KpsF is the arabinose-5-phosphate isomerase required for 3-deoxy-D-manno-octulosonic acid biosynthesis and for both lipooligosaccharide assembly and capsular polysaccharide expression in Neisseria meningitidis.

We have identified and defined the function of kpsF of Neisseria meningitidis and the homologues of kpsF in encapsulated K1 and K5 Escherichia coli. KpsF was shown to be the arabinose-5-phosphate isomerase, an enzyme not previously identified in prokaryotes, that mediates the interconversion of ribulose 5-phosphate and arabinose 5-phosphate. KpsF is required for 3-deoxy-d-manno-octulosonic acid (Kdo) biosynthesis in N. meningitidis. Mutation of kpsF or the gene encoding the CMP-Kdo synthetase (kpsU/kdsB) in N. meningitidis resulted in expression of a lipooligosaccharide (LOS) structure that contained only lipid A and reduced capsule expression in the five invasive disease-associated meningococcal serogroups (A, B, C, Y, and W-135). The step linking meningococcal capsule and LOS biosynthesis was shown to be Kdo production as the expression of capsule was wild type in a Kdo transferase (kdtA) mutant. Thus, in addition to lipooligosaccharide assembly, Kdo is required for meningococcal capsular polysaccharide expression. Furthermore, N. meningitidis, unlike enteric Gram-negative bacteria, can survive and synthesize only unglycosylated lipid A.

Neisseria meningitidis, an exclusive pathogen of humans, is the cause of epidemic bacterial meningitis and sepsis. Capsular polysaccharide and lipooligosaccharide (LOS 1 or endotoxin) are critical virulence factors in meningococcal pathogenesis. For example, both structures contribute to the resistance of menin-gococci to bactericidal activity of human sera (1,2). Structural differences in capsule and LOS determine meningococcal serogroups and immunotypes, respectively. Of the 13 different capsule serogroups so far defined, 5 serogroups (A, B, C, Y, and W-135) cause the majority of cases of invasive meningococcal disease. With the exception of the serogroup A capsule, (␣136)linked N-acetylmannosamine 1-phosphate, meningococcal capsules associated with invasive diseases contain sialic acid as follows: serogroup B, ␣238-linked N-acetylneuraminic acid (NeuAc); serogroup C, ␣239-linked partially O-acetylated NeuAc; serogroup Y, an alternating sequence of D-glucose and partially O-acetylated NeuAc; and serogroup W-135, an alternating sequence of D-galactose and NeuAc (3)(4)(5)(6). Each of the capsular polymers also contains a 1,2-diacylglycerol phospholipid anchor (7).
Meningococcal LOS consists of a lipid A moiety, a conserved inner core composed of two heptoses linked to two Kdo sugars, and an outer core with variable oligosaccharide composition. The meningococcal lipid A is distinct from that of Escherichia coli and is composed of a ␤1Ј,6-linked disaccharide of glucosamine acylated with ␤-hydroxymyristates and ␤-hydroxylaurates at 2-and 2Ј-, and 3-and 3Ј-positions respectively, and symmetrical acyloxyacyl linkages of laurate residues are located at 2-and 2Ј-positions (8). More than 30 genes involved in the biosynthesis of lipid A, heptose, Kdo, and the outer core polysaccharides have been identified (9), but the genes required for the assembly of Kdo are not completely defined (10).
The genetic basis of meningococcal capsule expression has been studied extensively. The capsule (cps) locus consists of a four-gene biosynthesis operon that encodes the production of sialic acid or N-acetylmannosamine and the formation of capsule polymers, whereas the divergently transcribed ctrABCD transport operon encodes proteins responsible for capsule translocation (11)(12)(13). Nearby are two genes, lipA and lipB, proposed to be involved in lipidation of capsule polymers (14). No genes outside the cps locus have been shown to participate in meningococcal capsule expression. E. coli K1 strains also express a capsule composed of ␣238-linked polysialic acid. The kps locus of K1 E. coli capsule expression has also been well characterized and, when compared with the meningococcal locus (Fig. 1A), contains several "extra" genes including kpsF, kpsD, kpsU, neuD, and neuE (15). KpsU has been shown to encode a second copy of the CMP-Kdo synthetase, KdsB (16). KpsD is a periplasmic protein, and mutation of kpsD results in periplasmic accumulation of capsular polysaccharide (17). The functions of KpsF, NeuD, and NeuE are currently unknown. Furthermore, it is not known whether these genes are present in meningococci and whether they play a role in meningococcal capsule expression.

Materials and Bacterial Strains
Bacterial strains, plasmids, and primers used in this study are described in Table I. Monoclonal antibodies for meningococcal serogroup B (2-2-B), C (4-2-C), Y (5-2-Y), and W-135 (7-1-W) capsular polysaccharides were kindly provided by Wendell Zollinger (Water Reed Army Institute of Research, Washington, D. C.). Monoclonal antibody M2 against FLAG epitope and antibiotics were obtained from Sigma. Restriction enzymes were purchased from New England Biolabs. Polyclonal antiserum to the KpsF protein was raised in rabbits (Covance Co.).

