N-acetylmuramic acid as capping element of alpha-D-fucose-containing S-layer glycoprotein glycans from Geobacillus tepidamans GS5-97T.

Geobacillus tepidamans GS5-97(T) is a novel Gram-positive, moderately thermophilic bacterial species that is covered by a glycosylated surface layer (S-layer) protein. The isolated and purified S-layer glycoprotein SgtA was ultrastructurally and chemically investigated and showed several novel properties. By SDS-PAGE, SgtA was separated into four distinct bands in an apparent molecular mass range of 106-166 kDa. The three high molecular mass bands gave a positive periodic acid-Schiff staining reaction, whereas the 106-kDa band was nonglycosylated. Glycosylation of SgtA was investigated by means of chemical analyses, 600-MHz nuclear magnetic resonance spectroscopy, and electrospray ionization quadrupole time-of-fight mass spectrometry. Glycopeptides obtained after Pronase digestion revealed the glycan structure [-->2)-alpha-L-Rhap-(1-->3)-alpha-D-Fucp-(1-->](n=approximately 20), with D-fucopyranose having never been identified before as a constituent of S-layer glycans. The rhamnose residue at the nonreducing end of the terminal repeating unit of the glycan chain was di-substituted. For the first time, (R)-N-acetylmuramic acid, the key component of prokaryotic peptidoglycan, was found in an alpha-linkage to carbon 3 of the terminal rhamnose residue, serving as capping motif of an S-layer glycan. In addition, that rhamnose was substituted at position 2 with a beta-N-acetylglucosamine residue. The S-layer glycan chains were bound via the trisaccharide core -->2)-alpha-L-Rhap-(1-->3)-alpha-L-Rhap-(1-->3)-alpha-L-Rhap-(1--> to carbon 3 of beta-D-galactose, which was attached in O-glycosidic linkage to serine and threonine residues of SgtA of G. tepidamans GS5-97(T).

Optical rotation measurement was performed as described previously (8). Chemical deglycosylation of S-layer glycoprotein was performed by the method of Edge et al. (16). SDS-PAGE with Coomassie Blue staining for protein and periodic acid-Schiff staining for carbohydrates was carried out as described previously (17). N-terminal sequencing of glycoproteins and glycopeptides according to standard protocols (7) was done after semidry blotting (3). Electron microscopy of thin-sectioned and freeze-etched preparations of bacterial cells as well as of negatively stained assembled S-layer glycoprotein was performed according to published procedures (18).
Isolation and Purification of S-layer Glycoprotein Species-The isolation of S-layer glycoprotein and the preparation of S-layer self-assembly products followed published methods (17). For separation of individual glycoprotein species, the S-layer glycoprotein (self-assembly products) was suspended at a concentration of 10 mg/ml in 60 mM Tris-HCl buffer, pH 6.8, containing 2.5% SDS, 5% ␤-mercaptoethanol, 10% glycerol, and 0.025% bromphenol blue, incubated for 2 min at 100°C, and pelleted. Separation of batches of 30 mg of S-layer glycoprotein by preparative SDS-PAGE with 25 mM Tris-HCl buffer, pH 8.3, containing 192 mM glycine and 0.1% SDS (running buffer), was performed over 21 h at a constant power of 50 watts at room temperature, using the Bio-Rad model 491 Prep Cell electrophoresis apparatus. The supernatant was loaded onto a gel tube (diameter ϭ 38 mm) consisting of 10 ml of a 4% stacking gel (in 125 mM Tris-HCl buffer, pH 6.8) and 60 ml of 6% resolving gel (in 325 mM Tris-HCl buffer, pH 8.8). Elution was done with running buffer without SDS at a flow rate of 0.8 ml/min; fractions of 8 ml were collected, and the total eluent volume was 1200 ml. Fractions were pooled according to the protein banding pattern on 7.5% SDS-polyacrylamide gels upon silver staining (19). Individual pools were dialyzed at 4°C against distilled water until salt-free and lyophilized.
