Structural determination of a 5-acetamido-3,5,7,9-tetradeoxy- 7(3-hydroxybutyramido)-L-glycero-L-manno-nonulosonic acid- containing homopolysaccharide isolated from Sinorhizobium fredii HH103

Biochem. J. (1999) 342, 527–535 (Printed in Great Britain) 527 Structural determination of a 5-acetamido-3,5,7,9-tetradeoxy-7- (3-hydroxybutyramido)-L-glycero-L-manno-nonulosonic acid-containing homopolysaccharide isolated from Sinorhizobium fredii HH103 Antonio M. GIL-SERRANO*1, Miguel A. RODRI ! GUEZ-CARVAJAL*, Pilar TEJERO-MATEO*, Jose !L. ESPARTERO†, Manuel MENENDEZ*, Javier CORZO‡, Jose !E. RUIZ-SAINZ§and Ana M. BUENDI ! A-CLAVERI ! A§ *Departamento de Quı!mica Orga !nica, Facultad de Quı!mica, Universidad de Sevilla, 41071 Sevilla, Spain, †Departamento de Quı!mica Orga !nica y Farmace !utica, Facultad de Farmacia, Universidad de Sevilla, 41071 Sevilla, Spain, ‡Departamento de Bioquı!mica y Biologı!a Molecular, Facultad de Biologı!a, Universidad de La Laguna, 38206 Tenerife, Spain, and §Departamento de Microbiologı!a, Facultad de Biologı!a, Universidad de Sevilla, 41080 Sevilla, Spain. The structure of a polysaccharide from Sinorhizobium fredii HH103 has been determined. This polysaccharide was isolated by following the protocol for lipopolysaccharide extraction. On the basis of monosaccharide analysis, methylation analysis, fast atom bombardment MS, matrix-assisted laser desorption ionization MS, electron-impact high-resolution MS, onedimensional "H-NMR and "$C-NMR and two-dimensional NMR experiments, the structure was shown to consist of a homopolymer of a 3:1 mixture of 5-acetamido-3,5,7,9-tetradeoxy-7-[(R)- and (S)-3-hydroxybutyramido]--glycero--mannoINTRODUCTION Sinorhizobium fredii is a Gram-negative soil bacterium that induces nodules on the roots of various legumes, such as Glycine max (soya bean). Within these nodules the bacteria, differentiated into bacteroids, fix atmospheric nitrogen. The development of legume nodules is controlled largely by reciprocal signal exchange between the symbiotic partners. Legume roots secrete specific flavonoids that induce the transcription of bacterial genes essential for nodule formation. Many nodulation (nod) genes are involved in the synthesis and secretion of lipochitin oligosaccharides, known as LCOs or Nod factors. These bacterial signal molecules trigger the nodule development programme of the leguminous plant [1]. In addition to bacterial genes involved in nitrogen fixation (nif and fix genes), other bacterial determinants are required for the formation of a functional nodule. These determinants include certain polysaccharides that are present in the bacterial cell wall: β-glucans, acidic exopolysaccharides, capsular polysaccharides (CPSs) and lipopolysaccharides (LPSs) [2]. The cell surface of Rhizobium, as in other Gram-negative bacteria, consists of a cytoplasmic (inner) membrane and an outer membrane, separated by the periplasmic space containing the peptidoglycan. The outer membrane consists of proteins, LPSs and phospholipids. β-Glucans are located mainly in the periplasmic space. The inner membrane is composed of phospholipids and lipoproteins. The extracellular matrix of the bacterium (commonly termed the capsule) is composed of an acidic exopolysaccharide that is released into the cell’s milieu. In the genus Sinorhizobium, the outer membrane and the surrounding capsule Abbreviations used: CPS, capsular polysaccharide; FAB, fast atom bombardment; HSQC, 1H-detection mode via single-quantum coherence; Kdo, 3deoxy-D-manno-2-octulosonic acid; Kdx sugar, 1-carboxy-2-oxo-3-deoxy sugar; LPS, lipopolysaccharide; MALDI–TOF-MS, matrix-assisted laser desorption ionization–time-of-flight MS. 1To whom correspondence should be addressed (e-mail agil!cica.es). nonulosonic acid. The sugar residues are attached via a glycosidic linkage to the OH group of the 3-hydroxybutyramido substituent and thus the monomers are linked via both glycosidic and amidic linkages. In contrast with the Sinorhizobium Kantigens previously reported, which are composed of a disaccharide repeating unit, the K-antigen polysacharide of S.fredii HH103 is a homopolysaccharide. Key words: capsular polysaccharide, K-antigen, pseudaminic acid. are composed in part of a complex array of LPS and a 3-deoxy- -manno-2-octulosonic acid (Kdo)-containing polysaccharide that is structurally analogous to the constituents of the group II K-antigen polysaccharides of Escherichia coli [3]. The structures of different K-antigens of several strains of S.fredii and Sinorhizobium meliloti have been reported previously [3–5]. Although possessing structural variability, it seems to be a conserved structural motif of a disaccharide repeating unit involving a variable glycosyl residue bonded to a 1-carboxy-2-oxo-3-deoxy sugar (Kdx). In all these structures studied, the Kdx sugar is either Kdo (in S.fredii strains) or a type of nonulosonic acid (in S.meliloti AK631, Rhizobium sp. NGR234 or S.fredii SVQ293) [3–5]. Kannenberg et al. [3] have proposed a consensus structure [(R")sugar (R#)Kdx ]nfor most of the K-antigen CPSs of Sinorhizobium strains. Here we show that S.fredii strain HH103 produces a Kantigen CPS that is not in accordance with the consensus structure. It consists of a homopolymer of a 3:1 mixture of 5-acetamido-3,5,7,9-tetradeoxy-7-[(R)-and (S)-3-hydroxybutyramido]--glycero--manno-nonulosonic acid. The sugar residues are attached via a glycosidic linkage to the OH group of the 3hydroxybutyramido substituent; monomers are thus linked via both