Construction of a Deep-rough Mutant of Burkholderia cepacia ATCC 25416 and Characterization of Its Chemical and Biological Properties*

Burkholderia cepacia is a bacterium with increasing importance as a pathogen in patients with cystic fibrosis. The deep-rough mutant Ko2b was generated from B. cepacia type strain ATCC 25416 by insertion of a kanamycin resistance cassette into the gene waaC encoding heptosyltransferase I. Mass spectrometric analysis of the de-O-acylated lipopolysaccharide (LPS) of the mutant showed that it consisted of a bisphosphorylated glucosamine backbone with two 3-hydroxyhexadecanoic acids in amide-linkage, 4-amino-4-deoxyarabinose (Ara4N) residues on both phosphates, and a core oligosaccharide of the sequence Ara4N-(1 → 8)d-glycero-d-talo-oct-2-ulosonic acid (Ko)-(2 → 4)3-deoxy-d-manno-oct-2-ulosonic acid (Kdo). The mutant allowed investigations on the biosynthesis of the LPS as well as on its role in human infection. Mutant Ko2b showed no difference in its ability to invade human macrophages as compared with the wild type. Furthermore, isolated LPS of both strains induced the production of tumor necrosis factor α from macrophages to the same extent. Thus, the truncation of the LPS did not decrease the biological activity of the mutant or its LPS in these aspects.

Taxonomically, Burkholderia cepacia is a complex of eight species or genomovars (1), which are all characterized as nutritionally versatile as well as notoriously drug-resistant (2) and all display a complex genome (3). Strains have been described to live on penicillin G as the sole carbon source (4), others are able to degrade chlorinated aromatic substrates, such as the principal ingredient of the herbicide Agent Orange (5). Furthermore, the organism can antagonize and repress many soil-borne plant pathogens and is considered as a plant growth-promoting rhizobacterium (6). However, regarding the pathogenic potential of B. cepacia, there has been a controversial discussion about the use and especially the liberation of this organism for biological control and soil decontamination purposes (6 -9).
First described as a phytopathogen associated with soft rot on onion bulbs (10), B. cepacia has recently emerged as an opportunistic pathogen in humans (3). Most important is its role as a pulmonary pathogen in patients with cystic fibrosis (CF) (2,11,12) or chronic granulomatous disease (CGD) (13,14) and in immunocompromised patients. The so-called cepacia syndrome occurs in ϳ20% of CF patients infected with B. cepacia, and is characterized by fever, acute necrotizing pneumonia, and bacteremia with a high mortality (12). In CGD patients B. cepacia pneumonia is the second leading cause of death (13,14).
The genome of B. cepacia consists of two to four chromosomes with an overall genome size of 5-9 Mb and is characterized by a large number of insertion elements (15,16). These mobile elements could easily facilitate the horizontal transfer of virulence genes from other pathogenic bacteria and thus enhance the pathogenicity of B. cepacia (11). Previous studies have demonstrated that B. cepacia strains are able to invade macrophages and respiratory epithelial cells (17,18) and can persist in macrophages; however, without significant bacterial replication (19). This property could contribute to the pathogenicity of the bacteria in lung damage and their systemic spread. A number of virulence factors are known for B. cepacia. In addition to extracellular products such as hemolysins, proteases, and lipases, iron-binding siderophores play a major role for virulence (20) as well as mucin binding cable pili, which adhere to mucin in lungs of CF patients (21). Another important virulence determinant is the lipopolysaccharide (LPS), 1 which is likely to contribute to the inflammatory nature of B. cepacia infections in cystic fibrosis patients. Clinical isolates as well as environmental strains possess LPS of high inflammatory potential (22), which induces much higher levels of tumor necrosis factor ␣ (TNF-␣) in monocytes than the LPS from Pseudomonas aeruginosa (23,24) and stimulates the up-regulation of other proinflammatory cytokines (25,26).
The common architecture of LPS and most steps of its biosynthesis are well known (27). One sugar constituent of the inner core region, found in all LPS structures, is 3-deoxy-Dmanno-oct-2-ulosonic acid (Kdo), which is directly linked to the lipid A. Most LPS contain an ␣(2 3 4)-linked Kdo disaccharide in their inner core region; however, several bacteria like Haemophilus, Vibrio, and Bordetella depict one phosphorylated Kdo in this position (28). In case of LPS from B. cepacia ATCC 25416 it was revealed that the inner core region consists of a * This work was supported by Deutsche Forschungsgemeinschaft Grants SFB 470/A1 (to H. B. and S. G.) and LI 448/1-1 (to B. L.). 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 (29). The unusual sugar Ko has been found earlier as a substitute for Kdo in the LPS of Acinetobacter (30 -32) and Tatlockia (33) and was recently identified in the LPS of Yersinia pestis (34), but so far nothing is known about the significance of this substitution. Interestingly, the core region of Burkholderia caryophylli does not contain Ko (35). A lot of data exist about the biosynthesis and transfer of Kdo into the LPS (36); however, no data are available about the incorporation of Ko. One could assume that a specific enzyme introduces a hydroxyl function at position 3 of the side branch Kdo. On the other hand, Ko could be synthesized by an independent pathway and transferred by a specific transferase. To answer these questions we investigated the early glycosyltransferases involved in the biosynthesis of the LPS in B. cepacia ATCC 25416. First, the Kdo transferase (WaaA) of this strain was cloned and characterized to determine whether it is a mono-or bifunctional enzyme. We show here that B. cepacia harbors a bifunctional Kdo transferase comparable to WaaA from Escherichia coli. Second, the gene encoding heptosyltransferase I (waaC) was cloned from B. cepacia, expressed in Corynebacterium glutamicum and characterized as a monofunctional heptosyltransferase in vitro and in vivo.
