Enzymatic synthesis of lipid A molecules with four amide-linked acyl chains. LpxA acyltransferases selective for an analog of UDP-N-acetylglucosamine in which an amine replaces the 3"-hydroxyl group.

LpxA of Escherichia coli catalyzes the acylation of the glucosamine 3-OH group of UDP-GlcNAc, using R-3-hydroxymyristoyl-acyl carrier protein (ACP) as the donor substrate. We now demonstrate that LpxA in cell extracts of Mesorhizobium loti and Leptospira interrogans, which synthesize lipid A molecules containing 2,3-diamino-2,3-dideoxy-d-glucopyranose (GlcN3N) units in place of glucosamine, do not acylate UDP-GlcNAc. Instead, these LpxA acyltransferases require a UDP-Glc-NAc derivative (designated UDP 2-acetamido-3-amino-2,3-dideoxy-alpha-d-glucopyranose or UDP-GlcNAc3N), characterized in the preceding paper, in which an amine replaces the glucosamine 3-OH group. L. interrogans LpxA furthermore displays absolute selectivity for 3-hydroxylauroyl-ACP as the donor, whereas M. loti LpxA functions almost equally well with 10-, 12-, and 14-carbon 3-hydroxyacyl-ACPs. The substrate selectivity of L. interrogans LpxA is consistent with the structure of L. interrogans lipid A. The mechanism of L. interrogans LpxA appears to be similar to that of E. coli LpxA, given that the essential His(125) residue of E. coli LpxA is conserved and is also required for acyltransferase activity in L. interrogans. Acidithiobacillus ferrooxidans (an organism that makes lipid A molecules containing both GlcN and GlcN3N) has an ortholog of LpxA that is selective for UDP-GlcNAc3N, but the enzyme also catalyzes the acylation of UDP-GlcNAc at a slow rate. E. coli LpxA acylates UDP-GlcNAc and UDP-GlcNAc3N at comparable rates in vitro. However, UDP-GlcNAc3N is not synthesized in vivo, because E. coli lacks gnnA and gnnB. When the latter are supplied together with A. ferrooxidans lpxA, E. coli incorporates a significant amount of GlcN3N into its lipid A.

We now demonstrate that UDP-GlcNAc3N is selectively utilized by LpxA orthologs present in cell extracts of Leptospira interrogans and Mesorhizobium loti and by the cloned LpxA proteins of A. ferrooxidans (AfLpxA) and L. interrogans (LiLpxA). All three of these organisms synthesize lipid A molecules containing GlcN3N units (5)(6)(7). L. interrogans LpxA displays no measurable activity with UDP-GlcNAc, and it uses R-3-hydroxylauroyl-ACP in absolute preference to all other acyl-ACPs. These findings are consistent with the proposed structure of L. interrogans lipid A, which is presented in the following paper (7). AfLpxA can also acylate UDP-GlcNAc at a slow rate, which is consistent with the reported structure of lipid A in this organism ( Fig. 1) (5). Escherichia coli LpxA (8 -11) utilizes both UDP-GlcNAc and UDP-GlcNAc3N with equal efficiency under standard assay conditions. However, the latter sugar nucleotide is not synthesized in wild-type cells. We have therefore constructed a novel strain of E. coli harboring the temperature-sensitive chromosomal lpxA2 mutation (12,13) while simultaneously expressing the cloned gnnA, gnnB, and lpxA genes of A. ferrooxidans. The lipid A backbone of this organism is partially substituted with GlcN3N when the cells are grown at elevated temperatures, consistent with our proposed pathway for the biosynthesis of GlcN3N-containing lipid A molecules (4). * This work was supported by National Institutes of Health Grants GM-51310 and GM-51796 (to C. R. H. R.), and GM-54882 (to R. J. C.) and by Grant PTR94 form the Institut Pasteur (to C. W.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the DDBJ/GenBank TM

EXPERIMENTAL PROCEDURES
Materials-All growth media, materials, and fine chemicals were the same as in the preceding manuscript (4) or were purchased from Sigma-Aldrich. PerkinElmer Life Sciences was the source of [␣-32 P]UTP. The LpxC inhibitor L-573,655 was kindly provided by Dr. A. Patchett (Merck Research Laboratories) (14).
