A Chlamydia trachomatis UDP- N -Acetylglucosamine Acyltransferase Selective for Myristoyl-Acyl Carrier Protein EXPRESSION IN ESCHERICHIA COLI AND FORMATION OF HYBRID LIPID A SPECIES*

Chlamydia trachomatis lipid A is unusual in that it is acylated with myristoyl chains at the glucosamine 3 and 3 * positions. We have cloned and expressed the gene encoding UDP- N -acetylglucosamine 3- O -acyltransferase of C. trachomatis ( CtlpxA ), the first enzyme of lipid A biosynthesis. C. trachomatis LpxA displays ; 20-fold selectivity for myristoyl-ACP over R/S -3-hydroxymyris-toyl-ACP under standard assay conditions, consistent with the proposed structure of C. trachomatis lipid A. CtLpxA is the first reported UDP- N -acetylglucosamine acyltransferase that prefers a non-hydroxylated acyl-ACP to a hydroxyacyl-ACP. When CtlpxA was expressed in RO138, a temperature-sensitive lpxA mutant of Escherichia coli , five new hybrid lipid A species were made in vivo afte r 2 h at 42 °C, inplace of Escherichia coli lipid A. These compounds were purified and analyzed by ma-trix-assisted laser desorption ionization/time of flight mass spectrometry. In each case, a myristoyl chain replaced one or both of the ester

bacteria display high selectivity for the lengths and stereochemistry of their R-3-hydroxyacyl-ACP 1 substrates, and are the primary determinants of the fatty acyl chains that are incorporated at the 3 and 3Ј positions of lipid A (9,(11)(12)(13)(14). The crystal structure of E. coli LpxA has been solved; it is a homotrimeric enzyme with an unusual left-handed parallel ␤-helix secondary structure (15). The location of the active site has been surmised based on site-directed mutagenesis (16), but crystal structures of enzyme-substrate complexes are not yet available.
In most bacteria examined to date, only R-3-hydroxyacyl chains are attached to the glucosamine disaccharide backbone of lipid A (1,3,17). However, the 3 and 3Ј acyl chains of lipid A are not hydroxylated in Chlamydia trachomatis (Fig. 2), an intracellular Gram-negative pathogen with a biphasic life cycle and a chronic inflammatory pathology (3,18,19). Ocular infections with C. trachomatis are the leading worldwide cause of preventable blindness (20,21). Urogenital infections with C. trachomatis are a major cause of female infertility (20,21). C. trachomatis also causes neonatal pneumonia, and may have a role in the etiology of certain types of arthritis (20,21). During the course of an infection, C. trachomatis exchanges some of its membrane lipids, including lipopolysaccharide, with those of its mammalian host cell (22) and vice versa (23)(24)(25). Understanding the biosynthesis of C. trachomatis lipid A may provide insights into the life cycle of this pathogen.
The C. trachomatis genome contains single copy homologues of most of the genes required for lipid A biosynthesis in E. coli, including CtlpxA, which encodes a protein of 280 amino acid residues with 42% identity and 50% similarity to the 262 amino acids of E. coli LpxA (26). Given that C. trachomatis makes a penta-acylated lipid A molecule (Fig. 2), a homologue of the E. coli msbB gene ( Fig. 1) (27,28) is missing in C. trachomatis (26).
We now show that the presence of non-hydroxylated acyl chains at the 3 and 3Ј positions in C. trachomatis lipid A (Fig.  2) is explained by the unusual substrate specificity of C. trachomatis LpxA, which is highly selective for myristoyl-ACP. Expression of C. trachomatis lpxA in E. coli mutant RO138, which is defective in its own lpxA gene (14,29) at 42°C, results in the production of structurally novel lipid A species. Following many generations of growth at 42°C, all of the lipid A molecules synthesized in this construct contain myristate residues at positions 3 and 3Ј. However, when analyzed 2 h after a shift from 30 to 42°C, residual ester-linked 3-hydroxymyristate moieties are present only at position 3Ј. Genetically engineered bacteria containing modified lipid A structures may prove useful for the production of novel vaccines and adjuvants.

