The Obligate PredatoryBdellovibrio bacteriovorusPossesses a Neutral Lipid A Containing α-D-Mannoses That Replace Phosphate Residues: SIMILARITIES AND DIFFERENCES BETWEEN THE LIPID As AND THE LIPOPOLYSACCHARIDES OF THE WILD TYPE STRAINB. BACTERIOVORUSHD100 AND ITS HOST-INDEPENDENT DERIVATIVE HI100

Abstract Bdellovibrio bacteriovorus are predatory bacteria that penetrate Gram-negative bacteria and grow intraperiplasmically at the expense of the prey. It was suggested that B. bacteriovorus partially degrade and reutilize lipopolysaccharide (LPS) of the host, thus synthesizing an outer membrane containing structural elements of the prey. According to this hypothesis a host-independent mutant should possess a chemically different LPS. Therefore, the lipopolysaccharides of B. bacteriovorus HD100 and its host-independent derivative B. bacteriovorus HI100 were isolated and characterized by SDS-polyacrylamide gel electrophoresis, immunoblotting, and mass spectrometry. LPS of both strains were identified as smooth-form LPS with different repeating units. The lipid As were isolated after mild acid hydrolysis and their structures were determined by chemical analysis, by mass spectrometric methods, and by NMR spectroscopy. Both lipid As were characterized by an unusual chemical structure, consisting of a β-(1→6)-linked 2,3-diamino-2,3-dideoxy-d-glucopyranose disaccharide carrying six fatty acids that were all hydroxylated. Instead of phosphate groups substituting position O-1 of the reducing and O-4′ of the nonreducing end α-d-mannopyranose residues were found in these lipid As. Thus, they represent the first lipid As completely missing negatively charged groups. A reduced endotoxic activity as determined by cytokine induction from human macrophages was shown for this novel structure. Only minor differences with respect to fatty acids were detected between the lipid As of the host-dependent wild type strain HD100 and for its host-independent derivative HI100. From the results of the detailed analysis it can be concluded that the wild type strain HD100 synthesizes an innate LPS.

signed for the isolation of bacteriophages in soil samples (1). They are small, motile Gram-negative bacteria and possess a predatory lifestyle that includes an obligate growth and replication phase taking place in the periplasm of the prey. Bdellovibrionaceae are widely spread in the environment, e.g. soil, marine sediments, rhizosphere of plants, sewage, etc. (1)(2)(3)(4). In the latter environment it was found that Bdellovibrionaceae are involved in the reduction of bacterial counts thus supporting the self-purification of domestic waste waters (4,5). Furthermore, Bdellovibrionaceae were found in the intestinal tract of mammals and might be important for the reduction of pathogenic bacteria in this environment (6,7).
Despite the unique predatory lifestyle and common morphological features, Bdellovibrionaceae show a great phylogenetic diversity based on 16 S rRNA analyses. Bdellovibrionaceae are divided into the species Bdellovibrio bacteriovorus, Bacteriovorax stolpii, Bacteriovorax starrii, and some strains yet to be assigned (3,7,8). The strain B. bacteriovorus HD100 1 investigated in this work is a reference strain, which was isolated from soil (1).
B. bacteriovorus possesses a life cycle consisting of an attack phase followed by attachment and invasion of the periplasmatic space of the prey bacterium. After penetration into the periplasmatic space of the prey the peptidoglycan layer of the host is partially degraded in a short period of time leading to swelling of the prey bacteria and the formation of spherical bdelloplasts. The former outer membrane of the prey forms a barrier against the surrounding environment and thus retains the available nutrients in a confined space (9,10). Inside the prey elongation and multiplication take place and finally the prey bacteria are lysed (11,12). B. bacteriovorus wild type strains solely grow on living bacteria. However, in a multistep selection procedure including streptomycin tolerance as marker host-independent (HI) mutants can be isolated that grow slowly on rich media but show a number of aberrant morphological features (13,14).
In former studies enzymatic activity of host-dependent (HD) B. bacteriovorus against the cell wall of Gram-negative bacteria including the LPS of the outer membrane was detected (15,16). However, in addition to the degradation of macromolecular compounds of the prey several studies also indicated that B. bacteriovorus reutilizes outer membrane proteins, lipid A, and fatty acids of the prey bacteria by integration into its membrane system (15)(16)(17)(18)(19). To understand the cell wall degrading mechanisms detailed information about its own membrane system is needed, as the predatory lifestyle requires that its own cell wall is protected against degradation. Nelson and Rittenberg (18) detected two different lipid A species in LPS preparations from host-dependent B. bacteriovorus 109J by thin layer chromatography (TLC). One lipid A showed more similarity to that of the prey bacteria whereas the other shared common features with the lipid A from a host-independent B. bacteriovorus strain. Chemical analysis revealed a nonadecenoic acid and (OH)-13:0 as characteristic fatty acids, and furthermore, glucosamine was determined as constituent of the lipid A backbone. However, these results did not include a complete structural description of lipid A. The structural determination of LPS and lipid As derived form wild type B. bacteriovorus (HD100) and its derivative (HI100) is a prerequisite to elucidate the molecular interaction between predator and prey bacteria. The structures of the lipid As of the host-dependent strain B. bacterivorus HD100 and its hostindependent mutant HI100 were determined in detail and cytokine release were measured for assessment of biological activity.