Bacterial Strains and Plasmids-A. ferrooxidans ATCC 23270 and M. loti ATCC 700743 were purchased from the American Type Culture Collection. A. ferrooxidans cells were grown on modified 9K medium (15). M. loti cells were grown on yeast mannitol agar or liquid medium (6). E. coli XL1-Blue (Stratagene) was used for cloning and maintenance of plasmids. E. coli BL21(DE3)/pLysS (Stratagene) was used for overexpression of cloned genes, typically using the Novagen plasmids pET23cϩ and pET30aϩ. The E. coli temperature-sensitive mutant RO138 (lpxA2 recA rpsL Tet r ), a recA derivative of SM101 (12), was provided by Dr. M. Anderson (Merck Research Laboratories). The vector pBluescript II SK(ϩ) (Stratagene) was used to express foreign genes in R0138, as described below. Plasmid constructs are summarized in Table I. E. coli cells were generally grown on LB agar or in LB broth, adjusted to pH 7.4 (16). Bacteria harboring hybrid plasmids were selected using ampicillin at 100 g/ml. The recombinant plasmids pTO1 (17), pTO5 (17), and pCS355 (4) have been previously described.
The plasmid pCS61 was constructed by subcloning the E. coli lpxA gene from pTO5 (17) into the low copy vector pNGHamp (18), using the SacI restriction site.
Recombinant DNA Techniques-Transformation of competent cells, nucleic acid purification, and electrophoresis were carried out according to established procedures (19,20). The plasmids were purified using the Qiaprep miniprep spin column kit (Qiagen). Digested PCR products and plasmid DNA were purified by phenol/chloroform/isoamyl alcohol extraction and ethanol precipitation (19,20) or by agarose gel electrophoresis in conjunction with the Qiagen Qiaquick gel extraction kit (4).
Isolation and Cloning of the lpxA Gene from A. ferrooxidans-Genomic DNA was prepared from A. ferrooxidans as described (4). The lpxA gene was PCR-amplified with the Pfu polymerase using the oligonucleotide primers 5Ј-GGA ATT CCA TAT GAC GGT GCA GAT TCA TCC GCT GG-3Ј (N-terminal primer) and 5Ј-CGG GAT CCC CCC GAT GAC CCC GGT TCT CAT ATG G-3Ј (C-terminal primer  (5). The 1-and 4Ј-positions (X and Y, respectively) lack phosphate moieties (5). C, L. interrogans and M. loti lipid A contain GlcN3N exclusively (3-and 3Ј-nitrogen atoms in red). The structure of M. loti lipid A is not fully characterized (6). That of L. interrogans is presented in the following paper (7). In lipopolysaccharide, 3-deoxy-D-manno-octulosonic acid would be attached at position 6Ј.
Generation of Plasmid Constructs Containing lpxA, gnnA, and gnnB from A. ferrooxidans-The AflpxA gene was subcloned from pCS311 along with the pET23cϩ ribosome-binding site into pBluescript II SK(ϩ) using XbaI and HindIII to generate pCS421 and then from this construct into pNGH100 (17) using SacI and BamHI to generate pCS449. The gnnA and gnnB genes were likewise cloned into pBluescript II SK(ϩ) by insertion of the XbaI/HindIII fragment of the bicistronic construct pCS355 (4) into similarly treated pBluescript vector. The desired hybrid plasmid expressing the two genes was designated pCS411. In addition, a pBluescript II SK(ϩ)-based construct was constructed bearing the A. ferrooxidans gnnA, gnnB, and lpxA genes. For this purpose, a PCR fragment bearing AflpxA was first generated using pCS311 as the template, and primers were designed to create a PCR product bearing the entire lpxA gene preceded by the pET23cϩ ribosome-binding site. The N-terminal primer was 5Ј-CCC AAA AAG CTT GGG AGA CCA CAA CGG TTT CCC-3Ј. The C-terminal primer was 5Ј-CCC GCC TCG AGG TCG ACG GAG CTC GAA TTC GGA TCC-3Ј. This fragment was then cloned into pCS411 using HindIII and XhoI digestion, which inserts the lpxA fragment behind the gnnA and gnnB genes. As noted above, these genes are themselves behind a ribosome-binding site from pET23cϩ (subcloned from pCS355). The hybrid plasmid derived from this ligation was designated pCS439.
