Flagellin Glycans from Two Pathovars of Pseudomonas syringae Contain Rhamnose in d and l Configurations in Different Ratios and Modified 4-Amino-4,6-Dideoxyglucose

Abstract

Flagellins from Pseudomonas syringae pv. glycinea race 4 and Pseudomonas syringae pv. tabaci 6605 have been found to be glycosylated. Glycosylation of flagellin is essential for bacterial virulence and is also involved in the determination of host specificity. Flagellin glycans from both pathovars were characterized, and common sites of glycosylation were identified on six serine residues (positions 143, 164, 176, 183, 193, and 201). The structure of the glycan at serine 201 (S201) of flagellin from each pathovar was determined by sugar composition analysis, mass spectrometry, and ¹H and ¹³C nuclear magnetic resonance spectroscopy. These analyses showed that the S201 glycans from both pathovars were composed of a common unique trisaccharide consisting of two rhamnosyl (Rha) residues and one modified 4-amino-4,6-dideoxyglucosyl (Qui4N) residue, β-d-Quip4N(3-hydroxy-1-oxobutyl)2Me-(1→3)-α-l-Rhap-(1→2)-α-l-Rhap. Furthermore, mass analysis suggests that the glycans on each of the six serine residues are composed of similar trisaccharide units. Determination of the enantiomeric ratio of Rha from the flagellin proteins showed that flagellin from P. syringae pv. tabaci 6605 consisted solely of l-Rha, whereas P. syringae pv. glycinea race 4 flagellin contained both l-Rha and d-Rha at a molar ratio of about 4:1. Taking these findings together with those from our previous study, we conclude that these flagellin glycan structures may be important for the virulence and host specificity of P. syringae.

Glycosylation of pathogenic bacterial cell surface proteins, such as flagellin and pilin, has recently been recognized as an important factor in host-pathogen interactions (3, 19). Flagellin glycosylation is found in animal pathogens, and the genes required for glycosylation and glycan structure have been characterized for several bacteria, such as Campylobacter jejuni (10, 32), Pseudomonas aeruginosa (1, 25, 34), and Helicobacter pylori (24). Similarly, flagellins of some plant pathogens, including Pseudomonas syringae pv. tabaci 6605, Pseudomonas syringae pv. glycinea race 4, and Pseudomonas syringae pv. tomato DC3000 (28), as well as Acidovorax avenae (4), have been found to be glycosylated. Although the possible biological significance of flagellin glycosylation is frequently discussed, experimental evidence has been restricted to our study on P. syringae (13, 30, 31) and studies by other groups on P. aeruginosa (2, 33) and on C. jejuni and Campylobacter coli (11).

The phytopathogenic bacterium P. syringae is classified as a pathovar by its virulence toward different host plant species. In our previous study, flagellin from P. syringae was found to be an elicitor that causes a hypersensitive reaction (HR) of nonhost plants (28). Moreover, the HR-inducing activity is thought to be dependent on glycosylation (29). The significance of glycosylation is particularly notable for the two pathovars of P. syringae, P. syringae pv. glycinea race 4 and P. syringae pv. tabaci 6605, because although the respective flagellins display absolute amino acid sequence conservation, the HR-inducing activities are different. Recently, we found that a flagellin glycosylation island, which possesses putative glycosyltransferase genes, is required for flagellin glycosylation in P. syringae pv. glycinea race 4 and P. syringae pv. tabaci 6605 and that deletion of these genes reduced both virulence for their respective host plants and HR-inducing activity for nonhost plants (13, 30, 31). These results demonstrate that flagellin glycosylation plays an important role in determining the host specificity of each pathovar of P. syringae. We have identified six glycosylated serine residues in flagellin from P. syringae pv. tabaci 6605 (30). These serine residues are all localized on the predicted surface-exposed domain when the flagellin folds as a monomer in the assembled filament. Based on studies of mutants in which Ser was replaced with Ala and of glycosylation island deletion mutants, we demonstrated that flagellin glycosylation is essential for bacterial adhesion, swarming motility, and virulence on host tobacco leaves. Thus, flagellin glycosylation plays a key role not only as the determinant of HR induction activity but also in virulence-related bacterial characteristics.

Although biological and mutational studies of P. syringae emphasized the importance of flagellin glycosylation for bacterial virulence and host specificity, there was no direct structural information on the flagellin glycans. Here we report the structural characterization of the flagellin glycans in P. syringae.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

The bacteria used in this study are listed in Table 1. P. syringae pv. glycinea race 4 and P. syringae pv. tabaci 6605 and their derivative mutants were maintained in King's B medium at 27°C. Escherichia coli strains were grown at 37°C in Luria-Bertani (LB) medium.

TABLE 1.

Bacterial strains used in this study

Bacterial strain or plasmid	Relevant characteristic(s)a	Reference or source
Strains
E. coli
DH5α	F− λ− φ80dlacZΔM15 Δ(lacZYA-argF)U169 recA1 endA1 hsdR17(rK− mK+) supE44 thi-1 gyrA relA1	Takara, Kyoto, Japan
S17-1	thi pro hsdR hsdM+recA (chr::RP4-2-Tc::Mu-Km::Tn7)	23
P. syringae pv. glycinea
race 4	Wild type	A. Collmer
race 4-d1	race 4 mutant; Δorf1	31
race 4-d2	race 4 mutant; Δorf2	31
race 4-d3	race 4 mutant; Δorf3	31
race 4-dfliC	race 4 mutant; ΔfliC	This study
race 4-S143A	race 4 mutant; S143A	This study
race 4-S164A	race 4 mutant; S164A	This study
race 4-S176A	race 4 mutant; S176A	This study
race 4-S183A	race 4 mutant; S183A	This study
race 4-S193A	race 4 mutant; S193A	This study
race 4-S201A	race 4 mutant; S201A	This study
race 4-6 S/A	race 4 mutant; S143A, S164A, S176A, S183A, S193A, S201A	This study
P. syringae pv. tabaci isolate 6605	Wild type	30
Plasmids
pMC	1.8-kb chimeric PCR product deleting fliC cloned into pK18mobsacB at the EcoRI site; Kmr	This study; 26
pK18mobsacB	Small mobilizable vector; Kmr, sucrose sensitive (sacB)	23

^aKm^r, kanamycin resistance.

