Biosynthesis and Structure of the Burkholderia cenocepacia K56-2 Lipopolysaccharide Core Oligosaccharide

Burkholderia cenocepacia is an opportunistic pathogen that displays a remarkably high resistance to antimicrobial peptides. We hypothesize that high resistance to antimicrobial peptides in these bacteria is because of the barrier properties of the outer membrane. Here we report the identification of genes for the biosynthesis of the core oligosaccharide (OS) moiety of the B. cenocepacia lipopolysaccharide. We constructed a panel of isogenic mutants with truncated core OS that facilitated functional gene assignments and the elucidation of the core OS structure in the prototypic strain K56-2. The core OS structure consists of three heptoses in the inner core region, 3-deoxy-d-manno-octulosonic acid, d-glycero-d-talo-octulosonic acid, and 4-amino-4-deoxy-l-arabinose linked to d-glycero-d-talo-octulosonic acid. Also, glucose is linked to heptose I, whereas heptose II carries a second glucose and a terminal heptose, which is the site of attachment of the O antigen. We established that the level of core truncation in the mutants was proportional to their increased in vitro sensitivity to polymyxin B (PmB). Binding assays using fluorescent 5-dimethylaminonaphthalene-1-sulfonyl-labeled PmB demonstrated a correlation between sensitivity and increased binding of PmB to intact cells. Also, the mutant producing a heptoseless core OS did not survive in macrophages as compared with the parental K56-2 strain. Together, our results demonstrate that a complete core OS is required for full PmB resistance in B. cenocepacia and that resistance is due, at least in part, to the ability of B. cenocepacia to prevent binding of the peptide to the bacterial cell envelope.

ionic amphipathic molecules that kill bacteria by membrane permeabilization. In response to a series of environmental conditions such as low magnesium or high iron, bacteria can express modified LPS molecules that result in a less negative surface. This reduces the binding of APs and promotes resistance to these compounds. Previous studies have shown that Burkholderia LPS molecules possess unique properties. For example, Kdo cannot be detected by classic colorimetric methods in LPS from Burkholderia pseudomallei and Burkholderia cepacia, and the covalent linkage between Kdo and lipid A is more resistant to acid hydrolysis than in conventional LPS molecules (7). In B. cepacia, 4-amino-4-deoxy-L-arabinose (L-Ara4N) is bound to the lipid A by a phosphodiester linkage at position 4 of the nonreducing glucosamine (GlcN II) (8) and is also present as a component of the core OS. Also, instead of two Kdo molecules, the B. cepacia core OS has only one Kdo and the unusual Kdo analog, D-glycero-D-talo-octulosonic acid (Ko), which is nonstoichiometrically substituted with L-Ara4N forming a 138 linkage with ␣-Ko (7,9). Although this is also the case for the inner core OS of B. cenocepacia J2315 (10), it is not a common feature for the core OS in all Burkholderia. For example, the inner core of Burkholderia caryophylli consists of two Kdo residues and does not possess L-Ara4N (11).
Burkholderia species, including B. cenocepacia, are intrinsically resistant to human and non-human APs such as these produced by airway epithelial cells (12,13), human ␤-defensin 3 (14), human neutrophil peptides (15), and polymyxin B (PmB) (15,16). The minimum inhibitory concentration determined for some of these peptides in several Burkholderia species is greater than 500 g/ml, which could aid these microorganisms during colonization of the respiratory epithelia (13). It has been proposed that the resistance of B. cepacia to cationic APs stems from ineffective binding to the outer membrane, as a consequence of the low number of phosphate and carboxylate groups in the lipopolysaccharide (17), but a systematic analysis of the molecular basis of AP resistance in B. cenocepacia and other Burkholderia is lacking. We have previously reported that a heptoseless B. cenocepacia mutant (SAL1) is significantly more sensitive than the parental clinical strain K56-2 to APs (15). This mutant has a truncated inner core and lacks the outer core, suggesting that a complete core OS is required for resistance of B. cenocepacia to APs.
Apart from heptoses, the role of other sugar moieties of the B. cenocepacia core OS in AP resistance is not known. In this study, we report the structure of the core OS for B. cenocepacia strain K56-2 and its isogenic mutants XOA3, XOA7, and XOA8, which carry various core OS truncations. The structural analysis, combined with mutagenesis, allowed us to assign function to the majority of the genes involved in core OS biosynthesis and ligation of the O antigen and to establish that the degree of truncation of the core OS correlates with increased binding and bacterial sensitivity to PmB in vitro and reduced bacterial intracellular survival in macrophages.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-Strains and plasmids used in this study are listed in Table 1. B. cenocepacia strain K56-2 was grown at 37°C in LB medium supplemented, as required, with 100 g/ml trimethoprim and 50 g/ml gentamicin. This strain is a clinical isolate that belongs to the same clonal group as the type strain J2315 (18). Escherichia coli strains were grown at 37°C in LB medium supplemented with trimethoprim (50 g/ml), kanamycin (40 g/ml), or chloramphenicol (30 g/ml), as required. Conditional mutants were grown at 37°C in M9 medium supplemented with 5 mg/ml yeast extract and 0.5% (w/v) rhamnose (permissive conditions) or 0.5% (w/v) glucose (nonpermissive conditions) as described previously (19).
