A Haemophilus influenzae Gene That Encodes a Membrane Bound 3-Deoxy-d-manno-octulosonic Acid (Kdo) Kinase

The lipopolysaccharide of Haemophilus influenzae contains a single 3-deoxy-d-manno-octulosonic acid (Kdo) residue derivatized with either a phosphate or an ethanolamine pyrophosphate moiety at the 4-OH position. In previous studies, we identified a kinase unique to H. influenzae extracts that phosphorylates Kdo-lipid IVA, a key precursor of lipopolysaccharide in this organism. We have now identified the gene encoding the Kdo kinase by using an expression cloning approach. A cosmid library containing random DNA fragments from H. influenzae strain Rd was constructed in Escherichia coli. Extracts of 472 colonies containing individual hybrid cosmids were assayed for Kdo kinase activity. A single hybrid cosmid directing expression of the kinase was found. The kinase gene was identified by activity assays, sub-cloning, and DNA sequencing. When the putative kinase gene was expressed inE. coli behind a T7 promoter, massive overproduction of kinase activity was achieved (∼8000-fold higher than in H. influenzae membranes). The catalytic properties and the product generated by the overexpressed kinase, assayed with Kdo-lipid IVA as the substrate, were the same as observed withH. influenzae membranes. Unexpectedly, the kinase gene was identical to a previously characterized open reading frame (orfZ), which had been shown to be important for establishing bacteremia in an infant rat model (Hood, D. W., Deadman, M. E., Allen, T., Masoud, H., Martin, A., Brisson, J. R., Fleischmann, R., Venter, J. C., Richards, J. C., and Moxon, E. R. (1996) Mol. Microbiol.22, 951–965). However, based solely on the genome sequence of H. influenzae Rd, no biochemical function had been assigned to the product of orfZ, which we now designate kdkA(“Kdo kinase A”). Although Kdo phosphorylation may be critical for bacterial virulence of H. influenzae, it does not appear to be required for growth.

The Gram-negative pathogen Haemophilus influenza is a common cause of otitis media, upper respiratory infections, and meningitis in children (1)(2)(3)(4). Like other Gram-negative bacte-ria (5,6), the outer surface of the H. influenzae outer membrane consists predominantly of lipopolysaccharide (LPS) 1 (7). LPS provides the organism with a permeability barrier to certain antibacterial agents (5,6). LPS is anchored into the outer leaflet of the outer membrane by its lipid A moiety (Fig. 1). Lipid A is an acylated disaccharide of glucosamine that is usually phosphorylated at positions 1 and 4Ј and triggers many of the inflammatory responses associated with infections (6,8,9). The enzymatic steps of lipid A biosynthesis have been fully delineated in Escherichia coli (6,10). Most of the relevant structural genes are required for viability (6,9). The current surge in bacterial genome sequencing projects has made it possible to identify homologues of the E. coli lipid A genes (6,9) in H. influenzae (11) and other pathogens (12)(13)(14).
In E. coli and Salmonella typhimurium, lipid A is glycosylated with a non-repeating oligosaccharide known as the core (6,15), which begins with the unusual sugar 3-deoxy-D-mannooctulosonic acid (Kdo) (Fig. 1). Many strains are further glycosylated with a distal repeating oligosaccharide, known as the O-antigen (not shown) (6,15,16). H. influenzae lacks the Oantigen but contains a more extensively branched core oligosaccharide than E. coli (7,17). Portions of the H. influenzae core are highly variable from strain to strain (17). The phenomenon of phase variation provides a mechanism for core alteration within a given strain (18 -20). The variability in core structures may provide a means for the bacteria to evade host immune responses (18 -20). Recent studies have established an important role for proper core biosynthesis in the virulence of H. influenzae. Hood et al. (7) prepared a series of mutants with insertions in the genes postulated to be involved in core biosynthesis. The virulence of the mutant strains was then tested by their ability to cause bacteremia in an infant rat model (7). A correlation was found between the extent of core glycosylation and virulence (7). For instance, mutants lacking heptose, which have the minimal LPS structure capable of supporting growth (Fig. 1), did not cause significant bacteremia (7).
The first step of core biosynthesis in all Gram-negative bacteria is the addition of Kdo to the 6-OH of the precursor, lipid IV A (Fig. 1) (6,10). In both E. coli and H. influenzae, the transfer of Kdo to lipid IV A is essential for viability (7,21). The E. coli Kdo transferase, encoded by the kdtA gene (22), is an unusual bi-functional enzyme ( Fig. 1) that catalyzes the sequential addition of two Kdo residues in distinct glycosidic linkages to lipid IV A (23). Most other Gram-negative bacteria similarly contain at least two Kdo residues in their inner core (24). In H. influenzae, however, only a single Kdo is present * This work was supported by National Institutes of Health Grants GM-51310 (to C. R. H. R.) and GM-54882 (to R. J. C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) HI0260. 1.
By using a non-typeable strain of H. influenzae, we previously demonstrated that the H. influenzae Kdo transferase is mono-functional (28), i.e. capable of adding only a single Kdo residue to lipid IV A . In addition, we provided the first evidence for the presence of a Kdo kinase unique to extracts of H. influenzae (28) (Fig. 1). A homologue encoding a protein with 70% predicted similarity to E. coli KdtA was readily apparent by inspection of the H. influenzae genome (11). Analysis of the reaction product generated by the overexpressed recombinant H. influenzae KdtA confirmed this genomic assignment and the mono-functional activity of the protein. 2 However, since no protein sequence was available for the Kdo kinase, the genome sequence alone was insufficient to permit identification of the kinase gene.