Site-directed mutagenesis of glycosylated residues of flagellin in P. syringae pv. glycinea race 4.

Ser-to-Ala mutants of P. syringae pv. glycinea race 4 were obtained by first generating a deletion mutant of the flagellin coding region (fliC). The resultant ΔfliC mutant was then complemented with the fliC region possessing the desired point mutation(s). The ΔfliC mutant of P. syringae pv. glycinea race 4 was made using a previously reported method (26) with a slight modification. One of the primers for the downstream region of fliC, designated PC4, was modified to 5′-GATCGCGTAAGTACCGTTGA-3′. The methods for site-directed mutagenesis and complementation by homologous recombination have been described previously (30). The Ser-to-Ala mutants were designated as follows: race 4-S143A, race 4-S164A, race 4-S176A, race 4-S183A, race 4-S193A, and race 4-S201A. A mutant with six serine substitutions (S143A, S164A, S176A, S183A, S193A, and S201A), designated race 4-6 S/A, was also constructed by the same method.

Purification of flagellin and preparation of glycosylated peptides.

P. syringae was incubated in LB medium containing 10 mM MgCl₂ for 48 h at 25°C. The cells were harvested by centrifugation, resuspended in 1/3 volume of minimal medium [50 mM potassium phosphate buffer, 7.6 mM (NH₄)₂SO₄, 1.7 mM MgCl₂, and 1.7 mM NaCl (pH 5.7)] supplemented with 10 mM (each) mannitol and fructose, and then incubated for 24 h at 23°C. Flagellin was purified by the method of Taguchi et al. (28). For the identification of glycan components of flagellin proteins, purified proteins were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and a band at 32 kDa was excised, crushed, and mixed with distilled water. The extracted flagellin was then concentrated using spin columns (Vivaspin VS0403; molecular weight cutoff, 10,000; Vivascience, Hannover, Germany). For purification of glycosylated peptides, purified flagellin from each pathovar was digested with aspartic N-peptidase (Boehringer Mannheim, Mannheim, Germany) at 35°C for 20 h in Tris-HCl buffer (pH 8.0). The resultant peptides with 0.1% (vol/vol) trifluoroacetic acid (TFA) were subjected to reverse-phase high-performance liquid chromatography (HPLC) using a 2.0- by 250-mm TSKgel octyldecyl silane-80TS column (Tosoh, Tokyo, Japan) as reported by Taguchi et al. (30). For large-scale preparation of the peptides, approximately 4 mg of digested flagellin was applied to a TSKgel octyldecyl silane-120TS column (4.6 by 150 mm; Tosoh) and eluted at a flow rate of 1.0 ml/min with a linear gradient of 9 to 90% aqueous acetonitrile (0.1% TFA) for 87 min. UV detection was carried out at 210 nm, and fractions were collected every minute. The target peptide (comprising amino acids D200 to A211; designated D200-A211) was identified by N-terminal amino acid sequencing using a protein sequencer (Procise 494 HT protein sequencing system; Applied Biosystems, Tokyo, Japan).

Mass spectrometry (MS).

For comparison of mass spectra for the intact flagellins and N136-K255 peptides from P. syringae pv. tabaci 6605 and P. syringae pv. glycinea race 4, flagellins from these wild-type strains and their mutants were digested with lysyl endopeptidase (Wako, Osaka, Japan) at 37°C overnight in 10 mM Tris-HCl buffer (pH 9.0). Each intact or digested protein was dissolved in water with 0.1% TFA, mixed with an equal volume of matrix solution (a saturated solution of sinapinic acid in 33% acetonitrile-water with 0.1% [vol/vol] TFA), and deposited on a target plate. Samples were analyzed using a Biflex III spectrometer (Bruker Daltonik GmbH, Bremen, Germany), and matrix-assisted laser desorption ionization-time-of-flight (MALDI-TOF) mass spectra were recorded in a linear, positive-ion mode with a mass accuracy of 0.1%.

MALDI-QIT-TOF MS analysis of the D200-A211 glycopeptide.

The HPLC eluate of the D200-A211 glycopeptide from each pathovar (0.5 μl) was mixed with an equal volume of matrix solution (10 mg/ml of 2,5-dihydroxybenzoic acid in 0.06% [vol/vol] TFA and 40% [vol/vol] acetonitrile) and deposited on a sample target plate. The mass spectra and tandem MS (MS-MS) spectra of the D200-A211 glycopeptide were recorded on an AXIMA quadrupole ion trap (QIT) MALDI-TOF MS (Shimadzu, Kyoto, Japan). Both MALDI-TOF MSs were calibrated using a standard mixture of peptides (Bruker Daltonics, Billerica, MA).

ESI-Q-TOF MS analysis of the D200-A211 glycopeptide.