Although most clinical isolates belong to genomovar III, so far the type strain ATCC 25416 (genomovar I) is the only strain of the B. cepacia complex with a known LPS core structure (29). Whether Ko is a constituent of the LPS in other genomovars, as well, remains to be investigated. We were interested in whether it was possible to generate a deep-rough, heptose-less mutant of the type strain in order to find out whether the mutant also contained a Ko residue in its truncated LPS. To incorporate this unusual sugar the bacterium has to use an additional biosynthetic pathway or at least an additional modifying enzyme. To learn something about the reasons why bacteria replace the conserved Kdo residues partially with Ko, the biological activity of the mutant was compared with that of the wild type. The answers to some of these questions are reported here.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Growth Conditions-All bacterial strains and plasmids used are listed in Table I. B. cepacia was grown at 30°C in nutrient broth, which was supplemented with 150 g/ml kanamycin for growth of the mutant B. cepacia Ko2b. Recombinant C. glutamicum strains were cultivated at 30°C in 2ϫ YT medium supplemented with kanamycin (30 g/ml). E. coli was grown at 37°C in Luria-Bertani medium supplemented with the appropriate antibiotics.
General Cloning Techniques, Gene Library Construction, and DNA Sequence Analysis-Manipulation of recombinant DNA was according to standard protocols (37). DNA was transformed by electroporation using a BioRad Gene Pulser. For gene library construction, chromosomal DNA of B. cepacia ATCC 25416 was partially digested with Sau3AI. Fragments ranging from 1.5 to 6 kb were ligated into the BamHI site of cloning vector pZErO2.1 (Invitrogen) and transformed into E. coli TOP10FЈ, yielding ϳ6 ϫ 10 5 colonies from which recombinant plasmids were prepared. The digoxygenin-11-dUTP system (Roche Diagnostics) was used for DNA labeling and detection in Southern experiments according to the instructions of the manufacturer. Determination of the sequence was carried out by cycle sequencing with custom-made primers (MWG-Biotech) and fluorescent dye terminators on an ABI 377 sequencing instrument (PerkinElmer Life Sciences). The computer programs Geneworks (Intelligenetics) and Basic Local Alignment Search Tool (TBLASTN; National Center for Biotechnology Information) were used to analyze DNA and deduced amino acid sequences.
Cloning of waaC and Plasmid Constructions-Approximately 2 g of the gene library were used for in vivo complementation of Salmonella enterica serovar Typhimurium LT2 deep-rough strain SA1377 (waaC630) (38) as described previously (39), and recombinant clones were tested for the expression of smooth type LPS (S-LPS). From this library, plasmid pBC1/1377 was isolated, and waaC was identified within its insert. To inactivate waaC by inserting a kanamycin-resistance cassette, the kanamycin resistance of the vector part of pBC1/ 1377 had to be exchanged for a chloramphenicol resistance. Thus, the chloramphenicol acetyltransferase gene was amplified with primers CAT1Spe (5Ј-CGCAGAATACTAGTATCCTGGTGTCC-3Ј, SpeI site in boldface type) and CAT2BspHI (5Ј-TTTGCGTTCATGAGCTTGGCAC-3Ј, BspHI site in boldface type) using plasmid pHSG397 as a template and ligated with two fragments of pBC1/1377: one HindIII-SpeI fragment, containing the waaC gene, and one HindIII-BspHI fragment containing the origin of replication. After ligation of the three fragments, the resulting plasmid was opened with EcoRI cut inside of waaC, and a non-polar kanamycin resistance cassette, obtained by EcoRI digestion of plasmid pUC4K, was inserted. The final plasmid was termed pBCwaaC::kan. The inactivated waaC gene was excised as a EcoRV-SphI fragment from plasmid pBCwaaC::kan and inserted into the suicide plasmid pCVD442 digested with SmaI and SphI. The resulting plasmid pCVDwaaC::kan was amplified in E. coli SM10pir. For subcloning, the waaC gene was amplified by PCR with primers BcewaaC1 (5Ј-GGCGGATCCTCAGCGTGCAAAAGAT-3Ј, BamHI site in boldface type) and BcewaaC2 (5Ј-GTATCTGCAGGGCGTTACAG-GAGACCG-3Ј, PstI site in boldface type) using plasmid pBC1/1377 as template. After digestion with BamHI and PstI the amplificate was ligated into shuttle vector pCB20, digested with the same enzymes. The resulting plasmid pBCC was used for transformation of S. enterica mutants SA1377 (waaC630) and SL3789 (waaF511) and C. glutamicum R163 for complementation and preparation of cell-free extracts, respectively.