LPS Isolation-The bacterial pellets were washed twice with 1% phenol, ethanol, and acetone and than dried at room temperature. For the host-dependent strain the yield was 4 g (pooled from a 22-liter culture) and for the host-independent strain 3.8 g (pooled from a 7-liter culture). After enzymatic degradation of nucleic acids and incubation with proteinase K the phenol/chloroform/light petroleum, 2:5:8 (v/v/v), method was used for LPS extraction (20).
Water was added to the extract of the host-dependent strain HD100 until a fraction H1 precipitated. The precipitate was collected by centrifugation. The remaining supernatant was concentrated until the phenol crystallized on ice water. A second LPS fraction, H2, was obtained by adding ethanol to the remaining solution on ice water and the precipitate was collected by centrifugation. The LPS of the host-independent strain HI100 could only be precipitated with ethanol. All three LPS fractions were washed with acetone three times and dried. The yield was 206 mg for H1 and 110 mg for H2. From the host-independent strain HI100 190 mg of LPS was isolated. For further purification the crude LPS fractions were resuspended in twice distilled water and proteinase K was added to a final concentration of 100 g ml Ϫ1 and incubated at 37°C overnight. These suspensions were dialyzed against twice distilled water and the LPS solutions of about 12 mg ml Ϫ1 were centrifuged at 100,000 ϫ g. The sediments were lyophilized giving a yield of 161.5 mg of LPS of H1, 79.6 mg of LPS of H2, and 111.4 mg of LPS of B. bacteriovorus HI100.
Fatty Acid Analysis-Fatty acids were released and converted to their methyl esters with HCl in methanol (for determination of total fatty acid: 2 M HCl, 24 h, 120°C; for ester-linked fatty acids: 0.5 M HCl, 30 min, 85°C). Fatty acid residues were extracted with chloroform and incubated with diazomethane to obtain the methylated products. After treatment with bis(trimethylsilyl)trifluoroacetamide the trimethylsilylethers of the fatty acids were analyzed by GC-MS. The absolute configuration of the hydroxy fatty acids was determined by GC-MS of the 1-phenylethylamide derivatives (21). GC-MS was performed on a Hewlett Packard mass spectrometer 5989A equipped with a fused silica capillary column (HP-5MS: 30 m, inner diameter: 0.25 mm, film thickness 0.25 m). Helium served as carrier gas and the GC temperature was initially 150°C for 3 min, then raised to 320°C at 5°C min Ϫ1 . Electron impact was carried out at 70 eV and chemical ionization mass spectra (chemical ionization-MS) were recorded with ammonia as reactant gas (0.1 kilopascal).
Sugar Analysis-Qualitative examination of the sugar portion of LPS was performed by methylation (0.5 M HCl/CH 3 OH, 45 min, 85°C, and 2 M HCl/CH 3 OH, 16 h, 85°C) followed by peracetylation (22). The absolute configuration of the mannose and 2,3-diamino-2,3-dideoxyglucose of the lipid A from B. bacteriovorus HD100 was determined with the acetylated (R)-2-butylglycosides (23). The parameters of the GC-MS analysis were identical as described for fatty acid analysis.
Matrix Electrospray Ionization Fourier Transform-Ion Cyclotron Resonance Mass Spectrometry (ESI FT-ICR MS)-ESI FT-ICR MS was performed in the negative and positive ion mode using an APEX II Instrument (Bruker Daltonics, Billerica, MA) equipped with a 7-tesla magnet and an Apollo ion source. Samples were dissolved at a concentration of about 10 ng l Ϫ1 in a 50:50:0.001 (v/v/v) mixture of 2-propanol, water, and triethylamine and sprayed at a flow rate of 2 l min Ϫ1 . Capillary voltage was set to 3.8 kV, and drying gas temperature to 150°C. Capillary skimmer dissociation was induced by increasing the capillary exit voltage from Ϫ100 to Ϫ350 V. Infrared multiphoton dissociation of isolated parent ions was performed with a 35 W, 10.6-m CO 2 laser (Synrad, Mukilteo, WA). The unfocused laser beam was directed through the center of the ICR cell for 40 ms and the fragment ions were detected after a delay of 0.5 ms.
