Lipid A Precursor from Pseudomonas aeruginosa Is Completely Acylated Prior to Addition of 3-Deoxy-~-manno-octu~osonate*

Inhibition of lipopolysaccharide (LPS) synthesis in Pseudomonaa aeruginosa at the stage of incorporation of 3-deoxy-~-manno-octu~osonate (KDO) caused accu- mulation of a lipid A precursor which contained all of the fatty acids present on the lipid A of mature LPS. The enzyme CTP:CMP-3-deoxy-~-manno-octuloson-ate cytidylytransferase (CMP-KDO synthetase) from P. aeruginosa is inhibited by the KDO analog a-C-[ 1,5-anhydro-8-amino-2,7,8-trideoxy-~-manno-octopyra- nosyl]carboxylate (I), and I is effectively delivered to P. aeruginosa following attachment by amide linkage to the carboxyl terminus of alanylalanine. Intracellular hydrolysis releases the free inhibitor (I) which then inhibits activation of KDO by CMP-KDO synthetase causing accumulation of lipid A precursor and subse- quent growth stasis. The major lipid A precursor species accumulated was purified and found to contain glucosamine, phosphate, C12:0,20H-C12:0 and 30H-C1O:O (in ester linkage), and 30H-C12:0 (in amide linkage) in molar ratios of


anhydro-7-amino-2,7-dideoxy-~-manno-heptopyranosyl]car~xyla~
IV, I1 linked to the carboxyl terminus of alanylalanine; HPLC, high performance liquid chromatography. This yields UDP-2,3-diacyl-GlcN (2). A portion of UDP-2,3diacyl-GlcN is hydrolyzed to UDP + 2, 3-diacyl-GlcN then 2,3-diacyl-GlcN condenses with a molecule of UDP-2,3-diacyl-GlcN, and the product is phosphorylated (3, 4) to yield the basic disaccharide structure of lipid A, 0-(2-amino-2deoxy-~-~-glucopyranosyl)-(l~)-2-amino-2-deoxy-a-D-glucose, acylated at positions 2, 3, 2', and 3' with 3-OH-C14:0, and bearing phosphate at positions 1 and 4' (IVA). Compound IVA is one of the lipid A precursor species which accumulated when a temperature-sensitive kdsA mutant of Salmonella typhimurium was incubated at the restrictive temperature (5). Thermal inhibition of KDO-8-phosphate synthetase, the product of the kclsA gene, effectively blocks further synthesis of KDO and causes accumulation of four major species of putative lipid A precursors (5). Since precursor species IVA accumulates first and to the greatest extent in both S. typhimurium and Escherichia coli, it likely represents the normal in vivo acceptor for KDO.' The other precursor species which appear only after IVA has accumulated several hundred-fold in the inner membrane probably represent aberrant reactions occurring in response to inhibition of metabolite flow through the pathway. ' Lipid A precursor from S. typhimurium is an acceptor of two KDO molecules in vitro (6), and recently a species with a single KDO was detected both in uitro and in vivo (7). Although the species containing a single KDO was not a precursor to the species containing two KDO molecules, it was a transient metabolite. Although purely speculative at present, the species containing a single KDO could be a direct precursor to LPS and thus represent a subfraction of LPS containing a single KDO. A single KDO apparently links lipid A to other core sugars in some Gram-negative bacteria (8,9). Our knowledge of the early stages of LPS assembly is limited to S. typhimurium and E. coli for two major reasons: (i) the genetic approach is limited due to lack of methods for directly selecting for mutations in the pathway, and (ii) the direct biochemical analysis of the pathway is difficult because intermediates are present in low levels and are rapidly turning over. However, we recently reported a new class of synthetic antibacterial agents (10) which specifically inhibit 3-keto-Dmanno-octulosonate incorporation in Gram-negative bacteria. The new agents consist of competitive inhibitors of CMP-KDO synthetase a-C-[1,5-anhydro-7-amino-2,7-dideoxy-~manno-heptopyranosyl]carboxylate (11) and a-C-[1,5-anhydro-8-amino-2,7,8-trideoxy-~-manno-octopyrano~yl]carboxylate (I) which are delivered to bacteria as peptide prodrugs.
These prodrugs are transported by the oligopeptide permease system and hydrolyzed by intracellular peptidases which release the free inhibitor.
In this report, we have used these new antibacterial agents * Goldman, R., Doran, C., and Capobianco, J., J. Bacteriol., in press.

