The Escherichia coli gene encoding the UDP-2,3-diacylglucosamine pyrophosphatase of lipid A biosynthesis.

UDP-2,3-diacylglucosamine hydrolase is believed to catalyze the fourth step of lipid A biosynthesis in Escherichia coli. This reaction involves pyrophosphate bond hydrolysis of the precursor UDP-2,3-diacylglucosamine to yield 2,3-diacylglucosamine 1-phosphate and UMP. To identify the gene encoding this hydrolase, E. coli lysates generated with individual lambda clones of the ordered Kohara library were assayed for overexpression of the enzyme. The sequence of lambda clone 157[6E7], promoting overproduction of hydrolase activity, was examined for genes encoding hypothetical proteins of unknown function. The amino acid sequence of one such open reading frame, ybbF, is 50.5% identical to a Haemophilus influenzae hypothetical protein and is also conserved in most other Gram-negative organisms, but is absent in Gram-positives. Cell extracts prepared from cells overexpressing ybbF behind the T7lac promoter have approximately 540 times more hydrolase activity than cells with vector alone. YbbF was purified to approximately 60% homogeneity, and its catalytic properties were examined. Enzymatic activity is maximal at pH 8 and is inhibited by 0.01% (or more) Triton X-100. The apparent K(m) for UDP-2,3-diacylglucosamine is 62 microm. YbbF requires a diacylated substrate and does not cleave CDP-diacylglycerol. (31)P NMR studies of the UMP product generated from UDP-2,3-diacylglucosamine in the presence of 40% H(2)180 show that the enzyme attacks the alpha-phosphate group of the UDP moiety. Because ybbF encodes the specific UDP-2,3-diacylglucosamine hydrolase involved in lipid A biosynthesis, it is now designated lpxH.

0 show that the enzyme attacks the ␣-phosphate group of the UDP moiety. Because ybbF encodes the specific UDP-2,3-diacylglucosamine hydrolase involved in lipid A biosynthesis, it is now designated lpxH.
The outer surface of the outer membrane of Gram-negative bacteria contains a unique glycolipid known as lipopolysaccharide (LPS) 1 (1)(2)(3)(4). LPS is responsible for the structural integ-rity of the outer membrane and for maintaining a permeability barrier against hydrophobic and large hydrophilic compounds, including many antibiotics (5,6). Lipid A is the hydrophobic portion of LPS that serves as the outer membrane anchor (1,3,4,7). Lipid A (also known as endotoxin) is the active component of LPS responsible for much of the pathophysiology seen during Gram-negative infections, and it is a potent stimulator of the innate immune response of animals (7)(8)(9)(10)(11).
A residual UDP-2,3-diacylglucosamine hydrolase activity distinct from Cdh is present in extracts of Cdh-deficient E. coli. 2 We have now identified the gene encoding this hydrolase by screening the Kohara library (25), a method employed previously (26,27) to identify the structural genes for the lauroyltransferase and the 4Ј-kinase of the lipid A pathway. The Kohara library contains 476 mapped hybrid bacteriophage clones and covers 99% of the E. coli genome (25). An enzymatic activity assay was used to screen lysates prepared with the Kohara clones in an E. coli mutant lacking cdh (17) for overexpression of UDP-2,3-diacylglucosamine hydrolase activity. Of the clones that were assayed, only one not harboring the known cdh gene (18) was found to promote overproduction of UDP-2,3-diacylglucosamine cleavage activity that was 5-10-fold above background. This clone contains DNA from the minute 11.6 to 12.1 region and includes a gene of unknown function, previously designated ybbF, that is present in many Gram-negative bacteria (28). Extracts of cells containing ybbF expressed from the T7 promoter exhibit ϳ540-fold overproduction of hydrolase activity. We describe a partial purification of the overproduced protein (designated LpxH) and the first characterization of its catalytic properties.

EXPERIMENTAL PROCEDURES
Materials-32 P i was purchased from PerkinElmer Life Sciences. Tryptone, yeast extract, and agar were obtained from Difco. Silica gel 60 (0.25 mm) and polyethyleneimine-cellulose thin layer chromatography plates were from EM Separation Technology, Merck. Triton X-100 was purchased from Pierce. Restriction enzymes were obtained from New England Biolabs. T4 DNA ligase, shrimp alkaline phosphatase, dNTPs, and custom-made primers were purchased from Invitrogen. Pfu DNA polymerase was obtained from Stratagene. Solvents used for thin layer chromatography were from Mallinckrodt. All other solvents and chemicals were obtained from Sigma or Aldrich.