Nuclear Magnetic Resonance (NMR) Spectroscopy-For NMR spectroscopy all exchangeable lipid A protons were replaced with deuterium by dissolving 5.0 mg of lipid A in 1.0 ml of chloroform-d/methanol-d 4 , 1:1 (v/v). After evaporating it in a slight N 2 stream, lipid As were dissolved in 0.5 ml of chloroform-d/methanol-d 4 7:3 (v/v) and transferred into 5-mm high precision NMR sample tubes (Promochem, Wesel, Germany). Tetramethylsilane was used as internal standard for 1 H (0.00 ppm) and 13 C (0.00 ppm) spectra.
Proton and all two-dimensional spectra were recorded at 600.1 MHz on a 14-tesla AVANCE DRX-600 (Bruker, Rheinstetten, Germany) and 13 C spectra were measured on a 8.4-tesla AVANCE DPX-360 (Bruker) at 90.6 MHz. All spectra were measured at 310 K and performed with Bruker XWINNMR 2.6 software. One-dimensional measurements were recorded with a 90°pulse angle, an acquisition of 16,384 data points, and a relaxation delay of 1.0 s. After zero filling to 32,768 data points they were Fourier transformed to spectra with a range of 7200 Hz ( 1 H) or 22,000 Hz ( 13 C). To determine DQF-COSY, TOCSY (100-ms mixing time), ROESY (250-ms mixing time), HMQC (coupled and decoupled), HMQC-TOCSY, and HMBC, between 256 and 512 experiments in the F 1 -dimension were recorded, each with 2048 data points in the F 2dimension. Sinusoidal multiplication and Fourier transformation led to two-dimensional spectra with a range of 4800 Hz in the proton dimension as well as 22,000 Hz and 32,000 Hz in the carbon dimension for HMQC and HMBC, respectively.
Immunological Characterization of LPS-Purified LPS was separated by SDS-PAGE gels and either stained with alkaline silver nitrate (24) or detected with monoclonal antibodies (mAb) after electrotransfer onto polyvinylidene difluoride membranes by tank-blotting (Bio-Rad Mini Trans-Blot cell). Prior to use, polyvinylidene difluoride membranes were wetted in methanol and rinsed carefully in distilled water (at least 10 min), where they were kept until further use. Blotting was carried out at 4°C for 16 h at 10 mA, all following steps were performed at room temperature. After transfer, membranes were placed in dis-tilled water for 30 min, washed six times for 5 min each in blot buffer (50 mM Tris-HCl, 0.2 M NaCl, pH 7.4), blocked 1 h in blot buffer supplemented with 10% nonfat dry milk, and incubated for 1 h with mAb A6, directed against a bisphosphorylated lipid A backbone (25), A20, recognizing a terminal Kdo residue (26), or mAb WN1, reacting with the core region of E. coli LPS (27). Antibody A6 was used as the cell culture supernatant (RPMI, supplemented with 10% fetal calf serum), A20 (2 g ml Ϫ1 ) and WN1 (5 g ml Ϫ1 ) were diluted in the same medium. Blots were washed six times (5 min each in blot buffer) to remove the primary antibody, followed by incubation for 1 h with alkaline phosphatase-conjugated goat anti-mouse immunoglobulin G (HϩL, Dianova, diluted 1:2000 in blot buffer supplemented with 10% nonfat dry milk), washed as before, and developed with 5-bromo-4-chloro-3-indolyl-phosphate and p-nitro blue tetrazolium chloride as substrates according to the supplier's instructions.
Release of Cytokines from Human Mononuclear Cells-Human mononuclear cells (hMNC) were isolated from healthy donors. Heparinized blood (20 IU ml Ϫ1 ) was processed directly by mixing with an equal volume Hanks' balanced salt solution and centrifugation on Ficoll density gradient for 40 min (21°C, 500 ϫ g). The interphase layer of mononuclear cells was collected and washed twice in serum free Hanks' solution and once in serum-free RPMI 1640 containing 2 mM L-glutamine, 100 units ml Ϫ1 penicillin, and 100 g ml Ϫ1 streptomycin. Cells were resuspended in serum-free medium, and the cell number was adjusted to 5 ϫ 10 6 ml Ϫ1 .
For the stimulation experiment 200 l of hMNC were transferred to each well of 96-well culture plates and LPS and lipid A were added to give final concentrations of 100 ng ml Ϫ1 , 10 ng ml Ϫ1 , and 1 ng ml Ϫ1 . The samples were incubated for 4 h at 37°C and 5% CO 2 . Supernatants were collected after centrifugation of the culture plates for 10 min at 400 ϫ g and stored at Ϫ20°C until further use.
