High Heterogeneity of the Exopolysaccharides of Pseudomonas solanacearum Strain GMI 1000 and the Complete Structure of the Major Polysaccharide*

The exopolysaccharide of Pseudomonas solana- cearum, which is believed to play an important role in bacterial virulence, was considered by most authors as a homogeneous entity essentially composed of N-ace- tylgalactosamine. The present work demonstrates the high degree of heterogeneity of this exopolysaccharidic material, which consists of a high molecular weight acidic polysaccharide and a mainly noncarbohydrate structure as major subfractions. Rhamnose-rich polyo-side and glucan fractions are also present as minor components. We report the complete structure of the acidic heteropolymer involving, in addition to N-ace- tylgalactosamine, equimolar ratios of two rare amino sugars, 2-N-acetyl-2-deoxy-~-galacturonic acid and 2-N-acetyl-4-N-(3-hydroxybutanoyl)-2,4,6-trideoxy-~-

including specific recognition steps. The final wilting of the infected plant as a result of the plugging of the xylem vessels by the abundant mucus excreted by the bacteria is the most striking evidence of the direct mechanical involvement of this EPS in the pathogenesis (5). In addition, numerous studies relevant to the interactions of both phytopathogenic and phytosymbiotic bacteria with their host plants have revealed direct links between the EPS structure and the establishment of either the disease or the symbiosis (6-9), suggesting a more specific involvement of the EPS in the bacteria-plant relationship. The actual demonstration of such a precise correlation in P. solanacearum pathogenicity could never be achieved due to the lack of structural knowledge of the exopolysaccharidic material. In the most recent works, the EPS was assumed to be a homopolymer of N-acetylgalactosamine (10, ll), and the last studies concerning the relationship of EPS to virulence were based on this assumption, developing the examination of point-insertion Tn5 mutants impaired in both virulence and production of N-acetylgalactosamine (12, 13). They led to conflicting conclusions because of the lack of precise chemical definition of the EPS.
We report in this paper the highly heterogeneous character of the exopolysaccharidic material from P. solanacearum and the complete structure of the major polysaccharide component containing, in equimolar ratios, N-acetylgalactosamine and two unusual amino sugars, 2-N-acetyl-2-deoxy-~-galacturonic acid and 2-N-acetyl-4-N-(3-hydroxybutanoyl)-2,4,6-trideoxy-D-glucose. This work focuses on the chemical characterization of the EPS of P. solanacearum as an essential prerequisite in the establishment of its potential roles in the development of the pathogenesis pathway. EXPERIMENTAL PROCEDURES'

RESULTS
Purification and Sugar Composition of the Total EPS-Ethanol precipitation and subsequent dialysis of the precipitated material using an Amicon XM50 membrane yielded an average of 700 mg of crude extract/liter of culture. Lipid contaminants, less than 2% in mass, were removed by chloroform extraction. Final purification was then carried out by cation exchange chromatography, which eliminated more * Portions of this paper (including "Experimental Procedures," part of "Results," Fig. 8, and Table 4) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press. than 50% in mass, due almost entirely to the removal of amino acids and/or peptides, as revealed by amino acids analysis, without any significant loss of sugar components. This ion exchange step led to a total exopolysaccharidic extract (total EPS) that contained no detectable nucleic acids and less than 4% proteins or amino acids in mass. Different preparations of total EPS were obtained from separate culture batches. They all contained neutral and amino sugars as revealed by the phenol-sulfuric and ninhydrin assays, respectively. Colorimetric determinations for ketoses (14) and uronic acids (15,16) were both negative. IR spectra showed typical amide I and amide I1 bands at 1550 and 1650 cm", and a carbonyl band at 1730 cm" indicative of acid or ester groups. We found that methanolysis (MeOH/HCl, 1 M, 20 h, 80 "C) resulted in the best qualitative and quantitative recovery of all sugars as compared to the other tested solvolysis conditions. GLC analysis of the resultant methylglycosides as peracetate derivatives led to the identification of rhamnose, mannose, galactose, glucose, and N-acetylgalactosamine in the relative ratio of 1:l:traces:l:g (Fig. 1). Peaks A, B, and C were further identified as N-acetylgalactosaminuronic acid forms as described below.
Fractionation of the EPS and Analysis of Subfractwns-Initial fractionations of the total EPS were attempted by anion exchange and gel filtration chromatographies with water as eluent, but only the use of these two techniques under dissociative conditions allowed a distinct separation of different subfractions.