Mammalian Cell Lines-Murine macrophage (RAW 264.7; ATCC TIB-71) and human lung epithelial cell lines (A549; CCL-185) were obtained from the American Type Culture Collection (Manassas, VA). Cells were maintained in Dulbecco's Construction of Mutants in B. cenocepacia K56-2-An internal 300-bp fragment near the 5Ј end of the gene targeted for mutagenesis was PCR-amplified and cloned into the XbaI and EcoRI restriction sites of the suicide vector pGP⍀Tp to provide the homology region for recombination. Vector and recombinant plasmids were maintained in E. coli SY327 (araD ⌬lac-pro argE recA56 nalA pir Rif R ). The integration vector has a replication origin that cannot function in B. cenocepacia and also contains transcriptional and translational stops causing polar effects on genes downstream from the insertion point (20). Similarly, the suicide plasmid pGPApTp (21) was used to construct nonpolar insertional mutants, and the plasmid pSC200 was used to construct conditional mutants as described (19). These plasmids were introduced into B. cenocepacia strain K56-2 by triparental mating (22) using E. coli MM294 (endA hsdR pro) containing the helper plasmid pRK2013 (23). Exconjugants were selected on LB agar plates supplemented with 50 g/ml gentamicin (to kill the E. coli donor and helper bacteria) and 100 g/ml trimethoprim for selection of K56-2 exconjugants. Exconjugants were screened for plasmid integration into the chromosome by colony PCR using a primer annealing to a region of the mutagenesis vector and another primer annealing to chromosomal sequences upstream of the insertion site. Integration was confirmed by Southern blot hybridization using a probe corresponding to a DNA fragment spanning the homology region used for recombination. The sequences of the DNA primers used for mutagenesis are listed in supplemental Table S1.
Molecular Cloning for Complementation of LPS Production and PmB Sensitivity-To construct the complementing plasmids pXO13, pXO14, pXO20, pXO21, and pXO22, the corresponding genes were PCR-amplified using ProofStart polymerase (Qiagen Inc., Valencia, CA). The PCR products were digested with XbaI and EcoRI and ligated into pAP20, which was also cut with the same restriction enzymes. Ligation mix- ) competent cells by the calcium chloride method (24), and transformants were plated on LB agar plates supplemented with 30 g/ml chloramphenicol. The correct DNA inserts were verified by colony PCR using primers 1630 (5Ј-CGCAGCAGGGTAGTCGCCCT-3Ј) and 1631 (5Ј-ACTCTC-GCATGGGGAGACCC-3Ј), which anneal to sequences flanking the cloning sites of pAP20. Plasmids from PCR-positive colonies were isolated, initially confirmed by restriction digestion and ultimately by DNA sequencing of the DNA insert. These plasmids were mobilized by conjugation into the corresponding B. cenocepacia mutant strains as described above.
Electrophoretic Analysis of LPS-For electrophoresis analysis, LPS samples were extracted as described previously (25). LPS was resolved by electrophoresis in 16.4% polyacrylamide gels using a Tricine-SDS system and visualized by silver staining. In the case of conditional mutants, the LPS was extracted from bacteria grown in liquid cultures at nonpermissive conditions.
Purification and Compositional Analysis of LPS-Large scale LPS preparations for structural analysis were obtained by the method of Westphal and Jann (26) from 1-liter cultures of B. cenocepacia strains XOA3, XOA7, and XOA8. The quality of the purified LPS samples was confirmed by Tricine-SDS-PAGE as described above. Fractions containing the core OS were obtained by mild acid hydrolysis in sodium acetate buffer, pH 4.4, for 3 h at 100°C as described previously (10). The determination of sugar residues, their absolute configuration, and linkage analysis, as well as the characterization of total fatty acids content and their absolute configuration were all carried out as described previously (10).
NMR Analysis-For structural assignments of OS fractions, one-dimensional and two-dimensional 1 H NMR spectra were recorded on Bruker 600 DRX equipped with a cryoprobe on a solution of 300 l of D 2 O using Shigemi tubes. T-ROESY experiments were recorded using data sets (t 1 ϫ t 2 ) of 4096 ϫ 256 points with mixing times between 100 and 400 ms. The spin lock field was attenuated (ϳ 4000 Hz) with respect to that employed for the hard pulses. No correction for Hartmann-Hahn effects was applied, because T-ROESY effectively removed most of these effects. Double quantum-filtered phasesensitive COSY experiments were performed using data sets of 4096 ϫ 512 points. TOCSY were performed with spin lock times from 20 to 100 ms, using data sets (t 1 ϫ t 2 ) of 4096 ϫ 256 points. In all homonuclear experiments the data matrix was zero-filled in both dimensions to give a matrix of 4000 ϫ 2000 points and was resolution-enhanced in both dimensions by a cosine-bell function before Fourier transformation. Coupling constants were determined on a first order basis from high resolution one-dimensional spectra or by two-dimensional phasesensitive DQF-COSY. HSQC and HMBC experiments were measured in the 1 H-detected mode via single quantum coherence with proton decoupling in the 13 C domain, using data sets of 2048 ϫ 256 points. Experiments were carried out in the phase-sensitive mode. A 60-ms delay was used for the evolution of long range connectivity in the HMBC experiment. In all heteronuclear experiments the data matrix was extended to 2048 ϫ 1024 points using forward linear prediction extrapolation.
MALDI-TOF MS Analysis-MALDI-TOF mass spectra were recorded in the negative and positive polarity in linear mode on a Voyager STR from PerSeptive Biosystems (Framingham, MA) equipped with delayed extraction technology. Ions formed by a pulsed UV laser beam (nitrogen laser, ϭ 337 nm) were accelerated at 24 kV. The mass spectra reported are the result of 256 laser shots. Resolution was about 1500.