For complementation of mutant Ko2b the waaC gene was excised as a BamHI-PstI fragment from plasmid pBCC and ligated into pBBR1Tp, a vector with a broad host range carrying a trimethoprim resistance. Deep-rough mutant Ko2b was transformed with the resulting plasmid pBBR1Tp/waaC, and recombinant clones were analyzed for expression of the S-form LPS in silver-stained SDS-polyacrylamide gels.
A PCR cloning strategy based on data from the ongoing genome sequencing project of B. cepacia genomovar III strain J2315 at the Sanger Center (www.sanger.ac.uk) was employed for cloning of the Kdo transferase (waaA). Based on sequence homologies to other waaA genes, primers BcwaaA1 (5Ј-TTGAGCGCGCGCCTGCTTCG-3Ј) and BcwaaA2 (5Ј-ACGCGAGTCCGGGCGGCATCG-3Ј) were designed to amplify by PCR a waaA-encoding fragment from chromosomal DNA of B. cepacia ATCC 25416. The amplificate was directly cloned into vector pT-Adv using the AdvanTAge PCR cloning kit (Clontech) and expressed in E. coli XL1-Blue employing blue/white screening. Sequence analysis was carried out on both strands of five different clones as well as on a direct amplificate of waaA from chromosomal DNA. For in vitro analysis waaA was amplified by PCR using primers BcwaaAATG (5Ј-GC-ATGCTGCGGGTGATCTATCGCG-3Ј), BcwaaAStop (5Ј-CGTCTATG-CGTCGTCGGCAGCGTC-3Ј) and Pfu polymerase. After purification the amplificate was phosphorylated with polynucleotide kinase (Stratagene) and ligated as a blunt-end fragment into vector pJKB20 (40), digested with EcoRV, and transformed into E. coli XL1-Blue. Resulting transformants were analyzed for correct orientation by restriction with BamHI. One construct was termed pQB2, transformed into C. glutamicum R163 and sequenced.
Construction and Genetic Characterization of the Chromosomal Knockout Mutation-B. cepacia was transformed with plasmid pCVDwaaC::kan by electroporation, and recombinant clones were selected on agar plates supplemented with 200 mg/liter kanamycin sulfate. The resulting chromosomal knockout mutant was termed B. cepacia Ko2b and was characterized by Southern hybridization and analytical PCR using chromosomal DNA as template. For this purpose the primer pair BCCKO1 (5Ј-GTGCAAAAGATCCTGATCG-3Ј) and BC-CKO2 (5Ј-TTACAGGAGACCGAAGCC-3Ј) was used to amplify the complete waaC gene, whereas primers SacB1 (5Ј-TAACCTTTACTACCG-GACTG-3Ј) and SacB2 (5Ј-TTGTCCTTGTTCAAGGATGC-3Ј) were used to identify remaining vector sequences of pCVDwaaC::kan by amplification of the sacB gene. Three different probes were used for Southern experiments. The first probe was an EcoRI fragment of pUC4K containing the kanamycin resistance cassette introduced into the waaC gene, the second harbored the sacB gene from pCVD442 on a PstI fragment and the third contained nucleotides 73-724 of the waaC gene including the insertion site of the kanamycin resistance cassette as EcoRV-EcoRI fragment. The chromosomal DNA isolated from B. cepacia Ko2b was digested with EcoRI, PstI, or EcoRV-SphI, respectively.
Nucleotide Sequence Accession Numbers-The sequences encoding the Kdo transferase (WaaA) and the heptosyltransferase I (WaaC) of B. cepacia ATCC 25416 were deposited at GenBank TM /EBI and are available under accession numbers AJ420777 and AJ302064, respectively.
In Vitro Analysis of Cloned Glycosyltransferases-Cell-free extracts containing Kdo transferase or heptosyltransferase I from B. cepacia were prepared from C. glutamicum strains R163/pQB2 and R163/ pBCC, respectively, as described previously (41). The in vitro tests for Kdo transfer were carried out in a total volume of 20 l containing HEPES (50 mM, pH 7.5), MgCl 2 (10 mM), and Triton X-100 (3.2 mM). CMP-Kdo was generated in situ from Kdo (2 mM) and CTP (5 mM) using purified CMP-Kdo synthetase as described (40). The synthetic tetraacylated lipid A precursor of E. coli, compound 406 (42), was used as lipid acceptor for Kdo transfer at a concentration of 100 M. As Kdo transferases for control reactions cell-free extracts of R163/pJKB16 (E. coli) or R163/pCB23 (Haemophilus influenzae) were used (41). A crude cellfree extract obtained from B. cepacia ATCC 25416 served as additional control. For heptose transfer, reactions were carried out in two steps; first, Kdo 2 -406 was generated by incubation with WaaA from E. coli for 1 h at 37°C and stopped by boiling for 10 min. Thereafter, ADP-heptose and cell-free extract containing a heptosyltransferase were added to the mixture, another incubation of 1 h at 37°C followed, and reactions were stopped by freezing at Ϫ20°C. The heptose donor ADP-heptose was provided by a partially purified (43) cell extract of the heptosyltransferase-deficient E. coli strain WBB06 (44). The control heptosyltransferase was prepared from R163/pJKB10 (E. coli) (40). Aliquots were spotted on silica-60 TLC plates (Merck) and developed in a solvent mixture of chloroform/pyridine/88% formic acid/water (30:70:16:10, v/v). Blotting and immunological identification of the reaction products was performed as described (40) using monoclonal antibodies A20 (45), A17 (46), and S36-20 (47), recognizing a terminal Kdo residue, a Kdo disaccharide (Re-type LPS), and Rd 2 -type LPS, respectively.