Growth Conditions
Meningococcal strains were grown with 3.5% CO 2 at 37°C unless specified otherwise. GC base agar (Difco), supplemented with 0.4% glucose and 0.68 mM Fe(NO 3 ) 3 , or GC broth with same supplements and 0.043% NaHCO 3 was used. BHI medium (37 g/liter brain heart infusion) with 1.25% fetal bovine serum was used when kanamycin selection was required. Antibiotics concentrations (in g/ml) used for E. coli strains were ampicillin, 100, kanamycin, 50, spectinomycin, 100, and erythromycin, 300; and for N. meningitidis were kanamycin, 80, spectinomycin, 60, erythromycin, 3, and tetracycline, 5. E. coli strain DH5␣ cultured on Luria Bertani (LB) medium was used for cloning and propagation of plasmids. Meningococci were transformed by the procedure of Janik et al. (18). E. coli strains were transformed by electroporation with GenePulser (Bio-Rad) according to the manufacturer's protocol.

Construction of Meningococcal Nonpolar Mutants
kpsF-An internal 793-bp fragment of kpsF (NMB0352) was produced by PCR amplification using primers YT60 and YT61 and cloned into pCR2.1 to yield pYT203. A SmaI-digested aphA-3 cassette (19) was subsequently inserted into the unique AscI site (blunted with Klenow) to generate pYT206. ScaI-linearized plasmid was used to transform  (20) was used to design primers YT84 and YT85. A 746-bp PCR product was amplified from chromosomal DNA of strain NMB using Vent DNA polymerase (New England Biolabs), phosphorylated with T4 kinase, and then cloned into HincII-SmaI-digested pUC18 to yield pYT256. The aphA-3 cassette released by SacI-HincII digestion was subsequently inserted into the unique EcoRV site within kdsB to generate pYT259. Colony PCR using KanC (21) and YT85 primers confirmed the correct insertion of the cassette. ScaI-digested pYT259 was used to transform strain NMB, and kanamycin-resistant colonies were selected at 30°C.
tal-PCR amplification using primers YT68 and YT69 and NMB chromosomal DNA as a template yielded a 715-bp internal fragment of tal, and the PCR product was subsequently cloned into pCR2.1. A unique ClaI site was used to insert the aphA-3 cassette released by EcoRI-BamHI double digestion and blunted with Klenow. The resulting plasmid, pCAS5, with correct orientation and in-frame fusion of the aphA-3 in tal was linearized with ScaI digestion and used for transforming meningococcal strain NMB to generate the CAS5 mutant.
kdtA-The construction of this mutant has been described previously. 2

Overexpression and Purification of the Meningococcal KpsF Protein
The coding sequence of KpsF was amplified with primers YT70 and YT71. The PCR product was digested with NdeI and XhoI and then cloned into pET20b(ϩ) which had been digested with the same enzymes, yielding pYT225. The plasmid was purified and transformed into E. coli expression strain, BLR21(DE3)pLysS. One liter of LB culture of the KpsF overexpression strain was induced with 1 mM IPTG. The harvested cells were resuspended in 15 ml of lysis buffer (50 mM sodium phosphate, pH 8.0; 300 mM NaCl; 10 mM imidazole; 1 mM phenylmethylsulfonyl fluoride) and sonicated 10 times for 30 s with 30-s cooling intervals. The cell debris was removed by centrifugation at 14,000 ϫ g for 15 min. The crude extract was then incubated with 2 ml of 50% suspension of nickel-nitrilotriacetic acid resin (Qiagen) for 2 h before packing into a column. The column was washed with 10 ml each of 20 and 50 mM imidazole in lysis buffer and then eluted with 10 ml of 250 mM imidazole. The fractions were pooled after SDS-PAGE analysis, concentrated through a Centricon-3 filter (Amicon), and dialyzed in storage buffer (50 mM HEPES, pH 7.5; 100 mM NaCl; 5 mM MgCl 2 ; 1 mM EDTA). Protein concentration was determined by the Bradford assay (Bio-Rad) with bovine serum albumin as standards.

Complementation of the NMB206 Mutant by a Second Copy of Meningococcal kpsF
A plasmid containing an intact copy of kpsF controlled by the lac promoter and ermC gene inserted downstream of kpsF was constructed. Full-length kpsF was amplified with primers CAS1 and CAS2. The PCR product was digested with HindIII and BamHI and cloned into pEGFP cut with the same enzymes. An erythromycin-resistant cassette (ermC) obtained from pAErmCЈG (22) with EcoRI digestion was subsequently cloned downstream of the kpsF gene using SmaI-EcoRI sites. This kpsF/ermC construct was amplified by PCR with primers CAS4 and YT59 and then cloned into the unique HincII site within a 1-kb chromosomal sequence of meningococcal 120A1 locus, which is located about 85 kb from the kpsF locus, in pYT109 (23). Transformation and homologous recombination of the flanking sequences of the 120A1 locus introduce the kpsF/ermC fragment into this site and resulted in erm R transformants. Subsequently, a PCR product (primers YT69 and CAS3) encompassing the sequence flanking the entire kpsF coding sequence with the aphA-3 insertion within kpsF was used to transform this strain. Erm R /Kan R transformants were then selected. A panel of PCR analyses and Southern blots with aphA-3 cassette and kpsF internal fragment (YT60-YT61) as probes confirm that allelic exchange of the aphA-3 cassette occurred at the wild type kpsF locus, and the second copy of kpsF at the 120A1 locus was intact (data not shown).