The release of TNF-␣ in the cell supernatants was performed in a sandwich enzyme-linked immunosorbent assay as described elsewhere (28). Briefly, microtiter plates (Greiner, Solingen, Germany) were coated with a mAb against TNF-␣ (Intex AG, Switzerland). Cell culture supernatants and standard (recombinant TNF-␣, rTNF-␣, Intex) were added as appropriately diluted test samples and serial dilutions of rTNF-␣. The horseradish peroxidase-conjugated rabbit anti-rTNF-␣ antibody was added and the plates were incubated while shaking for 16 to 24 h at 4°C. After six washings with distilled water the color reaction was initiated by addition of tetramethylbenzidine/H 2 O 2 in alcoholic solution and stopped with the same amount of 1 M H 2 SO 4 . The resulting yellow oxidation product was determined at a wavelength of 450 nm on an enzyme-linked immunosorbent assay reader (Rainbow, Tecan, Crailsheim, Germany). Interleukin-6 enzyme-linked immunosorbent assay was performed according to the manufacturer's recommendation (Intex AG, Switzerland).
Determination of (␤ 7 ␣) Gel to Liquid-Crystalline Phase Transition-The phase behavior was characterized by Fourier transform infrared (FT-IR) spectroscopy on a IFS55 (Bruker, Karlsruhe, Germany) using 10 mM lipid A and LPS suspensions isolated from HI100 and from E. coli K12 according to methods as described in the literature (29). The peak position of the symmetric stretching vibration of the methylene groups was taken as a measure of the acyl chain order, lying below 2850 cm Ϫ1 in the highly ordered gel phase (low fluidity) and above 2852 cm Ϫ1 in the less ordered liquid-crystalline phase (high fluidity).

LPS Isolation and SDS-PAGE-Extraction
of dried bacterial cells using a modified phenol/chloroform/light petroleum method gave LPS preparations, which were analyzed by SDS-PAGE (Fig. 1). The sample H2 and HI100 LPS showed characteristic patterns for S-form LPS (lanes 1, 4, and 5). The sample H1 and E. coli K12 LPS preparations showed profiles typical for R-form LPS (lanes 2, 3, and 6 -8). As can be deduced from the obvious similarity between H1 of HD100 (lanes 2 and 3) and the LPS of E. coli K12 (lanes 6 and 7) the waterprecipitated LPS fraction H1 contains mainly the R-form LPS of E. coli K12. In contrast, the ethanol-precipitated fraction H2 of HD100 (lanes 4 and 5) as well as the LPS of HI100 (lane 1) showed ladders of repeating units typical for S-form LPS. However, the distances of the bands are larger for the host-independent strain HI100 indicating a size difference of the repeating units of both strains. Small amounts of the LPS of B. bacteriovorus HD100 were also observed in H1 (lanes 2 and 3).
Immunological Characterization of LPS of B. bacteriovorus-For further characterization of the different LPS preparations three monoclonal antibodies (A20, A6, and WN1) were used for Western blots. The mAb A20 recognizes a terminal Kdo residue, mAb A6 reacts specifically with bisphosphorylated lipid A backbone, and mAb WN1 identifies the core region of E. coli K12 (Fig. 2). With mAb A20 (panel B) with water and ethanolprecipitated fractions H1 and H2 from strain HD100 (lane 6 and 1, respectively), LPS from strain HI100 (lane 3), and LPS from E. coli K12 LPS (lane 5) showed reactivity. This shows that B. bacteriovorus possess Kdo residues, which are accessible to mAb A20. However, it does not allow a differentiation of B. bacteriovorus LPS and that of E. coli K12. To detect residual host LPS in preparations of B. bacteriovorus strain HD100 mAb WN1 was used (Fig. 2, panel A). Strong reactivity was visible with the control LPS of K12 (lane 5) as well as with H1 of strain HD100 (lane 6). However, the fraction H2 of HD100 contained considerably less of the host LPS (lane 1). In contrast, purified LPS of the host-independent strain HI100 did not show any reaction with mAb WN1 (lane 3), indicating structural differences to the core region of E. coli. A structural difference of the lipid A backbones could be demonstrated with mAb A6 (panel C). In addition to E. coli K12 LPS (lane 5) and both B. bacteriovorus HD100 preparations (lanes 1 and 6), the purified lipid A of B. bacteriovorus HD100 showed a positive reaction with this mAb (lane 2). Neither the LPS of strain HI100 nor the isolated lipid A was recognized by mAb A6 (lanes  3 and 4). Taken together, these results indicate that B. bacteriovorus LPS possesses Kdo, that its core region and the lipid A backbone are different from that of E. coli, and that residual LPS from the host E. coli K12 is found in the HD100 LPS preparation. The free lipid A of B. bacteriovorus HI100 did not react with any available antibody (lane 4).