Preparation of S-layer Glycopeptides-S-layer glycopeptides were obtained after proteolytic degradation (20 h) of S-layer self-assembly products with V8 protease (Sigma). Purification of glycopeptides is described in detail elsewhere (20), except that chromatofocusing was performed using a pH gradient between 8.5 and 5.5. The glycopeptide pools (designated CF I through CF VI) were desalted using Sephadex G-10 and finally purified by semi-preparative RP(C18)-HPLC (8).
HPAEC Analysis of MurNAc-To distinguish which isomer of Mur-NAc was present in the S-layer glycan, hydrolysates were analyzed by HPAEC according to Ref. 21 with slight modifications. Briefly, ϳ0.5 mg of S-layer glycopeptide preparations CF II and CF IV were hydrolyzed using 25%(v/v) trifluoroacetic acid at 110°C for 4 h. The samples were analyzed on a PA-1 column (Dionex); elution was started isocratically in 16 mM NaOH, containing 2 mM sodium acetate (0 -21 min), followed by a linear increase to 100 mM NaOH (21-27 min). Then in 100 mM NaOH, the sodium acetate concentration was raised linearly to 75 mM sodium acetate (27-32 min), before isocratic elution of muramic acid in 100 mM NaOH, containing 75 mM sodium acetate (32-59 min). Subsequently, the column was washed and equilibrated in 200 mM NaOH (59 -65 min) and in 16 mM NaOH, containing 2 mM sodium acetate (65-80 min), respectively. For identification and verification of the peaks the sample was spiked with authentic (R)-N-acetylmuramic acid standard (Sigma).
Mass Spectrometry-Positive mode electrospray Q-TOF mass spectrometry (ESI Q-TOF MS) was performed on a Waters Micromass Q-TOF Ultima Global apparatus (Waters Micromass, Manchester, UK). For instrument tuning and calibration, the typical fragment ions of desialylated, dabsyl-labeled glycohexapeptide from bovine fibrin were used (22). The sample was subjected to off-line infusion ESI Q-TOF MS after dilution in 50% acetonitrile containing 0.5% formic acid. Spectra acquisition was performed over an m/z range from 900 to 3000 Da using 3.1-kV capillary and 100-V cone voltage. Desolvation gas flow was set at 300 liters/h and cone gas at 50 liters/h. Sample was injected at a flow rate of 5 l/min. The instrument was controlled by MassLynx 4.0 software (Waters Micromass, Manchester, UK).
NMR Spectroscopy-NMR spectra were recorded on a Bruker Avance DRX 600 NMR spectrometer at resonance frequencies of 600.13 MHz for 1 H and 150.90 MHz for 13 C (3). For structural assignment ϳ5 mg of purified glycopeptide CF II dissolved in 0.6 ml of D 2 O was used. All experiments were done at a temperature of 300.0 K, stabilized by a Bruker BCU05 cooling unit. Experiments included gradient selected double-quantum filtered correlation spectroscopy, TOCSY (10 -120 ms MLEV-17 spin-lock), NOESY (150 -300 ms mixing time including a bipolar purge gradient), gradient selected HSQC and HMBC (5 Hz typical long range coupling constant), gradient selected HSQC-TOCSY and HSQC-NOESY, and one-dimensional diffusion difference spectra. Detailed parameters for the used pulse sequences have been reported previously (20,23), and the diffusion difference procedure was followed (24).
For analysis of the peptide and the amino sugar portion, most of the experiments were additionally performed on solutions in 0.6 ml of 90% H 2 O, 10% D 2 O using pulsed field gradients for coherence selection and 3-9-19 WATERGATE with magic angle gradients for solvent suppression as described previously (23). After running the spectral series in H 2 O, the sample was again lyophilized and re-dissolved in D 2 O.
Processing was done off-line on Silicon Graphics workstations using the Bruker software XWIN-NMR 3.1. Detailed spectral analysis was accomplished with the NMR assignment and integration software Sparky 3 (provided by T. D. Goddard and D. G. Kneller, University of California, San Francisco). The chemical shifts are average values from all assigned two-dimensional cross-peaks with standard deviations of 0.003 ppm for 1 H and 0.015 ppm for 13 C. The proton-proton coupling constants were extracted from experimental or calculated (program XSIM 971120, provided by K. Marat, University of Manitoba, Winnipeg, Manitoba, Canada) 1 H NMR spectra. 1 H-13 C coupling constants were taken from appropriate slices of coupled two-dimensional HSQC spectra.

