Characterization of a Novel Branched Tetrasaccharide of 3-Deoxy-d-manno-oct-2-ulopyranosonic Acid

 For the first time, the tetrasaccharide Kdoα2→5Kdoα2→5(Kdoα2→4)Kdo (Kdo is 3-deoxy-d-manno-oct-2-ulopyranosonic acid) has been identified in a bacterial lipopolysaccharide (LPS),i.e. in the core region of LPS from Acinetobacter baumannii NCTC 10303. The LPS was analyzed using compositional analysis, mass spectrometry, and NMR spectroscopy. The disaccharided-GlcpNβ1→6d-GlcpN, phosphorylated at O-1 and O-4′, was identified as the carbohydrate backbone of the lipid A. The Kdo tetrasaccharide is attached to O-6′ of this disaccharide and is further substituted by shortl-rhamnoglycans of varying length and by the disaccharided-GlcpNAcα1→4d-GlcpNA (GlcpNA, 2-amino-2-deoxy-glucopyranosuronic acid). The core region is not substituted by phosphate residues and represents a novel core type of bacterial LPS. The complete carbohydrate backbone of the LPS is shown in Structure I as follows: Kdo α 2 → 4 Kd 5 ↑ Glc p NAc α 1 → 4 Glc p NA α 1 → 4 Kdo α 2 5       ↑       L Rha p ∗ α 1 → 3 L Rha p ∗ α 1 → 3 L Rha p α 1 → 3 L Rha p α 1 → 8 Kdo α 2       o α 2 → 6 Glc p N β 1 → 6 Glc 4   ↑   P   p N α 1 → P STRUCTURE I where Rha is rhamnose. Except were indicated, monosaccharides possess thed-configuration. Sugars marked with an asterisk are present in non-stoichiometric amounts.

The abbreviations used are: COSY, correlated spectroscopy; FAB-MS, fast atom bombardment-mass spectrometry; GlcNA, 2-amino-2deoxyglucuronic acid; HPAEC, high performance anion exchange chromatography; Kdo, 3-deoxy-D-manno-oct-2-ulopyranosonic acid; LPS, lipopolysaccharide; NOE, nuclear Overhauser effect; TOCSY, total correlation spectroscopy; NOESY, nuclear Overhauser spectroscopy; HMQC, heteronuclear multiple quantum coherence. references therein) have determined several structures of Ospecific polysaccharides from various Acinetobacter genomic species, and we have further shown that rabbit antisera raised against a given LPS react highly specifically. It is still not fully understood why Acinetobacter S-form LPS cannot be stained using silver nitrate in SDS-polyacrylamide gel electrophoresis.
In general, the core region and lipid A represent a common structural unit occurring in all LPS; thus they are thought to be important for viability and membrane function of Gram-negative bacteria (11). Compared with the O-specific polysaccharide structures, the structural variability is more restricted within LPS from one genus. All core regions that have been investigated so far possess at least one residue of 3-deoxy-D-mannooct-2-ulopyranosonic acid (Kdo) which links the core region to the lipid A. This Kdo residue may be substituted at O-4 by a second ␣-linked Kdo or an ␣-234-linked Kdo disaccharide (e.g. in LPS from Salmonella or Escherichia coli), and, in most cases that have been analyzed, it is elongated at O-5 with the rest of the core region, either via a manno-configured heptose or (in heptose-deficient cores) a hexose residue. Unique structures were identified in the core region of chlamydial LPS where either the trisaccharide Kdo␣238Kdo234Kdo (in Chlamydia trachomatis) or the tetrasaccharide Kdo␣238(Kdo␣234)-Kdo␣234Kdo (in Chlamydia psittaci) is present and no substitution at O-5 of any of the Kdo residues occurs (12). The branched Kdo tetrasaccharide has to date been the largest Kdo oligosaccharide identified.