EXPERIMENTAL PROCEDURES
Materials-Agar, tryptone, and yeast extract were from Difco. Acetic acid, chloroform, methanol, and sodium hydroxide were from Mallinckrodt Chemical Works. Acyl carrier protein, antibiotics, myristic acid, (R,S)-3-hydroxymyristic acid, (R,S)-3-hydroxypalmitic acid, and glucosamine 1-phosphate were from Sigma. Silica Gel 60 thin layer plates (0.25 mm) were purchased from E. Merck, Darmstadt, Germany. Restriction enzymes, pfu polymerase, and cloning vectors pBluescript II SK(ϩ) and pUC18 were obtained from Stratagene. The cloning vector pET23cϩ was purchased from Novagen. T4 ligase and preparative grade phenol/chloroform/isoamyl alcohol (25 coli LpxA selectively acylates the glucosamine 3 position of UDP-GlcNAc with (R)-3-hydroxymyristate, as indicated in red, eventually resulting in the incorporation of (R)-3-hydroxymyristate residues at positions 3 and 3Ј of E. coli lipid A (1, 51). The lengths of the acyl chains are indicated by the numbers at the bottoms of selected structures. About one-third of the E. coli K12 lipid A species are modified with a diphosphate unit (not shown) at position 1 in place of the more abundant monophosphate substituent (40). The Kdo disaccharide of the inner core of E. coli lipopolysaccharide is required for the completion of lipid A acylation by HtrB and MsbB. was cloned either into pET23cϩ (pTO1) or into pBluescript II SKϩ (pTO5) were described previously (16,32). Cells were generally grown at 30, 37, or 42°C, as indicated, in LB medium (10 g/liter Tryptone, 10 g/liter NaCl, and 5 g/liter yeast extract adjusted to pH 7.4 with NaOH) (33), or on LB plates containing 1% (w/v) agar. Antibiotics were used when appropriate at the following final concentrations: 100 g/ml ampicillin, 20 g/ml chloramphenicol, and/or 12 g/ml tetracycline (33).
Recombinant DNA Techniques-Preparation of competent cells, transformation, nucleic acid purification, and electrophoresis were performed according to published procedures (33,34). Plasmids were purified using the Qiaprep miniprep spin column kit (Qiagen), and DNA was extracted from agarose gels using the Qiaquick gel extraction spin column kit (Qiagen). All restriction enzymes, pfu polymerase, and T4 ligase were used according to the manufacturers' specifications. E. coli cells were made competent for transformation by the CaCl 2 method (33). DNA sequencing was performed on an ABIprism 377 instrument at the Duke University DNA Analysis Facility.
Plasmid Constructs-Plasmids used in this work are shown in Table  I. C. trachomatis lpxA (CtlpxA) was initially cloned into the Novagen vector pET23c(ϩ). The full-length sequences of CtlpxA at the DNA and protein levels were found by sequence comparisons using the BLAST algorithms (35) with E. coli lpxA (36) as the probe against the C. trachomatis D/UW-3/CX genome (26). One highly homologous gene was identified which displayed 42% identity and 50% similarity at the protein level over the entire length of E. coli LpxA. Oligonucleotide primers to the ends of the CtlpxA gene were synthesized with the flanking restriction sites NdeI at the 5Ј end and BamHI at the 3Ј end of the PCR product.
The forward primer sequence was 5Ј-GGAATTCCATATGACCAA-CATTCATCCTACAGCG-3Ј, and the reverse primer sequence was 5Ј-GCGGATCCTTAAGACTCAACGAAAGCTCCTTC-3Ј. The NdeI and BamHI sites, respectively, are underlined. These primers were used at a final concentration of 0.25 M in 100 l of pfu polymerase PCR reaction mixture containing 10 ng of genomic DNA. The reaction conditions were as follows: a 94°C melt for 5 min followed by 25 cycles of 94°C (melt) for 45 s, 55°C (annealing) for 45 s, and 72°C (extension) for 2 min. This was followed by a 10-min run-off at 72°C. Finally, 1 l of this reaction mixture was subjected to a second round of PCR, using the same conditions as the first.