RESULTS
General Description of the Organism-Electron microscopy of freeze-etched cells of G. tepidamans GS5-97 T revealed that the bacterium is completely covered with an oblique S-layer glycoprotein lattice. The recently determined lattice parameters (a ϭ 11.2 nm, b ϭ 7.9 nm, ␥ ϳ 80°) were confirmed in that culture (14). The overall degree of glycosylation of the S-layer protein is ϳ8.5% (w/w). On SDS-PAGE gels, the S-layer glycoprotein showed a four-banded protein staining pattern, with the individual bands corresponding to apparent molecular masses of 106, 123, 140, and 166 kDa. The three high molecular mass bands, but not the 106-kDa band, gave an intensive periodic acid-Schiff staining reaction on the gel. Upon chemical deglycosylation of the S-layer glycoprotein, a single band corresponding to the 106-kDa band remained visible on the gel, indicating that this band represents the nonglycosylated protomeric unit of the S-layer glycoprotein lattice of G. tepidamans GS5-97 T .
Characterization of S-layer Glycoprotein Species-Starting with ϳ350 g (wet pellet) of biomass of G. tepidamans GS5-97 T , ϳ500 mg (dry weight) of S-layer glycoprotein was obtained. Negative staining of the self-assembly products formed upon dialysis of the guanidinium hydrochloride-extracted S-layer glycoprotein revealed predominantly open-ended cylinders of glycoprotein monolayers, with an average diameter of 200 nm and an average length of 1.6 m. 60 mg of that material was used for purification of the nonglycosylated S-layer species GPS I and the three glycosylated forms GPS II through GPS IV by preparative electrophoresis. The obtained yields were 2.1 (GPS I), 5.0 (GPS II), 12.3 (GPS III), and 22.9 mg (GPS IV). N-terminal sequencing revealed identical N termini with the sequence ATNVDAVVN on the individual S-layer species. Concerning the S-layer glycan portion, GPS II, GPS III, and GPS IV revealed Rha and Fuc in an approximate molar ratio of 1:1 as major carbohydrate constituents.
For characterization of the S-layer glycoprotein, from the proteolytic degradation mixture of 1 g (dry weight) of S-layer self-assembly products, six crude glycopeptide pools, designated CF I through CF VI, were obtained after chromatofocusing. Yields of glycopeptides were 1.6 mg for CF I (pH range of elution 8.9 -8.3), 9.8 mg for CF II (pH range 8.1-7.9), 10.4 mg for CF III (pH range 7.8 -7.6), 5.3 mg for CF IV (pH range 7.2-7.0), 8.9 mg for CF V (pH range 6.9 -6.8), and 1.7 mg for CF VI (pH range 6.7-6.0). After desalting and final purification by RP(C18)-HPLC, the CF pools were analyzed by HPAEC/PAD for their carbohydrate composition, indicating rhamnose and fucose as major components in addition to minor amounts of galactose, GlcNAc, and N-acetylmuramic acid in each pool. Based on amino acid analysis and initial NMR measurements, CF II and CF IV were selected for detailed investigations. NMR Studies-Structural investigations by NMR spectroscopy were performed on a sample of 5.3 mg of purified CF II. In the 1 H NMR spectrum, two intense signals were observed in the anomeric region ( Fig. 1, ␦ ϭ 5.143 and 4.995) as well as two intense aliphatic CH 3 groups (␦ ϭ 1.312 and 1.195) suggesting two deoxy sugars for the repeating unit. Detailed analysis of these spin systems revealed a rhamnose (sugar E; for the indication of the residues see Fig. 2) and a fucose (sugar F) residue, both as pyranosyl units, and according to the measured coupling constants for the anomeric signals (Table I, 3 J H,H ϭ 1.0 and 3.7 Hz and 1 J H,C ϭ 170.0 and 169.7 Hz) both in ␣-configuration. In two-dimensional HMBC spectra, crosspeaks via the glycosidic bonds ( Fig. 2 and Table II) proved a 133 linkage from Rhap to Fucp and a 132 linkage from Fucp to Rhap. Regarding the absolute stereochemistry of the two sugar units, the observed glycosylation effects on the 13 C chemical shifts of the anomeric carbons (Table III) clearly required a consecutive change between the D-and the L-series of the two moieties. In addition, the measured optical rotation of ϩ65°r esulted in a calculated molar rotation of ϩ189°for the glycopeptide preparation, which fits to the calculated value of ϩ232°f or the molar rotation of an L-Rhap-D-Fucp polysaccharide (due to the strong levorotation of L-Fuc, an inverse assignment would lead to a pronounced negative value for the optical rotation). Consequently, the repeating unit of the S-layer glycan possesses the structure [32)-␣-L-Rhap-(133)-␣-D-Fucp- (13] n . As derived from NMR integration, the average number (n) of repeating units of the mature glycan chain is about 20. To our knowledge, this is the first report on an S-layer glycoprotein glycan containing ␣-D-Fuc residues.
