A Mutant of Escherichia coli Defective in the First Step of Endotoxin Biosynthesis*

Using localized mutagenesis of whole cells, we have isolated a temperature-sensitive UDP-N-acetylgluco- samine acyltransferase mutant of Escherichia coli that loses all detectable acyltransferase activity and quickly dies after a shift from 30 to 42 OC. Acyltransferase activity and temperature resistance are restored by transforming the mutant with a hybrid plasmid containing the E. coli gene for UDP-GlcNAc acyltransferase (ZpxA). In addition, a new assay has been devel- oped for quantitating the amount of lipid A (the active component of endotoxin) in E. coli and related Gram- negative strains. Cells are labeled with 32Pi and extracted with to remove glycerophospholipids. The residue is then hydrolyzed with HCl to liberate the phoryl” lipid A degradation

is reduced lo-fold when compared to wild-type after 60 min at 42 "C. These results provide physiological evidence that UDP-N-acetylglucosamine acyltransferase is the major committed step for lipid A biosynthesis in E. coli and that lipid A is an essential molecule.
Lipid A is the endotoxically active component of the lipopolysaccharides of Escherichia coli and related Gram-negative bacteria (1,2). It is also the hydrophobic anchor of lipopolysaccharide that constitutes the outer monolayer of the outer membrane.
Structural determination (2)(3)(4)(5) has shown that the predominant species of E. coli and Salmonella typhimurium lipid A is a p(l-6)-linked glucosamine disaccharide with (R)-3-hydroxymyristoyl substitutions at positions 2, 3, 2', 3', and phosphate at positions 1 and 4'. The biosynthesis of E. coli lipid A (Fig. 1) has been elucidated (l), and several early enzymes in the pathway have been isolated and partially characterized (1,(6)(7)(8)(9). UDP-N-acetylglucosamine acyltransferase and lipid A disaccharide synthase (Fig. l), are coded for by the 1~x4 and 1pxB genes, respectively (10 (11,12) and appear to be part of an operon that maps in the minute 4 region of the E. coli chromosome. E. coli strains that contain point mutations in 1pxA (11) or lpxB (10,13) have been isolated and shown to have decreased acyltransferase (11) and disaccharide synthase (6) activity, respectively. However, none of the available mutations results in complete inactivation of enzymatic activity or of lipid A biosynthesis (6,11,14). We now describe a new mutant defective in IpxA, in which UDP-iV-acetylglucosamine acyltransferase activity is absent in cell extracts. Growth and lipid A biosynthesis are temperature-sensitive.
This mutant provides a model with which to study the effect of acyltransferase deficiency on lipid A levels, outer membrane assembly, and cell physiology.
Conditions for the extraction of lipopolysaccharide (LPS)' from rough and smooth enteric Gram-negative bacteria with phenol-containing solvents have been established (15,16), and have facilitated the structural determination of lipid A and LPS (2). However, no simple, quantitative assay exists for determination of the lipid A content of E. coli. In order to characterize the lipid A content and the rate of lipid A synthesis in our mutants, we have developed a new radiochemical assay in which cells are labeled with 32Pi, followed by chloroform/methanol/water extraction of the glycerophospholipids.
Lipid A fragments are then released from the cell residue by mild acid hydrolysis, and are recovered by reextraction with appropriate chloroform/methanol/water proportions. Our method is rapid, applicable to many samples, and avoids the use of phenol.  was defective in UDP-N-acetylglucosamine acyltransferase activity. SMlOl was compared with mutant RX800 (&Al) (ll), isolated previously in our laboratory with the same procedure. This is illustrated in Fig. 2 with [p-32P]UDP-GlcNAc as the substrate. Unlike SMlOl, RX800 contains residual enzymatic activity. RX800 also has an unacceptably high reversion rate, and was not characterized further. The genetic lesion in SMlOl was designated lpxA2.
As described under "Experimental Procedures," the acyltransferase assay was simplified by using UDP-[6-3H]GlcNA~ as the radiolabeled substrate and by separating the acylated product from the starting material by reverse phase chromatography on a Cla Sep-Pak. This procedure avoids the use of thin layer chromatography and autoradiography.
