A special acyl carrier protein for transferring long hydroxylated fatty acids to lipid A in Rhizobium.

Lipid A, the hydrophobic anchor of lipopolysaccharides in the outer membranes of Gram-negative bacteria, varies in structure among different Rhizobiaceae. The Rhizobium meliloti lipid A backbone, like that of Escherichia coli, is a β1′-6-linked glucosamine disaccharide that is phosphorylated at positions 1 and 4′. Rhizobium leguminosarum lipid A lacks both phosphates, but contains aminogluconate in place of the proximal glucosamine 1-phosphate, and galacturonic acid instead of the 4′-phosphate. A peculiar feature of the lipid As of all Rhizobiaceae is acylation with 27-hydroxyoctacosanoic acid, a long hydroxylated fatty acid not found in E. coli. We now describe an in vitro system, consisting of a membrane enzyme and a cytosolic acyl donor from R. leguminosarum, that transfers 27-hydroxyoctacosanoic acid to (Kdo)2-lipid IVA, a key lipid A precursor common to both E. coli and R. leguminosarum. The 27-hydroxyoctacosanoic acid moiety was detected in the lipid product by mass spectrometry. The membrane enzyme required the presence of Kdo residues in the acceptor substrate for activity. The cytosolic acyl donor was purified from wild-type R. leguminosarum using the acylation of (Kdo)2-[4′-32P]-lipid IVA as the assay. Amino-terminal sequencing of the purified acyl donor revealed an exact 19-amino acid match with a partially sequenced gene (orf*) of R. leguminosarum. Orf* contains the consensus sequence, DSLD, for attachment of 4′-phosphopantetheine. When the entire orf* gene was sequenced, it was found to encode a protein of 92 amino acids. Orf* is a new kind of acyl carrier protein because it is only ∼25% identical both to the constitutive acyl carrier protein (AcpP) and to the inducible acyl carrier protein (NodF) of R. leguminosarum. Mass spectrometry of purified active Orf* confirmed the presence of 4′-phosphopantetheine and 27-hydroxyoctacosanoic acid in the major species. Smaller mass peaks indicative of Orf* acylation with hydroxylated 20, 22, 24, and 26 carbon fatty acids were also observed. Given the specialized function of Orf* in lipid A acylation, we suggest the new designation AcpXL.

Lipopolysaccharides, or endotoxins, comprise the outer leaflet of the outer membranes of Gram-negative bacteria (1)(2)(3)(4)(5). Lipid A, the hydrophobic moiety that attaches lipopolysaccha-ride to the membrane, is of special interest because it is essential for bacterial growth (6), and its biosynthesis is a target for the design of new antibacterial agents. 1 In addition, lipid A is a potent stimulant of mammalian immune cells (3,5,8). The overproduction of cytokines and inflammatory mediators by macrophages upon stimulation by lipid A during severe Gramnegative infections is thought to cause some of the clinical complications of Gram-negative sepsis (3, 5, 8 -10). Escherichia coli lipid A, one of the best studied examples, consists of a glucosamine disaccharide backbone that is linked ␤1Ј-6 ( Fig. 1), is acylated with R-3-hydroxymyristate at positions 2, 3, 2Ј, and 3Ј, and is phosphorylated at positions 1 and 4Ј (1,3,5,(11)(12)(13). The two R-3-hydroxy moieties of the distal unit are further acylated with laurate and myristate ( Fig. 1), forming acyloxyacyl groups (1,3,5,(11)(12)(13)(14)(15). The latter are structural hallmarks of lipid A moieties from diverse sources, and they are critical for the immunostimulatory activity of endotoxins (3,5,16,17). Lipid A analogs with a reduced number of acyloxyacyl moieties are of interest because some are potent endotoxin antagonists with possible utility for treating the complications of Gramnegative sepsis (3,(17)(18)(19)(20).
Given the importance of lipid A analogs as endotoxin antagonists, we have recently become interested in elucidating the enzymatic synthesis of lipid A in Rhizobium leguminosarum (21,22). Lipid A of R. leguminosarum lacks the phosphate groups (23) found in E. coli or Rhizobium meliloti lipid A (3,24,25). R. leguminosarum lipid A contains an acylated aminogluconate in place of the proximal glucosamine 1-phosphate, and a galacturonic acid residue in place of the 4Ј-phosphate ( Fig. 1) (23). Most remarkably, R. leguminosarum lipid A appears to lack acyloxyacyl residues, as judged by the absence of myristate, laurate, and other common short acyl chains (23). However, lipid As of R. leguminosarum and most other Rhizobiaceae contain an unusual, long hydroxylated acyl moiety, 27hydroxyoctacosanoic acid, not found in enterobacterial lipid A (26,27). The exact location of the ester-linked 27-hydroxyoctacosanoic acid moiety in R. leguminosarum lipid A is uncertain (23). It is apparently not attached to one of the R-3-hydroxy acyl chains (Fig. 1), assuming that acyloxyacyl moieties containing 27-hydroxyoctacosanoic acid have the same chemical reactivities as ordinary acyloxyacyl groups (23,28). The function of 27-hydroxyoctacosanoic acid is not known. Its length is about twice that of the normal fatty acids that are usually attached to enterobacterial lipid As. Other examples of related, long chain oxygen-containing fatty acids are 25-hydroxyhexacosanoic acid in Pseudomonas carboxydovorans (27) and 27keto-octacosanoic acid in Legionella (29).
We previously described enzymes in E. coli extracts that incorporate laurate and myristate into the key lipid A precur-sor, (Kdo) 2 -lipid IV A 2 ( Fig. 2) (30). The enzymes catalyzing these reactions are the products of the htrB and msbB genes, respectively (31,32). They have a remarkable requirement for the Kdo domain, as they do not acylate lipid IV A (30,32). This specificity explains why tetra-acylated lipid IV A rather than hexa-acylated lipid A accumulates in cells subjected to inhibition of Kdo biosynthesis or transfer (33)(34)(35)(36)(37)(38). The E. coli Kdo dependent late acyltransferases do not function with fatty acids longer than 14 carbons, and they require acyl chain activation by acyl carrier protein (30,32).
Given that R. leguminosarum extracts contain all the enzymes needed to synthesize the conserved precursor, (Kdo) 2lipid IV A (Fig. 2) (21), we examined the possibility that (Kdo) 2 -[4Ј-32 P]-lipid IV A may function as an acceptor for the 27hydroxyoctacosanoic acid moiety. In preliminary experiments (39), described in the accompanying manuscript (62), we ob-served a putative acylation reaction of (Kdo) 2 -[4Ј-32 P]-lipid IV A in crude extracts of R. leguminosarum. We now demonstrate that this reaction requires both a membrane component and a cytosolic factor, and that it indeed represents the addition of 27-hydroxyoctacosanoic acid to (Kdo) 2 -lipid IV A . It occurs in extracts of several strains of R. leguminosarum and R. meliloti, but not E. coli. The R. leguminosarum cytosolic factor has been purified, cloned, and sequenced. It is a new member of the acyl carrier protein family, designated AcpXL, that functions in transfers of long hydroxylated fatty acids. AcpXL is distinct from both the constitutive acyl carrier protein (40), AcpP, involved in synthesis of 12-18 carbon acyl chains and from the inducible acyl carrier protein (41,42), NodF, required for the generation of polyunsaturated fatty acids found in the nod factors of R. leguminosarum.

