The Biosynthesis of Gram-negative Endotoxin A NOVEL KINASE IN ESCHERICHIA COLI MEMBRANES THAT INCORPORATES THE 4’-PHOSPHATE OF LIPID A*

Extracts of Escherichia coli an enzyme that generates the B,l+S linkage of lipid A from fatty-acylated monosaccharide precursors, We now de- scribe a membrane-bound kinase that phosphorylates the 4‘-position of the above tetraacyldisaccharide 1- phosphate product. The lipid A 4”kinase is distinct from the diglyceride kinase of E. coli. When crude membrane preparations are employed, several nucleo- side triphosphates are able to support the phosphorylation of the tetraacyldisaccharide 1-phosphate, but ATP is the most efficient. The 4”kinase requires M 8 + and is stimulated by phospholipids, especially cardiolipin. Under optimal conditions the specific activity in crude extracts is 0.5 nmol/min/mg. The enzyme is rap- idly inactivated by preincubation in the presence of detergents, such as Nonidet or octylglucoside, but

The lipid A domain of lipopolysaccharide is a hydrophobic molecule that constitutes the outer monolayer of the outer membrane of Gram-negative bacteria and causes shock when introduced into the circulation of most mammals (1)(2)(3). The predominant molecular species of lipid A found in the Escherichia coli envelope (4) is shown in Fig. 1. Prior to 1983 (5,6), the biosynthesis of lipid A was unknown, since the true covalent structure of lipid A ( Fig. 1) (4,(7)(8)(9)(10) was not correctly established in earlier investigations (3,7).
An important clue to the biosynthetic pathway for lipid A was provided by the finding that certain E. coli mutants deficient in phosphatidylglycerol accumulate novel phospholipids that are derived from glucosamine (11,12). Structural *This research was supported by National Institute of Health Grants AM 19551 and AM 21722 (to C. R. H. R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Recipient of a Procter & Gamble predoctoral fellowship from the Department of Biochemistry, University of Wisconsin-Madison.
To whom correspondence should be addressed.
studies (13,14) revealed that the simplest of these is 2,3diacylglucosamine 1-phosphate, also referred to as lipid X ( Fig. 1). Inspection of the structure of 2,3-diacylglucosamine 1-phosphate suggested that it might be the precursor of the reducing end unit of lipid A (13,15,16). The further speculation that there might exist a nucleotide derivative of 2,3diacylglucosamine 1-phosphate to serve as the precursor of the nonreducing end unit of lipid A (15) led us to identify an enzyme in extracts of wild-type cells capable of generating the p J -6 linkage of lipid A by catalyzing the condensation of 2,3-diacylglucosamine 1-phosphate and UDP-2,3-diacylglucosamine ( X ) , as shown in Fig. 1. This enzyme, termed the lipid A disaccharide synthase, is coded for by the lpzB (pgsB) gene that maps near minute 4 on the E. coli chromosome (12,16). The verification of the existence of UDP-2,3-diacylglucosamine and 2,3-diacylglucosamine 1-phosphate as metabolites in living cells of wild-type E. coli (15) raised the question of how these compounds are generated from known biosynthetic intermediates. Recently, we have discovered enzymes in E. coli, distinct from the glycerol-3-phosphate acyltransferase (17,18), that are capable of acylating UDP-GlcNAc, as shown in the upper half of Fig. 1. The evidence from the enzymological and radiochemical labeling studies clearly demonstrates that UDP-2,3-diacylglucosamine is a precursor of 2,3-diacylglucosamine 1-phosphate ( Fig. 1) (17).
We now present further support for this biosynthetic scheme by demonstrating the existence of a 4"kinas.e. This enzyme catalyzes the ATP-dependent phosphorylation of the tetraacyldisaccharide 1-phosphate intermediate ( Fig. 1) generated by the lipid A disaccharide synthase. Under certain conditions the 4' phosphorylation reaction proceeds to about 60% of completion, generating a tetraacyldisaccharide-1,4'bisphosphate ( Fig. 1). Fast atom bombardment mass spectrometry and NMR spectroscopy have been employed for the structural analysis. The product generated by the 4'-kinase in vitro is identical to the predominant lipid A precursor that accumulates in vivo at 42 "C in temperature-sensitive mutants of Salmonella typhimurium, deficient in .