glycosidic and amidic linkages. Different pseudaminic acid derivatives have been described previously as components of the polysaccharides isolated from Pseudomonas aeruginosa,Shigella boydii and Vibrio cholerae [6–11] but this is the first report of the completely determined structure of a pseudaminic acid-containing homopolysaccharide in a rhizobial bacterium. Although different homopolysaccharides composed of a higher sugar of this class have been reported [12–15], and a polysaccharide with #1999 Biochemical Society 528 A. M. Gil-Serrano and others this class of glycosidic linkage has been described [9], this is the first report of a homopolymer of pseudaminic acid derivative with this type of glycosidic linkage. EXPERIMENTAL General methods GLC–MS was performed with a Micromass AutoSpec-Q instrument fitted with an OV-1 column (25 mi0.25 mm). The temperature programme for separating the trimethylsilylated methyl glycosides was isothermal at 150 mC for 2 min, followed by a 10 mC\min gradient up to 250 mC, whereas that for the partly methylated alditol acetates was isothermal at 120 mC for 1 min, followed by an 8 mC\min gradient up to 250 mC. The protocol for the trimethylsilylated 2-octyl esters of 3-hydroxybutyric acid was isothermal at 70 mC for 3 min, followed by a gradient of 10 mC\min up to 170 mC and then 25 mC\min up to 280 mC. The ionization potential was 70 eV and the mass spectra were performed with both low-resolution and high-resolution electronimpact MS. Positive-ion fast atom bombardment MS (FAB-MS) spectra were obtained with a Kratos MS80RFA instrument with a xenon atom beam at an accelerating voltage of 4 kV. The probe was loaded with 1 µl of sample solution in water with a matrix of glycerol. Positive-ion matrix-assisted laser desorption ionization– time-of-flight MS (MALDI–TOF-MS) was performed with a Micromass TofSpec mass spectrometer equipped with a 337 nm nitrogen laser, with 2,5-dihydroxybenzoic acid as matrix. A mixture of maltose oligosaccharides was used as an external calibration standard. Sugars were identified on GLC–MS separation of their trimethylsilylated methyl glycosides obtained as described [5]. The absolute configuration of 3-hydroxybutyric acid was assigned after the GLC–MS analysis of their trimethylsilylated (S)- and (R,S)-2-octyl esters. The polysaccharide was hydrolysed with 2 M trifluoroacetic acid for 1 h at 120 mC; 3-hydroxybutyric acid was extracted with ethyl acetate and esterified by heating at 140 mC overnight in the presence of a drop of trifluoroacetic acid [16]. The octyl ester was trimethylsilylated conventionally. Optical rotation was measured with a Perkin-Elmer model 241 MC polarimeter at 25 mC. Monosaccharides and oligosaccharides were obtained by mild acid hydrolysis of the polysaccharide with 1%(v\v) acetic acid at 100 mC for 1 h, followed by gel-permeation chromatography on Sephadex G-50 and G-15, with water as eluent. Autohydrolysis of fraction F3 was performed by heating a1%(v\v) aqueous solution at 60 mC for 2 h. Bacterial strains and isolation of the polysaccharide Wild-type S.fredii HH103 was grown routinely at 28 mCinTY medium as described by Beringer [17]. When required, media were supplemented with streptomycin (400 µg\ml). Earlystationary-phase cultures (300 ml, TY medium) of S.fredii HH103 were used to inoculate 30 litres of TY liquid medium supplemented with the antibiotic streptomycin. After inoculation, cultures were incubated for 3 days at 28 mC. Bacterial cultures were aerated by pumping sterile air through at a flow of 50 l\min. After incubation, the cells were harvested by low-speed centrifugation. The bacterial pellets were washed with 0.9% NaCl, freeze-dried and stored in sealed bottles at room temperature. The polysaccharide was extracted from the freeze-dried bacterial cells (5 g) with 100 ml of hot phenol\water (1:1, v\v) [18] and the two phases were separated. The aqueous phase was dialysed against water, concentrated and passed through an IRA 400 Amberlite (AcO−) anion-exchange resin column (30 cmi1.6 cm), with water as eluent. The eluate was freezeFigure 1 Silver-stained (A) and Alcian Blue/silver-stained (B) SDS/PAGE of extracts of S. fredii HH103 and comparison with LPS from Salmonella typhimurium Lanes 1, whole-cell extracts; lanes 2, F1 (very-high-molecular-mass fraction); lanes 3, F2 (highmolecular-mass fraction); lanes 4, F3 (lower-molecular-weight fraction obtained by GPC of the polysaccharide); lanes 5, LPS from Salmonella typhimurium. dried and redissolved in 10 mM MgSO%\50 mM Tris\HCl (100 ml, pH 7.0); DNase (1 mg) and RNase A (1 mg) were added and the solution was stirred overnight at 5 mC. Proteinase K (10 mg) was added; the solution was shaken for 24 h at 37 mC, dialysed and then freeze-dried. The polysaccharide was chromatographed on Sephacryl S-500 (60 cmi2.6 cm) with 0.2 M NaCl as eluent; carbohydrates were detected with a refractive-index detector and the orcinol\sulphuric acid method on TLC plates. Fractions containing carbohydrates were dialysed and freezedried. SDS/PAGE Three different protocols have been used to analyse the bacterial LPS and K-antigen polysaccharide profiles. SDS\PAGE electrophoresis of crude bacterial extracts (Figures 1A and 1B, lanes 1) was performed as described previously [5]. In brief, bacterial cultures of S.fredii HH103 were grown on solid TY medium. Bacterial cells were resuspended in 0.9%NaCl and pelleted by centrifugation. The bacterial pellet was resuspended and lysed by heating at 100 mC in 125 µl of 60 mM Tris\HCl (pH 6.8)\2% (w\v) SDS\1 mM EDTA for 5 min and then diluted to 1 ml with the same buffer without SDS. The bacterial crude extract was treated with RNase, DNase and proteinase K as described by Ko $plin et al. [19]. The electrophoresis of crude bacterial extracts (Figures 1A and 1B, lanes 1) or purified bacterial polysaccharide (Figures 1A and 1B, lanes 2–4) was performed on a 16.5% (w\v) polyacrylamide gel with the Tricine buffer system described by Lesse et al. [20]. For the detection of LPSs, gels (Figure 1A) were stained with silver as described by Kittelberger and Hilbink [21]. For the detection of LPSs and K-antigens, gels (Figure 1B) were fixed with Alcian Blue as described by Corzo et al. [22] and then stained with silver by the method described by Tsai and Frasch [23]. PAGE without detergent was also performed to facilitate the detection of K-antigens (Figure 2). Under the same conditions, only the K-antigens (and not the LPS) can migrate into the gel. Samples of the purified K-antigen polysaccharide were analysed by vertical electrophoresis (Hoefer-Pharmacia Biotech), as described for extracellular polysaccharides from Bradyrhizobium #1999 Biochemical Society 529Structure of a polysaccharide isolated from Sinorhizobium fredii HH103 Figure 2 Alcian Blue/silver-stained PAGE, densitometric tracing and band numbering of fraction F3 (lower molecular mass) performed in the absence of SDS The absence of SDS meant that only K-antigen non-hydrophobic polysaccharides could migrate into the gel. The electrophoresis was stopped when the Bromophenol Blue reached 10 mm from the bottom of the gel. No bands were detected below band 1. [24]. A continuous system was employed, with slab gels 0.75 mm thick. The acrylamide concentration was 17%(w\v), the acrylamide-to-N,Nh-methylenebisacrylamide ratio being 30:0.8. The electrophoresis buffer was 50 mM Tris\HCl (pH 8.5)\ 13 mM EDTA\15 mM boric acid. Samples were dissolved in the same buffer and diluted (1:2) in 1 M sucrose in deionized water. Gels were fixed by Alcian Blue in acetic acid [22] and stained with silver [23]. After being dried, the stained gels were scanned with a Pharmacia-LKB Ultroscan XL laser densitometer. Methylation The vacuum-desiccated sample of monosaccharide was carbonylreduced by dissolving in 1 M NH%OH (400 µl) containing 10 µg\µl NaB#H%for 2 h at room temperature. The borate was eliminated with acetic acid and by co-evaporation with methanol. The resulting polyol was esterified with 0.625 M HCl in methanol at 20 mC for 2.5 h and then carboxy-reduced by treatment with 2 mg of NaBH%dissolved in 500 µl of ethanol\water (3:1, v\v) at room temperature overnight. Finally, the carboxy-reduced polyol was methylated by the method of Ciucanu and Kerec [25]. For the oligosaccharide, the sample was carbonyl-reduced, methylated and carboxy-reduced as above, then hydrolysed with 1%(v\v) trifluoroacetic acid for 1 h at 100 mC and finally reduced and acetylated. NMR spectroscopy The samples were deuterium-exchanged several times by freezedrying from #H#O and then examined in solution (4 mg\ml) in 99.98%#H#O. Spectra were recorded at 303 K on a Bruker AMX500 spectrometer operating at 500.13 MHz ("H) and 125.75 MHz ("$C). Chemical shifts are given in p.p.m., with the H#HO signal (4.75 p.p.m.) ("H) and external DMSO (39.5 p.p.m.) ("$C) as references. In one experiment, "$C data were recorded with inverse gated decoupling to measure the integral of the "$C signals correctly. The two-dimensional homonuclear proton double-quantum-filtered COSY experiments were performed in the phase-sensitive mode with the Bruker standard pulse sequence. A data matrix of 512i1024 points was used to digitize a spectral width of 4500 Hz; 32 scans were used per increment with a delay between scans of 1 s. Before Fourier transformation, zero-filling was used in F"to expand the data to 1024i1024 points. The two-dimensional heteronuclear one-bond proton– carbon COSY experiment was registered in the "H-detection mode via single-quantum coherence (HSQC). A data matrix of 256i1024 points was used to digitize a spectral width of 4000 and 20000 Hz in F#and F"; 64 scans were used per increment with a delay between scans of 1 s and a delay corresponding to aJvalue of 150 Hz. "$C decoupling was achieved by the globally optimized alternating-phase rectangular pulses (‘GARP’) scheme. Squared cosine-bell functions were applied in both dimensions and zero-filling was used to expand the data to 512i1024 points. This experiment was slightly modified by the implementation of an editing block in the sequence [26]. The two-dimensional HSQC-TOCSY spectra were recorded with the conventional phase-cycle HSQC pulse sequence concatenated to a WALTZ 16-pulse train with a spin-locking πpulse of 98 µs. Two spectra were recorded with 20 and 140 ms of isotropic mixing time respectively. RESULTS Isolation and purification of the polysaccharide The polysaccharide was extracted from bacterial pellets of the parental strain S.fredii HH103 with hot phenol\water, treated with different enzymes and finally purified and fractionated by gel-permeation chromatography. Three fractions were isolated on the basis of their molecular masses: a very high-molecularmass fraction, F1, which was excluded from Sephacryl S-500, a high-molecular-mass fraction, F2, which was partly excluded from S-500, and a lower-molecular-mass fraction, F3, which was further chromatographed on Sephacryl S-400 with the same eluent and whose molecular mass was estimated to be 26 kDa. These