Electrospray ionization (ESI)-TOF MS experiments were conducted using a quadrupole TOF (Q-TOF) MS (QSTAR XL; Applied Biosystems) equipped with a nanospray ESI source. The ion spray voltage was set to 1,000 V. For accurate mass measurements, the instrument was calibrated using y-series fragment ions (22) of Glu-fibrinopeptide B, and mass accuracy was within 5 ppm. Prior to analyses, samples were prepared by being dissolved in 30% acetonitrile and 0.1% formic acid. All mass spectra were obtained in positive-ion mode.

Sugar composition analysis of flagellin glycans.

The monosaccharide composition of glycans from purified flagellin proteins was analyzed using an ABEE (p-aminobenzoic acid ethyl ester) labeling kit (J-Oil Mills, Tokyo, Japan). Sialic acid, since it is a nonreducing sugar, is not converted by ABEE. To assess the presence or absence of sialic acid in the flagellin preparation, purified flagellin was treated with N-acetylneuraminic acid aldolase in order to release sialic acid residues prior to acid hydrolysis, thereby enabling the detection of sialic acid as N-acetylmannosamine. The subsequent processes of acid hydrolysis, N acetylation, and conversion with ABEE were carried out according to the method of Yasuno et al. (36). The resultant ABEE-converted monosaccharide(s) in the aqueous layer was analyzed by reverse-phase HPLC using a Honenpak C₁₈ column (inner diameter, 75 mm by 4.6 mm; J-Oil Mills) according to the manufacturer's instructions. For quantification of monosaccharides, a set of monosaccharides including glucose, galactose, mannose, arabinose, ribose, fucose, xylose, rhamnose (Rha), N-acetylglucosamine, N-acetylgalactosamine, and N-acetylmannosamine was used as standards.

Determination of d-Rha/l-Rha ratios in flagellins.

Enantiomeric ratios of the Rha residues in glycopeptides and intact flagellins were determined using gas chromatography (GC) according to the method of Gerwig et al. (8) with a slight modification. The D200-A211 glycopeptide and intact flagellin protein from each pathovar were subjected to acidic solvolysis with 1 N HCl in (S)-2-butanol for 16 h at 80°C. The (S)-2-butyl glycosides formed were then converted into their trimethylsilyl (TMS) derivatives and analyzed by GC (GC-17A; Shimadzu) using a DB-1 column (30 m by 0.25 mm; J&W Scientific, Folsom, CA) (37). Because d-glycosides of (S)-2-butanol and l-glycosides of (R)-2-butanol have the same retention time by GC analysis (nonchiral stationary-phase separation), l-rhamnosides of (R)- and (S)-2-butanol were prepared as standards for determination of (S)-2-butyl-d-rhamnoside and (S)-2-butyl-l-rhamnoside, respectively. To confirm two peaks assigned as d- and l-rhamnosides of (S)-2-butanol from P. syringae pv. glycinea race 4 flagellin, GC-MS was performed according to the method of McNeil and Albersheim (20) with a slight modification. A JMS DX-303 MS (JEOL, Tokyo, Japan) was interfaced with a Hewlett-Packard (Palo Alto, CA) 5890 GC using an SPB-1 column (30 m by 0.32 mm; Supelco Inc., Bellefonte, PA). GC-MS was performed by Toray Research Center Inc., Kamakura, Japan.

NMR spectroscopy.

Lyophilized glycopeptide D200-A211 prepared from P. syringae pv. tabaci 6605 flagellin was dissolved in 300 μl of D₂O to give a final concentration of 150 nmol/liter (pH 4.2). Glycopeptide D200-A211 from P. syringae pv. glycinea race 4 was dissolved in 300 μl of D₂O to give a final concentration of 50 nmol/liter (pH 2.9). ¹H nuclear magnetic resonance (NMR) spectra of the glycopeptides, including ¹H-¹H correlation spectra (double quantum filtered correlation spectroscopy [DQF-COSY], total correlation spectroscopy [TOCSY], and nuclear Overhauser and exchange spectroscopy [NOESY]) and ¹H-¹³C correlation spectra (heteronuclear single-quantum coherence [HSQC] and heteronuclear multiple bond connectivity [HMBC]), were obtained at 800.33 MHz on a Bruker Avance 800 spectrometer with a three-channel inverse (¹H/¹³C[¹⁵N]) CryoProbe (Bruker Biospin, Karlsruhe, Germany) at a temperature of 298 K. ¹³C NMR spectra were obtained at 125.76 MHz on a Bruker Avance 500 spectrometer with a dual ¹³C[¹H] CryoProbe (Bruker Biospin) at 298 K. The methyl signals of 2-methyl-2-propanol, δ_H at 1.23 ppm and δ_C at 31.3 ppm, were used as references for ¹H and ¹³C chemical shifts.

RESULTS

Identification of the glycosylated amino acid residues in P. syringae pv. glycinea race 4 flagellin.

Monomeric flagellin of P. syringae pv. glycinea race 4 and the corresponding lysyl endopeptidase-digested peptides were subjected to MALDI-TOF MS in order to characterize the modification pattern (Table 2). Intact P. syringae pv. glycinea race 4 flagellin purified from the wild-type strain exhibited [M + H]⁺ around m/z 32,380 as a broad peak reflecting quantitative heterogeneity in glycosylation (Table 2; Fig. 1C). The m/z values of the three main species (32,380, 32,515, and 32,668) within the broad peak were larger than the predicted values for [M + H]⁺ (i.e., m/z 29,148) based on amino acid sequences by 3,232 Da, 3,367 Da, and 3,520 Da, respectively. The molecular mass of flagellin from the Δorf1 mutant, which lacks the ability to glycosylate (31), was 29,154 Da. This value corresponds, within the margins of error of the system, to the molecular mass predicted from the deduced amino acid sequence, confirming that the flagellin from the Δorf1 mutant is not glycosylated.