The biosynthesis of lipid A and the inner core region has been investigated in detail in E. coli, and it has been found that the first two Kdo residues are transferred by a single Kdo transferase which is thus bifunctional (13). Similar results were obtained for the chlamydial Kdo transferases which transfer either three (in C. trachomatis and Chlamydia pneumoniae) or four (in C. psittaci) Kdo residues (12). Although these Kdo transferases are multifunctional enzymes, their primary structures have little similarity. In a recent investigation, the Kdo transferase genes from different strains of A. baumannii and Acinetobacter haemolyticus have been characterized (14). In in vitro assays it was shown that each of these enzymes transfers up to three Kdo residues to a synthetic lipid A precursor from E. coli, establishing their trifunctionality.
We have investigated the structure of the carbohydrate backbone of LPS from A. baumannii strain NCTC 10303 (ATCC 17904, formerly Acinetobacter calcoaceticus NCTC 10303), in which the presence of the trisaccharide L-Rhap␣133L-Rhap␣138Kdo had been reported in an earlier investigation (15). We now present the complete structure of this carbohydrate backbone, establishing a novel LPS core region comprising a novel branched Kdo tetrasaccharide that will help to extend our knowledge of multifunctional Kdo transferases.

EXPERIMENTAL PROCEDURES
Bacteria and Bacterial LPS-A. baumannii strain NCTC 10303 was cultivated as described (8). The LPS was obtained (yield: 3% of dry mass) from the lyophilized culture supernatant by a modified phenol/ chloroform/light petroleum extraction as described (8).
Preparation of Oligosaccharides-The LPS (80 mg) was hydrolyzed in 5% acetic acid (100°C, 5 h), and the precipitate was removed by ultracentrifugation (100,000 ϫ g, 4 h). The supernatant was lyophilized (30 mg, 37.5% of the LPS) and contained a mixture of oligosaccharides which was separated on a column of Dowex 50-X4 (H ϩ ). Two fractions were obtained as follows: a neutral fraction, eluting with water, and a basic fraction, eluting with 5% aqueous ammonia. The basic fraction contained oligosaccharides 1 and 2, which were separated by using high performance anion exchange chromatography (HPAEC). The neutral fraction was reduced with NaBH 4 and separated using gel permeation chromatography on TSK-40 in water, which yielded as one fraction oligosaccharide 3 and a second, low molecular mass fraction that was not further investigated.
In a third experiment, the LPS (80 mg) was dissolved in anhydrous hydrazine (3 ml), kept at 60°C for 120 h, and evaporated to dryness. Water was added to this sample, and the insoluble fatty acids were removed by centrifugation (2,500 ϫ g, 30 min), the supernatant of which was injected onto a reversed-phase HPLC Delta-Pak C18 (Waters) column (2 ϫ 30 cm) that was eluted using water and from which oligosaccharide 7 was isolated (8 mg, 10% of the LPS).
Mass Spectrometry-Gas chromatography and gas chromatography mass spectrometry were performed as described (17). Positive ion fast atom bombardment(FAB)-mass spectra were obtained using MS1 of a JEOL JMS-SX/SX102A tandem mass spectrometer operated at ϩ or Ϫ10 kV accelerating voltage. The FAB gun was operated at 6 kV accelerating voltage with an emission current of 10 mA and using xenon as the bombarding gas. Spectra were scanned at a speed of 30 s for the full mass range specified by the accelerating voltage used and were recorded and averaged on a Hewlett-Packard HP9000 data system running JEOL COMPLEMENT software. Collision induced dissociation mass spectra were recorded using the same machine, with helium as the collision gas in the third field free region collision cell, at a pressure sufficient to reduce the parent ion to one-third of its original intensity. The oligosaccharide was dissolved in 50 l of water, and 1-l aliquots of sample solution were loaded into a matrix of mono-thioglycerol, to which 1 l of trifluoroacetic acid was added for the analysis of the untreated oligosaccharide. Negative ion mode electrospray mass spectrometry was performed on a VG Platform II single quadrupole mass spectrometer. Five l of the oligosaccharide solution made for the FAB mass spectrometric analysis were diluted by addition of 50 l of water, and 10 l of this solution was infused into a mobile phase of acetonitrile/ water (50:50, v/v) and introduced into the electrospray source at a flow rate of 5 l min Ϫ1 . Spectra were scanned at a speed of 10 s for m/z 700 -1700, with a cone voltage of 60 V and recorded and processed using the MassLynx software, version 2.0.
