Accumulation of the lipid A precursor UDP-2,3-diacylglucosamine in an Escherichia coli mutant lacking the lpxH gene.

The lpxH gene encodes the UDP-2,3-diacylglucosamine-specific pyrophosphatase that catalyzes the fourth step of lipid A biosynthesis in Escherichia coli. To confirm the function of lpxH, we constructed KB21/pKJB5. This strain contains a kanamycin insertion element in the chromosomal copy of lpxH, complemented by plasmid pKJB5, which is temperature-sensitive for replication and harbors lpxH(+). KB21/pKJB5 grows at 30 degrees C but loses viability at 44 degrees C, demonstrating that lpxH is essential. CDP-diglyceride hydrolase (Cdh) catalyzes the same reaction as LpxH in vitro but is non-essential and cannot compensate for the absence of LpxH. The presence of Cdh in cell extracts interferes with the LpxH assay. We therefore constructed KB25/pKJB5, which contains both an in-frame deletion of cdh and a kanamycin insertion mutation in lpxH, covered by pKJB5. When KB25/pKJB5 cells are grown at 44 degrees C, viability is lost, and all in vitro LpxH activity is eliminated. A lipid migrating with synthetic UDP-2,3-diacylglucosamine accumulates in KB25/pKJB5 following loss of the covering plasmid at 44 degrees C. This material was converted to the expected products, 2,3-diacylglucosamine 1-phosphate and UMP, by LpxH. Pseudomonas aeruginosa contains two proteins with sequence similarity to E. coli LpxH. The more homologous protein catalyzes UDP-2,3-diacylglucosamine hydrolysis in vitro. The corresponding gene complements KB25/pKJB5 at 44 degrees C, but the less homologous gene does not. The accumulation of UDP-2,3-diacylglucosamine in our lpxH mutant is consistent with the observation that the lipid A disaccharide synthase LpxB, the next enzyme in the pathway, cannot condense two UDP-2,3-diacylglucosamine molecules, but instead utilizes UDP-2,3-diacylglucosamine as its donor and 2,3-diacylglucosamine 1-phosphate as its acceptor.

Lipid A structures resembling the one found in E. coli are synthesized by many other Gram-negative bacteria (1,4,12). In general, diverse Gram-negative bacteria harbor single copies of the key genes that encode the enzymes for lipid A biosynthesis (4) (Fig. 1, see accompanying article (1)). Interestingly, only about 50% of all Gram-negative organisms sequenced to date lack clear orthologs of LpxH (Table II), formerly designated ybbF in E. coli (13). However, weak LpxH orthologs with E values ranging from 0.003 to 0.2 (designated LpxH2 below) do exist in some bacteria lacking LpxH, including Agrobacterium tumefaciens (14,15), Caulobacter crescentus (16), and Sinorhizobium meliloti (17) (Table I). Orthologs of both LpxH and LpxH2 are present in Pseudomonas aeruginosa (18) and Ralstonia solanacearum (19). Some Gram-negative bacteria (20) lack orthologs of both LpxH and LpxH2 (Table I).
To confirm the function of lpxH in E. coli, we now describe the construction of two E. coli mutant strains deficient in lpxH. KB21/pKJB5 harbors a kanamycin insertion mutation in the lpxH gene and is complemented by a plasmid carrying a temperature-sensitive origin of replication and lpxH ϩ . KB25/pKJB5 also bears an in-frame deletion of the chromosomal copy of cdh. At 44°C, pKJB5 is lost, and the lpxH gene and protein product disappear. Under these conditions, both mutant constructs lose viability, demonstrating the requirement of lpxH for E. coli growth, and the inability of cdh to compensate for loss of lpxH. The precursor UDP-2,3-diacylglucosamine accumulates from undetectable levels to about 10% of the total phospholipid in KB25/ pKJB5 at 44°C, providing further genetic validation for the lipid A pathway (Fig. 1, see accompanying article (1)). To evaluate the function of lpxH2, crude extracts of E. coli cells that overexpress P. aeruginosa LpxH or LpxH2 (18) were assayed for UDP-2,3diacylglucosamine hydrolase activity. The complementation of the lpxH-deficient E. coli mutant strains was also investigated. Only P. aeruginosa LpxH catalyzes UDP-2,3-diacylglucosamine hydrolysis in vitro and can substitute for E. coli lpxH in vivo. The function of LpxH2 remains unknown. Bacteria lacking LpxH presumably contain a distinct UDP-2,3-diacylglucosaminespecific pyrophosphatase that remains to be identified.

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) TLC plates were from EM Separation Technology, Merck. 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 TLC 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 II. Cells were grown at 37°C in LB medium, containing 10 g/liter tryptone, 5 g/liter yeast extract, and 10 g/liter NaCl (21). Cells harboring the temperature-sensitive plasmids pMAK705 (22,23), pKO3 (24), or their derivatives were grown in LB medium at 30°C, unless otherwise stated. Antibiotics were used, as needed, in the concentrations of 100 g/ml ampicillin, 30 g/ml chloramphenicol, 20 g/ml kanamycin, 30 g/ml streptomycin, and 12 g/ml tetracycline.
General Recombinant DNA Techniques-Most recombinant DNA techniques were performed as described by Sambrook et al. (26). E. coli W3110 chromosomal DNA was isolated according to Ausubel et al. (27). P. aeruginosa PAO1 chromosomal DNA was isolated using the Easy DNA Kit (Invitrogen). Plasmids were prepared using the QIAprep Spin Miniprep Kit (Qiagen). Restriction enzymes, T4 DNA ligase, and shrimp alkaline phosphatase were utilized according to the manufacturer's instructions. The Qiaex II Gel Extraction Kit was used to extract DNA from agarose gels (Qiagen). CaCl 2 -competent cells were utilized for the transformation of plasmid DNA (26).