fractions were analysed for the presence of LPS and\or Kantigens by using two different protocols for PAGE. SDS\PAGE with the Tricine buffer system and the silver-staining method as described by Kittelberger and Hilbink [21] were used to stain and detect the LPS component of the samples. The second method included a fixation treatment with Alcian Blue before LPS and K-antigens were stained with silver as described [23]. The treatment with Alcian Blue was developed by Corzo et al. [22] to detect K-antigen polysaccharides. Apparently, the fixation step with Alcian Blue prevents the escape of K-antigens from the gel so that they can be detected by silver staining. Figure 1 shows that fractions F1 (lanes 2) and F2 (lanes 3) still contained LPS, which gives a ladder pattern resembling the characteristic LPS profile of the crude cell-wall extract of strain HH103 (lanes 1). The higher mobility of the bands in samples subjected to purification procedures (Figures 1A and 1B, lanes 2 and 3) was probably due to partial hydrolysis of the LPS. The upper part of those lanes containing S.fredii samples (Figure 1B, lanes 1–3) in the gel treated with Alcian Blue were more heavily stained than the corresponding pairwise lanes (Figure 1A, lanes 1–3) in the gel without Alcian Blue. This difference in staining intensity was probably due to the differences in the silverstaining protocols used and\or the presence of high-molecularmass material (probably K-antigens) that did not escape from the gel because of the fixation with Alcian Blue. #1999 Biochemical Society 530 A. M. Gil-Serrano and others Figure 3 1H-NMR, 13C-NMR and multiplicity-edited two-dimensional HSQC experiment at 500 MHz obtained for fraction F3 Positive peaks (solid lines) correspond to methine and methyl groups; negative peaks (broken lines) correspond to methylenic groups. The numbers followed with an apostrophe correspond to the 3-hydroxybutyryl moiety. The lower-molecular-mass fraction (F3) apparently did not contain significant amounts of LPS, because the LPS ladder was absent. No material was silver-stained in the gel that was not treated with Alcian Blue (Figure 1A, lane 4) and only a weakly stained smear was visible in the gel treated with Alcian Blue (Figure 1B, lane 4). NMR analyses of fraction F3 did not detect the presence of LPS; the major component was another polysaccharide. Because of its apparent homogeneity, fraction F3 was chosen for NMR studies. Fractions F2 and F3 were used for sugar and methylation analyses. NMR analysis of the polysaccharide The "H-NMR and "$C-NMR spectra (edited HSQC in Figure 3) of the homogeneous lower-molecular-mass fraction F3 showed two signals (at δC18.3 and 19.9 p.p.m.) that revealed the presence of two methyl groups, and another (at δC23.4 p.p.m., correlated with the singlet at δH2.01 p.p.m.) that was assigned to the methyl group of an acetyl group. The methylenic nature of the carbons at δC37.2 and 45.2 p.p.m. is demonstrated unequivocally by the negative sign (broken lines) of their one-bond proton–carbon correlation cross peaks in this spectrum. The signals at δC49.8 and 55.1 p.p.m. were tentatively assigned to methine carbons carrying nitrogen. In the region of δC65–75 p.p.m. four carbon signals appeared; these were correlated with "H signals in the region of δH4–4.5 p.p.m. and were tentatively assigned to methine carbons linked to oxygen. The broad signal at δC99.7 p.p.m. (which showed no proton correlation cross-peak) was assigned to a ketalic anomeric carbon. In addition, the low-field region of the "$C spectrum (results not shown) contained signals at δC 173.7, 175.7 and 178.0 p.p.m., which corresponded to three carbonyl carbons. On the basis of the COSY and HSQC-TOCSY experiments, the methyl group at δC18.3 and δH1.3 p.p.m. was correlated with at least three "H signals in the δH4–4.5 p.p.m. region and was therefore identified as the terminal methyl group of a deoxy #1999 Biochemical Society 531Structure of a polysaccharide isolated from Sinorhizobium fredii HH103 Figure 4 Positive-ion MALDI–TOF mass spectrum of the autohydrolysate of fraction F3 The indicated mass value over each peak corresponds to an average for all the ions in the cluster. The abscissa scale is in units of m/z. sugar. The methylenic group at δC37.2, δH1.52 and 2.11 p.p.m. was correlated with two "H signals in the same region as before. These two connectivity pathways are representative of the C-3 to C-9 carbon framework of a 3,9-dideoxynonulosonic acid. In addition, the methyl group at δC19.9 and δH1.21 p.p.m. was correlated with the methylenic signal at δC45.2, δH2.43 and 2.57 p.p.m. and with a methine group at δC68.0 and δH4.2 p.p.m. This set of signals was identified as corresponding to a 3hydroxybutyryl moiety. These data allowed us to conclude that the analysed sample consists of a homopolymer of a 3,9-tetradeoxydiaminononulosFigure 5 1H-NMR spectrum obtained at 500 MHz for the monosaccharide The inset shows the δl4.0–4.3 p.p.m. zone. Signals marked with an asterisk correspond to redisual ethanol. onic acid bearing an acetyl group and a 3-hydroxybutyryl group as substituents. Sugar analysis GLC–MS analysis of the trimethylsilylated methyl glycosides obtained after methanolysis of the heterogeneous highmolecular-mass fraction F2 revealed that the fraction contained glucose, Kdo and traces of galacturonic acid, galactose, mannose, glucuronic acid and rhamnose (results not shown). However, no major chromatographic peak was found arising from the main component of the