Cloning of the K1 E. coli kpsF Homologue
Primers YT77 and YT78 were used to amplify kpsF from K1 E. coli strain EV36 (24). The PCR fragment was digested with EcoRI and BglII and ligated with pFlag-CTC (Sigma), which has been cut with the same enzymes. kpsF was therefore under the control of the tac promoter and was fused to the FLAG epitope in the resulting plasmid, pYT239. A fragment, which contains the lacI repressor gene and the cloned kpsF, was released by BglI digestion and then subcloned into the EcoRV site of a meningococcal shuttle vector pYT250 to yield pYT240. The kpsFcoding sequence with in-frame FLAG fusion was confirmed with DNA sequencing analysis. The kpsF-encoding plasmid was methylated with HaeIII methylase according to the reported procedure (25) prior to transformation into meningococci. Erythromycin-resistant transformants were analyzed for the presence of E. coli kpsF by colony PCR.

LOS Extraction and Characterization
Twelve liters of meningococcal cultures were harvested, and the combined cell pellet was dried in a SpeedVac (Savant) overnight, and the dry weight was measured. The dried pellet was then extracted with phenol:chloroform:petroleum ether as described (21). The LOS samples were analyzed with 16.5% Tricine SDS-PAGE followed by silver staining (26). A micro phenol/water extraction was done as described below. A 2-ml aliquot of cultures at ϳ0.9 of A 600 reading was collected, and the bacterial pellet was resuspended in 0.5 ml of buffer A (50 mM Na 2 HPO 4 , pH 7.0; 5 mM EDTA; 0.05% NaN 3 ). A 0.5-ml aliquot of 90% liquefied phenol was added to the cell suspension and mixed by vortex. The mixture was incubated at 65°C for 15 min with vortex every 5 min and then cooled on ice for another 5 min. The aqueous phase and the phenol phase were separated by centrifugation. Both phases were dialyzed (6000 -8000 M r cut-off) against 5 changes of water and then lyophilized. An extraction method using a solution of 0.25 M EDTA and 0.25 M of TEA was adapted (27). Cells from 1.5 ml of overnight cultures were resuspended in 50 l of EDTA-TEA buffer or EDTA-TEA, 5% phenol and incubated at 60°C for 30 min. The crude LOS in the supernatant was collected after centrifugation.
A mini-scale LOS preparation was obtained by proteinase K treatment of whole cell lysates. Briefly, cells were suspended in water, and the protein concentrations were estimated by Bradford assay. A mixture of 8 l of whole cell lysate at a concentration of 1 g/l, 28 l of 2% SDS in TE buffer, and 8 l of proteinase K (25 mg/ml) was incubated at 60°C overnight. The digestion was quenched by adding 38 l of loading buffer (21). Aliquots of LOS samples were resolved on a Tricine SDS minigel, and the LOS migration pattern was visualized by silver staining (26).

Structural Analysis of LOS of the NMB206 Mutant
The contaminant phospholipids in the LOS samples extracted from the phenol:chloroform:petroleum ether method were removed by extracting the LOS with ethanol:water (9:1, v/v). The supernatant was removed, and the pellet was extracted repeatedly until no more phospholipid was found in the supernatant. The level of phospholipid was determined by the amount of C16:0, C16:1, and C18:1 fatty acids present because they are characteristic of meningococcal phospholipids (28). The resulting pellet was suspended in water and freeze-dried.
Compositional analysis was performed by the preparation and combined gas chromatographic/mass spectrometric analysis of trimethylsilylmethylglycosides with N-acetylation and of fatty acid methyl esters (29). For the determination of Kdo, lipid A was methanolyzed with methanolic 1 M HCl at 80°C for 4 h (30) prior to trimethylsilylation and GC-MS analysis. Ester-linked fatty acids were selectively liberated from a vacuum-dried sample by alkaline transesterification with sodium methoxide (0.25 M, 37°C, 15 h) (31). Combined GC-MS analysis was performed using a 50-m methylsilicone column from Quadrex Corp.
Methylation analysis was carried out using the method of Ciucanu and Kerek (32). The permethylated product was purified by using a Sep-Pak C18 cartridge (33), then hydrolyzed with 2 M trifluoroacetic acid (120°C, 2 h), reduced with NaB 2 H 4 , acetylated, and analyzed by combined GC-MS.

Mass Spectrometry of LOS
To dephosphorylate lipid A, the sample was treated with cold aqueous 48% hydrogen fluoride (HF) and kept for 48 h at 4°C. The HF was removed by flushing under a stream of air, followed by addition of diethyl ether (600 ml) and drying with a stream of air. This latter diethyl ether/drying step was repeated three times. The resulting residue was suspended in deionized water, dialyzed at 4°C for 48 h, and finally freeze-dried.
Oligosaccharides were analyzed by MALDI-TOF mass spectrometry using a Hewlett-Packard LD-TOF system. The oligosaccharides were dissolved in distilled water at a final concentration of 2 g/l, and 1 l was mixed with the dihydroxybenzoic acid in methanol matrix for analysis.
Tandem MS/MS analysis was performed using a Q-TOF hybrid mass spectrometer (Q-TOFII; Micromass, UK) equipped with an electrospray source (Z-spray) operated in the either the positive or negative mode. The samples were dissolved in 1:1 methanol and chloroform and infused into a mass spectrometer with a syringe pump (Harvard Apparatus Cambridge, MA) at a flow rate of 5 l/min. A potential of 3 kV (ϩ or Ϫ) was applied to the capillary, and nitrogen was employed as both the drying and nebulization gas. NaI and [Glu]fibrinopeptide B were used as calibration standards in the negative and positive modes, respectively. In the MS analysis the Q1 is operated in RF-only mode with all ions transmitted into the pusher region of the TOF analyzer, and the MS spectrum was recorded from m/z 400 -2000 with 1-s integration time. For MS/MS spectra, the transmission window of quadrupole (Q1) was set up to about 3 mass units, and the selected precursor ions were allowed to fragment in the hexapole collision cell. The collision energies (40 -55 eV) were optimized for maximized product ion yield, and argon was used as collision gas. The MS/MS data were integrated over a period of 4 -5 min for each precursor ion.
Protein samples for Western blots were resolved by 12% SDS-PAGE and transferred to polyvinylidene difluoride membranes. 10% bovine serum albumin in TTBS was used to block the membrane. Anti-FLAG monoclonal antibody was used at 10 g/ml in TTBS. Anti-KpsF polyclonal serum (Covance) was used at a 1:500 dilution.