Plasmids for Analysis of E. coli lpxH-All plasmids used in this study are listed in Table II. The pKJB3 and pKJB4 plasmids were constructed in the following manner. pKJB2 (1), a pET21a ϩ cloning vector (Novagen) carrying the lpxH gene, was digested with NsiI to linearize the 5.3-kb plasmid. The lpxH gene contains an internal NsiI restriction site. pUC4K (Amersham Biosciences) was digested with PstI to generate the 1.2-kb kanamycin cassette. The linearized pKJB2 plasmid and the kanamycin cassette were electrophoresed using a 1% agarose gel and extracted from the gel using the Qiagen Gel Extraction Kit. pKJB2 was dephosphorylated by shrimp alkaline phosphatase and ligated together with the kanamycin cassette using T4 DNA ligase at 16°C overnight. The ligation reactions were then transformed into competent E. coli XLIBlue (Stratagene) cells, and colonies resistant to ampicillin and kanamycin were selected. Plasmid DNA was isolated from these colonies using the QIAprep Spin Mini-Prep Kit (Qiagen). The plasmids were digested with XbaI and BamHI to excise the 1.9-kb lpxH::kan cassette. The digestion reactions were electrophoresed using a 1% agarose gel to identify those plasmids containing the insert. The desired plasmid was designated pKJB3. pKJB3 and pMAK705 (22) were digested with XbaI and BamHI. The lpxH::kan cassette and pMAK705 were extracted after gel electrophoresis. pMAK705 was dephosphorylated and ligated together with the lpxH::kan cassette. The ligation reactions were transformed into XL1Blue, and colonies resistant to chloramphenicol and kanamycin were selected. Plasmids were isolated from these colonies and digested with XbaI and BamHI to verify the presence of the correct insert. The resulting plasmid is pKJB4.
The lpxH gene was cloned into the low copy vector pNGH1-amp (23,28) to yield pKJB6. PCR was employed to amplify lpxH from W3110 chromosomal DNA. Pfu DNA polymerase (Stratagene) was used according to the manufacturer's specifications. The sequence of the forward primer introducing a SacI site is 5Ј-GGG GCC GAG CTC ATG GCG ACA CTC TTT ATT GC-3Ј. The sequence of the reverse primer, which introduces a BamHI site, is 5Ј-GGC GGG GAT CCT TTA AAA GGG AAA ATG-3Ј. The SacI site includes the start codon for lpxH, whereas the BamHI site is immediately downstream of the stop codon. PCR was electrophoresed using a 1% agarose gel and extracted from the gel. Both the PCR product and pNGH1-amp were digested with SacI and BamHI. The digestion reactions were electrophoresed on a 1% agarose gel and extracted from the gel. The plasmid was dephosphorylated. The digested lpxH PCR product and the dephosphorylated vector were ligated together; the ligation reactions were transformed into competent XL1Blue cells, and colonies resistant to ampicillin were selected. Plasmid DNA was isolated from these colonies, digested with SacI and BamHI, and analyzed by gel electrophoresis to identify those plasmids containing the lpxH insert. The desired plasmid was designated pKJB6, and its sequence was confirmed.
Plasmid for Inactivation of E. coli cdh-The cdh replacement vector, pKJB102, was constructed in the following manner. To assemble the insert cloned into pKO3 (Table II), crossover PCR was employed, as illustrated by Link et al. (24). In the first step, two PCRs were performed using Pfu DNA polymerase, supplemental MgCl 2 , and R477 (29) chromosomal DNA as a template. Primer A and primer B were used in PCR 1, whereas primer C and primer D were used in PCR 2. Primer A has a 5Ј-BamHI site and anneals to a region 500 bp upstream of cdh. Its sequence is 5Ј-AAG AGA GGA TCC TGT CGA GCG CGG AAT-3Ј. Primer B includes a 3Ј-end containing the first 18 bp of cdh and a 5Ј-end with a 21-bp region designed to complement the 5Ј-end of primer C. Its sequence is 5Ј-GGG TAG GTG ATA TGC ATT TGT AAG ACC CGC TTT TTT CAT-3Ј. Primer C includes a 3Ј-end containing the last 36 bp of cdh and a 5Ј-end with a 21-bp region designed to complement the 5Ј-end of primer B. Its sequence is 5Ј-ACA AAT GCA TAT CAC CTA CCC GAA ATT CAG GAT CAT CAG TGT GAG ATT TTG CGT TAA-3Ј. Primer D anneals to a region 500 bp downstream of cdh and has a 5Ј-end BamHI site. Its sequence is 5Ј-CGG CGC GGA TCC TGA TCG CGA AAA AAT-3Ј. The PCR products were gel-purified. The 1-kb cdh deletion fragment was then assembled by PCR under standard conditions with 200 nM each of primers A and D, 1 l each of PCR products 1 and 2, and an additional 2 mM MgCl 2 . The PCR product and pKO3 were digested with BamHI and gel-purified. The PCR product was ligated into the dephosphorylated vector; the ligation reactions were transformed into competent XL1Blue cells, and colonies resistant to chloramphenicol were selected. Plasmids were isolated from these colonies and digested with BamHI to identify those containing the insert. The desired plasmid was designated pKJB102.
Plasmids Harboring Pseudomonas lpxH and lpxH2-The expression vector pKJB130 was constructed as follows. PCR was used to amplify lpxH from P. aeruginosa chromosomal DNA PAO1 with Pfu DNA polymerase. The sequence of the forward primer introducing an NdeI site is 5Ј-CCC CCC CAT ATG AGC GTC CTG TTC ATC-3Ј. The sequence of the reverse primer introducing a BamHI site is 5Ј-AAA GGA GGA TCC TCA GAG CGG GAA GGG-3Ј. The PCR was gel-electrophoresed, and the lpxH gene was extracted from the gel. Both the PCR product and pET21a ϩ were then digested with NdeI and BamHI. The digestion reactions were gel-extracted. The plasmid was dephosphorylated and ligated together with the digested PCR product. The ligation reactions were then transformed into competent XL1Blue cells, and colonies resistant to ampicillin were selected. The plasmids were digested with NdeI and BamHI and electrophoresed to identify those plasmids containing the lpxH insert. The desired plasmid was designated pKJB130, and its sequence was confirmed.