Mass Spectrometry of LPS of B. bacteriovorus-The negative ion MALDI LIN-TOF mass spectra of the complete LPS of B. bacteriovorus HD100 and B. bacteriovorus HI100 (Fig. 3, A  and B) show typical patterns of S-form LPS with a series of molecular ion peaks, representing LPS species with different numbers of repeating units (M 0 , M 1 , M 2 , M 3 , . . . ) and laserinduced in-source fragment ions originating from the cleavage of the labile linkage between lipid A and core oligosaccharides. Fragments representing the core oligosaccharide (O 1 , O 2 ) were identified by accompanying fragments (-44 Da) originating from decarboxylation of Kdo I . These results are in good agreement with the data from experiments using ␣-Kdo-mAb A20 (Fig. 2B). Although the spectra are not well resolved because of the heterogeneity and adduct ion formation, they reveal important data concerning the differences of LPS HD100 and HI100. The R-form LPS (M 0 ) can be assigned to the peaks at m/z 4759 (HI100) and m/z 4689 (HD100). The same mass differences were observed for the core oligosaccharides (O 1 ). The repeating unit of HI100 has an average mass of 950 Da whereas HD100 has an average mass of 716 Da. Laser-induced cleavage in the core oligosaccharide lead to fragments at m/z 3831 and m/z 3783 for HD100 and HI100, respectively. Furthermore, the host-dependent strain HD100 shows up to three peaks in the LPS population attributing to heterogeneity in the core oligosaccharide. High resolution ESI FT-ICR MS of the R-form LPS (M 0 ) gave a complex pattern of peaks differing by 14 Da combined with species at Ϫ2 and Ϫ4 Da indicating that the fatty acid composition must be very heterogeneous with respect to chain length and the degree of saturation (data not shown).
Chemical Analysis of B. bacteriovorus LPS (GC-MS)-Kdo, Glc, GlcN, and Hep were detected as part of the oligosaccharide portion of the complete LPS of B. bacteriovorus by GC-MS after methylation. The sugar portion of the lipid A was identified as 2,3-diamino-2,3-dideoxy-D-glucopyranose (GlcpN3N) and Dmannopyranose (Manp).
Fatty acid analysis revealed only hydroxylated fatty acids for both strains. The most abundant fatty acid component of B. bacteriovorus HD100 was (R)-3-hydroxy-11-methyl-dodecanoic acid (iso-C13:0(3-OH)). In small amounts unbranched 3-hydroxy fatty acids were detected. In the host-independent strain HI100 the major components could be identified as the branched and unbranched 13:0(3-OH) in nearly equal amounts. For both strains longer and shorter homologues were detected and the absolute configurations of all 3-hydroxy fatty acids were determined as R-form. Furthermore, in both strains additional dihydroxy fatty acids (14:0(3,4-OH) and 15:0(3,4-OH)) as well as unsaturated 3-hydroxy acids were present.
Because of rearrangement, reactions and the formation of byproducts during derivatization of the unsaturated and dihydroxy fatty acid molar ratios of the identified fatty acids could not accurately be determined. Thus a detailed study of the fragmentation of the lipid A was performed using mass spectrometry and NMR spectroscopy.
Mass corresponds to hexa-acylated lipid A carrying two or one hexose residues, respectively. The inlets in Fig. 4, A and B, represent enlargements of the respective molecular ion region each showing more than 14 different molecular peaks expressing the heterogeneity in the fatty acid composition. These findings were in good agreement with the data of the chemical component analysis. A comparison of the lipid As of both B. bacteriovorus strains revealed that the average of the fatty acids of wild type strain HD100 possess longer chains length (ϩ14, ϩ2 ϫ 14 Da) and a higher portion of double bonds (-2 Da) than HI100.
Lipid As from both strains were O-deacylated by hydrazine treatment and then mass analyzed. For both lipid As the groups of molecular species were reduced mainly by Ϫ424.6 Da (spectra not shown) demonstrating that two fatty acid (mainly (OH)-13:0) are ester-linked thus confirming that the backbone consists only of a GlcpN3N-disaccharide.
Capillary skimmer dissociation generated fragment ions of diagnostic importance (Fig. 5, A and B). By cleavage the glycosidic linkage between the two GlcpN3Ns fragment ions (B Hexa , B Penta ) were formed representing the non-reducing GlcpN3N II moiety (B-fragments according to the nomenclature of Domon and Costello (30)) followed by the subsequent loss of hexose and, to a lesser extent, also by the cleavage of one (OH)-13:0 fatty acid (-230 Da, peaks at m/z 800). From the masses of the B-fragments and known fatty acid composition of the GC-MS it is evident that the non-reducing GlcpN3N II of the hexa-acylated lipid A carries four residues, the reducing GlcpN3N I only two fatty acid residues. Furthermore, a comparison of the Bfragments (B Hexa -Hex) of the two lipid As missing the hexose (enlargements shown in Fig. 5, C and D) demonstrate that both samples comprise identical fragment ions, however, with different intensities. An unambiguous determination of the four fatty acid residues linked to GlcpN3N II is not possible. Therefore, the molecular masses of all possible GlcpN3N I , carrying two fatty acid residues besides hexose, were calculated from the mass differences between all measured molecular ions (Fig.  4A) and the B-fragments observed in Fig. 5C. These differences were compared with the masses calculated of all combinations of fatty acids detected by the component analysis (see Table I).