TABLE 2.

Mass values of intact flagellin and N136-K255 peptide fragments from P. syringae pv. glycinea race 4

Flagellin or peptide fragmenta	[M + H]+		Δ[M + H]+ (observed − calculated)
	Observedb	Calculatedc
Intact flagellin (A2-Q282)
Wild type	32,380, 32,515, 32,668	29,148	3,232, 3,367, 3,520
Δorf1	29,154	29,148	6
Peptide fragment (N136-K255)
WT	15,296, 15,444, 15,591	12,074	3,222, 3,370, 3,517
S143A	14,742, 14,890, 15,036	12,058	2,684, 2,832, 2,978
S164A	14,746, 14,894, 15,040	12,058	2,688, 2,836, 2,982
S176A	14,748, 14,893, 15,038	12,058	2,690, 2,835, 2,980
S183A	14,747, 14,894, 15,044	12,058	2,689, 2,836, 2,986
S193A	14,751, 14,899, 15,045	12,058	2,693, 2,841, 2,987
S201A	14,747, 14,894, 15,038	12,058	2,689, 2,836, 2,980
6 S/A	11,976	11,978	−2
Δorf1	12,072	12,074	−2
Δorf2	13,292-14,858	12,074	1,218-2,784
Δorf3	15,298, 15,446, 15,591	12,074	3,224, 3,372, 3,517

^aThe N-terminal methionine of intact flagellin is subject to posttranslational cleavage (data not shown). For the Δorf2 peptide fragment, peaks were observed as broad peaks by heterogeneity.

^bMajor peaks in MALDI-TOF mass spectra.

^cCalculated value by deduced amino acid sequence.

FIG. 1.

MALDI-TOF MS analysis of intact flagellin (A and C) and the N136-K255 peptide (B and D) from the wild-type strains P. syringae pv. tabaci 6605 (A and B) and P. syringae pv. glycinea race 4 (C and D). (Panel B is reprinted from Cellular Microbiology [30] with permission of the publisher.)

In a lysyl endopeptidase-digested peptide mixture from the wild-type strain, three sharp peaks were observed at m/z 15,296, 15,444, and 15,591 (Table 2; Fig. 1D). These corresponded to N136-K255 peptides, which without modification should give [M + H]⁺ of m/z 12,074. Thus, our results indicate that the N136-K255 peptide is modified with glycans with total molecular masses of 3,222 Da, 3,370 Da, or 3,517 Da. Because the Δ values [(observed m/z value) − (m/z value calculated from the peptide sequence)] for intact flagellin and the N136-K255 peptide are almost the same (Table 2), the sites of modification appear to be located between N136 and K255 in the primary amino acid sequence of flagellin.

Previous studies on flagellin from P. syringae pv. tabaci 6605 identified six serine residues (S143, S164, S176, S183, S193, and S201) as sites of glycosylation (30). To evaluate the effects of point mutations at each of the six serine residues on the glycosylation status of flagellin in P. syringae pv. glycinea race 4, one or all of the six serine residues were replaced with alanine. The molecular mass of flagellin from each alanine mutant was then determined. Replacement of any serine by alanine decreased the molecular mass of the modified peptide (Table 2, Δ value) by an average of 534 Da. The mutated form of flagellin in which all six serine residues had been replaced by alanine (i.e., 6 S/A) gave a molecular mass corresponding to that of the unmodified peptide (Table 2). These results suggest that the sites of glycosylatiaon in P. syringae pv. glycinea race 4 and P. syringae pv. tabaci 6605 flagellin are identical. Mass spectra of the N136-K255 peptides from each of the six Ser-to-Ala mutants showed the same heterogeneity at three major peaks as was observed for the wild-type strain (Table 2). The mass average of the intervals between the three peaks in these seven strains was 147 Da. This mass difference suggests that the heterogeneity is derived from the number of deoxyhexose units, which is predicted to give rise to a molecular mass difference of 146 Da per unit.

The mass spectra of intact flagellin or peptide N136-K255 from P. syringae pv. glycinea race 4 and P. syringae pv. tabaci 6605 were compared (Fig. 1). In our previous study (30), the mass spectra of intact flagellin and the peptide from the wild-type strain P. syringae pv. tabaci 6605 showed heterogeneity at two to three major peaks (Fig. 1A and B). However, for the wild-type strain P. syringae pv. glycinea race 4, although the positions of the peaks indicating glycosylation heterogeneity are similar to those of P. syringae pv. tabaci 6605, their relative intensities at higher mass values are significantly greater (Fig. 1C and D). The molecular mass of flagellin from the Δorf2 mutant of P. syringae pv. tabaci 6605 was reported to be quite variable, with a value intermediate between those of the wild type and the Δorf1 mutant (30). The molecular mass of flagellin from the Δorf2 mutant of P. syringae pv. glycinea race 4 also showed heterogeneity in the m/z range of 13,292 to 14,858, with more than 15 peaks.

Peptide mapping of the P. syringae pv. glycinea race 4 flagellin.