Sensitivity to PmB-Bacteria were grown overnight in LB medium or LB medium supplemented with the appropriate antibiotics as required. The next day cultures were diluted to an A 600 of 0.01, and 50 l of this suspension were added to LB medium or LB medium supplemented with 100 g/ml trimethoprim to make a final volume of 5 ml. 500 l of this bacterial suspension were aliquoted into Eppendorf tubes, and 10 l of buffer or PmB (5 mg/ml stock) were added to each tube to reach a final concentration of 100 g/ml. Cells were incubated with PmB at 37°C for 22 h with constant rotation using a Barnstead Thermolyne LABQUAKE (Barnstead International, Dubuque, IA), and the A 600 was recorded. To determine the MIC 50 (concentration of PmB causing 50% reduction in bacterial growth), cells were incubated as described above with PmB at final concentrations of 0, 25, 50, 100, and 200 g/ml.
Binding of Dansyl-PmB to Whole Cells-Dansyl/PmB was prepared from PmB and dansyl/chloride by the method of Schindler and Teuber (27) and quantified by the dinitrophenylation assay (28). For the binding assay bacteria were grown overnight in LB medium or LB medium supplemented with 100 g/ml trimethoprim, as required. The overnight cultures were diluted in LB to an A 600 of 0.1 and grown at 37°C for an additional 4 h. Bacteria were collected by centrifugation, suspended in 1 ml of 5 mM Hepes, pH 7.4, 10 mM sodium azide buffer, and diluted to an A 600 of 0.5. Aliquots (80 l) of this suspension were added to 96-well plates (Microfluor 2 white, flat bottom 96-well microtiter plates, ThermoLabsystems, Franklin, MA) and mixed with 20-l aliquots of Hepes/sodium azide buffer (negative control) or the corresponding solutions of dansyl-PmB prepared by dilution of a 1.7-g/l stock solution in Hepes/sodium azide buffer. Fluorescence was read in a Varian Cary Eclipse fluorescence spectrofluorometer using an excitation wavelength of 340 nm and an emission wavelength of 485 nm as described (29).
Macrophage Infection Assays-Macrophage infections were performed as described previously (30). Briefly, bacterial suspensions were added to RAW 264.7 cells grown on glass coverslips at a multiplicity of infection of 50 and incubated at 37°C in 5% CO 2 for 4 h. When needed, 0.5 mM LysoTracker Red DND-99 (Invitrogen) was added for 1 min prior to visualization. Fluorescence and phase contrast images were acquired using a Qimaging (Burnaby, British Columbia, Canada) cooled, charged-coupled device camera on an Axioscope 2 (Carl Zeiss) microscope. Images were digitally processed using the Northern Eclipse version 6.0 imaging analysis software (Empix Imag-ing, Mississauga, Ontario, Canada). Each experiment was independently repeated at least three times.
Adhesion Assays-Monolayers for adhesion assays were prepared by seeding 7 ϫ 10 4 human lung epithelial A549 cells in DMEM, 10% FBS into a 48-multiwell plate and incubating at 37°C for 20 h in a humidified atmosphere containing 5% CO 2 . Overnight bacterial cultures were washed and resuspended in DMEM, 10% FBS and added to the cells at multiplicity of infection of 50, centrifuged for 2 min at 300 ϫ g, and incubated for 30 min at 4°C. Nonadherent bacteria were removed by rinsing five times with ice-cold phosphate-buffered saline. Cells were lysed with 100 ml of 0.5% sodium deoxycholate. Serial dilutions were performed in LB and plated in duplicated. The percentage of adhesion was calculated as follows: 100 ϫ (number of cell-associated bacteria/initial number of bacteria added). Data were calculated from at least three independent experiments performed in triplicate and are expressed as means Ϯ S.E.
Statistical Analyses-The statistical significance of differences in the data were determined using the one-way analysis of variance test and the Tukey post-test, provided in the Prism GraphPad software version 4.0.

Organization of Core OS Synthesis Gene Loci in B. cenocepacia Strains J2315 and K56-2-
We examined the genome of strain J2315 (31) for genes predicted to encode enzymes for the synthesis of the core OS. Unlike enteric bacteria, the B. cenocepacia core OS genes are not found within a single cluster but rather dispersed into three different locations in chromosome 1 ( Fig. 1). One of these regions is located between nucleotides 2,656,960 and 2,669,740 and contains eight genes named BCAL2402 to BCAL2409. The last two genes, BCAL2409 and BCAL2408, have also been annotated as dnaE (encoding DNA polymerase C or PolC) and msbA (encoding the ATPase transported for lipid A-core oligosaccharide (6)), respectively (31). According to the established norms for nomenclature of bacterial polysaccharide genes (32) and in consultation with the Bacterial Polysaccharide Gene Data Base, we have annotated the remaining genes as wabO, wabP, wabQ, waaL, wabR, and wabS (Fig. 1A). The first gene of the cluster, wabO, is transcribed in the opposite direction relative to the other seven genes. There are 66 nucleotides separating the coding sequences of wabP and wabQ, but this sequence is too short to harbor a promoter region. The coding sequences of wabQ, waaL, and wabR overlap suggesting these three genes may be cotranscribed. In contrast, 260 bp separate wabR from wabS, suggesting this region could accommodate a promoter region. Also, it would appear that the dnaE-msbA-wabS genes are transcribed from a pro-  (34). The location of the IS402 insertion element interrupting the continuity of the wbxE gene in the type ET12 strain J2315 is indicated. However, this gene is functional in the strain K56-2 (34). C, region containing BCAL0967, which was identified as a waaF homolog, which encodes the heptosyltransferase II that is predicted to add the second heptose residue onto the core OS.
moter region upstream of dnaE. This organization suggests a complex transcriptional regulation of these gene loci, which we are currently investigating. Bioinformatic analyses indicate that the products from wabP and wabR are heptosyltransferases, because they have strong similarities to the E. coli WaaQ (heptosyltransferase III) and proteins of the GT1_LPS-heptosyltransferase family, respectively. The products of wabO, wabQ, and wabS have homologies with various glycosyltransferases. BCAL2405 was assigned as waaL based on the characteristics of its predicted product containing the typical topological features observed in O antigen ligases, including 12 predicted transmembrane domains and a large periplasmic loop, which contains highly conserved arginine and histidine residues (33). As expected, the insertional inactivation of waaL results in loss of O antigen surface expression in B. cenocepacia K56-2 (see below).