Whole Cell ELISA
The previously published protocol of whole cell ELISA (12) with minor modification was employed. A 50-l aliquot of a 1:3 dilution of cell suspensions at A 650 of 0.1 was applied and dried at 37°C overnight. Antibodies specific for serogroup B (2-2-B) and serogroup A (14-1-A) were used at 1:2000 and 1:30,000 dilution, respectively. All incubations were performed at 37°C.

Determination of Ketopentoses
The procedure of Dische and Borenfreund (34), as modified by Bigham et al. (35), was used to determine the presence of ketopentoses. Briefly, enzyme was incubated at 37°C in the presence of various aldopentoses (ribose 5-phosphate, erythrose 4-phosphate, glucose 6-phosphate, and arabinose). The reaction solution was quenched with the addition of 1.5% cysteine solution followed immediately by concentrated H 2 SO 4 . An aliquot of 0.12% carbazole in 95% ethanol was then added, and the absorbance was read at 540 nm. As reported by Ray and co-workers (35), the conversion of 1 mol of arabinose 5-phosphate (Ara-5-P) to 1 mol of ribulose 5-phosphate gave an ⌬A of 8.2.

P NMR Analysis
To a solution of phosphorylated pentose (2.7 mM Ara-5-P or 5.4 mM ribulose 5-phosphate), 100 mM Bis-tris propane-HCl, pH 7.5, and 10% D 2 O (for NMR lock) in a 3-mm NMR tube was added to ϳ25 pmol of Ara-5-P isomerase. This solution was then monitored by 31 P NMR until equilibrium (no further change in peak ratio) was achieved. Samples are referenced to an external standard of neat phosphoric acid (0 ppm). Spectra were obtained on a Brucker Avance DRX-500 operating at 202.46 MHz for 31 P with WALTZ16 proton decoupling. Each spectrum represents 64 scans.

Statistical Analysis
Student's t test with a two-tailed hypothesis was used to determine the significant difference (p Յ 0.05) between two variables in this study.

RESULTS
Identification of kpsF Homologue in N. meningitidis-The first gene of region 1 of the E. coli K1 kps locus, kpsF, is absent in the meningococcal cps locus (Fig. 1). A search of the complete serogroup B (MC58) and A (Z2491) genome data bases (20,39) with the K1 KpsF protein sequence revealed a single highly homologous gene in each genome, NMB0352 and NMA2135, respectively. In contrast to K1 E. coli, the kpsF homologues in both genomes were not associated with the capsule locus. For example, in serogroups B MC58 genome, the capsule locus (NMB0067-NMB0083) is centered around 80 kb, whereas NMB0352 is located near 360 kb. A divergently transcribed gene encoding a putative transaldolase (tal) was located 96 bp upstream of the kpsF homologue, and a 221-bp intergenic space separated kpsF from a gene (NMB0353) downstream predicted to encode a conserved hypothetical protein (Fig. 1). NMB0352 (kpsF) was predicted to encode a 34-kDa protein of 324 residues and to be a cytoplasmic soluble protein by topology prediction programs, TopPredII and Psort (40,41). The NMA2135 homologue of serogroup A genome was in an identical organization to that of NMB0352 (Fig. 1) and shared 98% identity. NMB0352 is annotated as a sugar isomerase in the MC58 genome because of the presence of a sugar isomerase domain (42) between residues 38 and 172. A putative Walker A box is also located within the sugar isomerase domain. Unlike N. meningitidis, E. coli contained two additional kpsF homologues, gutQ and yrbH.
An alignment of the five predicted proteins is shown in Fig.  2. The homology was shared throughout the entire protein sequence, especially the sequence surrounding the Walker A box within the sugar isomerase domain. Both meningococcal KpsF homologues have slightly higher similarity to YrbH than to either E. coli KpsF or QutQ (67 versus 60%).
Mutation of kpsF Yielded a Defect in Capsule Expression in the Five Invasive Disease-associated Meningococcal Serogroups-Because kpsF has been proposed to participate in K1 capsule expression (43), the meningococcal kpsF homologue was mutated in the serogroup B meningococcal strain NMB. The capsule phenotype in multiple independent nonpolar kpsF::aphA-3 transformants (n ϭ 4) was assayed by a serogroup B capsule-specific whole cell ELISA. Only 10 -20% of the serogroup B capsule expressed by the parent strain was expressed by the transformants. The data for one of these transformants, designated NMB206, are shown in Fig. 3A. The reduction of capsule in the NMB206 mutant and other kpsF::aphA-3 transformants was further confirmed by colony immunoblots (Fig. 3B). The capsule-deficient phenotype was also reproduced in four independent transformants of strain NMB using a PCR product containing only the aphA-3 cassette and kpsF-flanking DNA, amplified from the NMB206 mutant using primers YT60 and YT61 (Fig. 1).
Cieslewicz and Vimr (43) reported that mutation of kpsF results in an intracellular accumulation of capsular polymers in a K1-K12 hybrid E. coli strain. To determine whether the meningococcal kpsF::aphA-3 mutant accumulated intracellular capsular polysaccharide, mutant NMB206 was lysed by freezethaw treatment or by an EDTA-HEPES method described by Moe et al. (44). An E. coli strain, EV94 (kpsS::Tn10), which has been shown to accumulate intracellular capsule (45), was included as a positive control for lysis. Capsular polysaccharide released into the supernatant was quantified by ELISA. Additional capsular polysaccharide was not detected by either lysis method when compared with the whole cell ELISA, indicating intracellular capsule polymers did not accumulate in a meningococcal kpsF mutant (data not shown). The gene for a putative transaldolase (tal) was found to be immediately upstream of kpsF and was transcribed divergently from kpsF with a 96-bp intergenic space. A nonpolar aphA-3 insertional mutation was created in tal, whose product is predicted to function in the pentose phosphate metabolic pathway (46), and the resulting mutant expressed a wild type level of capsule as measured by capsule whole cell ELISA (data not shown).
To assess whether the meningococcal KpsF homologue was required for meningococcal capsule expression of the major disease-associated serogroups, strains of serogroup A (strain F8229), C (FAM18), Y (GA0929), and W-135 (GA1002) of N. meningitidis were transformed with linearized pYT206, and the mutation within kpsF in all strains was confirmed by PCR. Capsule expression was then assessed by whole cell ELISA (serogroup A) or colony immunoblots (serogroups C, Y, and W-135) by using capsule serogroup-specific monoclonal antibodies. Reduced capsule expression was observed in strains of all serogroups (Fig. 3B). These results demonstrated that KpsF was required for the expression of either the sialic acid (serogroups B, C, Y, and W-135) or non-sialic acid (serogroup A) meningococcal capsules.

