The envA Permeability/Cell Division Gene of Escherichia coli Encodes the Second Enzyme of Lipid A Biosynthesis

The envA gene of Escherichia coli has been shown previously to be essential for cell viability (Beall, B. and Lutkenhaus, J. (1987) J. Bacteriol. 169, 5408–5415), yet it encodes a protein of unknown function. Extracts of strains harboring the mutant envA1 allele display 3.5– 18-fold reductions in UDP-3- O -acyl- N -acetylglucosamine deacetylase specific activity. The deacetylase is the sec- ond enzymatic step of lipid A biosynthesis. The structural gene coding for the deacetylase has not been as- signed. In order to determine if the envA gene encodes the deacetylase, envA was cloned into an isopropyl-1- thio- (cid:98) - D -galactopyranoside-inducible T7-based expression system. Upon induction, a protein of the size of envA was highly overproduced, as judged by SDS-PAGE. Direct deacetylase assays of cell lysates revealed a con- comitant (cid:59) 5,000-fold overproduction of activity. Assays of the purified, overproduced EnvA protein demon- strated a further (cid:59) 5-fold increase in specific activity. N -terminal amino acid sequencing of the purified residues found in membrane glycerophospholipids. Since the equilibrium constant for UDP-GlcNAc acylation is unfavorable (Anderson et al. , 1993), the second reaction of the lipid A pathway (the deacetylase) appears to function as the first committed step, and it may be regulated. U , uridine; ACP , acyl carrier protein; PtdEtn , phosphatidylethanolamine. and by separated (cid:97) - 32 P]-UDP-GlcNAc and (cid:97) 32 P]-UDP-GlcN the substrate and product, respectively. - 32 P]-UDP-GlcN and counted by liquid in II International). Polyacrylamide Gel Electrophoresis— SDS-Polyacrylamide gel elec- trophoresis was used to identify the overproduced envA gene product and to estimate its molecular weight. Electrophoresis was carried out with the mini-Protean II gel apparatus (Bio-Rad) at 200 V or with the Protean II xi apparatus (Bio-Rad) at 150 V constant voltage, using the discontinuous system of Laemmli (1970). Separation and stacking gels were 13 and 3.2%, respectively.

The outer membrane of Gram-negative bacteria provides a formidable permeability barrier to the entry of large and hydrophobic compounds, primarily because of the unique lipopolysaccharide (LPS) 1 associated with the outer leaflet of the outer membrane (Nakae, 1986;Nikaido, 1976;Nikaido and Vaara, 1985). Over the years, a large number of mutant strains of Escherichia coli and other Gram-negative bacteria that exhibit defects in this barrier have been isolated and characterized (Weigand and Rothfield 1976;Roantree et al., 1977;Scudamore et al., 1979;Coleman and Leive, 1979;Angus et al., 1982;Vaara, 1990;Vuorio et al., 1991). For many of these mutants the molecular defect has not yet been determined.
One well studied example of such mutations is envA1, originally described in the late 1960s by Normark and colleagues (1969). The envA1 mutation was obtained in a study of penicillin resistance by screening ethyl methanesulfonate-treated, penicillinase-producing E. coli for increased ampicillin sensitivity and smooth colony morphology. Strains harboring envA1 exhibit a complex phenotype, including permeability to a wide variety of antibiotics and dyes, suggesting that their EDTAsensitive outer membrane surface layer is defective (Normark, 1970;Normark et al., 1969;Young and Silver, 1991). Mutants bearing envA1 also display a morphological defect in the completion of cell septation, resulting in growth as short chains (Normark et al., 1969;Normark et al., 1971). In addition, there is a slight (ϳ30%) reduction in the content of apparently normal LPS (Grundstrom et al., 1980).