We now report the expression cloning and biochemical char-acterization of the Kdo kinase structural gene of H. influenzae. A cosmid library containing DNA fragments of the H. influenzae strain used for the genome project (Rd) (11) was constructed in E. coli. Lysates of single colonies harboring individual hybrid cosmids of the library were assayed for the presence of Kdo kinase activity, which is absent in E. coli (28) (Fig. 1). A single cosmid that directs the expression of the kinase was found. Interestingly, the gene encoding the kinase had previously been described as a possible open reading frame of unknown function, termed orfZ (7), in H. influenzae. Although OrfZ had been shown to be essential for virulence, no biochemical function could be assigned to the protein based solely upon its sequence and the apparently normal electrophoretic properties of the LPS isolated from orfZ mutants (7). In light of our discovery that orfZ encodes the Kdo kinase, the genetic and pathogenic studies by Hood et al. (7) can now be re-interpreted to suggest that the absence of Kdo phosphorylation in H. influenzae dramatically reduces virulence but does not stop bacterial growth. Given the identification of the biochemical function of orfZ, we suggest that the gene now be designated kdkA (for 2 K. A. White and C. R. H. Raetz, manuscript in preparation. The phosphate residue (red) incorporated by the Kdo kinase (KdkA) is proposed to be located at the Kdo 4-position, based on the most recent structural characterization of intact H. influenzae LPS (17). The structures of the mature lipid A (black) and Kdo (blue) domains of E. coli and H. influenzae lipopolysaccharides are shown at the right. These LPS substructures are the predominant species present in L-glycero-D-manno-heptosedeficient mutants of E. coli or H. influenzae (6,26,52). In the case of H. influenzae, an alternative molecular species (not shown) containing a phosphate residue at the 5-rather than the 4-position of Kdo was proposed in earlier studies by some authors (17,26). However, in wild type LPS of both E. coli and H. influenzae, the first heptose residue of the core (not shown) is attached to the 5-OH position of the inner Kdo (6,8,17). In E. coli the substituent X denotes partial substitution with a phosphate residue (53,54), whereas in H. influenzae Y denotes partial substitution with an ethanolamine phosphate moiety (7,17). The LPS substructures found in heptose-deficient mutants (above right) are sufficient to support bacterial growth, but such mutants are not virulent (7). "Kdo kinase A") ( Fig. 1). Consistent with the occurrence of phosphorylated Kdo residues in some Gram-negative bacteria (29 -32), significant homologues of kdkA are present in Bordetella pertussis, Vibrio cholerae, Actinobacillus actinomycetemcomitans, Shewanella putrefaciens, and Pateurella multocida.

EXPERIMENTAL PROCEDURES
Materials-[␥-32 P]ATP was purchased from NEN Life Science Products. Kdo, HEPES, EDTA, EGTA, NAD ϩ , heme, CTP, ATP, and other nucleotides were purchased from Sigma. Triton X-100 was Surfact-Amps grade from Pierce. Yeast extract, tryptone, and brain heart infusion were obtained from Difco. All other chemicals and solvents were reagent grade. DEAE-cellulose (DE52) was purchased from Whatman. The 0.25-mm glass backed Silica Gel 60 thin layer chromatography plates were from Merck.
Bacterial Strains and Growth Conditions-The various strains and plasmids utilized for the experiments described are detailed in Table I. H. influenzae strain Rd (catalog number 51907) was purchased from ATCC. The H. influenzae cells were grown at 37°C in brain heart infusion medium (37 g/liter) supplemented with heme (10 g/ml) and NAD ϩ (10 g/ml) (33). E. coli strains were grown at 37°C on Luria broth, consisting of 10 g of NaCl, 10 g of tryptone, and 5 g of yeast extract per liter (34). When applicable, the cultures were supplemented with 50 g/ml ampicillin and/or 10 g/ml chloramphenicol.
Preparation and Isolation of Substrates-Milligram quantities of the precursor lipid IV A were prepared as described previously (35). Prior to use in assays and the synthesis of Kdo-lipid IV A , the lipid was subjected to reverse phase chromatography (36). Unlabeled Kdo-lipid IV A was prepared by a previously described method (28). Both unlabeled and Kdo 2 [4Ј-32 P]lipid IV A were prepared by the published methods (37,38). The [4Ј-32 P]lipid IV A was prepared by the method of Brozek et al. (39), using membranes isolated from the E. coli strain pJK2/BLR(DE3), which overexpresses the 4Ј-kinase (40). The Kdo[4Ј-32 P]lipid IV A was synthesized by a slight modification of the published method (28). Briefly, [4Ј-32 P]lipid IV A was prepared as usual (38), but at the end of the reaction, the volume was adjusted to 180 l with water. The solution was then converted to a two-phase Bligh/Dyer system, consisting of CHCl 3 /methanol/H 2 O (2:2:1.8, v/v) (41,42), by the addition of 200 l of both CHCl 3 and methanol. The tube was mixed thoroughly and centrifuged at 20,800 ϫ g for 5 min at room temperature to separate the phases. The chloroform-rich lower phase was removed and transferred to a fresh tube. The upper phase was then washed twice with preequilibrated acidic lower phase, i.e. a lower phase generated by mixing chloroform/methanol/0.1 M HCl (2:2:1.8, v/v). The resulting lower phases were pooled with the initial lower phase and dried under a stream of nitrogen. The reaction components for the mono-functional Kdo transferase reaction (28) were then added to the tube. Following the Kdo transferase reaction, the Kdo[4Ј-32 P]lipid IV A was isolated by preparative thin layer chromatography as described (28). All lipids were stored as aqueous dispersions at Ϫ20°C and were dispersed again after thawing by sonic irradiation in a bath for 30 -60 s prior to use. Recombinant E. coli CMP-Kdo synthetase (43) was partially purified as described by Brozek et al. (37).
Recombinant DNA Techniques-H. influenzae Rd genomic DNA was prepared as described previously (33). Plasmid DNA was isolated using the Qiagen Mini-Prep purification system (Qiagen). Restriction endo-nucleases (New England Biolabs), T4 DNA ligase (Life Technologies, Inc.), and shrimp alkaline phosphatase (U. S. Biochemical Corp.) were used according to the manufacturers' instructions. DNA sequencing was performed at the Duke University Medical Center shared DNA sequencing facility.