5217
Lipid A Precursor from P. aeruginosa to examine the early stages in assembly of LPS in diverse species of Gram-negative bacteria and to show that, although similarities exist, there are species-specific differences. In particular, the major lipid A precursor species from Pseudomonas aerugimsa is completely acylated with all of the fatty acids present on mature LPS prior to addition of KDO. In contrast, the major species from other enteric Gram-negative species contains only 3-OH-C14:0, and lacks the other nonhydroxy fatty acids characteristic of mature LPS.

MATERIALS AND METHODS
Bacterial Strains and Growth Conditions-Bacterial strains used in this study are listed in Table I. Bacteria were grown at 37 "C in defined medium (11) containing 0.2% glucose, 0.5 mM N-acetylglucosamine and 1 mM leucine, except where indicated.
Induction of Lipid A Precursor and Analysis by Radiolubeling-Bacteria from an overnight culture were inoculated into fresh medium to an A4m value of 0.05. When the culture attained an A42(J value of 0.3-0.5, drug was added to a final concentration of 50 or 100 pg/ml, and incubation was continued for 3 h. Radiolabel was added at the same time as drug using the following protocols: (i) N-acetyl-D-[l-3H]glucosamine (Amersham TRK .376) was added to 4 &i/ml, (ii) [33P]phosphoric acid (New England Nuclear NEZ 080) was added to 10 pCi/ml, and (iii) sodium acetate (1 mM final) was added to the medium, and [3H]acetate (Amersham TRK .12) or [2-"Clacetate (Amersham CFA.14) was added to 20 pCi/ml. Cells were harvested by centrifugation and washed twice with 0.05 M 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid-NaOH buffer, pH 7.4. Cell pellets were resuspended in 1.25 ml of methanol containing 0.1 N HCl, followed by addition of 0.5 ml of H20 and 0.625 ml of chloroform.
Samples were mixed intermittently at room temperature for 20 min, after which 0.625 ml of H20 and 0.625 ml of chloroform were added with mixing. Phases were separated by centrifugation at 5000 X g for 5 min, and the lower chloroform phase was saved. Chloroform (1.25 ml) was added to the methanol phase and interface followed by mixing and centrifugation. The first and second chloroform phases were combined and dried under a stream of Nz. The residue was resuspended in chloroform for application to Silica Gel H plates. Plates were developed in a solvent of chloroform, pyridine, 88% formic acid, H,O (4060165) for 2 h and air-dried. Fractions (0.5 cm) were scraped into scintillation vials and radioactivity determined following addition of 100 pl of H20 followed by 10 ml of Instagel (Packard).
Analysis of Radiotubeled Fatty Acids-Fatty acids attached to lipid A precursor were radiolabeled by protocol (see ii above), except 100 pCi of labeled acetate was added to 10-ml cultures. Precursors were extracted from bacteria as described above and purified by chromatography on Silica Gel H. Specific precursor peaks were extracted from silica gel using chloroform/methanol/H,O (63:33:4), containing 0.1 N HCI, which was evaporated to dryness under a stream of nitrogen. Fatty acids were released by acid hydrolysis (0.5 ml of 4 N HCl at 100 "C in a sealed, nitrogen-flushed tube). Free fatty acids were extracted twice into equal volumes of hexane and stored at room temperature.
Samples from hexane were dried under nitrogen, resuspended in methanol, and analyzed by HPLC using a Cl8 column (10 cm X 4.5 mm). Two solvent elution programs were used. AI1 samples, except those from P. aeruginosa, were run on a column equilibrated with acetonitrile/H,O (6040, v/v) and were eluted with a linear gradient (to acetonitrile/H20, 80:20, v/v) over 30 min, followed by isocratic elution for 40 min. Fatty acids from P. aeruginosa were run on a column equilibrated with acetonitrile/H2O (5050, v/v) and eluted with a linear gradient (to acetonitrile/H20, 9010, v/v) over 15 min, followed by isocratic elution for 30 min. AII H20 solutions contained 20 mM phosphoric acid, and solvent flow was 1 ml/min. Radioactivity in the column effluent was monitored with a Radiomatics Flow Detector (RadioAnalytic, Inc., Tampa, FL) using a 2.5-ml liquid cell and Flow Scint I11 at 3 ml/min. The counting efficiency for tritium was 33%.
Preparation of Lipid A from P. aeruginosa-P. aeruginosa K799 was grown in LB medium (GIBCO), and LPS was extracted with aqueous phenol as described (12). Lipid A was prepared by acid hydrolysis (1% acetic acid) and collected by centrifugation.