Most possible combinations consist of one dihydroxy fatty acid and one hydroxy fatty acid. Based on this calculation and results of O-deacylation by hydrazine treatment it can be concluded that four hydroxylated fatty acids are linked to GlcpN3N II . Furthermore, up to two double bounds are detectable in the B-fragments (Fig. 5, C and D).
To prove which fatty acid combinations are realized, MS/MS experiments were performed. As an example, the infrared multiphoton dissociation-MS/MS spectra of the hexa-acyl lipid A molecular ions at m/z 1992 and 1952 from strains HD100 and HI100 are given in Fig. 6, A and B, respectively. The enlargements of the B Hexa -Hex fragment regions clearly demonstrate that in both cases three prominent fragments at m/z 1009.81, 1021.81, and 1035.82 were generated that differ in the acyl chain length and correspond to the expected B-fragments deduced from the reducing ends given in Table I. The enlargements of the isolated molecular ion regions show that also species with one and two unsaturated acyl chain bonds were selected as parent ions in small quantities.
NMR Spectroscopy of Lipid A-The lipid A configurations of both B. bacteriovorus strains were studied by NMR spectroscopy at 310 K using about 5.0 mg of lipid A in 7:3 chloroformd/methanol-d 4 (v/v) (31). The backbone of both compounds was found to be identical but the fatty acid substitution pattern is slightly different.
The backbone consists of two ␤-GlcpN3N with an (1Ј 3 6) interglycosidic linkage. ␤-Anomeric conformations in both glycoside rings were determined by the 3 J H1-H2 coupling constants of the anomeric protons, being 8.0 and 8.2 Hz in GlcpN3N I and GlcpN3N II , respectively, and by the 1 J C-H coupling constants of the anomeric protons, both being 163 Hz. Proton H-1Ј in HD100 shows two signals at 4.385 and 4.381 ppm, indicating two stable conformations of this lipid A. 3 J H-H couplings to subsequent protons in both GlcpN3N were determined from DQF-COSY and TOCSY spectra. As all coupling constants are in the range of about 9 Hz, all proton in the two pyranose rings are in axial orientation. Carbon chemical shifts (Table II) confirm these findings and indicate the presence of acylated amino groups in positions 2 and 3 in both rings. The (1Ј 3 6) interglycosidic linkage was determined from the downfield shift of C-6 in GlcpN3N I , an ROE between H-1Ј and H-6a/b, and two 3 J C-H couplings between C-6 and H-1Ј as well as between C-1Ј and H-6a.
Furthermore, the backbone carries two ␣-Manp linked to positions O-1 and O-4Ј. ␣-Anomeric conformations of these two mannoses have been determined from the 1 J C-H coupling constants, being 173 and 170 Hz in Manp I and Manp II , respectively. The other protons in the two mannoses were assigned by DQF-COSY and TOCSY spectra. The ␤-GlcpN3N I -(1 3 1)-␣-Manp I linkage was identified by 3 J C-H couplings from H-1 in Manp I to C-1 in GlcpN3N I and from H-1 in GlcpN3N I to C-1 in Manp I . The ␣-Manp II -(1Ј 3 4Ј)-␤-GlcpN3N II linkage was deduced by the 3 J C-H couplings between H-1 in Manp II and C-4Ј in GlcpN3N II and between the respective coupling between H-4Ј and C-1Ј. Both linkages were confirmed by the ROEs between the respective protons. Summed 1 H-and 13 C-chemical shifts of the tetrasaccharide are listed in Table II. Investigations with one-dimensional 31 P NMR revealed no signal in the lipid As of both strains HD100 and HI100.
As derived from the mass spectrometry the acylation pattern of both lipid As consisted of four amide-linked primary and two ester-linked secondary fatty acids. NMR spectroscopic results identify the latter as 3-hydroxy fatty acids, which are located on the GlcpN3N II . In strain HD100 one of these two is partially unsaturated (ϳ40% Fa bb ). The primary fatty acids on GlcpN3N II are one esterified 3-hydroxy fatty acid and one partially (ϳ80% The exact acylation pattern of both B. bacteriovorus strains cannot been assigned by NMR, as a direct identification of all ester and amid linkages by HMBC was hindered by fast transversal relaxation (31). Determination by ROESY and COSY spectra in Me 2 SO-d 6 as used for a hepta-acyl lipid A from a Salmonella enterica strain (32) was impossible because of low solubility of the investigated lipid As. However, taking other analytical results into account the fatty acids of strain HD100 can be identified and the chemical shifts are assigned (Table  III, part A). In the HI100 lipid A only the chemical shifts of the fatty acid Fa bc differed significantly from the shifts in the HD100 lipid A (Table III,

part B).
