid A or Lipid A Samples-Spectra were acquired in the negative-ion linear mode using an AXIMA-CFR (Kratos Analytical, Manchester, UK) matrix-assisted laser desorption ionization/time of flight (MALDI-TOF) mass spectrometer with a 337-nm nitrogen laser. The instrument was operated with a 20-kV extraction voltage and timedelayed extraction, providing a mass resolution of about Ϯ1 atomic mass unit for compounds with a M r of ϳ2, 000. Each spectrum was the average of 100 laser shots. Samples were prepared for MALDI-TOF analysis by depositing 0.3 l of the lipid dissolved in chloroform/methanol (4:1, v/v) followed by 0.3 l of a saturated solution of 6-aza-2-thiothymine in 50% acetonitrile and 10% tribasic ammonium citrate (9:1, v/v) as the matrix. The samples were dried at room temperature before the spectra were acquired.
Mass Spectrometry of the Released 3Ј-Acyloxyacyl Moiety-Spectra were acquired on a QSTAR XL quadrupole TOF tandem mass spectrometer (ABI/MDS-Sciex, Toronto, Canada) equipped with an electrospray ionization (ESI) source. Spectra were acquired in the negative-ion mode and typically were the accumulation of 60 scans collected from 200 -2000 atomic mass units. For mass spectrometry (MS) analysis, the extracted lipids were dissolved in 200 l of chloroform/methanol (2:1, v/v) and infused into the ion source at 5-10 l/min. The nega-  ). B, S. typhimurium LT2 membranes (0.1 mg/ml) from the "no Ca 2ϩ " culture were assayed with 10 M Kdo 2 -4Ј-32 P-labeled lipid A for the indicated times. C, Kdo 2 -lipid A and its deacylated product were partially purified from a large-scale reaction mixture, prepared with 1 mg/ml S. typhimurium LT2 membranes. The lipids were analyzed by negative-ion mode MALDI-TOF mass spectrometry. amu, atomic mass units.
tive-ion ESI was carried out at Ϫ4200 V. Data acquisition and analysis was performed using the Analyst QS software.
NMR Spectroscopy of 3Ј-O-Deacylated Lipid A-The lipid A of E. coli W3110 harboring pLpxR2 was isolated as described above and further purified by DEAE cellulose anion exchange chromatography (44). Fractions containing 3Ј-O-deacylated lipid A were pooled, and the lipid was extracted by conversion into a two-phase Bligh/Dyer system (43). The lower phase was dried under a stream of N 2 , and 0.5 mg of the dried lipid was dissolved in 0.35 ml of CDCl 3 /CD 3 OD/D 2 O (2:3:1, v/v) in a 3-mm NMR tube. NMR spectroscopy was carried out at the Duke University NMR Spectroscopy Center. Proton chemical shifts are reported relative to tetramethylsilane at 0.00 ppm. NMR spectra were obtained on Varian Inova 800 or 500 MHz NMR spectrometers, each equipped with a Sun Ultra 10 computer and 5-mm Varian probe. 1 H NMR spectra at 800 MHz were obtained with a 7.2-kHz spectral window, a 67°pulse field angle (4.5 sec), a 4.5-s acquisition time, and a 1-s relaxation delay. The spectra were digitized using 64,000 points to obtain a digital resolution of 0.225 Hz/point. Two-dimensional NMR experiments with solvent suppression (correlation spectroscopy, two-dimensional nuclear Overhauser effect spectroscopy, total correlation spectroscopy, multiple quantum coherence) were performed as previously described (50,51). Directly detected, 1 H-decoupled 31 P NMR spectra were recorded at 202.37 MHz with a spectral window of 12143.3 Hz digitized into 25,280 data points (digital resolution of 1 Hz/point or ϳ0.005 ppm/point), a 60°pulse flip angle (8 s), and a 1.6-s repeat time. 31 P chemical shifts were referenced to 85% H 3 PO 4 at 0.000 ppm. Inverse decoupled difference spectra were recorded as 1 H-detected, 31 P-decoupled heteronuclear NMR experiments as previously described (50,51). Figs. 3, A and B, membranes of S. typhimurium LT2 efficiently convert Kdo 2 -4Ј-32 P-labeled lipid A to an unknown product, designated A, when 5 mM Ca 2ϩ is included in the assay (lanes 4 and 5 versus lanes 2 and 3). Formation of A is not strictly dependent upon the addition of Ca 2ϩ to the growth medium (lane 4 versus lane 5). The activity is localized primarily to S. typhimurium outer membranes, as discussed in detail below (data not shown).