Kdo Transferase Assay-Kdo transfer from the donor, CMP-Kdo, to the acceptor, [4Ј-32 P]lipid IV A , was assayed as described previously (28). The reaction mixtures (typically 10 -20 l) contained 50 mM HEPES, pH 7.5, 2 mM Kdo, 0.1% Triton X-100, 100 M [4Ј-32 P]lipid IV A (3000 -6000 cpm/nmol), 5 mM CTP, 10 mM MgCl 2 , and 1.8 milliunits of partially purified, recombinant CMP-Kdo synthase. Assays (at 30°C) were initiated by the addition of enzyme, usually H. influenzae membrane preparations, as indicated. The reactions were terminated by spotting 5-l portions onto a thin layer plate. The plate was dried under a cool air stream and developed in the solvent chloroform/pyridine/88% formic acid/water (30:70:16:10, v/v). The solvent was evaporated with a hot air stream, and the plate was exposed to a PhosphorImager screen for 12-16 h. The conversion of 32 P-labeled substrate to product was quantified using a Molecular Dynamics PhosphorImager equipped with ImageQuant software.
Kdo Kinase Assay-The Kdo kinase was assayed using the acceptor Kdo[4Ј-32 P]lipid IV A , as described previously (28). The conditions were very similar to those used for the Kdo transferase. Briefly, reaction mixtures (10 -20 l) contained 50 mM HEPES, pH 7.5, 0.1% Triton X-100, 10 mM MgCl 2 , 5 mM ATP, and 100 M Kdo[4Ј-32 P]lipid IV A (3000 -6000 cpm/nmol). The reactions were initiated with enzyme and incubated for designated times at 30°C. The assays were terminated, and the substrate and product were resolved by thin layer chromatography, as described above for the Kdo transferase. When membranes from pKdkA/BLR(DE3)/pLysS were used in the assays, 1 mg/ml bovine serum albumin was included in the reaction mixture. Other minor modifications to the standard reaction conditions are noted in the figure legends.
Construction of H. influenzae Rd Genomic Library-A cosmid library of H. influenzae strain Rd (11) was constructed in E. coli XL1-Blue MR (Strategene), utilizing the Gigapack III XL packaging system (Strategene). Briefly, 100 g of genomic DNA was partially digested with Sau3A1 (New England Biolabs) until the predominant DNA fragments were approximately 20 kb in size, as judged by agarose gel electrophoresis. The fragments were then ligated into the cosmid pWE15 (Stratagene). Prior to ligation, pWE15 was digested with BamHI, gel-purified (Qiagen Gel Extraction Kit), and treated with phosphatase, according to standard procedures (44). The ligation mixtures were then packaged into recombinant phage using the Gigapack III XL packaging system. To determine the colony-forming units per l, the packaging extracts were titered using strain XL1 Blue-MR. The library was subsequently amplified in XL1 Blue-MR, and aliquots were frozen as glycerol stocks at Ϫ80°C.
Preparation of Cosmid Library Lysates and Initial Screening for Kdo Kinase Expressing Clones-Individual colonies from the library were grown in microtiter plates, and lysates were prepared by the method of Dotson et al. (45). A portion of the amplified library was thawed and diluted appropriately (1:1 ϫ 10 6 ). Then, 100-l portions of the diluted library were plated onto LB agar plates containing 50 g/ml ampicillin. The plates were incubated overnight at 37°C. Single colonies were picked from these plates and used directly to inoculate six 96-well Intergenic region between orfM and opsX cloned into pBluescript, kdkA ϩ , Amp r This work pET21a Vector containing a T7 promoter, Amp r Novagen pKdkA pET21a containing Kdo kinase (orfZ) coding region, Amp r This work microtiter dishes (containing 150 l of LB medium with ampicillin per well). The dishes were placed in an air shaker and incubated overnight at 37°C. The overnight dishes were used to inoculate fresh microtiter plates (containing 200 l of LB medium with ampicillin per well) using a sterile 96-prong apparatus (Nalge Nunc International). Sterile glycerol (60% v/v) was added to the overnight plates (to achieve a final concentration of 20%). After mixing, the plates were frozen at Ϫ80°C for later use. Meanwhile, the freshly inoculated plates were grown for 6 h at 37°C with rotary shaking at 200 rpm. The cultures were then centrifuged at 3,600 ϫ g for 20 min at 4°C, and the supernatants were decanted. The microtiter plates were placed on ice, and the cell pellets were resuspended in 25 l of 33 mM Tris-HCl, pH 8.0. Following resuspension, 25 l of 33 mM Tris-HCl, pH 8.0, containing 0.2 mg/ml lysozyme and 5 mM EDTA was added to each well. The plates were incubated for 5 min on ice and then frozen at Ϫ80°C. Just prior to assay, the cells were lysed by thawing the plates at room temperature for 10 min. The lysates of the cells containing the individual cosmids were assayed by the Kdo kinase assay of White et al. (28) with slight modification. A 12-l portion of each lysate was preincubated with 100 M lipid IV A (final volume of 14 l) in a microtiter plate for 10 min at room temperature. This preincubation helped to prevent the interference of other reactions, such as the E. coli Kdo transferase, in the subsequent step. At the end of the preincubation, 7 l of the lysate/lipid IV A mixture was used to initiate a Kdo kinase assay (final volume 10 l) in a fresh microtiter plate. The kinase reaction mixture contained 50 mM HEPES, pH 7.5, 0.1% Triton X-100, 10 M Kdo[4Ј-32 P]lipid IV A (ϳ10,000 cpm/ nmol), 5 mM ATP, 10 mM MgCl 2 , and 70 M lipid IV A (resulting from the dilution of the preincubation mixture). The microtiter plate containing the radioactive reactions was incubated for 30 min at 30°C. The reactions were terminated by spotting 4 l from each well onto a silica gel TLC plate. The plates were air-dried and then developed in the solvent chloroform/pyridine/88% formic acid/water (30:70:16:10, v/v). The plates were dried and exposed to a PhosphorImager screen for 12-16 h. The percent conversion of the radiolabeled substrate to product was calculated using a Molecular Dynamics PhosphorImager equipped with ImageQuant software.
