Oxygen Requirement for the Biosynthesis of theS-2-Hydroxymyristate Moiety in Salmonella typhimurium Lipid A

Lipid A molecules of certain Gram-negative bacteria, including Salmonella typhimurium andPseudomonas aeruginosa, may contain secondaryS-2-hydroxyacyl chains. S. typhimurium has recently been shown to synthesize itsS-2-hydroxymyristate-modified lipid A in a PhoP/PhoQ-dependent manner, suggesting a possible role for the 2-OH group in pathogenesis. We postulated that 2-hydroxylation might be catalyzed by a novel dioxygenase. Lipid A was extracted from a PhoP-constitutive mutant of S. typhimurium grown in the presence or absence of O2. Under anaerobic conditions, no 2-hydroxymyristate-containing lipid A was formed. Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry of lipid A from cells grown in the presence of 18O2confirmed the direct incorporation of molecular oxygen into 2-hydroxyacyl-modified lipid A. Using several well characterized dioxygenase protein sequences as probes, tBLASTn searches revealed unassigned open reading frame(s) with similarity to mammalian aspartyl/asparaginyl β-hydroxylases in bacteria known to make 2-hydroxyacylated lipid A molecules. The S. typhimuriumaspartyl/asparaginyl β-hydroxylase homologue (designatedlpxO) was cloned into pBluescriptSK and expressed inEscherichia coli K-12, which does not containlpxO. Analysis of the resulting construct revealed thatlpxO expression is sufficient to induce O2-dependent formation of 2-hydroxymyristate-modified lipid A in E. coli. LpxO very likely is a novel Fe2+/α-ketoglutarate-dependent dioxygenase that catalyzes the hydroxylation of lipid A (or of a key precursor). The S. typhimurium lpxO gene encodes a polypeptide of 302 amino acids with predicted membrane-anchoring sequences at both ends. We hypothesize that 2-hydroxymyristate chains released from lipopolysaccharide inside infected macrophages might be converted to 2-hydroxymyristoyl coenzyme A, a well characterized, potent inhibitor of protein N-myristoyl transferase.

infection causes gastroenteritis, but in mice, the outcome is a fatal, typhoid-like sepsis, characterized by dissemination of bacteria into spleen, liver, and blood (1). S. typhimurium initially invade intestinal epithelial cells and M cells of Peyer's patches and then pass into the lymphatic system by colonizing phagocytic cells (1). The bacteria survive and multiply within modified vacuoles of macrophages (2) and gradually induce macrophage apoptosis (3)(4)(5).
The ability of S. typhimurium to adapt to the acidic pH and the low divalent cation concentrations found inside macrophage vacuoles is critical to the infection process (6). The low Mg 2ϩ concentration within phagolysosomes activates the PhoP/PhoQ two-component signal transduction system of S. typhimurium, triggering numerous responses needed for survival and persistence within macrophages (7). Phosphorylation of the transcriptional regulator PhoP (8) by the sensory kinase PhoQ under such conditions results in the activation or repression of as many as 40 S. typhimurium genes (9). The low pH of the phagolysosome, together with PhoP/PhoQ, also activates the PmrA/PmrB two-component system (10). The latter confers resistance to polymyxin and to many cationic antibacterial peptides (11).
Lipopolysaccharide (LPS) 1 is the principal constituent of the outer leaflet of the outer membranes of Gram-negative bacteria (12)(13)(14). In addition to its function as a protective permeability barrier (15), recognition of LPS by mammalian cells activates innate immune responses, including synthesis of cell adhesion proteins in endothelial cells (16) and of proinflammatory cytokines, like tumor necrosis factor-␣ and interleukin-1␤, in monocytes (17,18). Lipid A, the hydrophobic membrane anchor of LPS ( Fig. 1), triggers most of these responses (13,14). The acylated glucosamine disaccharide backbone of lipid A (Fig. 1A) is highly conserved in diverse Gram-negative bacteria (12,19) and is detected by the pattern recognition receptor TLR4 of animal cells (20,21).
Although previously thought to be a relatively static structure, recent studies of S. typhimurium and Pseudomonas aeruginosa have demonstrated that lipid A may be modified in a PhoP/PhoQ-dependent manner under conditions that mimic the phagolysosomal environment (22,23). A palmitate group can be added in acyloxyacyl linkage to the R-3-hydroxymyristate residue at position 2 on S. typhimurium lipid A, and the amount of S-2-hydroxymyristate at position 3Ј can be greatly increased (Fig. 1B) (22). Furthermore, 4-amino-4-deoxy-L-arabinose (L-Ara4N) and phosphoethanolamine (pEtN) groups may be attached to the 4Ј-and/or 1-phosphates (Fig. 1B) once the PmrA/PmrB system is activated (22). While not required for bacterial growth in culture, these modifications may facilitate host-pathogen interactions. For instance, S. typhimurium mutants that are defective in the PhoP/Q-activated gene pagP do not incorporate the palmitate moiety into their lipid A and are more susceptible to the NP-1 defensin (24). Strains that cannot make L-Ara4N are unable to acquire resistance to polymyxin (25). In Escherichia coli K-12, modification of lipid A with L-Ara4N, pEtN, and/or palmitate is seen in polymyxin-resistant mutants (26) or in wild type cells treated with metavanadate (27), but 2-hydroxymyristate is not made (28).