Lipo-oligosaccharide Was Truncated in the kpsF Mutant and Contained Only Unglycosylated Lipid A-
The kpsF (NMB206) mutant formed small crinkled colonies that had a dry-rough appearance. To access whether other outer membrane structures were altered in the mutant, whole cell lysates and outer membrane preparations were examined. No major alterations in proteins were observed in these preparations. However, proteinase K-treated whole cell lysates examined by Tricine SDS-PAGE followed by silver staining revealed no LOS. To understand these observations, we used four different LOS extraction methods as follows: phenol/chloroform/petroleum ether, hot phenol/water, EDTA-TEA/proteinase K, and EDTA-TEA, 5% phenol. As shown in Fig. 4, no silver-staining bands corresponding to LOS of the wild type parent were observed in material obtained from any of these extractions. As a LOS structure containing lipid A-Kdo 2 (mutant 469 (1, 47)) can be detected by silver staining, these data suggested that the LOS structure was further truncated or that no LOS was produced by the meningococcal kpsF mutant.
Fatty acid analysis of the material extracted by phenol/chloroform/petroleum ether revealed the presence of approximately equal molar amounts of dodecanoic acid (C12:0, 980 nmol/mg), 3-hydroxydodecanoic acid (3-OHC12:0, 965 nmol/mg), and 3-hydroxytetradecanoic acid (3-OHC14:0 940 nmol/mg). A small amount of palmitic acid (C16:0) was also observed which was not part of the LOS structure (see below) and, perhaps, was due to the presence of a low level of contaminating phospholipids. The same fatty acyl residues were present in the same ratio in HF-treated LOS, except that in this case a significant level of GlcN was also detected (925 nmol/mg). Assuming a normal lipid A structure which would have 2 mol of GlcN per mol of lipid A, it can be concluded that there are a total of 6 mol of fatty acid per mol of lipid A, i.e. ϳ2 mol each of C12:0, 3-OHC12:0, and 3-OHC14:0. After treatment of lipid A with sodium methoxide, C12:0 and 3-OHC12:0 were quantitatively liberated as methyl esters, showing that they had been exclusively ester-linked. The mild alkaline-treated LOS was subjected to strong alkaline hydrolysis which released only 3-OHC14:0 and proved that this was the amide-bound fatty acyl residue. Thus, composition analysis suggests that NMB206 produces a LOS with the expected lipid A for N. meningitidis. However, what proved to be very unusual was that the LOS contained no detectable glycosyl components other than the GlcN that is derived from the lipid A. In fact, none of the glycosyl residues typical of LOS from the wild type NMB or other mutants (8) was detected including the inner core sugar residues, heptose and Kdo. These results suggested that the LOS from NMB206 consisted only of unglycosylated lipid A because no glycosyl residues could be detected and because it contained the typical fatty acylation pattern for meningococcal lipid A.
LOS from NMB206 was further analyzed by mass spectrometry, MALDI-TOF MS. The results are shown in Fig. 5A. The [M Ϫ H] Ϫ ion of major intensity was m/z 1633, and those of minor intensities were m/z 1756, 1451, 1435, and 864. The m/z 1756 ion was not present in the spectrum shown in Fig. 5 but did occur in a second preparation as a minor ion together with the other ions mentioned. These different molecular ions were due to variations in phosphate, phosphoethanolamine, and fatty acyl substitution patterns. Except for m/z 864, all of the molecular species observed were consistent with the conclusion that the LOS consisted only lipid A and did not contain any detectable Kdo or core glycosyl residues. The minor ion at 864 was consistent with a mono-phosphorylated triacylglucosoamine equivalent to one-half of a lipid A molecule. Mild acid hydrolysis, which would remove glycosidically linked phos-  (lanes 1, 3, 5, 7, and 9) and the kpsF (NMB206) mutant (lanes 2, 4, 6, 8, and 10). Labeled molecular weight markers are on the left. The crude extracts from EDTA/TEA extraction before (lanes 7 and 8) and after (lanes 9 and 10) proteinase K treatment are shown for comparison. phate, did not alter the MALDI-TOF spectrum and indicated that the single phosphate group was most likely not glycosidically linked and was, therefore, located at the 4Ј-position. Much of the heterogeneity in the NMB206 LOS was removed by treatment with aqueous HF, which removes all phosphate substituents. MALDI-TOF MS analysis in the positive mode of the HF-treated LOS (Fig. 5B)  The above MALDI-TOF results showed that NMB206 produced one major LOS molecule, i.e. at m/z 1633, which consisted of N. meningiditis lipid A with only one phosphate. This LOS was completely devoid of any of the core glycosyl residues including Kdo. The results also suggested that the one phosphate group is located at the 4Ј-position and that there is no glycosidically linked phosphate. In order to confirm the location of the phosphate, the LOS was methylated, and partially methylated alditol acetates were prepared and analyzed by GC-MS. In this procedure the GlcN residues that are phosphorylated at the 4Ј-position retain the phosphate in their partially methylated alditol acetate derivative and are not observed during GC-MS analysis, whereas the reducing-end GlcN, or GlcN-1phosphate, residues of the lipid A are observed as the partially methylated alditol acetate of 6-linked GlcN (8). Because there was no detectable terminally linked GlcN, these results support the conclusion that the 4Ј-position in the LOS is phospho-rylated and, therefore, the single phosphate group on this LOS must be located at the 4Ј-position.