EXPERIMENTAL PROCEDURES
Materk~ls-~~P~ and [y32P]ATP were products of New England Nuclear. Glass-backed plates (5 X 20 cm), coated with a 250-pm layer of Silica Gel 60, were obtained from E. Merck, Darmstadt, West Germany. Glass-backed plates (5 X  Postulated scheme for the biosynthesis of lipid A in E. coli. In the figure, R designates a @-hydroxymyristoyl moiety and U designates uridine. Evidence for the fatty acylation of UDP-GlcNAc and for the precursor-product relationship between UDP-2,3-diacyl-GlcN and lipid X has been presented by Anderson et al. (17). The tetraacyldiaaccharide-1-phosphate synthase is the product of the &xB (pgsB) gene (16), as indicated. The tetraacyldisaccharide 1,4'-bisphosphate is generated by the 4"kinase reported in this manuscript, and it is also the predominant acidic lipid A precursor (termed polar lipid W A ) that accumulates in temperature-sensitive mutants of S. typhimuriurn defective in KDO biosynthesis (22,23). The transfer of two KDO moieties from two molecules of CMP-KDO to this precursor by membranes of wild-type cells has been claimed (24), but the product(s) have not been characterized by structural methoda. In a portion of the lipid A molecules, additional polar moieties are present, and these may be added at the same time as, or layer of Anasil H, were obtained from Analabs Inc., North Haven, CT. Silicic acid (Bio-Si1 A, 200-400 mesh) was a product of Bio-Rad. Yeast extract, tryptone, and agar were products of Difco. Deuterated compounds for nuclear magnetic resonance spectroscopy studies were obtained from Aldrich. Tetraacyldisaccharide 1-phosphate was preparedas previously described (16). Phosphatidylserine (bovine brain), phosphatidylcholine (egg yolk), and cardiolipin (bovine heart) were obtained from Sigma. Phosphatidylethanolamine and phosphatidylglycerol were isolated from E. coli (18). All other compounds were obtained from Sigma.
Preparation of Cell Extracts for Enzymatic Assays-E. coli strains RZ60 (dgk-6) and R477 (dgk+) were described previously (30). Cells (1 liter) were grown in LB broth (31) at 37 "C, harvested at the end of log phase (Am s 4), resuspended in 15 ml of 50 mM Tris-HC1, pH 8, and disrupted using a French pressure cell at 18,000 p.s.i. Unbroken cells were removed by centrifugation at 8,000 X g. , for 10 min. Membranes were prepared by ultracentrifugation of the French press extracts at 150,000 x g . , for 1 h. Membranes were resuspended in 3 mi of 50 mM Tris-HC1, pH 8, yielding a concentration of 15-20 mg/ ml of protein. All steps in the preparation were performed at 4 "C. Samples were stored at -70 "C. Protein concentration was determined by the method of Lowry et ul. (32) using a bovine serum albumin standard.
Conditions for Fast Atom Bombardment Mass Spectrometry and NMR Spectroscopy-Fast atom bombardment mass spectrometry was done at the National Science Foundation Regional Instrumentation Facility at The Johns Hopkins University, as described previously (13)(14)(15)(16). 'H NMR spectra were acquired at 270 MHz on a Bruker WH-270 Fourier transform superconducting spectrometer interfaced with a Nicolet 1180 computer. 31P NMR spectra were acquired at 80 MHz on a Nicolet NT-200 Fourier transform superconducting spectrometer equipped with a Nicolet 1280 computer. The spectra were usually recorded at 30 "C at a sample concentration of 5-10 mg/ml in CDC1,:CD30D (101, v/v). 'H decoupling experiments were performed at 500 MHz on a Bruker AM500 Fourier transform superconducting spectrometer.