Bacterial Strains and Growth Conditions-The bacterial strains used in this study are listed in Table I. Cells were grown at 37°C in Luria broth medium, containing 10 g/liter tryptone, 5 g/liter yeast extract, and 10 g/liter NaCl (29). Antibiotics were used, as needed, in the concentrations of 100 g/ml ampicillin, 30 g/ml chloramphenicol, and 12 g/ml tetracycline.
Preparation of Lipid Substrates-Carrier UDP-2,3-diacylglucosamine was prepared according to Radika and Raetz (30). The [␤-32 P]UDP-2,3-diacylglucosamine was synthesized by coupling [ 32 P]2,3-diacylglucosamine 1-phosphate to UMP-morpholidate (20), with modifications. Briefly, the [ 32 P]2,3-diacylglucosamine 1-phosphate was prepared according to Radika and Raetz (30) at a yield of 2 ϫ 10 7 cpm of [ 32 P]2,3-diacylglucosamine 1-phosphate per mCi of input 32 P i used to label a culture of mutant MN7 (31,32), and the material was dried under a stream of N 2 . The [ 32 P]2,3-diacylglucosamine 1-phosphate was resuspended and dispersed by sonic irradiation in 1 ml of anhydrous pyridine (99.8%; Aldrich), containing 1.5 mg of UMP-morpholidate. After overnight incubation at 37°C, the reaction went to 60 -90% completion. The reaction volume was reduced to 200 l under a stream of N 2 and applied in a thin line to a 20 ϫ 20-cm Silica Gel 60 thin layer chromatography plate. The line was dried with a cold air stream. The plate was developed in chloroform/methanol/water/acetic acid (25:15:4:2, v/v) and dried again under a stream of cold air for about 20 min. The plate was then exposed to autoradiography film (Kodak BioMax MR Film) for 15 min to locate the [␤-32 P]UDP-2,3-diacylglucosamine band. The plate was sprayed with a fine mist of water, and the silica containing the [␤-32 P]UDP-2,3-diacylglucosamine was scraped off with a sharp blade. The silica powder was placed into a 10-ml glass pipette that was plugged with glass wool and washed with 20 ml of CHCl 3 . The [␤-32 P]UDP-2,3-diacylglucosamine was then eluted with 12 ml of a single phase Bligh-Dyer mixture (33), consisting of chloroform/ methanol/water (1:2:0.8, v/v), and the effluent was collected in 12 1-ml fractions. The amount of radioactivity in each fraction was monitored with a Geiger counter. The most radioactive fractions (typically the first ten) were pooled, mixed, and centrifuged at 5000 rpm in a clinical centrifuge. The supernatant was separated from the pellet of residual silica particles and placed in a fresh tube. It was then divided equally into three 10-ml-thick walled Pyrex glass tubes, cooled in a dry ice/ acetone mixture, and dried down using a vacuum pump. The lipid in each tube was dispersed in a small volume of 10 mM BisTris, pH 6, to obtain 20,000 cpm/l and stored at Ϫ80°C. Prior to assay it was dispersed by sonic irradiation in a bath apparatus. Approximately 1 ϫ 10 7 cpm of [␤-32 P]UDP-2,3-diacylglucosamine was usually obtained (ϳ50% yield).
Assay of LpxH Activity with [␤-32 P]UDP-2,3-diacylglucosamine-The conversion of [␤-32 P]UDP-2,3-diacylglucosamine to [ 32 P]2,3-diacylglucosamine 1-phosphate was monitored by thin layer chromatography followed by PhosphorImager analysis. The standard reaction mixture contained 25 mM HEPES, pH 8.0, and 100 M [␤-32 P]UDP-2,3-diacylglucosamine (1000 cpm/nmol). The lipid substrates were subjected to sonic irradiation for 2 min prior to their addition to the reaction mixture. Protein in the amount of 0.1-500 g/ml (depending on the extract) was added, and the reaction mixture was incubated for various times at 30°C. The reactions were stopped by spotting portions of the reaction mixtures onto Silica Gel 60 thin layer chromatography plates. The plates were dried under a cold air stream, developed in the solvent chloroform/methanol/water/acetic acid (25:15:4:2, v/v), dried again, and exposed to a Molecular Dynamics PhosphorImager screen overnight at room temperature. The screen was analyzed using ImageQuant software (Amersham Biosciences) for the Macintosh.