Only the one original allele of the envA locus, the envA1 mutation, has been described (Normark et al., 1969). envA1 has since been used to locate precisely the mutated envA reading frame to minute 2 on the E. coli chromosome at the end of a large gene cluster required for cell division and cell wall biogenesis (Sullivan and Donachie, 1984). The sequence of the envA gene predicts a protein of 34 kDa. The gene itself has been demonstrated to be essential by insertional mutagenesis (Beall and Lutkenhaus, 1987), implying that the envA1 allele encodes a protein that retains some residual activity (Beall and Lutkenhaus, 1987). However, no further hints as to the function of envA have, as yet, been reported (Beall and Lutkenhaus, 1987), and no related sequences of known function from other organisms are available for comparative studies.
LPS is an essential component of the outer membrane and its associated barrier function (Morrison and Ryan, 1992;Rietschel, 1984). The lipid A component of LPS is the basic building block that comprises most of the outer leaflet (Raetz, 1990(Raetz, , 1993. The biosynthesis of lipid A in E. coli has been described (Raetz, 1990(Raetz, , 1993. It begins with the 3-O-acylation of UDP-GlcNAc with R-3-hydroxymyristate (Fig. 1), followed by deacetylation and reacylation at the glucosamine ring nitrogen with a second R-3-hydroxymyristate moiety (Raetz, 1990(Raetz, , 1993Williamson et al., 1991). The genes encoding the UDP-GlcNAc acyltransferases of E. coli have been identified as lpxA and lpxD (firA), respectively (Coleman and Raetz, 1987;Kelly et al., 1993), but the structural gene encoding UDP-3-O-acyl-GlcNAc deacetylase (Anderson et al., 1988) has remained elusive. The deacetylase has been implicated as a point of regulation with respect to the biosynthesis of lipid A Mohan et al., 1994). Identification of the deacetylase structural gene could shed light on the nature of this regulation.
ing the envA1 allele possess diminished levels of UDP-3-O-acyl-GlcNAc deacetylase activity, that T7-based overexpression of the envA reading frame causes a massive induction of deacetylase activity in cell extracts, and that UDP-3-O-acyl-GlcNAc deacetylase activity purifies with the EnvA protein. Thus, envA is an essential gene (Beall and Lutkenhaus, 1987), encoding the second enzymatic step unique to lipopolysaccharide biosynthesis. The intriguing questions relating to the molecular basis for the pleiotropic phenotypes associated with envA1 may now be studied at a biochemical level.  Table I. Strains MB5503, MB5504, LS583, and LS584 have been described previously (Young and Silver, 1991). DH5␣ was purchased from Life Technologies, Inc. Strains BL21(DE3) and BL21(DE3)pLysS were purchased from Novagen. Strains MB4926 and LS822 were created by bacteriophage P1 transduction (Miller, 1972) of MB2884 and BL21(DE3), respectively, with phage grown on a strain carrying a Tn10 element closely linked to the envA gene (Young and Silver, 1991). Strain JB1104 has been described previously (Bulawa and Raetz, 1984).

Materials-Sequencing
Preparation of Cell Extracts-For the purpose of assessing cellular deacetylase activity in various genetic backgrounds, cells were grown as follows.
Cultures of MB5503 and MB5504, inoculated from small overnight cultures into 125 ml of LB broth, were grown at 37°C to late log phase. MB5503 and MB5504 transformed with the pENV15 plasmid were grown in parallel but under selection with 20 g/ml streptomycin. Cells were harvested by centrifugation at 8000 ϫ g for 10 min, washed once in 1 volume of 10 mM sodium phosphate, pH 7.0, and resuspended in 2 ml (62-fold concentration) of the same buffer.
For T7-based deacetylase activity induction, log phase cultures of BL21(DE3)/pLysS alone or in combination with pET11a or pET11a-EnvA, were grown with shaking in LB broth at 37°C. The overnight cultures used for inoculation had been grown with appropriate selection (50 g/ml ampicillin for the pET vectors and 25 g/ml chloramphenicol for pLysS). When A 620 reached 0.15-0.2, the cells were induced by the addition of IPTG to a final concentration of 1 mM. The cells were incubated with shaking for another 2 h. Cells were harvested by centrifugation at 6000 ϫ g for 10 min., washed in 1 ⁄2 volume of 10 mM sodium phosphate, pH 7.0, containing 20% glycerol and 0.2 mM dithiothreitol (DTT), and resuspended in this buffer (containing 0.5 mM DTT) to yield a 25-fold concentrate relative to the original culture.