The amplified PCR product was then purified from the reaction mixture using phenol/chloroform/isoamyl alcohol (25:24:1, v/v). The aqueous phase of this extraction was washed two times with an equal volume of phenol/chloroform/isoamyl alcohol (25:24:1, v/v), and then mixed with 0.1 volume of 3 M sodium acetate and 2.5 volumes of 100% ethanol at 0°C for 30 min to precipitate the DNA. The PCR product was then dried and re-dissolved in 50 l of sterile water, digested with NdeI and BamHI, purified again, and ligated using T4 DNA ligase into similarly treated pET23c(ϩ) vector DNA. The ligation mixture was then transformed into CaCl 2 -competent XL1Blue cells. Plasmid-containing transformants were selected on LB agar plates supplemented with 100 g/ml ampicillin and grown overnight at 37°C. Putative  (19), as shown in blue. The lipid A structure indicated for C. trachomatis is the predominant molecular species (18,19). Based on their genomic DNA sequences (26,36), we propose that the pathways for lipid A biosynthesis are generally the same in C. trachomatis and E. coli, indicating that C. trachomatis LpxA should be selective for myristoyl-ACP. C. trachomatis lacks the msbB gene (26,36), which encodes the enzyme that attaches the secondary myristate chain present at position 3Ј in E. coli lipid A ( Fig. 1), consistent with the structure of C. trachomatis lipid A (19). Numbers at the bottom of the structure indicate the lengths of the acyl chains. For the hybrid lipid A species generated by RO138/pCS131 at 42°C, myristate chains are blue, whereas (R)-3-hydroxymyristate chains are red. Positions 3 and 3Ј are indicated to emphasize that any remaining (R)-3-hydroxymyristate chains present in lipid A species of RO138/pCS131 are selectively incorporated into the distal glucosamine unit. No wild type E. coli lipid A molecules are present in RO138/pCS131. The B1 ϩ and B2 ϩ fragments (53) are indicated for species EC1 only. Palmitate is present on the proximal unit of EC1, which is a minor component, and is incorporated by the regulated enzyme PagP (54) (not shown in Fig. 1).
clones were re-purified on LB ampicillin plates, inoculated into 5 ml of LB broth containing ampicillin, and grown in shaking culture overnight at 37°C. Plasmid DNA was then isolated and subjected to screening by NdeI and BamHI digestion for the presence of the insert. One positive clone containing the CtlpxA gene was designated pCS125 and verified by DNA sequencing.
For studies of the enzymatic activity associated with CtlpxA, the high copy lac vector pBluescript II SK(ϩ) was used. CtlpxA was subcloned by restriction endonuclease digestion of pCS125 with XbaI and HindIII, followed by ligation into these restriction sites of pBluescript II SK(ϩ). One positive clone was designated pCS131. This plasmid was transformed into RO138 and verified by restriction digestion.
Cell Extract Preparation-The strains pCS131/RO138, pTO5/RO138, and pBluescript II SK(ϩ)/RO138 were grown overnight in 5 ml of LB broth containing 100 g/ml ampicillin, at 30°C with 225 rpm rotary shaking. The culture was then diluted 1:100 into 100 ml of fresh medium, and allowed to reach A 600 of ϳ0.6. At this point, the cultures were shifted to 42°C and allowed to grow for another 2 h. At the end of this period, the cultures were chilled on ice for 10 min and then harvested by centrifugation at 4,000 ϫ g at 4°C for 10 min. The cell pellets were resuspended in 10 ml of ice-cold buffer (sterile 100 mM potassium phosphate at pH 7.5 with 200 mM NaCl and 20% glycerol) per 50 ml of original culture, and the cells were centrifuged again at 4,000 ϫ g at 4°C for 10 min. Each washed cell pellet was resuspended in 2 ml of the same buffer and frozen at Ϫ20°C. The cells were then thawed on ice and broken by one passage through a French pressure cell at 14,000 psi. The cell lysates were centrifuged at 10,000 ϫ g for 20 min at 4°C to remove large debris, and the supernatants (extracts) were collected and stored at Ϫ80°C. Before use in the activity assays, the protein concentrations of these extracts were quantified using the Pierce bicinchoninic acid assay kit (37). The data were collected using a Molecular Devices SpectraMax 250 plate reader, and analyzed relative to a bovine serum albumin standard with the software package Softmax Pro. The extracts were also characterized by 15% SDS-PAGE (Bio-Rad mini-protean system) to assess the extent of LpxA overproduction.
Substrate Preparation and Assay Conditions-The [␣ 32 P]-UDP-GlcNAc and (R,S)-3-hydroxymyristoyl-ACP substrates were prepared as previously described (13,14). Myristoyl-ACP (kindly provided by Dr. S. Carty of Duke University Medical Center) was prepared by the same enzymatic method as the hydroxymyristoyl-ACP.