polysaccharide identified by NMR, probably owing to the known lability of this kind of compound under acidic conditions [16,27]. FAB-MS and MALDI–TOF-MS analyses To determine fully the structure of the monosaccharidic repeating unit, a mild hydrolysis [1%(v\v) acetic acid] was performed and the monosaccharide was isolated by gel-permeation chromatography. On positive-ion FAB-MS analysis of the monosaccharide, [MjH]+,[MjNa]+and [Mj2NakH]+pseudomolecular ions were observed at m\z379, 401 and 423 respectively. The observed mass was thus consistent with an acetyl, 3-hydroxybutyryl derivative of a tetradeoxydiaminononulosonic acid. In addition, positive-ion MALDI–TOF mass spectra of the fraction F3 and its autohydrolysate were performed. The two spectra were very similar. Figure 4 shows the spectrum of the hydrolysate. Clusters of ions arising from different sodium salts and adduct ions were detected at the following mass values: m\z 401, assigned as before to the sodium adduct of the monomer (reported m\z400); m\z761, sodium adduct of the dimer (reported m\z762); m\z1165, sodium adduct of the disodium salt of the trimer (reported m\z1165); m\z1547, sodium adduct of the trisodium salt of the tetramer (reported m\z1547); m\z 1929, sodium adduct of the tetrasodium salt of the pentamer (reported m\z1930). The presence of the sodium adduct of the #1999 Biochemical Society 532 A. M. Gil-Serrano and others Table 1 1H[NMR and 13C-NMR chemical shifts (δ) and coupling constants (J) for the CPS from S. fredii and the derived monosaccharide Compound Parameter Position…12 3 456789 Polysaccharide δH(p.p.m.) – – 1.52 (ax) 4.20 4.21 4.07 4.19 4.30 1.30 2.11 (eq) δC(p.p.m.) 173.7 99.7 37.2 66.4 49.8 73.1 55.1 69.2 18.3 3-Hydroxybutyryl residue δH(p.p.m.) – 2.57 (a) 4.20 1.21 2.43 (b) δC(p.p.m.) 178.0* 45.2 68.0 19.9 Acetyl residue δH(p.p.m.) – 2.01 δC(p.p.m.) 175.7* 23.4 Monosaccharide δH(p.p.m.) – – 1.81 (ax) 4.18 4.25 4.06 4.19 4.14 1.13 1.94 (eq) J(Hz) – – k12.7 (3ax,3eq) 5.0 (3eq,4) 1.7 (5,6) 10.5 (6,7) 3.3 (7,8) 6.5 (8,9) 12.7 (3ax,4) 4.2 (4,5) δC(p.p.m.) 175.0†97.7 36.0 66.6 50.1 71.3 54.2 68.4 16.8 3-Hydroxybutyryl residue δH(p.p.m.) – 2.41 (a) 4.23 1.26 2.39 (b) J(Hz) – k14.5 (2ha,2hb) 4.9 (2hb,3h) 8.1 (2ha,3h) 6.3 (3h,4h) δC(p.p.m.) 177.4†46.4 66.2 23.37 Acetyl residue δH(p.p.m.) – 2.03 δC(p.p.m.) 175.9†23.2 *,†These assignments could be interchanged. sodium salt has been reported for sialylated oligosaccharides [28]. These experiments confirmed the homopolymeric character of the polysaccharide and the molecular mass of the repeating unit. NMR analysis of the monosaccharide The "H and "$C chemical shift resonances from the monosaccharide, obtained as described above, were assigned from "H (Figure 5) and "$C-NMR spectra, COSY and multiplicity-edited HSQC spectra (Table 1). Data for C-5 and C-7 indicated that these positions were aminated. The difference between chemicals shifts of the signals for H-3ax and H-3eq was only 0.13 p.p.m., indicating the equatorial disposition of the vicinal carboxy group [13]. The JH,H pattern found for all the protons in the monosaccharide residue, and the δCvalue for C-9 (16.8 p.p.m.) [7,8,13] suggested that the sugar was the -glycero--manno isomer or its enantiomer. The large negative value for the optical rotation, [α]Dk45 m(c1.1 in water), allowed us to conclude that the monosaccharide residue was a diacyl derivative of 5,7diamino-3,5,7,9-tetradeoxy--glycero-α--manno-nonulosonic (pseudaminic) acid. This result is in accordance with previous results reported for pseudaminic acid derivatives [7,8,10,11]. Electron-impact high-resolution MS analysis of the monosaccharide The positions of the acyl substituents were determined by an electron-impact high-resolution mass spectrum of a derivative of the monosaccharide repeating unit. The sugar was carbonylreduced, carboxy-reduced and methylated. GLC–MS analysis of the resulting monosaccharide derivative showed the presence of (2Ror 2S)-3,5,7,9-tetradeoxy-2-deutero-1,2,4,6,8-penta-Omethyl-5-(N-methylacetamido)-7-[N-methyl-(3-methoxybutyramido)]--glycero--manno-nonitol (Figure 6). The fragmentation scheme was confirmed by accurate mass measurements in a high-resolution electron-impact mass spectrum. The results for the diagnostic fragments were as follows: M+,m\z479.331282, C#$H%& #HN#O);m\z464.307513, C##H%# #HN#O);m\z 448.313575, C##H%# #HN#O(;m\z434.295509, C#"H%! #HN#O(; m\z420.281651, C#!H$) #HN#O(;m\z331.223404, C"'H$"N#O&; m\z277.187385, C"$H#& #HNO&;m\z246.170080, C"#H#%NO%; m\z233.161089, C""H#" #HNO%; and m\z202.143719, C"!H#!NO$. The base peak at m\z102.091674, C&H"#NO, was assigned to a secondary fragment arising from the loss of CH$CH(OMe)CH lClO from the primary fragment at m\z 202.143719. Linkage analysis The position of the glycosidic linkage of the pseudaminic acid derivative in the homopolysaccharide was determined by methylation analysis of a disaccharide isolated from the mild hydrolysis of the polysaccharide. The oligosaccharide was carbonyl-reduced, methylated, carboxy-reduced, hydrolysed, reduced and acetylated. Only the derivative arising from the reduced reducing-end unit, (2Ror 2S)-1-O-acetyl-3,5,7,9tetradeoxy-2-deutero-2,4,6,8-tetra-O-methyl-5-(N-methylacetamido)-7-[N-methyl-(3-O-acetylbutyramido)]--glycero-- manno-nonitol, was detected and identified by GLC–MS (Figure 7). The fragmentation scheme was confirmed by accurate mass measurements. The results for the diagnostic fragments were as follows: M+,m\z535.320986, C#&H%& #HN#O"! ;m\z520.301017, C#%H%# #HN#O"! ;m\z504.303188, C#%H%# #HN#O*;m\z 