In each of the above experiments, cell extracts were prepared by one passage of the washed, concentrated cells through a French pressure cell at 18,000 p.s.i. Unbroken cells and debris were removed by centrifugation at 8000 ϫ g for 10 min. Protein concentrations were determined by the method of Bradford (Bio-Rad protein assay), using bovine serum albumin as the standard (Bradford, 1976).
UDP-3-O-acyl-GlcNAc Deacetylase Activity Assay-The substrate [␣-32 P]UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc was prepared and stored as described previously . To assure that product formation was a linear function of protein concentration, enzyme samples were diluted, as appropriate, in a buffer consisting of 10 mM sodium phosphate, pH 7.0, containing 1 mg/ml fatty acid-free bovine serum albumin and 2 mM DTT. Assays were performed in a 600-l microcentrifuge tube at 30°C in a final volume of 20 l with the following components: 3 M [␣-32 P]-UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc(5 Ci/mmol), 40 mM bis-tris, pH 5.5 (or other buffers if indicated), 1 mg/ml fatty acid-free bovine serum albumin, 0.2 mM DTT, and enzyme sample. After 2 and 5 min, a 5-l sample was removed, mixed with 1 l of 5 M KOH in a second microcentrifuge tube in order to stop the reaction, and incubated for an additional 10 min at 30°C to remove ester-linked fatty acids. These samples were neutralized with 2 l of 7.5 M acetic acid, 30% (w/v) trichloroacetic acid (1:1, v/v). The tubes were centrifuged, and 2 l of supernatant was spotted onto a plastic-backed FIG. 1. Relationship of key precursors of fatty acid, lipid A, and peptidoglycan biosynthesis. Three major cell envelope components arise from two key precursors of E. coli metabolism. UDP-GlcNAc serves as the glucosamine source in both peptidoglycan and lipid A biosynthesis. The lipid A pathway begins with the acylation at the 3-OH moiety of the glucosamine ring of UDP-GlcNAc with R-3-hydroxymyristate derived from R-3-hydroxymyristoyl acyl carrier protein. The latter is also the precursor of palmitate residues found in membrane glycerophospholipids. Since the equilibrium constant for UDP-GlcNAc acylation is unfavorable , the second reaction of the lipid A pathway (the deacetylase) appears to function as the first committed step, and it may be regulated. U, uridine; ACP, acyl carrier protein; PtdEtn, phosphatidylethanolamine.
polyethyleneimine cellulose plate (PEI-plate, Merck Darmstadt). The plate was then dried, washed in methanol for 10 min, dried again, and developed in 0.2 M guanidinium hydrochloride. The developed plate was dried and analyzed by autoradiography to locate the separated [␣-32 P]-UDP-GlcNAc and [␣-32 P]-UDP-GlcN spots, derived from the substrate and product, respectively. The [␣-32 P]-UDP-GlcN spot was cut out and counted by liquid scintillation spectrometry in Bio-Safe II liquid scintillation mixture (Research Products International).
Polyacrylamide Gel Electrophoresis-SDS-Polyacrylamide gel electrophoresis was used to identify the overproduced envA gene product and to estimate its molecular weight. Electrophoresis was carried out with the mini-Protean II gel apparatus (Bio-Rad) at 200 V or with the Protean II xi apparatus (Bio-Rad) at 150 V constant voltage, using the discontinuous system of Laemmli (1970). Separation and stacking gels were 13 and 3.2%, respectively.
Construction of pENV15 and pET11a-EnvA-The envA gene of E. coli is known to reside on a 2.5-kilobase EcoRI genomic DNA fragment (Sullivan and Donachie, 1984) and is not stably maintained in multiple copies (Sullivan and Donachie, 1984). For this reason, it seemed likely that successful cloning of this locus might be possible only in a low copy number vector. Based on these observations, we cloned the wild type gene using a strategy that allowed genomic fragments to ligate into either a medium or a low copy number vector. Selection was then performed for both complementation of the EnvA Ϫ phenotypic defect and for the two possible vector products (Fig. 2).