NMR Spectroscopy-Structural assignments were made from sample solutions in 2 H 2 O at 27 or 37°C. For compounds 1-6, 1 H and 13 C NMR spectra were recorded at 600.13 MHz for 1 H and 150.90 MHz for 13 C with a Bruker AMX 600 spectrometer, and chemical shifts were reported relative to internal acetone (2.225 ppm for 1 H and 31.5 ppm for 13 C). The 1 H, 31 P COSY spectrum was recorded at 101.25 MHz on a Bruker DRX 250 instrument at ambient temperature. Correlation spectroscopy (COSY), total correlation spectroscopy (TOCSY), nuclear Overhauser spectroscopy (NOESY), and heteronuclear multiple quantum coherence (HMQC) spectra were assigned using the computer program PRONTO (20). For compound 7, spectra were recorded on a solution of 5 mg in 2 H 2 O with a Varian INOVA 750 spectrometer at 750.04 MHz for 1 H and at 188.6 MHz for 13 C at 27°C. The double quantum-filtered phase-sensitive COSY experiment was performed using the Varian standard program TNDQCOSY (21,22), with 0.34 s acquisition time and 4 K data points. F1 was zero-filled to give a matrix of 4 ϫ 2K points and was resolution-enhanced in both dimensions by a shifted sine bell. NOESY was performed using the Varian standard program TNNOESY (23) with a mixing time of 200 ms. TOCSY was performed using the Varian standard program TNTOCSY (24,25), with a spin-lock time of 120 ms. The heteronuclear single quantum coherence (HSQC) spectros-copy was performed using the pulse field gradient standard Varian program GHSQC (26), with a gradient strength of 4, 4, and 2 G/cm and gradient time of respectively 2, 2, and 0.5 ms, respectively.
Conformational Analysis-Conformational analysis was performed using the hard sphere exoanomeric-based GEGOP program (27,28) at 800 and 1200 K with Metropolis Monte Carlo simulations with up to 3 ϫ 10 6 steps. The ␣and ␤-D-Kdo coordinates were taken from a published x-ray structure (29). For a semi-quantitative comparison with the observed NOE proton-proton distances, Ͻr Ϫ6 Ͼ were monitored. Furthermore, minimization and molecular dynamics were carried out for some oligosaccharides fragments using the computer program DISCOVER with the consistent valence force field (30,31).

RESULTS AND DISCUSSION
Isolation and Chemical Analysis of the LPS-Most of the LPS from A. baumannii NCTC 10303 was released from the bacteria into the culture supernatant, from which it could be isolated using a modification of the phenol/chloroform/petroleum ether method (8). The additional extraction of dried bacteria gave less than 0.1% of the total LPS yield. Monosaccharide analysis of the LPS and determination of the absolute configurations of monosaccharides revealed the presence of L-Rha, D-Kdo, and D-GlcN. The configuration of Kdo was determined by measuring the optical rotation of its methyl ester, ␣-methylketoside derivative, isolated after mild methanolysis of the LPS. Further compositional analyses of the LPS identified phosphorus, dodecanoic acid, 2-hydroxydodecanoic acid, and 3-hydroxydodecanoic acid, the last of which was the only amide-bound fatty acid in the lipid A, as revealed by fatty acid analysis of de-O-acylated LPS.