Cloning of lpxA from L. interrogans Serovar Icterohemeorrhagiae (Strain Verdun)-The whole genome sequence of the L. interrogans serovar Lai (21) revealed one gene (LilpxA) of 780 bp that displayed 41% identity and 59% similarity at the protein level with E. coli LpxA (9). The LilpxA gene of L. interrogans serovar Icterohemeorrhagiae (strain Verdun, virulent isolate) (22) was PCR-amplified with the Pfu polymerase using the oligonucleotide primers 5Ј-GGA ATT CCA TAT GAA AAT ACA TCC GAC TGC TAT TA-3Ј (N-terminal primer) and 5Ј-GCG GAT CCT CAC CTG TGA TTT GTA ACT CCCC-3Ј (C-terminal primer). The NdeI and BamHI sites, respectively, are underlined. The amplified gene was digested with NdeI and BamHI and then ligated with T4 DNA ligase into similarly digested pET30aϩ. The ligation mixture was transformed by electroporation into supercompetent XL1-Blue (Stratagene). Plasmid-containing transformants were selected at 37°C on LB agar plates supplemented with kanamycin (30 g/ml). Putative clones were repurified on LB kanamycin plates. Plasmid DNA was then isolated and screened for the presence of the desired insert by BamHI and NdeI digestion. One positive clone containing the LilpxA gene was designated pLP3 and confirmed by DNA sequencing. Only two silent changes, specific for the Verdun strain, were detected in comparison with the nucleotide sequence of lpxA from the Lai strain.
For preparation of cell extracts and assays, the LilpxA gene of pLP3 was overexpressed in E. coli BL21(DE3)/pLysS (Stratagene). The LilpxA gene was also subcloned from pLP3 into pBluescript II SK(ϩ) using XbaI and XhoI to generate pCS611, as described for pCS421. The LilpxA gene was subcloned into the low copy vector pNGH100 to make pCS623, using SacI and BamHI as described for pCS449.
Preparation of Cell-free Extracts-To prepare E. coli extracts for assays, the constructs were grown at 30°C in 50 ml of LB broth with 100 g/ml ampicillin to A 600 ϭ 0.5, shifted to 18°C for 10 min, induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside, and then grown overnight. The extracts were prepared by passage through a French pressure cell, as described (4). The protein concentrations were quantified using the Pierce bicinchoninic acid assay kit (23) with bovine serum albumin as the standard.
Cell-free extracts and membranes of L. interrogans strain Verdun (avirulent derivative) were prepared from a 4.3-g frozen cell pellet derived from a 10-liter culture (2 ϫ 10 9 bacteria/ml) grown at the Institut Pasteur (Paris, France) (22). All of the steps were conducted at 0 -4°C. The pellet was resuspended in 20 ml of 50 mM HEPES, pH 7.5, containing 10% glycerol. A 5-ml portion of the cell suspension was diluted to 10 ml with the same buffer, and the remaining 15 ml was stored at Ϫ80°C for later use. The cells were broken by three passages through a French pressure cell at 10,000 p.s.i. The lysate was centrifuged at 12,600 ϫ g for 10 min to remove unbroken cells and large debris. The protein concentration of this extract was 6.9 mg/ml (23). A 7-ml portion of the extract was centrifuged at 100,000 ϫ g for 1 h. The soluble fraction was stored at Ϫ80°C, and the membranes were homogenized in ϳ1 ml of 50 mM HEPES, pH 7.5, to yield a final protein concentration of about 12 mg/ml. The membrane suspension was then divided into aliquots and stored at Ϫ80°C.