With the new assay, the specific activity of acyltransferase in wild-type extracts was 1.8 nmol/min/mg, in reasonable agreement with previous determinations (7). The acyltransferase specific activity of extracts of RX800 (&AI) was 0.11 nmol/min/mg UDP-N-acetylglucosamine acyltransferase activity was measured in cell extracts of SMlOl, RX800, and the isogenic wild-type by the method of Anderson and Raetz (7). Unreacted ["'PIUDP-GlcNAc was separated from the product by thin layer chromatography on Silica Gel 60 plates in chloroform/methanol/water/acetic acid (25:15:4:2, v/v). Lanes 1 and 2, reaction containing extract RX800 (IpxAZ ) spotted at 2 and 10 min, respectively. Lanes 3 and 4, reaction containing extract of SM105 (IpxA+) at 2 and 10 min. Lanes 5 and 6, reaction containing extract of SMlOl (IpxAP) at 2 and 10 min. Autoradiography was for 12 h at room temperature. sitivity of SMIOl by Various Hybrid Plasmids-Hybrid plasmids encoding fragments of the minute four region of the E. coli chromosome were tested for their ability to correct the temperature sensitivity and the enzymatic defect of SMlOl. The previously described plasmids pDC2, pCR9, pDC29, pDC24, and pLPXA (10, 11) were used to transform SMlOl. Except for pDC24, all of the plasmids relieved the temperature-sensitive phenotype of SMlOl. The four plasmids that restored the growth of SMlOl at 42 "C all harbor the 1pxA gene. pDC24 contains the gene IpxB, which codes for the disaccharide synthase, but only contains a small portion of the adjacent 1pxA gene. pLPXA bears the 1pxA gene as its only insert (ll), and this plasmid restores the wild-type level of UDP-N-acetylglucosamine acyltransferase to SMlOl (data not shown). Taken together, these results demonstrate that the lesion in SMlOl that confers temperature sensitivity is in the same gene (1~~4) encoding the acyltransferase, and that Cells were grown in G-56 medium and shifted from 30 to 42 "C at time 0. Cell growth was followed by measuring the absorbance of SMlOl ( IpxAZ) and the isogenic strain ( IpxA') at 550 nm. Viable cells, expressed as colonyforming units (CFU)/ml, were determined by diluting cultures of SMlOl (lpxA2) and the isogenic strain (IpxA') in LB and plating each dilution on LB plates at 30 "C. with the new assay, but no activity was detected in SMlOl.
To determine whether the &A2 mutation caused cell death under nonpermissive conditions, or merely arrested cell growth, cells of SMlOl growing exponentially at 30 "C on LB medium were shifted to 42 "C. The optical density of the culture measured at 550 nm plateaued 2 h after the shift, and then slowly declined (Fig. 3). However, cell viability decreased several orders of magnitude starting 1 h after the temperature shift (Fig. 3) as judged by plating efficiency on LB agar at 30 "C. Loss of viability was also observed when SMlOl was shifted to 42 "C on minimal G-56 medium (data not shown).
The lpxA2 mutation did not cause any gross abnormalities in the morphology of the outer membrane. After 1 h at 42 "C, cells of mutant SMlOl were indistinguishable from wild-type, as determined by phase contrast and scanning electron microscopy (data not shown).
Correction of the Enzymatic Defect and Temperature Sen-the 1pxA gene is essential for growth. Extraction and Quantitation of Lipid A-The temperature sensitivity of strains harboring mutations in the gene coding for UDP-GlcNAc acyltransferase raised several important questions. Is UDP-GlcNAc acyltransferase the sole committed enzyme responsible for the generation of lipid A in E. coli? What is the rate of lipid A synthesis and the content of lipid A in mutant and isogenic wild-type cells at 30 and 42 "C? Can cells lacking lipid A grow under some conditions?