Structural Analysis of Oligosaccharides 1 and 2 and Mixture 3-In an earlier investigation (15), the trisaccharide L-Rhap-␣133L-Rhap␣138Kdo had been obtained from one LPS of A. baumannii NCTC 10303 after mild acidic hydrolysis, dialysis, and separation of the dialysate on Bio-Gel P2. However, a second fraction possessing a higher molecular mass had additionally been isolated which contained Rha, Kdo, and phosphate but had not been investigated further. Consequently, the first step of the current investigation comprised hydrolysis of the LPS with 5% acetic acid in order to characterize the complete spectrum of products. Separation of the hydrolysate using ion exchange chromatography gave a basic fraction that was separated further using HPAEC that yielded oligosaccharides 1 and 2 and a neutral fraction that contained oligosaccharide mixture 3. The structures of oligosaccharides 1 and 2 were established using NMR spectroscopy (Tables I and II). The 1 H assignments are based on two-dimensional phase-sensitive COSY and TOCSY and that of 13 C on HMQC experiments. Anomeric configurations are assigned on the basis of the chemical shifts observed, and J 1,2 values were determined from the COSY experiment (Table I). Oligosaccharide 1 consists of one residue each of 2,7-anhydro-3-deoxy-D-manno-oct-2-ulofuranosidic acid (2,7-anh-Kdof), a hexosaminuronic acid ( 13 C NMR data, C-6 at 174.1 ppm in oligosaccharide 1) and a hexosamine. 2,7-anh-Kdof represents an artifact originating from acid hydrolysis of Kdo (32). It is identified by its 1 H and 13 C NMR chemical shifts and J H, H coupling constants that are in full agreement with those published (32). The hexosamine possesses the gluco-configuration, as revealed by its coupling constants J 2,3 and J 3,4 (ϳ10 Hz). The coupling constants of the hexosaminuronic acid could not be determined from this sample due to signal overlapping; however, this was possible with oligosaccharide 2 which was identified as a partial structure of oligosaccharide 1, namely the reducing disaccharide hexosamine-hexosaminuronic acid. Thus, the hexosaminuronic acid also possesses the gluco-configuration (J 2,3 , 10 Hz; J 3,4 , 9 Hz). In both amino sugars, the amino group is located at C-2 as follows from the chemical shifts for C-2 at 54 -59 ppm. Both, oligosaccharides 1 and 2 possess one N-acetyl group. As re-vealed by the chemical shifts of H-2, the amino group of the glucosamine is acetylated (3.814 ppm in oligosaccharide 1 and 3.822 ppm in 2), whereas that of the glucosaminuronic acid is free (2.71 ppm in oligosaccharide 1, 2.74 (␣) and 3.02 (␤) ppm in 2). The absolute configuration of glucosaminuronic acid could be determined as D using conformational analysis (see below); thus, the amino sugars in oligosaccharides 1 and 2 are 2-acetamido-2-deoxy-D-glucopyranose (D-GlcpNAc) and 2-amino-2-deoxy-D-glucopyranosuronic acid (D-GlcpNA). The sequence of monosaccharides in oligosaccharides 1 and 2 was determined by NOE measurements that gave NOE contacts between G1 and F4, and between F1 and E4 and E5. 2,7-anh-Kdof cannot be substituted at O-5 and hence must be glycosylated at O-4. This agrees with the 13 C NMR chemical shift data that indicate that GlcpNAc is a terminal residue (G in Fig. 1), and that GlcpNA (F) and 2,7-anh-Kdof (E) are each substituted at O-4 (signals for C-4 of GlcpNA at 76.1-76.8 ppm and of 2,7-anh-Kdof at 74.6 ppm). Taken together, the data allow the structures of oligosaccharides 1 and 2 to be assigned as depicted in Fig. 1.
Since the Kdo residue E is substituted by GlcpNA, the amino group that stabilizes the glycosidic bond, it cannot be liberated under conventional hydrolysis or methanolysis conditions. Consequently, its absolute configuration was determined by measuring the optical rotation of 2,7-anh-Kdof, obtained after deamination of oligosaccharide 1 and paper chromatography. The optical rotation of the product was ␣ D ϩ63°which is close to that of the reference compound prepared from authentic Kdo (␣ D ϩ68°) and allowed assignment of the D-configuration.
The neutral fraction obtained on ion exchange chromatography of the LPS hydrolysate contained a mixture of several oligosaccharides with similar compositions. It was reduced with NaBH 4 to give mixture 3, the NMR spectra of which contained signals for terminal and 3-substituted ␣-Rhap, and various signals of lower intensity, originating from Kdo derivatives (data not shown). Monosaccharide analysis of mixture 3 identified Rha and polyols and anhydropolyols of Kdo. Methylation analysis identified terminal and 3-substituted Rha in the molar ratio of ϳ1:2.5. The ratio Rha:Kdo could not be determined because of the presence of multiple forms of Kdo. However, it may be concluded that mixture 3 consists of several oligosacharides (predominantly tri-and tetrasaccharides) with structures similar to that of the previously identified trisaccharide L-Rhap␣133L-Rhap␣138Kdo (15).
