An Inner Membrane Enzyme in Salmonellaand Escherichia coli That Transfers 4-Amino-4-deoxy-l-arabinose to Lipid A

Attachment of the cationic sugar 4-amino-4-deoxy-l-arabinose (l-Ara4N) to lipid A is required for the maintenance of polymyxin resistance inEscherichia coli and Salmonella typhimurium. The enzymes that synthesize l-Ara4N and transfer it to lipid A have not been identified. We now report an inner membrane enzyme, expressed in polymyxin-resistant mutants, that adds one or twol-Ara4N moieties to lipid A or its immediate precursors. No soluble factors are required. A gene located near minute 51 on theS. typhimurium and E. coli chromosomes (previously termed orf5, pmrK, oryfbI) encodes the l-Ara4N transferase. The enzyme, renamed ArnT, consists of 548 amino acid residues in S. typhimurium with 12 possible membrane-spanning regions. ArnT displays distant similarity to yeast protein mannosyltransferases. ArnT adds two l-Ara4N units to lipid A precursors containing a Kdo disaccharide. However, as shown by mass spectrometry and NMR spectroscopy, it transfers only a single l-Ara4N residue to the 1-phosphate moiety of lipid IVA, a precursor lacking Kdo. Proteins with full-length sequence similarity to ArnT are present in genomes of other bacteria thought to synthesizel-Ara4N-modified lipid A, including Pseudomonas aeruginosa and Yersinia pestis. As shown in the following article (Trent, M. S., Ribeiro, A. A., Doerrler, W. T., Lin, S., Cotter, R. J., and Raetz, C. R. H. (2001) J. Biol. Chem. 276, 43132–43144), ArnT utilizes the novel lipid undecaprenyl phosphate-α-l-Ara4N as its sugar donor, suggesting that l-Ara4N transfer to lipid A occurs on the periplasmic side of the inner membrane.

Polymyxin is a cationic lipopeptide antibiotic produced by Gram-positive bacteria (21). It binds to lipid A and kills Gramnegative bacteria in a manner that shares some common features with the cationic antibacterial peptides of the innate immune system (21). Substitution of lipid A with L-Ara4N units is greatly elevated in polymyxin-resistant mutants of S. typhimurium (22,25) and E. coli K-12 (23). The polymyxin resistance phenotype is usually due to mutation(s) in the pmrA locus (26), which encodes a transcription factor that is activated during growth under mildly acidic conditions (pH Ͻ 6), in a PhoP/PhoQ-dependent manner during Mg 2ϩ limitation (10 M), or by exposure to ferric ions (20,27,28). The first two of these conditions exist within the phagolysosomes of macrophages, which engulf S. typhimurium during the course of an infection (19,27,29).
PmrA activation by environmental stimuli or by appropriate point mutations within pmrA induces the expression of genes needed for polymyxin resistance and covalent modification of lipid A with L-Ara4N (24,27,30,31). A cluster of genes mapping near minute 51 in S. typhimurium and E. coli (24) have been proposed to encode a set of enzymes required for the biosynthesis of the L-Ara4N moiety and its attachment to lipid A ( Fig. 2) (10). However, in vitro assays have not yet been developed to validate the functions of these putative enzymes.
We now report a novel enzyme, present in inner membranes of polymyxin-resistant mutants of S. typhimurium, that can transfer one or two L-Ara4N moieties to lipid A and certain lipid A precursors. Transferase activity is dependent upon activation of the pmrA gene, either directly by mutation or secondarily by growth under appropriate conditions. The enzyme adds a single L-Ara4N unit to the 1-phosphate moiety of the tetraacylated lipid A precursor, lipid IV A , which lacks Kdo (Fig. 3). However, lipid A molecules containing a Kdo disaccharide are modified with two L-Ara4N units, indicating that the addition of L-Ara4N to the 4Ј-position is Kdo-dependent. Transferase activity is greatly elevated when the orf5(pmrK) gene of S. typhimurium (6,24), now renamed arnT (Fig. 2), is expressed behind a T7lac promoter in E. coli BLR(DE3), which is itself shown here to be a polymyxin-resistant strain containing L-Ara4N modified lipid A. The L-Ara4N transferase is not dependent upon added soluble factors. As shown in the following article (32), ArnT utilizes the novel lipid, undecaprenyl phosphate-␣-L-Ara4N, as its donor substrate (Fig. 2).

EXPERIMENTAL PROCEDURES
Materials-32 P i and [␥-32 P]ATP were obtained from PerkinElmer Life Sciences. Silica Gel 60 (0.25-mm) thin layer plates were purchased from EM Separation Technologies. Tryptone and yeast extract were from Difco. Triton X-100 and bicinchoninic acid were from Pierce. CDCl 3 , CD 3 OD, and D 2 O were purchased from Aldrich. All other chemicals were reagent grade and were purchased from either Sigma or Mallinckrodt.
Bacterial Strains-Bacterial strains are described in Table I. Typically, the bacteria were grown at 37°C in LB broth, which contains 10 g of NaCl, 10 g of tryptone, and 5 g of yeast extract per liter (33). For experiments requiring Mg 2ϩ limitation or exposure to low pH, cells were grown as described previously on a defined medium (34). When required for selection of plasmids, cells were grown in the presence of 100 g/ml ampicillin, 12 g/ml tetracycline, 30 g/ml chloramphenicol, or 30 g/ml kanamycin.
