Structural Studies of Lipopolysaccharide-defective Mutants from Brucella melitensis Identify a Core Oligosaccharide Critical in Virulence*

The structures of the lipooligosaccharides from Brucella melitensis mutants affected in the WbkD and ManBcore proteins have been fully characterized using NMR spectroscopy. The results revealed that disruption of wbkD gives rise to a rough lipopolysaccharide (R-LPS) with a complete core structure (β-d-Glcp-(1→4)-α-Kdop-(2→4)[β-d-GlcpN-(1→6)-β-d-GlcpN-(1→4)[β-d-GlcpN-(1→6)]-β-d-GlcpN-(1→3)-α-d-Manp-(1→5)]-α-Kdop-(2→6)-β-d-GlcpN3N4P-(1→6)-α-d-GlcpN3N1P), in addition to components lacking one of the terminal β-d-GlcpN and/or the β-d-Glcp residues (48 and 17%, respectively). These structures were identical to those of the R-LPS from B. melitensis EP, a strain simultaneously expressing both smooth and R-LPS, also studied herein. In contrast, disruption of manBcore gives rise to a deep-rough pentasaccharide core (β-d-Glcp-(1→4)-α-Kdop-(2→4)-α-Kdop-(2→6)-β-d-GlcpN3N4P-(1→6)-α-d-GlcpN3N1P) as the major component (63%), as well as a minor tetrasaccharide component lacking the terminal β-d-Glcp residue (37%). These results are in agreement with the predicted functions of the WbkD (glycosyltransferase involved in the biosynthesis of the O-antigen) and ManBcore proteins (phosphomannomutase involved in the biosynthesis of a mannosyl precursor needed for the biosynthesis of the core and O-antigen). We also report that deletion of B. melitensis wadC removes the core oligosaccharide branch not linked to the O-antigen causing an increase in overall negative charge of the remaining LPS inner section. This is in agreement with the mannosyltransferase role predicted for WadC and the lack of GlcpN residues in the defective core oligosaccharide. Despite carrying the O-antigen essential in B. melitensis virulence, the core deficiency in the wadC mutant structure resulted in a more efficient detection by innate immunity and attenuation, proving the role of the β-d-GlcpN-(1→6)-β-d-GlcpN-(1→4)[β-d-GlcpN-(1→6)]-β-d-GlcpN-(1→3)-α-d-Manp-(1→5) structure in virulence.


GlcpN-(133)-␣-D-Manp-(135) structure in virulence.
Brucella is a genus of Gram-negative facultative intracellular coccobacilli that causes brucellosis in humans and animals. Although its true extent is not known and the disease is largely unreported (1), it has been estimated that there are half a million new cases every year, most of them located in the poorest rural areas of the world (2). Humans can acquire brucellosis by ingestion of unpasteurized milk from infected animals or by contact with their secretions, but generally they are not themselves a source of contagion. The species within this genus were originally differentiated on the basis of their primary host preferences, with Brucella melitensis (sheep and goat), Brucella suis (pig), and Brucella abortus (cattle) being the most common in domestic livestock. Epidemiological evidence shows that, among them, B. melitensis is the most virulent species for humans (3).
The lipopolysaccharides (LPS) of Gram-negative bacteria are exposed on the cell surface, and three different regions with different chemical and biological properties can be identified as follows: the lipid A, the core oligosaccharides, and the polysaccharide, which in most cases represents the O-specific polysaccharide (O-PS, 2 O-antigen) (4,5). The LPS of Brucella shows very low endotoxicity, which illustrates poor detection by innate immunity. Thus, it is considered one of the virulence factors that allow the pathogen to escape early detection by the host immune system (6,7). Whereas this characteristic is related to the structure of the lipid A and core oligosaccharide of LPS, the O-PS (8,9) also plays a major role in virulence, because it has been repeatedly observed that mutants lacking the O-PS (i.e. producing a rough (R) type LPS, also termed lipooligosaccharide (LOS)) are attenuated (10). Brucella strains with R-LPS are often caused by spontaneous mutations (11)(12)(13) and can also be due to mutation in genes encoding proteins involved in the biosynthesis of the monosaccharide components of the O-PS, its polymerization or transport, or the core oligosaccharide. In a previous report, the genes involved in the biosynthesis of the LPS of B. melitensis were screened, and several mutants affected in the biosynthesis of both the core oligosaccharide or O-PS were obtained (10). To assign mutations to the biosynthetic pathways, the mutants were classified as R1 (complete core), R2 (defective core), and R3 (deep R core), respectively, according to the decrease in their LPS apparent molecular mass (10). For instance, mutations in wbkD, wadA, and manB core gave rise to R1, R2, and R3 LPS core glycoforms, respectively (10). Based on sequence homology comparisons, wbkD was proposed to code for a putative epimerase/dehydratase involved in the biosynthesis of quinovosamine, the monosaccharide located at the reducing end of the O-PS (10,14,15). In contrast, wadA was proposed to encode a glycosyltransferase involved in the biosynthesis of the core, whereas manB core was proposed to encode a phosphomannomutase involved in the biosynthesis of GDP-mannose, used as a precursor in the synthesis of both the core oligosaccharide and the perosamine residues found in the O-PS (10, 16). Moreover, it was shown in later works that deletion of genes encoding glycosyltransferases WadB and WadC creates severe but uncharacterized defects in the core without affecting the section linked to the O-PS. Remarkably, B. abortus wadC mutants are attenuated despite carrying an intact O-PS (17). An overview of the proposed pathways involved in the biosynthesis of the smooth LPS of B. melitensis is shown in Fig. 1.
In this study, we describe the structural elucidation of the LOS from B. melitensis strain EP, a spontaneous mutant producing O-PS and increased amounts of LOS (18), Bm_wbkD and Bm_manB core , two mutants in wbkD and manB core , disrupted in the corresponding biosynthetic steps that produce R1 and R3 LOS, respectively (Table 1 and Fig. 1). We also report that deletion of B. melitensis wadC removes the core oligosaccharide section not linked to the O-PS. We furthermore show that, despite carrying the O-PS essential in virulence, the lack of this core section results in a marked change in bacterial surface physicochemical properties, a more efficient detection by innate immunity, and attenuation in both cellular and animal models.