An HPLC profile of the proteolytic fragments of flagellin was generated by digestion with endoproteinase Asp-N. When flagellin from P. syringae pv. tabaci 6605 was digested with this endoproteinase, peptides containing glycosylation sites were mapped to three fractions (fraction 41 for D200-A211, fraction 50 for D168-T187, and fraction 66 for D139-F167 and E189-I199) (30). Here we performed the same analysis on flagellin from P. syringae pv. glycinea race 4. N-terminal amino acid sequencing confirmed the presence of the former two peptides in fraction 43 (D200-A211) and fraction 51 (D168-T187). In fraction 67 of P. syringae pv. glycinea race 4, we detected two peptide sequences, DGSAXTMTFQVGS and ETNFXAAIAA (where X stands for an unidentified residue), corresponding to the N-terminal amino acids of D139-F167 (i.e., D139 to S151) and almost the entire sequence of E189-I199, respectively. It was not possible to determine the residual C-terminal sequence of D139-F167 (i.e., N152-F167), because the peptide concentration was too low. During sequence analysis, S143 (fraction 67), S176 (fraction 51), S183 (fraction 51), S193 (fraction 67), and S201 (fraction 43) were found to have anomalous retention times, suggesting that the serine residues had undergone modification. These results are consistent with those of the MS analysis of Ser-to-Ala substitution mutants. We were unable to verify whether S164 also runs anomalously, because this residue is located too far from the N terminus (D139) of the peptide for analysis.

Among these peptides, D200-A211 bears a single glycosylation site at S201, and a sufficient amount of material for structural analysis could be obtained by preparative chromatography. Therefore, peptide D200-A211 derived from either P. syringae pv. tabaci 6605 or P. syringae pv. glycinea race 4 was analyzed further by MS-MS and NMR spectroscopy in order to determine the structure of the modification site.

Sugar composition analysis.

Sugar composition analysis was carried out on intact flagellin from P. syringae pv. tabaci 6605 and P. syringae pv. glycinea race 4. In both pathovars, Rha was identified by the correspondence of the retention time of its derivative with that of the Rha standard as described in Materials and Methods (data not shown).

Determination of d-Rha/l-Rha ratios in glycopeptide D200-A211 and intact flagellin proteins.

For determination of the enantiomeric ratio of Rha residues on S201, glycopeptide D200-A211 was treated with HCl in (S)-2-butanol to form diastereomeric glycosides. For both pathovars, GC analysis of TMS derivatives of (S)-2-butyl rhamnoside yielded a peak corresponding to the l-rhamnoside diastereomer, showing that Rha residues on this peptide were exclusively of the l form (data not shown).

To elucidate the enantiomeric ratio of the Rha residues of the whole flagellin protein in each pathovar, the intact flagellin proteins were also subjected to solvolysis and converted into TMS derivatives of (S)-2-butyl rhamnoside. For P. syringae pv. tabaci 6605, the configuration of Rha was shown to be solely of the l form, with a retention time of 22.0 min (Fig. 2A). By contrast, the flagellin glycan from P. syringae pv. glycinea race 4 yielded two peaks corresponding to d- and l-Rha at a molar ratio of about 1:4, with retention times of 21.5 min and 22.0 min, respectively (Fig. 2B). The assignment of these two peaks was confirmed by GC-MS analysis. MS fragmentation patterns of these derivatives were identical to those of the l-Rha standard. Figure 2C shows the fragmentation patterns from a GC peak with a retention time of 21.5 min.

FIG. 2.

Determination of d-Rha/l-Rha ratios in intact flagellin proteins. (A and B) GC patterns of the trimethylsilylated (S)-2-butyl rhamnosides obtained from P. syringae pv. tabaci 6605 (A) and P. syringae pv. glycinea race 4 (B) flagellins. (C) GC-MS fragmentation patterns from a GC peak with a retention time corresponding to that of authentic trimethylsilylated (R)-2-butyl l-rhamnose in P. syringae pv. glycinea race 4 flagellin. (Inset) Structure of the rhamnose derivative and the expected primary fragment ions.

Structural characterization of glycopeptide D200-A211 by MS analysis.

To characterize the structure of flagellin glycan, the D200-A211 glycopeptides from P. syringae pv. tabaci 6605 and P. syringae pv. glycinea race 4 were subjected to MALDI-QIT-TOF MS and MS-MS analyses. The mass spectra of the two pathovars' glycans were essentially identical. Figure 3 shows the mass spectrum of D200-A211 from P. syringae pv. tabaci 6605. The [M + H]⁺ of the peptide was observed at m/z 1,814 (Fig. 3 inset), while the corresponding calculated value of the sequence DSALQTINSTRA is 1,276. Thus, modification of this peptide increased the molecular mass by 538 Da. This result is consistent with the difference in mass between the modified part of the wild type and that of the S201A mutant (see Table 2 for P. syringae pv. glycinea race 4 and Taguchi et al. [30] for P. syringae pv. tabaci 6605). Figure 3 shows the MS-MS spectrum of the peak at m/z 1,814 ([M + H]⁺). An ion observed at m/z 1,277 corresponds to the calculated value for the DSALQTINSTRA ion. Those found at m/z 1,699 ([M-115 + H]⁺) and 1,162 ([DSALQTINSTRA-115 + H]⁺) are presumably generated by cleavage of the N-terminal Asp. Ions observed at m/z 1,423 and 1,569 correspond to peptides with one and two glycosyl residues with a molecular mass of 146 Da, respectively. These results suggest that the 538-Da glycan is composed of three residues with masses of 246, 146, and 146 Da. The molecular mass difference of 146 Da suggests the presence of a deoxyhexose, consistent with the result from monosaccharide analysis in which only Rha was detected. Thus, the glycan is linked to the serine via two Rha residues. The product with a molecular mass of 246 Da does not correspond to a known saccharide.

FIG. 3.