The expression vector pKJB133, a pET21a ϩ vector harboring the P. aeruginosa lpxH2 gene, was constructed as described above, with modification. The sequence of the forward primer introducing an NdeI site is 5Ј-CCA CAA CAT ATG AGC ACC GCG CAA CGC-3Ј. The sequence of the reverse primer introducing a BamHI site is 5Ј-AAA GAA GGA TCC TCA TGC CGC CGG CTC-3Ј.
The low copy plasmid pSK1 was constructed according to the following procedure. PCR was used to amplify lpxH from P. aeruginosa chromosomal DNA PAO1 with Pfu DNA polymerase. The sequence of the forward primer introducing an NcoI site is 5Ј-CTC GAG CCA TGG CAA TGA GCG TCC TGT TCA TC-3Ј. The sequence of the reverse primer, which introduces a BamHI site, is 5Ј-GCG CCC TTC CCG CTC TGA GGA TCC AAG CTT-3Ј. The PCR was gel-electrophoresed, and the lpxH gene was extracted from the gel. Both the PCR product and pNGH1-amp were then digested with NcoI and BamHI. The digestion reactions were gel-extracted. The vector was dephosphorylated and ligated together with the digested PCR product. The ligation reactions were then transformed into competent XL1Blue cells, and colonies resistant to ampicillin were selected. The plasmids were digested with NcoI and BamHI and electrophoresed to identify those containing the lpxH insert. The desired plasmid was designated pSK1, and its sequence was confirmed.
The low copy plasmid pSK2 was constructed as described above, with modification. The sequence of the forward primer, which introduces an NcoI site, is 5Ј-AAG CTT CCA TGG CAA TGA GCA CCG CGC AAC GC-3Ј. The sequence of the reverse primer, which introduces a BamHI site, is 5Ј-GTC GAG CCG GCG GCA TGA GGA TCC CAT ATG-3Ј.
Construction of KB22, an E. coli cdh In-frame Deletion Mutant-A cdh in-frame deletion in E. coli R477 was constructed according to the methods of Link et al. (24), Hamilton et al. (22), and Garrett et al. (23), with minor modifications. pKJB102 was transformed into E. coli R477 competent cells, and colonies resistant to chloramphenicol were selected for at 30°C. One colony was used to inoculate a 25-ml culture grown to an A 600 of 0.7 in LB containing chloramphenicol. Next, 1 ϫ 10 5 cells were plated onto prewarmed plates containing chloramphenicol and incubated at 44°C. Cells with the plasmid integrated into the chromosome were selected at this step. A single colony was used to inoculate 1 ml of LB containing chloramphenicol, which was then grown for 6 h at 30°C. The culture was transferred to a flask containing 100 ml of LB with chloramphenicol, and this was grown to stationary phase overnight at 30°C. In a fraction of the cells, the integrated plasmid is excised from the chromosome under these conditions (22), yielding a plasmid carrying either cdh ϩ or ⌬cdh. Cells were plated on chloramphenicol at 30°C to obtain 50 cells/plate. Sixty-four colonies were restreaked onto two LB agar plates containing chloramphenicol. Both plates were incubated overnight, one at 30°C and the other at 44°C. Seventeen colonies did not grow at 44°C, indicating that plasmid excision had occurred in these strains. The plasmids were isolated from these strains grown at 30°C, digested with BamHI, and analyzed by gel electrophoresis to determine the size of the insert. Four plasmids contained the 1.7-kb insert harboring cdh ϩ , establishing that ⌬cdh has replaced the wild type gene in these strains. One such strain was designated KB22/pKJB103 (pKJB103 is the excised plasmid). Growth on LB agar containing streptomycin at 44°C overnight cured KB22/ pKJB103 of the plasmid, confirming the non-essential nature of cdh. The resulting strain was designated KB22.
Construction of KB21/pKJB5 and KB21/pKJB6 Harboring Chromosomal lpxH::kan Insertions-pKJB4 was transformed into E. coli R477 competent cells, and colonies resistant to chloramphenicol and streptomycin were selected for at 30°C. R477/pKJB4 was used to create the chromosomal lpxH::kan insertion mutation, KB20/pKJB5 (Table II), as described above, with slight modification. A 1-ml portion of LB containing chloramphenicol and streptomycin was inoculated with the cointegrant and grown at 30°C for 7 h. The entire culture was used to inoculate 25 ml of LB containing chloramphenicol and streptomycin, which was then grown to an A 600 of 0.3. The cells were plated on LB containing chloramphenicol and kanamycin to yield 500 cells/plate. Eighty colonies were restreaked on two LB agar plates containing chloramphenicol. One plate was incubated overnight at 30°C, and the other was incubated overnight at 44°C. Eleven colonies were temperature-sensitive at 44°C. The plasmids from these strains were isolated  (21). The resistance of the strain to tetracycline and sensitivity to UV light verified the recA Ϫ phenotype (27). The resulting strain is designated KB21/pKJB5 (Table II). The low copy vector pKJB6 was transformed into competent KB21/ pKJB5 cells and selected for at 30°C on LB agar containing ampicillin and chloramphenicol. Single colonies were then restreaked on LB agar containing ampicillin and incubated at 44°C, curing the strain of the temperature-sensitive plasmid, pKJB5. The resulting strain is KB21/pKJB6. R477/pKJB5 was constructed by transforming pKJB5 into competent R477 cells and selecting for colonies at 30°C on LB agar containing chloramphenicol. R477/pKJB6 was constructed by transforming pKJB6 into competent R477 cells and selecting for colonies at 37°C on LB agar containing ampicillin.
Construction of KB25/pKJB5 Deficient in Both lpxH and cdh-pKJB5 was transformed into competent KB22 cells, and colonies of KB22/pKJB5 were selected for at 30°C on LB agar containing chloramphenicol. The lpxH::kan cassette from KB21/pKJB5 was then transduced into KB22/pKJB5, yielding KB23/pKJB5. This procedure was performed according to Miller (21), and the incubation temperature for all strains was 30°C. KB21/pKJB5 lysates were made by infecting the strain with P1 vir bacteriophage (7 ϫ 10 9 plaque-forming units/ml; generously provided by R. Webster, Duke University). Transduction was confirmed by growth of KB23/pKJB5 on LB plates containing chloramphenicol and kanamycin at 30°C.