EXPERIMENTAL PROCEDURES
Materials-[␥-32 P]ATP and 32 P i were products of DuPont NEN. HEPES, MES, Kdo, and trypsin immobilized on inert beads were obtained from Sigma. The following materials and kits purchased were: Centricon, Centriprep, and Microcon centrifugation devices from Amicon; Silica Gel-60 thin layer plates, 0. 25  Bacterial Strains and Growth Conditions-Bacterial strains are listed in Table I. All cells were grown on TY medium, containing 5 g of tryptone and 3 g of yeast extract per liter, and supplemented with 10 mM CaCl 2 . Rhizobia were selected with 20 g/ml nalidixic acid (all strains) and 200 g/ml streptomycin sulfate (R. leguminosarum 8401, R. leguminosarum CE3, R. meliloti 1021, and R. meliloti GMI255). All strains were grown at 30°C. All strains are essentially wild-type (21,22), except that 8401 is lacking a symbiotic plasmid (J. A. Downie, John Innes Institute, Norwich, United Kingdom), and GMI255 carries a deletion of 280 kilobases in its symbiotic plasmid (43). Both 8401 and GM1255 are nod Ϫ .
Preparation of Cell-free Extracts-Bacterial cultures were harvested in late logarithmic phase (A 550 ϭ 0.6 -1.0) by centrifugation at 8000 ϫ g av for 15 min, and the cell pellet was resuspended in 50 mM HEPES, pH 7.5, to give a final protein concentration of 5-15 mg/ml. The cells were broken by passage through a French pressure cell at 18,000 p.s.i. Unbroken cells and debris were removed by another centrifugation at 2 The abbreviations used are: Kdo, 2-keto-3-deoxyoctonate; MES, 4-morpholineethanesulfonic acid; CAPS, 3-(cyclohexylamino)propanesulfonic acid; ORF, open reading frame; HPLC, high performance liquid chromatography; TMSO, trimethylsiloxy; FAME, fatty acid methyl esters; GLC-MS, gas liquid chromatography-mass spectrometry; nod factor, nodulation factor; ACP, acyl carrier protein.
FIG. 1. The unusual lipid A of R. leguminosarum contains a 27-hydroxyoctacosanoic acid moiety. Carbons of the E. coli glucosamine residues are numbered, as are the proposed corresponding carbons in the R. leguminosarum structure. The location of the esterlinked 27-hydroxyoctacosanoic acid moiety is not known, but based on the observation that it is not removed by Kraska methylation, direct attachment to the aminogluconate moiety is one possibility (23). The stereochemistry of the C27-hydroxyl group is not known. The esterlinked hydroxy fatty acids at positions 3 and 3Ј tend to be shorter than the N-linked acyl chains at positions 2 and 2Ј (23). Small amounts of 15-carbon ester-linked acyl chains are also present (not shown) (23). Based on the fatty acid composition, it is likely that some lipid A species contain R-3-hydroxymyristate at all four positions (i.e. 2, 3, 2Ј, and 3Ј) (23) (21,22). 8000 ϫ g av for 15 min. Extracts were prepared and handled at 0 -4°C. Protein concentrations were determined with bicinchoninic acid (44), using bovine serum albumin for the standard curve. Subcellular fractions were prepared by centrifugation of the crude extract in 25 mM HEPES, pH 7.5, at 150,000 ϫ g av for 60 min. The membrane pellet was removed, and resuspended in the original volume of buffer. The centrifugation was repeated on both the cytosol and the resuspended membranes. These final preparations are referred to as "cytosol" and "washed membranes." When strain 8401 was used for large scale preparations of cytosolic acyl donor, frozen cells were employed. A 150-liter culture was grown in a 200-liter New Brunswick fermenter, and cells were harvested at A 550 ϭ 1.0 using a Sharples centrifuge. The cell paste was stored at Ϫ80°C, and portions were thawed as needed in 2-3 ml of 50 mM HEPES, pH 7.5, per gram of cell paste. Cells were then broken by passage through a French pressure cell at 18,000 p.s.i., and debris was removed by centrifugation at 8000 ϫ g av for 15 min. The crude extract and acyl donating cytosolic fractions were then prepared as above.
Preparation of Radioactive Substrates-(Kdo) 2 -[4Ј-32 P]-lipid IV A and [4Ј-32 P]-lipid IV A were prepared as described previously (22,45). Aqueous dispersions of these lipids were stored at Ϫ20°C, and they were subjected to sonic irradiation in a bath sonicator for 2 min prior to use.
Assay of (Kdo) 2 -lipid IV A Acylation-The conditions for observing the acylation of (Kdo) 2 -[4Ј-32 P]-lipid IV A (formation of product a) were optimized. Reactions contained 50 mM HEPES, pH 8.2, 0.2% Triton X-100, and 10 M (Kdo) 2 -[4Ј-32 P]-lipid IV A , at approximately 20,000 -50,000 cpm/nmol, in 10 l. Crude extracts, membranes, and cytosolic fractions were present at the concentrations indicated. Acylation reactions were carried out at 30°C for 60 min, or as otherwise indicated. Following incubation, 5-l samples were withdrawn and spotted onto thin-layer chromatography plates that were then developed in the solvent chloroform, pyridine, 88% formic acid, water (30:70:16:10, v/v). Exposure to imaging screens at room temperature was carried out overnight. Extent of conversion of substrate to product(s) was measured using a Molecular Dynamics PhosphorImager operated with ImageQuant software.
Polyacrylamide Gel Electrophoresis-Polyacrylamide gels of 15% were cast in a Bio-Rad Protean II apparatus, according to the conditions of Laemmli (46), but without SDS. Samples were not denatured or reduced prior to loading in glycerol-containing buffer. Gels were run at constant current of 25 mA, using a buffer consisting of 3.1 g of Tris base and 14.4 g of glycine per liter. Gels were stained with Coomassie Blue.