Growth Curve Determination-The permissive temperature for RO138 is 30°C. The non-permissive temperature of 42°C inhibits growth and reduces cell viability within one doubling time. The ability of different constructs to grow at 42°C was determined both on plates and in liquid broth to assess complementation. Growth curves, following a mid-log phase temperature shift from 30 to 42°C, were performed for a more detailed analysis of complementation. Growth was followed by measuring turbidity at 600 nm on a Spectronic 21D spectrometer (Milton Roy) or a Beckman DU-640 spectrophotometer.
Assays of Antibiotic Resistance-Outer membrane integrity was evaluated by examining resistance to selected antibiotics. The assay was performed by placing 8-mm filter discs containing various antibiotics onto a lawn of cells, freshly plated using a saturated cotton swab from a culture at A 600 ϳ 0.2. After 16 h of growth, the zones of inhibition around each disc were measured, providing an assessment of relative antibiotic sensitivity (38,39). In our experiments, the diameter of each zone of inhibition was measured in two dimensions, and the average diameter was used to calculate the area of clearing (mm 2 ). The antibiotics used were rifampicin (30 g/disc) and erythromycin (25 g/disc).
Purification and Analysis of Hybrid Lipid A Species-For structural studies, we isolated hybrid lipid A species generated by the E. coli lpxA2 mutant RO138 complemented with CtlxpA, using previously established techniques (14,40). One liter of cells was grown in LB broth at 30°C to A 600 ϭ 0.7, and then shifted to 42°C for 2 h before being harvested at A 600 ϭ 1.4.
Glycerophospholipids were removed by Bligh-Dyer extraction of the cell pellet (14,40). To release the lipid A, the residue was hydrolyzed at 100°C in acetate buffer, pH 4.5, in the presence of SDS (14,40). This crude lipid A fraction was first purified by chromatography on a DEAEcellulose column (14,40). To resolve pure EC1-EC5, several hundred micrograms of the lipid A mixtures were spotted in a line onto two 20 ϫ 20-cm Silica Gel 60 plates (0.25 mm thickness), which were developed in chloroform/pyridine/formic acid/water (50:50:16:5, v/v), as described previously (40). A second small DEAE-cellulose column was run after extraction of the pure lipids from the silica plates to remove particulate and metal impurities prior to mass spectrometry (40).
Mass Spectrometry of Hybrid Lipid A Species-Spectra were acquired in the negative-ion linear mode by using a Kratos Analytical (Manchester, England) matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometer, equipped with a 337-nm nitrogen laser, a 20 kV extraction voltage, and time-delayed extraction (40). Each spectrum was the average of 50 shots. The matrix was a mixture of saturated 5-aza-2-thiothymine in 50% aqueous acetonitrile and 10% tribasic ammonium citrate (9:1, v/v). The lipid A samples were allowed to dry at room temperature prior to mass analysis. The hexa-acylated lipid A 1,4Ј-bis-phosphate form from E. coli (purchased from Sigma) was used as an external standard.
Mild Base Hydrolysis of Lipid A Species EC3 Prior to Mass Spectrometry-Lipid A samples were dissolved in 100 l of 3% aqueous triethylamine and subjected to sonic irradiation for 3 min before being incubated at 37°C. Portions (0.3 l) of the reaction mixtures were taken at 0, 2, 16, and 40 h for analysis by negative mode MALDI-TOF mass spectrometry, as described above.

RESULTS
Catalytic Activity and Acyl Chain Specificity of CtLpxA-The structure of C. trachomatis lipid A suggests that CtLpxA should transfer myristate specifically to the glucosamine 3-OH of UDP-GlcNAc (Figs. 1 and 2). To examine the activity of the C. trachomatis enzyme, we cloned the CtlpxA gene into pBluescript II SK(ϩ) to generate plasmid pCS131. The pBluescript II SK(ϩ) vector (500 -700 copies per cell) allows for overexpres- sion of the CtlpxA gene behind the lac promoter. The use of strain RO138 as the host for pCS131 facilitated direct measurement of the acyl chain specificity of CtLpxA without interference from E. coli LpxA (29,41).