The stimulation of S-2-hydroxymyristate biosynthesis at low Mg 2ϩ concentrations and its absence in PhoP null mutants suggest a function for 2-hydroxylation in pathogenesis (22,29). Although 2-hydroxy fatty acids have been used as taxonomic markers (28,30), the enzymatic pathway for the biosynthesis of the S-2-hydroxymyristate moiety in lipid A of S. typhimurium and other organisms is unknown. Early studies of hydroxyacyl composition, conducted prior to the elucidation of the covalent structure and biosynthesis of lipid A, suggested that the 2-OH (but not the 3-OH) groups of Pseudomonas might be derived from O 2 (31,32).
We now show that the presence of the 2-hydroxymyristate residue in S. typhimurium lipid A is O 2 -dependent and that 18 O 2 is directly incorporated into 2-hydroxymyristate-containing lipid A. We also report the discovery, cloning, and heterologous expression of a novel gene from S. typhimurium, designated lpxO, encoding a 302-amino acid polypeptide with significant sequence similarity to mammalian aspartyl/asparaginyl ␤-hydroxylase, an Fe 2ϩ /␣-ketoglutarate-dependent dioxygenase (33,34). LpxO shares more subtle structural features with other Fe 2ϩ /␣-ketoglutarate-dependent dioxygenases, including deacetoxycephalosporin C synthase from Streptomyces clavuligerus (35). Heterologous expression of S. typhimurium lpxO in E. coli K-12, which does not contain the gene, results in the aerobic biosynthesis of 2-hydroxymyristate-modified lipid A. The lpxO gene is highly homologous to a family of bacterial genes that may be responsible for the biosynthesis of 2-hydroxy fatty acids found in lipid A molecules of certain other Gram-negative pathogens, including Klebsiella pneumonia, P. aeruginosa, Bordetella pertussis, Legionella pneumophila, and all types of Salmonella. The release of 2-hydroxyacyl chains from lipid A within phagolysosomes, known to be catalyzed by mammalian acyloxyacyl hydrolase (36), might allow animal cells to synthesize 2-hydroxyacyl-coenzyme A species, some of which are very potent inhibitors of protein N-myristoylation (37).

EXPERIMENTAL PROCEDURES
Materials-32 P i was purchased from NEN Life Science Products. 18 O 2 (97% isotopic enrichment) was purchased from Isotec. Pyridine, methanol, 88% formic acid, and KH 2 PO 4 were from Mallinckrodt, while chloroform, KCl, and (NH 4 ) 2 SO 4 were purchased from EM Science. MES buffer and sodium fumarate were from Sigma. A dissolved oxygen test kit was purchased from Lamotte. Glass-backed Silica Gel thin layer chromatography plates (0.25 mm) were obtained from Merck. Stainless steel tubing and brass fittings used in the 18 O 2 delivery system were from Supelco.
Bacterial Strains-All bacterial strains used in this study are described in Table I. E. coli XL1-BlueMR was from Stratagene. S. typhimurium (CS14028) was obtained from ATCC. The S. typhimurium phoP c (CS022) strain was kindly provided by Dr. Samuel I. Miller (University of Washington) (38). Except where stated, bacterial shaking cultures were grown at 37°C in LB medium containing 10 g of Tryptone, 5 g of yeast extract, and 10 g of NaCl per liter (39). When needed, the concentration of ampicillin was 100 g/ml.
Isolation of Genomic DNA from S. typhimurium-Genomic DNA was isolated according to the method of Meade et al. (40). Briefly, 4 ml of an overnight culture of S. typhimurium (CS14028) was centrifuged and resuspended in 2 ml of TE buffer (41). A 100-l portion of 2 mg/ml lysozyme stock solution was added, and the mixture was incubated at 37°C for 15 min. Next, 180 l of 10% SDS, 45 l of 20 mg/ml proteinase K solution, and 3 l of 500 g/ml RNase were added. The final mixture was incubated an additional 1 h at 37°C. The solution was then transferred to a glass vial, and 2.5 ml of chloroform/phenol/isoamyl alcohol (25:24:1, v/v/v) was added. The tube was inverted gently 15-20 times and centrifuged briefly at room temperature to separate the phases. The lower phase was removed, and the upper phase was reextracted three times with fresh lower phase. After the third extraction, the upper phase was transferred into a fresh tube and extracted six times (until the interface was clear) with chloroform/isoamyl alcohol (24:1, v/v). Finally, 150 l of 3 M sodium acetate at pH 5.0 and 4 ml of 100% ethanol were added to the final upper phase. The DNA was allowed to precipitate at Ϫ20°C overnight. The precipitate was collected by centrifugation at 4°C for 5 min at 4000 rpm in a Beckman JS4.3 rotor. The pellet was air-dried, redissolved in 300 l of TE buffer, checked for purity based on the A 280 /A 260 ratio, and stored at Ϫ20°C.