Chromatography over DEAE-Trisacryl with a 7 M urea, 0.05 M Tris/HCl (pH 6.5) buffer as eluent led to the isolation of two main subfractions, each accounting for 40% in mass of the total EPS, and of a minor fraction.
The nonretained fraction contained mannose as the major sugar component (molar ratios: Man, 4; Glc, 1; Rha, traces), but the total sugar content of this fraction did not exceed 10% in mass of the total EPS. Its infrared spectrum showed two amide bands of strong intensity, but the amino acids analysis of this material after hydrolysis revealed less than 1% amino acids. A weakly acidic, minor fraction was eluted from the DEAE with 0.01-0.03 M NaCl concentration; it accounted for less than 5% in mass of the total EPS, and its sugar composition was strongly dominated by rhamnose (molar ratios: Rha, 7; GlcNAc, l ; Man, traces).
The second main subfraction of the total EPS was released from the DEAE with a 0.2-0.4 M NaCl concentration. This acidic polymer was found to contain N-acetylgalactosamine, N-acetylgalactosaminuronic acid, and rhamnose in the molar ratio 6:3:1 according to the GLC analysis of the methylglycosides released by methanolysis of this fraction. Infrared, 13C NMR, and 'H NMR analyses of this polymer were consistent with an exclusively polyglycosidic structure. This acidic fraction will be later referred to as the "acidic-EPS." Gel permeation over Sephacryl S-500 of the total EPS with a 2 M guanidinium chloride, 0.5 M sodium acetate (pH 6.8) buffer as eluent led to the separation of the acidic heteropolymer, as a broad peak in the high molecular weights range, and of the mannose-rich fraction. The rhamnose-rich minor fraction was not individually separated, but at the lowest molecular weights range, an additional subfraction was isolated, which accounted for less than 10% by weight and was found to be exclusively constituted of glucose.
Structural Characterization of 2-N-Acetyl-2-deoxy-~-galacturonic acid-The three methylglycosides corresponding to GLC peaks A, B, and C ( Fig. 1) were individually isolated by Cls reverse-phase HPLC after methanolysis of the total EPS. The EI/MS analyses of either the alditol acetate or carboxylreduced derivatives of A, B, and C allowed their identification as a-and @-methylfuranosides and @-methylpyranoside of a 2-N-acetyl-2-deoxyhexuronic acid (see Miniprint).
We have carried out the 'H NMR analysis of the methylglycosides A, B, and C either in their native or carboxylreduced forms. The chemical shifts and coupling constants that were observed for the protons of the carboxyl-reduced forms of A, B, and C were identical with those of methyl-GalNAc-a-furanoside, -a-pyranoside, and -@-furanoside, respectively (Table 1). Due to carboxylate deshielding effects, shifts to low fields by 0.5-1.0 ppm were recorded for the H-4 and H-5 protons of the native methylglycosides.
The L absolute configuration of this sugar was determined by optical rotation measurements of the carboxyl-reduced forms of A, B, and C methylglycosides as compared with the corresponding forms of standard D-GalNAc. The recorded values in methanol for the reduced compounds A and C respectively). The L configuration was further confirmed by GLC analysis of the corresponding 2-(R)-butylglycoside derivatives (17).

Identification of 2-N-Acetyl-4-N-(3-hydroxybutanoyl)-2,4,6trideoxy-D-glucose As an Additional Sugar Component of the
Acidic EPS-An analysis by gel permeation over Bio-Gel P-4 of a mixture of methylglycosides obtained from the purified acidic EPS fraction revealed the occurrence of an additional sugar component X whose peracetate derivative was never detected in the GLC analysis. A direct chemical ionization NH3/mass spectrometry study of its perdeuterioacetate deriv-

H NMR data of A, B, and C methylglycosides either under native (n) or carboxyl-reduced forms (r) and of the corresponding cyclic forms of standard D-GalNAc
GalNAc-a-f  The I3C NMR signals at 6C = 57.6,57.0,49.9, and 49.2 were assigned to four nitrogen-carrying carbons. Three methyl proton resonances of acetamide groups were observed at 6H = 2.02, 1.96, and 1.91; the corresponding carbons gave collapsed resonances at 6C = 23.7 in the I3C NMR spectrum. In the 'H NMR spectrum, two methyl resonances at 6H = 1.23 and 1.19 could be ascribed to the H-6 protons of the bacillosamine residue (the corresponding carbon was identified at 6C = 18.4 in the 13C NMR spectrum) and the methyl group of its 3-hydroxybutyramide substituent, respectively (21).