MALDI-QIT-TOF MS-MS spectrum of the D200-A211 glycopeptide from P. syringae pv. tabaci 6605 flagellin with the respective fragmentation scheme. The D200-A211 glycopeptide, showing [M + H]⁺ at m/z 1,814 (inset), corresponds to a glycopeptide with the sequence ²⁰⁰DSALQTINSTRA²¹¹, in which S201 is modified with a 538-Da moiety. The MS-MS experiment gave product ions at m/z 1,796 ([M + H-H₂O]⁺), 1,699 ([M-Asp + H]⁺), 1,569 ([M-246 + H]⁺), 1,423 ([M-246-Rha + H]⁺), 1,308 ([M-246-Rha-Asp + H]⁺), 1,277 ([M-246−2Rha + H]⁺) and 1,162 ([M-246−2Rha-Asp + H]⁺). In the fragmentation scheme, X stands for a substructure of 246 Da. This substructure was assigned to Qui4N(3-hydroxy-1-oxobutyl)2Me by subsequent NMR experiments.

To obtain structural information on this unidentified residue, accurate mass measurements by ESI-Q-TOF MS analysis were performed. The mass spectra of the peptides derived from the two pathovars were identical (Fig. 4 shows data for P. syringae pv. tabaci 6605). The initial mass spectrum of the D200-A211 glycopeptide showed that the glycan of 538 Da is composed of two deoxyhexose residues and one unknown residue with a mass of 246 Da (Fig. 4A). Furthermore, the unit of two deoxyhexose residues is directly attached to the peptide (at S201). This result is in accordance with the results of the MALDI-QIT-TOF MS-MS analysis (Fig. 3). The fragmentation pattern suggests that these residues are located linearly from the distal end, with the unknown residue first, followed by two deoxyhexose residues. The MS-MS spectrum of the [M + 2H]²⁺ observed at m/z 907.5 for the D200-A211 peptide (Fig. 4) yielded an intense product ion peak at m/z 246.1 (Fig. 4A). The MS-MS analysis of [M + 3H]³⁺ observed at m/z 605.3 for the D200-A211 peptide (Fig. 4) also yielded an intense product ion at m/z 246.1 (data not shown). These data are consistent with the proposal that the glycopeptide (1,813 Da) includes a residue of 246 Da. By contrast, MS-MS spectra of fragment ions at m/z 784.9, 711.9, and 638.8 did not include the product ion at m/z 246.1 (Fig. 4B to D). These observations confirm that the unit of 246.1 Da is located at the distal end of the glycan. Accurate mass analysis within 5 ppm of the calculated value of this unknown unit exhibited m/z 246.134 (C₁₁H₂₀NO₅, calculated 246.1335) for the ion.

FIG. 4.

ESI-Q-TOF mass spectra of the D200-A211 glycopeptide from P. syringae pv. tabaci 6605 flagellin. (A to D) MS-MS spectra of ions observed at m/z 907.5 ([M + 2H]²⁺), 784.9 ([M-246 + 2H]²⁺), 711.9 ([M-246-Rha + 2H]²⁺), and 638.8 ([M-246−2Rha + 2H]²⁺), respectively. The peak observed at m/z 605.3 corresponds to [M + 3H]³⁺. The b- and y-series ions, shown both in the sequence of this peptide and in the MS-MS spectra, originated from the N and C termini (22), respectively. A substructure of 246 Da was assigned to Qui4N(3-hydroxy-1-oxobutyl)2Me by subsequent NMR experiments.

Determination of glycan structure on D200-A211 by NMR analysis.

The structure of glycan on D200-A211 from P. syringae pv. tabaci 6605 flagellin was elucidated by NMR spectroscopy. Signals of two Rha residues and the unidentified substructure were observed in addition to those derived from the peptide backbone. These signals were assigned based on ¹H-¹H correlations on DQF-COSY and TOCSY and on ¹H-¹³C correlations on HSQC and HMBC spectra (Table 3; Fig. 5 and 6B). The binding of an l-Rhap residue to S201 was confirmed by HMBC correlation between the β-carbon of S201 and H-1 of the l-Rhap (Fig. 5 and 6B). The second l-Rhap is attached at C-2 of the l-Rhap linked to S201, as deduced by HMBC correlation of C-2-H-1 (Fig. 5 and 6B) and low-field shift of C-2 (79.9 ppm) (Table 3). The structure of the terminal saccharide was identified as 4-amino-4,6-dideoxyglucose (Qui4N; trivial name, viosamine) from the presence of a C-6 methyl group, large H-H coupling constants, and high-field shift of C-4 (58.3 ppm) (Table 3). The attachment position of Quip4N to l-Rhap was shown to be C-3 by HMBC correlations of l-Rhap C-3 to Quip4N H-1 and of Quip4N C-1 to l-Rhap H-3 (Fig. 5 and 6B). Further modification of the Quip4N residue by O methylation of C-2 was indicated by HMBC correlations (Fig. 5 and 6B). The presence of these links was supported by data from the NOESY experiments (Fig. 5 and 6A). The presence of the 3-hydroxybutyryl group and its attachment to Quip4N through an amide link was revealed by COSY and HMBC analyses (Table 3; Fig. 5 and 6B). Moreover, the elemental formula of the modified Quip4N residue (C₁₁H₂₀NO₅) coincides with the result of accurate mass analysis.

TABLE 3.