Substrate Preparation-The [␣-32 P]UDP-GlcNAc and all of the acyl-ACP substrates were prepared as previously described (24). To make [␣-32 P]UDP-GlcNAc3N from [␣-32 P]UDP-GlcNAc, 250 Ci of [␣-32 P] UDP-GlcNAc was dissolved in 138 l of deionized water. This radiolabeled substrate (final concentration, 1.5 M) was incubated with an additional 48.5 M UDP-GlcNAc carrier, 50 M HEPES, pH 8.0, 100 mM L-glutamic acid, 1 mM NAD ϩ , and 0.5 mg/ml pCS355 cell-free extract (4) in a reaction volume of 200 l for 2.5 h at 30°C. About 50% of the [␣-32 P]UDP-GlcNAc is converted to [␣-32 P]UDP-GlcNAc3N under these conditions. This reaction was then chilled on ice and diluted 3-fold with cold 100% ethanol. After 10 min, the debris was removed by centrifugation. The supernatant was diluted 4-fold with cold water.
Site-directed Mutagenesis of LilpxA-Using pLP3 as the template, the QuikChange site-directed mutagenesis kit (Stratagene) was used to generate the H120A substitution, which corresponds to the H125A mutation in E. coli (11). The structure of the mutated plasmid, designated pLP3-1, was confirmed by DNA sequencing. The pLP3-1 plasmid was then transformed into competent cells of BL21(DE3)/pLysS.

Structural Analysis of Lipid A Species Isolated from Various E. coli Constructs-Lipid
A was isolated from the E. coli lpxA2 mutant RO138 complemented with different sets of genes. The first construct expressed gnnA, gnnB, and AflpxA off of a single hybrid plasmid (RO138/ pCS439) (4), whereas the second construct expressed gnnA, gnnB, and LilpxA on separate hybrid plasmids (RO138/pCS411/pCS623) (Table I).
For preparation of the lipid A, RO138/pCS439 was grown without shaking in 1 liter of LB medium for 48 h at 42°C in the presence of 50 g/ml ampicillin and 12 g/ml tetracycline. A 100-ml culture of RO138/ pCS411/pCS623 was grown with shaking at 250 rpm in LB medium at 30°C. The latter construct was not viable at 42°C. A 100-ml culture of RO138/pCS61 also was grown at 42°C in LB medium with shaking, given that it grows normally and produces normal amounts of wild-type lipid A.
Extraction, hydrolysis at pH 4.5, and purification of lipid A 1,4Јbisphosphate species by DEAE-cellulose chromatography was carried out as previously described (17,26). Prior to mass spectrometry, the purified lipid A was subjected to base hydrolysis by incubation in freshly made chloroform, methanol, 1.7 M NaOH (2:1:0.4, v/v/v) at room temperature (Ϸ25°C) for 2 h, with occasional mixing of the two phases (27). Following hydrolysis, the organic phase was dried under N 2 , and the lipid was redissolved in chloroform, methanol (4:1, v/v). A portion was spotted onto a silica gel 60 TLC plate, developed in chloroform, pyridine, formic acid, water (50:50:16:5, v/v/v/v). After spraying with 10% sulfuric acid in ethanol, the lipids were visualized by charring.
Mass Spectrometry of Base-treated Lipid A Samples-MALDI-TOF mass spectra were acquired in the negative-ion linear modes using a Kratos Analytical (Manchester, UK) MALDI-TOF mass spectrometer, operated with a 337-nm nitrogen laser, a 20-kV extraction voltage, and time-delayed extraction (24).

Selectivity of LpxA Acyltransferases in Extracts of M. loti and
L. interrogans-Of the three organisms used for studying the origin of GlcN3N-containing lipid A molecules, M. loti is easiest to grow (6). As shown in Fig. 2A and Table II, Table II). The specific activity of M. loti LpxA was the highest with 3-hydroxymyristoyl-ACP (Table II), consistent with the reported fatty acid composition of M. loti lipid A (6). Prolonged incubation (Fig. 2B, lanes 10  and 14) resulted in nearly quantitative conversion of [␣-32 P]UDP-GlcNAc3N to acylated product, indicating that Nacylation by M. loti LpxA is thermodynamically favorable. This contrasts with the 3-O-acylation of [␣-32 P]UDP-GlcNAc by E. coli LpxA, which is thermodynamically unfavorable (11,28).