A I3C NMR experiment Distortionless Enhancement by Polarization Transfer mode allowed the identification of two methylene groups. The first, at 6C = 46.3 was assigned to the C-2 of the 3-hydroxybutyramide residue (24) (the related protons appeared as a pseudo-doublet in the 'H NMR spectrum at 6H = 2.30. The other at 6C = 62.5 was consistent with the C-6 of the N-acetylgalactosamine moiety (25).
De-0-acetylation induced no change in the 'H NMR and I3C NMR characteristic resonances of the N-acetyl and N-(3-hydroxybutanoyl) groups. Their relative integration, as determined from the 'H NMR signals (3H for each acetamide group; 2H and 3H for methylene and methyl groups of the 3hydroxybutyramide, respectively) demonstrated all amino sugars to be stoichiometrically N-acylated.
In contrast, the de-0-acetylation step induced the disappearance of resonances at 6H 2.17 (3H, singlet) and 6H = 5.63 (1H, multiplet) in the 'H NMR spectrum. These signals could be consequently assigned to the methyl group of an 0-acetyl residue and to the proton carried by the 0-acetylated carbon  (26). According to the relative intensity of these resonances, one of the sugars could be assumed to be stoichiometrically mono-0-acetylated. A parallel disappearance of the signal at 621.5 in the I 3 C NMR spectrum was also observed accompanying the de-0-acetylation process (Fig. 4).
Structure of Acidic EPS-deriued Oligosaccharides-After a short methanolysis of the native polysaccharide (MeOH/HCl, 1 M, 15 min, 80 "C) and gel permeation over Bio-Gel P-4 of the reaction products, two major methyloligoglycosides could be isolated.
The EI/MS and 'H NMR analysis of these oligosaccharides allowed their identification as GalANAc-Bac2NAc4N (30HBut) (I) and GalNAc-GalANAc-Bac2NAc4N(3OHBut) (11) oligoglycosides. The field desorption/mass spectrometry of the perdeuterioacetate derivative of trisaccharide I1 showed a molecular ion at m/z = 961. The EI/MS spectrum (Fig. 5) was poorly From this primary ion pattern, the sequence GalNAc-GalANA~-Bac2NAc4N(30HBut) with a nondeuteriated 0acetyl group carried by the GalANAc residue, and consequently preexistent on the native oligosaccharide 11, was deduced. From the molecular ion and from the fragment ions mentioned above, secondary ions were observed, resulting from individual or sequential eliminations of trideuterioacetic acid, acetamide, methylformiate, or ketene. Similarly, the disaccharide I was characterized as GalANAc-Bac2NAc4N (SOHBut), but no constitutive 0-acetyl group was detected in this case.
Complete assignments of the signals in the 'H NMR spectrum of the trisaccharide I1 were made possible through a COSY experiment (Fig. 6). All the resonances of the Bac2NAc4N(30HBut)residue were split due to the coexistence of the a and anomeric forms of this "reducing end" sugar residue ( Table 2). The anomeric doublets were observed at 6H carbon atom had its resonance at 5.62 ppm (multiplet, IH), showing on the COSY spectrum two cross-peaks with resonances of protons at 6H = 4.05 (1H) and 4.60 (1H). S' lnce no additional connectivities were observed with the latter signal, and according to the nature and the sequence of the sugar residues in the trisaccharide, it could only be ascribed to the H-5 of the GalANAc residue. Consequently, the 0-acetyl residue was located at the C-4 position of the GalANAc residue thus glycosylated at the C-3 position by the GalNAc residue; the sequence of connectivities of this GalANAc led to the identification of its anomeric doublet at 6H = 5.23, which exhibited a small coupling constant (J1,z = 4.0 Hz) corresponding to an a-glycoside. All the signals of this sugar were weakly split due to the two anomeric configurations of the terminal BacZNac4N(30HBut); this effect was not extended to the GalNAc residue whose resonances were found to be homogeneous and whose anomeric doublet at 5.0 ppm defined a low coupling constant (X,* = 3.72 Hz) characteristic of the CY anomeric configuration of the sugar in the trisaccharide. The 'H NMR data of the disaccharide I showed a strong homology with those of the trisaccharide I1 with respect to the bacillosamine and the GalANAc residues ( Table 2) s~n t i a t e d by the low field resonance of the anomeric carbon atom in I3C NMR (28,29) and the high field resonance of the anomeric proton in the 'H NMR spectrum (30).