SDS-PAGE-Proteinase K-digested cell-free lysates were prepared from recombinant strains (44), separated by SDS-PAGE, and detected with alkaline silver nitrate (48).
Preparation and Chemical Analysis of LPS-B. cepacia wild-type LPS was extracted according to the phenol/water method (49). B. cepacia Ko2b was cultivated for 24 h at 30°C in nutrient broth supplemented with 150 mg/liter kanamycin sulfate. The cells were dried as described (50), and LPS was extracted using a phenol/water/sarkosyl procedure followed by treatment of the aqueous phase material with phenol/chloroform/light petrol ether (51). Analyses of glucosamine, Kdo, and phosphate were performed as described (52). Analysis of complete fatty acids was carried out according to Ref. 53. LPS was de-O-acylated by mild hydrazinolysis (54), phosphate groups were removed with 48% hydrofluoric acid (4°C; 48 h), and the resulting product was dialyzed, reduced with NaBH 4 , and subjected to strong alkaline hydrolysis to achieve de-N-acylation (54). After neutralization and desalting single oligosaccharides were separated by analytical high performance anion exchange chromatography on a Dionex DX300 system as described previously (50). De-O-acylated LPS (LPS de-O-Ac ) and isolated oligosaccharides were subjected to mass spectrometry (MS). To determine the position of Ara4N in the core oligosaccharide of mutant Ko2b, LPS de-O-Ac was subjected to mild methanolysis followed by N-acetylation and permethylation for analysis by gas chromatography-MS (GC-MS) as described by Isshiki et al. (29).
Mass Spectrometry-Mass spectrometric analysis was performed in the negative ion mode using an electrospray Fourier-transform ion cyclotron resonance (ESI-FT-ICR) mass spectrometer (APEX II, Bruker Daltonics, Billerica, MA) equipped with a 7 Tesla actively shielded magnet and an Apollo ion source. Mass spectra were acquired using standard experimental sequences as provided by the manufacturer. Samples were dissolved at a concentration of ϳ20 ng/l in a 50:50:0.001 (v/v/v) mixture of 2-propyl alcohol, water, and triethylamine and sprayed at a flow rate of 2 l/min. Capillary entrance voltage was set to 3.8 kV, and dry gas temperature to 150°C. Capillary skimmer dissociation (CSD) was induced by increasing the capillary exit voltage from Ϫ100 V to Ϫ350 V. Infrared-multiphoton dissociation of isolated parent ions was performed with a 35 watt, 10.6 m CO 2 laser (Synrad, Mukilteo, WA). The unfocused laser beam was directed through the center of the ICR cell for 40 ms, and the fragment ions were detected after a delay of 0.5 ms.
Macrophage Invasion Assay-Invasion assays were carried out as described earlier (18). Briefly, B. cepacia was grown at 30°C to lateexponential phase, cells were washed and diluted in phosphate-buffered saline, and then added to 4 ϫ 10 5 human macrophages (at a concentration of 8 ϫ 10 5 per ml) at a multiplicity of infection (m.o.i.) of 1-7 bacteria per macrophage. After 1 h of incubation at 37°C macrophages were washed vigorously with warm Hank's balanced salt solution (HBSS) to remove extracellular bacteria. Fresh medium containing ceftazidime (1 mg/ml) and amikacin (1 mg/ml) was added to kill extracellular bacteria, and cultures were incubated for an additional 2 h at 37°C. To determine bacterial uptake macrophages were lysed after 1 and 3 h of incubation by addition of 0.1% saponin. Lysates were serially diluted and plated on agar. As controls, bacteria were cultivated without macrophages and antibiotics to monitor growth rates at 37°C.

RESULTS
Cloning and Sequence Analysis of waaA from B. cepacia-DNA sequences from the ongoing genome sequencing project of B. cepacia genomovar III strain J2315 were searched for homologies with known Kdo transferase genes. Two contigs were identified including the start and the end of a putative Kdo transferase gene. The putative waaA gene was amplified by PCR from chromosomal DNA of B. cepacia ATCC 25416 (type strain; genomovar I). The resulting amplificate of ϳ1500 bp was directly cloned into vector pT-Adv and sequenced. Sequence analysis of both strands revealed an open reading frame encoding a protein of 452 amino acids and a predicted molecular mass of 48 kDa. Its amino acid sequence showed highest homology to the Kdo transferases from Bordetella pertussis and Bordetella bronchiseptica (43% identity, 54% similarity). The protein contained a glycosyltransferase I motif at position 309 -413 according to the Pfam data base, which is present in all Kdo transferases investigated so far (39,59). After subcloning into the shuttle vector pJKB20 both strands of five different clones were sequenced. For comparison, a direct amplificate of waaA from chromosomal DNA was sequenced. One clone showed a point mutation and was discarded; all other sequences were identical.