KB22/pKJB6 was constructed by transforming pKJB6 into competent KB22 cells and selecting for colonies at 37°C on LB agar containing ampicillin and streptomycin.
KB25/pKJB6 was constructed by transforming pKJB6 into competent KB25/pKJB5 cells and selecting for colonies at 30°C on LB agar containing ampicillin and chloramphenicol. Single colonies were restreaked on LB agar containing ampicillin at 44°C, curing the strain of the temperature-sensitive plasmid, pKJB5.
Preparation of Cell-free Crude Extracts of KB25/pKJB5 and KB25/ pKJB6 Grown at 44°C-KB25/pKJB5 and KB25/pKJB6 were grown in 45 ml of LB with appropriate antibiotics (see Table II) to mid-log phase at 30°C. Portions of each culture were diluted to an A 600 of 0.005 into a flask containing 360 ml of LB (prewarmed to 30°C) and a flask containing 400 ml of LB (prewarmed to 44°C). The 360-ml cultures were grown at 30°C, whereas the 400-ml cultures were grown at 44°C. When the cultures grown at 30°C reached an A 600 of 0.2-0.3, the cells were collected by centrifugation at 10,000 ϫ g for 15 min at 4°C. Each cell pellet was washed with 36 ml of cold 50 mM HEPES, pH 7.4, and the centrifugation was repeated. Each cell pellet was then resuspended in 2 ml of the same buffer. Whenever the 44°C growth of KB25/pKJB6 reached an A 600 of 0.2-0.3, a 40-ml portion was diluted into 360 ml prewarmed LB, and the culture was grown at 44°C. The remaining 360 ml was harvested as described above. This process was repeated until the cells had been growing 8 h after the initial dilution. When the 44°C growth of KB25/pKJB5 reached an A 600 of 0.2-0.3 after 3 h, 40-ml portions were diluted into 3 flasks containing 360 ml of prewarmed LB, and the cultures were grown at 44°C. The remaining 280 ml was harvested as described above. Each time a portion of the KB25/pKJB6 culture grown at 44°C was harvested, one culture containing KB25/ pKJB5 grown at 44°C was also harvested. This method was utilized because KB25/pKJB5 loses viability after ϳ4.5 h and therefore would yield few cells after repeated dilutions.
To prepare cell-free crude extracts of each sample, cells were broken by one passage through a French pressure cell at 18,000 pounds/square inch, and the suspension was centrifuged at 4,000 ϫ g for 15 min at 4°C. Concentrations of soluble protein were determined using the Bradford assay (Bio-Rad).
Extraction and Detection of an Unknown Lipid Accumulating at 44°C in KB25/pKJB5-Single colonies of KB25/pKJB5 and KB25/ pKJB6 were inoculated into separate 10-ml cultures of LB with the appropriate antibiotics (Table II). The cultures were grown to an A 600 of 0.2-0.3 at 30°C. Each culture was then diluted to an A 600 of 0.01 into two flasks containing LB broth. One culture containing KB25/pKJB6 was grown at 44°C, whereas the other was grown at 30°C (both flasks held 25 ml of LB). One culture containing KB25/pKJB5 and 125 ml of LB was grown at 44°C, whereas the other holding 25 ml of LB was grown at 30°C. When the cultures grown at 30°C reached an A 600 of 0.15, 32 P i (5 Ci/ml) was added, and the cells were grown to an A 600 of 0.5. The cultures grown at 44°C were diluted 10-fold into fresh prewarmed media whenever the A 600 was 0.2-0.3. When the cumulative growth yield reached 0.9 for KB25/pKJB5 (the point at which its growth begins to slow) and 5.0 for KB25/pKJB6, the cultures were labeled with 32 P i (5 Ci/ml). KB25/pKJB5 was labeled for 3 h, and KB25/pKJB6 was labeled for 2 h. The three 25-ml cultures were each split evenly into 2 ϫ 15-ml conical tubes and centrifuged at 7000 rpm in the clinical centrifuge. Each cell pellet was resuspended in 1.5 ml of LB broth, and corresponding samples were pooled. A 1-ml portion of each was added to three separate tubes, and they were centrifuged at 7000 rpm in the clinical centrifuge. The pellets were frozen at Ϫ80°C for further analysis. The 125-ml culture was centrifuged at 3,000 ϫ g and resuspended in 10 ml of LB broth. Next, two portions of 1.5 ml and one portion of 7 ml were added to glass tubes, and they were centrifuged at 7000 rpm in the clinical centrifuge. The pellets were frozen at Ϫ80°C for further analysis.
To analyze the lipid profiles of each strain, 0.5 ml of a single phase Bligh-Dyer solution (30), consisting of chloroform/methanol/water (1:2: 0.8, v/v), was added to the pellets from the above 1-or 1.5-ml samples. The samples were vortexed and centrifuged at 5000 rpm for 15 min in the clinical centrifuge. The supernatant was centrifuged again to eliminate any cell debris. Approximately 40,000 cpm of each sample and 5,000 cpm of a [␤-32 P]UDP-2,3-diacylglucosamine standard were spot-
ted onto a 20 ϫ 20-cm Silica Gel 60 TLC plate. The spots were dried under a cold air stream. The plate was developed in chloroform/pyridine/formic acid/water (50:50:16:5, v/v) and dried again under a stream of cold air. The plate was exposed to an Amersham Biosciences Phos-phorImager screen overnight at room temperature. The screen was analyzed using ImageQuant software (Amersham Biosciences).
Isolation of the Unknown Lipid Accumulating at 44°C in KB25/ pKJB5-A single colony of KB25/pKJB5 was used to inoculate 20 ml of LB broth containing chloramphenicol, and the culture was grown to mid-log phase. A portion of the culture was diluted to an A 600 of 0.01 in 200 ml of LB broth, which was then grown at 44°C. When the A 600 reached 0.2-0.3, the culture was diluted 10-fold into fresh prewarmed LB. Once the cumulative growth yield reached 0.7, the culture was labeled with 32 P i (5 Ci/ml) and grown for 3 h. The cells were centrifuged at 4,000 ϫ g; the pellet was resuspended in 10 ml of LB, and the sample was transferred to a glass tube. It was then centrifuged in the clinical centrifuge at 7000 rpm for 15 min, and the pellet was frozen at Ϫ20°C for further analysis.