The linkage region to the peptide served as starting point for the more elaborate analysis of the distal and proximal parts of CF II. Typical amino acid signals of a serine residue, O-glycosidically linked to a hexose, could be identified. The corresponding two-dimensional HMBC and two-dimensional NOESY cross-peaks between the anomeric center of the linkage sugar and the serine CH 2 group are listed in Table II. Compared with the about 20 times more abundant signal intensity of the disaccharide repeating unit, these signals as well as those for all other terminal residues were very weak in all NMR spectra. A detailed NMR analysis was only performed with CF II. In CF IV, threonine signals were identified; however, the homogeneity of the sample was not sufficient to get reliable linkage information. In CF II, the monosaccharide attached to the serine residue turned out to be a galactopyranose (sugar J) in ␤-configuration (Table I, 3 J H,H ϭ 7.2 Hz and 1 J H,C ϭ 161.1 Hz), connected to the polymer at position 3 according to the 13 C chemical shift of 81.01 ppm. A two-dimensional HMBC crosspeak from carbon 3 of that residue led to an anomeric proton at 5.026 ppm. The trace at this chemical shift in a two-dimensional TOCSY spectrum (mixing time ϭ 120 ms) revealed a signal in the aliphatic region indicative of a 6-deoxy sugar, which was identified as ␣-Rha (sugar I, 1 J H,C ϭ 171.4 Hz). In addition, three further rhamnose residues (sugars C, G, and H) and one additional ␣-Fuc residue (sugar D) could be assigned besides the repeating unit constituents. The very small variability in the carbohydrate building blocks of the S-layer glycopeptide obscured the NMR analysis because of severe signal overlap in the spectra.
The comparison of the NMR data from CF II with those of the recently published S-layer glycopeptide from G. stearothermophilus NRS 2004/3a (3) enabled further signal assignments. In that organism, the polyrhamnan S-layer glycan chain is connected via a 33)-␣-L-Rhap-(133)-␣-L-Rhap(13 core either to a ␤-D-Galp-Ser or a ␤-D-Galp-Thr linkage unit. In analogy to these data, the same rhamnose disaccharide could be found 133-linked to the terminal galactose of CF II (sugars H and I).