Experimental Procedures
Bacterial Strains-B. melitensis 16M (biovar 1 reference strain) and B. melitensis H38 are wild type virulent strains com- . The steps leading to N-formylperosamine synthesis and to its polymerization are indicated in blue, and those leading to bactoprenol priming for N-formylperosamine polymerization are depicted in green. Subsequently, the O-PS is translocated to the periplasm by the Wzm/Wzt ABC transporter (also in blue) and ligated to the core oligosaccharide that results from the pathways marked in red. The LPS phenotypes obtained by disrupting the different steps (indicated by black triangles) are annotated (R1, R2, or R3). A white-filled triangle (GDP-mannose pathway) indicates a mutation that does not generate an R phenotype, and gray triangles marked with S (middle) indicate a mutation that, while blocking the synthesis of a core lateral branch, does not prevent O-PS linkage to the core. Enzymes disrupted in this study (WadC, Per, WbkD, and ManB core ) are indicated with an ellipse. monly used in virulence studies that are identical with regard to LPS genes (10). B. melitensis EP (henceforth BmEPR) is a strain that, although virulent and able to synthesize O-polysaccharides, produces comparatively large proportions of R-LPS (18) and is thus suitable to obtain R-LPS in large amounts for chemical analyses. Mutants with defects in LPS genes (Table 1 and Fig. 1) were obtained, characterized, and stored according to the procedures described previously (10,17).
Growth of Bacteria and Isolation of the LOS-Bacteria were propagated in tryptic soy broth either in a Biostat fermentor (BmEPR, Bm_wbkD, and Bm_manB core ) or in 2-liter flasks (Bm_wadC and Bm_wadC_per) on an orbital shaker in a BSL3 facility. After 36 h of incubation, bacteria were inactivated with phenol, harvested by tangential flow filtration, and washed twice with saline (19). For LOS extraction (BmEPR, Bm_wbkD, Bm_manB core , and Bm_wadC_per), bacteria were first acetonedried and then extracted following the phenol/chloroform/ light petroleum method (20), and the overall defect (R1, R2, and R3) in the R-LPS was confirmed by SDS-PAGE and silver staining and absence of reactivity with O-PS specific monoclonal antibodies (10). The LPS of Bm_wadC was obtained from the phenol phase of a water/phenol extract and further purified as described before (21).
Compositional Analyses-Quantitative analyses of sugars, organic bound phosphate, and fatty acids were performed as described previously (22).
Preparation of the Deacylated Oligosaccharides-The LOS from the different B. melitensis mutants were O-deacylated using hydrazine, followed by N-deacylation with hot KOH (23) and purified by high performance anion-exchange chromatography.
Mass Spectrometry-Electrospray ionization high resolution mass spectra (ESI-HR-MS) were recorded in negative ion mode using a MicrOTOF TM mass spectrometer (Bruker Daltonics). Nitrogen was used as the collision gas.
The diffusion-filtered 1 H NMR spectrum of the deacylated LOS from Bm_manB core was recorded on a Bruker Avance 500 MHz spectrometer equipped with a 5-mm Z-gradient (53.0 G⅐cm Ϫ1 ) TCI ( 1 H/ 13 C/ 15 N) CryoProbe using the one-dimensional stimulated spin-echo pulse sequence with bipolar gradients and LED (ledbpgp2s1d) (24). Diffusion-encoded sinusoidal gradients pulses (␦/2) of 1.8 ms and a strength of 70% of the maximum were used; the diffusion time was set to 100 ms. The one-dimensional diffusion-filtered 1 H NMR spectra of the deacylated LOS from BmEPR and Bm_wbkD were recorded on a Bruker Avance III 700 MHz spectrometer equipped with a 5-mm Z-gradient (53.0 G⅐cm Ϫ1 ) TCI ( 1 H/ 13 C/ 15 N) CryoProbe, using the same pulse sequence as described above and diffusion-encoded smoothed square-shape gradient pulses (␦/2) of 1.8 ms and a strength of 50% of the maximum; the diffusion time was set to 120 ms. 1 H and 13 C NMR chemical shifts assignments of the three LOS materials referred to in the preceding paragraph were obtained from experiments recorded at a magnetic field strength of 16.4 tesla. 1 H chemical shift assignments were obtained using 1 H, 1 H TOCSY experiments (25) employing the States-TPPI method, an MLEV-17 spin-lock of 10 kHz, and different mixing times (between 20 and 100 ms). To identify correlations from possible impurities of lower molecular mass than the deacylated LOSs from Bm_manB core , a diffusion-filtered 1 H, 1 H TOCSY experiment ( m ϭ 100 ms) was acquired employing the pulse sequence "ledbpgpml2s2d" from the standard Bruker library; diffusion-encoded smoothed square-shape gradients pulses (␦/2) of 1.8 ms and strength of 70% of the maximum were used and a diffusion time of 150 ms. 13 C NMR chemical shifts were obtained from the respective 13 C spectra, and the assignments were carried out using multiplicity-edited 1 H, 13 C HSQC experiments (26) employing the echo/anti-echo method. Adiabatic pulses (27,28) were used for 13 C inversion (smoothed CHIRP, 20%, 80 kHz, 500 s, Q ϭ 5.0) (29) and refocusing (composite smoothed CHIRP, 80 kHz, 2.0 ms). The 1 H, 13 C-H2BC experiments (30) were recorded with a constant-time delay of 33 ms, and the 1 H, 13 C HSQC TOCSY experiments were acquired using an MLEV-17 spin-lock of 10 kHz employing mixing times ranging from 20 to 100 ms.