Chemical Hydrolyses of (Kdo) 2 -[4Ј-32 P]-lipid IV A and Its Acylated Derivative (Product a)-Acylation reactions (10 l) were set up as described above, except that carrier-free (Kdo) 2 -[4Ј-32 P]-lipid IV A was used (1 ϫ 10 7 cpm/nmol), the chemical concentration of which was about 0.2 M. Reaction I contained no membranes or cytosolic factor. Reaction II contained 0.1 mg/ml 8401 membranes, and 0.3 mg/ml of DEAEpurified cytosolic factor (see below). Reaction III contained a system for generating (Kdo) 2 -[lauroyl]-[4Ј-32 P]-lipid IV A as a control, using 45 g/ml of an extract that overproduces HtrB but lacks MsbB (32), (Kdo) 2 -[4Ј-32 P]-lipid IV A , and 182 M lauroyl-ACP. Acylation reactions were incubated at 30°C for 60 min. Three 1-l portions were withdrawn from each reaction. Each 1-l portion was subjected to a different treatment. (a) For mild base hydrolysis, 1 l of reaction mixture was combined with 9 l of chloroform/methanol (2:1, v/v), and 1 l of 1.25 M NaOH. The mixture was incubated at room temperature for 30 min, after which time 1 l of 1.25 M HCl was added. After mixing, a 5-l portion was applied to a thin-layer plate. (b) For acid hydrolysis, 1 l of reaction mixture was combined with 9 l of 0.1 M HCl in a 0.65-ml Microfuge tube, which was sealed with a boiling clip. The tube was floated in a boiling water bath for 30 min. Next, 0.8 l of 1.25 M NaOH and 1 l of 10% SDS were added. The latter helped recover all the radioactivity from the tube, without affecting the hydrolysis. A 5-l sample was then applied to a thin-layer plate. (c) For mild acetate hydrolysis, the 1-l sample of reaction mixture was combined with 5 l of 50 mM sodium acetate, pH 4.5, 1 l of 10% SDS, and 3 l of H 2 O. The tube was heated in a boiling water bath as described for mild acid hydrolysis. Following incubation, a 5-l portion was applied directly to a thin-layer plate. The plate, containing the samples of all hydrolysis conditions, was then developed in chloroform, pyridine, 88% formic acid, water (30:70:16:10, v/v). Exposure to imaging screens at room temperature was carried out overnight.
Protease Treatment of Membranes and Cytosolic Factor-Either 20-l samples of 8401 membranes (0.5 mg/ml) or DEAE-Sepharose purified cytosolic factor (0.13 mg/ml) were combined with 20 l of immobilized trypsin or immobilized Pronase, used as supplied by the manufacturers. The mixtures were gently agitated on an automatic inverter for 60 min at room temperature. The beads were then lightly centrifuged to the bottom of the tubes, and the supernatants were removed. The treated membranes and cytosolic factor were assayed for activity in comparison to untreated controls. In addition, trypsin and Pronase beads were each added to the cytosolic factor, mixed, and immediately removed, so that the proteases did not have time to act. These control assays showed that the procedure itself did not leave residual protease or other material that would interfere during the assay itself.
Hydroxylamine Treatment of Membranes and Cytosolic Factor-Membranes and cytosolic factor were subjected to treatment with hydroxylamine at pH 6.5, to determine if either one contained a sensitive thioester linkage necessary for the observed activity (47). To this end, 0.35 mg of membrane protein or 0.14 mg of cytosolic factor, fractionated through the Centricon-50 step, were each treated with 1.2 M hydroxylamine-hydrochloride (derived from a 3.08 M stock solution and titrated to pH 6.5 with 10 M NaOH) in a final volume of 50 l. These reactions were incubated at room temperature for 2 h. As controls, identical samples were incubated at room temperature without hydroxylamine for 2 h. Then, all samples, including the untreated incubated ones, were exchanged twice with 0.5 ml of 50 mM HEPES, pH 7.5, using Centricon-30 devices, to remove the excess hydroxylamine prior to the acylation assays (described above). The final volumes were about 50 l.
Preparation of Cytosolic Factor for Microsequencing-Approximately 200 pmol of DEAE-purified factor was loaded onto a 15% polyacrylamide gel prepared without SDS or urea, using four lanes. Prestained polypeptide standards were run along the side of the gel as electrophoresis and transfer controls. Electrophoresis was carried out at 25 mA for 35-40 min using a running buffer consisting of 3.0 g of Tris base and 14.4 g of glycine/liter. Following electrophoresis, the gel was soaked at 4°C in 10 mM CAPS, pH 11.0, in 10% methanol. A piece of polyvinylidene difluoride membrane was prepared by first soaking in methanol, then water, and then the above CAPS/methanol solution. A semidry blotter (Bio-Rad) was used according to the manufacturer's directions, at 10 V for 30 min. The protein bands were visualized by Ponceau S staining, and the band of interest was cut out. NH 2 -terminal amino acid sequencing of intact protein, as well as trypsin digestion followed by amino acid sequencing of separated peptides, were carried Preparation of Cytosolic Factor for Mass Spectrometry-A sample of cytosolic factor, stored at Ϫ80°C for 2 weeks following the Superose-12 column (see below), was thawed and concentrated approximately 6-fold using a Microcon-30 device at 12,000 rpm at 4°C, according to manufacturer's directions. The final volume was 60 l, and it contained about 100 pmol of acyl donor. The sample was analyzed at the Harvard Microchemistry Facility by Dr. William S. Lane, using electrospray ionization mass spectrometry performed on a Finnigan TSQ700 triple quadrupole mass spectrometer (48). Prior to mass analysis, the sample was fractionated further on an 18-cm self-packed microcapillary Porose (reverse-phase type) column. Individual fractions, indicated on the tracing, were then analyzed by mass spectrometry. Molecular weights for each sample were calculated by averaging three to five of the most prominent peaks in the spectrum for each fraction. The experimental error is estimated as 1 atomic mass unit in 10,000 under optimal circumstances (48). For small samples, however, in which mixtures of related proteins are present, an error of 5-10 atomic mass units in 10,000 is not unusual. 3 DNA Sequencing-DNA was prepared from a culture of S17-1/ pCS115 by the BIGGER prep reagent kit (5Ј Ј 3Ј, Inc.), which yielded 400 g of plasmid from 500 ml of culture. Sequencing of open reading frame ORF * was carried out following the Sequenase Version 2.0 protocol with alkaline denaturation for double-stranded DNA sequencing. Five primers, 20 -22 bases in length, were used to confirm and extend the NH 2 -terminal portion of the Orf * sequence previously reported by Colonna-Romano et al. (49). The COOH-terminal end of Orf * was identified and validated by sequencing both strands. The primers used for DNA sequencing were: 1) ATCGACAAATTGCCATAGTG; 2) TATTG-CAGAAACCAGCGAGATC; 3) CCTCGGTATCGACAGCCTCG; 4) ACG-CAGGAAGTCAACGAAGGC; and 5) ATGCACTAGTTAGGAACGAA.

Preparation of the Enzymatic Acylation Product of (Kdo) 2 -[4Ј-32 P]lipid IV A for Fatty Acid Analysis-
The reaction mixture contained 50 mM HEPES, pH 8.2, 0.2% Triton X-100, 10 M (Kdo) 2 -[4Ј-32 P]-lipid IV A (50,000 cpm/nmol), 0.2 mg/ml 8401 washed membranes, and 0.75 mg of DEAE-purified cytosolic factor in 5 ml. The reaction was incubated in 10 portions in 1.5-ml microcentrifuge tubes at 30°C for 2 h. Thin layer chromatography analysis of the reaction showed that about 20% of the (Kdo) 2 -lipid IV A had been converted to the desired acylation product.