Recombinant DNA Techniques-Plasmids were prepared using the Qiagen Spin Prep kit. DNA fragments were isolated from agarose gels using the Qiaex II gel extraction kit. T4 DNA ligase (Life Technologies, Inc.), restriction endonucleases (New England Biolabs), and shrimp alkaline phosphatase (U. S. Biochemical Corp.) were used according to the manufacturer's instructions.
Overxpression of the L-Ara4N Lipid A Transferase (ArnT) Behind a T7lac Promoter-The arnT(orf5) gene of S. typhimurium (Fig. 2) was cloned into pET21 (Novagen) behind the T7lac promoter. The gene was amplified by PCR using S. typhimurium 14028 genomic DNA as the template. The forward primer contained a clamp region, an NdeI site (underlined), and the arnT coding region with its start codon. The reverse primer contained a clamp region, a BamHI site (underlined), and the coding region with its stop codon. Sequences of the S. typhimurium forward and reverse primer were 5Ј-GCGCGCCATATGAT-GATGAAATCGATA-3Ј, and 5Ј-GCGCGCGGATCCTCATTTAGGC-CGATA-3Ј, respectively. The PCR contained 140 ng of genomic DNA template, 0.15 g of each primer, 200 M each of dNTP, 100 mM Tris-HCl, pH 8.8, 35 mM MgCl 2 , 250 mM KCl, and 2.5 units of Pfu DNA polymerase (Stratagene) in a reaction volume of 0.05 ml. The reaction mixture was subjected to a 2-min denaturation at 95°C followed by 25 cycles of 95°C for 45 s, 60°C for 45 s, 72°C for 120 s and a final extension for 10 min at 72°C, using the PerkinElmer Life Sciences GeneAmp PCR system 2400. The PCR product and the vector were digested with NdeI and BamHI, ligated together, and transformed into Xl-1 Blue cells (Stratagene) for propagation of the plasmid, designated pArnTSt. The plasmid was then transformed into BLR(DE3) or Nova-Blue(DE3) ( Table I) with or without the pLysS plasmid for overexpression of the protein. First, a single colony of E. coli containing pArnTSt was inoculated into 20 ml of LB broth and grown in a rotary shaker at 37°C to A 600 ϭ 0.8. The culture was then used to inoculate 1 liter of fresh LB medium, and when A 600 reached ϳ0.6, the culture was induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 4 h. Crude extracts, membrane-free cytosol, and washed membranes were prepared as described below.
Preparation of Cell-free Extracts and Membranes-Typically, 100-ml cultures of bacteria were grown to A 600 ϭ 1.0 at 37°C and harvested by centrifugation at 7,000 ϫ g for 15 min. All steps were carried out at 0 -4°C. Cells were resuspended in 50 mM Hepes, pH 7.5, at a protein concentration of 5-10 mg/ml, and broken by one passage through a French pressure cell at 18,000 pounds/square inch. The crude lysate was centrifuged at 7,000 ϫ g for 15 min to remove unbroken cells. Membranes were prepared by two successive centrifugations at 149,000 ϫ g for 60 min, with a washing step inbetween to remove residual soluble components. The final membrane pellet was resuspended in 50 mM Hepes, pH 7.5, at a protein concentration of 5-10 mg/ml. Cytosol from the first 149,000 ϫ g centrifugation step was subjected to a second centrifugation to remove any remaining contaminating membranes. All membrane and cytosol preparations were stored in aliquots at Ϫ80°C, and protein concentrations were determined with bicinchoninic acid (35), using bovine serum albumin as the standard.
Isolation and Analysis of Lipid A Species from 32 P i -Labeled Cells-Cells were labeled uniformly with 5 Ci/ml 32 P i in LB broth, starting at an initial A 600 of ϳ0.05. Cells were then grown at 37°C for several hours, as indicated, and harvested when A 600 reached ϳ1.0. The 32 Plabeled cells were collected using a clinical centrifuge and washed with 5 ml of phosphate-buffered saline, pH 7.4 (36). The final cell pellet was resuspended in 3 ml of a single-phase Bligh/Dyer mixture (37), consisting of chloroform/methanol/water (1:2:0.8, v/v). After 60 min, the insoluble material, which still contains the 32 P-labeled lipid A covalently linked to the lipopolysaccharide core via its Kdo residues, was released by hydrolysis at 100°C in the presence of 1% SDS at pH 4.5, as described previously (10,38). The 32 P-labeled lipid A species were recovered by two-phase Bligh/Dyer extraction (10,38) and spotted onto a Silica Gel 60 TLC plate (ϳ10,000 cpm/lane). The plate was developed in the solvent chloroform, pyridine, 88% formic acid, water (50:50:16:5, v/v). The plate was dried and exposed to a PhosphorImager Screen overnight to visualize the resolved 32 P-lipid A species.