For assignments of inter-residue correlations, 1 H, 1 H NOESY and 1 H, 13 C heteronuclear multiple-bond correlation (HMBC) experiments were utilized. The gradient-selected 1 H, 1 H NOESY experiments (31) were recorded with a mixing time of 100 ms, whereas the gradient-selected 1 H, 13 C HMBC experiments (32) were carried out with an evolution time of 65 ms. Band-selective constant-time 1 H, 13 C HMBC experiments (33) with 2-fold low pass J-filters (hmbcctetgpl2nd) were also employed to improve spectral resolution in the anomeric region. The experiments were recorded over a spectral region of 5.4 ϫ 9.0 ppm with 2048 ϫ 256 data points, using an 80-ms delay for evolution of the long range couplings. A selective 13 C excitation pulse (Q3 gaussian cascade) of 2.5 ms was applied at the center of the anomeric region.  31 P-based NMR experiments were obtained on a Bruker Avance III 600 MHz spectrometer equipped with a 5-mm Z-gradient (55.7 G⅐cm Ϫ1 ) inverse TXI ( 1 H/ 13 C/ 31 P) probe. The one-dimensional 1 H-decoupled 31 P NMR spectra were recorded with a spectral width of 396 ppm, and the 31 P chemical shifts assignments were obtained from gradient selected 1 H, 31 P HMBC (32,34) and 1 H, 31 P-hetero TOCSY experiments (35). The 1 H, 31 P HMBC spectra were recorded with an evolution time of 100 ms. The 1 H, 31 P-hetero TOCSY experiments were carried out with mixing times of 23 and 46 ms, using a DIPSI2 mixing sequence set at 5 kHz on both channels or with a mixing time of 92 ms using a DIPSI2 mixing sequence set at 2.5 kHz.
Additionally, NMR experiments selected from those described above were carried out on deacylated tetra-and pentasaccharides from Bm_wadC_per on 0.28 and 0.26 mg, respectively, in D 2 O (0.55 ml) at pD 7, a temperature of 15°C, and a 1 H frequency of 500 MHz.
Potential-The surface charge density was measured as the electrophoretically effective potential ( potential, sm ) as described previously (10). For this, bacteria were grown in tryptic soy broth, inactivated with 0.5% phenol, washed, and resuspended in 1 mM CsCl, 10 mM HEPES (pH 7.2) at an A 600 of 0.2. Measurements were performed at 25°C in a Zetamaster instrument using the PCS 1.27 software (Malvern Instruments Ltd., Malvern, UK).
Virulence Assays-Bone marrow cells were isolated from femurs of a 7-8-week-old C57Bl/6 female mice and differentiated into dendritic cells (BMDCs) as described by Inaba et al. (36). Infections were performed by centrifuging the bacteria onto the differentiated cells (400 ϫ g for 10 min at 4°C; bacteria/cell ratio of 20:1 followed by incubation at 37°C for 30 min under a 5% CO 2 atmosphere). BMDCs were gently washed with medium to remove extracellular bacteria before incubating in a medium supplemented with 100 g⅐ml Ϫ1 gentamicin for 1 h to kill extracellular bacteria. Thereafter, the antibiotic concentration was decreased to 20 g⅐ml Ϫ1 . To monitor Brucella intracellular survival, BMDCs were lysed with 0.1% (v/v) Triton X-100 in H 2 O, and serial dilutions of lysates were rapidly plated onto tryptic soy agar plates to enumerate the colony-forming units (cfu).
Seven-week-old female BALB/c mice (Charles River, Elbeuf, France) were kept in cages with water and food ad libitum and accommodated under BSL3 biosafety containment 2 weeks before and during the experiments in the facilities of the "CIMA" (registration code ES31 2010000132). The animal handling and other procedures were in accordance with the current European (directive 86/609/EEC) and Spanish (RD 53/2013) legislations, supervised by the Animal Welfare Committee of the University of Navarra (CEEA 045/12). Inocula were prepared in sterile 10 mM PBS (pH 6.85), and ϳ5 ϫ 10 4 cfu in 0.1 ml (the exact dose was assessed retrospectively by plating dilutions of the inocula) were administered intraperitoneally to each mouse. For each strain, 10 mice were inoculated, and the number of cfu in spleen was determined at 2 and 8 weeks after inoculation as described previously (37). The individual data were normalized by logarithmic transformation.
Flow Cytometry-To analyze activation and maturation, BMDCs were analyzed for surface expression of classical mat-uration markers at 24 h post-treatment with the different LPS variants. Cells were labeled with fluorochrome-conjugated antibodies specific for mouse CD11c:PE-Cy7 (clone N418), IA-IE:PE (MHC class II clone M5/114.15.2) (PE), CD86:FITC (Clone GL-1), CD40:AlexaFluor 647 (clone 3/23), and CD80: PE-Cy5 (clone 16-10A1), all from BioLegend. Labeled cells were then subjected to multicolor cytometry using an LSR II UV spectrophotometer (BD Biosciences), and the data were analyzed using FlowJo software by first gating on the CD11c ϩ population (100,000 events) prior to quantifying expression of receptors. Cells were stimulated with 10 g⅐ml Ϫ1 of LPS purified from the B. melitensis wild type or Bm_wadC mutant strain or with Escherichia coli LPS (O55:B5) as a positive control.