Summary of Structural Analysis- Fig. 7 shows probable molecular species (1992. 38  HD100 and the host-independent derivative HI100. Minor differences are only found in the fatty acid portion as previously mentioned. Release of Cytokines from Human Mononuclear Cells-To quantify the potential biological effects of the described LPS and lipid A preparations, endotoxin-induced TNF-␣ and interleukin-6 release of hMNC were measured. The stimulation experiments clearly show differences in the amount of cytokines released by E. coli K12 LPS on the one hand and lipid A and LPS of B. bacteriovorus on the other hand. LPS and lipid A of E. coli F515 served as reference. The TNF-␣ release of three 10-fold dilutions of stimuli (0.1, 1, and 10 ng/ml) is shown in Fig. 8. In the case of all B. bacterivorus components the response of the hMNC is heavily dependent on the stimuli concentration. In contrast, in the case of E. coli K12 even the lowest LPS concentration resulted in a TNF-␣ release near the maximum level achievable under the chosen conditions. Because of these results up to a 100-fold decrease of TNF-␣ release can be assumed for the LPS and lipid As of both B. bacteriovorus strains. The stimulation experiments with B. bacterivorus components also gave a decreased interleukin-6 response (data not shown) in contrast to E. coli LPS confirming a lower potential for inducing toxic effects on mammalian cells.
Determination of (␤ 7 ␣) Gel to Liquid Crystalline Phase Transition-In FT-IR experiments we evaluated the temperature-dependent band position of the symmetric stretching vibration within liposomes made from HD100 lipid A and LPS in comparison to E. coli K12. In contrast to the pronounced gel to liquid crystalline (␤ 7 ␣) phase transition observed for lipid A and LPS of E. coli K12 (T c ϭ 46 and 35°C, respectively), a broad transition was detected for HD100 LPS at T c ϭ 10°C. The acyl chains of HD100 lipid A remained down to even Ϫ5°C in the liquid crystalline state. These results clearly show that the fluidity of the acyl chains within membranes made from B. bacteriovorus LPS is considerably higher than that within E. coli K12 membranes. DISCUSSION The structural characterization of the lipid As of B. bacteriovorus strains HD100 and HI100 revealed a high similarity. In both cases the lipid As showed the same backbone consisting of ␣-D-Manp II -(1 3 4)-␤-D-GlcpN3N II -(1 3 6)-␤-D-GlcpN3N I -(1 3 1)-␣-D-Manp I . The asymmetric acylation pattern (Fig. 7) of both lipid As concordantly comprised five hydroxy fatty acids and one dihydroxy fatty acid, whereby on average two of the six fatty acids were unsaturated.
These structural features are unusual because of the absence of negatively charged residues within the polar head group of the lipid As influencing not only the biophysical properties of the outer membrane (33) but also causing a decreased endotoxic activity (34 -36). Therefore, we additionally investigated the ability of B. bacteriovorus LPS and lipid A to induce cytokine production in hMNC and determined the fluidity of the acyl chains within liposomes made from strain HD100 lipid A and LPS. Studies concerning the endotoxic activity of B. bacteriovorus LPS by stimulation of hMNC confirmed a low induction level of TNF-␣ (Fig. 8) and interleukin-6 (data not shown) for both strains. The amount of stimuli had to be 2 orders of magnitude higher for B. bacteriovorus than E. coli to obtain a comparable release of cytokines. This significant decrease of biological activity was attributed to lower binding affinity to LPS receptors of human cells (34 -36)

Structure of B. bacteriovorus Lipid A
chain length of the fatty acids were comparable with the potent cytokine inducer E. coli hexa-acyl lipid A. As expected, the induction level of TNF-␣ for the complete S-form LPS of B. bacteriovorus HD100 and HI100 was higher compared with pure lipid A preparations (37). With respect to the uncharged backbone of strain HD100 lipid A the fluidity of acyl chains within liposomes was affected. The results revealed that the acyl chains within HD100 LPS are in a significantly less ordered state most probably because of the missing negative charges inhibiting cation bridging within the outer membrane and because of the presence of at least two unsaturated acyl chains per lipid A molecule. This increased fluidity may lead to an increased permeability of HD100 outer membrane (38) meaning an advantage for consumption of degraded prey components during the intraperiplasmatic growth phase at a temperature range typical for soil and aquatic habitats.