Structural Analysis of Oligosaccharides 4 -6 -Complete deacylation yielded oligosaccharides 4 -6, isolated as major components from the resulting oligosaccharide mixture using HPAEC. NMR spectroscopy of tetrasaccharide 4 demonstrated its identity to the previously described structure Kdo␣234-Kdo␣236GlcpN4P␤136GlcpN␣13 P (33). NMR analysis (COSY, TOCSY, NOESY, HMQC, 31 P, 31 P, 1 H COSY) of oligosaccharides 5 and 6 ( Fig. 1) established them to be hexa-and heptasaccharides, respectively (Tables I and II, Fig. 2). Heptasaccharide 6 was shown to be composed of D-GlcpNAc134D-GlcpNA134Kdo (modified trisaccharide 1) which is ␣235linked to tetrasaccharide 4, as demonstrated by the downfield shift of the signal for C-5 of Kdo C at 68.0 ppm (compared with about 65.6 ppm in tetrasaccharide 4). Hexasaccharide 5 comprises basically the same structure; however, the terminal GlcpN is absent, and GlcpNA is converted to its 4,5-unsaturated derivative as the result of alkaline ␤-elimination ( 1 H and 13 C NMR data, signals for H-4 at 5.723 ppm, C-4 at 107.5 ppm, and C-5 at 146.5 ppm). This was supported on prolonged alkaline hydrolysis (24 h at 120°) which converted purified oligosaccharides 6 to 5. Thus, the harsh alkaline conditions used for de-N-acylation induce the rarely occurring cis-elimination at the 4-substituted GlcpNA residue.
Both oligosaccharides contain three Kdo residues, of which residues C and D were identified on the basis of the NOE connectivities observed between protons C3 and D6 that allowed identification of their ␣234-linkage (34). Kdo residue E is substituted by GlcpNA, as indicated by the NOE connectivities between protons F1 and E4 and E5. Determination of the anomeric configuration of residue E was not straightforward. When compared with the chemical shifts of ␤-methyl Kdo (H-3eq, 2.388 ppm) (29), the low-field position of the E3 eq proton  Table I) suggest an ␣-anomeric configuration. The coupling constants identified between the ring protons do not differ from those observed for a Kdo pyranoside in a 5 C 2 conformation; however, the signals for protons E3ax and E3eq appear broadened which may be the result of a conformational equilibrium. As discussed below, it was finally concluded on the basis of the conformational analysis of oligosaccharide 7 that residue E possesses the ␣-configuration.
In the 13 C NMR spectrum, the downfield shift of C-5 (ϳ4 ppm compared with that in tetrasaccharide 4) indicates the attachment of residue E to O-5 of residue C. The NOE connectivities between residues E, D, and C (Table III) also could be explained on the basis of conformational analysis as resulting from the ␣235-linkage between residues E and C. In summary, oligosaccharides 5 and 6 possess the structures as depicted in Fig. 1.