476.271663, C##H$) #HN#O*;m\z359.216887, C"(H$"N#O';m\z 305.182540, C"%H#& #HN#O';m\z274.165515, C"$H#%NO&;m\z 261.157209, C"#H#" #HNO&and m\z230.139220, C""H#!NO%. The base peak m\z102 had the same origin as before. These results showed clearly that the glycosidic linkage was established between the hemiketalic OH and the hydroxyl group of the 3hydroxybutyramido residue. In consequence, the sugar repeating units were attached via two linkages: one ketalic and the other amidic. Additionally, the small difference between H-3ax and H3eq δvalues (0.59 p.p.m.) in the "H-NMR spectrum (Table 1) indicated that the orientation of the carboxy group of the pseudaminic diacyl derivative in the polysaccharide is equatorial #1999 Biochemical Society 533Structure of a polysaccharide isolated from Sinorhizobium fredii HH103 Figure 6 Electron-impact MS and fragmentation scheme for (2Ror 2S)-3,5,7,9-tetradeoxy-2-deutero-1,2,4,6,8-penta-O-methyl-5-(N-methylacetamido)-7- [N-methyl-(3-methoxybutyramido)]-L-glycero-L-manno-nonitol Figure 7 Electron-impact MS and fragmentation scheme for (2Ror 2S)-1-O-acetyl-3,5,7,9-tetradeoxy-2-deutero-2,4,6,8-tetra-O-methyl-5-(Nmethylacetamido)-7-[N-methyl-(3-O-acetylbutyramido)]-L-glycero-L-manno-nonitol (reported ∆δl0.6 p.p.m. for the equatorial carboxy group and ∆δl0.9 p.p.m. for the axial carboxy group) [8,13]. Moreover, the chemical shift for C-2 is reported to be in the range δl 99–100 p.p.m. when the carboxy group is equatorial and in the range δl101–104 p.p.m. when it is axial [10]. Our δvalue (99.7 p.p.m.) indicates that the carboxy group was equatorial. Because the configuration of the sugar residue is -glycero--manno, and the carboxy group was equatorial, the anomerism should be defined as α, in accordance with the recommendation for the nomenclature of carbohydrates. Determination of the absolute configuration of the 3-hydroxybutyramido group Hydrolysis of the polysaccharide followed by extraction of the hydrolysate with ethyl acetate gave free 3-hydroxybutyric acid, #1999 Biochemical Society 534 A. M. Gil-Serrano and others Figure 8 Proposed structure of the CPS from S. fredii HH103 which was converted into the (R,S)-oct-2-yl or the (S)-oct-2-yl ester, trimethylsilylated and analysed by GLC–MS. The chromatogram showed a 3:1 diasteroisomeric mixture obtained in the esterification with (S)-octan-2-ol. These results allowed us to propose that 75%of the 3-hydroxybutyric acid in the polysaccharide had the configuration Rand 25%the configuration S. On the basis of the above results, we propose that the CPS produced by S.fredii HH103 has the structure shown in Figure 8. Figure 1 indicates that fraction F3 does not contain LPS, or that its presence is negligible. To improve the detection of the Kantigens, we performed PAGE in the absence of SDS to prevent the migration of LPS molecules into the gel. Figure 2 shows the electrophoretic profile of the K-antigen polysaccharides of strain HH103 when the gel was fixed with Alcian Blue and silverstained. Under these conditions, at least 26 discrete bands and a trailing smear were present. The CPS in the sample was composed of small molecules with a polymerization degree ranging from 1 to more than 26 monosaccharide units. There was no predominant size because the quantities of molecules composed of 3–20 units were similar (Figure 2). In the absence of Alcian Blue fixation, the K-antigen polysaccharide was hardly visible, even with a large amount of sample loaded on the gel. These results are in accordance with those described for Sinorhizobium Kantigens [4]. However, the sensitivity of the silver-staining method after fixation with Alcian Blue was rather low because it was necessary to apply 5–10-fold the amount of sample normally used to reveal polysaccharides. The structure of the K-antigen of strain HH103 does not contain cis-diol groups that could reduce the silver ion during the silver-staining procedure. This explains why fraction F3 (which contained only the K-antigen polysaccharide) was so poorly silver-stained when standard protocols were used for the detection of K-antigen polysaccharides. It is important to determine the size of oligosaccharides because this can be an important factor in their biological activity [29]. The sizes of small molecules can be very accurately determined by different techniques such as FAB or MALDI–TOF. However, these techniques are not useful for the quantification of each component in a polydisperse sample, because their sensitivity decreases sharply as polymer size rises and because oligosaccharides that are larger than a few carbohydrate units are too big to be resolved adequately by these techniques. Nevertheless, the range of polymerization and the relative abundance of each molecule in the whole sample can easily be obtained from the electrophoretic profile (Figure 2). The absolute sizes of the molecules cannot be directly obtained by electrophoresis but in this case it was possible to ascribe the fastest band to the monomer because we knew from MALDI–TOF that in the CPS sample the smallest molecule was the monomer. Electrophoresis of whole-cell extract in Tris\borate buffer showed a CPS profile similar to that of the purified F3 sample (results not shown), so the small size of the purified CPS of S.fredii HH103 was not due to its degradation during purification. DISCUSSION The structures of K-antigen polysaccharides of different S.fredii and S.meliloti strains have recently been reported [3–5]. There