Cloning of lpxO from S. typhimurium Genomic DNA-Primers corresponding to the 5Ј-and 3Ј-ends of the lpxO open reading frame coding for the putative lipid A 2-hydroxylase were designed as follows: for the 5Ј-end, BHyd5 (5Ј-CCGCCGGAATTCCATATGTTCGCAGCAATCATT-ATCGG-3Ј); for the 3Ј-end, BHyd3 (5Ј-CCGCTCGAGTCAGAGGAGGC-TGAAAAGGAT-3Ј). The lpxO open reading frame was amplified by polymerase chain reaction from genomic DNA under the following conditions: 50-l total reaction volume, 200 nM each primer, 200 M dNTPs, 1.5 g of genomic DNA, 4 mM MgCl 2 , 2.5 units of Pfu DNA polymerase (Stratagene) with buffers supplied by the manufacturer. The temperature program was as follows: 94°C for 7 min, a cycle of 45 s each at 94, 50, and 72°C repeated 25 times, followed by 7 min at 72°C. The polymerase chain reaction product and the pBluescriptSK vector DNA were then digested with EcoRI and XhoI, gel-purified with the QIAEX-II kit (Qiagen), and ligated together with T4 DNA ligase (Life Technologies, Inc.) to form pHSG1, which was transformed into CaCl 2competent cells of E. coli strain XL1-BlueMR. Plasmid DNA from the resulting colonies (selected on ampicillin) was prepared from overnight cultures in LB with 100 g/ml ampicillin (QIAQUICK Spin Miniprep kit, Qiagen) and screened by restriction enzyme analysis, using EcoRI and XhoI. The nucleotide sequence of the final construct was confirmed from four candidate clones at the Duke University Nucleotide Sequencing Core Facility using the T7 promoter and T3 promoter primers to the pBluescriptSK plasmid (Stratagene). 32 P i Labeling of E. coli and S. typhimurium Cultures-Radioactive labeling of lipid A species was performed as described previously (27) at 5 Ci of 32 P i /ml of culture medium. E. coli was grown on LB broth (39), and S. typhimurium was grown on minimal G56F (see below) with 10 mM MgCl 2 . After extraction and 100°C hydrolysis at pH 4.5 (27), 2000 cpm of the released lipid A species were spotted onto a 10 ϫ 20-cm silica gel TLC plate, which was developed in chloroform/pyridine/88% formic To exclude oxygen, sterile 50-ml polypropylene centrifuge tubes were filled to the top with culture medium and inoculated using a 1:500 dilution of an overnight culture of S. typhimurium phoP c CS022 grown on LB broth. The tubes were tightly capped and incubated at 37°C without shaking. The concentration of dissolved oxygen was determined at 1-h intervals using the azide-modified Winkler titration (Lamotte) (43). After 1 h of growth under these conditions, no dissolved oxygen was detected. For lipid A preparations from anaerobic cells, four 50-ml culture tubes were used. Aerobic cultures were grown in 200 ml of the same medium in a 1-liter culture flask at 37°C with shaking at 200 rpm.
Isolation of the Lipid A 1,4Ј-Bisphosphate Components of CS022 by Chromatography on DEAE-cellulose-In our protocol, 200 ml of bacteria grown either aerobically or anaerobically to A 600 ϳ 0.6 were harvested by centrifugation at 4°C. Each cell pellet was resuspended in 80 ml of phosphate-buffered saline to which was added 100 ml of chloroform and 200 ml of methanol to make a single phase Bligh/Dyer mixture (44). After extraction of the glycerophospholipids and centrifugation, crude lipid A components were released from the Bligh/Dyer insoluble pellet by hydrolysis at pH 4.5 in the presence of SDS (27,45). To purify the major lipid A 1,4Ј-bisphosphate species (lacking the L-Ara4N or pEtN substituents), the crude lipid A samples from CS022 were fractionated based on charge using DEAE-cellulose column chromatography (27,45). Typically, the crude lipid A from a 200-ml culture was redissolved in 5 ml of chloroform/methanol/water (2:3:1, v/v/v) with the aid of a bath sonicator. A 1-ml DEAE-cellulose column in the acetate form (Whatman DE52) (46) was prepared and washed with 20 ml of chloroform/methanol/water (2:3:1, v/v/v) prior to loading the sample. After application of the sample at the natural flow rate, the column was eluted with increasing concentrations of ammonium acetate (45). The major, unmodified lipid A 1,4Ј-bisphosphate components emerged with chloroform/methanol/240 mM aqueous ammonium acetate (2:3:1, v/v/v) (45). The desired components were detected by thin layer chromatography in the solvent chloroform/pyridine/88% formic acid/water (50:50: 16:5, v/v/v/v) and charring either with 10% sulfuric acid in ethanol or with ethanol/p-anisaldehyde/H 2 SO 4 /acetic acid (89:2.5:4:1, v/v/v/v) (45,47). Fractions containing the lipid A 1,4Ј-bisphosphate components were pooled and converted to two-phase Bligh/Dyer systems by the addition of the appropriate amounts of chloroform and water. The lower phases were collected, dried under nitrogen, and stored at Ϫ80°C.