From rhamnose H heteronuclear long range as well as nuclear Overhauser effect correlations established an additional rhamnose G in an ␣-133 linkage (Fig. 2). This latter Rhap residue provides the attachment site to the polysaccharide via an  Table I and Fig. 5). Table I and Fig. 5) and from GalpJ to the serine K. ␣-132 linkage to fucose F of the repeating unit. The absolute stereochemistry could be verified by analysis of the glycosylation effects on the 13 C chemical shifts of the anomeric carbons as well as of the carbons at the linkage position (Table III). Starting from the established general configuration of the L-Rhap-D-Fucp disaccharide, the three core rhamnoses belong to the L-series and the galactose to the D-series. To observe the amide proton signals and hence the full amino acid spin systems of the peptide moiety, all necessary experiments were repeated in a solution of 90% H 2 O, 10% D 2 O (Fig.  3). The NMR spectra after this solvent exchange showed significant 1 H and 13 C chemical shift differences for the terminal residues and are listed as data set II in Table I. Essentially, nuclear Overhauser effect and cross-peaks arising from Overhauser effects in the rotating frame between the NH and CH␣ protons of the adjacent amino acid (Table II)  The structural investigation of the nonreducing end of CF II turned out to be rather surprising. No signals for common nonreducing terminating elements, such as methyl groups, phosphates, or glycerol (12), could be found in the NMR spectra. As seen in Fig. 1, there is one small anomeric signal left for a rhamnose residue (sugar C). This monosaccharide is 133linked to the tiny fucose spin system D showing only very small 1 H and 13 C chemical shift deviations, mainly at position 2 and 3, compared with repeating unit fucose F. The terminal rhamnose residue (sugar C) revealed substitutions at positions 2 and 3, as both carbons are deshielded to 77.5 and 80.1 ppm, respectively (Table I).

FIG. 2. HMBC spectrum of glycopeptide CF II showing the glycosidic linkages between the sugar residues (A-F, see
In the 1 H anomeric region there is a well separated signal at a chemical shift of about 5.3 ppm (Fig. 1). The sample in H 2 O gave a TOCSY trace for this monosaccharide in the chemical shift region of the NH protons indicating an amino sugar (Fig.  3). Additionally, there was a second sugar spin system in that area including an anomeric signal at 4.5 ppm, partly overlapping with galactose J-H1 (Fig. 1). Both amino sugars could be assigned to be of the 2-acetamido-2-deoxyglucose type, the first one in ␣-( 3 J H,H ϭ 3.1 Hz and 1 J H,C ϭ 170.5 Hz) and the latter in ␤-configuration ( 3 J H,H ϭ 7.9 Hz and 1 J H,C ϭ 162.2 Hz). Additionally, the first glucosamine is substituted at position 3 with an 3O-1-carboxyethyl residue, thus corresponding to muramic acid. Diffusion difference experiments (20,24) clearly proved these amino sugars as being constituents of CF II, as no relative intensity changes were seen between these signals and the rest of the glycan chain upon alteration of the pulsed field gradient amplitude in the stimulated echo experiments. Detailed analysis was finally possible following the aforementioned observation of significant shift differences for this terminal part after exchanging the solvent from D 2 O to H 2 O. Repetition of the NMR experiments after re-exchanging the solvent back to D 2 O moved these signals even further, resulting in a third data set III (Table I). These normally undesirable effects on the chemical shifts, possibly due to changes in the pH or salt concentration, allowed for the elucidation of the complete spin system of the two amino sugars and consequently for determination of the connectivity (Table II). The ␤-D-GlcpNAc residue A is 132-linked to the ␣-L-Rhap C, as inferred from a long range cross-peak between its anomeric proton and carbon 2 of that rhamnose residue. The second set of signals corresponding to the ␣-configured GlcpNAc residue revealed the presence of a lactyl ether group at carbon 3. This ␣-D-MurNAc B is 133-linked, as shown by the appropriate HMBC crosspeak from its H-1 to C-3 of sugar C (Fig. 2).
N-terminal Sequencing of CF IV-As mentioned previously, full NMR analysis of CF IV was impossible because of insufficient homogeneity of the sample. Only the presence of a Thr residue could be identified unambiguously. Therefore, this sample was split, and one part was subjected to N-terminal sequencing and the other to MS analysis. The major portion of this fraction possessed the sequence TA, but there was also a minor portion present with the sequence TQ.