Assay Conditions for Detecting L-Ara4N Transferase Activity-The L-Ara4N transferase was assayed under optimized conditions in a 10-l reaction containing 50 mM MES, pH 6.5, 0.2% Triton X-100 and either 10 M [4Ј-32 P]lipid IV A or Kdo 2 [4Ј-32 P]lipid IV A (each at 20,000 cpm/ nmol) as the acceptor substrate. Washed membranes (0.5-1.0 mg/ml) were employed as the source of enzyme and L-Ara4N donor, as indicated. Reaction mixtures were incubated at 30°C for various times. The phosphate residues and acyl chains of lipid A in S. typhimurium may be modified in a regulated fashion (30), as indicated by the dashed bonds. The phosphates moieties may be substituted with L-Ara4N and/or pEtN groups, both of which are under PmrA control (blue substituents) (11,22,24). Minor species are present in which the locations of the L-Ara4N and pEtN groups are reversed or in which both phosphates are modified with the same substituent (11). Addition of the palmitoyl chain is catalyzed by the outer membrane enzyme PagP (12), and formation of 2-hydroxymyristate (ϫ) requires a novel hydroxylase homologue, encoded by lpxO (14). Substituents that are incorporated in a PhoP/PhoQ-dependent manner are shown in red.
of the initial experiments, the PmrA C S. typhimurium strain JSG435 (Table I) was used for the characterization of the assay conditions, because of its apparently high levels of endogenous L-Ara4N transferase activity and the presence of the donor substrate.
When [4Ј-32 P]lipid IV A was employed as the acceptor, the substrate and reaction products were separated using the solvent chloroform, pyridine, 88% formic acid, water (50:50:16:5, v/v). For assays containing Kdo 2 [4Ј-32 P]lipid IV A as the acceptor substrate, plates were developed in chloroform, pyridine, 88% formic acid, water (30:70:16:10, v/v). Following chromatography, the plates were dried and analyzed using a Molecular Dynamics PhosphorImager equipped with ImageQuant software. The enzyme activity was calculated by determining the percentage of the substrate converted to product. The apparent specific activities were expressed in units of nmol/min/mg, recognizing that the membranes supply not only the enzyme but also the L-Ara4N donor substrate.
Separation of Inner and Outer Membranes-Membranes isolated from the PmrA C S. typhimurium strain JSG435 (Table I) were separated by isopycnic sucrose gradient centrifugation as described previously (34). First, washed membranes were prepared as described above and then were resuspended at a concentration of ϳ5 mg/ml in 10 mM Hepes, pH 7.0, containing 0.05 mM EDTA. Membranes (2.2 ml) were applied to a 9.6-ml, seven-step sucrose gradient (43), and subjected to ultracentrifugation at 35,000 rpm in a Beckman SW40.1 rotor for 19 h at 3°C. The gradient was collected in ϳ0.5 ml fractions by piercing the bottom of the tube. Each fraction was assayed for NADH oxidase as the inner membrane marker and for phospholipase A as the outer membrane marker (38). The amount of protein in each fraction was determined using the bicinchoninic acid assay (35). Finally, each fraction was assayed for L-Ara4N transferase activity, using the optimized assay conditions described above.
Purification of the L-Ara4N-modified Reaction Products Generated in Vitro from Lipid IV A -The lipid IV A reaction products, generated in vitro with membranes of strain JSG435, were purified by preparative thin layer chromatography. A 68-ml L-Ara4N transferase reaction mixture (see above), containing 100 M lipid IV A and 1 mg/ml JSG435 membranes, was incubated overnight at 30°C. The reaction mixture was then converted into a two-phase acidic Bligh/Dyer system, consisting of chloroform, methanol, 0.1 M HCl (2:2:1.8, v/v), by the addition of 76 ml of chloroform, 76 ml of methanol, and 0.6 ml of 12 M HCl. After mixing, the phases were separated by centrifugation at 5,000 ϫ g for 15 min. The lower phase was removed, and the upper phase was extracted a second time by the addition of 76 ml of a fresh lower phase derived from a two-phase Bligh/Dyer mixture of chloroform, methanol, and 0.1 M HCl. The lower phases, containing the residual substrate and the modified lipid IV A species, were pooled. Next, 4 ml of pyridine was added to neutralize remaining HCl carried over during the extraction process. The sample was dried by rotary evaporation, dissolved in 8 ml of chloroform/methanol (4:1, v/v), and spotted as lines onto the origins of 14 individual 20 ϫ 20-cm Silica Gel 60 TLC plates (0.25-mm thickness) (ϳ0.57 ml per plate). The lipids were separated using the solvent system chloroform, pyridine, 88% formic acid, water (50:50:16:5, v/v). While the plates were drying at room temperature, the bands of residual lipid IV A and its modified products could be seen transiently as white zones. The latter was marked with a pencil, and the plates were dried for an additional 30 min with a cold air stream. Regions containing the putative L-Ara4N-modified lipid IV A derivatives (i.e. presumed to be lipids II A and II B based on their migration) (Fig. 4) were removed separately with clean razor blades and then stored in separate thickwalled glass tubes at Ϫ80°C (10).