Briefly, this cloning was accomplished by preparing a pool of 2.5kilobase EcoRI fragments from the envA ϩ E. coli strain LS584. This pool was ligated to EcoRI-digested and phosphatase-treated vector pLL24, a hybrid of pBR322 and a pSC101 replicon, joined at their respective EcoRI sites. This ligation mixture subsequently was used to transform the envA Ϫ E. coli strain, LS583. Selection was for complementation of the envA1 allele (growth in the presence of rifampicin at 1 g/ml) in combination with either streptomycin or spectinomycin, each at 5 g/ml (resistance associated with the pSC101 replicon) or ampicillin at 10 g/ml (resistance associated with the pBR322 replicon). A single, low copy, pSC101 origin-containing transformant was obtained. This construct, conferring resistance to rifampicin in an envA Ϫ host as well as streptomycin/spectinomycin resistance, and consisting of the anticipated 2.5-kilobase fragment with the restriction sites expected for the envA region (Sullivan, 1984), was named pENV15.
Construction of the EnvA-overproducing Plasmid pET11a-EnvA-The envA gene product was overproduced using the T7 polymerase expression system of Studier et al. (1990). The envA gene was cloned into the pET11a vector by polymerase chain reaction (PCR) amplification using Pyrococcus furiosus DNA polymerase (Stratagene). Restriction sites were introduced on either side of the coding sequence to facilitate cloning. The N-terminal primer consisted of a G/C clamp, an NdeI restriction site overlapping the initiation codon, and the first 26 base pairs of the envA coding sequence. The sequence of this primer was 5Ј-CCC CGG GCG GCC GCC ATA TGA TCA AAC AAA GGA CAC TTA AAC G-3Ј. The C-terminal primer consisted of a G/C clamp, a BamHI restriction site, two termination codons, and the last 21 base pairs of the envA anticoding sequence. The sequence of this primer was 5Ј-CCG GGC CCT AGG GCC GGA TCC TTA TTA TGC CAG TAC AGC TGA AGG CGC-3Ј. PCR was performed on plasmid template DNA (pENV15) previously linearized by restriction with the restriction enzyme BamHI, which cuts the vector once. PCR reactions contained 200 M dNTPs, 2 ng template DNA, 3 g of each primer, 10% Me 2 SO, and 5 units of P. furiosus DNA polymerase in a buffer of 20 mM Tris-Cl, pH 8.8, 10 mM KCl, 6 mM (NH 4 ) 2 SO 4 , 2 mM MgCl 2 , 0.1% Triton X-100, and 100 g/ml nuclease-free bovine serum albumin (Stratagene) in a 100 l volume. Reactions were first incubated at 95°C for 5 min., after which the Me 2 SO and DNA polymerase were added, and the reaction was incubated at 72°C for 3 min., followed by 30 cycles of denaturation at 95°C for 30 s, primer annealing at 45°C for 30 s, and extension at 72°C for 3 min in a Perkin-Elmer DNA Thermocycler. The resulting PCR product of approximately 900 base pairs was recovered using GeneClean (Bio101) according to the manufacturer's instructions and cleaved with BamHI and NdeI.
The NdeI-BamHI-cut PCR product was ligated at room temperature (ϳ19°C) overnight into similarly cut and gel-purified pET11a (Studier et al., 1990). In order to facilitate the complete digestion of this vector, properly digested pET11a was actually obtained as the fragment released from NdeI and BamHI digestion of pET11a into which the 2415-base pair gyrB gene had been cloned. Upon gel purification, this doubly digested vector was easily separated from any incompletely digested products. This ligation mixture was used to transform DH5␣ cells made competent by the method of Hanahan (1983), and colonies resistant to ampicillin were selected. Plasmid was isolated from a number of putative clones by the method of Holmes and Quigley (1981) and analyzed for the presence of an insert of approximately 900 base pairs. One resulting plasmid with such an insert was designated pET11a-EnvA, and it was used to transform the envA Ϫ T7 polymeraseharboring host LS822 (BL21(DE3), envA1) to ampicillin resistance. The construct pET11a-EnvA proved capable of complementing the phenotypic sensitivity of this envA1 mutant to rifampicin at 1 g/ml and overproduced a protein band of appropriate molecular weight upon induction with 1 mM IPTG. Sequencing of the envA reading frame harbored by this clone by the dideoxy chain-termination method of Sanger (1977) verified the exact match of the cloned sequence to that published previously by Beall and Lutkenhaus (1987).