Assignment of NMR signals^a of glycan on S201 of the D200-A211 peptide

Residuea	Position	1H NMR			13C NMR δ (ppm)
		δ (ppm)	Multiplicityb (W1/2, Hz)	J (Hz)
Ser201	Amide carbonyl				172.5c
	α	4.599	dd	4.3, 3.7	55.53
	β	4.055	dd	10.0, 4.3	67.67
	3.732	dd	10.0, 3.7
α-l-Rhap 1	1	4.888	br s (4.5)		99.81
	2	3.977	br s (7.2)	79.90
	3	3.780	dd	9.7, 3.1	71.66
	4	3.440	dd	9.7, 9.6	73.71
	5	3.541	dq	9.6, 6.0	70.84
	6	1.273	d	6.0	18.51
α-l-Rhap 2	1	4.979	br s (4.5)		103.55
	2	4.264	br dd	3.1, 1.6	71.38
	3	3.900	dd	9.8, 3.1	81.13
	4	3.591	dd	9.8, 9.8	72.83
	5	3.756	dq	9.8, 6.1	71.04
	6	1.271	d	6.1	18.39
β-d-Quip4N (4-amino-4,6-dideoxy-β-d-glucopyranosyl)	1	4.726	d	8.0	105.20
	2	3.126	dd	9.2, 8.0	84.94
	3	3.517	dd	10.2, 9.2	74.50
	4	3.619	dd	10.2, 10.0	58.26
	5	3.535	dq	10.0, 6.0	72.56
	6	1.199	d	6.0	18.64
	2-O-Me	3.620	s	61.77
N-(3-hydroxy-1-oxobutyl)	1′	2.402	dd	14.1, 7.9	176.2c
	2′	2.422	dd	14.1, 5.5	46.87
	3′	4.188	ddq	7.9, 5.5, 6.3	66.73
	4′	1.218	d	6.3	23.77

^aObtained at 800.33 MHz for ¹H and 125.76 MHz for ¹³C in D₂O at 298 K. Numbers of rhamnosyl residues are based on the order from the reducing end.

^bs, singlet; d, doublet; t, triplet; q, quartet; br, broad; W_1/2, width at half height.

^cRead from a position of an HMBC cross peak.

FIG. 5.

Chemical structure and selected NMR correlations (HMBC and NOESY) of glycan attached to S201 of the D200-A211 peptide from P. syringae pv. tabaci 6605 and P. syringae pv. glycinea race 4.

FIG. 6.

Two-dimensional NMR spectra of β-d-Quip4N(3-hydroxybutyryl)2Me-(1→3)-α-l-Rhap-(1→2)-α-l-Rhap-(1→O^S201)-peptide D200-A211. (A) NOESY; (B) HMBC.

Quip4N was estimated to be in the d configuration based on the ¹³C NMR chemical shift of the C-1 signal (18, 35). The value of 105.20 ppm indicated an opposing absolute configuration of β-Quip4N and α-l-Rhap in (1→3) linkage. An alternative configuration of Quip4N would lead to a smaller chemical shift by ca. 3 ppm. Therefore, the structure of the glycan was determined to be β-d-Quip4N(3-hydroxy-1-oxobutyl)2Me-(1→3)-α-l-Rhap-(1→2)-α-l-Rhap (Fig. 5). The structural identity of the glycan on D200-A211 from P. syringae pv. glycinea race 4 with that from P. syringae pv. tabaci 6605 was confirmed by NMR analyses.

DISCUSSION

Bacterial flagellin is one of the best-studied molecules containing pathogen- or microbe-associated molecular patterns that can activate basal defense in the form of nonhost resistance in plants (14). The synthetic oligopeptide flg22, which was designed from an N-terminal conserved sequence in the D0 interior domain of flagellin from P. aeruginosa, elicits basal defense responses of plants (7). Thus, flg22 has been defined as a general elicitor. In Arabidopsis thaliana, recognition of flg22 is mediated by its binding to a leucine-rich repeat transmembrane receptor kinase, FLS2 (5, 9). We have demonstrated the importance of flagellin glycosylation, located on the surface of the flagellar filament, as a determinant of host specificity (29, 30, 31). Plant HRs caused by flagellin are attenuated when the flagellin is modified by glycosylation in the manner of a compatible bacterium (29). Thus, there is a plausible hypothesis that flagellin glycosylation plays an important role in masking this pattern in order to avoid recognition by the host plant.

Our analysis of P. syringae pv. glycinea race 4 flagellin shows that six serine residues at positions 143, 164, 176, 183, 193, and 201 are glycosylated. The mass value of the major glycan on each serine was about 534, while heterogeneity due to the addition of one or two units with a mass value of about 147 was also observed. These results are concordant with our previous work on P. syringae pv. tabaci 6605 flagellin (30). Analysis of intact flagellin showed that d-Rha is present in the glycan of P. syringae pv. glycinea race 4, whereas Rha from P. syringae pv. tabaci 6605 flagellin glycan is exclusively of the l form. Studies on the D200-A211 glycopeptide derived from either P. syringae pv. tabaci 6605 or P. syringae pv. glycinea race 4 flagellin revealed structural identity of the major glycan on S201, where only l-Rha was found. It is conceivable that minor glycan species comprising one or two more Rha residues might attach via S201. We believe that heterogeneity of glycosylation is not confined to one particular modification site, because it was present in each of the six Ser-to-Ala mutants. Differences between P. syringae pv. tabaci 6605 and P. syringae pv. glycinea race 4 may exist in the structure of the glycan with the extra Rha residue(s). Further investigation of glycan moieties attached to serine residues other than S201 is also required in order to explain the host specificity of these pathovars in terms of the structure of flagellin.