Additional loci for lipid A-core OS synthesis were previously reported in a 4-gene transcriptional unit next to the O antigen gene cluster (Fig. 1B) (34). They include waaC (encoding a heptosyltransferase I homolog) and waaA (encoding a Kdo transferase) genes. Also, there is a monocystronic gene locus between nucleotides 1,055,340 and 1,056,400 that corresponds to waaF (heptosyltransferase II) homolog (Fig. 1C). As expected, all of these loci were also present with an identical gene organization in the B. cenocepacia strain K56-2, which is from the same clonal lineage as J2315 (18).
Functional Assignment of Core OS Genes-To assign function and establish the order of synthesis of the core OS sugars, we constructed insertional mutants in each of the identified genes as described under "Experimental Procedures." Also, an insertional mutation in the wbxE gene (Fig. 1B), which produces a truncated O antigen (Fig. 2, lane 1), was constructed. The LPS banding pattern profiles of parental and mutant strains were examined by gel electrophoresis and silver stain-ing. Compared with the parental strain K56-2, the mutants XOA6 (wabP::pGP⍀Tp) and XOA9 (wabQ::pGP⍀Tp) did not compromise O antigen synthesis as evidenced by the detection of typical O-ladder-like banding patterns (Fig. 2, lanes 2-4). This suggests that wabP and wabQ gene products mediate lipid A-core OS modifications that are nonessential for the attachment of O antigen polysaccharide chains, or alternatively, their functions can be supplemented by unidentified genes in the B. cenocepacia K56-2 genome. The LPS from the strain XOA7 (waaL::pGP⍀Tp) displayed a lipid A-core OS band that comigrated with the corresponding band in the parental K56-2 LPS (Fig. 2, lanes 2 and 5). This is in agreement with the functional assignment of this gene as encoding the O antigen ligase. In contrast, insertional mutations in wabR, wabS, and wabO, as well as the mutations in waaC and waaF (Fig. 2, lanes 6 -9 and 11) resulted not only in loss of O antigen surface expression but also gave rise to progressive truncations of the lipid A-core OS, as determined by the presence of fast migrating bands of smaller mass than the band produced by XOA7 (waaL::pGP⍀Tp).
The CCB1 (waaC::pGP⍀Tp) mutant produced the shortest lipid A-core OS band (Fig. 2, lanes 9 and 12), which also comigrated with the lipid A-core OS of the SAL1 strain (data not shown). Because SAL1 has a defect in the synthesis of ADPglycero-manno-heptose precursors (15), these results confirm the functional assignment of WaaC as the heptosyltransferase I enzyme. The lipid A-core OS band produced by the waaF::pSC200 mutant XOA19 migrated between the lipid A-core OS from the waaC and wabO mutants, suggesting that waaF encodes the heptosyltransferase II. LPS production in all of the mutants examined except for XOA15 (wabR::pGP⍀Tp) was restored to parental levels after introducing complementing plasmids carrying the corresponding gene under the control of a constitutive promoter (data not shown). In the case of XOA15 (wabR::pGP⍀Tp), the complementing plasmid pXO22 could not be introduced by either conjugation or electroporation, suggesting this mutant has a defect preventing the stable maintenance of the complementing plasmid. Together, the mutagenesis experiments suggest that the lipid A-core OS synthesis in B. cenocepacia K56-2 requires the sequential glycosyltransferase activities of WaaC, WaaF, WabO, WabS, and WabR, although the participation of WabP and WabQ could not be deduced.
Structural Characterization of the Core OS of B. cenocepacia K56-2 LPS-To verify the functional assignments of the core OS genes and to characterize in more detail the structure-function of the core OS in B. cenocepacia, we determined the structure of the core OS produced by strains XOA3 (wbxE::pGP⍀Tp), XOA7 (waaL::pGP⍀Tp), and XOA8 (wabO::pGP⍀Tp). LPS fractions from these strains were extracted by large scale purification and confirmed by SDS-PAGE. Monosaccharide and fatty acid analyses showed the same content in fatty acids (data not shown) but different composition in sugars (Table 2).
To elucidate the primary structure of the core OS of the XOA3 strain, a mild hydrolysis with sodium acetate buffer was performed to split the lipid A from the core OS XOA3 fraction that by gel permeation chromatography yielded one main frac- Core ϩ Oag denotes the extra band in mutant XOA3 that corresponds to lipid A-core ϩ an incomplete O antigen repeat. Core denotes the region of the gel corresponding to the migration of the lipid A-core. The genes in parentheses indicate the gene that is mutated by insertional mutagenesis in the appropriate strain. Lanes 10 -13 correspond to a different gel in which the effect of the waaF mutation on the core OS LPS was tested using the other wild type and mutant strains as controls. The gels were run under identical conditions of voltage, and the migration of the lipid A-core OS band was compared with that of the wild type strain as standard. The relative migration of the mutant lipid A-core OS bands relative to the wild type positive control was reproducible in different runs.