harvested by centrifugation beginning 1.5 h after addition of IPTG. A total of 73.5 g of wet cell paste was resuspended in 365 ml of buffer (10 mM sodium phosphate, pH 7.0, 2 mM DTT), blended to yield a smooth paste, and rapidly frozen in liquid nitrogen in four centrifuge bottles. The cells were lysed and brought to an ammonium sulfate fraction in two batches. In each case, this was accomplished by thawing the cell paste on ice, followed by addition of Brij 58 to approximately 0.05% (w/v) and incubation in ice water until a viscous mixture was obtained as a result of autolysis mediated by the endogenous T7 lysozyme (encoded by the pLysS plasmid). This mixture was clarified to generate a cell-free extract by centrifugation in a Ti-45 rotor (Beckman) at 40,000 rpm for 60 min. Deacetylase activity was precipitated by the addition of 0.3 g of solid ammonium sulfate/ml of cell-free extract (approximately 50% saturation) over a 45-min period with a further incubation on ice with stirring for 20 min. The ammonium sulfate precipitate was collected by centrifugation in a Sorvall SS-34 rotor at 14,000 rpm for 20 min at 4°C. The ammonium sulfate pellets were resuspended in buffer A (25 mM imidazole, pH 7.0, 20% glycerol, 2 mM DTT) and dialyzed against this buffer at 4°C for 5 h, at which point the conductivity had decreased to a level appropriate for further chromatography. This fraction (40.5 ml, 2552 mg of protein) was rapidly frozen in two portions, which were further purified separately on a Q-Sepharose Fast Flow column.
Q-Sepharose Chromatography-All chromatographic steps were carried out at 4°C using a Pharmacia Bio-Pilot system (controller, LCC-500plus; valve, IMV-7; pump, P-6000; detector, Pharmacia LKB UV-1 280 nm single wavelength). A column (Pharmacia XK50, 5.0 ϫ 8.5-cm, approximately 167 ml) of Q-Sepharose Fast Flow resin was packed and equilibrated at an 8 ml/min flow rate with buffer A (25 mM imidazole, pH 7.0, 20% glycerol, 2 mM DTT). The dialyzed ammonium sulfate precipitate was diluted 1:10 in buffer A and filtered through a 0.22micron surfactant-free cellulose acetate membrane filter. This material (156 ml, 1455 mg of protein, 1.21 ϫ 10 9 units of deacetylase activity; 1 unit ϭ 1 pmol ϫ min Ϫ1 ) was applied to the column at 3 ml/min. The column was washed with 500 ml of buffer A and then developed with a 1000-ml linear gradient (total volume) of buffer A containing 0 -350 mM KCl (6 ml/min.), followed by 200 ml of buffer A containing 1 M KCl. The fractions (10 ml) were assayed for deacetylase activity as described above. The total deacetylase activity recovered on the column was 1.16 ϫ 10 9 units (1424 mg) representing 96% of applied activity. Approximately three-fourths of this activity eluted in a peak from 90 to 130 mM KCl. These fractions were rapidly frozen in liquid nitrogen and held at Ϫ80°C. (Fig. 1). Previous studies have suggested that this enzyme is subject to regulation Kelly et al., 1993;Mohan et al., 1994). In order to assess the level of deacetylase activity in various bacteria, we assayed extracts prepared from a selection of laboratory strains (Table II). In general, deacetylase-specific activity was of the same order of magnitude among the enteric isolates tested (Table II), and it was also relatively constant within the pH range 5.5-6.5. The deacetylase in extracts of Pseudomonas was unique in that it was not active at pH 5.5, suggesting that the Pseudomonas deacetylase is somewhat different from that of enteric bacteria. Interestingly, E. coli strain MB4926 (envA1) was consistently deficient in deacetylase specific activity (Table II) compared with other strains of E. coli K-12. Assay of an isogenic pair of E. coli strains differing only at the envA locus revealed an 18-fold reduction in deacetylase specific activity in the envA1 mutant, MB5504, as compared with the envA ϩ parent, MB5503, (Table III, Experiment 1). Further, this deficiency was eliminated when the envA1 mutation was complemented in trans with pENV15, a plasmid carrying the wild-type allele in a low copy vector (Table III, Experiment 1). A similarly trans-complemented envA ϩ chromosomal allele did not produce a measurable increase in expressed cellular deacetylase activity over wild type.