Ca 2ϩ -dependent Conversion of Kdo 2 -lipid A to a 3Ј-O-Deacylated Product-As shown in
Product A (Fig. 3A), generated in vitro with S. typhimurium membranes and Kdo 2 -lipid A, was partially purified by chromatography on DEAE cellulose, as described under "Experimental Procedures" and was analyzed by low resolution MALDI-TOF mass spectrometry in the negative ion mode. As shown in  To identify the gene encoding the enzyme, we assumed that an orthologue of the S. typhimurium 3Ј-O-deacylase was not present in E. coli K-12 and that it was annotated as an outer membrane protein in the S. typhimurium data base. Only one open reading frame, STM1328, within the relevant segment of the S. typhimurium genome met both criteria. The product of STM1328 has a predicted N-terminal signal peptide, consistent with outer membrane targeting (54). A tBLASTn (55) search with the predicted protein product of STM1328 also did not reveal any significant orthologues in E. coli K-12.
STM1328 (renamed lpxR) was subcloned into pET23a to create pLpxR1 and expressed in E. coli NovaBlue(DE3). As shown in Fig. 4 Table 2 lists some of the closest orthologues of LpxR present in the non-redundant data base. Helicobacter pylori and Yersinia enterocolitica, which have orthologues of LpxR, have been previously reported to contain lipid A species that lack the 3Ј-acyloxyacyl moiety (56,57). Ca 2ϩ -dependent 3Ј-O-deacylase activity is present in the membranes of several Gram-negative pathogens that contain LpxR orthologues, including E. coli 0157:H7, Y. enterocolitica, and Vibrio cholerae. 4 The complete nucleotide sequence of lpxR has been submitted to GenBank™ under accession number DQ272513.
LpxR-catalyzed Production of 3Ј-O-deacylated Kdo 2 -lipid A and Release of the Intact 3Ј-Acyloxyacyl Moiety-To confirm that the recombinant LpxR catalyzes the same 3Ј-O-deacylation of Kdo 2 -lipid A seen with Salmonella membranes (Fig. 3), a large-scale in vitro 3Ј-O-deacylase reaction mixture was prepared using E. coli NovaBlue(DE3)/pLpxR1 membranes as the enzyme source. After purification over DEAE-cellulose, negative-ion MALDI-TOF mass spectrometry was used to analyze the products eluting with 240 to 480 mM ammonium acetate. The major peak, interpreted as [M Ϫ H] Ϫ of 3Ј-O-deacylated Kdo 2 -lipid A, at m/z 1801.9 atomic mass units (data not shown) was the same within experimental error as that seen with wildtype Salmonella membranes (Fig. 3C).
High-resolution electrospray ionization (ESI) mass spectrometry was used to determine whether or not the 3Ј-acyloxyacyl moiety was released intact from Substrate Specificity of S. typhimurium LpxR-In vitro 3Ј-O-deacylation of Kdo 2 -4Ј-32 P-labeled lipid A by LpxR required the presence of the nonionic detergent Triton X-100 with optimal activity at 1%. LpxR also required Ca 2ϩ for activity, with an optimum at 5 mM (data not shown). Product formation by E. coli NovaBlue(DE3) membranes over-expressing LpxR, assayed with 10 M Kdo 2 -4Ј-32 P-labeled lipid A under optimized conditions, was linearly dependent upon both protein concentration and time (data not shown). The ability of LpxR to hydrolyze (Kdo 2 -4Ј-32 Plabeled lipid A, Kdo 2 -4Ј-32 P-labeled lipid IV A and 4Ј-32 Plabeled lipid IV A ) in vitro was compared. Assays with each of these substrates at 10 M revealed that 4Ј-32 P-labeled lipid IV A was a relatively poor substrate (1 nmol/min/mg of membrane protein) when compared with Kdo 2 -4Ј-32 P-labeled lipid IV A (132 nmol/min/mg) or Kdo 2 -4Ј-32 P-labeled lipid A (434 nmol/min/mg). The data show that the 3Ј-O-deacylase activity is strongly Kdo-dependent.
The outer membrane enzymes PagP and PagL have been previously shown to resist heat denaturation (26,27). Similarly, E. coli membranes over-expressing LpxR retained significant activity (ϳ67%) after a 10-min pre-incubation at 100°C (data not shown).