General Recombinant DNA Techniques-Most recombinant DNA techniques were performed as described by Sambrook et al. (34). E. coli W3110 chromosomal DNA was isolated according to Ausubel et al. (35). Plasmids were prepared using the QIAprep Spin Miniprep Kit (Qiagen). Restriction enzymes, T4 DNA ligase, and shrimp alkaline phosphatase were used according to the manufacturer's instructions. The Qiaex II Gel Extraction Kit was used to extract DNA from agarose gels (Qiagen). CaCl 2 -treated competent cells were used for the transformation of plasmid DNA (34).
Construction of Plasmid pKJB2-The gene encoding the UDP-2,3diacylglucosamine hydrolase, ybbF (28), was cloned into the pET21a ϩ cloning vector (Novagen). This vector has a T7lac promoter for tight transcriptional control. PCR was used to amplify ybbF from W3110 chromosomal DNA. Pfu DNA polymerase (Stratagene) was used according to the manufacturer's specifications. The sequence of the forward primer, introducing an NdeI site, is 5Ј-GCC CGC GCA TAT GGC GAC ACT CTT TAT TGC-3Ј. The sequence of the reverse primer, which introduces a BamHI site, is 5Ј-GGC GGG GAT CCT TTA AAA CGG AAA ATG-3Ј. The NdeI site includes the start codon for ybbF, whereas the BamHI site is immediately downstream of the stop codon. The PCR product was electrophoresed using a 1% agarose gel. The 723-bp piece was extracted from the gel using the Qiagen Gel Extraction Kit. Both the PCR product and pET21a ϩ were then digested with BamHI and NdeI. The digestion products were electrophoresed on a 1% agarose gel and extracted from the gel using the Qiagen Gel Extraction Kit. The plasmid was then dephosphorylated by shrimp alkaline phosphatase. The digested ybbF PCR product and the dephosphorylated vector were ligated together with T4 DNA ligase at 16°C overnight. The ligation reactions were then transformed into competent E. coli XL1Blue (Stratagene) cells, and colonies resistant to ampicillin were selected. Plasmid DNA was isolated from these colonies using the Qiagen Mini-Prep Kit. The plasmids were digested with NdeI and BamHI and electrophoresed again on a 1% agarose gel to identify those plasmids containing the 723-bp ybbF insert. The desired plasmid was designated pKJB2, and its sequence was confirmed.
Expression of the ybbF Gene Product-The pKJB2 plasmid was transformed into competent cells of E. coli strain BLR(DE3)pLysS (Novagen), and colonies resistant to ampicillin and chloramphenicol were selected. A 5-ml culture of BLR(DE3)pLysS/pKJB2 was grown overnight at 37°C. A portion of this was used to inoculate a 1-liter culture at an A 600 of 0.01. This was grown at 37°C. When the A 600 reached 0.6, a 50-ml portion was transferred to a 250-ml Erlenmeyer flask, and IPTG was added to the remaining 950 ml at a final concentration of 1 mM. The cells were grown for an additional 3 h. The control strain, BLR(DE3)pLysS/pET21a ϩ , was grown in a similar manner. Cells were harvested by centrifugation at 4000 ϫ g for 15 min. The cells were washed with 100 ml of 50 mM HEPES, pH 7.4, and resuspended in 30 ml of the same buffer. Cells were broken by one passage through a French pressure cell at 18,000 pounds/square inch. To obtain cell-free crude extracts, the suspension was centrifuged at 4,000 ϫ g for 15 min. The supernatant was stored at Ϫ80°C. Protein concentration was determined using the Bradford assay (Bio-Rad).
Partial Purification of LpxH-Cell-free crude extracts of BLR(DE3)pLysS/pKJB2 were obtained from 1 liter of cells induced with 1 mM IPTG. The membrane and soluble fractions were isolated from the crude extracts by centrifugation at 150,000 ϫ g for 1 h. The soluble fraction was placed into a fresh tube, and the membrane fraction was resuspended by homogenization in 5 ml of 50 mM HEPES, pH 7.4. The soluble and membrane fractions were centrifuged again under the same conditions. The final cytosol (twice centrifuged) and the washed membranes were stored in aliquots at Ϫ80°C. The washed membranes and the residual membranes recovered from the second centrifugation of the cytosol were resuspended by homogenization in 1.5 ml of 50 mM HEPES, pH 7.4, and stored in aliquots at Ϫ80°C.