It should be emphasized that our experiments cannot answer the question if B. bacteriovorus integrates complete lipid A or LPS molecules of the prey bacteria into its cell wall (39). We structurally determined the LPS of HD100 after completing its   (11) showing no membrane fusion process during the invasion of B. bacteriovorus with the outer membrane of the prey. It was proposed that B. bacteriovorus enzymatically hydrolyzes LPS of prey bacteria. However, the presence of high amounts of unmodified LPS of preys after complete degradation of the prey bacteria by B. bacteriovorus suggests that this does not occur to a great extent and may be restricted to the invasion process (12,15). Furthermore, we microscopically observed complete cell hulls of former prey bacteria suggesting that the outer membrane of the prey bacteria rather provides a protected environment (10) (18) could not be confirmed. We assume that an insufficient isolation procedure led to a contamination of the preparations with phospholipids, as we also noticed traces of phosphatidylethanolamine in the lipid A preparation of B. bacteriovorus HI100. Dissimilarities between the two B. bacteriovorus lipid As were different portions of iso-fatty acids and unsaturated fatty acids. In strain HD100 statistically five of the six fatty acids were branched, whereas strain HI100 only possessed about 3.6 branched fatty acids. Furthermore, a double bond in the secondary fatty acid Fa b (Table III, Fig. 7), which was present in about 40% of both lipid As was at a different position. Whereas the double bond was located between C-9 and C-10 in strain HI100, its position was between C-5 and C-6 in strain HD100.   (31), which was confirmed by NOEs from H-1Ј to H-6a/b and H-5. In strain HI100 that possessed the double bond in a different position, these dissimilar conformations could not be observed. Furthermore, the amount of unsaturated fatty acids was about 50% lower in strain HI100 than in strain HD100. With regard to these minor differences it can be assumed that wild type strain HD100 synthesizes an innate lipid A without generating hybrid forms by recycling components of the preys lipid A. The high portion of iso-fatty acids in strain HD100 further indicates that they are not consumed from the preys LPS because only unbranched fatty acids are known for E. coli K12 (40). How can the differences between the structure of the LPS of host-dependent strain HD100 and the host-independent strain HI100 be interpreted? To isolate host-independent strains of  4Ј. B, structure and distribution of the fatty acids for strain HD100. A 3-hydroxy-11-methyldedocanoic acid and a 3-hydroxy-11-methyldedodec-7-enoic acid (80% unsaturated) are amide linked to position 2Ј and 3Ј in ␤-D-GlcpN3N II . The two hydroxy functions of these primary acids were esterified with a 3-hydroxy-11-methyldodecanoic acid and a 3-hydroxy-11-methyldodec-5-enoic acid (40% unsaturated). Positions 2 and 3 in ␤-D-GlcpN3N I were substituted with a 3,4-dihydroxy-13-methyltetradecanoic acid and a 3-hydroxy-tetradecanoic acid. Chain length and branching of the fatty acids representing their predominant distribution were derived from the GC-MS analysis, infrared multiphoton dissociation-MS/MS, and NMR. The exact position of each fatty acid could not be determined from the analytical results.
B. bacteriovorus the selection of streptomycin-resistant mutants is necessary. Additionally, several more passages are necessary before the mutants are able to grow on enriched media (13). During this procedure the mutants probably acquire several mutations that enable them to grow without a host. Host-independent mutants show morphologically distinctive features like spheroplast forming, long spiral-shaped cells, and pigmentation (14). Although a genetic locus (hit) was identified that is involved in the conversion from the predatory to the host-independent lifestyle, all authors are in agreement that this phenotype cannot be explained by a single mutation in the hit locus (13,14,42). Our hypothesis is that the differences in the oligosaccharide portion of the HI100 LPS and HD100 LPS are caused by the selection procedure (13). It is likely that streptomycin resistance of HI strains is caused by alterations of the ribosomal protein S12 that interferes with protein synthesis leading to pleiotropic changes (43,44). As a consequence the proteins for the biosynthesis of the outer membrane might be affected causing morphological aberrations in host-independ-ent mutants. However, the main structural features of the HI100 lipid A are conserved in comparison to wild type strain HD100.
The unique predatory lifestyle of B. bacteriovorus requires the evolutionary development of special structures. The lipid A that was found in B. bacteriovorus and is described here for the first time may represent such a unique structure. Analyses of other Gram-negative bacteria living in highly specialized environments have indicated that an unusual lipid A structure is important for the lifestyle and may serve as an evolutionary marker. It was shown that in Aquifex pyrophilus, living in hot springs, the conserved phosphate substituents of the lipid A are replaced by galacturonic acid residues (45). Chlamydia trachomatis, an obligate intracellular pathogen, comprises a highly heterogeneous LPS in regard to fatty acid composition (41). This may be a parallel to the intracellular growth of B. bacteriovorus inside the prey bacteria. The correlation between an unusual lipid A structure and a specialized environment can now be extended to B. bacteriovous.