Silica chips from four separate thin layer plates containing bands with the same R f were processed together. The lipids were extracted from the chips with 6 ml of an acidic single-phase Bligh/Dyer mixture, consisting of chloroform, methanol, 0.1 M HCl (1:2:0.8, v/v). Following removal of the chips by low speed centrifugation, the supernatant was converted to a two-phase Bligh/Dyer system by adding 1.58 ml of both  53), a separate enzyme to catalyze the decarboxylation step is not necessary in our scheme (10). ArnB (previously Orf1 or PmrH) (6, 10, 24) then catalyzes a further transamination to form UDP-L-Ara4N (S. Breazeale, A. A. Ribeiro and C. R. H. Raetz, manuscript in preparation). Based upon its homology to dolichyl phosphate-mannose synthase of yeast, we propose that ArnC (PmrF) (6, 10, 24) transfers the L-Ara4N moiety to undecaprenyl phosphate, forming the novel compound undecaprenyl phosphate-␣-L-Ara4N, the existence of which is demonstrated in the following article (32). After translocation to the outer surface of the inner membrane by unknown mechanisms (10), ArnT (previously Orf5, PmrK, or YfbI) (6,24) transfers the L-Ara4N unit to lipid A. Other genes of the polymyxin operon (pmrJ, pmrL, and pmrM), as well as the adjacent pmrG gene (6,24), cannot yet be assigned specific enzymatic or transport functions in our scheme. The ArnA protein has a second catalytic domain (reaction not shown) that can transfer a formyl group from N-10formyltetrahydrofolate to UDP-L-Ara4N (S. Breazeale, A. A. Ribeiro, and C. R. H. Raetz, manuscript in preparation), but the significance of this modification is unclear. Addition of the L-Ara4N moiety to lipid A reduces the affinity of lipid A for polymyxin and other cationic anti-microbial peptides.
chloroform and water. The lower phase, which contained the desired lipid, was withdrawn, and the upper phase was re-extracted once more with an equivalent volume of fresh lower phase. The lower phases were pooled, and 6 drops of pyridine were added. The sample was dried under a stream of N 2 .
To remove minor breakdown products and contaminating silica particles, the individual lipids recovered from the TLC plates were subjected to anion exchange chromatography on small DEAE-cellulose columns (10). Each lipid sample was re-dissolved in 3 ml of chloroform/ methanol/water (2:3:1, v/v) and subjected to sonic irradiation for 30 s in a bath apparatus before application to a 1-ml DEAE-cellulose column, suspended in the same solvent mixture, and equilibrated with acetate as the counter ion (10,44). After application of the sample, the column was washed with 4 bed volumes of chloroform/methanol/water (2:3:1, v/v). The products were eluted with the same solvent system but with the aqueous portion consisting of 60, 120, 240, or 500 mM ammonium acetate in ascending order. For each elution step, four 1-ml fractions were collected. The L-Ara4N-modified species, lipids II A and II B , eluted with chloroform, methanol, 120 mM ammonium acetate (2:3:1, v/v), whereas the substrate lipid IV A eluted with 240 -500 mM ammonium acetate. Fractions from the DEAE columns containing the desired lipid products, which were detected by charring on a TLC plate with 10% sulfuric acid in ethanol, were pooled and converted to a two-phase Bligh/Dyer mixture by addition of appropriate amounts of chloroform and water. The upper phases were extracted once with fresh lower phases to maximize the recovery of each lipid. The appropriate com-bined lower phases for each lipid product were then dried under a stream of N 2 and stored at Ϫ80°C.
Mass Spectrometry of the L-Ara4N-modified Reaction Products-Spectra of the purified lipids were acquired in the negative and positive linear modes, using a matrix-assisted laser desorption-ionization/time of flight (MALDI/TOF) instrument (Kompact MALDI 4, Kratos Analytical, Manchester, UK), equipped with a nitrogen laser (337 nm) (10). By using 20-kV extraction voltage and time-delayed extraction, each spectrum represented the average of 50 laser shots. The instrument was operated at a resolution of about Ϯ1 atomic mass units for compounds with M r ϳ2000. Saturated 6-aza-2-thiothymine in 50% acetonitrile and 10% tribasic ammonium citrate (9:1, v/v) served as the matrix in negative and positive ion modes. The samples were dissolved in chloroform/ methanol (4:1, v/v) and mixed with an equal portion of matrix. The sample was dried at 25°C prior to mass analysis.
Nuclear Magnetic Resonance Spectroscopy of the L-Ara4N-modified Reaction Products-Approximately 1.5 and 1 mg, respectively, of the putative lipid II A and lipid II B , generated in vitro and purified by TLC as described above, were dissolved in 0.6 ml of CDCl 3 /CD 3 OD/D 2 O (2:3:1, v/v) and transferred into 5-mm NMR tubes. The NMR spectra were recorded at 25°C using a Varian Unity 500 spectrometer equipped with a Sun Ultra 5 data system and a 5-mm Varian inverse probe. The 2 H signal of CD 3 OD was used for a field frequency lock. Both the homoand heteronuclear NMR experiments were performed as described previously (11,45,46).

A Possible L-Ara4N Transferase in Membranes of PhoP and
PmrA Constitutive Mutants of S. typhimurium-Although the L-Ara4N moiety of S. typhimurium lipid A was first reported in 1970 (7), its enzymatic synthesis has remained obscure until recently. S. typhimurium mutants harboring constitutively active forms of the PhoP or PmrA transcription factors synthesize large amounts of L-Ara4N-modified lipid A (22,30). Accordingly, an in vitro assay was developed with membranes from such mutants, using the precursor lipid IV A (Fig. 3) as a possible acceptor for L-Ara4N transfer. Membranes of S. typhimurium CS022 (47), in which PhoP is constitutively active (PhoP C ), converted a measurable portion of the [4Ј-32 P]lipid IV A probe to a more hydrophilic product migrating like lipid II A (Fig. 4, lane 3), a well characterized substance that accumulates in Kdo-deficient mutant of S. typhimurium (9,46,48). As shown by the structural formulas in Fig. 3, lipid II A has the same the tetra-acylated glucosamine disaccharide backbone as does lipid IV A , but it is modified with the L-Ara4N unit on its 1-phosphate moiety (9,46,48).