Mass Spectrometry-Mass spectra were recorded for glycopeptides CF II and CF IV and interpreted on the basis of the known S-layer glycan structure elucidated for CF II. Based on the NMR and on the compositional data, the theoretical average masses of the glycopeptide molecules, including the expected numbers of disaccharide repeating units, were calculated and compared with the masses calculated from the experimentally acquired data under consideration of the different charge-specifying adducts (Table IV). Charge states from ϩ3 to ϩ5 were observed with the ϩ4 state giving the best signals, which were therefore used to calculate glycopeptide masses ( Fig. 4 and Table IV). Multiple peaks of the same charge state were observed for each glycopeptide due to different combinations of the charge-specifying adducts with hydrogen, sodium, and potassium ( Fig. 4 and Table V). In CF II, the overall degree of polymerization of disaccharides varies between n ϭ 18 and n ϭ 22 with n ϭ 20 being the most abundant form. The spectrum of CF IV, which was insufficient for NMR investigations, was evaluated in analogy to the CF II spectrum. The masses exactly matched with the S-layer glycans composed of 18 to 22 disaccharide repeats bound to a threonine residue, with the second amino acid being a glutamine. Additionally, a set of signals corresponding to a minor form of the glycopeptide containing an alanine instead of a glutamine was detected as well. This concurred with the compositional data. Thus, it is conceivable to assume that on SgtA two types of threoninelinked S-layer O-glycans are present with the sequence signa-TABLE II NMR connectivity cross-peaks between the individual residues within the S-layer glycopeptide CF II as derived from two-dimensional HMBC as well as from two-dimensional NOESY or two-dimensional ROESY spectra For the indication of the residues see Table I  tures TQ and TA. As a result, the proposed structure of the different glycopeptides from G. tepidamans GS5-97 T , derived from one-and two-dimensional NMR investigations and ESI Q-TOF MS measurements, is shown in Fig. 5.

DISCUSSION
S-layer glycoproteins are one of the best investigated systems of prokaryotic protein glycosylation. If present, they are among the most abundant cellular proteins, indicating their pivotal role for bacterial organisms. Although no precise functions of S-layers in general, and S-layer glycoproteins in particular, have been determined thus far, it can be assumed that they provide a selection advantage for bacteria under the competitive conditions of the natural habitat (e.g. provision of a hydrophilic coat comparable with LPS to bacteria; involvement in general cell surface phenomena). This is supported by the observation that the S-layer glycans present in fresh isolates may be lost after prolonged cultivation of the bacteria in rich laboratory media (2). Usually bacterial S-layer glycans consist strain-specifically of long, highly variable polysaccharide chains, made of identical repeating units with up to 100 -120 monosaccharide residues, O-glycosidically linked to the S-layer polypeptide (2).
In this contribution, we report on the unusual S-layer glycan structure of G. tepidamans GS5-97 T . The carbohydrate chains consist of disaccharide repeating units with the structure [32)-␣-L-Rhap-(133)-␣-D-Fucp- (13] nϭ18 -22 . L-Rhamnose is a frequent constituent of bacterial polysaccharides, including Slayer glycans (2,25). D-Fucose, on the other hand, is a rare constituent of bacterial polysaccharides, which is almost exclusively present in LPS O-antigens (26,27). Because in S-layer glycans of Gram-positive bacteria typical sugar constituents of LPS are frequently found, it is reasonable to assume that in Gram-positive bacteria the ability to glycosylate proteins was acquired during evolution by lateral gene transfer from Gramnegative organisms. This is supported by the presence of transposases within recently sequenced chromosomal S-layer glycan biosynthesis (slg) gene clusters of different bacteria (9), allowing the division of the cluster into distinct groups of genes with similar G ϩ C contents. Depending on the structure of the encoded S-layer glycan, polycistronic slg clusters are ϳ16 -25 kb in size, and based on data base comparison, they contain typical glycan biosynthesis components, including nucleotide sugar pathway genes that are clustered in an operon, sugar transferase genes, glycan-processing genes, and transporter genes (9). slg gene clusters from Bacillaceae seem to be much less organized than the clusters encoding the biosynthesis of other bacterial polysaccharides, such as LPS O-antigens of Gram-negative bacteria (28) or exopolysaccharides of lactic acid bacteria (29). Housekeeping genes that map outside the slg gene clusters are additionally required for S-layer protein glycosylation.