A portion of the final cytosol containing 50.4 mg of protein was adjusted to 40 ml to yield a final concentration of 50 mM HEPES, pH 7.5, 0.2% Triton X-100, and 20% glycerol. A 15-ml column (1.5-cm diameter) of Reactive Green 19 (Sigma) was washed at 4°C with 20 column volumes of a buffer containing 50 mM HEPES, pH 7.5, 2.5 M NaCl, 0.2% Triton X-100, and 20% glycerol. The flow rate was 1.5 ml/min. The resin was then equilibrated with 20 column volumes of a solution containing 50 mM HEPES, pH 8, 0.2% Triton X-100, and 20% glycerol at 1.5 ml/min. The cytosol described above was applied to the column at the same flow rate, and the column was washed with 10 volumes of the equilibration buffer. The protein was eluted by applying 5 column volumes each of the above buffer, containing either 0.5, 1.0, or 2.5 M NaCl. Sixty fractions (about 7 ml each) were collected. The peak of enzyme activity was determined by assaying each fraction under standard conditions for the conversion of [␤-32 P]UDP-2,3-diacylglucosamine to [ 32 P]2,3-diacylglucosamine 1-phosphate. In addition, the proteins in the samples were visualized by 12% SDS-PAGE, using the Laemmli buffer system (36). Fractions 40 -44 contained most of the active LpxH and were pooled. The protein concentration was determined using the Pierce BCA Protein Assay Enhanced Standard Protocol. The partially purified protein was stored in aliquots at Ϫ80°C. 31 2 18 O (95%, Isotec) and 2 g/ml partially purified LpxH was incubated at 30°C for 5 h. A second reaction mixture with the same components in 100% H 2 16 O was incubated in parallel. Complete conversion of UDP-2,3-diacylglucosamine to 2,3-diacylglucosamine 1-phosphate and UMP was achieved in both cases. Each reaction was converted to a single phase Bligh-Dyer mixture, consisting of CHCl 3 / CH 3 OH/H 2 O (1:2:0.8, v/v), by addition of 1.5 ml of CHCl 3 and 3.0 ml of CH 3 OH. These were converted into acidic two-phase Bligh-Dyer mixtures by the further addition of 1.5 ml of CHCl 3 , 1.5 ml of H 2 O, and 75 l of concentrated HCl. In an acidic two-phase Bligh-Dyer system, the 2,3-diacylglucosamine 1-phosphate partitions into the lower phase, and the UMP is recovered in the upper phase (20,31). The 2,3-diacylglucosamine 1-phosphate-containing lower phase of each sample was washed with the upper phase of a new pre-equilibrated acidic two-phase Bligh-Dyer mixture and then transferred to a fresh tube. Both samples were then dried under a stream of N 2 gas, 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 UMP-containing upper phase of each sample was washed with the lower phase of a fresh pre-equilibrated acidic two-phase Bligh-Dyer mixture and transferred to a new tube. Both UMP samples were then dried on a vacuum pump, dissolved in 0.6 ml of 100% D 2 O, and transferred into 5-mm NMR tubes. Measurement with a 3-mm pH electrode revealed that the UMP prepared in 40% H 2 18 O and the UMP prepared in H 2 16 O had pD values of 0.98 and 0.53, respectively. The NMR spectra of the UMP samples generated by the LpxH-catalyzed hydrolysis of UDP-2,3-diacylglucosamine were compared with the spectrum of a 5 mM UMP standard (Aldrich) dissolved in 0.6 ml of 100% D 2 O. 31 P and 1 H NMR spectra were recorded on a Varian Unity 500 spectrometer equipped with a Sun Ultra 5 computer and 5-mm Varian probe. 1 H-decoupled 31 P NMR spectra were recorded with a spectral

Screening for Overproduction of UDP-2,3-diacylglucosamine Hydrolase Activity in Kohara Library
Lysates-The lpxH gene was identified by screening Kohara library lysates for overproduction of UDP-2,3-diacylglucosamine activity. The Kohara library consists of 476 mapped hybrid bacteriophage clones covering 99% of the E. coli chromosome (25). Each phage contains a different segment of the E. coli genome and is able to infect E. coli. When the E. coli lysate generated from a phage carrying the gene of interest is assayed for enzymatic activity, a 2-10-fold overproduction of activity is usually observed (26,27).