Purification of Unmodified and 2-Hydroxymyristate-modified Lipid A from E. coli Cells Expressing S. typhimurium lpxO-Crude lipid A species from 1 liter of E. coli XL1-BlueMR(pHSG1) grown on LB medium to A 600 ϳ 2.0 were extracted following pH 4.5 hydrolysis as described above. Total lipid A 1,4Ј-bisphosphates were first prepared by chromatography on a 2-ml DEAE-cellulose column, equilibrated and eluted proportionally as described above. To separate the lipid A species with the S-2-hydroxymyristate residue at the 3Ј-acyloxyacyl position from the unmodified lipid A, half of the DEAE-purified sample was redissolved in ϳ100 l of chloroform/methanol (4:1, v/v) and spotted in 2-l portions along the origins of two 20 ϫ 20-cm silica gel TLC plates. The plates were then developed in chloroform/pyridine/88% formic acid/ water (50:50:16:5, v/v/v/v). As the plates were drying at room temperature, two closely migrating lipid A species were resolved and could be seen transiently as distinct white zones on the plates. The more slowly migrating component contains the 2-hydroxymyristate substituent. Both of the zones were marked with a pencil. The plates were dried for another ϳ20 min to remove solvents. After moistening the surface with a water spray, the silica within the zones was scraped off with a scalpel, and the chips were collected in separate glass tubes. The chips in each tube were extracted four times with 6-ml portions of chloroform/methanol/50 mM aqueous ammonium acetate adjusted to pH 1.5 with HCl (1:2:0.8, v/v/v). Following each extraction, large chips were removed by low speed centrifugation, and the supernatants were passed through a column of glass wool in a Pasteur pipette. The filtered samples were converted into two-phase Bligh/Dyer systems, consisting of chloroform/ methanol/water (2:2:1.8, v/v/v), by the addition of appropriate amounts of water and chloroform. The pooled lower phases were neutralized with 24 drops of high pressure liquid chromatography grade pyridine, cleared by adding 30 drops of methanol, and evaporated under nitrogen.
To remove any remaining fine silica particles, each of the lipid A components was redissolved in 5 ml of chloroform/methanol/water (2: 3:1, v/v/v) and purified over another 2-ml DEAE-cellulose column, as described above. The intact lipids were analyzed by TLC and matrixassisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry. Fatty acid compositions were determined by preparation of methyl esters, which were resolved by gas chromatography and detected with flame ionization. 18 O 2 Labeling of Lipid A-The procedure for growing bacteria in an 18 O 2 atmosphere was adapted from that described for Pseudomonas ovalis by Kawahara and co-workers (32). A sealed 1-liter culture flask, containing 200 ml of LB medium inoculated by 1:100 dilution from an overnight culture of the PhoP-constitutive S. typhimurium mutant CS022, was evacuated to 27 inches of mercury, as determined on a Marsh Instrument Co. model 1305-12 in line vacuum gauge. The flask was flushed with nitrogen and reevacuated twice to eliminate residual oxygen in the culture medium. Following a third application of the vacuum to 27 inches of mercury, the headspace was brought to 20 inches Hg with 18 O 2 and then to ambient pressure with nitrogen (resulting in an atmosphere of approximately 26% 18 O 2 at 97% isotopic enrichment). The sealed flask was removed from the apparatus and shaken at 37°C and 220 rpm for 4 h until the culture reached an A 600 of ϳ0.8, after which the lipid A 1,4Ј-bisphosphate species were isolated as described above. A control culture was prepared by sealing a flask that was not evacuated immediately after inoculation, so that the bacteria could grow under otherwise identical conditions with ambient air.
Fatty Acid Compositions of Purified Lipid A Components-Purified lipid A samples were sent to Microbial ID, Inc. (Newark, DE) for fatty acid analysis according to standard procedures. In brief, samples were esterified in 2 M methanolic HCl, extracted into petroleum ether, and concentrated. Next, 2-l samples were injected with a 100:1 split onto a gas chromatography column. Fatty acid methyl esters were detected by flame ionization, and their elution times were compared with a mixture of known fatty acid methyl ester standards run prior to analysis. Retention times were analyzed by the Sherlock software package. Peaks appearing outside of a defined retention time window were labeled as unknowns.
MALDI-TOF Mass Spectrometry-Spectra were acquired in the negative ion linear mode by using a Kratos Analytical (Manchester, United Kingdom) MALDI-TOF mass spectrometer equipped with a 337-nm nitrogen laser, a 20-kV extraction voltage, and time-delayed extraction. Each spectrum was the average of 50 shots. The matrix was a mixture of saturated 6-aza-2-thiothymine in 50% acetonitrile and 10% tribasic ammonium citrate (9:1, v/v). The lipid A samples were dissolved in a mixture of chloroform/methanol (4:1 v/v) and mixed with the matrix on a slide. The sample mixtures were allowed to dry at room temperature prior to mass analysis. Hexa-acylated lipid A 1,4Ј-bisphosphate from wild-type E. coli (purchased from Sigma) was used as an external standard for calibration. (38) and efficiently incorporates L-Ara4N, pEtN, palmitate, and 2-hydroxymyristate moieties into its lipid A (Fig. 1B), even in the presence of 10 mM MgCl 2 in the medium. To test the effects of O 2 on the formation of 2-hydroxymyristate-containing lipid A, CS022 cells were grown either aerobically or anaerobically in 200 ml of G56F medium. The cells reached an A 600 of ϳ0.8 after 6 h of anaerobic growth. Under these conditions, approximately 90% of the cell mass was generated after all measurable O 2 had been depleted from the medium (data not shown).