The pellet was resuspended in 1 ml of a single phase Bligh-Dyer solution (30), consisting of chloroform/methanol/water (1:2:0.8, v/v). The sample was incubated at room temperature for 25 min, vortexed, and centrifuged at 7000 rpm for 15 min. The supernatant was transferred to a glass tube, and the pellet was washed with 1 ml of a single phase Bligh-Dyer solution. The two supernatants were combined and converted to a two-phase Bligh-Dyer system (30), consisting of chloroform/ methanol/water (2:2:1.8, v/v). The sample was vortexed and centrifuged for 5 min at 5000 rpm. The upper phase (which would contain any UDP-2,3-diacylglucosamine that might be present) (10) was transferred to another tube, reduced in volume under a stream of N 2 to 300 l, and then applied in a thin line to a 10 ϫ 20-cm Silica Gel 60 TLC plate. The line was dried under 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. The radiolabeled lipid was isolated according to the procedure routinely used for making the [␤-32 P]UDP-2,3-diacylglucosamine substrate for LpxH, as described previously (1)).
Expression of E. coli LpxH, P. aeruginosa LpxH, and P. aeruginosa LpxH2-Five-ml cultures of BLR(DE3)pLysS/pKJB130, BLR(DE3)-pLysS/pKJB133, and BLR(DE3)pLysS/pKJB2 were grown overnight at 37°C. Portions of each were used to inoculate 250-ml cultures at an A 600 of 0.01. These were grown at 37°C. When the A 600 of each culture reached 0.6, a 50-ml portion was transferred to a 250-ml Erlenmeyer flask, and IPTG was added to the remaining 200 ml at a final concentration of 1 mM. The cells were grown for an additional 3 h. The cultures were harvested by centrifugation at 4000 ϫ g for 15 min at 4°C. The cell pellets were each washed with 20 ml of 50 mM HEPES, pH 7.4, collected again by centrifugation, and resuspended in 8 ml of the same buffer.  (Table II). The cultures were diluted into prewarmed LB (without antibiotics) to an A 600 of 0.1 and grown at 44°C. To maintain logarithmic growth, the cells were diluted 10-fold into fresh prewarmed LB medium whenever they reached an A 600 of 0.2-0.3. The A 600 values of the cultures were measured every 30 min to determine their cumulative growth yields at 44°C. B, the cumulative growth yields at 44°C of E. coli strains KB24/pKJB5, KB24/pKJB6, KB25/pKJB5, and KB25/ pKJB6 were examined as described above.
FIG. 3. Plating efficiency of KB25/pKJB5 does not increase at 44°C in liquid culture. Strains KB24/pKJB5, KB24/pKJB6, KB25/ pKJB5, and KB25/pKJB6 were grown at 44°C to maintain logarithmic growth, as described in Fig. 2. Portions of the cultures were diluted and plated at various times on LB agar with or without antibiotics, as indicated. The plates were incubated overnight at 30°C, and the colonyforming units were determined.
FIG. 4. UDP-2,3-diacylglucosamine hydrolase activity disappears in extracts of KB25/pKJB5 after growth at 44°C. KB25/ pKJB5 and KB25/pKJB6 were grown at 44°C to maintain logarithmic growth as described in Fig. 2. Portions of the cells were harvested at intervals throughout the experiment, and the cell-free crude extracts were assayed for UDP-2,3-diacylglucosamine hydrolase activity under standard, linear conditions for up to 45 min (1). The final protein concentration in reaction mixtures containing cell-free crude extracts of KB25/pKJB5 was 0.15 mg/ml, whereas the reactions incubated with extracts of KB25/pKJB6 had a final protein concentration of 0.5 mg/ml. Cells were broken by one passage through a French pressure cell at 18,000 p.s.i. To obtain cell-free crude extracts, the suspensions were centrifuged at 4,000 ϫ g for 15 min. The supernatants were stored at Ϫ80°C. Protein concentrations were determined using the Bradford assay (Bio-Rad).

UDP-2,3-diacylglucosamine Hydrolase Activity in Extracts of Wild Type Cells and cdh Mutant KB22-In vivo experiments
designed to characterize the phenotype of an E. coli lpxH mutant can be carried out using KB21/pKJB5 (Table II), a strain with a chromosomal lpxH kanamycin insertion mutation covered by a temperature-sensitive plasmid carrying lpxH ϩ . However, to measure LpxH activity in cell extracts following loss of the lpxH gene, a strain bearing mutations in both cdh and lpxH must be used. Cdh encodes an enzyme catalyzing the same conversion of UDP-2,3-diacylglucosamine to 2,3-diacylglucosamine 1-phosphate and UMP as LpxH in vitro (10), and therefore a high background activity exists when extracts of KB21/pKJB5 are assayed. Cdh does not play a role in lipid A biosynthesis, given that it is a non-essential gene and its active site likely faces the periplasm (9,11).
To circumvent this problem, an E. coli strain harboring a cdh in-frame deletion was first constructed (24), as described under "Experimental Procedures." The resulting cdh mutant is designated KB22. To confirm the loss of Cdh activity in this strain, cell-free crude extracts were prepared and assayed for the conversion of [␤-32 P]UDP-2,3-diacylglucosamine to [ 32 P]2,3-diacylglucosamine 1-phosphate. As shown in Fig. 1, hydrolysis of UDP-2,3-diacylglucosamine was markedly reduced in KB22 compared with its parental strain R477. However, the somewhat less robust LpxH activity was still measurable. Hydrolase activity in extracts of KB22 was comparable with that of JB1104 (11), a different strain with a cdh::Tn10 mutation (Fig. 1).