To isolate the lipids from the aqueous reaction mixture, 6.0 ml each of chloroform and methanol, and 75 l of HCl were added to the combined 5-ml reaction in a glass tube. Following vigorous mixing, the phases were separated using a clinical centrifuge. The lower phase was removed (6.0 ml), and the upper phase and interface were washed once with 6.0 ml of a pre-equilibrated acidic lower phase. The mixing and centrifugation were repeated, and the resulting lower phase was added to the previous one (total volume of 12 ml). To neutralize the sample, 30 drops of HPLC-grade pyridine were added to the lower phase, and then it was dried under a stream of nitrogen.
The sample was applied to a 2.5-ml silica gel column equilibrated in CHCl 3 , pyridine, methanol, 88% formic acid, water (55:60:2.5:15:3, v/v), which was then washed with 40 ml of the same solvent. Next, the column was washed with 10 ml of CHCl 3 , pyridine, 88% formic acid, water (30:70:16:10, v/v). Fractions of 1 ml were collected. Those containing the desired lipid were identified by thin layer analysis, using the CHCl 3 , pyridine, 88% formic acid, water (30:70:16:10, v/v) solvent system, and exposure of the thin-layer plate to an imaging screen to locate the radioactive spots. Fractions 10 -27 contained mainly the acylated metabolite of interest (product a) and only a small amount of contaminating (Kdo) 2 -[4Ј-32 P]-lipid IV A . The fractions were pooled (ϳ20 ml), and 40 ml of CHCl 3 /CH 3 OH (95:5, v/v) was added. Since the mixture was cloudy, a few drops of methanol was added until it cleared.
This sample was applied onto a 2-ml silica gel column equilibrated in CHCl 3 /CH 3 OH (95:5, v/v). The column was washed with 10 ml of CHCl 3 / CH 3 OH (95:5, v/v) to remove residual pyridine and formic acid. The radioactive lipid was then eluted with 7.6 ml of a single-phase, acidic Bligh and Dyer mixture (chloroform, methanol, 0.1 M HCl, 1:2:0.8 v/v). Next, 2 ml each of CHCl 3 and water were added to generate a two-phase system. After vigorous mixing, the phases were separated using a clinical centrifuge, and the lower phase was placed into a fresh tube. The upper phase was washed once with 2 ml of a pre-equilibrated, acidic lower phase. The mixing and centrifugation were repeated, and the resulting lower phase was combined with the first one. To this solution were added 15 drops of HPLC-quality pyridine, and it was then dried under a stream of nitrogen. The final yield of acylated (Kdo) 2 -[4Ј-32 P]lipid IV A was about 5 nmol. It was stored dry at Ϫ80°C.

Fatty Acid Analysis of Acylated (Kdo) 2 -[4Ј-32 P]-lipid IV A Generated in Vitro-
The purified, acylated (Kdo) 2 -[4Ј-32 P]-lipid IV A was subjected to fatty acid analysis by conversion of the hydroxy fatty acyl substituents to trimethylsiloxy (TMSO) fatty acid methyl esters (FAMEs), followed by analysis using combined gas liquid chromatography (GLC) and mass spectrometry (MS). The procedure was as described previously (23). Briefly, the sample was dissolved in 200 l of methanol containing 1 M HCl and heated to 80°C for 18 h. After cooling, the solvent was evaporated using a stream of air. The resulting hydroxy FAMEs were converted to TMSO-FAMEs by adding 200 l of Tri-Sil (Pierce Chemical Co.) and heating at 80°C for 20 min. The solvent was evaporated using a stream of air, and the resulting TMSO-FAMEs were dissolved in hexane and analyzed by combined GLC-MS using a 30-m DB1 capillary column from J & W Scientific (Folsom, CA).

RESULTS
In preliminary studies (39) that are reported in the accompanying manuscript (62), we demonstrated the conversion of (Kdo) 2 -[4Ј-32 P]-lipid IV A to a more hydrophobic derivative, designated a, that was formed in crude extracts of R. leguminosarum but not E. coli. Product a was proposed to represent a novel acylation of (Kdo) 2 -[4Ј-32 P]-lipid IV A , possibly an addition of 27-hydroxy-octacosanoic acid (62). The formation of product a required both a membrane associated and a cytosolic component (62). We now demonstrate that the cytosolic factor is a new member of the acyl carrier protein family and is acylated predominantly with 27-hydroxyoctacosanoic acid.
Purification of the Cytosolic Factor from R. leguminosarum-A cytosolic fraction was prepared from 36.5 g of frozen R. leguminosarum 8401 cells by preparation of a crude extract in 25 mM HEPES, pH 7.5, followed by centrifugation at 150,000 ϫ g av for 60 min. The membrane pellet was removed, and the ultracentrifugation was repeated to remove residual membranes. This cytosolic fraction, consisting of 1.2 g of protein in 92 ml, was then divided and centrifuged through eight Centriprep 100 units at 500 ϫ g av , according to the manufacturer's directions. When the retentate became very viscous, a total of 40 ml of 25 mM HEPES, pH 7.5, was added with gentle stirring. The centrifugation was then continued until a total of 70 ml had passed through the Centri-prep 100 membranes. The process took several hours and was conducted at 4°C. The pooled material was referred to as the "C100 filtrate." The C100 filtrate was then centrifuged through two Centri-prep 50 units at 1500 ϫ g av . The upper chambers were refilled with C100 filtrate as they emptied. The final "C50 retentate" was collected in a total volume of 10.0 ml. Approximately 5.0 ml of the C50 retentate was incubated at 65°C for 15 min, causing about half of the protein to precipitate without loss of acyl donating activity. The heat-treated material was clarified in a table top microcentrifuge for 4 min.
A 4.0-ml portion of the heat-treated material (11.4 mg of protein) was applied to a 1.8-ml DEAE-Sepharose column that was equilibrated in 25 mM HEPES, pH 7.5. Since the active cytosolic factor is relatively stable at this stage, the column was run at room temperature. After application of the sample, the column was washed with 4 ml of 25 mM HEPES, pH 7.5. Next, the column was washed stepwise with 6 ml of 0.1 M NaCl, 6 ml of 0.2 M NaCl, and 6 ml of 0.3 M NaCl, all buffered with 25 mM HEPES, pH 7.5. Fractions of approximately 1 ml were collected and assayed for acyl donor activity, using 4 l of each column fraction in a 10-l reaction mixture, as described under "Experimental Procedures." The run-through and wash were collected as a batch of 8 ml and contained 5.2 mg of protein. The profiles of acyl donor activity and protein eluting from the DEAE-Sepharose column are shown in Fig. 3A. The activity is expressed as nanomoles of cytosolic factor per fraction, obtained by calculating the nanomoles of acylated (Kdo) 2 -[4Ј-32 P]-lipid IV A generated with excess enzyme. We assume that the cytosolic factor is limiting in the acylation reactions and that it is completely consumed (see below). The cytosolic acyl donor eluted with 0.3 M NaCl (fractions 19 and 20). This pool contained 0.33 mg of protein in 2.0 ml. A small amount of additional protein, but no activity, eluted with a subsequent 1.0 M NaCl wash (not shown). NaCl concentrations of 0.05-0.20 M in the acylation assay were not inhibitory (not shown). The active cytosolic factor consistently eluted at 0.3 M NaCl with the above protocol, or late in the 0.2 M wash, if a prolonged 0.2 M wash was used.