Effects of O 2 on Growth and Biosynthesis of Lipid A Species
Lipid A was isolated from solvent-extracted cells by hydrolysis at 100°C in SDS at pH 4.5. To simplify the interpretation of the mass spectra, the crude lipid A released from the cells was first fractionated on DEAE-cellulose columns (data not shown). Only those lipid A species containing unsubstituted 1and 4Ј-bisphosphate moieties, which elute with chloroform/ methanol/240 mM aqueous ammonium acetate (2:3:1, v/v/v) (27), were analyzed further. Two major and one minor lipid A 1,4Ј-bisphosphate species were resolved from cells grown in the presence of O 2 by TLC (Fig. 2, lane 1). The slowly migrating major component (A) was not seen when lipid A was prepared from CS022 cells grown without O 2 (Fig. 2, lane 2).
Mass Spectrometry of the Lipid A 1,4Ј-Bisphosphates Isolated from S. typhimurium CS022 Grown with or without O 2 -Negative mode MALDI-TOF mass spectrometry of the lipid A 1,4Јbisphosphates, isolated by DEAE chromatography from aerobically grown cells, showed major peaks at m/z 1796.9 and 1813.0 (Fig. 3A). The peak at m/z 1796.9 is seen in lipid A of both E. coli and S. typhimurium. It corresponds to [M Ϫ H] Ϫ of a hexa-acylated lipid A 1,4Ј-bisphosphate bearing laurate and myristate as secondary acyl chains at positions 2Ј and 3Ј, re-spectively (Fig. 1A). The peak at m/z 1813.0 is 16.1 atomic mass units higher, and it presumably corresponds to [M Ϫ H] Ϫ of the species bearing a secondary 2-hydroxymyristoyl chain in place of myristate at position 3Ј, as is seen in PhoP-constitutive S. typhimurium (Fig. 1B). In addition, minor species are observed  (27,76). Lipid A species are released from cells (or purified LPS) by hydrolysis at pH 4.5 in the presence of SDS at 100°C (27,76). The same 1,4Ј-bisphosphate is also a major component of S. typhimurium lipid A, but it may be modified with additional substituents (see below), the biosynthesis and attachment of which are regulated by the PhoP/PhoQ (red) and PmrA/PmrB (blue) two-component systems (22,26). The pyrophosphate variant is seen only in PhoP -S. typhimurium grown at high Mg 2ϩ concentrations (Z. Zhou and C. R. H. Raetz, unpublished observations). B, S. typhimurium lipid A becomes heavily modified with covalently attached pEtN and/or L-Ara4N moieties, as indicated, when grown at low Mg 2ϩ concentrations (ϳ10 M), a condition that resembles the intraphagosomal environment and activates PhoP/PhoQ (22). In addition, the lipid A is modified by incorporation of an extra palmitoyl chain and/or a 2-hydroxymyristoyl group (in place of myristate) at the 2-and 3Ј-acyloxyacyl positions, respectively (22,24). Different combinations of these substituents account for the 16 or more lipid A molecular species seen in PhoP-constitutive S. Positive mode MALDI-TOF mass spectrometry is consistent with the results shown in Fig. 3 and confirms that the extra oxygen (when present) is located on the distal glucosamine unit (data not shown). Gas chromatography of the fatty acid methyl esters prepared from the DEAE-cellulose-purified lipid A 1,4Јbisphosphates confirmed the presence of 2-hydroxymyristate in the aerobically grown cells and its absence in anaerobic cells (data not shown).
18 O 2 Labeling of the S. typhimurium Lipid A 1,4Ј-Bisphosphates-To confirm the hypothesis that molecular oxygen is directly incorporated into the 2-hydroxymyristoyl chain of S. typhimurium lipid A, CS022 cells were grown in the presence of 18 O 2 . In a procedure adapted from an earlier study with P. ovalis (32), 200-ml LB broth cultures of CS022 were grown at 37°C in sealed 1-liter Erlenmeyer flasks with either ambient air or a mixture of 26% 18 O 2 and 74% N 2 in the headspace. The cells were harvested at A 600 ϳ 0.8. The lipid A 1,4Ј-bisphosphates were purified by DEAE chromatography and analyzed by MALDI-TOF mass spectrometry. In the negative mode, the mass difference (⌬) between the hydroxylated and nonhydroxylated lipid A species of the culture grown with ambient air (Fig.  4A) was 15.9 atomic mass units for the hexa-acylated and 16.1 atomic mass units for the hepta-acylated subtypes, respectively. When grown under 18 O 2 , the difference (⌬) increased to 17.7 atomic mass units and 17.9 atomic mass units, respectively (Fig. 4B), as expected if molecular O 2 is incorporated into the 2-OH group.