An E. coli strain suitable for the screening method as applied to lpxH was first constructed. This strain is -sensitive and lacks Cdh (17). The latter gives rise to high background activity in the assay in wild type strains, because it also catalyzes the conversion of [␤-32 P]UDP-2,3-diacylglucosamine to [ 32 P]2,3diacylglucosamine 1-phosphate (17,20). To construct an appropriate cdh-deficient strain, P1 vir transduction was used to move the cdh::Tn10 mutation from the -resistant strain, JB1104 (17), into the -sensitive strain, W3110. The resulting cdh::Tn10 transductant is designated KB1. The Kohara library was then propagated in the KB1 background using the techniques described by Clementz et al. (26). Fresh lysates of KB1 were assayed for UDP-2,3-diacylglucosamine hydrolase activity by monitoring the conversion of [␤-32 P]UDP-2,3-diacylglucosamine to [ 32 P]2,3-diacylglucosamine 1-phosphate. Of the 476 clones analyzed for overexpression of UDP-2,3-diacylglucosamine hydrolase, only one was identified that did not carry the known cdh gene. This clone, 157[6E7], overproduces the enzyme activity 5-10-fold above the background levels present in KB1 (Fig. 2).
Identification of ybbF as the Putative Structural Gene for UDP-2,3-diacylglucosamine Hydrolase-The fragment of the E. coli genome carried by clone 157[6E7] is ϳ20 kb long, contains 19 open reading frames, and spans minutes 11.6 -12.1 (28,37). To determine which gene on this fragment is responsible for the overexpression of the UDP-2,3-diacylglucosamine hydrolase activity, the proteins encoded by the fragment were analyzed for sequence similarity to predicted proteins in Haemophilus influenzae (38)  Expression Cloning and Characterization of E. coli LpxH thetical protein of unknown function unique to many Gramnegative bacteria. The amino acid sequence of YbbF is 50.5% identical to one H. influenzae protein (Fig. 3). YbbF orthologs are not present in Gram-positive bacteria or eucaryotic cells. Some Gram-negative bacteria, including Aquifex aeolicus (40), Chlamydia trachomatis (41), Synechocystis sp. (42), and Rickettsia prowazekii (43), lack ybbF. However, the ybbF gene is the only reasonable candidate for an E. coli UDP-2,3-diacylglucosamine hydrolase, given that other genes on clone 157[6E7] are not restricted to Gram-negatives.
In E. coli, ybbF is 723 bp long and is located upstream of ppiB (peptidyl-prolyl cis-trans isomerase) in an operon containing both genes (28). It encodes a putative 240-amino acid protein with an expected molecular mass of 26,893 daltons. YbbF is not homologous to other E. coli proteins and does not contain membrane-spanning domains.
T7 Promoter Driven ybbF Expression Causes Overproduction of LpxH Activity-PCR was used to isolate ybbF from E. coli, and the gene was subcloned into the T7lac expression vector pET21a ϩ . The plasmid carrying ybbF, designated pKJB2, was then transformed into BLR(DE3)pLysS cells. Upon addition of IPTG, ybbF was transcribed from the T7lac promoter by T7 RNA polymerase. Cell-free crude extracts of this strain were prepared and assayed for UDP-2,3-diacylglucosamine hydrolase. The cells overexpressing ybbF exhibit orders of magnitude more activity than uninduced cells or cells with the vector alone (Fig. 4). The apparent level of overexpression compared with the vector control (ϳ540-fold) is actually an underestimation, because Cdh activity is present in BLR(DE3)pLysS. Nevertheless, these results show that ybbF likely encodes a UDP-2,3diacylglucosamine hydrolase. Because ybbF represents the specific UDP-2,3-diacylglucosamine hydrolase involved in lipid A biosynthesis, as shown in the accompanying article (44), it is renamed lpxH.
Partial Purification of the lpxH-encoded UDP-2,3-diacylglucosamine Hydrolase-The pET21a ϩ vector harboring E. coli lpxH causes UDP-2,3-diacylglucosamine hydrolase activity to be overexpressed about 540-fold in BLR(DE3)pLysS cells upon induction with IPTG (Table II). The tightly controlled T7lac promoter of pET21a ϩ is necessary for successful LpxH overexpression. When lpxH was cloned into pET3a ϩ , a vector containing the leaky T7 promoter, the resulting plasmid could not be transformed into BLR(DE3)pLysS. These results suggest that lpxH overexpression is toxic to cells.
Approximately 60 -70% of the total LpxH activity is recovered in the 150,000 ϫ g membrane-free supernatant fraction in extracts of both wild type and overproducing cells, prepared by disruption through a French pressure cell (Table III). With the addition of 5 mM MnCl 2 to the cell resuspension/homogenization buffer, washed membranes contain 54.4% of LpxH activity, whereas the membrane-free supernatant accounts for 28.8% (Table III). These results suggest that manganese may be important for LpxH activity (see below) and/or aid in its membrane association. We used a membrane-free supernatant prepared without MnCl 2 or added NaCl as the starting material for our purification.