The active fractions eluting from DEAE-Sepharose at 0.3 M NaCl were concentrated to 250 l in a Centricon C30 unit. A 200-l portion of this concentrated pool was then applied to a Superose-12 size-exclusion column (27 ml), connected to a Waters 650 Chromatography Plus System, operated at 0.5 ml/min and 4°C. Absorbance was detected at 280 nm. After sample application, the column was washed with 25 mM HEPES, pH 7.5, and fractions of 2 ml were collected. The column was eluted for 70 min. Acylation activity was assayed using 4 l of each fraction per 10-l assay, as above. The absorbance at 280 nm and the activity profile are shown in Fig. 3B. Fractions 13-15 contained the desired activity. The active factor always eluted on the downward slope of the main protein peak of the Superose-12 column. Fractions 8 -9 (minutes 15-17) define the excluded volume, and fractions 32-34 (minutes 64 -68) the included volume. The cytosolic factor was rendered more labile once purified through the Superose column. Room temperature incubations and repeated freeze-thaw cycles were avoided after the Superose step.
Identification of a Protein Band Migrating with Acyl Donor Activity-Fractions 12-15 from the Superose column were analyzed by native polyacrylamide gel electrophoresis followed by staining with Coomassie Blue (Fig. 4). The band indicated by the arrow follows the acylation activity profile (Fig. 3B). Furthermore, acyl donor activity is present in this band, as judged by excising it from a preparative native gel, allowing it to diffuse into buffer solution, and assaying for acylation of (Kdo) 2 -[4Ј-32 P]-lipid IV A .
Characterization of the Acylation Assay with Purified Soluble Donor-The purified cytosolic factor, together with a washed membrane preparation from R. leguminosarum, represent an efficient in vitro system for generating metabolite a, presumed to be an acylated derivative of (Kdo) 2 -[4Ј-32 P]-lipid IV A (Fig. 5). The reaction depends absolutely on the presence of membranes and cytosolic factor. The reaction also requires 0.1-0.5% Triton X-100, but it is not stimulated by ATP, dithiothreitol, acyl coenzyme A, or various acyl ACPs of E. coli, such as lauroyl-or myristoyl-ACP. From 0 to 20 min at fixed acyl donor concentration, the conversion of (Kdo) 2 -[4Ј-32 P]-lipid IV A to a is relatively linear with respect to time and membrane protein concentration (Fig. 6). The cytosolic factor is stable under assay conditions for 60 min, as determined by a preincubation prior to the addition of membranes, but the membranes themselves lose 30 -40% of their activity in 1 h when subjected to preincubation under assay conditions in the absence of the cytosolic factor (data not shown). When larger amounts of cytosolic factor are present in the assay, 80 -90% conversion of (Kdo) 2 -[4Ј-32 P]-lipid IV A to a is observed in 15-20 min. In most of the assays used to follow the purification of the cytosolic factor, 20 -30% conversion of (Kdo) 2 -[4Ј-32 P]-lipid IV A to a was observed in 15 min, and the reaction did not proceed further when incubated for another 30 -60 min. We assume that the cytosolic factor was quantitatively consumed in the reaction, and thus calculate the amount of cytosolic factor in nanomoles. However, the actual amount might be underestimated slightly, depending on the equilibrium constant.
Protease Sensitivity of the Cytosolic Factor-To characterize further the chemical nature of the membrane component and the cytosolic factor, we subjected them both to proteolysis, as described under "Experimental Procedures." Washed membranes and DEAE-purified cytosolic factor were each treated with immobilized trypsin or immobilized Pronase for 60 min at room temperature, and the treated components were assayed to determine the extent of inactivation (Table II). The cytosolic factor was very sensitive to both proteases. The membrane component was also sensitive to trypsin but not to Pronase. It is possible that the membrane-bound component is not accessible to Pronase.
A Neutral Hydroxylamine Labile Linkage Is Required for Activity of the Cytosolic Factor-To test whether the cytosolic factor might have a labile thioester linkage needed for activity, as would an acyl ACP, we treated both the washed membranes and the cytosolic factor (purified to the C50 retentate stage) with hydroxylamine at pH 6.5, as described under "Experimental Procedures." Since the treatment involved a long incubation at room temperature and recovery by centrifugation through a membrane filtration device to remove the hydroxylamine, parallel samples without hydroxylamine were also processed as controls. The hydroxylamine-treated components were assayed along with their untreated counterparts to determine the extent of inactivation. As shown in Table III, only 20% of the membrane activity was inactivated by hydroxylamine (7.1 versus 8.9 pmol of product formed). In contrast, cytosolic factor activity was almost completely inactivated by hydroxylamine treatment (1.9 pmol of product generated versus 12.3 pmol for the control) (Table III). Therefore, the cytosolic factor may contain a thioester linkage, providing further support for the view that the formation of product a from (Kdo) 2 -[4Ј-32 P]-lipid IV A represents an acylation. Based on these results, we hypothesize that the cytosolic factor is a novel acyl ACP and that the membrane component is an acyltransferase.

Acylation of (Kdo) 2 -[4Ј-32 P]-lipid IV A in Other
Strains of Rhizobium-If the enzymatic reaction described above indeed represents the addition of 27-hydroxyoctacosanoic acid to (Kdo) 2 -[4Ј-32 P]-lipid IV A , it should be present in extracts of all strains of Rhizobium. To examine this issue, numerous Rhizobium strains, including R. meliloti and four biovars of R. leguminosarum were assayed for modification of (Kdo) 2 -[4Ј-32 P]-lipid IV A in crude extracts and for stimulation of the activity by partially purified R. leguminosarum cytosolic factor (Table IV). All Rhizobium strains tested showed the putative acylation of (Kdo) 2 -[4Ј-32 P]-lipid IV A in crude extracts in varying amounts, indicating the presence of both the membrane and cytosolic components. In addition, all extracts produced much more of product a when excess cytosolic factor was added to the assay. No product was detected in any of the E. coli extracts, either with or without the added cytosolic factor from R. leguminosarum. These results indicate that conversion of (Kdo) 2 -[4Ј-32 P]lipid IV A to product a occurs in extracts of at least six common strains of Rhizobium. In addition, the positive results obtained with R. meliloti GMI255 indicate that neither the membranebound nor the cytosolic components are encoded in the nod region of the R. meliloti chromosome.

Acid and Base Hydrolyses of Product a Derived from (Kdo) 2 -[4Ј-32 P]-lipid IV A in Extracts of R. leguminosarum-
To test the idea that product a is a novel acylated form of (Kdo) 2 -[4Ј-32 P]lipid IV A , we subjected the material generated with washed membranes of R. leguminosarum and purified acyl donor to mild base, acid, and pH 4.5 sodium acetate hydrolyses, as described under "Experimental Procedures." Incubation I in Fig. 7 is a reaction mixture containing (Kdo) 2 -[4Ј-32 P]-lipid IV A but no membranes or cytosolic factor, and it served as a control. Incubation II of Fig. 7 contained the complete, reconstituted R.