Membranes of wild-type S. typhimurium (Fig. 4, lane 2) generated very little lipid II A under these assay conditions. However, both wild-type and PhoP C membranes produced large amounts of lipid IV B (Fig. 4, lanes 2 and 3), a pentaacylated product containing a palmitoyl moiety (Fig. 3), which is incorporated by the outer membrane enzyme PagP (12). PagP has a very high specific activity under these assay conditions.
A further increase in the rate and extent of lipid II A formation was seen with membranes of the PmrA C S. typhimurium strain JSG435 (Fig. 4, lane 6), which displayed an apparent specific activity of ϳ0.06 nmol/min/mg at 1 mg/ml membrane protein. A substance migrating somewhat faster than the 4Ј-32 P-lipid II A product, designated lipid II B (9, 48) (Fig. 4, lanes 3  and 6), likely arises from the PagP-catalyzed addition of palmitate to lipid II A or, alternatively, by L-Ara4N transfer to lipid IV B (Fig. 3). As predicted, lipids IV B and II B largely disappear when membranes of a PhoP C strain that also harbors a pagP insertion mutation were used as the enzyme source (Fig. 4, lane  5). The putative L-Ara4N transferase was entirely absent in membranes of either phoP or pmrA knockout mutants (Fig. 4,  lanes 4 and 7), consistent with the absence of L-Ara4N-modified lipid A in these strains (6, 30, 49).
Membranes of wild-type S. typhimurium grown either in LB FIG. 3. Structure of lipid IV A and the products generated from it by membranes of pmrA C S. typhimurium. Lipid IV A and lipid II A are tetra-acylated species with the latter containing the L-Ara4N moiety attached to the 1-phosphate group (9,46,48). The numbering scheme shown for lipid II A is used for the NMR analysis. Lipids IV B and II B are further modified with a palmitoyl group at the 2-position, which is incorporated by the outer membrane enzyme PagP (12). broth or in minimal medium at pH 7.4 supplemented with 10 mM Mg 2ϩ (conditions under which the PhoP and PmrA systems are shut off) showed little or no L-Ara4N transferase activity (data not shown). However, in accordance with the well characterized behavior of PhoP and PmrA, membranes from bacteria grown in minimal medium in the presence of 10 M Mg 2ϩ or at pH 5.8 (conditions leading to activation of PmrA) (20,27,29) displayed transferase activity comparable with membranes of phoP C or pmrA C mutants grown on LB broth at pH 7.4 (data not shown).
Catalytic Properties and Inner Membrane Localization of the L-Ara4N Transferase-Transfer of the L-Ara4N unit from its putative isoprenoid carrier, undecaprenyl phosphate-L-Ara4N (Fig. 2), to the radiolabeled acceptor [4Ј-32 P]lipid IV A (Fig. 3) was dependent upon the presence of the nonionic detergent Triton X-100, with maximal activity observed at 0.2% in the assay system. The pH optimum was 6.5, and the enzyme did not require any soluble cytoplasmic components (Fig. 5). CaCl 2 and MgCl 2 were inhibitory above 1 mM. Product formation was dependent upon protein concentration up to 1 mg/ml (data not shown) and was linear for about 60 min (Fig. 5) using 10 M [4Ј-32 P]lipid IV A as the acceptor with 1 mg/ml membrane pro-tein from strain JSG435 (pmrA C ) as the source of enzyme and L-Ara4N donor.
Separation of inner and outer membranes by isopycnic sucrose gradient centrifugation showed that the transferase is located mainly in the inner membrane (Fig. 6). NADH oxidase served as the inner membrane marker. The peak of NADH oxidase and L-Ara4N transferase activity coincided at fraction 17 (Fig. 6). Phospholipase A (50) was used as the marker for outer membranes (Fig. 6). Supplementation of the assay with purified undecaprenyl phosphate-L-Ara4N (32) did not alter the distribution of L-Ara4N transferase activity (data not shown).