The structure of the LOS, after removal of the phosphate substituent, was further analyzed by tandem MS/MS analysis, The remaining fragments are due to the loss of ␤-hydroxylaurate, ␤-hydroxylauryl, or laurate from several of the primary fragments. This fragmentation pattern is completely consistent with the typical symmetrically fatty acylated lipid A reported for N. meningitidis.
From the above results, it is clear that the LOS from NMB206 consists primarily of lipid A that is not glycosylated and contains a single phosphate group at the 4Ј-position. There are minor species in which this phosphate is substituted by a phosphoethanolamine group (m/z 1756), lacks one of the fatty acyl substituents (m/z 1435 and 1451), or consists of a monophosphorylated triacylglucosamine residue (m/z 864). The structure of the major LOS from the kpsF mutant (NMB206) is shown in Fig. 6B.
Complementation of the Meningococcal kpsF Mutant-To confirm that the observed phenotypes were due to mutation of kpsF, complementation experiments were performed to introduce a second intact copy of kpsF into the meningococcal NMB206 (kpsF) mutant. To overcome decreased competence of the NMB206 mutant, we transformed strain NMB with the plasmid carrying a second copy of kpsF prior to the inactivation of the wild type gene with the aphA-3 insertion (see "Experimental Procedures"). Only transformants in which PCR and Southern blots confirmed that aphA-3 insertion occurred in the chromosomal copy but not the plasmid-encoded copies of kpsF were selected for further studies. When capsule expression was examined by whole cell ELISA and colony immunoblots, the complemented strain was identical to the wild type (data not shown), indicating mutation of the wild type copy of kpsF was complemented by the presence of the second copy of kpsF.
The meningococcal KpsF protein shares 64% similarity to the E. coli K1 KpsF protein. To determine whether these two proteins were homologues, we first constructed a strain (NMB240) carrying pYT240, an Erm R shuttle vector containing the E. coli K1 kpsF fused to a FLAG epitope and under the control of the tac promoter. We then disrupted the meningococcal kpsF using a PCR fragment (primers YT60 and YT61) amplified from the NMB206 mutant that contained the aphA-3 cassette within the kpsF coding sequence. Erm R and Kan R transformants (NMB240/206) were identified that contained the insertion of aphA-3 cassette into meningococcal kpsF and an intact copy of E. coli K1 kpsF on the shuttle vector. A strain (NMB250) containing the shuttle vector without the E. coli kpsF insert (pYT250) was used to generate a negative control strain (NMB250/206) for these experiments.
Whole cell lysates of strains NMB206, NMB240, NMB240/ 206, NMB250, and NMB250/206 were analyzed by Western blots probed with an anti-FLAG monoclonal antibody (Fig. 7A) and anti-KpsF polyclonal serum (Fig. 7B). K1 KpsF-FLAG proteins were expressed at a similar level in both the NMB240 and NMB240/206 strains without induction, indicating an incomplete suppression by lacI. Increased expression was ob-served in the presence of IPTG (Fig. 7A). The absence of reactive band in strains containing kpsF::aphA-3 mutations, when probed with antiserum against meningococcal KpsF, confirmed that these mutations eliminated expression of meningococcal KpsF (Fig. 7B). The colony morphology and growth rate of the NMB240/206 strain was similar to the wild type parent, whereas the NMB250/206 strain resembled the NMB206 strain. When these strains were examined by the capsulespecific whole cell ELISA, the NMB240/206 complemented strain yielded wild type level of capsule expression, whereas the NMB250/206 negative control produced a capsule-deficient phenotype similar to that of strain NMB206 (Fig. 7D). These data demonstrated that E. coli K1 KpsF could functionally replace the meningococcal KpsF and complement the capsuledeficient phenotype of the NMB206 mutant. LOS from these strains was also examined by silver-stained Tricine SDS-PAGE, and wild type LOS bands were restored in the NMB240/ 206 strain but not the NMB250/206 strain, indicating complementation of meningococcal kpsF mutation by E. coli K1 KpsF (Fig. 7C).
