N-Acetylglucosaminylation of Serine-Aspartate Repeat Proteins Promotes Staphylococcus aureus Bloodstream Infection*

Background: Staphylococcus aureus agglutinates in plasma in a manner that requires host fibrinogen and clumping factor A, a bacterial surface protein with serine-aspartate (SD) repeats. Results: SdgB modifies serine residues in SD repeats with GlcNAc, and this glycosylation contributes to the pathogenesis of sepsis. Conclusion: Glycosylation of SD repeats aids bacterial escape from host defenses. Significance: Interference with glycosylation may alter staphylococcal infections. Staphylococcus aureus secretes products that convert host fibrinogen to fibrin and promote its agglutination with fibrin fibrils, thereby shielding bacteria from immune defenses. The agglutination reaction involves ClfA (clumping factor A), a surface protein with serine-aspartate (SD) repeats that captures fibrin fibrils and fibrinogen. Pathogenic staphylococci express several different SD proteins that are modified by two glycosyltransferases, SdgA and SdgB. Here, we characterized three genes of S. aureus, aggA, aggB (sdgA), and aggC (sdgB), and show that aggA and aggC contribute to staphylococcal agglutination with fibrin fibrils in human plasma. We demonstrate that aggB (sdgA) and aggC (sdgB) are involved in GlcNAc modification of the ClfA SD repeats. However, only sdgB is essential for GlcNAc modification, and an sdgB mutant is defective in the pathogenesis of sepsis in mice. Thus, GlcNAc modification of proteins promotes S. aureus replication in the bloodstream of mammalian hosts.

Staphylococcus aureus is the most frequent cause of human bloodstream infection, sepsis, and endocarditis in the United States (1,2). This Gram-positive bacterium evolved to clot animal and human plasma, bind fibrinogen, and agglutinate within a meshwork of fibrin fibrils (3). S. aureus agglutination requires two secreted products, Coa (coagulase) and vWbp (von Willebrand factor-binding protein) (4), which each associate with and activate prothrombin to convert fibrinogen to fibrin (5,6).
ClfA (clumping factor A), a sortase-anchored surface protein (7), tethers bacteria to fibrin fibrils by binding to the C-terminal end of the ␥-chain, thereby blocking incorporation of additional fibrin subunits into agglutinated fibrils (4,8,9). Agglutination with fibrin fibrils protects staphylococci from phagocytes and promotes the formation of infectious thrombi that contribute to the lethal outcome of staphylococcal sepsis in mice (4,10). Staphylococcal agglutination is also essential for the pathogenesis of infectious endocarditis and the formation of purulent abscess lesions, which promote bacterial persistence and dissemination in host tissues (11)(12)(13).
ClfA is the founding member of the family of serine-aspartate (SD) repeat proteins that are synthesized and secreted as precursors with N-terminal signal peptides (14,15). Proteins with SD repeats are typically composed of three domains. The N-terminal A region provides for association with specific ligands, predominantly the ␣-, ␤-, or ␥-chain of fibrinogen (16 -18). For S. aureus SdrC, SdrD, and SdrE, but not for ClfA and ClfB, the A domain is followed by B domain repeats with additional ligand binding properties (19). SD repeats tether the N-terminal ligand-binding domains to the C-terminal LPXTG sorting signal, which triggers sortase A-mediated linkage to the cell wall envelope (20 -22). SD repeat proteins are also found in other pathogenic staphylococci, for example the opportunistic pathogen Staphylococcus epidermidis (23) and Staphylococcus saprophyticus, a frequent cause of urinary tract infection (24).
Hazenbos et al. (25) isolated a human monoclonal antibody that recognizes proteins with SD repeats from S. aureus and S. epidermidis in a manner requiring post-translational modification by the sdgB-encoded UDP-GlcNAc glycosyltransferase.
In vitro experiments with purified glycosyltransferases confirmed that proteins with SD repeats are substrates for SdgBand SdgA-mediated GlcNAc modification (25). Hazenbos et al. further proposed that glycosylation prevents proteolytic degradation of such proteins by human, but not mouse, neutrophils or cathepsin.
Here, we searched for factors that contribute to S. aureus agglutination with fibrin fibrils. We report that mutant alleles of aggA and aggC display reduced agglutination in human plasma, and aggC, but not aggA, affects the post-translational modification of ClfA, but not the staphylococcal coagulases Coa and vWbp (12). aggA and aggC flank the sdrCDE locus of S. aureus (see Fig. 1A), which prompted further examination of transposon insertions in neighboring genes (nwmn_0521 through nwmn_0528) and identified aggB as a second factor affecting ClfA modification. aggB and aggC encode the previously reported SdgA and SdgB proteins. We demonstrate that the aggC (sdgB) gene is required for GlcNAc modification of SD repeats of staphylococcal ClfA. Furthermore, S. aureus aggC (sdgB) variants, but not aggB (sdgA) variants, are defective in the pathogenesis of bloodstream infections in mice.