MALDI/TOF Mass Spectrometry of the L-Ara4N
Transferase Products Formed with Lipid IV A as the Acceptor-To confirm the presence of the L-Ara4N moiety in the in vitro products migrating with lipids II A and II B (Fig. 4), 1-1.5 mg of each was purified by ion exchange and thin layer chromatography, as described under "Experimental Procedures." MALDI/TOF mass spectrometry in the negative-ion mode of the product migrating with the lipid II A standard (Fig. 4) (Table I) were assayed for transfer of L-Ara4N moieties from endogenous sources to the acceptor substrate 4Ј-32 P-lipid IV A . The protein concentration was 1 mg/ml. Assays were carried out for 6 h at 30°C with 10 M 4Ј-32 P-lipid IV A (20,000 cpm/nmol). The products were separated by TLC and detected with a PhosphorImager. trometry has been documented previously (10,11). The small peak at m/z 1558.3 atomic mass units (Fig. 7A) is attributed to the sodium adduct [M ϩ Na Ϫ 2H] Ϫ . MALDI/TOF mass spectrometry of the additional reaction product, migrating with the lipid II B standard (Fig. 4), yielded a major ion at m/z 1774.1 atomic mass units (Fig. 7B) Positive-ion mode MALDI/TOF mass spectrometry (not shown) failed to reveal the site of attachment of the L-Ara4N substituent in the in vitro products, given the labile nature of the L-Ara4N phosphodiester linkage compared with the ␤,1Ј-6 linkage of the glucosamine disaccharide. However, the positiveion mode spectra (not shown) confirmed the location of the palmitate residue on the proximal glucosamine unit in the lipid II B -like product (9,48) (Fig. 3).
In summary, the mass spectrometry strongly supports the structural assignments, consistent with the initial TLC analysis (Fig. 4), as representing the in vitro formation of lipids II A and II B from lipid IV A . An enzyme that adds L-Ara4N units to lipid A or its precursors had not been reported previously. Evidence that the membrane bound donor of the L-Ara4N res-idue ( Fig. 2) is in fact undecaprenyl phosphate-L-Ara4N is presented in the following article (32).
Analysis of the L-Ara4N Attachment Site by Difference 1 H NMR Spectroscopy with Selective Inverse ( 31 P) Decoupling-As discussed previously (9,46), the precursor lipid II A that accumulates in Kdo-deficient S. typhimurium mutants is modified with a single L-Ara4N moiety via a phosphodiester linkage at the 1-phosphate position, as shown in Fig. 3. In contrast, most of the L-Ara4N attached to mature lipid A in wild-type cells is attached via a phosphdiester linkage to the 4Ј-phosphate ( Fig.  1) (11,46). As shown in the preceding article (11), however, a portion of the lipid A of S. typhimurium does contain two L-Ara4N moieties, implying that the attachment of the L-Ara4N unit to the 4Ј-position may be Kdo-dependent.
To establish the site of L-Ara4N attachment in the lipid II A -like substance generated in vitro by membranes of JSG435, the 31 P NMR spectrum of this substance was recorded at 202 MHz in CDCl 3 /CD 3 OD/D 2 O (2:3:1, v/v). As shown in Fig. 8A, well resolved 31 P resonances were detected at 2.14 and Ϫ2.55 ppm. A 1 H NMR spectrum of the same sample was recorded at 500 MHz in the absence of 31 P-continuous wave decoupling, and the sugar proton region is shown at the bottom of Fig. 8B. The tracing in Fig. 8B labeled 2.14 ppm shows the difference spectrum that resulted from subtracting the 1 H NMR spectra of the sample with selective on-and off-resonance decoupling of the 2.14 ppm 31 P resonance (Fig. 8A). This difference spectrum shows a "triplet" near 4.2 ppm, which can be assigned to the FIG. 6. Inner membrane localization of the L-Ara4N transferase of S. typhimurium. Membranes isolated from the pmrA C strain JSG435 were separated by isopycnic sucrose density gradient centrifugation, and ϳ0.5-ml fractions were collected (34). A, the outer membrane marker phospholipase A (% of total activity) and the protein concentration (mg/ml) were assayed for each fraction. B, the inner membrane marker NADH oxidase (% of total activity) and the L-Ara4N transferase activity displayed similar distributions. The L-Ara4N transferase profile was not changed significantly by addition of exogenous undecaprenyl phosphate-L-Arar4N (data not shown). H-4Ј-glucosamine proton (Fig. 3) based upon a complete 1 H-1 H COSY analysis of the II A -like product generated in vitro (see Fig. 1 in the Supplemental Material), which is the same as that of authentic lipid II A (46). This outcome clearly establishes the lower field 2.14 ppm phosphate resonance as arising from a monophosphomonoester linked to the C-4Ј of the distal glucosamine unit (Fig. 3), exactly as in authentic lipid II A isolated from Kdo-deficient S. typhimurium cells (46).
The top tracing in Fig. 8B, labeled Ϫ2.55 ppm, is the difference spectrum obtained upon subtracting 1 H NMR spectra with selective on-and off-resonance decoupling of the Ϫ2.55 ppm 31 P resonance (Fig. 8A). This spectrum reveals two overlapping proton "doublet" signals near 5.5 ppm (assigned to H-1 and H-1Љ by COSY analysis), and two smaller "doublet of doublets" at 4.22 and 3.70 ppm that are assigned to H-2 and H-2Љ, respectively (see Fig. 3 for the numbering scheme). These findings establish the upfield Ϫ2.55 ppm phosphorus resonance as arising from the bridging monophosphodiester group between C-1 of the proximal glucosamine and C-1Љ of the L-Ara4N ring in the in vitro synthesized material, further establishing its identity as lipid II A . Small differences in the actual 31 P chemical shifts between the in vitro product (Fig. 8A) and the lipid II A previously isolated from cells (2.14 versus 1.56 ppm for P-4Ј and Ϫ2.55 versus Ϫ2.62 ppm for P-1) (46) are attributed to slight differences in the pH, which was not rigorously controlled.