A crude extract of RO138/pCS131 harvested at 42°C was assayed for its ability to acylate UDP-GlcNAc in the presence of various acyl-ACPs ( Fig. 3 and Table II). Transfer of myristate from myristoyl-ACP to UDP-GlcNAc was detected in this extract, but not in a matched extract of the vector control RO138/ pBluescript II SK(ϩ) (Fig. 3) under conditions where product formation is linear with time. A quantitative analysis (Table II) of the specific activities demonstrates that CtLpxA is about 20-fold more active with myristoyl-ACP (19 Ϯ 2.5 pmol/min/mg, n ϭ 9) than with (R/S)-3-hydroxymyristoyl-ACP (1.1 Ϯ 0.27 pmol/min/mg, n ϭ 7). EcLpxA expressed in RO138 is ϳ700 times more active with (R/S)-3-hydroxymyristoyl-ACP (29,000 Ϯ 6,500 pmol/min/mg, n ϭ 10) than with myristoyl-ACP (42 pmol/min/mg) ( Table II). The low LpxA specific activity of RO138/pCS131 extracts (Table II) may reflect the use of the heterologous E. coli acyl-ACP as the donor substrate; it is not the result of poor protein expression, since C. trachomatis LpxA can be visualized in the cytoplasmic fraction of RO138/ pCS131 by SDS-PAGE (data not shown).
Partial Complementation of the Temperature Sensitivity of RO138 by CtlpxA-RO138/pCS131 was able to grow on plates at 42°C (not shown). However, the colonies were smaller than wild type E. coli. RO138/pCS131 was also able to grow slowly in liquid shaking culture at 42°C, but only reached A 600 of ϳ0.5 after 24 h.
To study the complementation of RO138 by CtlpxA in more detail, cell density (A 600 ) and plating efficiencies were measured after shifting cells from 30 to 42°C. RO138/pCS131 and other control strains were inoculated at A 600 ϭ 0.02 at 30°C, and were shifted to 42°C when A 600 had reached 0.2 (Fig. 4). These growth curves confirm the partial complementation of RO138 by CtlpxA, as indicated by the low A 600 of the RO138/ FIG. 4. Partial complementation of the temperature sensitive growth of RO138 by C. trachomatis lpxA on pCS131. The indicated strains were grown overnight at 30°C on LB broth containing 100 g/ml ampicillin, and then diluted to A 600 ϭ 0.02. The diluted cultures were grown for several more hours at 30°C until A 600 ϭ 0.2, and were shifted to 42°C at time 0. At 42°C, RO138 rapidly lost viability and gradually lysed (29). RO138/pCS131 did not undergo lysis or lose viability at 42°C, but it failed to reach the higher cell density characteristic of wild type cells, such as SM105, or RO138 complemented with wild type E. coli lpxA (not shown).

FIG. 5. Antibiotic sensitivity of RO138 and various strains.
Antibiotic susceptibility was determined by measuring the zones of clearing (mm 2 ) around antibiotic discs placed on lawns of the indicated strains, grown overnight on LB plates, as described previously (31). The strains are further described in Table I. FIG. 6. Ion exchange chromatography of lipid A species from RO138/pCS131 grown at 42°C. Crude lipid A from a 1-liter culture of RO138/pCS131, grown at 42°C for 2 h, was prepared as previously described by hydrolysis at pH 4.5 at 100°C in the presence of SDS (40). A 2-ml DE52-cellulose column (acetate form) was equilibrated with chloroform/methanol/water (2:3:1, v/v) (40,52), and following sample application in the same solvent mixture, was eluted by increasing the ammonium acetate concentration in the aqueous component, as indicated by the first number (mM) below each fraction (3 ml each). Lipid species were detected by spotting 20-l portions of each fraction onto a Silica Gel 60 TLC plate, which was developed in the solvent chloroform/ pyridine/formic acid/water (50:50:16:5, v/v), followed by charring with 10% sulfuric acid in ethanol. The five lipid A species, labeled EC1 to EC5, were further purified by preparative thin layer chromatography on the 100 -500-g scale (40). pCS131 culture in stationary phase at 42°C (Fig. 4). However, dilution plating of RO138/pCS131 after a shift to 42°C showed no decrease in viable cell count (not shown), in contrast to RO138, which loses about 4 orders of magnitude of viability after several hours at 42°C (29). Antibiotic Hypersensitivity of RO138 Complemented by CtlpxA-To assess the impact of heterologous expression of CtlpxA in RO138 on the integrity of the outer membrane, we tested the sensitivity of RO138/pCS131 and several control strains to rifampicin and erythromycin (38,39). Significant antibiotic hypersensitivity is observed in RO138 even at 30°C where the cells are able to grow, and may be the consequence of a slight reduction in lipid A content because of low LpxA activity (29). Restoration of wild-type antibiotic resistance in RO138 at 30°C by introduction of lpxA on a hybrid plasmid indicates that sufficient lipid A species is again produced to restore the outer membrane barrier. Reversal of the antibiotic hypersensitivity of RO138 at 30°C is seen with E. coli lpxA, as well as several heterologous lpxA genes, including lpxA from Neisseria meningitidis and Pseudomonas aeruginosa (13,14).