Isolation of the Product from the Lipid A 4'-Kinose Reaction-A 120-ml reaction mixture was prepared as described in the legend to Fig. 3, except that nonradioactive tetraacyldisaccharide 1-phosphate (0.25 mg/ml) was employed. The system was incubated for 2 h at 37 "C and then an additional 240 mg of P-enolpyruvate was added. Three hours later, the reaction was stopped by the addition of 300 ml of methanol. Next, 150 ml of chloroform was added. The resulting one-phase system was centrifuged at low speed to remove debris. The supernatant was converted to a two-phase system by the addition of 150 ml of chloroform and 150 ml of 0.3 M HCl. Centrifugation at low speed was required to separate the phases. The lower (chloroform) phase was concentrated to 15 ml on a rotary evaporator. This was diluted to 100 ml in chlorofomxmethanokwater (23:1, v/v). Then the sample was applied to a DEAE-cellulose column (125-ml bed volume) prepared in the same solvent, and the product was eluted as described by Raetz et al. (22) for the purification of the precursor IVA. The solvent and salts were removed by acidic Bligh-Dyer partitioning (22), followed by rotary evaporation. About 10 mg of kinase product was recovered. Thin layer chromatography, however, revealed that this material was contaminated to approximately 30% with a cardiolipin-like molecule.
Final purification was achieved by HPLC, using an Alltech CIS 10pm reversed phase column (250 X 6.5 mm), a Waters 6000A solvent delivery system, and a Waters 480 LC spectrophotometer. Elution of the kinase product was detected at 210 nm. Two solvent mixtures were employed (a) acetonitri1e:water (5050, v/v) containing 5 mM tetrabutylammonium phosphate; and (b) isopropanokwater (85:15, v/ v) containing 5 mM tetrabutylammonium phosphate. The partially shortly after, KDO (21, 23). Specifically, a portion of the lipid A molecules of E. coli bear a pyrophosphate residue at the reducing end (8, 25) (as indicated by the dashed bond), while in S. typhimurium additional polar substituents, including 4-amino-4-deoxy-~-arabinose and/or phosphoethanolamine, may be attached to lipid A (1,21,23). The "late acylations" by which the lauroyl and myristoyl residues (1,9,26) are incorporated have not been studied, but the presence of these residues is correlated with extreme toxicity in animals (27, 28). The final structure shown is the minimal unit thought to be required for cell growth and outer membrane biogenesis in E. coli (1,2). All other core sugars and 0-antigens of lipopolysaccharide are not essential for cell growth under laboratory conditions, and mutants lacking these are not temperature-sensitive (1,29). purified kinase product (-10 mg) was dissolved in 3.5 ml of solvent Asolvent B (21, v/v). Five individual injections (0.7 ml each) were required to purify all of the available material. After each injection, the column was eluted at a flow rate of 2.3 ml/min with a linear gradient, starting with solvent Asolvent B (21, v/v) and ending with solvent B. Each chromatography was completed in 60 min, and in each case the kinase product emerged at min 39, as did polar lipid IVA (determined in a separate experiment).
The peaks of the product recovered from the five individual injections were pooled, and the volume (-60 ml) was reduced to -4 ml by rotary evaporation. Next, 60 ml of CHC13:methanol:water (2:3:1, v/v) was added, and the sample was applied to a second column (40-ml bed volume) of DEAE-cellulose in the same solvent (22) to remove the tetrabutylammonium phosphate. After washing the column with 120 ml of CHC&:methanol:water (23:1, v/v), the kinase product was eluted with 100 ml of CHC13, methanol, 480 mM ammonium acetate (2:3:1, v/v). Fractions containing the kinase product were adjusted by adding appropriate amounts of CHCI3, water, and concentrated HCl to generate a two-phase, acidic Bligh-Dyer system (11,22). The lower phase was dried, and the residue was weighed. Approximately 3.5 mg of highly purified kinase product was recovered. Since the lipid is obtained in the free acid form under these conditions, it is redissolved immediately in 0.4 ml of CDC13:CDsOD (101, v/v) for NMR analysis. Prolonged storage of the kinase product (or lipid X), as the free acid, results in the gradual loss of the anomeric phosphate residue.