Extracts of E. coli Strains Harboring the envA1 Allele Are Deficient in UDP-3-O-acyl-GlcNAc Deacetylase Activity-UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc deacetylase is the second enzyme of lipid A biosynthesis in E. coli
A priori, these observations could be explained by a regulatory function for the EnvA protein. Alternatively, the EnvA protein could actually be the deacetylase. The deacetylase might be regulated in some manner to maintain a constant level of enzyme activity. The level of expression from each of the envA gene(s) in these complemented strains is not known, and such experiments do not allow us to distinguish between these two possibilities. However, the results do indicate that Str/Spc, streptomycin/spectinomycin; Tet, tetracycline. The NdeI -BamHI restriction fragment resulting from digestion of pET11a-gyrB was removed prior to subsequent ligation, as described under "Experimental Procedures." the level of deacetylase activity in these cells is, at a minimum, a function of the allelic state of the envA gene.
Overexpression of the envA Gene Product-In order to address more fully these two possibilities, we placed the wild-type E. coli envA gene from pENV15 into the T7-based expression vector pET11a (Studier et al., 1990), as described under "Experimental Procedures." The insert of the resulting construct, pET11a-EnvA, was verified by sequencing and found to match exactly that reported previously by Beall and Lutkenhaus (1987).
To verify the function of the reading frame in this construct, pET11a-EnvA was transformed into LS822 (BL21(DE3) envA1), a lysogenic strain containing the gene for T7 RNA polymerase under the control of tandem lac UV5 promoters and sensitive to low levels of rifampicin as a consequence of the envA1 allele. Transcription of T7 RNA polymerase in this strain is normally somewhat "leaky," and the small amount of polymerase produced results in low basal transcription and translation of the gene(s) downstream of the T7 promotor. The pET11a-EnvA construct proved capable of complementing the envA1 permeability defect in this strain in the presence of IPTG as evidenced by the ability to grow on LB plates containing 1 g/ml rifampicin.
The soluble extracts of these induced host, vector, and pET11a-EnvA-bearing strains were assayed for UDP-3-O-acyl-GlcNAc deacetylase activity under standard conditions (Table  III, Experiment 2). The extract made from cells harboring the pET11a-EnvA clone was found to overexpress greatly the deacetylase activity, consistent with massive induction of the envA gene product. These results indicate that deacetylase activity parallels the apparent level of envA gene expression but formally does not rule out the alternative possibility that the expressed envA gene product directs expression of a similarly sized protein, which itself is the deacetylase.
Purification of the envA Gene Product-In order to identify positively the overexpressed protein found in induced BL21(DE3)/pLysS/pET11a-EnvA cells as the EnvA polypeptide and to prove its function as UDP-3-O-acyl-GlcNAc deacetylase, we isolated this protein from an induced, large scale preparation as described under "Experimental Procedures." From a 15-liter culture of BL21(DE3)/pLysS/pET11a-EnvA cells, grown and induced as described under "Experimental Procedures," we obtained 73.5 g of cell paste. This material was conveniently lysed with a single freeze-thaw step and the subsequent addition of 0.05% (w/v) Brij detergent. The presence of T7 lysozyme in these cells, which aids in suppressing the expression of cloned protein during growth prior to induction, acts to autolyze the peptidoglycan layer upon freeze-thawing. Analysis of the high speed supernatant fraction by SDS-PAGE and deacetylase assay revealed the presence of a high level of a 34-kDa protein and of a correspondingly massive amount of enzymatic activity (Fig. 4, lane 1, and Table IV). We found that enzyme held at this stage of purification was unstable with a decay half-life of approximately 40 h (stability studies data not shown). This instability was essentially eliminated by the next step.