The 1 H NMR spectrum of OS XOA3 is shown in Fig. 3. A combination of homo-and heteronuclear two-dimensional NMR experiments, including double quantum-filtered correlation spectroscopy (DQF-COSY), total correlation spectroscopy (TOCSY), transverse rotating-frame Overhauser enhancement spectroscopy (T-ROESY), 1 H-13 C HSQC, and 1 H-13 C HMBC were performed to assign all the spin systems of OS XOA3 and the monosaccharide sequence. In the anomeric region of the 1 H NMR spectrum (Fig. 3), nine anomeric signals were identified (A-H, Table 3). Furthermore, the signals at 1.92/2.16 ppm were identified as the H-3 methylene of the Kdo residue. The NMR data indicated a mixture of two oligosaccharides with different lengths. Spin systems A, B, D, D, and F (Table 3) were all identified as ␣-heptose residues, as indicated by their 3 J H1,H2 and 3 J H2,H3 coupling constants (below 3 Hz) and by the intraresidual nuclear Overhauser effect (NOE) of H-1 with H-2.  The sugar content was obtained by performing a methanolysis (the sample was methanolyzed and acetylated) followed by analysis via gas chromatography-mass spectrometry. The ring size and the attachment points were determined by methylation analysis, and the absolute configuration of sugar residues has been determined by gas chromatography-mass spectrometry analysis of the acetylated O-(ϩ)-Oct-2yl glycoside derivatives. In all cases, both the fragmentation pattern and a systematic comparison with standards allowed the unequivocal identification and assignment of the sugar content. The results were reproducible over the course of three repeated experiments. X is the first letter defining the name of the mutant strains (e.g. XOA3); x indicates that the sugar is present in the analysis; Ϫ indicates that the sugar is absent in the analysis. AUGUST 7, 2009 • VOLUME 284 • NUMBER 32

JOURNAL OF BIOLOGICAL CHEMISTRY 21743
These data, together with the C-6 13 C chemical shift values (all below 70 ppm) allowed us to identify them as L-glycero-Dmanno heptose, in agreement with the chemical analysis. Spin systems C and E were identified as glucose residues, as indicated by their large ring 3 J H,H coupling constants (above 10 Hz). The strong intra-residue NOE contacts of H-1 E with H-3 and

H and 13 C (italics) NMR chemical shifts (ppm) of sugar residues of the core region of the oligosaccharide OS XOA3
The heptose residues possess a L-glycero-D-manno configuration. ND indicates not detectable. H-5 E together with the 3 J H1,H2 coupling constant (7 Hz) were diagnostics of ␤-configuration, whereas the intra-residue NOE contact of H-1 with H-2 and the 3 J H1,H2 coupling constant (3 Hz) were indicative of ␣-anomeric configuration of residue C.

Structure-Function of B. cenocepacia Core Oligosaccharide
Residue H was recognized as ␣-Rha. Actually, in TOCSY spectrum scalar correlations of the ring protons with methyl signals in the shielded region at 1.14 ppm were visible. The manno configuration of residue H was established from the 3 5.00 ppm). The scalar correlation found in the HMBC spectrum confirmed the OS sequence assigned so far. Thus, methylation analyses, glycosylation shifts, NOE, and HMBC data were all in agreement to indicate an OS XOA3 structure as depicted in Fig. 4A. An alternative glycoform of the ␣-heptose D was found, indicated as residue D (H-1 4.99 ppm; Fig. 3 and Table 3), which was identified as a 7-␣-Hep. The NOE contact of H-7 D with H-1 of ␤-D-QuiN G confirmed the glycosylation by G of residue D at position 7. Residue G was in turn substituted at O-3 by the terminal ␣-rhamnose H, as suggested by the NOE contact of H-1 G with H-3 H. These data validated the structure depicted in Fig. 4B, differing from the one in Fig. 4A by the presence of the additional terminal disaccharide (G-(133)-H-(13 in Fig. 4B).
A MALDI-TOF mass spectrum of the OS XOA3 mixture OS1 confirmed the above structural hypotheses (Fig. 5A). The negative ion mass spectrum showed two major ions at m/z 1565.9 and 1899.0 (⌬m/z ϭ 333). Species at m/z 1565.9 was identified as the octosaccharide built up of four Hep, two Hex, one Kdo, and one Ko residues. Species at m/z 1899.0 (⌬m/z 333) was consistent with the decasaccharide that differed by the presence of the additional dHex-dHexNAc disaccharide.
In addition to the molecular ions in the mass range 2500 -4000 Da, the negative ion MALDI mass spectrum of intact core OS XOA3 revealed ion peaks related to fragments arising from the very labile glycoside bond cleavage between Kdo and the lipid A moiety (Fig. 5B). Thus, at low molecular masses (Fig.  5B), the ion peak at m/z 1548.2 could be assigned to the octosaccharide described above. Ion peaks A-E were derived from the lipid A that was constituted by a mixture of tetra-and pentaacylated species differing by the phosphorylation pattern. Species A at m/z 1444.1 was identified as a tetra-acylated disaccharide backbone carrying in ester linkage one 14:0 (3-OH) chain and in amide linkage two 16:0 (3-OH) chains, one of which, on the GlcN II, was further substituted by a secondary fatty acid, a 14:0 residue. Species C at m/z 1670.8 (⌬m/z ϭ 131) was the corresponding penta-acylated species carrying two ester-linked 14:0 (3-OH) residues. The other species differed for the presence of one or two L-Ara4N residues linked to the phosphate groups. The core OS molecular ions, in the mass range 2500 -4000 Da (Fig. 3), were given by the combination of the peaks of the lipid A and the core region. Interestingly, peaks related to the presence of an additional pentosamine identified via gas chromatography-mass spectrometry as L-Ara4N linked at position O-8 of the Ko residue were also present (see ions at m/z 3254.5 and 3587.1). Peaks at m/z 3455.5 and 3587.1 (⌬m/z ϭ 332 with respect to 3123.5 and 3254.5) were consistent with the presence of the additional terminal dHex-dHexNAc disaccharide. The penta-acylated species was present in a very low amount, and peaks corresponding to penta-acylated lipid A in the core OS region were not detectable. Together, NMR and MS data validated the structure for the lipid A-core OS of strain XOA3 (wbxE::pGP⍀Tp) that is shown in Fig. 4C.