Crude extracts of L. interrogans serovar Icterohemeorrhagiae (strain Verdun) were prepared from frozen cells (22) and   (Table II). Prolonged incubation resulted in complete conversion of 10 M [␣-32 P]UDP-GlcNAc3N to product, as with the M. loti extracts. The quantification of LpxA specific activities observed in various cell extracts with different combinations of substrates is summarized in Table II. Unexpectedly, E. coli LpxA was found to acylate UDP-GlcNAc3N at a slightly faster rate than its natural substrate UDP-GlcNAc (10 M acceptor and donor substrates).
Mild Alkaline Hydrolysis of Acylated [␣-32 P]UDP-GlcNAc3N-To confirm that LpxA acylates [␣-32 P]UDP-GlcNAc3N on the nitrogen atom at the pyranose 3-position, a portion of the LpxA product generated with either E. coli or L. interrogans LpxA was subjected to mild alkaline hydrolysis (27). In contrast to [␣-32 P]UDP-3-O-acyl-GlcNAc synthesized by E. coli LpxA, which is deacylated by a 30-min exposure to dilute NaOH at room temperature (27), the acylated [␣-32 P]UDP-GlcNAc3N generated either by E. coli or L. interrogans LpxA was unaffected. These observations, together with the thermodynamically favorable acylation seen with UDP-GlcNAc3N versus UDP-GlcNAc, are consistent with the formation of an amidelinked acyl chain at the 3-position of the pryanose ring of UDP-GlcNAc3N (Scheme 1 in Ref. 4).
Cloning of LpxA Orthologs from L. interrogans and A. ferrooxidans-The lpxA genes of L. interrogans (21) and A. ferrooxidans were identified by probing their genomes (www.ncbi.nlm.nih.gov/BLAST/) with the E. coli LpxA sequence (9). One 780-bp gene, designated LilpxA, encodes a predicted protein of 259 amino acids that displays 41% amino acid identity and 59% similarity with E. coli LpxA (21), with an E value of about 4 ϫ 10 Ϫ51 in a pair-wise comparison (29). The nearly identical lpxA gene from L. interrogans strain Verdun (22) was cloned by PCR from its genomic DNA, inserted into pET30aϩ, and expressed in E. coli BL21(DE3)/pLysS. The lpxA gene of A. ferrooxidans encodes a protein of 260 amino acid residues with 49% amino acid identity, 69% similarity with an E value Ϸ 3 ϫ 10 Ϫ71 when compared with E. coli LpxA (29). This gene was first cloned into pET23cϩ and also into pBluescript II SK(ϩ), pNGH100, or pCS411, depending on the experiment (see below). A ClustalW alignment of EcLpxA, AfLpxA, and LiLpxA is shown in Fig. 3.
In Vitro Assays of the Cloned LpxA Orthologs-The pET vector constructs harboring either EclpxA, AflpxA, or LilpxA were expressed in E. coli BL21 (DE3)/pLysS, and the extracts were assayed for their ability to acylate either UDP-GlcNAc or UDP-GlcNAc3N (Table III). Extracts of all three constructs demonstrated significant overexpression of acyltransferase activity with at least one combination of acyl-ACP donor and sugar nucleotide acceptor, when compared with the vector controls (Table III). In the case of recombinant L. interrogans LpxA, acylation was not measurable for any combination except UDP-GlcNAc3N and 3-hydroxylauroyl-ACP, consistent with the assays shown in Table II. The recombinant A. ferrooxidans enzyme, although selective for UDP-GlcNAc3N, displayed low but measurable activity with UDP-GlcNAc (Table III). The recombinant AfLpxA displayed relaxed acyl donor chain length selectivity (Table III), when compared with E. coli or L. interrogans LpxA.