Outer Membrane Localization of Recombinant LpxR-Inner and outer membranes of E. coli W3110/pLpxR2 were separated by isopycnic sucrose gradient centrifugation (Fig. 7). The heavier outer membranes were detected by their phospholipase A activity, whereas the lighter inner membranes were detected by assaying for NADH oxidase (Fig. 7B). LpxR activity was pre-  dominantly localized to the outer membrane (Fig. 7A), although a portion was also recovered in the intermediate region (fractions 9 to 13), which was depleted of both marker enzymes (Fig. 7B). Total membranes, outer membranes, and inner membranes of W3110, harboring either the vector pWSK29 or pLpxR2, were analyzed by 12% SDS-PAGE. As shown in Fig. 8 To verify the presence of the signal peptide predicted by Signal-P (54) (http://www.cbs.dtu.dk/services/SignalP/) and to establish the exact cleavage site, LpxR from the outer membranes of W3110/ pLpxR2 was blotted to a polyvinylidene difluoride membrane and subjected to N-terminal micro-sequencing. The first ten amino acid residues were SSLAISVAND, indicating that the cleavage of the signal peptide occurred between residues 22 and 23 of the full-length protein, consistent with the prediction of the Signal-P program. Thus, mature LpxR is comprised of 296 amino acids with a molecular weight of 32,400.
Deacylation of Lipid A by LpxR in Living Cells-Although wildtype S. typhimurium membranes contain measurable LpxR activity (Fig. 3A), the lipid A of Salmonella is not appreciably 3Ј-O-deacylated under the commonly used growth conditions (58 -61). In contrast, heterologous over-expression of Salmonella LpxR in E. coli K-12 resulted in extensive 3Ј-O-deacylation of endogenous lipid A (Fig.  9). The lipid A species, released from cells by hydrolysis at pH 4.5, were purified and analyzed by lowresolution MALDI-TOF mass spectrometry. The endogenous lipid A of W3110/pWSK29 consisted primarily of the hexa-acylated bis-phosphate form, typically seen with wild-type E. coli K-12 strains. The peak at m/z 1798.4 atomic mass units in the negative-ion mode was interpreted as [M Ϫ H] Ϫ of lipid A (Fig.  9A). Upon over-expression of LpxR, an additional peak was seen at m/z 1361.6 atomic mass units in the negative-ion mode (Fig. 9B) (Fig. 10B).
NMR Spectroscopy of 3Ј-O-deacylated Lipid A Isolated from W3110/pLpxR2-The partial 800 MHz 1 H NMR spectrum of 3Ј-O-deacylated lipid A (Fig. 11), purified from W3110/pLpxR2, revealed well resolved resonances in the sugar and acyl chain regions (Table 3), similar to those previously reported for wildtype E. coli K-12 lipid A dissolved in CDCl 3 /CD 3 OD/D 2 O (2:3:1 v/v) (50,51). The sugar region was comprised of eight resolved single-proton resonances, three 2-proton envelopes (5.2, 3.97 and 3.65 ppm) and one 3-proton envelope (3.82 ppm), consistent with fourteen sugar and three ␤-hydroxymethine protons (Fig. 11A). The ␣ CH 2 resonances integrated to eight protons (Fig. 11B), indicating the existence of four acyl chains. Comparison with the ␣ CH 2 resonances of wild-type E. coli lipid A (50) showed the lack of the 2.7 ppm multiplet, previously assigned to the 3Ј acyl chain, and the loss of one ␣ CH 2 resonance near 2.3 ppm. The NMR data are therefore consistent with the removal of the 3Ј-acyloxyacyl moiety by LpxR.