Cloning and Sequence Analysis of waaC from B. cepacia-The waaC gene from B. cepacia was cloned from a gene library by functional complementation of S. enterica mutant SA1377 (waaC630) (38). Selection of transformants with novobiocin and phage Ffm (39) resulted in the discovery of plasmid pBC1/ 1377. As shown on silver-stained SDS-polyacrylamide gels of proteinase K-treated whole cell lysates, the plasmid was able to restore the biosynthesis of smooth LPS in S. enterica SA1377 (waaC630) (data not shown). Sequence analysis of pBC1/1377 revealed an open reading frame encoding a protein of 331 amino acids and a predicted molecular mass of 36 kDa. The deduced amino acid sequence contained a heptosyltransferase motif (Pfam data base accession no. PF01075, amino acids 76 -330) and possessed significant homology to other known heptosyltransferases I; highest homology was shown to WaaC of P. aeruginosa (42% identity, 57% similarity). Downstream of the waaC gene more putative open reading frames were identified; however, no functions could be assigned. After subclon-ing of the waaC gene in a shuttle vector the resulting plasmid pBCC was transformed into S. enterica mutants SA1377 (waaC630) and SL3789 (waaF511) for complementation studies. As shown in Fig. 1 the enzyme was able to restore expression of S-form LPS in strain SA1377 (lane 4) but not in SL3789 (lane 5), indicating that WaaC of B. cepacia is functionally equivalent to other heptosyltransferases I (39,40,43,60).

In Vitro Characterization of the Cloned Kdo Transferase and Heptosyltransferase I of B. cepacia-
The putative open reading frame of the waaA gene was amplified by PCR and cloned into the shuttle vector pJKB20 to give plasmid pQB2. Plasmids pQB2 and pBCC were used to transform C. glutamicum R163. Cell-free extracts were prepared from both strains and subjected to in vitro assays. The activity of WaaA was tested as described (39), using the synthetic lipid acceptor 406, Kdo, CTP, and CMP-Kdo synthetase from E. coli. As controls the Kdo transferases from E. coli and H. influenzae were included as well as the cell-free extract from B. cepacia ATCC 25416. All transferases converted acceptor 406 into one or more polar products, which reacted with mAb A20, indicating the transfer of at least one Kdo residue ( Fig. 2A). The cloned Kdo transferase of B. cepacia is clearly bifunctional comparable to that of E. coli and also the crude extract from B. cepacia generated a product that reacted with mAb A17, demonstrating the presence of a 2 3 4-linked Kdo disaccharide (Fig. 2B). Only the transferase from H. influenzae displayed mostly monofunctional activity.
The activity of WaaC from B. cepacia was investigated using synthetic lipid acceptor 406 in a coupled enzyme assay with Kdo transferase from E. coli and a low molecular mass filtrate from E. coli WBB06 (44) as the source of ADP-heptose (40). The cloned WaaC of E. coli was used as a reference (40). Both transferases led to the formation of a more polar product than the acceptor structure Kdo 2 -406 after separation of reaction products on TLC plates (Fig. 3, lanes 2 and 3). Bands were either visualized with mAb S36-20, specific for Rd 2 -type LPS (47), to identify reaction products (Fig. 3, panel B), or with mAb A20 (45), recognizing a terminal Kdo residue, to identify the reaction intermediate Kdo 2 -406 (Fig. 3, panel A). When either heptosyltransferase I or ADP-heptose were omitted from the reaction mixture no Rd 2 -type LPS was generated (Fig. 3, lanes  4 and 5).
Construction genome of B. cepacia. For this purpose, the cloned waaC gene on plasmid pBC1/1377 was inactivated by insertion of a nonpolar kanamycin resistance cassette derived from vector pUC4K. The resulting construct was subcloned into the suicide vector pCVD442, which possesses a pir-dependent origin of replication and needs the presence of the -protein of phage lambda to replicate (61). When introduced into B. cepacia, only those cells survived selection on kanamycin, which had integrated the plasmid into the chromosome by homologous recombination. To enforce excision of the wild type allele and remaining vector sequences from the chromosome, a counter selection on sucrose was carried out. The sacB gene from Bacillus subtilis, also present on plasmid pCVD442, encodes a levansucrase that synthesizes a polymer from sucrose, which is lethal in many bacteria (62). When cultivated on agar plates containing sucrose and kanamycin only those mutants survived in which the wild type allele had been replaced by the inactivated copy of waaC, and vector sequences were no longer present on the chromosome. One of the isolated mutants was termed B. cepacia Ko2b and further analyzed. To confirm the construction of mutant Ko2b analytical PCRs were carried out using primers BCCKO1 and BCCKO2 as well as SacB1 and SacB2. Primers BCCKO1 and BCCKO2 bound to the 5Ј-and 3Ј-region of the waaC gene, respectively, and led to amplification of the complete gene. Using chromosomal DNA from wild type B. cepacia as a template, an amplificate of ϳ1 kb was generated, consistent with the length of waaC (996 bp). However, when chromosomal DNA from mutant Ko2b was used as template, the sole amplificate showed a length of ϳ2.3 kb, indicating the presence of the kanamycin resistance cassette (1282 bp) within the sequence of waaC (996 bp) (data not shown). Primer pair SacB1 and SacB2 were designed for amplification of the sacB gene. Using DNA from plasmid pCVD442 as template, a PCR product of ϳ1.3 kb was obtained, consistent with the length of sacB (1347 bp). When either chromosomal DNA from B. cepacia wild type or mutant Ko2b was used as template, no amplificate could be obtained, indicating that the vector sequence had been excised from the chromosome (data not shown). To further verify the results, Southern hybridization was carried out using three different probes. Hybridization of chromosomal DNA from either wild type or mutant Ko2b with probes against (i) the kanamycin resistance cassette, (ii) the sacB gene, and (iii) the waaC gene gave signals (a) only with Ko2b DNA, (b) only with control plasmid, and (c) with all DNA. However, with Ko2b DNA the hybridizing fragment was ϳ1.3 kb larger than that of wild type DNA consistent with the insertion of the kanamycin resistance cassette (1282 bp) (data not shown). Taken together, the results indicate that the waaC gene on the B. cepacia chromosome was successfully replaced by an inactivated copy.