To answer these questions it was necessary to devise a simple, new assay for determining the content of lipid A in small samples of radiolabeled cultures. The existing methods for extracting LPS with phenol-containing solvents are not necessarily quantitative (l&16). Because of the structure and microheterogeneity of intact LPS (2) there is no unique radiochemical precursor with which to label and quantitate the number of lipid A molecules in cells.
We exploited the acid lability of LPS (1,2) in order to quantitate the amount of lipid A present. As shown by Takayama and coworkers (5), mild acid hydrolysis of isolated LPS cleaves the KDO-lipid A linkage and removes the anomeric phosphate. Partial loss of esterified fatty acids may also occur, accounting for the heterogeneity (5) of the observed products (Fig. 4)
Unlike intact LPS, they can be extracted quantitatively with chloroform/methanol/water mixtures using standard Bligh-Dyer proportions (18). The phosphorus content of the combined 4'-monophosphoryl lipid A units is an accurate measure to the molar amount of lipid A present in a LPS sample.
In a typical experiment, exponentially growing E. coli were either labeled uniformly or pulsed with "*Pi, as described under "Experimental Procedures." A sample of the culture (0.8 ml) was quenched with 2 ml of methanol and 1 ml of chloroform, yielding a single phase Bligh-Dyer mixture. The glycerophospholipids were extracted into the solvents, but the LPS remained associated with the cell debris and was collected by centrifugation.
Next, the cell debris was resuspended in aqueous 0.2 M HCl and boiled for 90 min. The 4'-monophosphoryl lipid A was then recovered by a second Bligh-Dyer extraction.
Time-course studies (not shown) revealed that most of the 4'-monophosphoryl lipid A was released from the cell debris in 30 min and that the recovery did not decline with prolonged hydrolysis. A portion of both the glycerophospholipids and the 4'-monophosphoryl lipid A species were then analyzed by thin layer chromatography and autoradiography, as shown in (lane 1). The ratio of lipid A to glycerophospholipid (determined by quantitation of the "*P in each fraction) in uniformly labeled, wild-type cells was approximately 0.12, consistent with one monolayer of lipid A in each cell (1). The same ratio was observed when wild-type cells were pulse-labeled with ""Pi for only 5 min.
The above assay is applicable to all smooth strains and to rough strains of the Rb and Ra chemotypes (data not shown). Because deep rough LPS is soluble in chloroform/methanol mixtures it is removed from the cells together with glycerophospholipids.
Consequently the above assay cannot be used with strains of the Re, Rd, and Rc chemotypes (2).
Characterization of the 32P-Labeled Lipid A Fragments Released from Cells by Acid Hydrolysis-The lipids released from solvent-extracted E. coli by acid hydrolysis appeared similar to the monophosphoryl lipid A degradation products obtained from purified S. typhimurium LPS (5) as judged by their mobility on thin layer silica plates. We also characterized their relationship to the standard compound, lipid IV* (Fig.  l), a tetraacyl disaccharide-1,4'-bisphosphate precursor of mature lipid A that can be purified from KDO-deficient mutants of S. typhimurium (21). When 4'-"'P-lipid IV* (22) was acid hydrolyzed and compared to the lipids released by acid hydrolysis of solvent-extracted E. coli, the 4'-32P-lipid IV*-derived material migrated more slowly than the major cell-derived bands (Fig. 6, lanes 2 and 4). This observation is consistent with the proposed structures ( Figs. 1 and 4), since acid hydrolysis should remove mainly the anomeric phosphate of lipid IV*, leaving a tetraacyl-disaccharide 4'-monophosphate. The corresponding derivative of mature lipid A would contain some normal fatty acids in acyl-oxyacyl linkage, in addition to the (R)-3-hydroxymyristoyl residues attached to the disaccharide backbone (Fig. 4). When the acid-hydrolyzed compounds were hydrolyzed further with mild alkali, however, both the lipid IV*-derived, and the cell-derived substances collapsed to identical, more slowly migrating species (Fig. 6 disaccharide backbone in both samples (Fig. 4).