The specific activity of the recombinant E. coli enzyme with 3-hydroxymyristoyl-ACP and UDP-GlcNAc is in agreement with published data (10,11). EcLpxA is ϳ100-fold selective for the 3-hydroxymyristoyl-ACP over 3-hydroxylauroyl-ACP or 3-hydroxypalmitoyl-ACP (Table III) (10). Cloned E. coli LpxA showed robust activity with UDP-GlcNAc3N (Table III) and retained the same pattern of 3-hydroxyacyl-ACP chain length selectivity as with UDP-GlcNAc (Table III). No LpxA protein tested showed measurable activity with decanoyl-ACP, lauroyl-ACP, myristoyl-ACP, or palmitoyl-ACP (data not shown).  (11). In many but not all bacteria that make lipid A containing GlcN3N, the lysine 76 equivalent is substituted with glycine. Lysine 76 is in the vicinity of the E. coli LpxA active site and may contribute to substrate binding (11). However, the G71K substitution in LiLpxA had no effect on sugar nucleotide selectivity (not shown).

Site-directed Mutagenesis of L. interrogans LpxA-Previous
studies demonstrated that His 125 of E. coli is required for catalysis (11), perhaps as a general base to activate the glucosamine 3-OH group of UDP-GlcNAc during acyl chain transfer (Fig. 4). His 125 is conserved and corresponds to His 120 in L. interrogans (Fig. 3, lower arrow). Substitution of His 120 with alanine inactivates the L. interrogans LpxA (Fig. 5), suggesting a similar function as in E. coli, despite the differences in acceptor and donor selectivity.
Partial Complementation of E. coli RO138 (lpxA2) by gnnA, gnnB, and/or AflpxA-The temperature-sensitive E. coli lpxA2 mutant RO138 (12,13) was transformed with hybrid plasmids harboring genes for the biosynthesis of UDP-GlcNAc3N and/or an LpxA ortholog specific for UDP-GlcNAc3N. The plasmids pCS411 (gnnA gnnB), pCS449 (low copy AflpxA), and pCS623 (LilpxA) did not complement RO138 at 42°C, but pCS421 (high copy AflpxA) by itself was effective. The latter observation suggests that the AfLpxA acyltransferase activity seen with UDP-GlcNAc as the acceptor substrate (Table III) is sufficient to restore the growth of RO138 at 42°C provided the AfLpxA protein is expressed at a high level. The transformation of RO138/pCS411 (gnnA gnnB) with pCS449 (low copy AflpxA) likewise did not rescue the temperature-sensitive phenotype of RO138, suggesting that there was not enough expression of AfLpxA from pCS449. To address this issue, RO138 was transformed with the pBluescript-derived high copy plasmid pCS439 (gnnA gnnB AflpxA). This strain grew slowly on LB plates or liquid medium at 42°C, provided the culture was not shaken.
Biosynthesis of GlcN3N-containing Lipid A Molecules in FIG. 4. Possible role of His 125 in the catalytic mechanism of E. coli LpxA. In our previously published model (11), N ␦1 of histidine 125 (in the alternative tautomeric representation from the one shown above) was proposed to function as the general base. However, the His 125 side chain of LpxA was flipped in the published x-ray structure (9) (S. Roderick, personal communication). With the revised conformation, N ⑀2 of His 125 is proposed to activate the glucosamine 3-OH group in the acceptor substrate, and the conserved Asp 126 side chain would then be in a good position to orient and/or stabilize the His 125 residue by hydrogen bonding the N ␦1 proton.  RO138 Complemented with gnnA, gnnB, and AflpxA-Lipid A samples from RO138/pCS61 (EclpxA) and RO138/pCS439 (gnnA gnnB AflpxA) grown without shaking at 42°C were isolated and purified by ion exchange chromatography on DEAE cellulose. Each lipid A preparation was subjected to mild alkaline hydrolysis to remove all of the ester-linked acyl chains. TLC analysis of the hydrolyzed samples suggested that a portion (ϳ30%) of the RO138/pCS439 (gnnA gnnB AflpxA) lipid A contained three base-resistant acyl chains (Fig. 6A, Band I) under conditions that degrade all of the RO138/pCS61 (EclpxA) lipid A to a single, diacylated derivative (Fig. 6B, Band II).