Chemical Characterization of LPS from B. cepacia Wild Type and Strain Ko2b-B. cepacia wild type LPS was extracted according to the phenol/water method (49), whereas preliminary experiments had shown that extraction of mutant Ko2b gave highest yields when 2% N-lauroyl sarkosyl sodium was added to the phenol/water mixture (51). For further purification, the material was extracted with phenol/chloroform/light petrol ether, and the LPS obtained was dissolved in water and precipitated with ethanol/acetone. Separation of LPS on SDSpolyacrylamide gels (Fig. 4) clearly showed that B. cepacia wild type expressed S-form LPS with a typical ladder pattern (lane 2), whereas the mutant Ko2b appeared as deep-rough Re-type LPS (lane 3) comparable to that of S. enterica strain SA1377 (lane 1). When mutant Ko2b was transformed with plasmid pBBR1Tp/waaC, the ability to express complete S-form LPS was restored (data not shown).
Chemical analyses of Ko2b LPS identified GlcN, Kdo, and phosphorous, as well as the fatty acids 3-hydroxytetradecanoic acid, 3-hydroxyhexadecanoic acid, tetradecanoic acid, and hexadecanoic acid. To elucidate the chemical structure of the core region of the mutant LPS in more detail LPS was de-Oacylated by mild hydrazinolysis (LPS de-O-Ac ). Negative ion ESI-FT-ICR mass spectra of LPS de-O-Ac showed several single-, double-, and triple-charge molecular ion peaks. For easier interpretation the charge deconvoluted spectrum is given in Fig. 5A  pected molecular structure depicted in Fig. 6 consisting of one Ko, one Kdo, two glucosamines, two phosphate groups, two hydroxyhexadecanoic acids, and two or three Ara4N residues, respectively. Both peaks were accompanied by peaks of low intensity originating from the exchange of Ko by Kdo (⌬m/z Ϫ 16) and from sodium adduct ions (⌬m/z ϩ 22). Tandem mass spectrometric anaylsis using infrared multiphoton dissociation (IRMPD) of the two main components exhibited several fragment ions of diagnostic importance due to the cleavage of the linkage between lipid A and the core oligosaccharide, the loss of Ara4N units (⌬m/zϪ131), and the loss of phosphate residues (⌬m/zϪ98 and Ϫ80). The MS/MS spectrum (Fig. 5B) obtained from the LPS carrying three Ara4N residues (m/z 1856.8) as parent ion comprises four diacyl species (m/z 1856.8, 1725.7, 1594.7, and 1643.7), three diacyl lipid A fragment ions (m/z 1269.6, 1138.6, and 1007.5) and two core fragments (m/z 455.1 and 587.2), which differ from one another by ⌬m/zϪ131. Furthermore, cleavage of phosphate was observed from diacyl lipid A carrying no Ara4N substitution or one Ara4N residue but not from lipid A carrying two Ara4N residues indicating that both phosphates must have been substituted by Ara4N. MS/MS from LPS carrying only two Ara4N (spectrum not shown) exhibits the same lipid A fragment ions but only one core fragment ion at m/z 455.1 indicating that LPS carries three Ara4N residues, lipid A two, and the core, in non-stoichiometric amounts, one Ara4N residue. These results were further confirmed by capillary skimmer dissociation, which generated the same fragment ions (data not shown).
That the Ara4N linked to the core was exclusively linked to the terminal Ko is shown in Fig. 7. When analyzed by GC-MS after mild methanolysis followed by N-acetylation and permethylation of LPS de-O-Ac , fragment ions at m/z 291 3 259 (ϪMeOH) and 216 3 156 (-HOAc) indicated the presence of Kdo at the reducing end and Ara4NAc at the nonreducing end of the oligosaccharide. The two fragment ions at m/z 522 and 597 were in agreement with Ara4NAc 3 Ko and Ko 3 Kdo moieties of the trisaccharide, respectively. The same fragmentation pattern was previously described by Isshiki et al. (29). Chemical ionization MS revealed a pseudomolecular ion (MϩNH 4 ) ϩ at m/z 847, which was in accordance with the molecular mass of the structure depicted in Fig. 7 (data not  shown).