A Homologue of Bovine Aspartyl/Asparaginyl ␤-Hydroxylase in S. typhimurium and Some Other Gram-negative Bacteria-Several cloned, well characterized dioxygenases, representing different mechanistic families including cytochrome P450-, dinuclear non-heme iron-, and Fe 2ϩ /␣-ketoglutarate-dependent enzymes, were used as probes in tBLASTn searches (48) against the available microbial genome data bases. The rationale was to identify a unique class of dioxygenase homologues in those bacteria that synthesize 2-hydroxyacyl modified lipid A (28, 31, 49 -52). A tBLASTn search with the protein sequence of bovine aspartyl/asparaginyl ␤-hydroxylase (33, 34), a type of Fe 2ϩ /␣-ketoglutarate-dependent dioxygenase, as the probe fit this criterion. When compared by tBLASTn with the S. typhimurium genome, the catalytic domain of bovine aspartyl/asparaginyl ␤-hydroxylase revealed significant similarity (E value of ϳ10 Ϫ10 ) to a previously unidentified open reading frame on the S. typhimurium chromosome, mapping next to fdhF at 92.8 min (53). This novel gene (designated lpxO) codes for a putative 302-amino acid polypeptide (Fig. 5) with hydrophobic N-and C-terminal sequences (Fig. 5, shading). The predicted LpxO amino acid sequence shares several important features with the aspartyl/asparaginyl ␤-hydroxylase catalytic domain, including four conserved histidine side chains (red) and several aspartate and glutamate residues (blue). Histidine 675 of bovine aspartyl/asparaginyl ␤-hydroxylase (Fig. 5, arrow) has in fact been shown to be an iron ligand (34) and is conserved. LpxO also contains a His-X-Asp-X ϳ50 -His motif (Fig. 5, yellow) (54), resembling the Fe 2ϩ binding site deacetoxycephalosporin C synthase (35). The latter is the only Fe 2ϩ /␣-ketoglutaratedependent dioxygenase for which a high resolution crystal structure is available (35). The His-X-Asp-X ϳ50 -His motif is also present in TfdA (55) and TauD (56), two additional bacterial Fe 2ϩ /␣-ketoglutarate-dependent dioxygenases. The GC content of lpxO and its flanking DNA does not deviate from the S. typhimurium average of 52%. It is therefore unlikely that S. typhimurium lpxO resides within a pathogenicity island. As noted above, however, homologues of aspartyl/ asparaginyl ␤-hydroxylase are seen in other bacteria that synthesize 2-hydroxyacyl chains (Table II) and L. pneumophila, all of which make 2-hydroxyacyl-containing lipid A species (28, 31, 49 -52). P. aeruginosa contains two LpxO homologues, consistent with the fact that its lipid A contains two distinct 2-hydroxylaurate chains (50). The bacterial LpxO homologues are all closely related to each other (E values from ϳ10 Ϫ178 to 10 Ϫ23 when probed with S. typhimurium LpxO), and are all about the same length. LpxO homologues are not present in Gram-positive bacteria or in Gramnegatives that do not make 2-hydroxyacylated lipid A. Despite the presence of the His-X-Asp-X ϳ50 -His motif (54,57) in LpxO, which was noted by visual inspection (Fig. 5), PSI-BLAST searches (58) comparing S. typhimurium LpxO with the complete nonredundant data base failed to reveal significant similarity to other Fe 2ϩ /␣-ketoglutarate-dependent dioxygenase besides the mammalian aspartyl/asparaginyl ␤-hydroxylases.
Cloning of S. typhimurium lpxO and Expression in E. coli K12-The lpxO gene was amplified by polymerase chain reaction from genomic DNA prepared from wild type S. typhimurium. The polymerase chain reaction product, prepared with primers containing appropriate restriction sites, was digested with EcoRI and XhoI and ligated into pBluescriptSK so that expression of lpxO would be driven by the lac promoter. The resulting hybrid plasmid, pHSG1, was transformed into competent E. coli XL1-BlueMR. The correct coding sequence of the plasmid was confirmed by nucleotide sequencing in both directions.
Lipid A Hydroxylation in E. coli Cells Expressing S. typhimurium lpxO-Although not optimized for maximal expression of LpxO activity, the hybrid plasmid pHSG1 confers upon E. coli the ability to synthesize lipid A molecules containing 2-hydroxymyristate. 32 P-Labeled lipid A species purified from XL1-  (48,58), using the predicted S. typhimurium LpxO protein sequence as the probe. S. paratyphi and Salmonella enteritidis contain single homologues (not shown) that are essentially the same as the one in S. typhi.

FIG. 5. Bioinformatic identification of LpxO, a putative hydroxylase involved in lipid A biosynthesis, in S. typhimurium.
To find LpxO, tBLASTn searches (48,58) against the available microbial genomes were conducted using well characterized protein sequences of various cloned dioxygenases as probes. Only one enzyme, the aspartyl/asparaginyl ␤-hydroxylase of animal cells, a type of Fe 2ϩ /␣-ketoglutarate-dependent dioxygenase, yielded a distinct family of homologues restricted to those bacteria known to make 2-hydroxyacylated lipid A species. LpxO of S. typhimurium shown above displayed an E value of ϳ10 Ϫ10 against the catalytic domain of bovine aspartyl/asparaginyl ␤-hydroxylase. The proposed histidine 675 iron ligand of the bovine aspartyl/asparaginyl ␤-hydroxylase is indicated by the arrow and is present at position 155 in S. typhimurium LpxO and in the other bacterial homologues (not shown). Red letters show other conserved histidine residues, while blue letters indicate conserved aspartate and glutamate residues, some of which might function as iron ligands. Hydrophobic segments are shaded in gray, whereas the putative His-X-Asp-X ϳ50 -His motif (54,57), also seen in the other bacterial LpxO homologues, is shaded in yellow.