are structural variations between the K-antigen polysaccharide of S.meliloti and those of S.fredii, and also from strain to strain within a single species. Despite these variations, the structure seems to be a conserved structural motif of a disaccharide repeating unit composed of a variable glycosyl residue bonded to a Kdx sugar [3]. This Kdx sugar is Kdo in S.fredii strains and derivatives of the nonulosonic acid in S.meliloti and Rhizobium sp. NGR234 [3,4]. The repeating unit of the K-antigen polysaccharide of S.fredii strain HH103 is pseudaminic acid, an analogue of Kdo. Therefore it is most probable that the Kdo detected by GLC–MS analyses was due to the presence of trace amounts of LPS in the purified samples. In contrast with the Sinorhizobium K-antigens reported previously [3], which are composed of a disaccharide repeating unit, the K-antigen polysaccharide of strain HH103 is a homopolysaccharide in which the glycosidic linkage is established between the hemiketalic OH and the hydroxy group of the 3hydroxybutyramido residue. In consequence, the sugar repeating units are attached via two linkages: one hemiketalic and the other amidic. The monomer of the K-antigen polysaccharide of strain HH103 is structurally related to the Kdx sugar of the disaccharide repeating unit of the CPSs of S.meliloti AK631 and Rhizobium sp. NGR234. These two bacteria seem to be related to S.fredii strains. The species S.fredii and S.meliloti are very closely related [30] and some S.fredii strains are capable of establishing an effective nodulation with alfalfa cultivars [31]. Although Rhizobium sp. NGR234 does not nodulate on soya beans, this bacterium carrying the symbiosis plasmid of S.fredii USDA191 can form nitrogen-fixing nodules with soya beans [32]. HH103 and NGR234 show a very broad host range of nodulation, the legumes nodulated by HH103 being a subset of those nodulated by NGR234. The DNA sequence of genes located in the symbiotic plasmid of NGR234 is strikingly similar to that in HH103 [33]. Strain NGR234 is also closely related to S.fredii on the basis of serology, partial SSU (small subunit) rRNA nucleotide sequences, and FAME (fatty acid methyl ester) analysis [34]. All the reported K-antigens of S.meliloti strains contain Nacetyl groups. In contrast, this characteristic is not present in the S.fredii K-antigens, with the sole exception of strain HH303 [3]. Here we show that the K-antigen of S.fredii strain HH103 also contains N-acetyl groups. Interestingly, both strain HH103 and HH303 also have a host specificity different from those of other S.fredii strains in that they are able to nodulate, and fix nitrogen, with American soya-bean cultivars. The conserved structural motif of the K-antigen polysaccharides of Sinorhizobium strains [(R")sugar-(R#)Kdx] has been proposed as a result of a comparative study of the K-antigens of S.meliloti (unable to nodulate soya beans), Rhizobium sp. NGR234 (unable to nodulate soya beans), and S.fredii strains USDA201, USDA205, USDA208 and USDA257. All these S.fredii strains form effective symbioses with Asiatic soya-bean cultivars but they fail to nodulate with most American soya-bean cultivars [35]. In contrast, S.fredii strains HH103 and HH303 are able to nodulate with American soya-bean cultivars and they are the only two strains that do not #1999 Biochemical Society 535Structure of a polysaccharide isolated from Sinorhizobium fredii HH103 produce a K-antigen polysaccharide repeating unit involving a variable glycosyl residue bonded to Kdo or a derivative of Kdo. Although the number of S.fredii strains investigated is still too small to obtain a consistent conclusion, there is other evidence supporting the idea that K-antigen polysaccharides could have a role in the specificity of soya-bean cultivars\bacterial strains. For instance, mutations in the nolWXBTUV genes of S.fredii USDA257 result in extending the host range of this strain to include different American soya-bean cultivars that are not normally infected by it. These host-range mutants show clear differences in their LPSs and K-antigens [3]. Although no clear-cut functions have been assigned to rhizobial K-antigen polysaccharides, it has been shown recently that they could be involved in the onset of the alfalfa defence response [36]. The structural diversity of Sinorhizobium K-antigen polysaccharides makes it difficult to establish a relationship between their structure and their possible biological activity. In addition, the expression of rhizobial K-antigens has been shown to be affected by legume-derived compounds [3]. This report shows that the repeating unit of the K-antigen polysaccharide can also be a monosaccharide instead of a disaccharide. In addition, strain SVQ293 (a mutant derivative of HH103 that shows a pleiotropic phenotype) produces a K-antigen polysaccharide in which the repeating unit is a trisaccharide composed of galactose, ribose and 5-O-methyl-3-deoxynonulosonic acid [5]. S.fredii strain USDA257 can produce two different K-antigen polysaccharides, both containing Kdo in the disaccharide repeating unit. Our results show that, although strain HH103 apparently produces only one K-antigen polysaccharide, it has the intrinsic capacity to produce another CPS (in mutant SVQ293) that is very different in structure [5]. This situation resembles that described for S.meliloti Rm 1021 (also called Rm2011), in which two completely different types of exopolysaccharide can be produced. Exopolysaccharide I, or succinoglycan, is produced by