For subsequent studies of LpxH function in cell extracts, KB22 was transformed with pKJB5, transduced with the lpxH::kan allele, and then converted by transduction to recA Ϫ . The resulting strain is KB25/pKJB5 (Table II).
Temperature-sensitive Growth of KB21/pKJB5 and KB25/ pKJB5 on LB Agar-KB21/pKJB5 (Table II) was used to examine the requirement of lpxH for cell viability. Strain KB21/ pKJB6 (Table II) contains the same chromosomal lpxH insertion mutation but is covered by a non-temperature-sensitive plasmid (23) carrying lpxH ϩ . Both strains grow at the same rate at 30°C on LB agar containing kanamycin and tetracycline. However, at 44°C, only single colonies of KB21/ pKJB6 appear after overnight incubation (data not shown). These results demonstrate that LpxH is required for E. coli growth. Moreover, Cdh does not compensate for the loss the LpxH.
Strains KB25/pKJB5 and KB25/pKJB6 (Table II) are the same as KB21/pKJB5 and KB21/pKJB6, respectively, but also harbor a cdh in-frame deletion. These constructs are viable at 30°C on LB plates containing kanamycin and tetracycline, but only KB25/pKJB6 grows at 44°C (data not shown).
Temperature-sensitive Growth of KB21/pKJB5 and KB25/ pKJB5 in Liquid Medium-A more quantitative experiment to examine the viability of KB21/pKJB5 at 44°C in LB broth was carried out in parallel with the controls KB21/pKJB6, R477/ pKJB5, and R477/pKJB6 (Table II). Each construct was grown overnight at 30°C and then diluted to an A 600 of 0.01 into prewarmed LB broth lacking antibiotics. The cultures were grown at 44°C with shaking at 250 rpm. To maintain logarithmic growth, the cultures were diluted 10-fold into fresh prewarmed LB whenever the A 600 reached 0.2-0.3. Strains KB21/ pKJB6, R477/pKJB5, and R477/pKJB6 grew logarithmically for the duration of the experiment (10 h), but the growth of KB21/pKJB5 slowed at about 4.5 h and stopped after 6 h at 44°C (Fig. 2A), confirming the essential nature of lpxH.

FIG. 5. A lipid migrating with a [␤-32 P]UDP-2,3-diacylglu-
cosamine standard accumulates in 32 P i -labeled KB25/pKJB5 after growth at 44°C. KB25/pKJB5 and KB25/pKJB6 were grown at 30°C. A portion of each culture was diluted into 2 flasks containing LB medium. One culture was grown at 30°C, and the other was shifted to 44°C. As described under "Experimental Procedures," the cultures were labeled for several hours with 32 P i after an initial period of growth at 44°C, and the cells were harvested. The lipids were extracted from the cells and analyzed by TLC, followed by PhosphorImager analysis. Only KB25/pKJB5 grown at 44°C accumulates a lipid that migrates with synthetic [␤-32 P]UDP-2,3-diacylglucosamine.
FIG. 6. Conversion of the unknown lipid present in KB25/ pKJB5 at 44°C to 2,3-diacylglucosamine 1-phosphate and UMP by LpxH. The lipid accumulating in KB25/pKJB5 labeled with 32 P i after growth at 44°C was isolated as described under "Experimental Procedures." It was tested as a substrate for LpxH at 2000 cpm/nmol under assay standard conditions (1). Cell-free crude extracts of BLR(DE3)pLysS/pKJB2 (1) and BLR(DE3)pLysS/pET21a ϩ induced with IPTG were added to the reaction mixtures as the source of LpxH to obtain a final protein concentration of 0.05 mg/ml. The reactions were incubated at 30°C for 30 min. As a control, synthetic [␤-32 P]UDP-2,3diacylglucosamine was incubated in parallel with the same cell-free crude extracts, as described above. The reactions were spotted onto a thin layer plate, which was developed in chloroform/methanol/water/ acetic acid (25:15:4:2, v/v), and visualized with a PhosphorImager. Lanes labeled A represent reactions containing the unknown lipid, and lanes labeled B correspond to reactions including synthetic [␤-32 P]UDP-2,3-diacylglucosamine. The lane labeled C represents a [ 32 P]2,3-diacylglucosamine 1-phosphate standard (1). pKJB6, and KB25/pKJB6 all grew logarithmically for the duration of the experiment (8 h). However, KB25/pKJB5 stopped growing after about 4 h (Fig. 2B), consistent with the phenotype of KB21/pKJB5 ( Fig. 2A). No significant difference in growth phenotype resulted from the loss of cdh.
As shown in Fig. 2A, the R477 parental strain grows somewhat faster than the R477 construct harboring the lpxH::kan insertion mutation. The difference in growth rate probably is due to the fact that R477 is recA ϩ , whereas KB21 is recA Ϫ .
At regular intervals after the temperature shift to 44°C, portions of the KB25/pKJB5, KB25/pKJB6, KB24/pKJB5, and KB24/pKJB6 cultures (Table II) were diluted and plated onto LB agar (Fig. 3). In the case of KB25/pKJB6 and KB24/pKJB6, the LB agar was supplemented with ampicillin. The plates were grown at 30°C overnight, and the colonies were counted, and the number of colony-forming units was determined. The strains harboring chromosomal and/or plasmid-derived lpxH ϩ gained colony-forming units in parallel with the optical density for the duration of the experiment (Figs. 2 and 3). However, the number of colony-forming units of KB25/pKJB5 (seen at 30°C) remained relatively constant or declined slightly in the culture shifted to 44°C (Fig. 3), consistent with the inhibition of pKJB5 replication and death of those cells that had lost lpxH. As an additional control, KB24/pKJB5 and KB25/pKJB5 were also plated on LB agar containing chloramphenicol and grown at 30°C (data not shown) to confirm that pKJB5 is lost in these strains at 44°C. These results all confirm that lpxH is required for E. coli viability.