Purification of Lipid A Precursor from P. aeruginosa K799"Cells were grown in defined medium (330 ml) containing 0.2% glucose, 1 mM leucine, and 0.5 mM N-acetylglucosamine. When the A420 value reached 0.3, I11 was added to 100 pg/ml. After incubation for 3 h, cells were harvested and washed twice with 4-(2-hydroxyethyl)-lpiperazineethanesulfonic acid-NaOH buffer, 0.01 M, pH 7.4, and frozen at -20 "C. The culture supernatant was filter-sterilized and used to induce lipid A precursor in another batch of cells as follows. Cells in late log growth (A420 = 2) were harvested by centrifugation and diluted immediately into the filter-sterilized supernatant to an value of 0.2-0.3, and incubation was continued for 3 h. These cells were harvested, and the cycle was repeated. Preliminary small scale experiments using radiolabel analysis of lipid A precursor showed that equal amounts of precursor were produced during each of the three cycles.
Cell pellets were delipidated by a series of extractions with 95% ethanol followed by acetone, and then diethyl ether (13). Delipidated cells were extracted with chloroform by a modified Bligh-Dyer procedure (5, 14, 15). The chloroform extracts were pooled and a small amount of 90% phenol mixed into the chloroform. The chloroform was evaporated under nitrogen and 2 volumes of methanol added to the remaining phenol phase. The methanol-phenol mix was loaded onto a 2.5 X 11-cm DEAE-cellulose column equilibrated with 99% methanol, 1% acetic acid. The column was washed with equilibration buffer, then eluted with 500 ml linear gradient of 0-1 M ammonium acetate in equilibration buffer (6,13). Fractions of 4 ml were collected and tritium counts determined with a Packard Tri-Carb scintillation counter. Peak tubes were pooled and extracted into chloroform as outlined above.
Pulse-Chase Ezperiments-Cells were grown in defined medium at 37 "C to an kt420 value of 0.3-0.5, drug (III,50 pg/ml) was added for 5 min, followed by pulse labeling with [3H]N-acetylglucosamine (21 pCi/ml) for 15-30 min. Cells were rapidly chilled and washed twice with medium containing 5 mM unlabeled N-acetylglucosamine, and resuspended in medium lacking drug and radiolabel, but containing 5 mM unlabeled N-acetylglucosamine. Samples were taken at the end of the pulse period and at designated times following chase.
Samples were analyzed by the following methods: (i) extraction of lipid A precursor into chloroform and analysis by chromatography on Silica Gel H as given above; (ii) solubilization in 0.06 M tris (hydroxymethy1)aminomethane-HC1 buffer, pH 6.8, containing 3% sodium dodecyl sulfate (w/v) and 5% mercaptoethanol (v/v) at 100 "C for 5 min, followed by polyacrylamide gel electrophoresis (16) on a 15% acrylamide gel containing 0.1% sodium dodecyl sulfate and 6 M urea. Gels were sliced into 2-mm fractions, starting from the bottom, and each gel fraction was added to a scintillation vial containing 1 ml of 0.1% sodium dodecyl sulfate in H20. After 24 h at room temperature, scintillation fluid was added and samples were counted. LPS from strain K799 consists predominately of short chains lacking 0-antigen (12). Using this gel system, LPS migrated to fraction 9, while phospholipid and lipid A precursor migrated to fraction 6, fraction 1 being defined at the position of the bromphenol blue tracker dye; or (iii) preparation of inner and outer membrane fractions following lysis using French press and sucrose density gradient centrifugation (17). Fractions containing inner and outer membranes were pooled, dialyzed against H,O, lyophilized, and analyzed for lipid A precursor by extraction and chromatography on Silica Gel H (see above).
Analytical Procedures-Glucosamine content was determined after hydrolysis with 4 N HCl at 100 "C for 14 h by a modification of the Elson-Morgan reaction (18,19). Phosphate content was determined by the Ames procedure (20) and protein estimated by the method of Lowry et al. (21). Fatty acids were determined by gas chromatography (7) after conversion to their methyl esters with methanolic HCl (100 'C/17 h). A standard fatty acid mix was prepared from the lipid A of P. aeruginosa K799. Confirmation of peaks was accomplished by mass spectrum analysis on a Hewlett-Packard model 5985A, gas chromatography/quadrupole mass spectrometer (Packard Instrument Co.) operated at 70 eV ionization energy, 300 micro A filament, and a source temperature of 200 "C. Both electron and chemical (NH3) impact were utilized to generate mass molecular ions [M + HI' and [M + NH,]+.