Difference spectra like those shown in Fig. 8B were also acquired for the lipid II B product generated in vitro, again demonstrating that the 1-phosphate is the sole site of L-Ara4N attachment under these assay conditions (data not shown).
Overexpression of the S. typhimurium arnT Gene in a Polymyxin-resistant Strain of E. coli-Based on a PSI-BLAST anal-ysis (51), arnT of S. typhimurium (Fig. 2) is proposed to encode the L-Ara4N transferase, since ArnT displays distant similarity to the yeast protein mannosyltransferases, which utilize dolichyl phosphate-mannose as their donor substrate (52). The predicted ArnT protein is 548 amino acids long and has 12 membrane-spanning regions (24). To determine whether arnT encodes a functional L-Ara4N transferase, the gene was amplified by PCR from S. typhimurium genomic DNA and was subcloned behind the T7lac promoter in the expression vector pET21. The resulting hybrid plasmid was designated pArnTSt.
Prior to expression studies with pArnTSt, we were surprised to find that E. coli B derivatives, such as the BLR(DE3) host strains from Novagen, contain lipid A modifications similar to those found in pmrA C S. typhimurium (Fig. 9, lane 1). In contrast, E. coli K-12 strains do not contain any modified lipid A species (2, 10) when grown on LB broth (Fig. 9, lanes 2 and  4). Separation of the 32 P-labeled lipid A species from BLR(DE3)pLysS (Fig. 9, lane 3) clearly demonstrated the presence of L-Ara4N-and/or pEtN-modified lipid A species. These lipid A modifications in the host strains were not dependent upon the presence of the DE3 lysogen or on the plasmid pLysS that encodes the T7 lysozyme (data not shown). Furthermore, the BLR strains of E. coli are polymyxin-resistant when grown in the presence of 2 g/ml polymyxin B sulfate on LB agar plates (data not shown). Accordingly, a source of L-Ara4N donor substrate (i.e. the putative undecaprenyl phosphate-L-Ara4N) should be available in membranes prepared from these E. coli B derivatives, commonly used for overexpression of proteins.
Membranes isolated from strains of BLR(DE3)pLysS, containing either the vector control pET21 or pArnTSt, were assayed for L-Ara4N transferase activity using [4Ј-32 P]lipid IV A as the acceptor (Fig. 10). Massive overproduction of the FIG. 8. 31 P NMR spectrum of lipid II A synthesized in vitro and selective inverse 31 P-decoupled 1 H-detected difference spectroscopy. A, the 31 P NMR spectrum of the lipid II A -like material synthesized in vitro is consistent with the presence of one monophospho-monoester (2.14 ppm) and one monophosphodiester moiety (Ϫ2.55 ppm). B, the bottom tracing is a reference 1 H NMR spectrum, showing the relevant sugar protons from 3.6 to 5.8 ppm. The complete COSY analysis, which is the basis for assignment of key resonances as indicated, is shown in Fig. 1 of the Supplemental Material. The atom numbering is shown in Fig. 3. The middle tracing is a difference spectrum obtained with on and off resonance decoupling of the 31 P signal at 2.14 ppm. The only observed signal is a triplet near 4.2 ppm assigned to H-4Ј, demonstrating that the 4Ј-phosphate is not substituted with the L-Ara4N moiety. The top difference spectrum was obtained by selective on and off resonance decoupling of the 31 P signal at Ϫ2.55 ppm. It reveals two overlapping proton doublets near 5.5 ppm (H-1 and H-1Љ) and two smaller doublet of doublets at 4.22 and 3.7 ppm (H-2 and H-2Љ, respectively). This pattern confirms that the L-Ara4N unit is attached at position 1, as in authentic lipid II A (46). L-Ara4N transferase activity was seen in membranes from cells containing pArnTSt (Fig. 10, lane 5) as compared with the vector control (Fig. 10, lane 3). The membranes containing the overexpressed ArnT had a specific activity of 12 nmol/min/mg, an ϳ200-fold increase compared with membranes of the pmrA C S. typhimurium mutant JSG435 (Fig. 4), when assayed at a protein concentration of 0.5 mg/ml. However, membranes isolated from arnT overexpressed in the E. coli K12 strain NovaBlue(DE3)pLysS did not catalyze any transfer of the L-Ara4N moiety to [4Ј-32 P]lipid IV A (Fig. 10, lane 4), presumably because of the absence of an endogenous L-Ara4N donor substrate. As expected, the lipid A species isolated from NovaBlue(DE3)pLysS contained no L-Ara4N or pEtN substituents (Fig. 9, lane 4). The combined data strongly support the view that arnT (Fig. 2) is the structural gene for the L-Ara4N transferase.