MALDI-TOF Mass Spectrometry of the Base-hydrolyzed Lipid A Samples-To confirm the identities of the base-treated lipid preparations, MALDI-TOF mass spectrometry was performed in the negative mode. Hydrolyzed lipid A from RO138/ pCS61 (EclpxA) gave rise to a single peak at m/z 951.8 (Fig.  7A), interpreted as [M Ϫ H] Ϫ of a diacylated lipid A 1,4Јbisphosphate (Fig. 7A), formed during base hydrolysis by the loss of the four ester-linked acyl chains that are normally present in native E. coli lipid A (Fig. 1).
The base-treated lipid A from RO138/pCS439 (gnnA gnnB AflpxA) grown at 42°C likewise shows a major peak at m/z 952.1 (Fig. 7B), interpreted as [M Ϫ H] Ϫ of the same diacylated lipid A species seen with the RO138/pCS61 (EclpxA) lipid A (Fig. 7A). However, a second peak is present at m/z 1177.4 (Fig.  7B), which would be expected for [M Ϫ H] Ϫ of a lipid A 1,4Ј bisphosphate species with three N-linked hydroxymyristoyl chains (Fig. 7B). The latter would be formed by mild base hydrolysis of lipid A molecules consisting of one glucosamine residue and one GlcN3N unit. The negative mode data (Fig. 7B) do not reveal the extent to which the GlcN3N unit is incorporated into the proximal or distal positions. DISCUSSION LpxA is an essential cytoplasmic enzyme in E. coli that catalyzes the first step of lipid A biosynthesis according to the reaction: UDP-GlcNAc ϩ R-3-hydroxymyristoyl-ACP 3 UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc ϩ ACP (3, 8, 11, 30). The crystal structure of E. coli LpxA at 2.6-Å resolution suggests that the enzyme is homotrimer in which the active sites are situated between adjacent subunits (9 -11). Each LpxA monomer is constructed around an unusual left-handed parallel ␤-helix, which is conserved in all LpxA orthologs and in many other bacterial acetyl-and acyltransferases (9,(31)(32)(33). The crystal structure of E. coli LpxA has not been determined in the presence of bound substrates or substrate analogs, but sitedirected mutagenesis has demonstrated that histidine 125 is crucial for activity and that adjacent basic residues may contribute to substrate binding ( Fig. 4) (11). The recent structure of Helicobacter pylori LpxA with a detergent molecule bound at the proposed active site supports the above conclusions (34).
Although E. coli and related LpxA proteins have been characterized as UDP-N-acetylglucosamine 3-O-acyltransferases (8,28,35), their sugar nucleotide specificity has not been examined in depth (8). As proposed in Scheme 1 of the preceding manuscript (4), lipid A biosynthesis in bacteria that make GlcN3N-substituted lipid A molecules might start with the conversion of UDP-GlcNAc to the analog UDP-GlcNAc3N, followed by selective N-acylation catalyzed by special LpxA orthologs present in those organisms. The results shown in Fig. 2   FIG. 7. Negative ion mode MALDI-TOF mass spectrometry of base-hydrolyzed lipid A from RO138/pCS61 and RO138/pCS439. A 10-fold excess of RO138/pCS61 (EclpxA) lipid A (A) was hydrolyzed, when compared with RO138/ pCS439 (gnnA gnnB AflpxA) lipid A (B), to ensure that complete hydrolysis of all ester-linked acyl chains was taking place under all experimental conditions. The proposed structures of the mild alkaline hydrolysis products are indicated.
and Tables II and III demonstrate that LpxA orthologs from M. loti, L. interrogans, and A. ferrooxidans do indeed utilize UDP-GlcNAc3N in strong or absolute preference to UDP-GlcNAc.
L. interrogans and M. loti make lipid A molecules based exclusively on a ␤-1Ј,6-linked GlcN3N disaccharide (6, 7), whereas A. ferrooxidans lipid A contains both GlcN3N and GlcN units (5). The LpxA with the most stringent substrate specificity is that of L. interrogans (Tables II and III). In agreement with the absence of glucosamine in the lipid A backbone of this organism (7), LiLpxA does not utilize UDP-GlcNAc at a measurable rate with any acyl donor substrate (limit of detection Ͻ2 pmol/min/mg of extract). Furthermore, LiLpxA does not acylate UDP-GlcNAc3N with any donor other than 3-hydroxylauroyl-ACP. Although most other LpxA proteins display a high degree of specificity for a particular acyl chain length, they do in fact function with alternative acyl donors at slow rates (Tables II and III) (8,17,24,35,37).