TABLE II Specific activity of UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc deacetylase in membrane-free extracts of various bacteria
Strain Extract specific activity a pH 5.5 pH 6.0 pH 6.5  a Two-to three-fold variations in deacetylase specific activity are observed with different extracts and different strains of envA ϩ E. coli K12. Variation of duplicates of the same extract is approximately 5%.
b Performed at pH 6.5. c Performed at pH 5.5.
The protein band and deacetylase activity co-precipitated in high yield in a 50% ammonium sulfate fraction (Fig. 4., lane 2, and Table IV). The latter was dialyzed, filtered, and processed by anion ion exchange chromatography. The 34-kDa protein band and deacetylase activity were both recovered at approximately 110 mM potassium chloride (Fig. 4, lane 3, and Fig. 5). The isolated protein was judged to be Ͼ98% pure by SDS-PAGE (Fig. 4, lane 3).
A sample of similarly purified material was analyzed by Edman degradation in order to demonstrate that this isolated protein was indeed the expression product of the envA gene. The sequence obtained, Met-Ile-Lys-Gln-Arg-Thr-Leu-Lys-Arg-Ile-Val-Gln-Ala-Thr-Gly-Val-Gly-Leu-His, was exactly that predicted for the first 19 amino acids of the envA reading frame (Beall and Lutkenhaus, 1987). Further, the molar yield of phenylthiohydantoin-derivative products was consistent with the absence of measurable contaminating proteins, thus confirming the EnvA protein itself was the source of the comigrating deacetylase activity. We propose the designation lpxC for this gene.

DISCUSSION
Early studies of strains bearing the envA1 mutation indicated a slight reduction of lipopolysaccharide content compared with wild-type, possibly accounting for the antibiotic hypersensitivity associated with envA1 (Wolf-Watz and Normark, 1976;Grundstrom et al., 1980). We (Young and Silver, 1991) postulated that, in envA1 bearing mutants, increased entry of hydrophobic antibiotics might be due to the lowered LPS content but that increased entry of hydrophilic antibiotics and exit of periplasmic proteins could be due to breakage and rejoining of the outer membrane during inefficient separation at cell division. A number of pleiotropic mutants of E. coli and Salmonella exhibit similar "leaky" phenotypes with accompanying septal morphological defects (Weigand and Rothfield, 1976).
Wolf-Watz and Normark (1976) found that N-acetylmu-ramyl-L-alanine amidase activity, possibly involved in peptidoglycan remodeling at the septum, was 6-fold decreased in an envA1 strain. Since a 20-fold decrease of this amidase caused by a mutation at a different site in an envA ϩ background does not lead to increased outer membrane permeability (Tomioka et al., 1983), envA is unlikely to be the structural gene for this amidase.
Recently, it was noted that point mutations in the lpxA and lpxB genes, encoding the first acyl transferase (Galloway and Raetz, 1990) and the disaccharide synthase (Crowell et al., 1987) of lipid A biosynthesis, respectively, display increased antibiotic permeability (Vuorio and Vaara, 1992). 2 In addition, conditional mutants defective in lpxA (Galloway and Raetz, 1990) leak periplasmic enzymes 3 under permissive conditions. The fact that these mutations in lipid A biosynthesis exhibit a phenotype similar to that associated with envA1 and the finding that UDP-3-O-acyl-GlcNAc deacetylase activity is decreased in envA1-bearing strains suggested that the envA gene product might also play a role in lipid A biosynthesis.  Fig. 4). Lane W, Q-Sepharose wash. Lanes 2-20, column fractions as indicated. Lane S, protein molecular weight standards; phosphorylase b (97,400); bovine serum albumin (66,200); ovalbumin (45,000); carbonic anhydrase (31,000); soybean trypsin inhibitor (21,500); and lysozyme (14,400). When assayed in the range of pH 5.5-6.5, we found that crude extracts made from a number of common enterobacterial strains possess deacetylase specific activities that are within an order of magnitude of each other (Table II). However, extracts of envA1-bearing strains of E. coli K12 were 3.5-18-fold less active than closely related constructs (Tables II and III). The depression of deacetylase activity in strains harboring the envA1 mutation was especially apparent in an isogenic pair of E. coli strains (Table III). The deficit in deacetylase activity could be corrected specifically, although not increased above normal levels, by complementation in trans with the low copy vector pENV15 (envA ϩ ).