Investigation of the slg gene cluster of G. stearothermophilus NRS 2004/3a, which possesses a polyrhamnan S-layer glycan with about 15 repeating units, that is terminated by 2-Omethylrhamnose (3), revealed, among other proteins, an ABC transporter and a protein with a predicted methyltransferase    (3) and preliminary sequencing data indicate the presence of a similarly organized slg gene cluster, it is conceivable to assume that a comparable S-layer glycan biosynthesis pathway occurs in that strain. This would involve the successive addition of sugar residues from nucleotide-activated precursors to the nonreducing end of the nascent saccharide chain by specific glycosyltransferases, utilization of an ATP-binding cassette dependent export system, and transfer of the chain to the O-glycosylation site target sequences on the S-layer polypeptide by ligase. In analogy to the ABC transporter-dependent LPS O-antigen biosynthesis route of Gram-negative bacteria (10, 11), nonreduc-ing terminal modifications seem to play a pivotal role in S-layer glycan chain termination (12). In different LPS O-antigens of Gram-negative bacteria besides the common nonreducing terminal O-methyl groups (30 -32), unusual terminal residues such as 3-deoxy-D-manno-oct-2-ulosonic acid (33), or 2,3,4-triamino-2,3,4-trideoxy-␣-galacturonamide (34) have been described. Hence, the finding that in G. tepidamans GS5-97 T MurNAc and GlcNAc residues represented so far unknown capping elements of the S-layer glycan chain was quite unexpected. MurNAc and GlcNAc are the major glycan constituents of all bacterial peptidoglycans and form [34)-␤-D-MurNAc-(134)-␤-D-GlcNAc- (13] n oligosaccharide chains of variable length (13). Most interestingly, MurNAc is known to occur not only in peptidoglycan but has also been found as a rare constituent of repeating units in different LPS O-antigens and capsular polysaccharides (21,(35)(36)(37). In the S-layer glycan of G. tepidamans GS5-97 T , however, MurNAc and GlcNAc are exclusively present at the nonreducing end of the glycan chain and therefore seem to be involved in chain length determination. This is supported by the absence of a putative methyltransferase gene in the slg gene cluster of that organism. 2 Different from peptidoglycan, the MurNAc residue in the S-layer glycan is ␣-linked. This change of anomeric configuration may be explained by the action of a different glycosyltransferase, which facilitates the transfer of this predicted capping MurNAc residue to the terminal repeating unit. Similar observations have been made in LPS O-antigens (38,39).
Although not yet experimentally demonstrated, the cellular pool of the nucleotide-activated precursors UDP-MurNAc and UDP-GlcNAc can obviously be accessed by the organism for consecutive addition of MurNAc and Glc-NAc to the nonreducing end of the S-layer glycan chains. The mechanism of the S-layer glycosylation pathway is not known yet, but the parallels of this route and ABC transporter-dependent O-antigen biosynthesis are striking, which suggests

FIG. 4. ESI Q-TOF MS spectrum of glycopeptides CF II (top) and CF IV (bottom).
The peaks representing the ϩ4 charge state of the major forms containing 18 -22 (Rha-Fuc) repeats are shown. At this charge state, the mass difference corresponding to a (Rha-Fuc) disaccharide unit is 73.1 Da. Adducts containing 2 hydrogens, 1 sodium, and 1 potassium atoms are labeled (see also Table V). In CF II the Ser/Ala/Asp peptide backbone (m/z 1819.9 Da for n ϭ 20) predicted by NMR was confirmed; in CF IV a Thr/Ala peptide (m/z 1794.5 Da for n ϭ 20, designated TA) was found besides the major Thr/Gln peptide backbone (m/z 1808.9 Da for n ϭ 20, designated TQ).

TABLE V Detailed analysis of ESI Q-TOF MS results for the n ϭ 20 glycoform
Theoretical m/z values (m/z cal ) of the glycopeptides CF II and CF IVa,b containing 20 (Rha-Fuc) disaccharide repeating units with regard to the different charge specifying adducts were calculated based on the compositional and available NMR data using average masses. Only the data for the ϩ4 charge state is shown in detail. The signal derived from the molecule with two hydrogens (H), one sodium (Na), and one potassium (K) adducts was giving the most abundant signal. ND, not determined.