Assays of LpxH Activity in Extracts Prepared from KB25/ pKJB5 Grown at 44°C-As the temperature-sensitive plasmid bearing lpxH ϩ is lost from KB25/pKJB5 after growth at 44°C, the enzymatic activity of LpxH should also be eliminated. To confirm this idea, in vitro hydrolase activity was monitored in extracts of KB25/pKJB5 and KB25/pKJB6 prepared at various times after a shift from 30 to 44°C. As shown in Fig. 4, LpxH activity in KB25/pKJB6 extracts decreases only slightly after a shift to 44°C, from 1 nmol/min/mg at the time of the temperature shift to 0.5 nmol/min for the remainder of the experiment, presumably reflecting a small reduction in plasmid copy number. In KB25/pKJB5 extracts, however, LpxH activity decreases from ϳ2 nmol/min/mg to undetectable levels between 3 and 5 h at 44°C (Fig. 4). The loss of LpxH activity in KB25/ pKJB5 extracts grown under non-permissive conditions provides further evidence that lpxH is the structural gene for the enzyme.
Accumulation of a Novel Lipid in KB25/pKJB5 at 44°C-As shown in Fig. 5, a typical glycerophospholipid profile (consisting of cardiolipin, phosphatidylethanolamine, and phosphatidylglycerol) (31) is present in both KB25/pKJB5 and KB25/ pKJB6 grown at 30 or 44°C, as judged by labeling with 32 P i . However, in KB25/pKJB5 at 44°C, an additional lipid migrating with a UDP-2,3-diacylglucosamine standard accumulates to about 10% of the total chloroform-soluble material, consistent with the idea that LpxH is the specific pyrophosphatase required for lipid A biosynthesis in E. coli. This material also accumulates in KB21/pKJB5 at 44°C (data not shown), demonstrating that Cdh does not have full access to UDP-2,3diacylglucosamine in living cells, despite its efficient cleavage of this intermediate in cell-free extracts (10) (Fig. 1).
The Lipid That Accumulates in KB25/pKJB5 at 44°C Is a Substrate for LpxH-To confirm the identity of the lipid that accumulates in KB25/pKJB5 at 44°C as UDP-2,3-diacylglucosamine, this material was purified and treated with cell-free extracts of BLR(DE3)pLysS/pKJB2 (Fig. 6). pKJB2, a pET21a ϩ vector harboring E. coli lpxH, overexpresses LpxH activity about 540-fold in BLR(DE3)pLysS cells upon induction with IPTG (1). As a control, the unknown lipid was also treated with BLR(DE3)pLysS/pET21a ϩ crude extracts (Fig. 6). Furthermore, chemically synthesized [␤-32 P]UDP-2,3-diacylglucosamine (1) was treated with the extracts of these two strains in parallel (Fig. 6). The material isolated from the cells is completely converted to compounds migrating with [ 32 P]2,3diacylglucosamine 1-phosphate and [ 32 P]UMP, the expected products of LpxH hydrolysis, in the reaction with the BLR(DE3)pLysS/pKJB2 extracts (Fig. 6), whereas the standard only generates [ 32 P]2,3-diacylglucosamine 1-phosphate, consistent with the way it is synthesized (1). The identity of UMP was confirmed by thin layer chromatography using polyethyleneimine-cellulose plates (48). The vector controls catalyze very little hydrolysis of either the standard or the unknown under the conditions employed (Fig. 6).  Enzymatic Activity Assays of P. aeruginosa LpxH Orthologs Expressed in E. coli-Two orthologs of E. coli LpxH appear to be present in P. aeruginosa. The amino acid sequences are aligned in Fig. 7. The protein designated P. aeruginosa LpxH is 46% identical and 61% similar to E. coli LpxH and has an E value of 10 Ϫ59 . The protein designated P. aeruginosa LpxH2 contains a 119-amino acid region that is 28% identical and 39% similar to E. coli LpxH and has an E value of 0.009. P. aeruginosa LpxH and LpxH2 are 23% identical and 39% similar to each other and have an E value of 10 Ϫ4 over 215 amino acids.
To determine whether these orthologs are both functional UDP-2,3-diacylglucosamine hydrolases, PCR was used to isolate lpxH and lpxH2 from PAO1 genomic DNA. The genes were then subcloned into the T7lac expression vector pET21a ϩ . The plasmids, designated pKJB130 and pKJB133, were transformed into E. coli BLR(DE3)pLysS cells. Upon addition of IPTG, lpxH and lpxH2 are transcribed off the T7lac promoter by T7 RNA polymerase. Cell-free crude extracts of these strains were prepared, and protein overexpression was confirmed by SDS-PAGE analysis (data not shown). The extracts were then assayed for hydrolase activity using E. coli UDP-2,3-diacylglucosamine as the substrate. As shown in Table III, extracts of cells overproducing P. aeruginosa LpxH (BLR(DE3)pLysS/ pKJB130) exhibit 2 orders of magnitude more UDP-2,3-diacylglucosamine hydrolase activity than extracts of uninduced cells, whereas extracts of the cells overproducing P. aeruginosa LpxH2 (BLR(DE3)pLysS/pKJB133) have hydrolase activity comparable with the uninduced cells. Furthermore, the activity present in LpxH2-containing extracts is not stimulated by addition of MgCl 2 , CaCl 2 , MnCl 2 , Triton X-100, or EDTA (data not shown). These results suggest that LpxH is likely the functional UDP-2,3-diacylglucosamine hydrolase involved in P. aeruginosa lipid A biosynthesis, whereas LpxH2 is not.
The viability of these four strains was further examined by growing them at 44°C in LB liquid medium under the conditions described in Fig. 2. Strains KB25/pKJB5/pKJB6 and KB25/pKJB5/pSK1 continued to grow logarithmically for the duration of the experiment at 44°C (8.5 h) (Fig. 8). However, the growth of KB25/pKJB5/pSK2 and the vector control strain, KB25/pKJB5/pNGH1-amp, stopped after about 4 h at 44°C (Fig. 8), consistent with the plating experiments.