Interestingly, CtlpxA supplied to RO138 had only a small effect on the antibiotic hypersensitivity at 30°C (Fig. 5), suggesting that RO138/pCS131 generates insufficient or non-functional lipid A species (Fig. 5). We favor the former scenario, since the specific activities of RO138/pCS131 extracts are only 20 pmol/min/mg with myristoyl-ACP as the donor substrate, compared with 150 pmol/min/mg in wild-type vector control cells with hydroxymyristoyl-ACP as the donor (data not shown). As noted above, the low UDP-GlcNAc acyltransferase specific activity in extracts of RO138/pCS131 (Table II) may   FIG. 7. Negative ion MALDI-TOF mass spectrometry of hybrid lipid A species EC1 to EC5. The hybrid lipids were purified as indicated in the legend to Fig. 6, and analyzed as described under "Experimental Procedures." Proposed structures and predicted masses are given in Fig. 2 and Table III. a The data refer to the predominant lipid A 1,4Ј-bis-phosphate species found in wild-type E. coli K12 (14). None of this material is present in RO138/pCS131. reflect the fact that the C. trachomatis enzyme does not efficiently utilize the heterologous E. coli myristoyl acyl-carrier protein as the donor substrate.
Biosynthesis of Hybrid Lipid A Species in RO138/pCS131 at 42°C-Since RO138/pCS131 grows slowly and retains viability at 42°C (Fig. 4), it may be synthesizing novel lipid A species with myristoyl chains at position 3 and/or 3Ј of lipid A. We initially examined the structures of the lipid A species made by a mid-log phase culture of RO138/pCS131, after being shifted to 42°C for 2 h. Separation of the five major lipid A species present in this strain under these conditions required a combination of ion-exchange and thin layer chromatography. Purification of the lipid A obtained from RO138/pCS131 on a DEAE-cellulose column revealed that the five lipid A-like components, designated EC1 to EC5 in Fig. 6, eluted with 240 mM ammonium acetate in the aqueous component. Prior to mass spectrometry, EC1-EC5 were resolved from each other by preparative TLC (40).
Negative ion MALDI-TOF Mass Spectrometry of Lipid A Species in RO138/pCS131-The negative-ion MALDI-TOF spectra ( Fig. 7 and Table III) are consistent with the idea that EC1 and EC2 are hepta-and hexa-acylated lipid A 1,4Ј-bisphosphate species, respectively (Fig. 2). The molecular ions [M-H] Ϫ for these two compounds at m/z 2019.5 and m/z 1781.9, respectively, are 16 atomic mass units smaller than those expected for hepta-or hexa-acylated lipid A 1,4Ј-bisphosphates from wild-type E. coli, which would be observed at m/z 2035.8 and 1797.4, respectively. These results suggest that one hydroxymyristoyl chain has been replaced by a myr-istoyl moiety in both EC1 and EC2 (Figs. 2 and 7 and Table III).
EC3 is of particular interest as it is a penta-acylated lipid A molecule ( Fig. 6) Table III). The latter value is 32 atomic mass units lower than that expected for a penta-acylated E. coli lipid A species in which the secondary myristate chain at position 3Ј is missing, as would be the case for lipid A isolated in mutant defective in msbB (42) (Fig. 1). The molecular ion of EC3 (Fig. 7) therefore suggests that both the 3 and 3Ј-hydroxymyristoyl chains have been replaced with myristate (Fig. 2). A secondary myristate chain cannot be incorporated at position 3Ј into a lipid A species of this kind (Fig. 2).