A similar approach was used to elucidate the structure of the core OS from XOA7 (waaL::pGP⍀Tp). The NMR fraction isolated by gel filtration chromatography following the acid treatment was identical to the OS from XOA3 (wbxE::pGP⍀Tp) but lacked the terminal ␤-D-QuiNAc-(133)-␣-L-Rha disaccharide as demonstrated by the comparison of the NMR spectra of XOA3 and XOA7 (supplemental Fig. S1) and by the MALDI-MS spectrum (supplemental Fig. S2). The MALDI-MS spectrum of the intact OS (Fig. 5B) also showed, at low molecular masses, the ion peak derived from the core OS at m/z 1548.4 that was identified as an octosaccharide carrying four heptoses, one Kdo, one Ko, and two hexoses. The lipid A was constituted by a mixture of tetra-and penta-acylated species differing by the phosphorylation pattern (species A-E) and carrying from none to two L-Ara4N residues. The L-Ara4N residue on the Ko residue was present in nonstoichiometric amounts. Therefore, we concluded that the lipid A-core OS of the mutant XOA7 (waaL::pGP⍀Tp) has the structure indicated in Fig. 4D lacking the ␣-L-Rha-(133)-␤-D-QuiNAc disaccharide. This suggests that the addition of this disaccharide requires the activity of the WaaL protein.
In the case of the XOA8 (wabO::pGP⍀Tp), the ion peak derived from the core OS was found at m/z 839.4 in the MALDI-TOF MS spectrum of intact OS (Fig. 5D). This corresponded to a tetrasaccharide with two heptoses, one Kdo and one Ko. The lipid A was constituted by a mixture of tetra-and penta-acylated species differing by the phosphorylation pattern (species A-E) and carrying from none to two L-Ara4N residues. The L-Ara4N residue on the Ko residue was present in nonstoichiometric amount (see MALDI of intact core OS). Based on the results from MALDI and NMR (not shown), the core OS structure of the XOA8 (wabO::pGP⍀Tp) contains a heptose disaccharide attached to the Kdo (Fig. 4E). The structural data, together with the rapid banding pattern of the lipid A-core OS in SDS-PAGE, support the assignment of wabO as the gene encoding the glucosyltransferase responsible for the addition of ␤-D-Glc to HepI (Fig. 4, B and C).
PmB Sensitivity for B. cenocepacia K56-2 LPS Core Mutants-The sensitivity of the different mutants to PmB was determined as described under "Experimental Procedures." Except for mutants XOA9 (wabQ) and XOA6 (wabP), all the mutants with truncations in their lipid A-core OS (Fig. 2) grew significantly less (p Ͻ 0.001) than the parental strain in the presence of 100 g/ml PmB (Fig. 6A). To further investigate the role of the lipid A-core OS in the resistance of B. cenocepacia to PmB, we grew the mutants XOA7 (waaL), XOA8 (wabO), XOA15 (wabR), XOA17 (wabS), and CCB1 (waaC) in the presence of a range of PmB concentrations from 0 to 200 g/ml, and we determined the concentration for which the growth was reduced to 50% (MIC 50 ) when compared with a control culture grown in the absence of PmB. Deeper truncations of the lipid A-core OS correlated with increased sensitivity to PmB in this assay (Fig.  6B). The MIC 50 values for the different mutants ranged from 3.5-to 14.6-fold less than Ͼ512 g/ml, the MIC 50 value of the parental strain K56-2 (data not shown) (15). Furthermore, the  Fig. 3). C-E indicate the structures of the lipid A-core OS for the mutants XOA3, XOA7, and XOA8. The functional assignment for the proteins encoded by the wab genes, waaL ligase, waaC, waaF, and wbxE are based on the structural data and the migration profiles of purified lipid A-core OS in SDS-PAGE. Rectangles indicate the assignment is confirmed. ? indicates proposed functional assignments. were comparable with that reported for the heptoseless mutant SAL1 (15).
To determine whether the increased sensitivity to PmB correlated with increased binding of the mutants to this peptide, we performed binding assays using the fluorescence analog dansyl-PmB. This compound only fluoresces when bound to whole cells or purified LPS. As reported before for B. cepacia (29), B. cenocepacia K56-2 did not bind dansyl-PmB. In contrast, all the mutants bound dansyl-PmB with a degree of binding that was proportional to their sensitivity to PmB (supplemental Fig. S3). Thus, deeper truncations of the core lipid A-OS correlated with greater binding of the bacterial cells to dansyl-PmB.