FIG. 5. Conversion of (Kdo) 2 -[4-32 P]-lipid IV A to product a using washed membranes and DEAE-Sepharose purified cytosolic factor.
Reactions contained 50 mM HEPES, pH 8.2, 0.2% Triton X-100, and 10 M (Kdo) 2 -[4Ј-32 P]-lipid IV A , at approximately 50,000 cpm/nmol, in 10 l. The cytosolic factor was from the DEAE-Sepharose column peak, described in Fig. 3A, and was present at 0.1 mg/ml. Washed membranes were used as the source of enzyme at 0.1 mg/ml. Acylation reactions were allowed to proceed at 30°C for 60 min.

FIG. 6. Time course of the conversion of (Kdo) 2 -[4-32 P]-lipid IV A to product a at various membrane protein concentrations.
Reactions contained 50 mM HEPES, pH 8.2, 0.2% Triton X-100, and 10 M (Kdo) 2 -[4Ј-32 P]-lipid IV A , at approximately 50,000 cpm/nmol, as in Fig. 5. The reaction volumes were 50 l for each membrane protein concentration. The cytosolic factor was the DEAE-Sepharose peak fraction (Fig. 3A) at 50 g/ml. Samples of 5 l were withdrawn at 0, 10, 20, 30, 45, and 60 min. The total final conversion to product a is about half of that observed in Fig. 5, since only half the amount of the DEAE-Sepharose purified factor was employed.
The 32 P-labeled lipids were not purified from their respective incubation mixtures prior to the chemical hydrolyses shown in Fig. 7. Mild base treatment (Fig. 7, Base) of product a (incubation II) generated predominantly the same material as was generated from (Kdo) 2

-[4Ј-32 P]-lipid IV A (incubation I).
This result is consistent with the idea that product a contains an extra ester-linked fatty acid. However, acid hydrolysis of product a yielded a more hydrophobic radioactive lipid than was obtained with (Kdo) 2 -[4Ј-32 P]-lipid IV A (Fig. 7, Acid). This is the expected result if product a contains an additional ester-linked acyl chain that is attached to the glucosamine disaccharide moiety, but not to the Kdo or the phosphate residues, since the anomeric phosphate and the Kdo disaccharide are removed by acid treatment, and an acyl phosphate would not be stable to acid hydrolysis. Last, acetate hydrolysis at pH 4.5 in SDS (Fig.  7, Acetate), which removes the Kdo but not the anomeric phosphate, also yielded a more hydrophobic compound from product a than from (Kdo) 2 -[4Ј-32 P]-lipid IV A , consistent with the results of the acid treatment.
Formation of Product a Requires the Kdo Domain-Washed membranes and DEAE-Sepharose purified cytosolic factor were used to examine the acceptor specificity for the proposed acylation. As shown in Fig. 8, (Kdo) 2 -[4Ј-32 P]-lipid IV A and [4Ј-32 P]-lipid IV A were compared as substrates. Under conditions resulting in 35% conversion of (Kdo) 2 -[4Ј-32 P]-lipid IV A to product a only 0.3% of [4Ј-32 P]-lipid IV A was converted to a more rapidly migrating form. Therefore, there is at least a 100-fold preference for an acceptor bearing the Kdo disaccharide. In this respect, the enzymatic generation of product a in extracts of R. leguminosarum closely resembles the HtrB catalyzed transfer of laurate from lauroyl-ACP to (Kdo) 2 -[4Ј-32 P]lipid IV A in E. coli (32). However, washed R. leguminosarum membranes are inactive with lauroyl-ACP as the donor and (Kdo) 2 -[4Ј-32 P]-lipid IV A as the acceptor (data not shown).
Amino Acid Sequence of the Isolated Cytosolic Factor Reveals Identity with orf * -The protein band identified as the active cytosolic factor (Fig. 4) was transferred to a polyvinylidene difluoride membrane and subjected to amino acid sequencing, as described under "Experimental Procedures." The predominant NH 2 -terminal sequence (20 amino acids) was TATFDK-VADIIAETSEIDRA (Fig. 9). In addition, about 20% of the immobilized protein started with methionine and then continued with the above sequence. A sequence data base search (Blastp) revealed an exact match with an internal segment of a  theoretical, partially sequenced protein of R. leguminosarum, previously designated Orf * (49). Orf * is located next to Orf240, a transcription factor involved in the regulation of nitrogen fixation (49). Only the first 51 amino acids of Orf * were sequenced by Colonna-Romano et al. (49). The start codon proposed by Colonna-Romano et al. (49) for Orf * is located 14 codons upstream of our experimentally determined NH 2 terminus (Fig. 9). The published sequence specifies valine (a GTG codon) where 20% of our purified protein contains methionine (i.e. the likely start of our protein). Since there are numerous examples in which the codon GTG is used to initiate the synthesis of a protein with N-formylmethionine (50), our data strongly support the view that GTG is the true start codon of Orf * . It is very unlikely that Orf * contains an atypical leader sequence of 13-14 amino acids, which is subsequently removed during protein maturation.
We also obtained internal amino acid sequences of Orf * by trypsin digestion of a second polyvinylidene difluoride blot followed by HPLC separation of the major peptides. The fragments that were detected in this manner are underlined in Fig.  9. In addition to the three fragments near the NH 2 terminus, which confirmed the published DNA sequence, one unique fragment closer to the COOH terminus of Orf * was also detected. The latter was confirmed by our own sequencing of the complete Orf * DNA (see below).
DNA Sequence of Entire orf * Gene-The DNA coding for the entire Orf * protein was sequenced, using plasmid pCS115 (49), as described under "Experimental Procedures" (Fig. 9). Plas-mid pCS115 is about 18 kilobases in size and is a derivative of plasmid pACYC177 (49). The partial Orf * sequence reported previously (49) was confirmed, and an additional 123 bases were sequenced to the stop codon. The peptide fragment closest to the COOH terminus, as determined by internal peptide sequencing (see above), was identified in the DNA sequence. If the NH 2 -terminal methionine is included, Orf * is 92 amino acids long. Assuming no covalent modifications, the protein by itself has a predicted molecular weight of 10,277.53 (10, 146.34 without the NH 2 -terminal methionine).