Mass spectrometric analyses of the core oligosaccharides obtained after reduction, dephosphorylation and de-N-acylation of LPS de-O-Ac , and separation by high performance anion exchange chromatography revealed three fractions with molecular masses of [M-H] Ϫ1 at m/z 781.27, 797.27, and 928.33, representing a glucosamine-glucosaminol backbone substituted with a Kdo disaccharide, a Ko-Kdo disaccharide and an Ara4N-Ko-Kdo trisaccharide, respectively (data not shown).

Infection of Human Macrophages with B. cepacia Wild Type and Mutant
Ko2b-The capability of B. cepacia mutant Ko2b to invade human monocyte-derived macrophages was investigated in comparison to that of the wild type strain. Since both strains differ only in the architecture of their LPS we wanted to find out whether a deep-rough mutant was still able to invade macrophages and to survive at comparable rates. Human macrophages were infected and plated after 1 and 3 h of incubation post-infection (see Table II). The rate of infection, presented as the percentage of the bacterial inoculum that was detected in macrophage lysates after 1 h of infection, varied between 0.7 and 1.3% for the wild type and between 1.0 and 4.9% for the mutant. Furthermore, the intracellular survival of B. cepacia, depicted as percentage of bacteria that survived 2 h of extracellular antibiotic treatment, was between 7.5 and 46% and between 6.3 and 33% for the wild type and the mutant, respectively. The relatively large scatter regarding the survival rates was most likely due to the fact that no macrophage cell line was used but primary cells that were derived from isolated monocytes of individual blood donors. It is known that human individuals usually show a broad range of reaction profiles when tested in biological assays; however, data obtained from the same donor depict a high degree of consistency. From a total of four experiments our results clearly showed that the truncation of the LPS did not impair the mutant in its capacity to invade macrophages or to survive inside the macrophages when compared with the wild type strain. Thus, no major difference could be pointed out between the behavior of the wild type and the mutant B. cepacia strain.
TNF-␣ Release of Human Macrophages in Response to LPS of B. cepacia-To examine the potential of isolated wild type and mutant LPS to activate macrophages, the induction of TNF-␣ was determined. Isolated LPS from S. enterica serovar Friedenau served as control (Fig. 8). The results demonstrated that LPS of B. cepacia is a potent inductor of TNF-␣, which is in accordance with data obtained from other groups (19,23,26). In comparison to LPS from S. enterica the preparation of B. cepacia mutant Ko2b is somewhat less active; however, saturation of TNF-␣ induction is reached at a concentration of 10 ng of LPS/ml. The fact that the preparation of B. cepacia wild type LPS is about 10-fold less active than Ko2b LPS can be explained by the large difference in the molecular weight of both molecules. When fatty acid contents of similar amounts of LPS were determined, B. cepacia wild type LPS contained only one-tenth of the amount obtained from mutant Ko2b LPS (data not shown). This significant difference in molecular weight should make up the difference found in TNF-␣ induction as seen in Fig. 8. Our results indicate that the truncated LPS of the deep-rough mutant Ko2b can induce TNF-␣ release in human macrophages to a similar extent as the wild type LPS from B. cepacia.

DISCUSSION
The main chain of the inner core region in the LPS of B. cepacia consists of a trisaccharide with two heptose and one Kdo residues. However, the substitution at position 4 of the Kdo is a Ko residue and not a second Kdo residue as in Enterobacteriaceae, nor a phosphate group as in Vibrio, Haemophilus, or Bordetella. This sugar has been found in other species as well (30,34,63), but only in B. cepacia ATCC 25416 Ko was found to replace Kdo in stoichiometric amounts. So far it is not known whether Ko is the product of a separate biosynthetic pathway and is transferred by a specific Ko transferase or whether it is generated by oxidation of free Kdo, CMP-Kdo, or after incorporation into the LPS. To answer these questions, it was necessary to determine first whether the Kdo transferase of B. cepacia is a mono-or bifunctional enzyme, since a monofunctional Kdo transferase would strongly support the incorporation of Ko by a specific Ko transferase. Cell-free extracts, prepared from B. cepacia ATCC 25416, were used as a source for Kdo transferase and tested in in vitro experiments. These extracts exhibited a bifunctional action, generating a disaccharide linked to the lipid A acceptor. However, it was not possible to exclude the presence of a putative Ko transferase in these assays and because it is not known whether the antibodies used were able to distinguish between Ko and Kdo in the reaction product. To investigate the Kdo transferase in detail the enzyme was cloned and expressed in the Gram-positive host C. glutamicum. The cloned enzyme displayed a bifunctional activity when Kdo was included in the assay. Interestingly, it was not possible to clone waaA from a gene library of B. cepacia in analogy to other Kdo transferases (39,64,65) in E. coli   strain CJB26, harboring an intact copy of the waaA gene on a temperature-sensitive plasmid (64). For the monofunctional Kdo transferase from H. influenzae it was shown that the enzyme supported growth of E. coli CJB26 (66). However, in another system a viable mutant could only be generated when the Kdo kinase was also present (50). One explanation why the Kdo transferase from B. cepacia was not able to support growth could be the large phylogenetic difference between both species; E. coli belongs to the ␥-group of proteobacteria with a G ϩ C content of 52% and B. cepacia belongs to the ␤-group with a G ϩ C content of 64%. The waaA gene has a G ϩ C content of 73%.