Loss of O Antigen Production Increases the Adherence of B. cenocepacia to Lung Epithelial Cells-Airway epithelial cells play a key role in maintaining mucosal integrity, and they are the first cells to be challenged by airborne pathogens. We investigated the ability of our panel of mutants in LPS O antigen and lipid A-core OS synthesis to adhere to A549 human lung epithelial cells. Previous work has shown that B. cenocepacia can survive intracellularly in this cell line (38). Our results showed that XOA7 (waaL::pGP⍀Tp), which lost the ability to produce O antigen but produces a complete lipid A-core OS, exhibited 100-fold increased adhesion to A549 cells compared with the parental strain (31 Ϯ 18% adhesion for XOA7 versus 0.23 Ϯ 0.13% for K56-2). Similar results were obtained with all of the other mutants producing truncated core OS (data not shown). We conclude from these results that loss of O antigen production, but not core OS truncations, in B. cenocepacia correlates with increased bacterial adhesion to epithelial cells.
Lipid A-Core OS Heptoseless Mutant CCB1 Is Defective for Intracellular Survival in Macrophages-Previous work in our laboratory demonstrated that B. cenocepacia can survive intracellularly in macrophages (39). Survival occurs in B. cenocepacia-containing vacuoles (BcCVs) that delay normal phagosomal maturation by interfering with vacuolar acidification and the phagolysosomal fusion (40). We investigated the role of O antigen and the core OS in the intracellular behavior of B. cenocepacia. RAW 264.7 macrophages were infected with K56-2, XOA7 (waaL::pGP⍀Tp), and CCB1 (waaC::pGP⍀Tp) strains, and the colocalization of BcCVs with LysoTracker Red was assessed by fluorescence microscopy. LysoTracker Red is an acidotropic dye that preferentially accumulates in lysosomes FIGURE 5. A, negative ion MALDI mass spectrum of core OS product from XOA3 strain. B, negative ion MALDI-TOF mass spectrum of core OS from XOA3 obtained in linear mode. C, negative ion MALDI TOF mass spectrum of core OS from XOA7 obtained in linear mode. D, negative ion MALDI TOF mass spectrum of core OS from XOA8 obtained in linear mode. (41). At 4 h post-infection, less than 30% of the BcCVs containing either K56-2 or XOA7 B. cenocepacia colocalized with Lysotracker Red (Fig. 7, A and C, and data not shown). Similar results were obtained with mutants XOA8 (WabO::pGP⍀Tp), XOA15 (WabR::pGP⍀Tp), and XOA17 (WabS::pGPApTp) (Fig. 7C). In contrast, over 85.4 Ϯ 6.6% of BcCVs in macrophages infected with the heptoseless mutant CCB1 colocalized with the dye at 4 h (Fig. 7, A and C).
Given that intracellular CCB1 bacteria did not prevent phagolysosomal acidification, we also investigated their viability in RAW 264.7 macrophages. For these experiments we performed infections with bacteria expressing the monomeric red fluorescent protein 1 (mRFP1) encoded by pJR1 (Table 1). Using B. cenocepacia cells expressing mRFP1, we have previously demonstrated that intracellular bacteria trafficking into lysosomes rapidly lose cell envelope integrity and are destroyed, resulting in the leakage of mRFP1 into the vacuolar lumen (30,40,42). At 4 h post-infection, 68.9 Ϯ 4.7% of BcCVs containing CCB1(pJR1) bacteria were fluorescently labeled, suggesting that soluble mRFP1 had leaked from the bacterial cytoplasm into the phagosomal lumen (Fig. 7B). In contrast, 11 Ϯ 5% of the BcCVs containing K56-2 (pJR1) showed leakage of mRFP1 (p Ͻ 0.0001; Fig. 7D and data not shown). Furthermore, in contrast to the apparently normal bacterial morphology of intracellular K56-2 (pJR1), internalized CCB1(pJR1) exhibited a variety of abnormal morphologies such as rounding and a highly dense cytoplasm, further suggesting a compromise of their cellular envelope (data not shown). Together, these experiments indicate that the CCB1 mutant loses the ability to survive intracellularly, whereas mutants containing less severe core OS truncations or a complete core OS with no O antigen are not impaired in intracellular survival.

DISCUSSION
We have identified gene loci responsible for the biosynthesis of the core OS moiety in B. cenocepacia K56-2. This allowed us to create a set of core OS-deficient mutants, three of which were used to determine the structure of the core OS. The mutant XOA3 has an insertional mutation in the wbxE gene that encodes a glycosyltransferase involved in O antigen synthesis, resulting in the production of lipid A-core OS and a partial O antigen unit (34). This mutation recreates the same LPS phenotype as observed in strain J2315, whose structure has been recently reported (10). In J2315, the spontaneous insertion of the IS402 element in wbxE causes the formation of a lipid A-core OS with a partial O antigen repeat that cannot be polymerized (34). The core OS structures in J2315 and XOA3 strains are identical, except for the presence of ␣-galactose instead of ␣-Glc linked to the outer core branched 3,7-disubstituted Hep. The galactose in the J2315 strain has an additional ␣-glucose at the O-6 position. The mutant XOA7, which has an inactivated waaL gene, lacks the terminal Rha-QuiNAc disaccharide found in the outer core OS of strains XOA3 and J2315 (10). From these data, we conclude that the Rha-QuiNAc disaccharide is a remnant of the interrupted O antigen in these strains. The O antigen in B. cenocepacia K56-2 is synthesized via the ABC export pathway (4,34). This particular mode of O antigen synthesis requires an adaptor sugar bound to undecaprenyl-PP, to which the remainder of the O antigen repeating units become attached (4). Based on our structural information, combined with the mutagenesis data, we conclude that the QuiNAc residue is the adaptor sugar for the O antigen synthesis in B. cenocepacia K56-2. Furthermore, our data support the conclusion that the ␤-D-QuiNAc-(137)-␣-LD-Hep linkage is made by the WaaL O antigen ligase, explaining why the Rha-QuiNAc disaccharide is absent in strain XOA7. The terminal rhamnose in the core OS of the XOA3 mutant, is likely the first sugar of the repeating O unit, which we have previously established as a Rha-GalNAc-GalNAc trisaccharide (34), but which cannot be completed because of the mutation in the WbxE glycosyltransferase. Therefore, our data also suggest that WbxE encodes a GalNAc transferase. Current work in our laboratories is under way to resolve the complete biosynthesis pathway of the O antigen component of the B. cenocepacia LPS.