About one-third of E. coli lipid A molecules contain a diphosphate substituent at position 1 in place of the more predominant monophosphate group (not shown in Fig. 1) (40,43). Based on their mass spectra ( Fig. 7 and Table III), EC4 and EC5 are interpreted as diphosphate variants of EC2 and EC3, respectively (Fig. 2). No wild-type E. coli lipid A, with [M-H] Ϫ expected at m/z 1797.4 (40), was seen in any of the lipid A fractions isolated from RO138/pCS131 (Fig. 7).
Positive Ion MALDI-TOF Mass Spectrometry of Lipid A Species in RO138/pCS131-The positive ion mode spectra of EC1, EC2, and EC4 ( Fig. 8 and Table III) show that the B1 ϩ ions of these species are identical to those expected for the B1 ϩ ion of E. coli wild-type lipid A (14,40), whereas the B2 ϩ ions for these three compounds, like their [M-H] Ϫ value, are each 16 atomic mass units smaller. This finding demonstrates conclusively that myristate replaces hydroxymyristate only at position 3 on the proximal sugar of these species, and that the 3Ј position  (53). Proposed structures and predicted masses are given in Fig. 2 and Table III. occupied by (R)-3-hydroxymyristate as in wild-type E. coli lipid A (Fig. 2).
In EC3 and EC5 (Fig. 8), the size of B1 ϩ ions are each 16 atomic mass units smaller than what would be expected for a lipid A species isolated from an E. coli msbB mutant (Fig. 1) that is otherwise wild type with respect to lipid A biosynthesis (42). The B2 ϩ ions of EC3 and EC5 are each 32 atomic mass units smaller than would be the case if all four primary lipid A acyl chains were (R)-3-hydroxymyristate (Fig. 8). Therefore in these species there is a loss of 16 atomic mass units in both the proximal and distal units, consistent with the view that myristoyl groups are present both at positions 3 and 3Ј (Fig. 2).
To prove that the myristate chains in our hybrid lipid A molecules are indeed ester linked, EC3 (Fig. 2) was treated with 3% aqueous triethylamine for 40 h. At the times indicated in Fig. 9, the EC3 samples were analyzed by MALDI-TOF mass spectrometry in the negative mode. The gradual loss of two 210 atomic mass units fragments from EC3 clearly demonstrates that both of the myristate chains of EC3 are ester linked (Figs. 2 and 9). DISCUSSION UDP-N-acetylglucosamine 3-O-acyltransferase, the product of the lpxA gene, attaches the primary ester-linked hydroxyacyl chains found at positions 3 and 3Ј of E. coli lipid A (Fig. 1) (9 -11). LpxA is required for bacterial growth and is an important determinant of the structure of lipid A (13,14,29). The 3Ј (R)-3-hydroxyacyl chain is further acylated in wild-type E. coli and Salmonella typhimurium with a secondary myristoyl chain to generate the 3Ј-acyloxyacyl unit (Fig. 1) (28, 42). The absence of the secondary myristoyl chain, as seen in msbB mutants of E. coli and S. typhimurium (Fig. 1), greatly reduces virulence (44,45) and diminishes the ability of the mutant bacteria to stimulate cytokine production in human macrophages (42), a process mediated by TLR-4 (6). The lipid A of C. trachomatis is unusual in that it contains myristoyl instead of (R)-3-hydroxymyristoyl chains at positions 3 and 3Ј (Fig. 2) (18,19). C. trachomatis therefore cannot generate a 3Ј-acyloxyacyl unit (Fig. 2), and its lipid A lacks potency when tested as an inducer of cytokines in human cells (46).
We have now shown that the presence of the myristate chain at the 3 and 3Ј positions of C. trachomatis lipid A is due to the unusual substrate specificity of CtLpxA. Selectivity for myristoyl-ACP over (R)-3-hydroxymyristoyl-ACP is apparent both in vitro and in vivo, when CtlpxA is expressed in E. coli RO138 (Figs. 2, 3, and 6 -9). All previously characterized UDP-GlcNAc 3-O-acyltransferases display a high degree of selectivity for 3-hydroxyacyl-ACPs over nonhydroxylated acyl-ACPs (11)(12)(13)(14).
Although the x-ray structure of E. coli LpxA has been solved at 2.6-Å resolution (15), the substrate-binding pocket is not well characterized, since enzyme-acyl-ACP complexes have not yet been crystallized.