The mass molecular ion (M -H)for DEAE-cellulose peak 1 and for the lipid A standard were determined on a Kratos MS-50 mass spectrometer (AEI/Kratos) in the negative mode. Samples (10-20 pg) in chloroform/methanol (4:l) were mixed with tetraethylene glycol on the instrument probe. A neutral beam of xenon was used and the translational energy varied between 6-8 kV. Data were collected at 30 s/decade scan rate over a mass range of 500-1800 daltons.
Chemicals-All radiolabels were purchased from Amersham Corp. DEAE-cellulose (DE52) was obtained from Whatman and methanolic HC1 kits were purchased from Alltech Assoc. Inc. KDO analogs were

RESULTS AND DISCUSSION
Induction of Lipid A Precursor in P. aeruginosa-Growth ceased and new glucosamine containing metabolites accumulated (Fig. 1, A and B ) following treatment of P. aeruginosa K799 with I11 (I amide linked to the carboxyl terminus of alanylalanine). Intracellular hydrolysis of I11 releases the a CMP-KDO synthetase inhibitor I which inhibits CMP-KDO synthetase from P. aeruginosa with a k, of 0.5 p~. The major new component migrated on Silica Gel H between phospholipid and lipid A precursor species IVA prepared from S. typhinuriurn (Fig. 1B). Since both components were also radiolabeled with acetate and phosphate, they were likely acylated and phosphorylated derivatives of glucosamine. The major component was detectable 20 min after drug addition and reached peak levels by 50 min. Samples radiolabeled with [3H]acetate were recovered from Silica Gel H and analyzed for fatty acid content by HPLC. The major component contained 3-OH-10:0, 2-OH-12:0, 3-OH-12:0, and 12:O in the ratios 1:0.5:0.5:1, respectively, based on radiolabel ( Fig. 2A). The 3-OH-12:O moiety was amide-linked (not removed by base hydrolysis), whereas 12:0, 2-OH-12:0, and 3-OH-10:O were all apparently ester-linked (removed by base hydrolysis, Fig. 2B). The second, minor precursor species (Fig. 1B) also contained predominantly hydroxy fatty acids, but was not characterized further. Identical results were obtained using P. aeruginosa strain A5007 (data not shown).
Purification and Characterization of the Major New Lipid A Derivative-The major new lipid A derivative was purified by chromatography on DEAE-cellulose (Fig. 3), eluting between 0.2 and 0.3 M ammonium acetate. Chemical analysis gave glucosamine, phosphate, 3-OH-100, 3-OH-12:0, 2-OH-12:0, and 12:O in the same molar ratio as found in lipid A prepared from LPS (Table 11)   observed at m/z 826 and 842, which is consistent with the above hypothesis. These data are similar to those reported for lipid A of Chromobacterium viohceum (22). The identity of the fatty acids (2-OH-C12:0, 3-OH-C12:0, C12:0, and 3-OH-C100) were confirmed by tandem gas chromatography/fast atom bombardment mass spectroscopy (Table 111)   Control experiments demonstrated that the new phosphorylated and acylated glucosamine metabolites were only detectable following treatment with 111. No fatty acids derived from lipid A were detectable by direct gas chromatographic analysis of chloroform extracts from control cells, whereas 12:0, 3-OH-10:0, 3-OH-12:0, and 2-OH-12:O were all present from analogous extracts of drug-treated cells (data not shown). A second series of control experiments used radiola-beling of control cells to high specific activity with [3H]acetate (10 @Ci/ml) followed by extraction chromatography on Silica Gel H, and analysis of the chromatographic regions between the position of IVA and phospholipid. No 12:0,3-OH-1@.0,3-OH-12:0, or 2-OH-120 fatty acids were detected by HPLC analysis (data not shown). These control experiments show that the new lipid A derivative only appears following treatment with 111, and that it is not a cleavage product of mature LPS generated during extraction.
The New Lipid A Derivative Accumulates in the Inner Membrane and Is a Precursor to LPS-The new lipid A derivative accumulated predominantly in the inner membrane following pulse labeling with [3H1N-acetylglucosamine (Fig.   5, A and B), with only 5-10% being found in the outer membrane fractions. The membrane fractions observed were nearly identical to those previously reported (l?), consisting of inner membrane, M band, and outer membrane. Although we obtained slight fractionation of the outer membrane into subfractions (Fig. 5A), there was no significant difference in the small amounts of lipid A derivative present in the two outer membrane fractions analyzed. In addition, less than 5% of the pulsed radiolabel was present in the M band, and it was thus not investigated further. These data are consistent with the hypothesis that the new lipid A derivative is assembled in the inner membrane, but is not capable of rapid translocation to the outer membrane. This observation is similar to the results reported for S. typhimurium (24) and suggests that, in both Salmonellu and Pseudomonas, preassembled lipid A is not capable of rapid translocation to the outer membrane. If true, removal of drug should allow addition of KDO, completion of LPS synthesis, and effect a chase of the new lipid A derivative to complete LPS. Following removal of drug, 50-80% of the pulsed radioactivity present in the new lipid A derivative chased into LPS (Table IV), with a concurrent decrease of the lipid A derivative from the inner membrane (Fig. 5B). These data show that the new lipid A derivative is a precursor to LPS.