The nitrogen-carrying carbon resonances at 6C = 49.2 and 49.9 in the 13C NMR spectrum could only be ascribed to the C-2 of 2-acetamido-2-deoxy sugars with a-galacto configuration, since the C-2 of hexosamines with a-gluco, @-gluco, and f i -g a~c~o configuration have their resonances at a lower field (SC > 51.5) (28,29). This conclusion was in accordance with the CY configuration previously established for GalNAc and Resonances of the protons of this residue are identical for the (Y and 4 anomers of bacillosamine. GalANAc residues in the trisaccharide. The fl configuration was consequently ascribed to the bacillosamine residue. The resonances that could be ascribed to the carbons of an (3-D-Bac2NAc4N(30HBut)p glycosylated at C-3 by an a-L-GalpANAc residue (20) were distinctly present in the 13C NMR spectrum of the acidic EPS (Table 3). This I3C NMR spectrum exhibited a C-6 resonance of the GalNAc residue ( E = 62.5) that is characteristic of a nonsubstituted carbon at that position (25). Methylation of the polysaccharide by the Hakomori procedure and subsequent analysis of the partially methylated alditol acetates led to the identification of 4,6-di-O-methyl derivative of the galactosaminitol but failed in providing the expected derivatives from the GalANAc and the Bac2NAc4N(30HBut) residues. This result proved the glycosylation of the GalNAc residue at position 3 and suggested probable degradations of the other glycosyl residues during the methylation and/or hydrolysis steps. Since Gal-ANAc and BacZNAc4N(30HBut) are both substituted at positions 2 and 4, it follows that the only possible linkage is at position 3. In the 13C NMR spectrum of the de-0-acetylated acidic EPS (Fig. 4), the C-2 signal of both GalNAc and GalANAc are almost superimposed at 50 ppm, which is consistent with the almost identical resonances of their C-2 in their monosaccharide forms (25,31); this suggested the same glycosylation effect for both sugars in the polysaccharide. The proposed structure for the acidic EPS fraction (Fig. 7) was substantiated by a complete interpretation of the 13C NMR spectrum of the native and the de-0-acetylatedpolysaccharide based on literature data (Table 3).
In the acidic EPS preparation, rhamnose had been detected at an 8 times lower rate than GalNAc. No significant signal ascribable to rhamnosyl residue was detected in either the 'H or 13C NMR spectra of the polysaccharide. When the total exopolysaccharide material of P. solanacearum was fractionated over DEAE-Trisacryl, a weakly acidic minor fraction (5% by weight) was eluted at a NaCl concentration of 0.01-0.03 M; its sugar composition is dominated by the rhamnose (Rha, 7; GlcNAc, 1); therefore, it appears likely that the low rhamnose level detected in the acidic EPS compositional analysis originated from a slight residual contamination (3-5%) of this polysaccharide by the rhamnose-rich fraction.

DISCUSSION
This study reveals the high degree of complexity of the exopolysaccharidic material produced by P. solanacearum, complexity due both to its heterogeneity and to the nature of the components involved in the structure of its subfractions.
Since 1958, eight independent groups have published results dealing with the sugar composition of this EPS, but the overall findings were somewhat confusing (34). The main discordance was relative to the predominance of N-acetylgalactosamine (10,11,35,36), glucose (37), or glucose and mannose (38). The development of gel permeation and anion exchange chromatography with dissociative eluents was decisive in the  establishment of the heterogeneity of the EPS and in the extensive separation into subfractions. The high level of interactions between these subfractions could explain the fact that the EPS of P. soZunaceurum had so long been considered as a single entity.
The occurrence of three distinct subfractions involving, respectively, GalNAc, Man, or Glc as the major sugar component in their structures could explain the apparent discrepancies of the above mentioned previous reports relative to the EPS sugar composition. Indeed, according to the strain, the culture conditions, and the purification sequence, either one fraction or the other may appear as predominant in the final EPS preparation.
Our work has primarily been focused on the structural characterization of the major polyosidic fraction of the EPS and has led to the first identification of 2-N-acetyl-2-deoxy-L-galacturonic acid and 2-N-acetyl-4-~-~3-hydroxybutanoyi)-2,4,6-trideoxy-~-glucose as sugar components of the EPS of P. solanacearum.