with the monophosphoryl lipid A species from the whole cell The chemical basis for the apparent doublet after sequential suspensions. acid and base hydrolysis observed in both samples (Fig. 6) is The results of the FAB mass spectrometry confirm the uncertain.
conclusions of the radio-chromatographic analysis (Fig. 6), FAB Mass Spectrometry and 'H NMR Spectroscopy of the that sequential acid-base hydrolysis of solvent-extracted E. Products of Sequential Acid-Base Hydrolysis-The products coli and of pure lipid IV* generate the same predominant 4'of sequential acid-base hydrolysis of solvent-extracted E. coli, monophosphoryl lipid A species (Fig. 7). The identity of these and pure lipid IV* were compared by fast atom bombardment materials was also confirmed by 'H NMR spectroscopy (Fig. mass spectrometry in the positive and negative modes. The 8). Both preparations were converted to their pyridinium salts results indicate that the material derived from E. coti (Fig. 6, and dissolved in 0.6 ml of CDCl,/CDsOD (21, v/v). The major lane 3) is very similar to the comparable hydrolysis product features of their spectra (Fig. 8) are very similar, consistent of lipid IV* (Fig. 6, lane 5). Only one major species appeared with the scheme shown in Fig. 4. in both samples in the negative mode at m/z 871 (not shown), In a separate experiment, the lipids released by acid hyand this is attributed to [M -HI-, consistent with the drolysis of solvent-extracted E. coli (Fig. 6, lane 2) were structure shown at the bottom of Fig. 4. In the positive mode, analyzed directly by FAB mass spectrometry, without purifithe major species of the highest molecular weight were com-cation or additional base hydrolysis. In the negative mode mon to both preparations (Fig. 7). The peak at m/z 895.5 in (data not shown), prominent mass ions were observed at m/z the positive mode can be accounted for by [M + Na]+. The 1717 and m/z 1507. These correspond to a hexaacylated 4'other major peak in the positive mode at m/z 468.3 is probably monophosphoryl lipid A bearing laurate and myristate as the the oxonium ion derived from the nonreducing end, which normal fatty acyl moieties, and to a pentaacylated 4'-monoarises by cleavage of the glycosidic linkage. The peak at m/z phosphoryl lipid A bearing only laurate as its sole normal 550.6 has not been identified conclusively, but it is present in (non-hydroxy) fatty acyl chain. The simplicity of the mass both products. The smaller fragments (below m/z 468) present spectrum of this crude monophosphoryl lipid A preparation in the E. coli-derived sample have not been identified but are further validates our method for the determination of the lipid presumed to derive from minor impurities that copurified A content of E. coli.  Lipid A Content of Acyltransferase-deficient Mutants-The above data indicates that the lipids released by acid hydrolysis of solvent-extracted E. coli are a mixture of 4'-monophosphoryl lipid A fragments. Depending on the time of labeling with "P,, the 4'-monophosphoryl lipid A species prepared in this manner can provide a measure of the total lipid A content or the rate of lipid A synthesis. The amount of 32P in the monophosphoryl lipid A fraction can be normalized to the AssO or to the "P in the total glycerophospholipid fraction. Cells were usually labeled in G-56 medium, which has a lower phosphate content than LB, allowing greater incorporation of 32P into all lipids. As shown in Fig. 9, the banding pattern of the [32P]monophosphoryl lipid A species, recovered after pulse-labeling the cells, is the same in both the wild-type and mutant, indicating that the lipid A produced by the mutant is not structurally different from that of the wild-type. The same results are obtained with continuous labeling (data not shown). The rate of synthesis of all the labeled species derived from SMlOl is reduced to a similar extent after 1 h at 42 "C ( Fig. 9), consistent with a lesion in the first step of the biosynthetic pathway.