Compositional Analyses
The analysis of sugars, fatty acids, and organic bound phosphate (P) of the LOS from mutants Bm_manB core , Bm_wbkD, and BmEPR revealed the presence (in nmol⅐mg Ϫ1 LOS) of Glc GlcN3N was not quantified. Mannose was present in trace amounts in the LOS from BmEPR but was lacking in the other two LOSs, because it was not present at all in Bm_manB core and was not released under the acidic hydrolysis conditions used due to its substitution by GlcN (compare structures below).

Mass Spectrometry
HRMS of the LOS from Bm_manB core -The two sets of single negatively charged pseudo-molecular ions observed in the high resolution mass spectrum of the deacylated LOS from Bm_manB core ( Fig. 2A) are consistent with the presence of two different oligosaccharides. The ions at m/z 1099.3 and 937.2 correspond to species formed through the loss of a proton. The former ion is in agreement with the presence of a pentasaccharide composed of one hexose (Hex), two Kdo, two diaminohexoses (HexNN), and two phosphate groups (P), whereas the latter indicate the presence of a tetrasaccharide composed of two Kdo residues, two diaminohexoses, and two phosphate groups (see Table 2). Different singly negatively charged sodium adducts of the pentasaccharide are observed at m/z 1121.3, 1143.2, and 1165.2, whereas those of the tetrasaccharide can be found at m/z 959.2, 981.2, and 1003.2.
HRMS of the LOS from Bm_wbkD-The mass spectrum of the deacylated core oligosaccharide from Bm_wbkD ( Fig. 2B) is also consistent with the presence of two major oligosaccharides, which produce the doubly negatively charged ions at m/z 952.3 and 871.8 through the loss of two protons each. The ion with higher m/z value corresponds to a decasaccharide composed of two Hex, four hexosamines (HexN), two Kdo, two HexNN, and two phosphate groups (see Table 2); the different doubly charged sodium adducts of this oligosaccharide appear at m/z 963.3, 974.3, and 985.3. The ion at m/z 871.8 corresponds to a nonasaccharide composed of two Hex, three HexN, two Kdo, two HexNN, and two phosphate groups; ions from three different sodium adducts of this compound are present at m/z 882.8, 893.7, and 904.7.

NMR Spectroscopy
LOS from Bm_manB core -In the 1 H NMR spectrum, three resonances corresponding to anomeric protons were identified as follows: two doublets at 4.577 and 4.632 ppm ( 3 J H1,H2 ϭ 8.1 and 8.0 Hz, respectively) and one doublet of doublets at 5.430 ppm ( 3 J H1,H2 ϭ 3.2 Hz and 3 J P,H1 ϭ 8.2 Hz). The residues were named A, E, and B, in order of decreasing 1 H chemical shifts (Fig. 3A). Two major resonances were identified in the 31 P NMR spectrum at 2.6 and 3.2 ppm, suggesting the presence of two phosphomonoester groups, one of which is consistent with the splitting of the H1 resonance of residue A. All the protons from H1 to H6 could be traced using 1 Table 2). Note that the peaks observed in A and B correspond to singly and doubly negatively charged ion species, respectively.

TABLE 2 HR-MS data (negative ion mode) and proposed compositions of the LPS from Bm_manB core and Bm_wbkD
Summary of the diagnostic pseudo-molecular ions observed in the high resolution mass spectra of the deacylated LOS from Bm_manB core and Bm_wbkD and the proposed composition of the two main components present in each sample. The components are abbreviated as follow: Hex (hexoses), HexN (hexosamine), HexNN (diaminohexose), and P (phosphate group).