Preparation of Cell-free Extracts and Membranes for Quantitative Enzyme Assays-Typically, 1-liter cultures of either H. influenzae or E. coli were grown to late log phase (A 600 ϳ1.0 -1.5) and then harvested by centrifugation at 1,900 ϫ g for 10 min at 4°C. In the case of BLR(DE3)/ pLysS cells expressing the Kdo kinase behind the T7 promoter (pKdkA), the cells were grown to A 600 ϳ0.6 -0.8, induced with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h and harvested as described above. All cell pellets were washed with ϳ150 ml of 30 mM HEPES, pH 7.5, containing 2.5 mM EDTA and 1.0 mM EGTA. The washed pellet was resuspended in buffer (ϳ20 ml of buffer per liter of culture), consisting of 30 mM HEPES, pH 7.5, 1.0 mM EDTA, and 1.0 mM EGTA. The cells were ruptured by passage through an ice-cold French pressure cell (SLM Instruments, Urbana, IL) at 18,000 p.s.i. The unbroken cells and large debris were removed by centrifugation at 1,900 ϫ g for 15 min at 4°C. The supernatant (designated the cell-free extract) was stored in aliquots at Ϫ80°C.
The cell-free extract was separated into membrane and cytosolic fractions by ultracentrifugation at 150,000 ϫ g for 60 min at 4°C. The supernatant (cytosol) was centrifuged again to remove any remaining membrane fragments. The membrane pellet was resuspended in ϳ10 -20 ml of 30 mM HEPES, pH 7.5, 1.0 mM EDTA, and 1.0 mM EGTA and was centrifuged again to yield a washed membrane fraction. Like the cell-free extracts, the cytosol and the washed membranes were stored in aliquots at Ϫ80°C (10 -20 mg/ml).
Protein concentrations were determined using the bicinchoninic assay (Pierce) with bovine serum albumin as the standard (46).
Subcloning and Identification of the Gene Encoding the Kdo Kinase-Cosmid DNA was isolated from clone B67E (corresponding to box 6, well 7E), which directs the overexpression of the Kdo kinase in E. coli, using the Qiagen mini-prep system. The isolated cosmid DNA was subjected to digestion with EcoRI, resulting in the release of the H. influenzae genomic DNA (ϳ10-kb insert) from the vector pWE15. A total of five EcoRI fragments was produced (ϳ4.5, 2, 1.8, and 0.95 kb), including a fragment attributable to the cosmid (8.2 kb). The 4.5-kb fragment was resolved by gel electrophoresis and purified (44). The fragment was ligated into pBluescript IIKS, and the ligation mixture was used to transform Sure cells (Stratagene) (44). Plasmid-containing cells were selected by growth at 37°C on LB agar supplemented with ampicillin (50 g/ml). The presence of Kdo kinase activity in extracts of these cells was confirmed, using the assay described above. The pBluescript KSII containing the ϳ4.5-kb insert of H. influenzae DNA was designated pE3.2.
An additional subclone was constructed by digesting pE3.2 with BamHI. The BamHI cut once in the pBluescript vector and once within the insert, deleting a piece of ϳ1.2 kb. The digested vector was religated, transformed into Sure cells (Stratagene), and selected as described above (44). This smaller subclone, designated pB6A, also encoded the Kdo kinase, as judged by the presence of Kdo kinase activity in cell lysates.
Finally, pIB100, a pBluescript plasmid bearing only the putative Kdo kinase gene, was constructed. The region, previously designated orfZ or HI0260.1 and located between opsX (HI0261) and orfM (HI0260) (7,11), was amplified by the polymerase chain reaction (PCR). Primers for the reaction were designed based on the H. influenzae DNA data base (11). The primers were as follows: the forward primer, 5Ј GCG GCG AAG CTT CTG GGC TTT CAA TCG 3Ј, and the reverse primer, 5Ј GCA TCG GAT CCG CTA AGG CAT GAC AG 3Ј. The forward primer inserted an HindIII restriction site (shown in bold) approximately 100 base pairs from the start codon on of the putative Kdo kinase (orfZ) gene, and the reverse primer introduced a BamHI restriction site (shown in bold) ϳ130 base pairs downstream of the stop codon. The PCR reaction mixture contained 10 ng of pE3.2 template, 0.2 g of each primer, 20 mM Tris-HCl, pH 8.8, 10 mM KCl, 10 mM (NH 4 ) 2 SO 4 , 0.1% Triton X-100, 100 g/ml nuclease-free bovine serum albumin, 200 M each dNTP, 2 mM MgCl 2 , and 2.5 units of Pfu DNA polymerase (Stratagene) in a final volume of 0.05 ml. The mixture was subjected to 5 min of denaturation at 94°C and then 25 cycles of 94°C for 1 min, 50°C for 1.5 min, 72°C for 1.5 min, and ended with a 7-min extension at 72°C in a Perkin-Elmer GeneAmp PCR system 2400. The PCR product was digested with HindIII and BamHI, and ligated into pBluescript IIKS, which had been digested similarly. A portion of the ligation mixture was transformed into Sure cells (Stratagene). Plasmid containing colonies were selected as described above. The insert in pIB100 was confirmed by DNA sequencing.