The incoherence between the sugar composition of the acidic EPS, as determined after methanolysis and subsequent GLC analysis, and the proposed structure arose both from the resistance of the GalANAc-BacZNAc4N(3OHBut) linkage toward acid solvolysis and from the nondetection of the peracetate or pertrimethylsilyl derivatives of the bacillosamine residue under GLC analysis.
The occurrence in the acidic EPS structure of a glycosidic linkage acid-resistant between two acid-labile sugars could explain the fact that neither GalANAc nor BacZNAc4N (30HBut) had been detected in the previous analyses of the EPS from P. s o l u n~e a r~m , which revealed only the N-acetylgalactosamine residue (10).
In the methylation assays, we could not obtain the partially methylated alditol acetate derivatives from the GalANAc and Bac2NAc4N(30HBut) residues of the polysaccharide. GLC/ chemical ionization-MS analysis of the final reaction products showed numerous ions with molecular masses lower than 250 atomic mass units, suggesting degradation events either at the methylation step or during the acidic hydrolysis of the methylated polysaccharide. A preliminary carboxyl reduction of the polysaccharide by treatment with carbodiimide/Na-B2H4 was attempted, but three successive cycles gave a reduction rate lower than 20%. The glycosylation sites of both sugars, as well as the location of the 0-acetyl residue, were unambiguously assigned from the NMR data of the native polysaccharide and the derived oligosaccharides.
The complete assignment of the 13C NMR spectra of the native and de-0-acetylated acidic EPS fractions supported the proposed structure. The structure of the main polysaccharidic component of the EPS of P. s o~~e a r u m (Fig. 7) exhibits a striking analogy with the 0-antigen chain of the LPS from Pseudomonas ueruginosa strain Habs 0 3 (20). In addition, the occurrence of both L-GalANAc and Bac2NAc4N (30HBut) moieties has, up to now, exclusively been reported in 0-antigen chains of LPS of P. aeruginosa strains. Such a structural analogy might just be the consequence of the phylogenetic filiation of these pseudomonad strains. However, both EPS and LPS are exocellular materials and consequently at the interface between the bacteria and their respective eucaryotic hosts. The location, added to the structural complexity, of these polyosidic structures give them a particular interest in the study of the interactions between the pathogen and its host at the early stages, such as recognition steps, The investigation of such specific roles of the EPS in the pathogenesis induced by P. solanacearum requires a precise definition of the EPS structures. The chemical characterization of the acidic EPS fraction now provides a precise basis for the examination of the potential correlation of EPS with pathogenicity. Work in progress3 by using Tn3-induced avirulent mutants within a 6.5 kilobase genomic cluster reveals that mutants devoid of any acidic EPS are strictly avirulent, whereas mutants producing low amounts of acidic EPS exhibit reduced aggressiveness. Thus, the acidic EPS appears to be one of the determinants of the pathogenicity of P. solanacearum.
This work reveals the complexity of the exopolysaccharidic material from F. so~~nucearum in which the acidic EPS only accounts for 40% by weight, The second major fraction is a mainly nonglycosidic fraction (10% by weight of sugar components), and the hypothesis of a peptidic component has been discarded due to the low amino acid content of this material (<I%). The first 'H and I 3 C NMR data relative to the minor glucan fraction exclude the hypothesis of a 8-2glucan and suggest a complex mixture of low molecular weight L. Sequeira, unpublished work. glucans differing in their linkage modes.
These additional components of the EPS of P. solanacearum strain GMI 1000, as well as the rhamnose-rich fraction are currently under study.
The denomination "EPS" provides a general term for all forms of bacterial polysaccharides found outside the cell wall and secreted into the environment. The strain designation "EPS"' in P. solanacearum is extremely ambiguous since it is mostly based upon a smooth, fluidal colony morphology on agar medium. In contrast, the afluidal rough colony type is designated "EPS"'. This colony morphology designation cannot take into account qualitative and/or quantitative differences in the EPS, nor can it help in distinguishing between LPSor EPS-deficient mutants. The present work is the first step toward a precise biochemical definition of the EPS. It demonstrates that the EPS of P. solanacearum can no longer be considered as a whole entity and that the future studies on the correlation EPS with virulence will have to take into account the structural diversity of these different subfractions that could be related to differences in their biological functions.