A survey of several dye-ligand binding columns revealed that LpxH binds Reactive Green 19 Resin and elutes with 1-2.5 M NaCl. The addition of 0.2% Triton X-100 to the equilibration, wash, and elution buffers for this column improves the binding and recovery of LpxH. Glycerol is added to the buffers to increase enzyme stability. As shown in Table IV and Fig. 5, an ϳ60% pure preparation of LpxH was obtained with a 13.4-fold purification and a 42.8% overall yield when the enzyme was eluted with 1.0 M NaCl from the Reactive Green 19 column (Table IV), provided that the purified enzyme is preincubated with MnCl 2 . Stimulation by MnCl 2 following purification over Reactive Green suggests that manganese or some other divalent cations stimulate LpxH activity or may be required for catalysis. The Reactive Green 19 column may strip metal ions away from LpxH.
Catalytic Properties of the UDP-2,3-diacylglucosamine Hydrolase-UDP-2,3-diacylglucosamine hydrolase activity was assayed under conditions that were linear with respect to time and protein concentration in the range of pH 5.5-9.5 using Reactive Green 19 purified LpxH (Fig. 6). The pH optimum is 8.0 (Fig. 6B). Ionic strength has little or no effect on hydrolase activity in the range of 0 -200 mM NaCl after preincubation on ice for 10 min in 25 mM HEPES, pH 8.0 (data not shown).
The effect of detergents, like Triton X-100, on LpxH activity was analyzed using the partially purified protein. Fig. 6C demonstrates the striking inhibition of hydrolase activity in the presence of Triton X-100 at concentrations greater than 0.01% (near the critical micelle concentration). This property is unusual because most other E. coli phospholipid synthetic enzymes utilizing diacylated substrates have an absolute requirement for a nonionic detergent for in vitro activity (45)(46)(47), usually in the range of 0.1-0.2%. However, LpxB (the enzyme that follows LpxH in lipid A biosynthesis) (Fig. 1) is also an exception (30). It does not require detergent for activity and is inhibited slightly by Triton X-100 at concentrations greater than 1.0% (30). Inhibition of LpxH and LpxB by Triton may reflect surface dilution of UDP-2,3-diacylglucosamine in mixed micelles (48). In addition, UDP-2,3-diacylglucosamine aggregates are unusually small compared with other lipids, presumably forming micelles in water (30). Other long chain (Ͼ12) diacylated phospholipids usually form bilayers (49). Alternatively, Triton X-100 may bind to hydrophobic patches on the surface of LpxH, thereby affecting its structural stability or limiting access of the substrate to the active site.
Kinetic Studies of the UDP-2,3-diacylglucosamine Hydrolase-UDP-2,3-diacylglucoamine hydrolase activity is linearly TABLE III Localization of UDP-2,3-diacylglucosamine hydrolase activity in cells overexpressing LpxH Cells of strain BLR(DE3)pLysS harboring pKJB2 were subjected to lysis and ultracentrifugation with and without 5 mM MnCl 2 , as described under "Experimental Procedures." UDP-2,3-diacylglucosamine hydrolase-specific activities were determined under standard linear assay conditions. To determine the percentage of hydrolase activity in the various fractions, the total units of each was divided by the sum of the total units.