Construction of Plasmid (pKdkA) with the orfZ/kdkA Gene Behind the T7 Promoter-To overexpress the Kdo kinase to high levels, the orfZ/kdkA gene was cloned into pET21a (Novagen), under the control of the T7 promoter. The gene was amplified by PCR using pE3.2 as the template. The primers were as follows: the forward primer, 5Ј GCG CGC CAT ATG CAC CAA TTC C 3Ј, and the reverse primer, 5Ј GCG CGG ATC CGC TTT TAT TGA TG 3Ј. The forward primer introduces a NdeI restriction site (shown in bold) at the start codon of orfZ/kdkA, and the reverse primer creates a BamHI site downstream of the stop codon of the gene. The PCR was done under the conditions described above for the construction of pIB100. The PCR product was digested with NdeI and BamHI and ligated into pET21a cut with the same enzymes. The ligation mixtures were used to transform Sure cells, as described above. The presence of the appropriate insert was confirmed by restriction digestion of the plasmid with NdeI and BamHI. The desired plasmid was designated pKdkA. The insert in pKdKA was confirmed by DNA sequencing. For overexpression of active KdkA, pKdkA was transformed into BLR(DE3)/pLysS (Novagen).
Large Scale Preparation of Phospho-Kdo-Lipid IV A -The Kdo kinase reaction was optimized so that the substrate, Kdo-lipid IV A , was converted completely to the product, phospho-Kdo-lipid IV A . The high yield simplified the isolation of the phospho-Kdo-lipid IV A for further structural analysis. Briefly, two large scale reaction mixtures (10 ml each) were prepared in 16 ϫ 125 mm borosilicate tubes. The reaction mixtures contained 50 mM HEPES, pH 7.5, 0.2% Triton X-100, 5 mM ATP, 10 mM MgCl 2 , and 100 M Kdo-lipid IV A . Each reaction was initiated by addition of 10 g of washed membranes prepared from pKdkA/ BLR(DE3)/pLysS. The mixtures were incubated at 30°C for 15 min and then transferred to a 150-ml Corex bottle. The aqueous solution was converted to a single phase Bligh/Dyer system by the addition of 1.24 ml of chloroform, 2.5 ml of methanol, and 0.04 ml of concentrated HCl per ml of reaction mixture. The sample was thoroughly mixed and centrifuged at 3,000 ϫ g for 20 min at room temperature to remove precipitated proteins. The supernatant was divided between two fresh Corex bottles and converted to a two-phase Bligh/Dyer system by adding 0.263 ml of CHCl 3 and 0.263 ml of H 2 O per ml of supernatant. After mixing, the phases were separated by centrifugation, as described above. The CHCl 3 -rich lower phase was removed, and the upper phase was washed twice with ϳ25 ml of a pre-equilibrated neutral lower phase, i.e. a lower phase generated by mixing chloroform/methanol/H 2 O (2:2:1.8, v/v). The lipid-containing lower phases were pooled; ϳ100 l of high pressure liquid chromatography grade pyridine was added, and the solvent was removed by rotary evaporation. The sample was redissolved in ϳ20 ml of chloroform/methanol/H 2 O (2:3:1, v/v) and was loaded onto a 0.5-ml DEAE-cellulose column (Whatman DE52), equilibrated as the acetate form in the same solvent (35). The column was then washed with 6 ml of CHCl 3 /methanol/120 mM aqueous ammonium acetate (2:3:1, v/v), followed by 6 ml of CHCl 3 /methanol/240 mM aqueous ammonium acetate (2:3:1, v/v) and 6 ml of CHCl 3 /methanol/0.5 M aqueous ammonium acetate (2:3:1, v/v). The lipid product, phospho-Kdo-lipid IV A , was eluted with ϳ30 ml of CHCl 3 /methanol/1 M aqueous ammonium acetate (2:3:1, v/v). Fractions containing the product were identified by spotting 5 l of each fraction onto a silica thin layer chromatography plate, which was developed in the same solvent system described above for the Kdo kinase assay. The lipid was detected by spraying the dried TLC plate with 20% sulfuric acid in ethanol, followed by charring on a hot plate. Fractions from the DEAE column containing phospho-Kdo-lipid IV A were pooled and were converted to a two-phase Bligh-Dyer system by the addition of 0.17 ml of CHCl 3 and 0.28 ml of H 2 O per ml of pool. After thorough mixing in 25-ml Corex tubes, the phases were separated by centrifugation at 1,900 ϫ g for 20 min. The combined upper phase (ϳ20 ml) was washed three times with ϳ10 ml of neutral pre-equilibrated lower phase (prepared as described above). The lower phases were pooled, and 5-10 l of high pressure liquid chromatography grade pyridine was added, and the solvent was removed by rotary evaporation at room temperature. The pure phospho-Kdo-lipid IV A (ϳ2-3 mg) was stored dry at Ϫ20°C until further analysis.
Mass Spectrometry of the Product Generated by the Recombinant Kdo Kinase-Spectra were acquired in the negative-ion linear mode by using a Kratos Analytical (Manchester, UK) time of flight matrixassisted laser desorption/ionization (MALDI) mass spectrometer, equipped with a 337 nm laser, a 20-kV extraction voltage, and timedelayed extraction (47). Each spectrum was the average of 50 shots. The matrix was a mixture of saturated 6-aza-2-thiothymine in 50% acetonitrile and 10% tribasic ammonium citrate (9:1, v/v). The product generated by the recombinant Kdo kinase was dissolved in a mixture of chloroform/methanol (4:1, v/v) before being mixed with the matrix (1:1, v/v) on a slide. The sample mixtures were allowed to dry at room temperature prior to mass analysis. The hexa-acylated lipid A 1,4Ј-bisphosphate from E. coli K-12 (Sigma) was used as an external mass standard.

Kdo Transferase and Kdo Kinase Activities in Membranes of
H. influenzae Rd-Before constructing a genomic library, it was necessary to confirm that H. influenzae Rd membranes possessed the same Kdo transferase and Kdo kinase activities previously observed in strain 722 (28). Therefore, membranes from H. influenzae Rd and 722 were assayed in parallel for these enzymatic reactions (Fig. 2). The addition of only a single Kdo residue to [4Ј-32 P]lipid IV A was observed with both types of membranes (Fig. 2, lanes 2 and 3). The calculated specific activity of Kdo transfer by the Rd membranes was about the same as that seen with strain 722 membranes (1.7 and 2.4 nmol/min/mg, respectively).