the wild-type strain, whereas exopolysaccharide II, or galactoglucan, is produced by some specific mutants [37]. It seems that the presence of an acidic CPS could be a general feature of a broad range of plant symbiont and pathogenic bacteria [3]. Here we have shown that the K-antigen polysaccharide of strain HH103 is not easily detected by using the standard Alcian Blue method unless a very high concentration of the polysaccharide sample is used in the electrophoresis. This fact is a warning of the possibility that the presence of certain types of rhizobial CPS might not be detected in rhizobial strains. We thank Professor E. Mele !ndez for the use of his densitometric facilities. We thank the Comisio !n Interministerial de Ciencia y Tecnologı!a (grants BIO096-1469-C03-01 and 02) for financial support. The work of J. C. was supported by a grant from the Consejerı!a de Educacio !n of the Canarian Government. REFERENCES 1De !narie !, J., Debelle, F. and Prome !, J. C. (1996) Annu. Rev. Biochem. 65, 503–535 2 Gray, J. W., de Maagd, R. A., Rolfe, B. G., Johnston, A. W. and Lugtenberg, B. J. J. (1992) in Molecular Signals in Plant–Microbe Communications (Verma, D. P. S., ed.), pp. 359–376, CRC Press, Boca Raton, FL 3 Kannenberg, E. L., Reuhs, B. L., Fosberg, L. S. and Carlson, R. W. (1998) in The Rhizobiaceae (Spaink, H. P., Kondorosi, A., and Hooykaas, P. J. J., eds.), pp. 119–154, Kluwer Academic Publishers, Dordrecht Received 19 February 1999/13 May 1999; accepted 30 June 1999 4 Forsberg, L. S. and Reuhs, B. L. (1997) J. Bacteriol. 179, 5366–5371 5 Gil-Serrano, A. M., Rodrı!guez-Carvajal, M. A., Tejero-Mateo, P., Espartero, J. L., Thomas-Oates, J., Ruiz-Sainz, J. E. and Buendı!a-Claverı!a, A. M. (1998) Biochem. J. 334, 585–594 6 Knirel, Y. A., Vinogradov, E. V., L’vov, V. L., Kocharova, N. A., Shashkov, A. S., Dimitriev, B. A. and Kochetkov, N. K. (1984) Carbohydr. Res. 133, C5–C8 7 Knirel, Y. A., Vinogradov, E. V., Shashkov, A. S., Kochetkov, N. K., L’vov, V. L. and Dimitriev, B. A. (1985) Carbohydr. Res. 141, C1–C3 8 Knirel, Y. A., Kocharova, N. A., Shashkov, A. S. and Kochetkov, N. K. (1986) Carbohydr. Res. 145, C1–C4 9 Knirel, Y. A., Vinogradov, E. V., Shashkov, A. S., Dimitriev, B. A., Kochetkov, N. K., Stanislavsky, E. S. and Mashilova, G. M. (1986) Eur. J. Biochem. 157, 129–138 10 Knirel, Y. A., Kocharova, N. A., Shashkov, A. S., Dimitriev, B. A., Kochetkov, N. K., Stanislavsky, E. S. and Mashilova, G. M. (1987) Eur. J. Biochem. 163, 639–652 11 Kenne, L., Linberg, B., Schweda, E., Gustafsson, B. and Holme, T. (1988) Carbohydr. Res. 180, 285–294 12 Knirel, Y. A., Moll, H., Helbig, J. H. and Za $hringer, U. (1997) Carbohydr. Res. 304, 77–79 13 Knirel, Y. A., Rietschel, E. T., Marre, R. and Za $hringer, U. (1994) Eur. J. Biochem. 221, 239–245 14 Knirel, Y. A., Helbig, J. H. and Za $hringer, U. (1996) Carbohydr. Res. 283, 129–139 15 Helbig, J. H., Lu $ck, C. P., Knirel, Y. A. and Za $hringer, U.(1995) Epidemiol. Infect. 115, 71–78 16 Vinogradov, E. V., Shashkov, A. S., Knirel, Y. A., Kochetkov, N. K., Dabrowski, J., Grosskurth, H., Stanislavsky, E. S. and Kholodkova, E. V. (1992) Carbohydr. Res. 231, 1–11 17 Beringer, J. E. (1974) J. Gen. Microbiol. 84, 188–198 18 Westphal, O. and Jann, K. (1965) Methods Carbohydr. Chem. 5, 83–91 19 Ko $plin, R., Wang, G., Ho $tte, B., Priefer, U. B. and Pu $hler, A. (1993) J. Bacteriol. 175, 7786–7792 20 Lesse, A. J., Campagnari, A. A., Bittner, W. E. and Apicella, M. A. (1990) J. Immunol. Methods 126, 109–117 21 Kittelberger, R. and Hilbink, F. (1993) J. Biochem. Biophys. Methods 26, 81–86 22 Corzo, J., Pe !rez-Galdona, R., Leo !n-Barrios, M. and Gutie !rrez-Navarro, A. M. (1991) Electrophoresis 12, 439–441 23 Tsai, C. M. and Frasch, C. E. (1982) Anal. Biochem. 119, 115–118 24 Dı!az-Marrero, A. R., Santamarı!a, M., Poveda, A., Jime !nez-Barbero, J. and Corzo, J. (1998) Electrophoresis 19, 2621–2624 25 Ciucanu, I. and Kerec, F. (1984) Carbohydr. Res.131, 209–217 26 Parella, T., Sa !nchez-Ferrando, F. and Virgili, A. (1997) J. Magn. Reson. 126, 274–277 27 Knirel, Y. A., Vinogradov, E. V., Shashkov, A. S., Dimitriev, B. A., Kochetkov, N. K., Stanislavsky, E. S. and Mashilova, G. M. (1987) Eur. J. Biochem. 163, 627–637 28 Mock, K. K., Davey, M. and Cottrell, J. S. (1991) Biochem. Biophys. Res. Commun. 177, 644–651 29 Battisti, L., Lara, J. C. and Leigh, J. A. (1992) Proc. Natl. Acad. Sci. U.S.A. 89, 5625–5629 30 De Lajudie, P., Willems, A., Pot, B., Dewettinck, D., Mestrojuan, G., Neyra, M., Collins, M. D., Dreyfus, B., Kesters, K. and Gillis, M. (1994) Int. J. Syst. Bacteriol. 44, 715–733 31 Hashem, F. M., Kuykendall, L. D., El-Fadly, G. and Devine, T. E. (1997) Symbiosis 22, 255–264 32 Appelbaum, E. R., McLoughlin, T. J., O’Connell, M. and Chartrain, N. (1985) J. Bacteriol. 163, 385–388 33 Lamrabet, Y., Bellogı!n, R. A., Cubo, T., Espuny, R., Gil, A., Krishnan, H. B., Megı!as, M., Ollero, F. J., Pueppke, S. G., Ruiz-Sainz, J. E. et al. (1999) Mol. Plant–Microbe Interact. 12, 207–217 34 Jarvis, B. D. W., Downer, H. L. and Young, J. P. W. (1992) Int. J. Syst. Bacteriol. 42, 93–96 35 Keyser, H. H., Bohool, B. B., Hu, T. S. and Weber, D. F. (1982) Science 215, 1631–1632 36 Becquart-de Kozak, Y., Reuhs, B. L., Buffard, D., Breda, C., Kim, J. S., Esnault, R. and Kondorosi, A. (1997) Mol. Plant–Microbe Interact. 10, 114–123 37 Becker, A. and Pu $hler, A. (1998) in The Rhizobiaceae (Spaink, H. P., Kondorosi, A., and Hooykaas, P. J. J., eds.), pp. 97–118, Kluwer Academic Publishers, Dordrecht #1999 Biochemical Society