DISCUSSION
The lpxH gene encodes a specific UDP-2,3-diacylglucosamine hydrolase in E. coli (Fig. 1, see accompanying article (1)). To determine whether or not LpxH is required for lipid A biosynthesis, we constructed the E. coli mutant KB21/pKJB5 (Table  I). KB21 contains a kanamycin cassette inserted into the chromosomal copy of lpxH. The plasmid pKJB5 harbors lpxH ϩ and features a temperature-sensitive origin of replication (22). KB21/pKJB5 grows on nutrient medium at 30°C but not at 44°C when the plasmid carrying lpxH ϩ is lost ( Fig. 2A). We conclude that lpxH is essential in E. coli. The results are consistent with the requirement for E. coli viability of the other genes that encode enzymes catalyzing the early steps of lipid A biosynthesis (4). The chromosomally encoded Cdh (9), which can catalyze the same reaction as LpxH in vitro (10), apparently cannot compensate for the absence of LpxH in vivo.
To determine whether 20-fold overexpression of cdh can compensate for the loss of lpxH, a pNGH1-amp (23) derived plasmid harboring cdh was transformed into KB21/pKJB5 (data not shown). The resulting strain still did not grow at 44°C, indicating that even a 20-fold excess of Cdh does not hydrolyze sufficient UDP-2,3-diacylglucosamine to generate 2,3-diacylglucosamine 1-phosphate and UMP in vivo. Because the primary sequence of Cdh has a membrane-spanning segment at its N terminus and the rest of the protein is predicted to face the periplasm (9), Cdh may be unable to gain full access to UDP-2,3-diacylglucosamine, which is generated from acyl-ACP and UDP-GlcNAc in the cytoplasm. Conceivably, Cdh mutants that are inappropriately localized to the cytoplasm might function as suppressors of lpxH knockouts.
Strain KB25/pKJB5, which is deficient in both cdh and lpxH, FIG. 8. P. aeruginosa lpxH restores growth of KB25/pKJB5 at 44°C, whereas P. aeruginosa lpxH2 does not. E. coli strains KB25/pKJB5/pSK1, KB25/pKJB5/pSK2, KB25/pKJB5/pKJB6, and KB25/pKJB5/pNGH1-amp were grown at 44°C with appropriate back dilution, as described in Fig. 2. was constructed in order to examine the remaining UDP-2,3diacylglucosamine hydrolase activity in cell extracts without interference from Cdh when the lpxH containing covering plasmid is lost at 44°C. Like KB21/pKJB5, KB25/pKJB5 is not viable at 44°C on LB broth, and it stops growing at about 5 h after the temperature shift to 44°C (Fig. 2, A and B). UDP-2,3diacylglucosamine hydrolase activity in cell extracts decreases from about 2 nmol/min/mg to undetectable levels between 3 and 5 h after the temperature shift (Fig. 4), providing further evidence that lpxH is indeed the structural gene for the hydrolase and that enzymes other than LpxH and Cdh do not contribute significantly to UDP-2,3-diacylglucosamine hydrolysis activity in extracts of E. coli.
If LpxH alone catalyzes UDP-2,3-diacylglucosamine hydrolysis in living cells, the metabolite should accumulate in the mutant under non-permissive conditions. To test this, KB25/ pKJB5 grown at 44°C was labeled with 32 P i . The lipids were extracted and analyzed by TLC. As shown in Fig. 5, a lipid migrating with a [␤-32 P]UDP-2,3-diacylglucosamine standard was present in the lpxH cdh-deficient mutant under non-permissive conditions. The unknown compound was isolated and, upon incubation with extracts containing overexpressed LpxH, was cleaved to form [ 32 P]2,3-diacylglucosamine 1-phosphate and [ 32 P]UMP (Fig. 6). These results confirm that UDP-2,3diacylglucosamine does indeed accumulate in E. coli in the absence of LpxH and that LpxB (25,32) is unable to condense two molecules of UDP-2,3-diacylglucosamine directly. Similar results were obtained when KB21/pKJB5 was labeled with 32 P at 44°C (data not shown), validating the idea that Cdh does not have full access to UDP-2,3-diacylglucosamine in vivo.
Although we have identified the hydrolase required for lipid A biosynthesis in E. coli, we must still determine how other Gram-negative bacteria lacking orthologs of E. coli LpxH generate 2,3-diacylglucosamine 1-phosphate. LpxH is conserved with E values less than 10 Ϫ44 among only ϳ50% of the Gramnegative organisms sequenced to date (Table I). This finding is somewhat surprising because the other genes required for the early stages of Kdo 2 -lipid A biosynthesis are present in virtually all Gram-negatives (4). In searching with E. coli LpxH against the microbial data base using the gapped BLASTp algorithm (33), we discovered a family of weak homologs (E values between 0.003 and 0.2), which we designate LpxH2. LpxH2 exists in many organisms lacking LpxH, including A. tumefaciens (14,15), C. crescentus (16), and S. meliloti (17). Helicobacter pylori (34) contains a distinct ortholog that is equally distant from both LpxH and LpxH2 (Table I). Interestingly, LpxH2 is present in P. aeruginosa (18) and R. solanacearum (19), which also contain LpxH.
Whether or not LpxH2 is a functional UDP-2,3-diacylglucosamine hydrolase, or an unrelated protein sharing the phosphoesterase signature sequence, is essential to further our understanding of lipid A biosynthesis. To test if both P. aeruginosa LpxH and LpxH2 catalyze UDP-2,3-diacylglucosamine hydrolysis in vitro, they were overexpressed in E. coli, and the cell-free extracts were assayed for enzymatic activity. Whereas P. aeruginosa LpxH catalyzes the hydrolase reaction at about one-fifth the specific activity of E. coli LpxH under standard assay conditions, P. aeruginosa LpxH2 does not hydrolyze UDP-2,3-diacylglucosamine in vitro (Table III). In addition, we attempted to complement the lpxH insertion mutation with vectors carrying the two P. aeruginosa genes. LpxH compensated for the loss of E. coli LpxH, but LpxH2 did not (Fig. 8).
Although most Gram-negative bacteria contain LpxH and/or LpxH2 orthologs, at least seven organisms appear to lack both proteins (Table I). It will be necessary to study the biochemistry of these systems directly for the presence of additional hydrolase activities in cell extracts and to purify or expressionclone the relevant enzymes.