The Kdo kinase is assayed with purified Kdo[4Ј-32 P]lipid IV A as the acceptor and unlabeled ATP the donor. The addition of the hydrophilic phosphate group to Kdo[4Ј-32 P]lipid IV A generates a product that migrates more slowly than the substrate. Comparable kinase activities are present in both the Rd and the 722 membranes (Fig. 2, lanes 5 and 6), and the specific activities are 9.6 and 8.3 nmol/min/mg, respectively.

Screening of a H. influenzae Rd Genomic Library for a Cosmid That Directs Expression of Kdo Kinase in E. coli-A cosmid
library was constructed in E. coli XL1-MR using genomic DNA fragments of H. influenzae Rd as inserts in the cosmid vector pWE15. To determine the approximate sizes of the genomic inserts, 10 random single colonies were picked, and the cosmid from each colony was isolated. The cosmid DNAs were digested with EcoRI, liberating the inserted DNA from the vector. The insert sizes, estimated by gel electrophoresis, ranged from 10 to 20 kb (not shown).
Considering the size of the H. influenzae genome (1.8 megabases) (11) and the average sizes of the DNA inserts in the cosmid library, 472 individual colonies were initially picked from the cosmid library to generate a set of lysates suitable for screening for the expression of Kdo kinase activity. The lack of Kdo kinase in E. coli (28) made it a convenient background with which to search for this H. influenzae gene. However, in the concentrated lysates prepared in microtiter plates, some conversion (1-4%) of the Kdo[4Ј-32 P]lipid IV A to an unidentified product migrating like phospho-Kdo[4Ј-32 P]lipid IV A was observed irrespective of which DNA insert was present (Fig. 3). The nature of this background reaction was not characterized, as only those extracts capable of generating higher levels of phospho-Kdo[4Ј-32 P]lipid IV A -like material were of interest (Fig. 3). Accordingly, out of the 472 colony lysates, 10 possible candidates were identified for further evaluation, including the seven active extracts shown in Fig. 3.
Variability in protein concentrations, in conjunction with the ATP-independent modification(s) of Kdo-lipid IV A seen in the concentrated colony lysates (Fig. 3), necessitated the re-assay of the 10 active candidates under more controlled conditions. Larger cultures (100 ml) were grown to A 600 ϭ ϳ0.9 from the colonies harboring each of the candidate cosmids, and French press extracts were prepared. By using a final protein concentration of 0.5 mg/ml in each assay, the formation of phospho-Kdo[4Ј-32 P]lipid IV A and the ATP dependence of the reaction was re-evaluated for the 10 Kdo kinase candidates. The results for the seven active clones from Fig. 3 are shown in Fig. 4. The XL1 control samples (Fig. 4) illustrate the lack of Kdo kinase activity in extracts of the E. coli host strain. Only clone 7E demonstrated reproducible ATP-dependent Kdo kinase activity under these more rigorous assay conditions (Fig. 4).
Localization of the Kdo Kinase Gene-Cosmid DNA from clone 7E was prepared and digested with EcoRI. By gel electrophoresis analysis of the resulting fragments, it was determined that cosmid 7E contained ϳ10 kb of H. influenzae DNA. A total of five EcoRI fragments were produced (ϳ4.5, 2, 1.8, and 0.95 kb), including a fragment attributable to the cosmid (8.2 kb). The fragments derived from the insert were ligated into pBluescript IIKS. Extracts were prepared from Sure cells transformed with the ligations. The extracts were assayed for Kdo kinase activity (not shown). Only the 4556-base pair EcoRI fragment of H. influenzae DNA directed expression of the kinase. The plasmid encoding the kinase was designated pE3.2. The junctions between the vector and insert were sequenced using primers complementary to the vector. The resulting sequences (ϳ300 -400 bases from each end) were used to search the H. influenzae data base. The region encoding the Kdo kinase was narrowed to nucleotides 288512 to 293069 of the H. influenzae genome (11). The open reading frames located within this region are shown in Fig. 5. Both lgtC and hemR were ruled out as candidates for the kinase gene, since their sequences were partially deleted in pE3.2. Construction of the active subclone pB6A (Fig. 5) similarly eliminated orfM.
The only complete open reading frames on pB6A were orfZ (7) (a gene of unknown function also recently designated HI0260.1) and opsX (7, 11) (HI0260, thought to encode heptosyltransferase I of H. influenzae). Therefore, pIB100, was constructed by PCR (Fig. 5) to examine the function of orfZ. Transformation of pIB100 into Sure cells and assay of extracts demonstrated that orfZ and its immediate flanking DNA could indeed direct the expression of the kinase (not shown). The orfZ gene (Fig. 6) encodes for a protein of 241 amino acids. The sequence of the pIB100 insert (Fig. 6) is 100% identical to that in the H. influenzae genomic data base (11). Assay of pB100/ Sure cell membranes and cytosol demonstrated that ϳ80% of the Kdo kinase activity is localized to membranes (not shown), as seen with the wild type enzyme (28). However, hydropathy analysis of OrfZ revealed no obvious membrane spanning regions (not shown).
Although not initially designated as an open reading frame in the genome project, the interval containing orfZ was considered by Hood et al. (7) as a possible gene involved in LPS biosynthesis, since it is located between the core glycosyltransferase genes, lgtC and opsX (Fig. 5). No function could be ascribed to orfZ based on its sequence, and the LPS isolated from mutants containing insertions in orfZ did not appear to be dramatically altered (7). Given its function as the Kdo kinase gene, we propose the new designation kdkA in place of orfZ.