The absolute configuration of GlcpNA was deduced from conformational analyses of oligosaccharides 1 and 5-7 in which strong NOE connectivities between protons F1 and E4 and E5 were identified. The conformations of the disaccharides GlcpNA␣134(2,7-anh)-Kdof␣ and GlcpAN␣134Kdop␣ were calculated for both DD, as well as LD absolute configurations of the constituents. The NOE signals observed (between F1 and E4 and E5) are only consistent with the D-configuration of both sugars, since short distances between these protons were determined on conformational analysis. In the LD-configured disaccharide, the distance between protons F1 and E5 was much greater. Therefore, the NOE signals observed identify the Dconfiguration of GlcpNA. In the 31 P NMR spectra of oligosaccharides 5 and 6, two peaks of similar intensity at 2.6 and 4.3 ppm were identified which, on 31    The procedure used for the preparation of oligosaccharides 4 -6 does not allow the isolation of rhamnose-containing oligosaccharides. After desalting, the reaction mixture contains large amount of rhamnose, the majority of which appears on HPAEC in the solvent front as (Rha) n -Kdo oligosaccharides, whereas a minor amount elute throughout the whole gradient program. The products obtained after deacylation with hot KOH are different if analyzed on HPAEC either before or after neutralization. Thus the rhamnose-containing oligosaccharides appear to be unstable and to rapidly decompose under neutral or acidic conditions. The Structure of Oligosaccharide 7-In order to obtain rhamnose-containing oligosaccharides, LPS was deacylated with anhydrous hydrazine. Separation of the products on a C18 reversed-phase column yielded completely deacylated products, eluting with the solvent front, and oligosaccharide 7, containing one 3-hydroxydodecanoic acid residue. NMR spectroscopy of the deacylated products indicated structures similar to that of oligosaccharide 7; however, it was not possible to isolate a pure compound due to decomposition. Oligosaccharide 7 also decomposes at neutral pH over several days, mostly due to cleavage of Kdo J, but it is stable at pH ϳ8 and could thus be analyzed. 1 H, 13 C, and 31 P NMR spectroscopic analysis of oligosaccharide 7 showed that it contains the partial structure A-G, one additional Kdo (J), and three rhamnose residues (H, I, and L). The structure of the fragment H-I-L-J is identical to that of oligosaccharide 3. To determine the attachment site of the 3-hydroxydodecanoic acid, a NOESY spectrum of a solution of oligosaccharide 7 in H 2 O: 2 H 2 O (9:1, by volume) was recorded. A strong NOE signal between the amide proton and H-2 of GlcN B was observed, indicating its substitution by the fatty acid. Thus this acyl group is more stable to hydrazine deacylation than that attached to GlcN A.
In the 13 C NMR spectrum of oligosaccharide 7, a low-field shift of carbon E5 (69.7 ppm, compared with 63.5 ppm in oligosaccharide 6) indicated the attachment of J to O-5 of residue E. The NOESY spectrum indicated strong NOE connectivities between protons J3 and E7, and E3 and J6, in addition to those identified for oligosaccharides 5 and 6 ( Fig. 4, Table III). In order to determine the structure and explain the NOE connectivities observed, a conformational analysis of the structural elements GlcpN␣134Kdo␣235(Kdo␣234)Kdo␣23 6GlcpN␤ (F-E-D-C-B) and GlcpN␣134(Kdo␣235 or ␣237)-Kdo␣235(Kdo␣234)Kdo␣236GlcpN␤ (F-(J-)E-D-C-B, GlcpN was set as residue instead of GlcpNA) was performed using Metropolis Monte Carlo simulations and the program GEGOP (27,28). The structure with residue J ␣237-linked to E showed distances between protons of these residues not in agreement with observations. The calculations reveal mainly a single min-imumenergyconformationfortheglycosidiclinkageofKdo␣235-Kdo␣ (J-E) with ϳ20°and ϳ20°, where represents the dihedral angle C1-C2-O2-C5 and that of C2-O2-C5-H5 (Fig.  5). The calculated interatomic distances averaged as Ͻr Ϫ6 Ͼ are consistent with all NOE connectivities in which the protons of residue J take part. For the linkage Kdo␣235Kdo␣ (E-D), two main minima were observed. A calculation at 800K showed no transition between these two energy wells, and it is possible to obtain separate averaged distances for each conformation. One of these with and near 20°resembles that of the J-E linkage and does not explain the experimental results (800K-2 in Table III), in particular the NOE connectivities observed between D3eq and E5, C5, and E6, and the absence of connectivities between E3 and C7, and C3 and E6. In contrast, the experimental data obtained are consistent with another conformation with Ϫ90°, Ϫ20°for the E-C linkage (800K-1 in Table III). Calculations at higher temperature (1200 K) showed transitions between these two conformations and give average interatomic distances. Averaging as Ͻr Ϫ6 Ͼ emphasizes short distances, consistent with all NOE connectivities observed. However, averaging clearly predicted an NOE between E3 and C7, which is not observed. In conclusion, it is most likely that the linkage E-C exists in a conformation with approximately Ϫ90°and approximately Ϫ20°, which is different from that of J-E.