RESULTS
Identification and Properties of a Novel Kinase in Extracts of E. coli Specific for Lipid A Precursors-When tetraacyldisaccharide 1-phosphate is incubated with a crude extract or with membranes of E. coli in the presence of [-p32P]ATP and Mg+, as described in Fig. 2, a slowly migrating 32P-labeled lipid product is formed, having an RF upon thin layer chromatography that is identical to that of the tetraacyldisaccharide-l,4'-bisphosphate (Fig. 1, also designated precursor IVA), isolated from KDO-deficient mutants (22, 23). Little or P + S.F. FIG. 2. A novel kinase specific for lipid A precursors. The complete reaction mixture (50 pl) contained 0.25 mg/ml tetraacyldisaccharide I-phosphate, 2 mg/ml cardiolipin, 50 mM HEPES, pH 7.4, 1% Nonidet P-40, 5 mM MgC12, 4 mM [yR2P]ATP (lo5 cpm/nmol), and E. coli membranes (1.7 mg/ml) that were used to start the reaction. After 20 min at 37 "C, the lipids in the reaction mixture were extracted under acidic Bligh-Dyer conditions (ll), and a portion of each CHC13 phase was spotted on a Silica Gel 60 thin layer plate that was developed in CHC13, pyridine, 88% formic acid, water (40:6016:5, v/v). The plate was subjected to autoradiography, as shown. Lanes 2 and 4, complete system with membranes of R477 (dgk') and RZ60 (dgk-61, respectively; lanes I and 3, complete system minus tetraacyldisaccharide I-phosphate with membranes of R477 and RZ60, respectively; lane 5, complete system minus enzyme. Ori, no product is generated in the absence of the tetraacyldisaccharide 1-phosphate substrate (Fig. 2, lanes 1 and 3) or in the absence of enzyme (lane 5). Membranes from wild-type cells (lanes 1 and 2) also incorporate 32P into the glycerophospholipids (the more rapidly migrating components), independently of added tetraacyldisaccharide 1-phosphate. Presumably, the latter process is mediated by diglyceride kinase, since membranes prepared from E. coli mutants (30) lacking the diglyceride kinase (lanes 3 and 4) do not incorporate much 32P into the glycerophospholipid fraction. Further, the incorporation of 32P into the lipid A precursor is unaffected by the dgk-6 mutation (Fig. 2), demonstrating that the lipid A kinase is a distinct enzyme.
The formation of the putative tetraacyldisaccharide-1,4'bisphosphate was also demonstrable by incubation of tetraacyldisaccharide l-[32P]phosphate with unlabeled ATP (data not shown). CTP, UTP, and GTP could be substituted, but ATP supported approximately twice the rate of the others. As shown in Table I, cardiolipin, Nonidet P-40, and Mg2+ all stimulated the reaction rate. Efficient conversion of the lipid substrate to product at a linear rate was observed only if cardiolipin was included in the reaction mixture, together with an ATP regenerating system ( Fig. 3 and Table I). Other phospholipids, including phosphatidylethanolamine, phosphatidylserine, phosphatidylcholine, and phosphatidylglycerol, were not as effective as cardiolipin. The inclusion of a phospholipid "cofactor" and a nonionic detergent in the reaction mixture was essential, because of the intrinsic insolubility of the tetraacyldisaccharide 1-phosphate in water, especially in the presence of divalent cations. The conditions for the preparation of stable, aqueous emulsions of disaccharide/cardiolipin mixtures are described in the legend to Fig.  3.
Other Properties of the Kinase-In contrast to the other enzymes shown in the upper half of Fig. 1 (16, 17), most of the lipid A kinase sedimented with the membrane pellet when centrifuged at 150,000 X g , , for 1 h. Fig. 4 gives the results of a semiquantitative assay of the lipid A kinase in crude cellfree extracts or in membrane preparations. Fig. 44 shows the protein dependency of the activity. Although crude extracts and membrane preparations had approximately the same specific activities, Fig. 4B shows that the assay was linear for a much longer time (and consequently more reliable) with membrane preparations. Typically, the specific activity of a crude extract is 0.3-0.5 nmol/min/mg or about one-half of the specific activity of the lipid A disaccharide-1-phosphate Incubation conditions were essentially the same as described in Fig. 3, using 0.16 mg of membranes from strain Rz60 in 100 pl a t 37 "C for 20 min. The complete system generated 3.7 nmol of product a t a linear rate, representing an 18.5% conversion of the tetraacyldisaccharide l-[32P]phosphate substrate. Values reported are the averaees of tridicates.