The anomeric H-1 of the proximal sugar (5.45 ppm) and the H-1Ј of the distal sugar (4.54 ppm) (Fig. 11A) served as convenient entry points for full 2D NMR assignments (Table 3), which were based on correlation spectroscopy and total correlation spectroscopy analysis (not shown). The proximal sugar protons (H-1 to H-6) of the 3Ј-O-deacylated lipid A showed similar chemical shifts to those of wild-type E. coli lipid A (50, 51) (Table 3). Significantly, H-3Ј of the 3Ј-O-deacylated lipid A at 3.75 ppm (Table 3) was 1.44 ppm upfield compared with H-3Ј of wild-type E. coli lipid A (5.18 ppm). Other protons of the distal unit were not shifted by more than several tenth of a ppm (Table 3), thus confirming the selective deacylation of the 3Ј-position by LpxR.  . Isopycnic sucrose density gradient centrifugation of LpxR-containing membranes. Inner and outer membranes, isolated from W3110/ pLpxR2, were separated as described previously (45,46), and ϳ0.5-ml fractions were collected. A, protein concentration (open circles) and the 3Ј-Odeacylase activity of LpxR (closed circles) were determined. B, outer membrane phospholipase A activity (open squares) and inner membrane NADH oxidase activity (closed squares) were measured. FIGURE 8. Outer membrane localization of LpxR protein in membranes of E. coli W3110/pLpxR2. The inner and outer membranes from W3110 harboring either the empty vector pWSK29 or pLpxR2 were separated by isopycnic sucrose density ultracentrifugation. Next, 30-g samples of protein from the unfractionated membranes, the outer membranes, and the inner membranes were analyzed by 12% SDS-PAGE and Coomassie Blue staining. The molecular mass standards are indicated to the right of the gel.

DISCUSSION
Lipid A is present in virtually all Gram-negative bacteria and is detected by the innate immune system receptor TLR-4 (9, 10). Nine constitutive enzymes of lipid A biosynthesis have been identified in E. coli (15), and with few exceptions, single copies of the corresponding structural genes are present in all Gram-negative organisms. Despite the conservation of the biosynthetic pathway, which generates the lipid A 1, 4Ј-bis phosphate shown in Fig. 1A, the lipid A moiety of LPS displays significant structural diversity in many important pathogens. For instance, in Salmonella lipid A is covalently modified in response to environmental stimuli that activates the two-component regulatory systems, PhoP/PhoQ (63,64) and PmrA/PmrB (59,65,66) (Fig.  1B). Modification of lipid A is correlated with bacterial virulence, antimicrobial peptide resistance, and attenuation of TLR-4 activation (21-24, 66 -70). The enzymes catalyzing the covalent modifications of Salmonella lipid A (Fig. 1B) have recently been identified and characterized (18 -20, 25-27).
Here we report the discovery of LpxR, an outer membrane enzyme catalyzing a novel deacylation of lipid A at the 3Ј-position. Utilizing an in vitro assay system, we determined that membranes of wild-type S. typhimurium convert Kdo 2 -lipid A to 3Ј-O-deacylated Kdo 2 -lipid A (Fig. 3) in the presence of Ca 2ϩ . That membranes of S. typhimurium could catalyze 3Ј-O-deacylation of Kdo 2 -lipid A was surprising given that 3Ј-O-deacylated lipid A species have not been reported for this organism. However, 3Ј-O-deacylated lipid A species have been isolated from H. pylori (56), Y. enterocolitica (57), Francisella tularensis (71), and Porphyromonas gingivalis (72). Our data demonstrate that a lipid A-3Ј-O-deacylase is present in the outer membrane of S. typhimurium but that it is latent in vivo under all growth conditions examined to date. The previously reported 3-O-deacylase PagL, which is also found in the outer membrane, displays a similar latency (73). Endogenous inhibitors present in stoichiometric quantities relative to LpxR and PagL might account for these findings. We found that it is possible to generate significant amounts of 3Ј-Odeacylated lipid A in S. typhimurium LT2 through pWSK29-mediated overexpression of LpxR (Fig. 10B). We can, thus, hypothesize that an upregulation of lpxR expression in wild type Salmonella would lead to detectable 3Ј-O-deacylated lipid A species.
LpxR-catalyzed 3Ј-O-deacylase activity is absolutely dependent upon the presence of Ca 2ϩ in our in vitro system (Fig. 6). Sr 2ϩ and to a lesser extent Cd 2ϩ could also support LpxR activity. LpxR is the second example of an LPS-modifying enzyme that exhibits a requirement for Ca 2ϩ . Recently, our laboratory reported that EptB, a phosphoethanolamine transferase specific for the outer Kdo moiety of LPS, is a Ca 2ϩ -dependent enzyme (32). LpxR also shares some characteristics with the outer membrane phospholipase A. Outer membrane phospholipase A is an integral Ca 2ϩ -dependent membrane enzyme that participates in secretion of colicins and has been implicated in bacterial virulence (74 -76). The activity of outer membrane phospholipase A is regulated by a Ca 2ϩ -induced, reversible dimerization (74). It is possible that Ca 2ϩ may likewise be required to facilitate LpxR dimerization. Interestingly, PagL, which catalyzes a related hydrolysis reaction (3-O-deacylation), does not require Ca 2ϩ or other divalent cations in vitro (27). PagL and LpxR may utilize different mechanisms to deacylate the lipid A 3 or 3Ј positions, respectively. Studies of purified LpxR should provide insights into how Ca 2ϩ modulates the activity of this enzyme.