Conformational analysis of the structural element GlcpN␣-134Kdo␣235(Kdo␣234)Kdo␣236GlcpN␤ (F-E-D-C-B) yielded a similar conformation, showing that the presence/absence of residue J is of no significant conformational influence. The same two minima were obtained for the linkage E-C, of which only one gave interatomic distances consistent with the NMR data. The other interatomic distances are also consistent with the experimental observations. A similar conformation (Fig. 6) to that described above was obtained from another calculation using the consistent valence force field for the two structural elements possessing three (C, D, E) and four Kdo residues. Similarly, molecular dynamics simulations of the fragment 4Kdo␣235(Kdo␣234)Kdo␣ (E-D-C) for up to 400 ps in a water box at 298 K starting at either minimum did not show any transitions between minima.  Fig. 6). This effect is probably due to a rigid arrangement of the branched Kdo tetrasaccharide, and it can be compared with the general effect observed in ␣-D-pyranoside residues linked to an axial hydroxyl group that is in close proximity to an equatorial hydroxyl group, as is present for example in galabiose (35). Furthermore, the equatorial H-3 is shifted downfield, but less strongly, which is probably due to an interaction with the lone pair of the glycosidic oxygen possessing a fixed orientation. The slight differences between the chemical shifts of protons H-6 and H-3 of Kdo residues E and J can be explained by the different conformation of the ␣235 linkage, also reflected in the differences observed for the NOE connectivities (thus an analogue of the NOE connectivity between protons J3eq and E7 is not observed for the pair E-C).
In conclusion, we have successfully established the chemical structure of a novel core type of LPS from Acinetobacter (Fig. 1) which comprises the tetrasaccharide Kdo␣235Kdo␣235-(Kdo␣234)Kdo␣ (J-E-D-C) that has been identified for the first time. However, a Kdo␣235Kdo␣ disaccharide had been identified in the core region of Campylobacter lari strain PC 637 (36). The core region of LPS from A. baumannii NCTC 10303 includes two structural elements in which a Kdo residue is connected to O-5 of another Kdo residue. The chemical behavior of these 5-linked Kdo residues was found to be different. The linkage of Kdo J which is substituted by short rhamnose oligosaccharides is extremely acid-labile and is also cleaved by hot alkali, whereas that of residue E which is substituted by GlcNAc␣134GlcNA␣ possesses a similar stability to that of the 4-and 6-linked Kdo residues. Conformational analyses of these structural elements showed that the conformation of the glycosidic linkages is different in the two Kdo␣235Kdo␣ disaccharides, which could be a reason for the difference in linkage stability observed.
The herein reported Kdo tetrasaccharide is only the second to be identified in LPS. The first, having the structure Kdo␣238(Kdo␣234)Kdo␣234Kdo␣, was identified from a recombinant E. coli strain that expressed the Kdo transferase and thus furnished the core region of C. psittaci (39). Most FIG. 5. Rotamer population density map for the glycosidic linkages between Kdo residues. A, residues D and C; B, residues E and C; C, residues J and E (see Fig. 1 for letters). interestingly, the Kdo transferase transfers all four Kdo residues and is thus multifunctional. The Kdo transferase(s) of A. baumannii NCTC 10303 has not yet been investigated but that of strain NCTC 15308 was shown to be trifunctional (14). However, no complete structural analysis of this core region has been undertaken to date. Still the question remains whether one or two different Kdo transferases are present in Acinetobacter. Whereas the former case has been identified in LPS biosynthesis of E. coli (13) and Chlamydia (12), the latter was suggested for LPS of Rhizobium etli (40), in which a third Kdo residue that is substituted by the O-antigen is present at the non-reducing terminus of the core region (41). It was shown that R. etli contains all enzymes necessary to furnish a Kdo disaccharide that is linked to the E. coli lipid A precursor (42). Thus, a single Kdo transferase (as in E. coli) should transfer the first two Kdo residues, whereas a second Kdo transferase was proposed to be responsible for the transfer of the third Kdo residue. The herein reported structure of LPS from A. baumannii NCTC 10303 represents a prerequisite for investigating the question of whether one or more Kdo transferases are needed for core biosynthesis, since it is needed for the construction of defined mutants and/or for chemical synthesis of partial core structures that are employed in the production of monoclonal antibodies.