BlueMR containing either no vector, pBluescriptSK, or pHSG1 were analyzed by TLC and were compared with similarly radiolabeled lipid A species from S. typhimurium wild type or CS022 (Fig. 6). Lipid A from wild-type or vector control E. coli (lanes 1 and 2) exhibited only those spots characteristic of wild type E. coli K-12 grown under normal conditions, i.e. mostly the hexa-acylated lipid A 1,4Ј-bisphosphate with some 1-pyrophosphate and small amounts of 4Ј monophosphate (a byproduct of hydrolysis). E. coli containing pHSG1 (lane 3) showed at least two additional species, indicated by the arrows on the right, that migrated more slowly than the lipid A species in the vector control. Interestingly, these pHSG1-dependent lipid A variants migrated at the same R-factors as certain bands in crude S. typhimurium lipid A prepared by pH 4.5 hydrolysis (Fig. 6,  lanes 4 and 5).
The MALDI-TOF mass spectrometry of the DEAE-cellulosepurified lipid A 1,4Ј-bisphosphates from E. coli XL1-Blue containing pBluescriptSK shows a single major peak at m/z 1796.7 (Fig. 7A), typical of wild-type cells. Expression of the Salmonella lpxO on pHSG1, however, confers upon E. coli the ability to add an additional OH group to lipid A, as judged by the appearance of the peak at m/z 1813.7 (Fig. 7B).
Purification and Fatty Acid Composition of Nonhydroxylated and Hydroxylated Lipid A from E. coli Expressing S. typhimurium lpxO-For conclusive confirmation that lpxO encodes a protein that makes the 2-hydroxymyristate moiety of S. typhimurium lipid A, hydroxylated and nonhydroxylated lipid A species from E. coli cells expressing lpxO were separated from each other by preparative TLC and anion exchange chromatography, as described under "Experimental Procedures." The TLC plate of the purified lipid A species is shown in Fig. 8A. MALDI-TOF mass spectra of these species (Fig. 8, B and C) confirm their purity and identity. Gas chromatography of the corresponding fatty acid methyl esters (Fig. 8, insets) unambiguously demonstrates the presence of 2-hydroxymyristate only in the slowly migrating hydroxylated lipid A sample with [M Ϫ H] Ϫ at m/z 1813.5, derived from E. coli expressing pHSG1. The appearance of the 2-hydroxymyristate group is accompanied by a corresponding decline in myristate content. The analysis excludes the formal possibility that LpxO incorporates a hydroxyl group at some site other than carbon 2 of the myristate chain of E. coli lipid A.

DISCUSSION
Although 2-hydroxy fatty acids have long been recognized as components of lipid A molecules in a select subset of Gramnegative bacteria (28, 31, 49 -52), little is known about their biosynthesis and function (32). The recent discovery that formation of the 2-hydroxymyristate moiety in S. typhimurium lipid A is under the control of the PhoP/PhoQ two-component regulatory system raises the intriguing possibility that this substituent might play a role in pathogenesis (22). Although other lipid A modifying groups, such as L-Ara4N, pEtN, and C16:0 moieties, are similarly regulated (22) and are common to both E. coli and S. typhimurium (27), the 2-hydroxymyristate substituent is seen only in S. typhimurium. Mutants defective in PhoP/PhoQ do not make any of these substituents (22). Such strains are much less virulent than wild type and are susceptible to cationic anti-microbial peptides (24,25). The biological functions of each of the lipid A modifying groups are difficult to assess, however, because the enzymes that catalyze their formation have not yet been identified. We have now established that the formation of the 2-hydroxymyristate moiety in S. typhimurium lipid A is oxygen-dependent, as is the case for 2-hydroxylaurate in P. ovalis (32). When S. typhimurium cells are grown under anaerobic condi-tions, myristate is incorporated into lipid A in place of 2-hydroxymyristate. The 2-hydroxymyristate residue is probably generated by hydroxylation of the myristate chain after its incorporation into nascent lipid A by MsbB (13, 59). However, FIG. 9. A "Trojan horse" model for release of the 2-hydroxymyristate moiety and inhibition of host cell signaling. The diagram represents a macrophage or epithelial cell harboring S. typhimurium in its phagolysosome. Bacteria are represented as red ovals. AOAH, acyloxyacyl hydrolase; Protein NMT, myristoyl-coenzyme A/protein N-myristoyltransferase. The normal pathway of protein myristoylation is shown with black arrows, while the effect of the S. typhimurium Trojan horse is shown with red arrows. Given that the 2-hydroxymyristate substituent can probably be released in animal cells from S. typhimurium lipid A by the action of acyloxyacyl hydrolase (36), it is plausible that it might be converted to 2-hydroxymyristoyl coenzyme A, a very potent inhibitor of protein N-myristoyl transferase (37). The consequence might be suppression of host cell signaling functions, permitting a more prolonged survival of the bacteria inside the host cell. we cannot yet exclude the alternative possibility that 2-hydroxymyristate is formed while still attached to acyl carrier protein, i.e. by hydroxylation of myristoyl-acyl carrier protein. Development of an in vitro assay should resolve this question.