The incomplete complementation of the temperature sensitivity of RO138 by CtlpxA (Fig. 4) may be due to the low specific activity of CtLpxA (Table II) when E. coli myristoyl-ACP is the donor substrate. To confirm this idea, the specific activity of CtLpxA with C. trachomatis myristoyl-ACP as the substrate should be investigated. It may be possible to improve CtlpxA complementation of RO138 by simultaneous expression of C. trachomatis acpP, the structural gene for acyl carrier protein (26).
The presence of 3-hydroxymyristate at the 3Ј position of several of the hybrid lipid A species (EC1, -2, and -4) (Fig. 2) FIG. 9. Negative ion MALDI-TOF mass spectrometry of triethylaminetreated EC3. Purified EC3 was treated with 3% aqueous triethylamine at 37°C for the indicated times to cleave the esterlinked acyl chains. The pattern is consistent with loss of two myristoyl moieties, as expected for a compound with the structure of EC3 (Fig. 2). seen in RO138 complemented by CtlpxA is explained by the procedure used to grow this construct. In the experiment shown in Fig. 6, cultures were grown at 30°C to A 600 of ϳ0.7, and then shifted to 42°C for 2 h, during which time the A 600 increased to ϳ1. 4. Therefore, about half the lipid A species obtained for mass spectrometry were derived from RO138/ pCS131 growing at 30°C, where the E. coli LpxA2 mutant protein would be partially functional (29). We do not believe that CtlpxA itself incorporated the 3-hydroxymyristate found in EC1, -2, and -4 (Figs. 2 and 6 -9), since prolonged exponential growth of RO138/pCS131 at 42°C results in the complete disappearance of species EC1, -2, and -4, and their replacement by EC3 and EC5 (Fig. 2) together with small amounts of novel palmitate-containing variants of EC3 and EC5 (not shown).
The fact that 3-hydroxymyristate is present only in the distal glucosamine units of EC1, -2, and 4 ( Figs. 2 and 8) is very interesting. This finding suggests that LpxB, the lipid A disaccharide synthase (47)(48)(49) and/or LpxH, the UDP-2,3-diacylglucosamine hydrolase (50) are highly selective for the structures of their UDP-2,3-diacylglucosmaine substrates (Fig. 1). The incorporation of (R)-3-hdyroxymyristate exclusively into the 3Ј lipid A position of distal units of EC1, -2, and -4 ( Fig. 2) might occur in two ways. 1) E. coli LpxH (Fig. 1) might have a higher binding affinity for UDP-2,3-diacylglucosamine species containing myristate at the 3-position, thereby inhibiting cleavage of UDP-2,3-diacylglucosamine bearing hydroxymyristate at position 3, and resulting in selective formation of lipid X species that are acylated with myristate at position 3. 2) LpxB might select against UDP-2,3-diacylglucosamine species containing myristate at position 3, forcing this material to be cleaved by LpxH, and expanding the lipid X pool with molecular species containing myristate (Fig. 1). The advantage of both mechanisms over random incorporation is that the chances for generating a 3Ј-acyloxyacyl group are maximized, ensuring virulence during animal infections (44,45).
The relevance of the unusual specificity of its UDP-GlcNAc 3-O-acyltransferase to the biology of C. trachomatis is unknown. C. trachomatis lipid A is known to lack potent endotoxin activity (46). Being an obligate intracellular pathogen, C. trachomatis may require the presence of a lipid A species with low intrinsic endotoxin activity, so as not to activate the innate immune system excessively. To test this hypothesis, the feasibility of deleting C. trachomatis lpxA and replacing it with E. coli lpxA could be explored. Providing E. coli msbB in addition to E. coli lpxA might also be of interest. These changes in the lipid A pathway should convert C. trachomatis lipid A to a more potent endotoxin. However, we cannot exclude the possibility that the unusual C. trachomatis lipid A has some other biological function, such as the inhibition of phagolysosome fusion, which is a critical event for the intracellular survival of C. trachomatis (20,21).
Our ability to produce lipid A species in E. coli that bear some resemblance to those found in C. trachomatis now provides facile access to a new series of interesting endotoxin analogs. The immunostimulatory activities of these compounds need to be examined. More generally, the use of Gram-negative bacteria with modified lipid A structures may prove to be an effective approach to the development of new vaccines, justifying the exploration of additional lpx gene chimeras.