Proposed composition (exact molecular mass) Observed
Calculated Annotation

Molecular formula
Bm_manB core ( Fig. 2A ments, indicating that these residues have the gluco-configuration. 13 C NMR chemical shifts were assigned using multiplicityedited 1 H, 13 C HSQC (Fig. 3B) and 1 H, 13 C H2BC experiments. Both C2 and C3 resonances of residues A and B were found in the region between 54 and 59 ppm (Fig. 3B, right middle), which indicates that these are nitrogen-bearing carbons. 1 J C1,H1 couplings constants were extracted from a coupled 1 H, 13 C HSQC spectrum and revealed that residues B and E are ␤-linked ( 1 J C1,H1 ϭ 164 and 163 Hz, respectively), whereas residue A is In the 13 C NMR spectrum, three anomeric carbons were identified at 100.10, 100.15, and 100.46 ppm and attributed to C2 resonances in three different populations of Kdo residues (D*, D, and C, respectively).
In the 1 H, 13 C HMBC spectrum, the C2 carbon of each Kdo residue could be correlated to their respective H3 protons via two-bond heteronuclear correlations (Fig. 3C), and the H3 resonances were used as starting points for the assignments of the respective spin systems. The multiplicity-edited 1 H, 13 C HSQC spectrum (Fig. 3B, right top) showed cross-peaks of different relative intensities for each of the Kdo residues, which was confirmed by integration of the H3a resonances in the 1 H NMR spectrum (1.0 H, 0.6 H, and 0.4 H, in residues C, D, and D*, respectively) and indicated the presence of two different oligosaccharide components. 1 H and 13 C chemical shifts assignments are compiled in Table 3. The ␣-configuration of the anomeric center of the Kdo residues was established by comparison of the 13 C chemical shifts of the D*-C moiety with those of the same moiety found in a tetrasaccharide with similar structure, previously reported in the literature (38). Inter-residue correlations were observed in the 1 H, 13 C HMBC spectrum from all the anomeric carbons to the respective protons at the substitution positions (Fig. 3C), as well as from the anomeric protons (in residues E and B) to the respective glycosyloxylated carbons in the next residue (Table 3). These results are in agreement with a major (ϳ63%) pentasaccharide component (E-D-C-B-A) in which a Kdo residue (D) is 4-substituted with a Glc residue (E) and a minor (ϳ37%) tetrasaccharide component (D*-C-B-A) in which the Glc residue is absent, and consequently the corresponding Kdo residue (D*) is non-substituted. The percentages of each component were determined by integration of the C3/H3b cross-peaks of residues D and D* in the 1 H, 13 C HSQC spectrum. The substitution positions of the phosphomonoester groups were determined using 1 H, 31 P HMBC and 1 H, 31 Phetero TOCSY experiments and were found at C1 in residue A ( 3 J P,H1 ϭ 8.2 Hz, 2 J P,C1 ϭ 5.4 Hz, and 3 J P,C2 ϭ 6.1 Hz) and C4 in residue B (Fig. 3D). Three additional resonances of minor intensities were found in the 31 P NMR spectrum at ␦ P 4.0 (t, 3 (39,40). Diffusion-filtered experiments ( 1 H NMR and 1 H, 1 H TOCSY spectra) were employed to confirm that these compounds had lower masses than the deacylated LOS material of Bm_manB core and thus were not linked to LPS. The genome of B. melitensis contains an LptA homologue putatively involved in the transference of ethanolamine to an unknown position of the core-lipid A, which could account for the 2-aminoethyl phosphate group (16). This may have been substituting the O4 position of residue B (either as phosphodiester or diphosphodiester), but under the basic conditions used to prepare the oligosaccharides, this was probably lost (41)(42)(43). Additional spin systems of lower intensity, similar to A and B but with slightly different chemical shifts, were also identified in the 1 H, 1 H TOCSY spectra (A: 5.406, 2.848, 3.106, 3.480, and 4.093, and B: 4.516, 2.762, 3.152, 3.830, and 3.737, from H1 to H5, respectively), as well as resonances in the 1 H NMR at 1.286 (CH 2 ) and 0.869 (CH 3 ) ppm, attributed to residual acyl groups. The structure of the penta-and tetrasaccharide components of the LOS from Bm_manB core are shown in Fig. 4A and are consistent with the structure of the lipid A reported previously for B. abortus (44).
LOS from BmEPR-The 1 H NMR spectrum of the deacylated LOS (Fig. 5A, top) revealed a complex material, with several signals of different intensities in the anomeric region (between 5.42 and 4.43 ppm). Eight distinctive spin systems originating from the anomeric protons were identified in the 1 H, 1 H TOCSY spectra, and the different sugar residues were denoted A, F, G, E, B, J, H, and I, in order of decreasing 1 H chemical shifts. In addition, a minor spin system similar to residue F but with slightly different chemical shifts was also identified and denoted F*. For residues A, G, E, B, J, H, and I, all protons from H1 to H6 could be identified in the 1 H, 1 H TOCSY spectrum recorded with the longest mixing time (Fig. 5B), indicating that these monosaccharide components have the gluco-configuration. The distinctive downfield chemical shift of H2 in residues F and F* suggested that these are Man residues. The anomeric proton of residue A is a doublet of doublets ( 3 J H1,H2 ϭ 3.2 Hz and 3 J P,H1 ϭ 8.3 Hz) indicating that this could be the ␣-D-GlcN3N1P residue at the reducing end of the lipid A moiety. 13 C chemical shifts were assigned using multiplicity-edited 1 H, 13 C HSQC (Fig. 5C, left), 1 H, 13 C-H2BC, 1 H, 13 C HMBC, and 1 H, 13 C HSQC TOCSY experiments. 1 H and 13 C chemical shifts assignments are compiled in Table 4. Both C2 and C3 resonances of residues A and B were found in the region between 54 and 59 ppm, confirming that these are the two GlcN3N residues of the lipid A. The C2 resonances of residues G, J, H, and I were also found in the region of the nitrogen-bearing carbons, indicating that these are GlcN residues (Fig. 5C, left top); thus, residue E is Glc. 1 J C1,H1 couplings constants were extracted from a coupled 1 H, 13 C HSQC spectrum and revealed that residues A and F are ␣-linked ( 1 J C1,H1 ϭ 173-174 Hz), whereas residues B, E and G-J are ␤-linked ( 1 J C1,H1 ϭ 162-165 Hz). The spin systems of the Kdo residues were analyzed as described before, and four distinctive populations were found and denoted D and C

TABLE 3 NMR chemical shift assignments of the LOS from Bm_manB core and Bm_wadC_per and inter-residue correlations from 1 H, 13 C HMBC NMR spectra
The 1 H, 13 C, and 31 P NMR chemical shifts (ppm) of the LOS from Bm_manB core were determined at 25°C and pD 8, whereas the 1 H and 13 C NMR chemical shifts of the tetra-and pentasaccharide from Bm_wadC_per (OS1 and OS2, respectively) were obtained at 15°C and pD 7. 3 J H1,H2 values are given in Hz in parentheses and 1 J C1,H1 values in braces. For residue A, 3 J P,H1 ϭ 8.2 Hz, 2 J P,C1 ϭ 5.4 Hz, and 3 J P,C2 ϭ 6.1 Hz. ND means not determined. (major populations) and D* and C* (minor populations). Interresidue correlations from anomeric carbons and protons were extracted from both regular and band-selective 1 H, 13 C HMBC spectra and used for determination of the sequence of sugar residues in the LOS ( Table 4). The same two basic structures as for the LOS from Bm_manB core could be identified (i.e.