TABLE rv
Chase of lipid A precursor to LPS Cells (strain K799) growing in defined medium were treated with 50 pg/ml of I11 for 5 min, followed by pulse labeling with 13H)Nacetylglucosamine for 15 min. Cells were then chilled, washed twice with media containing 5 pM unlabeled N-acetylglucosamine, and resuspended in fresh medium lacking drug and radiolabel, but containing 5 m M unlabeled N-acetylglucosamine. Duplicate samples were taken at the end of the pulse and at 30 and 60 min following initiation of the chase. Samples were analyzed for lipid A precursor and LPS as described in the text. Results are the average of two determinations. ter species were treated with IV (I1 amide-linked to the carboxyl terminus of alanylalanine). In each case, the major component comigrated with lipid A precursor species IVA prepared from $. typhimurium or E, coli. The structure of IVA is ~-(2-amino-2-deoxy-~-~-g~ucopyranosyl)-(l~)-Z-amino-2-deoxy-a-D-glucose, acylated at positions 2,3,2', and 3' with 3-OH-el40 groups, and bearing phosphate at positions 1 and 4' (5). Species IIIA of S. typhimurium is a derivative of IVA which contains phosphoethanolamine attached to the 4'phosphate. Species IA is identical to IVA, except for attachment of phosphoethanolamine to the 4'-phosphate and attachment of aminopentose to the 1-phosphate. Species 11, is identical t o IvA except for attachment of aminopentose to the 1-phosphate. The minor components from Serratia and Providencia comigrated with species IIA of S. typhimurium, while those from E. coli, Enterobacter, and Citrobacter comigrated with species 111, from S. typhimurium. Both the major and minor components were radiolabeled with [33P]phosphate and [3H]acetate. The major components contained 3-OH-C140 as the only fatty acid (Fig. 7, A and B). The   ited. The complexity of lipid A precursor species, which accumulate in S. typhimurium upon inhibition of the KDO pathway, is thus far unique. A major component apparently identical to IvA from s. typhimurium does appear in other enteric Gram-negative species, however, only one other minor, more polar precursor species appeared. Species IVA accumulates first and to the greatest extent when S. typhimurium is treated with IV,2 indicating that it is the normal in uiuo acceptor of KDO. The much slower and less extensive accumulation of I*, IIA, and 111, may thus reflect aberrant reactions which only occur when IVA has accumulated due to inhibition of KDO addition. Kinetic analysis of appearance of the two lipid A derivatives in E. coli was similar in that IVA appeared first and to the greatest extent.2 Since the component comigrating with IVA was the most abundant in all enteric species examined, the appearance of minor species likely reflects formation of metabolic side products due to inhibition of the normal pathway. The single species of nonenteric Gram-negative bacteria examined, two strains of P. aeruginosa, revealed a significant difference compared to enteric organisms. The major precursor species accumulated was fully acylated, in contrast to the presence of only hydroxy fatty acids attached to the major precursor species of enteric species. These data are consistent with the following hypothesis. In P. aerugirwsa, UDP-GlcNAc is acylated at the 3 position with 3-OH-C100, at the 2-position (after deacylation) with 3-OH-C120, and finally the acyloxacyl fatty acid (2-OH-Cl2:O or (2120) is added to the 2-position 3-OH-C120 group. This would allow the observed heterogeneity in the structure of the nonreducing Gln. Hydrolysis of a portion of the derivatized Gln from UDP, yielding a source of the reducing Gln analogous to reactions in E. coli (1) would allow the observed heterogeneity in the reducing Gln. Alternatively, heterogeneity could arise following initial assembly of tetraacyldissacharide 1-4'-bisphosphate, with subsequent addition of either C12:O or 2-OH-C12:0 to both the reducing and nonreducing Gln residue. These differences likely reflect organism-specific pathways for the addition of fatty acids and KDO during the assembly of lipid A.