Incubation conditions
Relative amount of product formation

Complete system Omit enzyme
Enzyme heated for 10 min at 60 "C   Fig. 2, except that '*P-labeled tetraacyldisaccharide 1-phosphate (10' cpm/nmol) was employed in 50 mM EPPS at pH 8.5, together with unlabeled ATP and an ATP regenerating system, consisting of 10 mM P-enolpyruvate and 100 units/ml of pyruvate kinase. The reaction was initiated by the addition of E. coli membranes (1 mg/ml, final concentration) from strain RZ60 (dgk-6). The rate of the reaction waa about 2-fold greater at pH 8.5 than at 7.4. To quantitate product formation, 5 pl of the reaction mixture was spotted on a thin layer silica gel plate (Silica Gel 60), and the plate was developed in chloroform, pyridine, 88% formic acid, water (406016:5, v/v). The "P-labeled product (which migrates more slowly than the 32P-labeled substrate) was located by autoradiography and was quantitated by scraping it off the plate and counting it in 10 ml of Patterson-Green mixture (33). W, complete assay as described above; U , same but no cardiolipin included. In the above reaction, as in Table I and Fig. 2, the tetraacyldisaccharide 1phosphate and cardiolipin are added together from a more concentrated stock dispersion, which is made as follows. Substrate (0.5 mg) and cardiolipin (4 mg) are mixed in about 1 ml of chloroform:methanol (2:1, v/v). Next, they are dried under a stream of N2, and 1 ml of buffer (usually 100 mM EPPS, pH 8.5) is added. The mixture is dispersed for 5 min in a bath sonicator to give a nearly transparent, stable emulsion. This emulsion (presumably consisting of liposomes) is two-fold more concentrated than is required in the final reaction mixture. It can be stored for several weeks at 0-4 "C without significant chemical deterioration.
synthase (16). The dependency of the rate of product formation by membranes on the concentration of tetraacyldisaccharide 1-phosphate is shown in Fig. 5. Although product formation was linear for 1 h in the presence of detergent and an ATP regenerating system containing P-enolpyruvate (Figs. 3 and 4), the enzyme was rapidly inactivated upon preincubation with detergent alone, even at 0 "C (Table 11). Inclusion of Mg2+ (Table 11) had no effect. When P-enolpyruvate was present during preincubation, the enzyme was stabilized significantly (Table 11). Glycerol (25%, w/v) also greatly stabilized the kinase (data not shown).
When the tetraacyldisaccharide 1-phosphate substrate was subjected to mild alkaline hydrolysis in the presence of triethylamine (14) to remove the ester-linked fatty acids, the resulting diacyldisaccharide 1-phosphate was still a substrate FIG. 5. Rate of product formation as a function of tetraacyldisaccharide 1-phosphate concentration. The amount of product formed by the kinase was determined after 20 min at 37 "C under conditions identical to those of Fig. 3, except that the volume of the reaction mixture was 0.5 ml and the substrate concentration was varied. Under the standard assay conditions (Fig. 4) the tetraacyldisaccharide 1-phosphate is employed at a concentration of approximately 200 p~.
for the lipid A kinase, but the rate of phosphorylation was about two times slower (data not shown). The enzyme did not phosphorylate 2,3-diacylglucosamine 1-phosphate (lipid X) under the conditions shown in Fig. 2.

Analysis of the Reaction Product by Fast Atom Bombardment
Mass Spectrometry-To determine the nature of the product generated from the tetraacyldisaccharide 1-phosphate by the lipid A kinase, the product was isolated from a largescale reaction mixture as described under "Experimental Procedures" and analyzed by FAB mass spectrometry and NMR spectroscopy. Purity of the isolated kinase product was assessed by thin layer chromatography in chloroform, pyridine, 88% formic acid, water (40:6016:5, v/v), followed by sulfuric acid charring (22), or by HPLC.
The positive mode FAl3 mass spectrum of the in Vitro product, shown in Fig. 6, revealed a strong peak at m/z 1406, attributed to (M + H)+, as expected for a tetraacyldisaccharide bisphosphate product (22). The molecular formula is C,J-I,30N,0,3P2 with a predicted molecular weight of 1405.72. Other major peaks are observed at m/z 1309, corresponding shown in panel A of Fig. 7, and the corresponding spectrum Effect ofpreincubation of E. coli membranes with detergent on lipid of the in vitro product of the lipid A 4'-kinase is shown in A I'-kinase activity panel B. The spectra are virtually identical. The peak assign-E. coli membranes (3 mg/ml) were preincubated at 0 "c with 0.5% ments (Fig. 7A) for lipid lvA in CDC1,:cD3OD (10.1, v/v) (w/v) Nonidet P-40 for the times indicated. Assay conditions are were determined by proton decoupling experiments (data not described in the legend to Fig. 3, except that the final concentrations of tetraacyldisaccharide 1-phosphate and cardiolipin were 0.13 and 1 shown) and were entirely consistent with the 'H NMR specmg/ml, respectively.