The lipid A content of the temperature-sensitive, acyltransferase-deficient mutant (SMlOl) and the isogenic wild-type 4 (SMl05) were compared at 30 "C and after 1 h at 42 "C. The ratio of lipid A to phospholipid in wild-type strains (such as SM105) is consistently around 0.12 at both temperatures. In the mutant SMlOl the lipid A to phospholipid ratio was 0.08 at 30 "C (the permissive temperature), and it dropped to 0.05 after 1 h at 42 "C (the nonpermissive temperature). Fig. 10 shows the results of 5 min pulse labeling of SM105 and SMlOl with 32P, at 30 "C (time 0) and as a function of time following a shift to 42 "C. Lipid A and phospholipid syntheses rise in the wild-type at 42 "C when normalized to the Asso until the cells reach stationary phase (Fig. 10, A and  B). The ratio of lipid A to phospholipid in pulse-labeled SMI05 is around 0.14 at all times (Fig. lOC), approximately the same as the ratio seen with continuous labeling (0.12). In mutant SMlOl, lipid A synthesis fails to rise in the first 5 min after shifting from 30 to 42 "C, and then it gradually drops (Fig. 10A). Phospholipid synthesis, however, does rise in SMlOl after the shift, as in the wild-type (Fig. lOB), even though growth in SMlOl is inhibited and viability begins to decline after 60 min (Fig. 3). Fig. 1OC shows that the ratio of lipid A to phospholipid synthesis in SMlOl falls rapidly in the first 20 min after temperature shift, while the ratio for the wild-type strain is unaffected. From these results we conclude that there is a lo-fold inhibition of the rate of lipid A synthesis in the mutant at 42 "C relative to the rate of glycerophospholipid synthesis.

Effect of Restoration of Acyltransferase
Activity on Lipid A Biosynthesis-The presence of a plasmid encoding the 1pxA gene in SMlOl (SMlOl/pLPXA) restored temperature resistance and UDP-N-acetylglucosamine acyltransferase activity as noted above. SMlOl/pLPXA was also assayed for the rate of lipid A synthesis, as in Fig. 10 (data not shown). In SMlOl/ The transductants were analyzed for growth at 30 and 42 "C.
acyltransferase, it will be necessary to generate "tighter" alleles of IpxA, preferably ones that are under the control of a regulated promoter (29). The temperature sensitivity of SMlOl strongly suggests that lipid A is an essential molecule for the growth and replication of E. cob, most likely, because it is required for outer membrane assembly (1,2). Outer membrane proteins appear to have a high affinity for the lipid A domain of LPS (30). It will be of considerable interest to study the synthesis and export of outer membrane proteins when lipid A synthesis is inhibited. The rapid loss of viability of SMlOl at 42 "C ( Fig. 3) may be significant in this regard, since the accumulation of functional porin trimers within the inner membrane might be lethal. Inhibition of protein synthesis has no immediate effect on the synthesis and export of LPS (26).
In the coming years, it will be important to examine a large collection of temperature-resistant revertants of SMlOl. The few spontaneous revertants that we have studied so far (Table  II) appear to regain lipid A, most likely, because they have regained sufficient UDP-GlcNAc acyltransferase function to support a normal rate of lipid A synthesis in uiuo. It is conceivable, however, that cells could grow without lipid A, or possibly even without an outer membrane, provided that appropriate bypass mutation(s) are present. It may also be possible to bypass the requirement for UDP-GlcNAc acyltransferase by inducing an alternative pathway for lipid A synthesis, by activating a gene that codes for an isoenzyme, or by altering the regulation of the lpx/dnaE operon. The simple radiochemical assay for lipid A synthesis and content that we have devised will facilitate the analysis of additional mutants and bypass revertants.
The observation that mutants defective in UDP-GlcNAc acyltransferase lose viability under nonpermissive conditions (Fig. 3) suggests that inhibitors of the early steps of the lipid A pathway (Fig. 1) might have utility as antibiotics. Inhibitors of CMP-KDO biosynthesis have been explored in this context (31,32), but so far the available compounds lack potency and are not bacteriocidal. Nevertheless, the gram-negative spectrum of the CMP-KDO synthesis inhibitors (31,32) validates the generality of LPS biosynthesis as a pharmacological target.