To study the biosynthesis of the inner core region more closely it was necessary to generate a genetically defined deeprough mutant of B. cepacia. We decided to do so by inactivation of heptosyltransferase I. The waaC gene was cloned from a gene library by in vivo complementation and tested. It displayed a monofunctional activity in vivo and in vitro, comparable to all other WaaC enzymes investigated so far (39,40,43,(67)(68)(69). Thus, the heptosyltransferase of B. cepacia acted as well on acceptors containing Ko or Kdo, indicating that the enzyme is not able to distinguish between Kdo and Ko in the acceptor molecule. By inserting a kanamycin resistance cassette into waaC the gene was inactivated in vitro and introduced into B. cepacia where the plasmid was integrated into the chromosome by homologous recombination. The instable, merodiploid intermediate had to undergo a second recombination event, and thus the active copy of waaC and additional vector sequences were excised, leaving the inactivated copy of heptosyltransferase I on the chromosome. The resulting mutant Ko2b depicted a deep-rough LPS phenotype in silverstained gels (Fig. 4) and could be complemented to express complete S-form LPS by transformation with a plasmid harboring waaC from B. cepacia (data not shown).
We constructed a deep-rough mutant in order to obtain an LPS with the lowest complexity to facilitate the structural analysis, which indicated that the mutant also contains Ko as a substituent for Kdo. Therefore, it is concluded that the incorporation of Ko is achieved before the first heptose residue is added to the LPS. From earlier investigations on B. cepacia LPS (70) it was known that only minor amounts of Kdo were detectable and that at least one phosphate group of the lipid A was substituted with Ara4N. Our data showed that both phosphate groups are substituted with Ara4N, confirming assumptions by Helander et al. (71) that in B. cepacia also the glycosidically linked phosphate is substituted with Ara4N. Further structural investigations revealed the presence of the trisaccharide Ara4N (1 3 8) Ko (2 3 4) Kdo in the core region of strain ATCC 25416; however, the substitution of Ko with Ara4N was estimated to about 25% (29). We show here that the trisaccharide Ara4N-Ko-Kdo is also present in the deep-rough mutant Ko2b. In addition, an unsubstituted Ko-Kdo disaccharide was found as well as minor amounts of a Kdo disaccharide. For unknown reasons the latter appears in most but not all preparations investigated and may depend upon the growth conditions. The major fraction contains two Ara4N residues in the lipid A moiety, whereas about 40% carry a third Ara4N residue located at the Ko (Fig. 5). The higher degree of substitution in the deep-rough mutant in comparison to the wild type might enhance the stability of the LPS in the absence of an outer core region and O-antigen.
Attachment of Ara4N residues to the phosphate groups of lipid A leads to polymyxin resistance in a number of bacteria (72)(73)(74)(75). Recently, the enzyme responsible for the transfer of Ara4N to both phosphate groups of lipid A has been identified in S. enterica and E. coli and was termed ArnT (76). A homolog of the arnT gene could also be identified in the genome of B. cepacia; however, it remains to be elucidated whether the same transferase incorporates Ara4N also into the inner core region. Our results showed, that in B. cepacia the substitution with Ara4N always occurs at the outer sugar, mostly the Ko. In contrast, structural investigations of LPS on various serotypes of the genus Proteus showed a substitution of the inner Kdo residue with Ara4N (77). It will be interesting to investigate and compare all those Ara4N transferases for the substrate specificity they need to accomplish their different tasks. It also remains to be elucidated how Ko is incorporated into the LPS. This becomes even more challenging since Ko was found to be a component of the LPS more often in a variety of bacteria (30,34,63). Whether these organisms share a specific Ko transferase and a synthesis pathway to generate Ko or they possess machinery to transform Kdo into Ko will be the subject of future studies.
The fact that LPS is a pathogenicity factor and might play an important role during an infection with B. cepacia led us to compare the wild type strain ATCC 25416 and the mutant Ko2b in biological test systems. First, the capacity to infect human monocyte-derived macrophages was analyzed. It turned out that the mutant showed infection rates comparable to the wild type (Table II). Thus, the truncation of the LPS does not seem to impair the organism to interact with macrophages and infect them. However, growth rates were slightly lower at 37°C compared with the wild type. In a second set of experiments LPS preparations from both strains were tested for their ability to induce TNF-␣ release in human macrophages (Fig. 8). It is obvious that LPS from B. cepacia is a potent inducer of TNF-␣, although the control LPS of S. enterica is approximately two orders of magnitude more active. If corrected for the mass difference between wild type and mutant Ko2b LPS, there is no significant variation in the biological activity of the LPS of both strains. Considering these results, mutant Ko2b appears to be suited for further biological testing. However, whether the introduced mutation impairs its capacity to survive in an in vivo model has to be investigated.