trum of the tetramethyl derivative of IVA, previously examined in detail (23) by two-dimensional methods. Of interest is the peak tentatively assigned to H-4' of the in vitro kinase product (Fig. 7B, arrow) (16) are preincubation at 0 "C. Values are given as the percent activity re-Further evidence for the assignment of the resonance at maining in comparison to no preincubation f S.E.
subjecting this material to selective 'H decoupling at 500 MHz Preincubation also included 5 mM MgCl,. to the loss of a phosphate; at m/z 1180, corresponding to the loss of a 8-hydroxymyristoyl moiety; and at m/z 1082, corresponding to the loss of a @-hydroxymyristoyl moiety and a phosphate. Importantly, the predominant ion at m/z 694 is attributed to the oxonium ion that arises from the nonreducing end of the tetraacyldisaccharide bisphosphate after cleavage of the glycosidic linkage (22). The mass of this peak The sample, purified by HPLC, was prepared as described in the legend to Fig. 7. The insets are the relevant peaks observed when the resonances, indicated by the arrows, are selectively irradiated. The assignments are explained in the text. The chemical shift, 6, is expressed as parts/million relative to tetramethylsilane.

TABLE I11
Phosphorus-31 chemical shifts and 'H-31P coupling data for the kinase product and polar lipid IVA The data was taken from the 31P spectrum shown in Fig. 9 and from a comparable spectrum of polar lipid IVA. Chemical shifts are in parts/million relative to 10% phosphoric acid in D20. The estimated error of the chemical shifts is f O . l ppm. Coupling constants were assigned on the basis of previous studies of I VA (23) and tetraacyldisaccharide 1-phosphate (16) and are expressed in Hz with an estimated error of +0.4 Hz. H-4' as a complex multiplet in the kinase product (Fig. 8), in contrast to its appearance as a triplet in the substrate (16), is accounted for by the introduction of Jd,,p of 10 Hz (Table 111) following phosphorylation of the 4' hydroxyl.
31P NMR Spectroscopy-The product generated by the 4'kinase was analyzed by 31P NMR spectroscopy. As expected, a broad band 'H decoupled 31P spectrum (Fig. 9A) revealed the presence of two nonequivalent phosphorus resonances of equal intensity separated by about 1.6 ppm. Under the conditions employed (free acid form of the lipid in CDCl,:CD,OD (lO:l, v/v)) the spectrum of polar lipid IVA is identical (data not shown). The downfield resonance at -0.06 ppm is attributed to the 4'-monophosphate function on the basis of its relative chemical shift, by comparison to standards (23), and on the basis of the prominent JH,P values of 10 Hz determined from the proton coupled spectrum ( Fig. 9B and Table 111).
In summary, the 'H and 31P NMR spectra, taken together with the mass spectrum (Fig. 6), demonstrate that the kinase we have discovered is specific for position 4' and that other hydroxyl groups are not phosphorylated to a significant extent by the crude membrane preparations employed. 9. Analysis of the kinase product by "P NMR spectroscopy. The sample (-4 mg) was prepared as described in the legend to Fig. 7. Panel A shows the 31P NMR spectrum observed with broad band bilevel 'H decoupling. Panel B shows the 'H coupled spectrum. Chemical shifts are expressed as parts/million relative to a 10% phosphoric acid external standard to D20. H,P coupling constants are summarized in Table 111.

DISCUSSION
The kinase described in this article appears to be a novel enzyme that incorporates the 4'-monophosphate moiety of lipid A (Fig. 1). The existence of this kinase in extracts of E. coli supports the proposed biosynthetic scheme for lipid A shown in Fig. 1 and is consistent with the view that the tetraacyldisaccharide 1-phosphate intermediate generated by the lipid A disaccharide synthase (IpxB gene product) is a true biosynthetic precursor (Fig. 1). The lipid A 4"kinase is distinct from diglyceride kinase (30), since mutants lacking diglyceride kinase activity (such as RZ60) have normal levels of the 4"kinase (Fig. 2).