The structure of the core OS in the XOA8 strain revealed a major truncation, consistent with the migration pattern of the LPS in SDS-PAGE. The mutated wabO gene in XOA, encodes a putative glycosyltransferase, and based on the elucidated structure and the short lipid A-core OS band produced by the mutant strain, we predict that WabO protein is the glucosyltransferase responsible for the glucosylation of HepI (Fig. 4C). FIGURE 6. A, growth of mutants with truncated LPS in the presence of 100 g/ml PmB as compared with the wild type K56-2. Graph shows the % relative growth of the different strains in the presence of 100 g/ml PmB as compared with the growth without PmB. Error bars correspond to S.E. for three independent experiments done in triplicate. ***, denotes p Ͻ 0.001 for the pairwise comparisons of the % relative growth of each mutant versus the % growth of the parental strain, as determined by the Tukey post-test. B, dosedependent growth for mutants with truncated LPS under increasing concentrations of PmB ranging from 25 to 100 g/ml as compared with the growth without PmB. Error bars correspond to S.E. for three independent experiments done in triplicate.
we have determined for the B. cenocepacia K56-2 heptoseless mutant SAL1 6 (15). Thus, the inner core OS is highly conserved in Burkholderia species.
The panel of isogenic strains with gradual truncations in their lipid A-core OS allowed us to investigate the relationship between LPS and the extraordinary resistance of B. cenocepacia against cationic APs. Although, as it was shown previously with B. cepacia (49), PmB binds to P. aeruginosa much better than to B. cenocepacia K56-2, progressive truncation of the lipid A-core OS leads to increased PmB binding. These findings support the notion that the cell envelope of B. cenocepacia has unusual characteristics that enable it to act as a barrier against APs. Although significantly more sensitive than the wild type strain to PmB, our core mutants are still much more resistant to PmB than other organisms such as Salmonella and E. coli with intact core OS. Thus B. cenocepacia must possess additional mechanisms that make these bacteria extremely resistant to APs.
Taking advantage of the set of isogenic mutants in B. cenocepacia K56-2 that range from the formation of a full-length LPS O antigen (parental strain) to a mutant producing heptoseless lipid A-core OS (CCB1), we also investigated the biological role of LPS in adhesion to epithelial cells and intracellular survival in macrophages. Our results demonstrated that O antigen production by B. cenocepacia prevents bacterial adhesion to epithelial cells. This suggests that the O antigen in these bacteria masks bacterial surface molecules that can interact with epithelial cell receptors, or alternatively, the exposed core OS residues are themselves ligands for binding. We considered the latter hypothesis less likely given that all the core OS mutants with progressive truncations showed increased adhesion, suggesting that no specific sugar residue is required for adhesion, in contrast to recent observations in other bacteria (50). It has been previously shown that Bcc isolates can survive intracellularly within amoebae (51), respiratory epithelial cells (38), and macrophages (39,40). Others have reported that the LPS O antigen plays an essential role in internalization and survival of the related bacterium B. pseudomallei in macrophages (52). Our data investigating the ability of the various mutants with defects in lipid A-core OS production to survive intracellularly in macrophages revealed that only the heptoseless mutant CCB1 is impaired for survival. These results are somewhat surprising and indicate that the ability of B. cenocepacia to survive in macrophages does not correlate with the level of truncation of the core OS. B. cenocepacia can resist oxidative (53,54) and nonoxidative (55) intracellular killing mechanisms, and the latter mainly depend on APs. Therefore, the intracellular survival of the other mutants with core OS truncations, despite their increased sensitivity to PmB in vitro, suggest that either the AP concentration in BcCVs is not enough to compromise the viability of these mutants or other factors are involved. It is possible that the ability of B. cenocepacia to survive intracellularly is highly dependent on the stability of the outer membrane cell envelope, which may be only seriously perturbed in the presence of a drastically truncated lipid A-core OS. This is in agree-ment with other observations indicating that B. cenocepacia heptoseless mutants have defects in motility and increased permeability to other hydrophobic compounds in addition to antimicrobial peptides (15). 6 In conclusion, we have identified the genes involved in the biosynthesis of the core OS in B. cenocepacia, performed the structural analysis of the core OS, and assigned function to most of the genes of the core OS loci. We also demonstrated that progressive truncations of core OS are associated with a dramatic reduction in the resistance to PmB, which inversely correlates with increasing binding of this peptide to the bacterial cell envelope of the mutant strains. Finally, we also show that the majority of the core OS is expendable for intracellular survival of B. cenocepacia in macrophages, whereas the O antigen contributes to prevent bacterial adhesion to epithelial cells. Further investigations are underway in our laboratories to better elucidate the characteristic of the outer membrane and the LPS molecules that contribute to the extraordinary resistance of B. cenocepacia to a wide range of antimicrobial molecules, including APs.