TABLE IV Partial purification of LpxH from E. coli BLR(DE3)pLysS/pKJB2
LpxH was assayed with protein preincubated on ice for 1 h with 25 mM HEPES, pH 8, containing 5 mM MnCl 2 . After preincubation, the protein was diluted 1:10 into an assay mixture. The cytosolic fraction was assayed at 1 g/ml, and the fractions pooled from the Reactive Green 19 column (fractions 40 -44) were assayed at 200 ng/ml. Omission of MnCl 2 from the preincubation (not shown) had no effect on the activity of the cytosol, but it stimulated the Reactive Green fraction about 3-fold. Taking into account the overproduction of LpxH activity in E. coli BLR(DE3)pLysS/ pKJB2 (Table II), the actual purification compared with wild type cells is estimated as ϳ7.2 ϫ 10 3 -fold.  dependent on time for about 10 min, as shown in Fig. 6A. The reaction goes to completion in the presence of high enzyme concentrations (data not shown). Fig. 7 shows the steady state kinetics of the partially purified LpxH preparation with UDP-2,3-diacylglucosamine as the substrate in the absence of added Triton X-100 other than the small amount (0.0004%) carried over from the enzyme preparation. When the substrate concentration is varied from 5 to 400 M, the apparent K m is 61.7 M, and V max is 17.2 ϫ 10 3 nmol/min/mg. Substrate Specificity and Acyl Chain Requirements of the UDP-2,3-diacylglucosamine Hydrolase-The acyl chain requirement for the catalytic activity of LpxH was explored by assaying for pyrophosphate bond cleavage with [␣-32 P]UDP-GlcNAc or [␣-32 P]UDP-3-O-(R)-3-hydroxymyristoyl-GlcNAc as alternative substrates. These substrates are the unacylated and mono-acylated precursors of UDP-2,3-diacylglucosamine, respectively (13,21,22). The conversion of both [␣-32 P]UDP-GlcNAc and [␣-32 P]UDP-3-O-(R)-3-hydroxymyristoyl-GlcNAc to [␣-32 P]UMP was monitored by thin layer analysis. Neither substrate was utilized by LpxH under standard assay conditions at a protein concentration of 20 g/ml, demonstrating that this enzyme requires two acyl chains for catalysis (data not shown). Furthermore, LpxH does not cleave CDP-diacylglycerol (data not shown). 31 (16,50). The small magnitude of this perturbation requires a very high digital resolution to detect the shift effect. When the hydrolysis reaction is carried out in 40% H 2 18 O, the product that incorporates the water should appear as a doublet in the 31 P NMR spectrum with the smaller component shifted upfield, whereas the product that does not combine with water during hydrolysis gives rise to a singlet 31 P resonance. Fig. 8 shows the 31 P NMR spectra of the UMP preparations generated either in 40% H 2 18 O or 100% H 2 16 O from UDP-2,3diacylglucosamine by LpxH. In the experiments leading to Fig.  8, the UMP and 2,3-diacylglucosamine 1-phosphate products were separated using an acidic two-phase Bligh-Dyer system, as described under "Experimental Procedures." The 31 P NMR spectrum of the UMP product generated in the presence of 40% H 2 18 O appears as a distinct "doublet" near 0.84 ppm with the smaller component upfield, whereas the 31 P NMR spectrum of the UMP product generated in 100% H 2 16 O reveals a sharp singlet resonance at 0.76 ppm. The 31 P NMR spectrum of 2,3-diacylglucosamine 1-phosphate from the H 2 18 O experiment (not shown) reveals a single 31 P NMR resonance near 0.35 ppm. Because the exact position of the 31 P resonance of phosphomonoesters is influenced by slight changes in pH, the small change in the shifts of the UMP products in Fig. 8, A versus B, is presumably due to pH differences in the two Bligh-Dyer isolations, with measured pD values of 0.98 and 0.53 for the 40% 18 O-UMP and 100% 16 O-UMP products, respectively. The 31 P NMR results demonstrate unequivocally that LpxH hydrolysis incorporates H 2 18 O into UMP by catalyzing the attack of water on the ␣-phosphorus atom of UDP-2,3-diacylglucosamine. In this regard it has the same selectivity as demonstrated previously (16) for Cdh.

DISCUSSION
In Gram-negative bacteria like E. coli, nine constitutive enzymes are required to synthesize Kdo 2 -lipid A, the minimal LPS required for growth ( Fig. 1) (1, 4). All but one of the structural genes encoding these enzymes has been reported. The remaining unidentified gene codes for the enzyme, UDP-2,3-diacylglucosamine hydrolase, which catalyzes the conversion of UDP-2,3-diacylglucosamine to 2,3-diacylglucosamine 1-phosphate and UMP (12). We have now identified the gene encoding UDP-2,3-diacylglucosamine hydrolase of E. coli, and FIG. 6. Quantitative assay of purified LpxH, effects of pH, and inhibition by Triton X-100. A, LpxH-catalyzed conversion of UDP-2,3-diacylglucosamine to 2,3-diacylglucosamine 1-phosphate and UMP is linear for 9 min at pH 8.5 with either 100 (open circles) or 200 ng/ml (closed circles) protein purified on the Reactive Green 19 column. At higher enzyme concentrations, the reaction goes to completion (not shown). B, 200 ng/ml Reactive Green-purified LpxH was measured at the indicated pH values in 25 mM MES, pH 5.5, 6.0, or 6.5, 25 mM HEPES, pH 7.0, 7.5, or 8.0, 25 mM EPPS, pH 8.5, or 25 mM CAPSO, pH 9.0 or 9.5. C, 2000 ng/ml Reactive Green 19-purified LpxH was preincubated for 15 min with 0 -1% Triton X-100. Each mixture was then diluted to a final protein concentration of 200 ng/ml in a standard assay mixture containing the same Triton X-100 concentrations as in the preincubation. Similar results (not shown) were obtained when the assay was carried with increasing Triton X-100 concentrations in the assay but with no preincubation. we refer to it as lpxH. We have also purified LpxH to ϳ60% homogeneity and have characterized some of its catalytic properties.