Mass Spectrometry of Purified Orf * -To determine whether or not the purified cytosolic factor contained 4Ј-phosphopantetheine and 27-hydroxyoctacosanoic acid, mass spectrometry was performed using material purified through the Superose-12 stage. Just prior to mass spectrometry, the sample was subjected to reverse-phase chromatography on a microcapillary column (Fig. 10) (48). Despite the fact that the cytosolic factor appeared to migrate as a single band as judged by electrophoresis on a native gel (Fig. 4), the factor was resolved into several related components (Fig. 10). Seven fractions corresponding to major peaks were collected as indicated, and they were analyzed individually by electrospray ionization mass spectrometry (48). For each fraction, the molecular weights of the major component(s) were calculated (Table V), as described under "Experimental Procedures." The major molecular species, eluting last from the microcapillary column (scan 363 in Fig. 10), has a molecular weight of 10910.2, in excellent agreement with the calculated mass of 10909.4 for Orf * , lacking the NH 2 -terminal methionine but containing 4Ј-phosphopantetheine and 27-hydroxyoctacosanoic acid. Four smaller peaks (emerging at 359, 356, 352, and 328) are interpreted as Orf * derivatized with 4Ј-phosphopantetheine and progressively shorter hydroxylated fatty acids differing by two carbon increments from C26 to C20. The order of elution of the peaks (328, 352, 356, 359, and 363) in Fig. 10 is consistent with what is expected on a reverse-phase column for a family of fatty acids of increasing chain length. There is a minor peak that can be interpreted as Orf * derivatized with 4Ј-phosphopantetheine, but without an acyl chain (scan 328), and another peak that is interpreted as a disulfide-linked dimer between two Orf * molecules both lacking acyl chains (scan 336). All of the theoretical molecular weights are within 5 mass units of FIG. 9. Sequence of the complete orf * gene and experimental amino acid sequences obtained from the purified cytosolic factor. The DNA sequence was derived from pCS115 and begins with the previously proposed start codon for Orf * , based on the work of Colonna-Romano et al. (49). Our data indicate the real start site is the GTG located 13 codons downstream, corresponding to the experimentally determined NH 2 -terminal sequence of the gel purified factor shown in Fig. 4. The proposed GTG start codon was also verified by mass spectrometry of the purified protein. Underlined amino acids were those actually identified by sequencing of intact protein or isolated peptides. The DNA sequence between the asterisks was previously reported by Colonna-Romano et al. (49), and the rest was determined in the present study. The consensus sequence for 4Ј-phosphopantetheine attachment is enclosed in the box.
FIG. 10. Microcapillary HPLC fractionation of Superose-12 purified cytosolic factor. The sample was prepared as described under "Experimental Procedures," and the material was fractionated further into a family of related proteins on an 18-cm Porose (reverse phase) column just prior to mass spectrometry. Individual fractions (as indicated) representing samples of all the major peaks were analyzed. The calculated masses of the major components of each peak are summarized in Table V. Units of the x axis are seconds, whereas units on the y axis are relative current reflecting the presence of protein mass (48).
those that are actually observed (Table V). This reflects the accuracy anticipated for a complex mixture of related proteins, such as that shown in Fig. 10. The proposed covalent structures of each major component based the mass spectrometry analysis (Table V) are shown in Fig. 11.
The primary components in fractions 328, 336, 352, 356, 359, and 363 (Table V) are each accompanied by secondary mass peaks that are 131 mass units larger, comprising 15-20% of the total intensity (Table V, not shown). These are protein species containing an additional NH 2 -terminal methionine, consistent with the amino acid sequencing data (see above). All of these molecular weights are also interpretable as a family of related Orf * molecules containing 4Ј-phosphopantetheine and long chain hydroxy fatty acids, as shown in Fig. 11, with the exception of the first peak (scan 326). The latter probably represents an entirely different protein impurity, and it does not have a companion peak of ϩ131 atomic mass units (not shown).
Given the results of the mass spectrometry, it is apparent that the cytosolic factor exhibits anomalous behavior in the Centricon centrifugations used to isolate it (see above), where it displayed an apparent molecular mass of 50,000 -100,000 daltons. It is possible that acylated Orf * aggregates with itself in aqueous solutions because of its long acyl chain.
Fatty Acid Analysis of the Lipid Product-To determine if all of the molecular species of purified Orf * (Table V and Fig. 11) can donate acyl chains to (Kdo) 2 -[4Ј-32 P]-lipid IV A in cell extracts, product a was generated in vitro, isolated, and subjected to fatty acid analysis. Analysis of the TMSO-FAMEs derived from product a was accomplished by GLC-MS, using ion selective scanning. The GLC selective ion profiles are shown in Fig.  12. Product a contained 27-hydroxyoctacosanoic acid as the major acyl chain incorporated into (Kdo) 2 -[4Ј-32 P]-lipid IV A . The substrate did not contain any detectable fatty acid other than ␤-hydroxymyristate (Fig. 12). Small amounts of 23-hydroxytetracosanoic acid and 21-hydroxydocosanoic acid were observed in product a (Fig. 12), suggesting that some of the minor acylated forms of Orf * are competent as donors in vitro and are not just precursors of the major 28 carbon species. In vivo, however, additional control mechanisms may exist to prevent the incorporation of the shorter hydroxylated fatty acids into lipid A. The results of Fig. 12 show unequivocally that our system for generating product a represents the incorporation of long (1) hydroxylated fatty acids into (Kdo) 2 -[4Ј-32 P]-lipid IV A , predominantly 27-hydroxyoctacosanoate. DISCUSSION Long chain 1-hydroxylated fatty acids, usually 27-hydroxyoctacosanoate, are esterified to the lipid A moiety of Rhizobium lipopolysaccharides (26,27). We have discovered a system for converting the key lipopolysaccharide precursor, (Kdo) 2 -lipid IV A (3, 21, 54), to a more hydrophobic metabolite, designated product a, in extracts of R. leguminosarum. This reaction represents the first description of an enzymatic acylation involving 27-hydroxyoctacosanoate based on the following observations. 1) Product a formation occurs in extracts of all strains of Rhizobium examined. Product a formation requires a membrane-bound component and a cytosolic factor, neither of which are present in E. coli extracts. 2) Product a is significantly more hydrophobic than (Kdo) 2 -[lauroyl]-lipid IV A generated by the HtrB acyltransferase of E. coli (32), indicating that it is acylated with a longer fatty acid moiety. The hydrophobic modification present in product a is attached through an ester linkage to the lipid IV A region. 3) Fatty acid analyses and mass spectrometry demonstrate that 27-hydroxyoctacosanoate is indeed present in product a. 4) The cytosolic factor purified from R. leguminosarum is a new member of the acyl carrier protein family, designated AcpXL, that is distinct from AcpP (40) and NodF (41) of Rhizobium. The purified active protein factor is acylated mainly with 27-hydroxyoctacosanoate, but also with shorter chains ranging in size from C20 to C26. The acyl chains are probably attached as thioesters, given the sensitivity of acylated AcpXL to neutral hydroxylamine.