Mass spectrometry of the reaction products generated during the in vitro assay revealed that LpxR selectively cleaves the ester linkage at the 3Ј position of lipid A, as no 3-O-deacylated species was detected. NMR spectroscopy of the LpxR-modified lipid A confirmed the selective loss of the 3Ј-acyloxyacyl group (Fig. 11 and Table 3). Our data further demonstrate that LpxR releases the intact 3Ј-acyloxyacyl moiety (Fig. 5) and does not further cleave the acyloxyacyl-linkage. Accordingly, it is reasonable that LpxR does not share any sequence similarity with the acyloxyacyl hydrolase (77,78) of mammalian cells, which removes only the secondary acyl chains of lipid A.
Like the palmitoyltransferase PagP (26) and the 3-O-deacylase PagL (27), LpxR is associated with the outer membrane. Subcellular fractionation studies of membranes derived from LpxR-expressing E. coli W3110 cultures revealed that 3Ј-O-deacylase activity was present mostly in the outer membrane fractions (Fig. 7). SDS-polyacrylamide gel electrophoresis provided further evidence of LpxR localization within the outer membrane (Fig. 8). The overexpressed LpxR protein was missing its type I signal peptide, as shown by N-terminal sequencing of the outer membrane-associated band. LpxR (ϳ32 kDa) is larger than either PagP (ϳ20 kDa) (26) or PagL (ϳ20 kDa) (27). The crystal structure of PagP revealed that this protein is an eight-stranded, inside-out ␤-barrel (79). Topology analysis predicts a 14-stranded ␤-barrel structure for LpxR.
All S. typhimurium lipid A-modifying enzymes, with the exception of LpxO, are transcriptionally regulated by PhoP/PhoQ and/or PmrA/PmrB. Recently, SlyA (80,81), a transcription factor that is under the control of PhoP, was shown to regulate STM1328(lpxR) expression. Microarray analysis of a slyA deletion mutant of S. typhimurium revealed that the expression of STM1328 was reduced 3-fold (82). However, it is unclear whether or not SlyA regulation of lpxR via PhoP is functionally significant at the level of protein synthesis, because membranes isolated from a Salmonella phoP null mutant (83) display similar levels of 3Ј-O-deacylase activity as wild type (data not shown). Other SlyA-regu- lated genes have functions associated with the bacterial cell envelope, and some genes have been implicated in virulence and resistance to antimicrobial peptides (84 -87). SlyA mutants are profoundly attenuated for virulence, unable to replicate in host phagocytes, and sensitive to oxidative stress (80,87,88).
Orthologues of LpxR are found in the genomes of E. coli 0157:H7, Y. enterocolitica, H. pylori, and V. cholerae (Table 2). However, only Y. enterocolitica and H. pylori synthesize 3Ј-Odeacylated lipid A species (56,57). The lipid A of H. pylori is completely deacylated at the 3Ј-position, resulting in a tetraacylated lipid A structure (56). In comparison to the lipid A of the Enterobacteriaceae family, H. pylori lipid A shows very little immunological activity, presumably helping the organism to evade the innate immune response and contributing to its persistence (89). Our work suggests that LpxR may be important during H. pylori infection by reducing the number of lipid A acyl chains, thereby resulting in decreased TLR-4 activation. Removal of the R-3-hydroxymyristate chain at the 3-position of S. typhimurium lipid A by PagL likewise attenuates TLR-4 activation (69).
Surprisingly, we were unable to identify orthologues of LpxR in the genomes of F. tularensis and P. gingivalis even though these bacteria synthesize 3Ј-O-deacylated lipid A species (71,72). It is, therefore, likely that F. tularensis and P. gingivalis contain additional, structurally distinct 3Ј-O-deacylases.
LpxR is the sixth enzyme identified to date that is capable of modifying the lipid A component of Salmonella LPS (Figs. 1B and 2). Construction of lpxR null mutants in Salmonella and other pathogens will be necessary to evaluate the functions of LpxR in pathogenesis and/or outer membrane remodeling. The intriguing mechanisms by which LpxR and PagL are rendered latent in wild type Salmonella also need to be investigated.