E-D-C-B-A and D*-C-B-A), with the only difference being a branched oligosaccharide moiety composed of GlcN residues (I-H-[J]G),
in the case of the major component that extends from a Man residue (F) linked to position 5 of the Kdo residue C. Thus, the major decasaccharide and nonasaccharide components of the LOS (ϳ66 and ϳ34%, respectively, determined by integration of the characteristic C3/H3b cross-peaks of residues D and D* in the 1 H, 13

C HSQC spectrum) have the following sequence of sugar residues: E-D-[I-H-[J]G-F]C-B-A and D*-[I-H-[J]G*-F*]C*-B-A.
The substitution positions of the phosphomonoester groups were studied as described above. The two major resonances in the 31 P NMR spectrum (2.5 and 3.1 ppm) were attributed to ␣-D-GlcpN3N1P (residue A) and ␣-D-GlcpN3N4P (residue B); likewise, the minor resonances found at ␦ P 4.0, 4.1, and 4.4 were attributed to free ethanolamine phosphate, glycerol 2-phosphate, and glycerol 3-phosphate. Additional crosspeaks of lower intensity (ϳ19%) were also found in the multiplicity edited 1 H, 13 C HSQC spectrum and attributed to an oligosaccharide similar to the decasaccharide described above, but without the GlcN residue I linked to residue H. In this case, 1 H and 13 C chemical shift assignments were carried out by comparison with the chemical shifts of the same oligosaccharide found in the LOS from Bm_wbkD (ϳ48%) (see below), and the corresponding sugar residues were denoted with a double asterisk. The C4/H4 cross-peaks of residues G and G** in the 1 H, 13 C HSQC spectrum were used to estimate the percentage of these oligosaccharides in the sample. The structures of the deca-and nonasaccharide components of the LOS from BmEPR are shown in Fig. 4B.
LOS from Bm_wbkD-The 1 H NMR spectrum of the deacylated LOS from Bm_wbkD (Fig. 5A, bottom) appears quite similar to that of BmEPR (Fig. 5A, top), with the only difference being the relative intensities displayed by some of the resonances. 1 H and 13 C chemical shifts assignments were carried out as described previously; the same oligosaccharide structures as above were identified,

E-D-[I-H-[J]G-F]C-B-A and D*-[I-H-[J]G*-F*]C*-B-A, and particularly the nonasaccharide E-D-[H**-[J]G**-F**]C-B-A,
that was fully characterized due to the higher relative concentration of the component in the sample (ϳ48%, instead of 19% observed in the case of BmEPR). Comparison of the multiplicity-edited 1 H, 13 C HSQC spectrum of the deacylated LOS from BmEPR (Fig. 5C, bottom left) and  that of Bm_wbkD (Fig. 5C, bottom right) allowed for the identification of a conspicuous resonance at ␦ H /␦ C 4.440/103.48 (H1/C1 in residue H**) that is noticeably stronger in the latter spectrum. The 1 H and 13 C resonances of residue H** were assigned employing 1 H, 1 H TOCSY, 1 H, 13 C HSQC TOCSY, and 1 H, 13 C HMBC experiments. The H6a and H6b resonances (3.752 and 3.934 ppm, respectively) were identified in the 1 H, 1 H TOCSY spectrum ( m 100 ms) due to their significant chemical shifts differences with respect to those of H6a and H6b in residue H and were correlated to the C6 carbon in the multiplicityedited 1 H, 13 C HSQC spectrum. The differences in the chemical shifts of C6 in residues H** and H (61.46 and 69.66 ppm, respectively) indicated that the former is not 6-substituted and thus is a terminal residue. Another conspicuous signal was observed in the multiplicity-edited 1 H, 13 C HSQC spectrum at ␦ H /␦ C 3.736/ 79.12 (H4/C4 in residue G**); the 1 H resonances in that spin system were identified using correlations from the carbon at 79.12 ppm in the 1 H, 13 C HSQC TOCSY spectrum ( m 100 ms), and the corresponding 13 C resonances were assigned using a multiplicity-edited 1 H, 13 C HSQC spectrum. Furthermore, an inter-residue correlation was observed in the band-selective constant-time 1 H, 13 C HMBC spectrum (recorded with enhanced resolution in the carbon anomeric region) between the anomeric carbon of residue H** and the proton at 3.736 ppm (H4 in residue G**). In addition, the regular 1 H, 13 C HMBC spectrum showed a correlation between the anomeric proton of residue H** and the C4 carbon in residue G** (Fig. 5D), thus confirming that the G** residue is 4-substituted with residue H**. Furthermore, the anomeric carbon of residue G** showed an inter-residue correlation in the band-selective constanttime 1 H, 13 C HMBC spectrum to a proton at 4.129 ppm, attributed to H3 of residue F** (overlapping with H3 of residue F). 1 H and 13 C chemical shifts assignments of the deca-and nonasaccharide components of the LOS from Bm_wbkD are compiled in Table 4; the corresponding structures and relative percentages are shown in Fig. 4B.
LOS from Bm_wadC_per-Two oligosaccharides were isolated from the deacylated LOS of the double mutant, viz. a tetra-and a pentasaccharide (OS1 and OS2, respectively), the structures of which correspond to those described for the oligosaccharide mixture from Bm_manB core . Their 1 H and 13 C NMR chemical shifts (Table 3) were assigned by two-dimensional NMR experiments (cf. 1 H, 13 C HSQC NMR spectra in Fig.  6, A and B, respectively), in good agreement with those from the Bm_manB core oligosaccharides. In particular, an interglycosidic heteronuclear correlation was present in the 1 H, 13 C HMBC spectrum of the pentasaccharide from the anomeric proton (H1) of the glucosyl residue to the glycosyloxylated carbon atom (C4) of the second Kdo residue (Fig. 6C), thus establishing and confirming the structural element E-D (Fig. 4).
In a recent investigation of oligo-and polysaccharides obtained by mild acid hydrolysis of LPS from different Brucella serotypes, structural elements consistent with core oligosaccharides presented herein were present (15). Notably, it was shown that D-QuiNAc, which is the primer for the O-chain polysaccharide, is ␤-(134)-linked to the glucose residue of the core (B. suis data), thereby defining the attachment site of the O-antigen to the core region of Brucella serotypes, a finding anticipated to be valid in the Brucella serotypes investigated herein.