Kdo Biosynthesis Is Involved in Meningococcal Capsule Expression-LOS is the major component of the outer leaflet of the outer membrane, and capsule is anchored on the outer membrane via a diacylglycerol moiety (7). Possible structural changes of the outer membrane produced by the markedly truncated LOS (unglycosylated lipid A) might affect capsule expression, although capsule expression was not influenced in previous meningococcal mutants with various truncations in the outer and inner core of LOS (1) or in a lpxA mutant that does not contain LOS. 3 To address this question, a kdtA mutation was created in meningococcal strain NMB yielding strain NMB249. KdtA is the CMP-Kdo transferase that catalyzes the transfer of Kdo to lipid A. The NMB249 mutant generated the same unglycosylated lipid A structure as that of the kpsF mutant. 2 However, this mutant, when analyzed by whole cell ELISA, produced wild type levels of capsule (Fig. 8). Thus, the reduced capsule expression caused by the kpsF mutation was not due to outer membrane alterations or other pleiotropic effects resulting from a truncated LOS.
The absence of Kdo in LOS and the reduction of capsule expression in the kpsF mutant suggested a role of Kdo in capsule expression. Kdo has been identified as a component of the E. coli K5 capsule at the reducing end of the polymer (48). To assess further the role of Kdo in meningococcal capsule expression, an insertional nonpolar mutation was created in the CMP-Kdo synthetase, kdsB, in order to eliminate the production of the activated Kdo sugar. When the kdsB mutant was assayed by whole cell ELISA, a significant reduction in capsule expression, similar to that of the kpsF mutant, was observed  :aphA-3). The A 405 nm reading of the wild type parent was normalized to 100% (n Ն 3). (Fig. 8). These data showed a role for Kdo in meningococcal capsule expression and further suggested that KpsF was involved in Kdo production.
KpsF Is the Arabinose-5-phosphate Isomerase of N. meningitidis-The predicted protein sequence of KpsF suggested a sugar isomerase activity. To determine whether KpsF had an isomerase activity, a colorimetric assay system for keto-pentoses (35) was first conducted using the purified KpsF protein, which contained a C-terminal His 6 tag (Fig. 9A). Addition of KpsF protein into the reaction mixture containing arabinose 5-phosphate resulted in an increase in color, whereas adding the protein to ribulose 5-phosphate caused a decrease in reading of A 540 (data not shown). The reaction was dependent on the concentration of KpsF protein. Other monosaccharides, such as erythrose 4-phosphate, glucose 6-phosphate, ribose 5-phosphate, and arabinose, did not serve as substrates for KpsF protein (data not shown). The NMR chemical shifts of the phosphoryl groups of arabinose-5-P and ribulose-5-P are different; thus 31 P NMR can monitor the interconversion between the two sugars. As shown in Fig. 9C, using arabinose-5-P as the substrate, a new peak corresponding to the phosphoryl group of ribulose-5-P appeared and increased over time, and the same phenomenon was observed for the reverse reaction. These data demonstrated that KpsF catalyzes the interconversion of ribulose 5-phosphate and arabinose 5-phosphate. The LOS defect in the kpsF mutant, but not the kdsB mutant, can be complemented by exogenous supplement of arabinose in the growth medium (Fig. 9B), further indicating the role of KpsF protein as an arabinose-5-phosphate isomerase. DISCUSSION By using a comparative genomic approach based on the two recently published meningococcal genome data bases (39, 40), we identified meningococcal homologues of kpsF and kpsU (kdsB) of K1 E. coli. In contrast to K1 E. coli, N. meningitidis contains only a single copy of the gene encoding the CMP-Kdo synthetase, kdsB, in a location distant from the capsule locus. We found that the meningococcal kpsF and kpsU (kdsB) homologues are both required for Kdo biosynthesis. KpsF was identified as the arabinose-5-phosphate isomerase. Mutations of kpsF and kpsU (kdsB) in N. meningitidis caused a significant reduction in surface capsule expression and resulted in a LOS structure composed only of unglycosylated meningococcal lipid A.
The condensation of arabinose 5-phosphate and phosphoenolpyruvate catalyzed by the Kdo synthase, KdsA, is usually considered to be the first step in Kdo biosynthesis (Fig. 10). However, arabinose 5-phosphate is not readily available from glycolysis, and an isomerase is required for the interconversion of ribulose 5-phosphate and arabinose 5-phosphate (Fig. 10). Although the enzymatic activity has been demonstrated in cell extract, the gene encoding this enzyme has not been identified previously (35). KpsF contains a sugar isomerase domain commonly found in a variety of proteins such as GlmS and LpcA (42) involved in phosphosugar isomerization. In addition, the sugar isomerase domain is also present in the transcriptional regulator, RpiR, of the ribose-phosphate isomerase, RpiA, which interconverts ribose 5-phosphate and ribulose 5-phosphate, a reaction that precedes the arabinose-5-phosphate isomerase. Interestingly, in addition to kpsF, region 1 of the capsule locus of E. coli strains expressing group II capsules (e.g. K1 and K5) encodes a second copy of the CMP-Kdo synthetase (kpsU). In E. coli, this may be an evolutionary acquisition of genes to ensure that the Kdo substrate required for capsule biosynthesis was not limited by requirements of the LPS bio-  31 P spectra of the Ara-5-P isomerase reaction starting with either Ara-5-P (a-c) or ribulose 5-phosphate (d-f) as substrates. Spectra a-c were taken at t ϭ 0, t ϭ 54 min, and t ϭ 250 min, respectively, with only Ara-5-P (␦ ϭ 4.9 ppm) present at t ϭ 0 min (a). Spectra f, e, and d were taken at t ϭ 0, t ϭ 180 min, and t ϭ 600 min, respectively, with only ribulose 5-phosphate (␦ ϭ 5.3 ppm) present at t ϭ 0 min (f). Note that spectra c and d represent equilibrium approached from either substrate and that the equilibrium ratio of Ara-5-P to ribulose 5-phosphate (65:35) is the same in either case. synthesis pathways. In fact in E. coli two other predicted proteins, GutQ and YrbH, are homologues of KpsF. GutQ is located within the glucitol operon, but its function has not been determined previously (49).