A. ferrooxidans LpxA displays some activity with all of the 3-hydroxyacyl-ACP donors tested and with both sugar nucleotides (Table III). M. loti cell extracts (Fig. 2) and the Bordetella pertussis and B. bronchiseptica LpxA orthologs (24) are likewise very relaxed with regard to acyl chain length selectivity. In the bordetellae, structural studies indicate that the acyl chains at the 3-and 3Ј-positions of lipid A can be different under different conditions, explaining the need for a relaxed LpxA ortholog (38 -40). Structural characterization of A. ferrooxidans lipid A shows no such heterogeneity (5); however, this work was done with A. ferrooxidans IFO 14262 and may not reflect the exact lipid A structure of A. ferrooxidans ATCC 23270 from which our AflpxA gene was cloned. In RO138/ pCS439, grown at 42°C (Fig. 7B), mainly 3-hydroxymyristate was incorporated into the lipid A 3-and 3Ј-positions by AfLpxA. AfLpxA may be more selective in vivo, or more R-3-hydroxymyristoyl-ACP may be available inside the cells.
The composition of the lipid A synthesized by AfLpxA in living cells may be affected by the relative sizes of the UDP-GlcNAc and the UDP-GlcNAc3N pools. Despite reports of a mixed lipid A backbone in A. ferrooxidans (5) and our demonstration of a mixed lipid A composition in RO138/pCS439 (Figs. 6A and 7B), AfLpxA is Ϸ100-fold selective for UDP-GlcNAc3N over UDP-GlcNAc when assayed in vitro at 10 M of each donor and acceptor substrate (Table III). Consequently, one could argue that a 100:1 concentration ratio of UDP-GlcNAc to UDP-GlcNAc3N might be needed in vivo to yield a mixed lipid A backbone containing equal amounts of GlcN3N and GlcN, as suggested in the literature (5). Because GlcN3N is roughly one-sixth of the total lipid A sugar in RO138/pCS439 (Fig. 7B), the in vivo ratio of UDP-GlcNAc to UDP-GlcNAc3N present in this E. coli construct might be Ϸ300:1.
The specific activity of AfLxpA-expressing E. coli extracts (36 pmol/min/mg) was significantly lower than that of the empty vector control (51 pmol/min/mg), when assayed with UDP-GlcNAc and 3-hydroxymyristoyl-ACP (Table III). Apparent suppression of the background chromosomal activity by heterologous overexpression of foreign LpxA orthologs has been noted previously (10,17). This phenomenon may be due to the formation of inactive heterotrimers between EcLpxA and the foreign LpxA proteins. Because there was no measurable chromosomal activity with any acyl donor other than 3-hydroxymyristoyl-ACP in the vector control extracts under these assay conditions (Table III), the low but significant activity seen with UDP-GlcNAc when AfLpxA-expressing extracts were assayed with C-12 and C-16 hydroxyacyl-ACPs (Table III) likely represents true AfLpxA catalytic function.
E. coli LpxA has robust activity with UDP-GlcNAc3N and can tolerate the presence of some GlcN3N in its lipid A, as in RO138/pCS439 grown on plates or in nonshaking liquid culture medium at 42°C (Fig. 7B). However, the complementation of the lpxA2 phenotype was not complete. Very small colonies formed on plates, and the maximal A 600 reached in nonshaken liquid medium was only Ϸ0.2. It may be that lipid A substituted with GlcN3N is toxic to E. coli. Whatever the explanation for the slow growth of RO138/pCS439 at 42°C, it might yet be possible to substitute E. coli lipid A completely with GlcN3N, either by using alternative GnnA, GnnB, and/or LpxA orthologs or by introducing second site suppressor mutations into constructs like RO138/pCS439. Engineered strains containing GlcN3N-based lipid A molecules might be useful for the preparation of new endotoxin antagonists (41,42) or novel vaccines.