Role of Brucella Melitensis Core in Virulence
For an unambiguous analysis of the role of the LPS core in the virulence of B. melitensis, the use of core-defective O-PS-bearing bacteria is necessary. To this end, a mutant in wadC was constructed. As shown in Fig. 7, SDS-PAGE and Western blot analyses show that this wadC mutant carries a core defect but keeps the O-PS, a result in agreement with previous work in B. abortus (17). This phenotype is consistent with the putative role of WadC as a mannosyltransferase because this enzymatic activity would be necessary to create the mannose-Kdo linkage in the structure shown in Fig. 4B. To verify this, a double Bm_wadC_per mutant was constructed and its LOS analyzed. Consistent with the role of Per in O-PS synthesis (Fig. 1) and the need of the O-PS for export to the periplasm and subsequent linkage to the core oligosaccharide (5), the LOS of this double mutant lacked the full O-PS section (quinovosamine, mannose, and N-formyl perosamine polymer). Moreover, this LOS also completely lacked the mannose/glucosamine-containing oligosaccharide linked to Kdo (residue C), being similar to that of the manB core mutant (blocked in mannose synthesis). The lack of aminosugars should increase the negative charge of the inner sections of LPS, and this was shown to occur in sm potential measurements of the Bm_wadC_per mutants and Bm_per (Fig. 8).  13 C, and 31 P NMR chemical shifts (ppm) at 25°C and pD 8 of the resonances of the components of the deacylated LOS from BmEPR and Bm_wbkD and inter-residue correlations from the 1 H, 13 C HMBC spectra. The different oligosaccharide components of the samples differ in the presence/absence of residues E (Glc) and/or I (GlcN) (see Fig. 4). For residue A, 3 J P,H1 ϭ 8.3 Hz, 2 J P,C1 ϭ ϳ4.5 Hz, and 3 J P,C2 ϭ ϳ7.4 Hz. Resonances from H1 and C1 in residue G* are found at 4.626 and 102.12 ppm, respectively (overlapping with H1 and C1 in residue G**). Resonances from H3 and C3 in F** are found at 4.129 and 80.07 ppm (overlapping with H3 and C3 in residue F). ND means not determined. NR means not resolved. To assess the biological effect of the core, the ability of Bm_wadC to multiply in BMDCs was studied in comparison with wild type B. melitensis. Mutant bacteria displayed a comparatively reduced ability to multiply in these cells suggestive of attenuation (Fig. 9A). When the virulence was assessed in vivo using the mouse model of brucellosis, Bm_wadC and the parental strain yielded similar cfu at post-infection week 2 (Fig. 9B).

Sugar
At post-infection week 8, however, the mutant was present in comparatively reduced numbers in the spleens (Fig. 9C).
To determine whether the reduced virulence of Bm_wadC was associated with an alteration in the immunogenicity of its LOS structure, further experiments were performed to assess its potential to induce pro-inflammatory responses in BMDCs. Unlike the Bm_wt LPS, which induced no secretion of inflammatory cytokines, Bm_wadC LPS induced the release of the pro-inflammatory cytokines IFN-␥, IL-12p40, IL-6, and TNF-␣ at high levels that were comparable with those obtained with LPS from E. coli (Fig. 10A). Similarly, with E. coli LPS, Bm_wadC-stimulated BMDCs underwent maturation as judged by the surface expression of MHC-II and the co-stimulatory markers CD86, CD80, and CD40. By contrast, BMDCs treated with Bm_wt LPS maintained an immature phenotype with no evident up-regulation of these surface receptors (Fig.  10B). Taken together, these results demonstrate that an intact LPS core is not only required for full virulence of B. melitensis, but it also contributes to limiting the activation and maturation of dendritic cells while undergoing replication in these target cells. An alteration in the LPS core, as demonstrated here with Bm_wadC, would appear to confer a more endotoxigenic phenotype rendering the pathogen more visible to host target cells, attenuating its intracellular replicative capacity and virulence in mice.