Forced expression of the envA ϩ reading frame using the T7 system of Studier (1987) resulted in massive overexpression of deacetylase activity in broken cell preparations (Table III). The demonstration that essentially all the deacetylase activity in these extracts purified with the expressed EnvA protein verified that the envA locus is indeed the structural gene encoding UDP-3-O-acyl-GlcNAc deacetylase. We propose the designation lpxC to replace envA, given its function in lipopolysaccharide biosynthesis.
In the lipid A pathway, the deacetylase functions between two acyltransferases, encoded by the lpxA and lpxD (firA) genes, respectively (Fig. 1). lpxA and lpxD map to a macromolecular synthesis operon at minute 4, containing genes involved in DNA, phospholipid, lipid A, and outer membrane protein biosynthesis ( Fig. 6) Raetz, 1993). In contrast, the lpxC (envA) gene near minute 2 resides at the 3Ј end of a large cluster of murein and cell division genes, the regulation of which is complex and not fully understood (Donachie, 1993, Errington, 1993. LpxC (envA) appears to have its own promoter (Sullivan and Donachie, 1984), but transcripts arising from upstream promoters could also contribute to deacetylase expression under some conditions. Whether the up-regulation of the deacetylase  under conditions of limited lipid A biosynthesis occurs at the level of transcription or by some other mechanism remains to be established. The association of lpxC with genes involved in peptidoglycan biosynthesis and cell division may reflect the operation of a novel, global regulatory network for envelope assembly.
Genetic interruption of the early steps in LPS biosynthesis is bactericidal (Galloway and Raetz, 1993;Kelly et al., 1993). Mutants in these genes must be studied as conditional lethals (Raetz, 1990;Raetz, 1993). The previous demonstration by Beall and Lutkenhaus (Beall and Lutkenhaus, 1987) of lethality associated with insertional inactivation of the envA reading frame is consistent with these observations. We have confirmed that extracts of the envA1-bearing point mutant do indeed possess residual enzymatic function (Tables II and III), as postulated by Beall and Lutkenhaus (1987). The question of why the depletion of lipid A deacetylase activity has such pleiotropic effects on cell morphology and outer membrane permeability requires further examination.
The apparent instability of the lpxC/envA ϩ gene, when introduced into E. coli on medium or high copy number plasmids (Sullivan and Donachie, 1984), also requires further investigation. Overexpression of the deacetylase might consume too much R-3-hydroxymyristoyl acyl carrier protein (Fig. 1) resulting in depletion of membrane glycerophospholipids.
The lpxC (envA) gene bears no significant relationship to other reading frames in GenBank TM . It contains no structural motifs that would suggest to what family of amidase (i.e. metallo, serine, or cysteine) it belongs. Mechanistic studies will be facilitated by the availability of large amounts of deacetylase. Cloning of the lpxC gene from other bacterial species, like Pseudomonas, would also serve to identify critical amino acid residues and might shed light on the mechanism of the deacetylase.
FIG. 6. Organization of the genes in the minute 2 and minute 4 regions of the E. coli chromosome. These regions have been completely sequenced, and the functions of most reading frames have been assigned as indicated (Errington, 1993;Tomasiewicz, 1987;Kelly et al., 1993). All genes shown at both minute 2 and minute 4 are transcribed from left to right (clockwise).