The exact location of the 27-hydroxyoctacosanoate that is incorporated into the acceptor, (Kdo) 2 -lipid IV A , remains to be established. Based on the available free OH groups, it must be linked either to one of the four R-3-hydroxymyristoyl moieties or to the 4-OH of the proximal glucosamine residue (compare Figs. 1 and 2). In previous studies with mature lipid A isolated FIG. 11. Proposed composition of the acylated Orf * (AcpXL) and related molecules in the cytosolic factor isolated from R. leguminosarum. The mass spectrometry data supporting the above assignments is summarized in Table V. Based on the relative intensities of the peaks in Fig. 10, it appears that the 27-hydroxyoctacosanoic acid derivative of Orf * is probably the most abundant in the isolated mixture. The peak of comparable size at 328 in Fig. 10 is a mixture of unacylated Orf * and a C20 derivative. from R. leguminosarum, Bhat et al. (23) concluded that the 27-hydroxyoctacosanoate was not part of an acyloxyacyl unit, since it was not released by Kraska methylation (28). This conclusion was based on the assumption that acyloxyacyl moieties containing such long hydrocarbon chains have the same chemical reactivities as do the more typical, shorter acyloxyacyl residues found in enteric Gram-negative bacteria (23,28). Accordingly, Bhat et al. (23) favored the view that the 27hydroxyoctacosanoate was esterified to the 4 or 5-position of the aminogluconate moiety of the R. leguminosarum lipid A. Acylation at position 5 is not possible with (Kdo) 2 -lipid IV A as the acceptor, since O-5 is part of the proximal pyranose ring (Fig. 2). Unequivocal identification of the site of 27-hydroxyoctacosanoate attachment will require NMR spectroscopy of milligram quantities of pure product a. An intriguing possibility is posed by the observation that the acylation of the acceptor is absolutely dependent upon the presence of the Kdo domain. Lipid IV A is not a substrate for the transfer of 27-hydroxyoctacosanoate (Fig. 8). A similar Kdo requirement is observed in E. coli extracts for the incorporation of laurate by HtrB (30,32). Perhaps, the membrane enzyme of Rhizobium that we have identified is homologous to HtrB. Polymerase chain reaction analysis with conserved sequences from diverse htrB genes might permit the cloning of the R. leguminosarum membrane component that is required for product a formation.
AcpXL contains a DSLD motif, the site for 4Ј-phosphopantetheine attachment in the better characterized acyl carrier proteins, such as the constitutive ACPs of E. coli (51)(52)(53) and Rhizobium (40) (Fig. 13). The overall sequence identity between AcpXL and Rhizobium AcpP is only 26% (Fig. 13). The overall sequence identity between AcpXL and NodF (the inducible ACP required for the incorporation of polyunsaturated acyl chains into nod factors) (41,42) is 22% (Fig. 13). Since NodF and AcpP are themselves only 23% identical (Fig. 13), it appears that Rhizobium NodF, AcpP, and AcpXL are each distinctly different members of the ACP family, as judged by their sequence diversity. In contrast, the constitutive AcpP proteins of E. coli (53) and Rhizobium (40) are 59% identical (Fig. 13). Furthermore, the E. coli and Rhizobium AcpPs are more related to the ACP of Neurospora mitochondria (55) or to chloroplast ACPs (56) than to AcpXL. Data base searches using the whole AcpXL sequence as the probe do not show it to be homologous to any other proteins besides those of the ACP family. The chromosome of Hemophilus influenzae contains only the acpP gene (57).
Long chain (C18-C26) fatty acyl moieties, hydroxylated at the 1 position, are found in the nod factors of R. meliloti but they are not incorporated into R. meliloti lipid A (58). Conversely, the hydroxylated C28 acyl chains, found in all Rhizobial lipid As, are not incorporated into the nod factors (26,27,4 O. Geiger, personal communication. FIG. 12. Analysis of the fatty acyl chains present in the substrate (Kdo) 2 -lipid IV A and in product a. Approximately 5 nmol of product a was isolated and subjected to fatty acid analysis as described under "Experimental Procedures." Fatty acid analysis of the substrate control, (Kdo) 2 -lipid IV A , was also performed. The TMSO-FAME derivatives were analyzed by GLC-MS, and due to the small sample size, selective ion scanning was used to monitor the elution of the fatty acid derivatives. Panel A shows the ion scan profile at m/z 175. This ion is characteristic of the TMSO-FAMEs of ␤-hydroxy fatty acids. Both the retention time and the fragment ions at m/z 175 and 315 (mass spectrum not shown) were consistent with the TMSO-FAME of ␤-hydroxymyristic acid. This fatty acid was found in both product a and the substrate control. Panel B shows the ion scan profile at m/z 117. This ion is characteristic of the TMSO-FAMEs of 1-hydroxylated fatty acids. There were no 1-hydroxylated fatty acids in the substrate control (the peak marked with an asterisk (*) is not due to a fatty acid derivative). The TMSO-FAME of 27-hydroxyoctacosanoic acid was present in greatest amount. Its retention time and characteristic fragment ions (spectrum not shown) of m/z 117 and 511 were identical to those observed for this fatty acid derivative from R. leguminosarum lipid A (23,27). Smaller amounts of TMSO-FAMEs of C22 (fragment ions of m/z 117 and 427) and C24 (fragment ions of m/z 117 and 455) homologs were also detected in product a.

FIG. 13. Relationship of AcpXL to other acyl carrier proteins.
The sequences used for these comparisons are: 1) the constitutive AcpP of R. meliloti (40); 2) the constitutive AcpP of E. coli (53); 3) the inducible NodF of R. leguminosarum (41); and 4) AcpXL of R. leguminosarum. The constitutive ACPs of R. meliloti and R. leguminosarum are greater than 90% identical, but the sequence of the latter is not published. 4 58). The C28 chains are present in R. meliloti lipid A even when the C18-C26 1-hydroxylated fatty acids are eliminated from the nod factors by the nodD3 mutation (58). The N-acyltransferases that are involved nod factor biosynthesis (42,59) are almost certainly distinct from the O-acyltransferase that incorporates 27-hydroxyoctacosanoate into lipid A. However, both nod factor and lipid A biosynthesis may share the common donating protein, AcpXL, since 1-hydroxyacyl fatty acids ranging from C20 to C28 are attached to AcpXL (Fig. 10 and Table V). AcpXL is not encoded in the nod region of the R. meliloti chromosome. However, other genes, such as nodD3 and syrM that determine the appearance of the 1-hydroxylated fatty acids in the nod factors, are located in the nod region (58). Isolation of mutants defective in AcpXL and/or the acyltransferase that incorporates 27-hydroxyoctacosanoate into lipid A of R. leguminosarum may shed light on the function of this unusual fatty acid.
The stereochemistry and origin of the 1-hydroxyl group remain to be established (23,26,27). If the penultimate hydroxyl is somehow left in place during the biosynthesis of long fatty acids on AcpXL, then something unusual must be happening during the first cycle of fatty acid elongation on AcpXL. One possibility is that there is a distinct set of enzymes that carry out the conversion of C2 to C4 on AcpXL. Alternatively, ␤-hydroxybutyrate may be transferred directly from ␤-hydroxybutyryl-coenzyme A to free AcpXL. Perhaps, the constitutive dehydrases of the fatty acid biosynthesis cycle (analogous to FabZ in E. coli) (52, 60) cannot act on 4-carbon hydroxy acyl chains attached to AcpXL, resulting in the failure to remove the 1-hydroxyl. Further studies of the structure and function of acylated AcpXL should answer these questions.