Discussion
Rough mutants of B. melitensis have been reported to show attenuated responses with respect to strains having a smooth LPS. In addition, phenotypes with different core defect structures have been observed to give rise to different attenuation patterns, and this has been associated with changes on the bacterial surface (10,17). Therefore, knowledge of the primary structure of the LOS may help not only to have a better understanding of the genetics involved in biosynthetic processes but also to comprehend the interaction of the bacteria with the host immune system. Herein, we present a comprehensive study of the structures of different LOSs extracted from four B. melitensis strains. These include the native R-LPS (LOS) obtained from a strain producing increased proportions of R-LPS (18) (BmEPR), a mutant in which the wbkD gene was disrupted (Bm_wbkD) and a mutant affected in the ManB core protein (Bm_manB core ) as well as the double mutant strain Bm_wadC_per.
The structural elucidation of the deacylated LOSs from BmEPR and Bm_wbkD, carried out using NMR spectroscopy, revealed that both mutants are capable of producing a complete core structure (Fig. 4B) as well as minor components lacking the terminal residue I (19 and 48%, respectively) and/or residue E (34 and 17%, respectively). Furthermore, the pseudo-molecular ions observed in the mass spectra of the deacylated LOS from Bm_wbkD ( Fig. 2B and Table 2) were in agreement with the two major components determined by NMR spectroscopy as follows: (i) a complete core decasaccharide with a molecular formula of C 64 H 116 N 8 O 53 P 2 that is consistent with a structure containing two Hex (Glc and Man), four HexN (GlcN), two Kdo, and two HexNN residues bearing phosphomonoester substituents (GlcN3N1P and GlcN3N4P), and (ii) a nonasaccharide with a molecular formula of C 58 H 104 N 7 O 49 P 2 that is consistent with the same components as above but having three GlcN residues instead of four. It needs to be stressed that the full structures obtained for BmEPR and Bm_wbkD are identical. B. melitensis EP was used in this work because it keeps the ability to synthesize S-LPS while producing increased amounts of R-LPS, which makes possible to obtain larger amounts of a native R-LPS for chemical analysis. Because a complete core structure was present in the LOS from Bm_wbkD, the wbkD gene can be anticipated to encode for a protein involved in the biosynthesis of the O-PS. This is in agreement with the predicted function of WbkD as involved in the biosynthesis of QuiNAc, the undecaprenol-priming residue and therefore the first sugar of the O-PS chain (Fig. 1) (10). In a smooth LPS the O-PS chain, having a 3-substituted QuiNAc residue located at its reducing end, would be extended from the O4 position of the Glc residue (15). Consequently, the pentasaccharide moiety  composed of GlcN residues linked to a mannosyl residue constitutes a core branch. The mutant in the manB core gene showed a deeply truncated core (Fig. 4A) in which the lateral branch is absent. This is in agreement with previous observations that such a kind of mutants (Bm_manB core ) have a lower molecular mass LOS (R3 phenotype) than those of BmEPR and Bm_wbkD (R1 phenotype) and fail to react with monoclonal antibodies specific for the outer core (10,16). The studies carried out using NMR spectroscopy showed that the LOS from Bm_manB core is com- posed of two major oligosaccharide components as follows: a pentasaccharide and a tetrasaccharide (63 and 37%, respectively), with the only difference being the presence or absence of the terminal Glc residue, respectively (Fig. 4A). These structures are in agreement with the two sets of pseudo-molecular ions observed in the mass spectrum that correspond to structures of general formula C 34 H 62 N 4 O 32 P 2 and C 28 H 52 N 4 O 27 P 2 ( Fig. 2A and Table 2), respectively. The ManB core has been previously predicted to be the phosphomutase involved in the biosynthesis of GDP-mannose (Fig. 1), which is used in the biosynthesis of monosaccharide components of the core and O-PS. Indeed, it is shown here that when the manB core gene is blocked the LOS oligosaccharide lacks the mannosyl residue and the branched structure that extends from it. From a biological point of view, the structure described here is not a trivial one as it explains a number of characteristics related to the virulence of an important pathogen. The brucellae are resistant to components of the innate immune system such as bactericidal peptides and complement while prompting a weak pro-inflammatory response during infection (7), characteristics that are in part related to the biological and physicochemical properties of its LPS (6,7,17). Many of these properties have been attributed to the poor cytokine-inducing characteristics of Brucella lipid A, on account of its very long chain fatty acids that may prevent effective recognition by TLR4-MD2 (6). However, it has been predicted that the core should play a complementary role whereby its structure allows it to conceal the inner sections of lipid A, making them less accessible for binding by the TLR4-MD2 complex or for binding bactericidal peptides or complement (15,17). The loss in monoclonal antibody reactivity observed with a B. abortus wadC mutant suggested that the core of smooth brucellae may atypically be composed of two separate oligosaccharide branches, one linking the lipid A to the O-PS and another that protrudes laterally thereby concealing the negatively charged groups essential for interaction with the TLR4-MD2 complex (17). The characterization presented here confirms these predictions as the GlcN-rich oligosaccharide stemming from the first Kdo residue (C) also creates a positively charged structure that can neutralize the negatively charged groups of Kdo and lipid A, the primary targets of innate immunity receptors and effector molecules. This is further supported by the attenuated intracellular replicative/virulence profile and change in surface charge exhibited by the Bm_wadC mutant strain, as well as the enhanced ability of its LPS to stimulate maturation and pro-inflammatory cytokine secretion (including Th1type cytokines IL-12 and IFN-␥) when compared with its WT counterpart.
In addition to reaffirming the above hypothesis, these observations provide new understanding of the important role of Brucella core-LPS in influencing key host-pathogen interactions during the early stages of infection and perhaps when chronic disease develops. Moreover, our findings have important implications as they identify Brucella LPS as a key target/protagonist for strategies aimed toward the development of safer attenuated brucellosis vaccines that can promote more efficient protective Th1 responses (45), thought to be critical for the control of intracellular brucellae.