Overexpression of the Kdo Kinase and Characterization of the Recombinant Enzyme-Once the identity of the Kdo kinase gene was established, the plasmid pKdkA was constructed in which the kinase gene was placed behind a T7 promoter. Next, pKdkA was transferred into E. coli BLR(DE3)/pLysS, which synthesizes T7 polymerase when induced with isopropyl-1thio-␤-D-galactopyranoside. Following induction, the Kdo kinase was greatly overproduced, representing about 60% of the total membrane protein, as judged by SDS gel electrophoresis and Coomassie Blue staining (not shown). The specific activity of the overexpressed kinase in membranes was about 70,000 nmol/min/mg, ϳ8,000-fold higher than in wild type H. influenzae Rd membranes (8.6 nmol/min/mg). Membranes isolated from control cells harboring the vector without the insert contained no measurable Kdo kinase activity.
The membranes isolated from induced cells of pKdkA/ The percent conversions of the Kdo[4Ј-32 P]lipid IV A substrate to a compound migrating like phospho-Kdo[4Ј-32 P]lipid IV A is shown for 96 colony extracts, representing one microtiter dish out of the six that were screened. The lysates were prepared from single colonies of the library and assayed as described under "Experimental Procedures." The seven active extracts were further evaluated by the quantitative assay shown in Fig. 4. This analysis showed that only the lysate marked with the asterisk expresses high levels of Kdo kinase activity.

FIG. 4. Re-assay of Kdo kinase activity in fresh extracts prepared from active candidates identified in the initial screening.
French press extracts were prepared from freshly grown cultures of each of the active candidate clones, as well as the host strain XL1, and were assayed at 0.5 mg/ml for Kdo kinase activity under standard conditions but with 50 M Kdo[4Ј-32 P]lipid IV A (3000 cpm/nmol). The ATP dependence of the reaction was assessed by assaying the extracts in the absence or presence of 5 mM ATP, as indicated. The reactions were incubated for 30 min at 30°C and analyzed by TLC analysis to determine the extent of phospho-Kdo[4Ј-32 P]lipid IV A formation. The opsX gene is thought to encode heptosyltransferase I, whereas lgtC putatively encodes an outer core glycosyltransferase that functions after heptose incorporation (7). The role of orfM is unknown. BLR(DE3)/pLysS were used to characterize some of the catalytic properties of the Kdo kinase. As seen with wild type H. influenzae membranes, the recombinant kinase activity was optimal in the presence of 0.1-0.2% Triton X-100 and had maximal activity at a pH of 7.5 in HEPES buffer (not shown). The reaction was linear with membrane protein from 0.02 to 0.4 g/ml and with time for up to 10 min at 0.02 g/ml (Fig. 7). The apparent K m for Kdo-lipid IV A at saturating ATP concentration (5 mM) was 11.6 Ϯ 0.8 M, with a V max of 73,625 nmol/min/mg (at a protein concentration of 0.04 g/ml).
Like the wild type H. influenzae kinase, the recombinant Kdo kinase in membranes of pKdkA/BLR(DE3)/pLysS preferred ATP over other nucleotide triphosphates (Fig. 8) (28). However, minor enzymatic activity was detected with GTP (Fig. 8). Furthermore, as shown in Fig. 8 [␥-32 P]ATP was utilized as the donor and unlabeled Kdo-lipid IV A as the acceptor. The reactions were then analyzed by thin layer chromatography, followed by PhosphorImager analysis, to demonstrate the incorporation of 32 P into the putative phospho-Kdo-lipid IV A (Fig. 9). Lanes 1-3 are control reactions under conditions similar to the standard kinase assay, in which Kdo[4Ј-32 P]lipid IV A is the acceptor and ATP is the co-substrate required for phospho-Kdo[4Ј- 32

P]lipid IV A formation (lane 3).
In lanes 4 -6 of Fig. 9, the substrate concentrations are the same as in lanes 1-3, but [␥-32 P]ATP is used in conjunction with unlabeled lipid acceptor. Lane 4 (the no enzyme control) shows the migration of the [␥-32 P]ATP substrate in this solvent system. Lane 5 is derived from a reaction mixture containing the recombinant kinase and [␥-32 P]ATP but lacking Kdo-lipid IV A . In this case, no 32 P-labeled lipid product is formed. Finally, lane 6 of Fig. 9 conclusively demonstrates that the 32 P of [␥-32 P]ATP is transferred to Kdo-lipid IV A when all components of the system are present. The lipid product obtained in this manner (Fig. 9, lane 6) migrates with the same Rf as that obtained with Kdo[4Ј-32 P]lipid IV A and unlabeled ATP (Fig. 9,  lane 3). In previous studies with wild type membranes of H. influenzae it was difficult to demonstrate such transfer of 32 P from [␥-32 P]ATP to Kdo-lipid IV A , probably because of interfering phosphatases and/or endogenous nucleotides at the rela-FIG. 6. Sequence of the H. influenzae orfZ/kdkA DNA fragment corresponding to IB100 subclone (see "Experimental Procedures"). In the first version of the H. influenzae data base (11), the interval containing orfZ/kdkA was not assigned an open reading frame. However, the predicted protein sequence corresponding to the Kdo kinase is now designated HI0260.1 in the revised data base. The DNA sequence shown is identical to the corresponding region in the H. influenzae data base. The start codon for kdkA is underlined, and the protein sequence is shown above the DNA coding strand. The partial sequences, both nucleotide and protein, of opsX and orfM encoded on the IB100 plasmid are also shown. The construction of IB100 and the expression of Kdo kinase activity by this plasmid demonstrated conclusively that kdkA is the structural gene for the H. influenzae Kdo kinase. tively high membrane concentrations employed as the enzyme source (28).
Lipid  (Fig. 10). The inability of [4Ј-32 P]lipid IV A to serve as an acceptor supports the idea that the enzyme phosphorylates the Kdo moiety. The lack of activity with Kdo 2 [4Ј-32 P]lipid IV A shows that derivatization of the 4-OH position of the inner Kdo interferes with kinase function, consistent with the hypothesis that the kinase phosphorylates the 4-OH of the inner Kdo.