We used a bioinformatic approach to identify the putative hydroxylase responsible for the formation of the 2-hydroxymyristoyl moiety of S. typhimurium lipid A. We first probed all of the available microbial genomic data bases for uncharacterized open reading frames with sequence similarity to well characterized, cloned dioxygenases, such as selected P450-, dinuclear iron-, and Fe 2ϩ /␣-ketoglutarate-dependent hydroxylases. Those dioxygenases with homologues in bacteria known to make 2-hydroxyacylated lipid A species were considered further. Of the various dioxygenases used to search the microbial data bases, only the mammalian aspartyl/asparaginyl ␤-hydroxylase yielded the desired pattern. This enzyme hydroxylates certain aspartyl and asparaginyl residues in clotting factors and other proteins (60,61). Aspartyl/asparaginyl ␤-hydroxylase (33,34,60,61) belongs to the larger family of Fe 2ϩ /␣-ketoglutarate-dependent dioxygenases, which include prolyl and lysyl hydroxylases (62), deacetoxycephalosporin C synthase (35), taurine hydroxylase (56), and thymine hydroxylase (63,64). In a single tBLASTn search (48,58), however, only the aspartyl/asparaginyl ␤-hydroxylase (33, 34) produced significant matches with the relevant bacterial species (K. pneumonia, P. aeruginosa, P. putida, B. bronchiseptica, L. pneumophila, and all types of Salmonella) ( Table II).
The bovine aspartyl/asparaginyl ␤-hydroxylase and S. typhimurium LpxO share several important conserved amino acid residues (Fig. 5). Of these, His 155 of the S. typhimurium LpxO is of special interest, since it corresponds to His 675 of bovine aspartyl/asparaginyl ␤-hydroxylase (Fig. 5), a residue identified by site-directed mutagenesis to be critical for iron binding and catalysis (34). The His-X-Asp-X ϳ50 -His motif (Fig. 5, yellow  shading), a structural feature of many non-heme iron active sites (54,57), appears to be present as well in S. typhimurium LpxO (Figs. 5) and all other bacterial LpxO homologues, with the possible exceptions of Legionella LpxO. Its presence in S. typhimurium LpxO was recognized by visual inspection. This motif is well characterized in the crystal structure of deacetoxycephalosporin C synthase, in which the two His residues and one Asp residue of the motif function as iron ligands (35). The crystal structure of the bovine aspartyl/asparaginyl ␤-hydroxylase itself has not yet been solved.
The ability of the S. typhimurium lpxO gene to enable the biosynthesis of 2-hydroxymyristate-modified lipid A in E. coli (Figs. 7 and 8) strongly suggests that lpxO is the structural gene for a novel membrane bound, ␣-ketoglutarate-dependent hydroxylase. The LpxO active site may face the cytoplasm. However, in view of the ability of S. typhimurium to secrete large amounts of ␣-ketoglutarate (Ͼ100 M) into the medium under conditions of iron stress, the possibility of a periplasmic or outer membrane localization for LpxO cannot be excluded (65).
Considerable effort has been devoted in recent years to the discovery of genes expressed in vivo during infection. Studies with Salmonella have utilized both signature-tagged mutagenesis and in vivo expression technology (66 -69). It is interesting that no S. typhimurium genes expressed in vivo have been mapped to minute 92.8. Furthermore, despite the dependence of 2-hydroxymyristate biosynthesis on the PhoP/Q system (22), no genes mapping to lpxO have been found in searches for PhoP/Q-regulated genes (70,71). It may be that these genetic screens were not fully saturated.
Despite the important roles that lipid A plays in host cell signaling and the fact that modified lipid A structures can elicit different host responses (13,14), the role of 2-hydroxylation of lipid A has not been investigated. We propose a "Trojan horse" hypothesis for 2-hydroxymyristate function during S. typhimurium infections (Fig. 9). Neutrophils and monocytes are known to deacylate purified LPS (72,73) and can even remove the secondary acyl chains from LPS in whole bacterial cells (74). The relevant acyloxyacyl hydrolase has been well characterized and is thought to detoxify LPS from diverse Gramnegative bacteria (36,72,73). During S. typhimurium infections, acyloxyacyl hydrolase would be expected to release 2-hydroxymyristate from LPS, which might allow mammalian cells to synthesize 2-hydroxymyristoyl coenzyme A, a potent inhibitor (K i ϳ 40 nM) of myristoyl-coenzyme A/protein Nmyristoyltransferase (37). Inhibition of the latter might result in mislocalization of numerous proteins that utilize myristoyl chains as membrane anchors (75), possibly interfering with signal transduction and/or vesicle trafficking. Our hypothesis is contingent upon the transport of the released 2-hydroxymyristate from the phagolysosome to the cytosol. Interference with protein myristoylation could provide S. typhimurium with a way to modify the intracellular environment and facilitate the infection process.
The cloning of the lpxO gene now provides the means to purify large quantities of 2-hydroxymyristate-modified lipid A and to construct mutants of S. typhimurium and other Gram-negative organisms lacking lpxO. Preliminary characterization of such mutants in our laboratory indicates that 2-hydroxymyristate-modified lipid A is absent in S. typhimurium mutants lacking lpxO. 2 The discovery and characterization of lpxO should advance our understanding of the enzymology of lipid A modification and its involvement in pathogenesis.