The E. coli lpxH gene was expression-cloned by screening Kohara library lysates (25,26), prepared from a mutant lacking the cdh gene, for a clone that directs the overproduction of UDP-2,3-diacylglucosamine hydrolase activity. Individual lysates were assayed for hydrolase activity by monitoring the conversion of [␤-32 P]UDP-2,3-diacylglucosamine to [ 32 P]2,3diacylglucosamine 1-phosphate using thin layer chromatography. Only one lysate (157[6E7]) that overproduced hydrolase activity was found not to harbor the known cdh gene (17, 18) (Fig. 2). The latter encodes a putative periplasmic enzyme catalyzing the same reaction as LpxH in vitro but is not required for lipid A biosynthesis or cell growth (17,18). The new candidate hydrolase gene ybbF(lpxH) (28,37) from clone 157[6E7], encoding a hypothetical protein of unknown function, was inserted into a T7 overexpression vector. Massive UDP-2,3-diacylglucosamine hydrolase activity was demonstrated when YbbF was overproduced and assayed in vitro (Table II).
The ybbF(lpxH) gene is 723 bp long, is located near minute 12 (28,37), and is not clustered with other lipid A biosynthetic genes (3,4). The lpxH gene is upstream of ppiB in an operon containing these two genes (28). LpxH consists of 240 amino acids with a predicted molecular weight of 26,893 and contains no obvious membrane-spanning domains or signal sequence.
We believe that LpxH is the functional hydrolase involved in E. coli lipid A biosynthesis. When lpxH is cloned behind the T7lac promoter, LpxH activity is overproduced ϳ540-fold (Table II). In addition, its amino acid sequence is well conserved among at least half of the Gram-negative organisms sequenced to date (44) but is absent in Gram-positive bacteria and eucaryotic cells. An insertion mutation defective in lpxH is not viable and accumulates UDP-2,3-diacylglucosamine, as shown in the accompanying article (44). Gram-negative organisms lacking obvious lpxH orthologs include A. aeolicus (40), C. trachomatis (41), Synechocystis sp. (42), and Rickettsia prowazekii (43). Different pyrophosphatase(s) must catalyze UDP-2,3-diacylglucosamine cleavage in these systems.
The subcellular localization of overexpressed LpxH was examined. The first three lipid A biosynthetic enzymes (LpxA, LpxC, and LpxD) are cytosolic (23, 24 ,51), whereas the last four (LpxK, KdtA, LpxL, and LpxM) are membrane-bound (26,(52)(53)(54). LpxB partitions to both the cytosol and the membrane during ultracentrifugation (30,55), and LpxH behaves in a similar manner. Ultracentrifugation of cell-free crude extracts containing overproduced LpxH showed that 66.3% of the hydrolase activity was localized to the membrane-free cytosol and 22.6% was associated with the washed membranes (Table III). Similar results were obtained when the extracts were prepared and ultracentrifuged with up to 200 mM NaCl (data not shown). However, upon addition of 5 mM MnCl 2 to the buffers used to prepare the crude extracts, the protein localization was reversed (Table III). These results suggest that manganese or other divalent cations may be involved in maintaining the structure, membrane association, or active site of LpxH.
A membrane-free cytosol prepared without added manganese ions was used to purify the overexpressed LpxH. A preparation of about 60% purity was obtained after passage over a Reactive Green 19 column (Fig. 5). However, following the purification LpxH was routinely preincubated in 25 mM HEPES, pH 8.0, containing 5 mM manganese chloride, prior to assay of hydrolase activity. This preincubation stimulated the activity of the purified material about 3-fold but had no effect on the cytosolic enzyme (not shown). Under these assay conditions, a 13.4-fold purification of LpxH was achieved with a FIG. 7. Steady state kinetics of LpxH. LpxH activity was assayed under standard linear assay conditions at the indicated UDP-2,3-diacylglucosamine concentrations. A small amount of Triton X-100 (0.0004%) was carried over from the enzyme preparation. Partially purified LpxH from the Reactive Green 19 column was used at 200 ng/ml. The apparent K m for UDP-2,3-diacylglucosamine was 61.7 M, whereas the apparent V max was 1.7 ϫ 10 4 nmol/min/mg. The lines are drawn using a non-linear least squares fitting to the equation: V ϭ (V max [