Mass Spectrometry of the Product Generated by the Recombinant Kdo Kinase-The reaction product generated by the recombinant Kdo kinase was isolated and subjected to mass spectrometry. The MALDI-time of flight mass spectrum in the negative mode is shown in Fig. 11. The structure of the proposed phospho-Kdo[lipid IV A product is shown in the inset. The prominent peak at m/z 1705.1 is interpreted as the (M-H) Ϫ of the parent compound, given the predicted molecular weight of 1705.86 for phospho-Kdo[lipid IV A . The observed (M-H) Ϫ confirms that only a single phosphate group is transferred to Kdo[lipid IV A by the recombinant kinase. The other major peak at m/z 1405.2 represents the anionic lipid fragment remaining after cleavage of the Kdo glycosidic linkage, consistent with the loss a phospho-Kdo unit from the parent compound. This lipid fragment is interpreted as the lipid IV A anion, which has a predicted molecular weight of 1404.7. The observed fragmentation pattern is very similar to that reported in previous studies with H. influenzae membranes (28) and further validates the proposal that the kinase phosphorylates only the Kdo residue of Kdo[lipid IV A . However, unequivocal identification of the 4-OH of Kdo as the site of phosphorylation will require additional structural studies.

DISCUSSION
In the present study we report the first identification and cloning of a structural gene encoding a Kdo kinase (28) (Fig. 1). The H. influenzae gene that we have found corresponds to a previously identified open reading frame, designated orfZ (or HI0260.1) (7). Hood and co-workers (7) postulated that orfZ might be involved in H. influenzae LPS biosynthesis because of its proximity to two other genes (opsX and lgtC) encoding putative core glycosyltransferases (Fig. 5). However, they were unable to assign a biochemical function to the orfZ gene because of its unique sequence and the apparently normal size of the LPS isolated from orfZ mutants (7). Our previous discovery and development of an assay for the Kdo kinase (28) in extracts of H. influenzae have now enabled the unambiguous identification of orfZ as the Kdo kinase structural gene. Expression cloning, in conjunction with the availability of the H. influenzae genomic data base (11), greatly accelerated the search for the kinase gene. As illustrated by our study, new biochemical assays will be very useful for elucidating the roles of the many open reading frames of unknown function, recently uncovered by genome sequencing. Development of new biochemical assays, however, hinges upon the elucidation of the structures of previously uncharacterized (or partially characterized) natural products like LPS (6,48) or the identification of novel physiological processes.
The cloning and overexpression of kdkA facilitated the characterization of several catalytic properties of the kinase that previously were difficult to evaluate using H. influenzae membranes (28) as the enzyme source. Specifically, the divalent cation requirement (Fig. 8) and the incorporation 32 P from [␥-32 P]ATP into the lipid product ( Fig. 9) were demonstrable with the overexpressed kinase. Mass spectrometry (Fig. 11) confirmed the incorporation of only a single phosphate residue by the recombinant Kdo kinase. However, the proposed location of the phosphate group at position 4-OH on the Kdo moiety remains to be confirmed.
The conclusive identification of the Kdo kinase structural gene has allowed for the search for related genes in other Gram-negative bacteria. Highly significant kdkA homologues were found in the partially sequenced genomes of B. pertussis, V. cholerae, A. actinomycetemcomitans, S. putrefaciens, and P. multocida. Strains of B. pertussis and V. cholerae contain a phosphate group attached to the single Kdo that is present in their LPS (29 -32), and Actinobacillus is closely related to Haemophilus. A distant homologue of kdkA is also present in Por-phyrmonas gingivalis, which contains a single Kdo in its LPS, thought to be phosphorylated at position 7 or 8 (49). No homologues of the kinase were found in the many Gram-negative bacteria that contain two or more Kdo residues in their inner cores, with the exception of Pseudomonas aeruginosa, which contains two weak kdkA homologues. P. aeruginosa LPS is modified with several distinct phosphate residues in its core (50), perhaps explaining the presence of weak kdkA homologues. Analysis of the kdkA sequence using the PSI-BLAST algorithm (51) showed the Kdo kinase is not member of a large protein family. In this kind of alignment, kdkA does display limited homology to a number of protein kinases over a segment of ϳ20 amino acid residues, possibly representing the location of the ATP]binding site.
The biological significance of Kdo phosphorylation versus the presence of multiple Kdo residues that are characteristic of the inner LPS cores of most other Gram-negative bacteria remains to be established. Based on our findings, it appears that Kdo phosphorylation in H. influenzae LPS plays a crucial role in modulating the virulence of the bacteria in certain animal models. Our identification of kdkA/orfZ as the structural gene for the Kdo kinase sheds new light on the intriguing genetic and biochemical data of Hood et al. (7), who found that when kdkA/orfZ was disrupted, the bacteria were viable and made LPS of relatively normal size but were rendered avirulent. Apparently, Kdo phosphorylation is somehow essential for bacterial pathogenesis. It would now be very interesting to introduce the bifunctional E. coli Kdo transferase (22) into a H. influenzae mutant defective in kdkA and determine whether or not the addition of a second Kdo in place of the phosphate can restore bacterial virulence.
As a next step, it will be important to demonstrate that the Kdo kinase is indeed missing in the orfZ-deficient mutants reported by Hood et al. (7). The absence of the Kdo[linked phosphate group in the LPS of this mutant in vivo also remains to be confirmed. We cannot yet exclude the scenario that H. influenzae contains more than one Kdo kinase, although this seems unlikely based on the genome sequence (11). Finally, the possibilities that a defect in Kdo phosphorylation causes secondary changes in LPS structure, outer membrane protein assembly, and/or growth rate need to be investigated before the biological significance of Kdo phosphorylation in pathogenesis can be fully evaluated.