Unlike the lipid A disaccharide synthase (16) or the enzymes responsible for the generation of UDP-2,3-diacylglucosamine from UDP-GlcNAc (17), the 4"kinase is predominantly membrane-bound. Its substrate, the tetraacyldisaccharide 1-phosphate, is extremely hydrophobic and insoluble in water (16). In the presence of Mg2+ ions, it forms a coarse precipitate in water that cannot be dispersed by sonic irradiation. Since the kinase requires magnesium for catalytic activity, however, it was necessary to develop conditions permitting the efficient dispersal of the tetraacyldisaccharide l-phosphate in mixed micellar form in the presence of magnesium. This was achieved by the co-dispersion of the tetraacyldisaccharide 1-phosphate with cardiolipin and by the inclusion of a nonionic detergent, such as Nonidet P-40 or octyl-P-Dglucoside, in the reaction mixture.
While the inclusion of 1% Nonidet P-40 was required to obtain a maximal initial rate of 4' phosphorylation, the 4'kinase was solubilized and rendered unstable by various detergents, particularly in the absence of the complete assay mixture and glycerol, as shown in Table 11. Since phosphoenolpyruvate stabilized the enzyme significantly against inactivation, the best conversions of the tetraacyldisaccharide 1phosphate substrate to the 1,4'-bisphosphate product were observed in the presence of phosphoenolpyruvate and an ATP regenerating system (Fig. 3). The reasons for the instability of the kinase in the presence of detergent and for the effects of P-enolpyruvate are not known, but the existence of an additional cofactor or labile phosphorylated intermediate cannot be excluded, given the crude enzyme system employed. In this regard, it is relevant that we have not yet established the stoichiometry of 4' phosphorylation and ADP formation because of the multiple fates of ATP in such cell-free extracts.
Other nucleoside triphosphates, although less active than ATP, also support the 4' phosphorylation. However, the crude enzyme preparation employed does have an absolute requirement for a nucleoside triphosphate, since no product formation is observed when an excess of hexokinase and glucose are added (data not shown).
The specificity of the 4"kinase for other lipid molecules will also require further investigation. In preliminary studies we have shown that the monosaccharide, lipid X, is not phosphorylated, suggesting that the enzyme is specific for disaccharides. When the ester-linked fatty acids are removed from the tetraacyldisaccharide 1-phosphate substrate by treatment with triethylamine (14), the resulting diacyldisaccharide I-phosphate is still a relatively good substrate for the 4"kinase. We believe that the 4"kinase acts on a tetraacyldisaccharide 1-phosphate intermediate in vivo, however, since mutants of S. typhirnuriurn deficient in KDO biosynthesis accumulate a tetraacyldisaccharide 1,4'-bisphosphate, having the structure shown in Fig. 1, that is identical to the 1,4'bisphosphate generated in vitro by the 4"kinase (Figs. 6 and 7) (22, 23). In addition, the disaccharide synthase (16) shows a 100-1000-fold kinetic selectivity for diacylated substrates relative to monoacylated or triacylated substrates.' The isolation of mutants deficient in the lipid A 4'-kinase would provide the strongest evidence that the tetraacyldisaccharide 1-phosphate is the true physiological substrate, since it would be expected to accumulate in such mutants. The effect of mutations in the 4"kinase on lipopolysaccharide assembly and cell viability would also be of considerable interest.
Whatever the biological role of the 4'-kinase, we anticipate several novel uses for this enzyme in studies of lipid A physiology. We have developed conditions (data not shown) for generating the tetraacyldisaccharide 1,4'-bisphosphate product in vitro using [y3'P]ATP of very high specific radioactivity. It may be possible to use these radiolabeled disaccharides as probes for the identification of lipid A receptors or other lipid A binding proteins in cell membranes. Probes of high specific radioactivity will also be useful for studying the metabolism of lipid A and lipid A precursors by bacterial and C. R. H. Ftaetz, unpublished observation.
animal cells. If it is possible to clone the gene for the 4'kinase and to purify the enzyme to homogeneity, the enzyme may be a useful reagent, in conjunction with chemical methods, for the synthesis of lipid A and lipid A-like molecules.