Although kpsF is conserved (98% amino acid identity) in the kps locus of E. coli strains expressing K1 and K5 capsular polysaccharides (50,51), the role of KpsF in capsule expression of these bacteria has not been firmly established. The expression of region 1 in E. coli is regulated by temperature at the transcriptional level; however, KpsF is not required for thermoregulation (50). A nonpolar kpsF mutation resulted in about 10-fold reduction of capsule translocation to the surface of K1 E. coli (43). Although Kdo is a component of K5 capsules at the reducing end of the polymers (48), KpsF is not required for K5 capsule expression because cloning the K5 capsule gene cluster lacking kpsF into a K12 strain produced a capsule comparable with that of the wild type K5 strain (52). The lack of clear phenotypes in E. coli kpsF mutants is likely due to the presence of gutQ and/or yrbH. The meningococcal genomes, however, have no other homologues of kpsF. We found that E. coli K1 kpsF complements the meningococcal kpsF mutation in both LOS biosynthesis and capsule expression. This suggests that E. coli K1 and K5 and meningococcal kpsF homologues perform analogous functions.
Our data indicate that Kdo is involved in meningococcal capsule expression. We found that inactivation of the arabinose-5-phosphate isomerase, KpsF, reduces expression of both sialic acid (serogroup B, C, Y, and W-135) and non-sialic acid (serogroup A) meningococcal capsules. Thus, Kdo is not likely to be involved in the activation of the different monomeric substrates (e.g. sialic acid and mannosamine) but may function in capsule assembly, such as an acceptor involved in the initiation of capsule polymerization.
Gotschlich et al. (7) characterized the structures of meningococcal capsular polysaccharides from serogroups A, B, and C. The authors reported a common diacylglycerol substitution at the reducing end of capsule polymers, but no Kdo residues were observed. Although K1 and K5 E. coli have an identical organization of region 1 of the kps locus (Fig. 1), Kdo is found as the reducing sugar in the E. coli K5 capsule (48) but thus far not in K1 polymers. Finke et al. (48) hypothesized that the biosynthesis of K5 capsule is initiated by substitution of an undecaprenol phosphate (UP) carrier with Kdo, which then acts as an acceptor for subsequent capsule polymerization. In the case of K1 E. coli, however, UP alone was proposed to be the acceptor in capsule polymer assembly (53). Our data suggest that Kdo-UP may act as a carrier for capsule polymer assembly in meningococci and may be subsequently replaced by the phospholipid substitution, thus removing Kdo from the final assembled capsule polymers. In both meningococcal kpsF and kdsB mutants, capsule expression was significantly reduced but not completely eliminated, suggesting an alternative acceptor other than Kdo-UP.
Re endotoxin, a Kdo 2 -lipid A structure, has been widely regarded as the minimal endotoxin structure that supports viability (54). However, meningococci were shown recently to be viable without any endotoxin (55). We found that a viable meningococcal kpsF mutant can produce fully acylated lipid A without Kdo glycosylation (Fig. 6). The kpsF mutant expresses primarily 4Ј-phosphorylated hexa-acyl lipid A, with minor components consisted of penta-acyl lipid A and triacylglucosamine. Interestingly, the glycosidic position of lipid A was unphosphorylated in the kpsF mutant. A negative charge provided by phosphorylation at this position has been proposed to be essential (10) or when absent is provided by the reducing sugar of the lipid A backbone, ␣-D-galacturonic acid, as in Rhizobium leguminosarum (31) or in Rhodospirillum fulvum, a galacturonic acid replaces the phosphate group at the glycosidic position (56). The absence of phosphorylation at the glycosidic position was also detected in the unglycosylated lipid A species expressed by the kdtA mutant. 2 These observations suggest that phosphorylation of the glycosidic position is optional when Kdo glycosylation is absent. The data also suggest that N. meningitidis can express a lipid A glycosidic phosphatase.
In summary, we have identified that kpsF in meningococci and its homologues in E. coli encode the arabinose-5-phosphate isomerases of prokaryotes. The unique phenotype of the kpsF mutant should allow us to understand better the genetic basis and biosynthesis of meningococcal LOS and capsule, two critical virulence factors in meningococcal pathogenesis.