It is intriguing to explore the relationship between the content of d-Rha and glycosylation heterogeneity. For example, the relative intensities of the peaks observed in MALDI-TOF MS analyses might reflect the content of d-Rha. In accordance with the heterogeneity of flagellin glycosylation, d-Rha may be dispersed among all or some of the six glycans. Alternatively, d-Rha may be attached to a specific serine residue(s) in P. syringae pv. glycinea race 4 flagellin. If so, such a residue(s) may be significant in determining host specificity. More-precise analysis of each glycan and utilization of Ser-to-Ala mutants will be helpful in determining the localization of d-Rha in P. syringae pv. glycinea race 4 flagellin.

Rha is also reported to be a common major component of lipopolysaccharides (LPS) in these two pseudomonad pathogens (17). Although naturally occurring Rha is present mainly in the l form, both the d and l forms have been found in O antigens of P. syringae. Furthermore, emerging patterns in the chain of rhamnan are thought to correlate with serogroups (17). This may explain the significance of the chirality of Rha in LPS. LPS is recognized by mammals differently according to its constituent parts. Internal/conserved domains, such as lipid A, are important for the innate immune response, whereas the surface-exposed/highly variable domains, such as O antigen, determine antigenic specificity (21). This prototype of domain distinction of activities is also applicable to flagellin. Thus, it is reasonable to propose that flagellin glycans on the putative surface-exposed domain are responsible for determining host specificity (38). A complete structural characterization of flagellin should help to elucidate how plants recognize flagellin glycan and how bacteria evade such recognition.

In animals, flagellin recognition is mediated by Toll-like receptor 5 (TLR5) (12). The conserved N-terminal region of flagellin is reported to be important for binding to TLR5 (27). Notably, the role of flagellin glycosylation of animal-pathogenic bacteria in innate immunity is just beginning to be elucidated. For P. aeruginosa strains PAK and PAO1, the virulence of flagellar glycosylation mutants in mice was significantly attenuated (2). In both of these strains, flagellin glycosylation plays an important role in the ability of flagellin to stimulate interleukin-8 release from human lung carcinoma cells (33). These results suggest that flagellin glycans might be responsible for the stimulation of inflammation. In P. aeruginosa strain PAK, a glycan consisting of 11 monosaccharides is linked to the flagellin protein through a Rha residue at the two glycosylation sites (25). Thus, Rha is a common component of flagellin glycan in P. syringae and P. aeruginosa.

The distal residue of the glycan on S201 is the modified unique amino sugar Qui4N. MALDI-TOF MS analyses of the six Ser-to-Ala mutants suggest that the major glycan on each serine residue is a trisaccharide composed of modified Qui4N and two Rha residues. Only Rha was detected by sugar composition and enantiomeric ratio analyses by GC despite the presence of the modified Qui4N. In Vibrio LPS, 4-amino-4,6-dideoxymannose was produced in abundance by mild acid hydrolysis (15). Thus, 4-amino-4,6-dideoxyhexoses may be destroyed by more vigorous chemical procedures. On the other hand, in Bacillus anthracis exosporium, Qui4N(3-hydroxy-3-methyl-1-oxobutyl)2Me was detected after methanolysis and acetylation (6). It might be possible to detect modified Qui4N in our study by adopting such methods. P. aeruginosa LPS is known to be rich in unusual amino sugars, some of which have hydroxybutyryl groups instead of acetylation (16). This suggests the existence of a common synthetic pathway for flagellin glycans and LPS. Qui4N was also detected as one of the components of P. aeruginosa PAK flagellin glycan (25). Thus, there is a clear structural similarity between the glycans of P. syringae and P. aeruginosa a-type flagellins, in addition to the attachment of Rha to the peptide backbone. The Orf1 and Orf2 products of the flagellin glycosylation island of P. syringae are similar to OrfN (FgtA [for flagellar glycosyltransferase]) of the P. aeruginosa a-type strain PAK (32 and 38% identity, respectively) (30, 31). Orf1 and Orf2 also show similarity to PA1091 (FgtA) in the P. aeruginosa b-type strain PAO (43 and 34% identity, respectively). OrfN and PA1091 are considered to transfer deoxyhexose to the protein backbone (25, 34), indicating similarity to the function of Orf1 in P. syringae. Thus, these putative glycosyltranferases may possess a common enzymatic activity in pseudomonad pathogens. We propose that the orf1 and orf2 genes in the P. syringae glycosylation island be renamed fgt-1 and fgt-2, respectively. Notably, the homologue of orfA, which belongs to the PAK glycosylation island but is not found in P. aeruginosa PAO1, is located upstream of the flagellar gene cluster in two P. syringae pathovars whose whole genomes have been sequenced, P. syringae pv. tomato DC3000 and P. syringae pv. phaseolicola 1448A. The homologues from both of these pathovars display 68% identity to PAK OrfA at the amino acid level. orfA possesses homology to the vioA gene, which is responsible for the synthesis of viosamine, i.e., Qui4N (1). Qui4N is one of the PAK strain-specific glycan components in P. aeruginosa flagellin but was not detected in strain PAO1 (25, 34). Thus, it will be interesting to investigate the role of the orfA homologue in flagellin glycan synthesis in P. syringae.

Although the structures of P. syringae pv. tabaci 6605 and P. syringae pv. glycinea race 4 flagellin glycans are similar, differences were observed in the content of d-Rha. The chirality of the Rha residues may be one of the significant determinants of the elicitor activity of flagellin. The interaction of P. syringae with its host or with a nonhost plant is the most advanced system for elucidating the biological significance of flagellin glycosylation in the interaction of bacteria with eukaryotes. Therefore, our findings are important in defining biological activity, such as bacterial virulence and host specificity, in terms of molecular structure.