Assay of ArnT with Kdo 2 [4Ј-32 P]Lipid IV A or Kdo 2 [ 32 P]Lipid A as the Acceptor-To determine whether modification of the 4Ј-phosphate with the L-Ara4N moiety is dependent upon the Kdo disaccharide, membranes from cells containing the overexpressed arnT gene were assayed using the precursor, Kdo 2 -[4Ј-32 P]lipid IV A (Fig. 11). The rate at which the L-Ara4N modification reaction occurred was comparable with that seen with [4Ј-32 P]lipid IV A as the acceptor, but a second more hydrophilic substance, designated Product C in Fig. 11, was observed, consistent with the incorporation of a second L-Ara4N unit. When assayed at various times, Product C appeared to be dependent upon the prior formation of Product B, a species proposed to contain only one L-Ara4N-derivatized phosphate. Unfortunately, it is impossible to determine from these experiments which phosphate residue of Kdo 2 [4Ј-32 P]lipid IV A is modified first with the L-Ara4N moiety. In fact, Product B is likely to be a mixture of singly modified species. The results do strongly suggest, however, that S. typhimurium ArnT is a bifunctional glycosyltransferase in that it can incorporate two L-Ara4N units. Product A presumably results from the PagPdependent palmitoylation (12) of the Kdo 2 [4Ј-32 P]lipid IV A acceptor (Fig. 3). Additional studies will be required to determine Membranes from BLR(DE3)pLysS cells expressing either pET21 or pArnTSt were assayed for L-Ara4N transferase under standard conditions at 0.5 mg/ml protein and 10 M Kdo 2 -4Ј-32 P-lipid IV A for the indicated times. Products were separated by TLC and visualized by PhosphorImager analysis. The migration of Products B and C is consistent with the addition of one or two L-Ara4N moieties, respectively. Product A is formed by incorporation of a palmitoyl group, catalyzed by PagP (12). The small amount of transferase in the vector control is due to chromosomal E. coli ArnT. the exact structures of the products generated by the action of ArnT on Kdo 2 [4Ј-32 P]lipid IV A .
In a separate set of assays, Kdo 2 [ 32 P]lipid A (i.e. the hexaacylated material) was used as the acceptor. As seen with Kdo 2 [4Ј-32 P]lipid IV A (Fig. 11), two hydrophilic products were generated by membranes that overexpressed ArnT, confirming that one of the L-Ara4N units is incorporated in a Kdo-dependent manner (data not shown).

DISCUSSION
The L-Ara4N substituent of S. typhimurium lipid A was discovered in 1970 (7), but its biosynthesis and function have remained elusive until recently. Increased substitution of lipid A with L-Ara4N moieties occurs in polymyxin-resistant mutants (22,23,25), suggesting a charge neutralization function that might protect cells against cationic antimicrobial peptides. Polymyxin resistance is caused by mutations in the transcription factor PmrA, which is rendered constitutively active in resistant strains (26). An important breakthrough was made in 1998 by Gunn et al. (24), who discovered an operon near minute 51 required for maintenance of polymyxin resistance in pmrA constitutive organisms. Sequence analysis of the protein products of this operon with the PSI-BLAST tool suggested a plausible hypothesis for the enzymatic synthesis of L-Ara4N from UDP-glucuronic acid, as shown in Fig. 2 (10, 53).
We have now developed the first in vitro system for L-Ara4N transfer to lipid A (Figs. 4, 10, and 11). Both the transferase (ArnT) and its L-Ara4N donor substrate are associated with the inner membrane of S. typhimurium (Figs. 5 and 6). This is consistent with the hydropathy profile of the transferase, which contains 12 membrane-spanning regions and with the proposed role of undecaprenyl phosphate-L-Ara4N as the donor (Fig. 2). Robust L-Ara4N transferase activity is seen only in membranes of cells in which PmrA is activated and is completely absent in membranes of pmrA Ϫ mutants (Fig. 4).
Overexpression of S. typhimurium arnT behind the T7lac promoter in E. coli BLR/DE3 results in a further 200-fold elevation of L-Ara4N transferase-specific activity, when compared with membranes of pmrA C S. typhimurium, supporting the assignment of arnT as the structural gene for the transferase (Fig. 2). Unexpectedly, however, the T7 expression strain BLR(DE3) is itself polymyxin-resistant and contains L-Ara4Nmodified lipid A (Fig. 9), whereas no L-Ara4N-modified lipid A is made in the polymyxin-sensitive E. coli K-12 expression host, NovaBlue(DE3) (Fig. 9). No L-Ara4N transferase activity is seen in membranes of NovaBlue(DE3) cells that overexpress S. typhimurium arnT (Fig. 10), because the L-Ara4N donor substrate is absent. Addition of purified undecaprenyl phosphate-L-Ara4N to such membranes reconstitutes full transferase activity (32).
ArnT of E. coli K-12 is 550 amino acid residues long, and it displays 71% identity and 82% similarity to the 548-residue S. typhimurium ArnT. E. coli K-12 furthermore contains all the other genes of the S. typhimurium operon required for polymyxin resistance (24). The regulation of these genes in E. coli is probably similar to that reported for the better characterized S. typhimurium system, but there are some differences. Unlike S. typhimurium, lipid A of E. coli K-12 does not contain the L-Ara4N moiety when grown in LB broth at pH 6.8 (10,11). Selection of polymyxin-resistant mutants (23) or treatment of wild-type cells with ammonium metavanadate (10) is required to induce L-Ara4N biosynthesis in E. coli K-12. As demonstrated in the following article (32), the L-Ara4N transferase utilizes the novel carrier lipid, undecaprenyl phosphate-L-Ara4N, as its donor substrate (Fig. 2). The utilization of this substrate by ArnT suggests that the transferase active site faces the periplasm, given that peptidoglycan (54) and cytosis and sensitizing the bacteria to attack by the innate immune system.