Plasmid pClfA SD5 provided for expression of a truncated clfA gene with an insertion of the coding sequences for the tobacco etch virus (TEV) protease cleavage site, Strep-tag, five tandem SD repeats, a hexahistidyl tag, and a stop codon at nucleotide 1677, i.e. following codon 559 of S. aureus Newman clfA. The clfA SD5 gene was generated via DNA synthesis (GeneArt GmbH) with the following nucleotide sequence at position 1688 of clfA: 5Ј-GAAAACCTGTATTTTCAGGGCTGG-AGCCACCCGCAGTTCGAAAAAGATTCAGATTCAGAT-CCTGGTTCAGATTCAGGTTCAGATTCAAATTCAGAT-TCAGGTTCACACCATCATCACCATCACTAGGGATCC (with the stop codon and BamHI site underlined). clfA SD5 was cloned into pWWW412 to generate pClfA SD5 . Chemicals were purchased from Sigma unless indicated otherwise.
Agglutination-Overnight cultures of S. aureus were washed with 1 ml of PBS and suspended to a final concentration of A 600 ϭ 4.0. Bacteria from 1 ml of culture were incubated with 1:500 SYTO 9 (Invitrogen) for 15 min, washed twice with 1 ml of PBS, and suspended in 1 ml of PBS. Bacteria were mixed 1:1 with lepirudin-treated human plasma on glass microscope slides and incubated for 15 min. Samples were viewed and images were captured on an IX81 live cell total internal reflection fluorescence microscope using a 20ϫ objective. For quantification of agglutination, 20 bright-field microscope images were obtained from each strain using a Nikon TE2000-U microscope with 20ϫ objective. The threshold function in ImageJ software was used to convert the image into a dichromatic format in which staphylococci are black and the background is white. The areas of agglutination complexes in each image were measured. The average area of S. aureus Newman in saline without plasma was subtracted from all data sets. To statistically compare the wild type and mutant, data were analyzed by two-way analysis of variance (ANOVA) using Prism (GraphPad Software); p values Ͻ0.05 were deemed significant.
Immunoblotting-S. aureus Newman and its mutants (aggA, aggB, or aggC) with or without complementing plasmids were grown overnight in TSB. For plasmid stability, chloramphenicol (10 g⅐ml Ϫ1 ) was added as needed. Overnight cultures were diluted 1:100 into fresh TSB and grown to A 600 ϭ 1.5. One ml of culture was treated with TCA (14%, 0°C, 30 min) and centrifuged (20,000 ϫ g, 15 min), and the pellet was washed with acetone (1 ml, 0°C, 15 min). The sediment was suspended in 1 ml of Buffer A (50 mM Tris-HCl (pH 7.5) and 150 mM NaCl) with 20 g of lysostaphin and incubated (37°C, 1.5 h). Solubilized proteins were precipitated with TCA, washed with acetone (as described above), and dried. Proteins were solubilized in 0.5 M Tris-HCl (pH 8.0) and 4% SDS, mixed and heated with sample buffer, and subjected to 8% SDS-PAGE with electrotransfer to PVDF membranes. Membranes were blocked by incubation in bovine serum albumin (3%, 1 h, room temperature). To prevent binding of primary antibodies to protein A, 0.8 mg of human IgG was added to 10 ml of blocking buffer. After incubation (1 h, room temperature), immune serum (primary antibody at 1:5000 dilution) was added, and the membranes were incubated for an additional hour. Membranes were washed three times for 10 min with TBS-T (50 mM Tris (pH 7.5), 150 mM NaCl, 0.1% and Tween 20) and incubated with Alexa Fluor 680-labeled goat anti-rabbit secondary antibody (1:20,000; LI-COR) for 1 h. Membranes were washed again with TBS-T, and fluorescence at 700 nm was measured with a LI-COR Odyssey infrared scanner.
Purification of ClfA SD5 -S. aureus Newman (wild-type) or its aggB (sdgA) and aggC (sdgB) mutants harboring plasmid pClfA SD5 were grown overnight in TSB with 10 g⅐ml Ϫ1 chloramphenicol. Cultures were diluted 1:100 into 2 liters of TSB and grown to A 600 ϭ 1.5. Cultures were centrifuged (8000 ϫ g, 15 min). Culture supernatants were cooled on ice, and ammonium sulfate was added to (w/v) 90%, followed by incubation for 1 h. The precipitate was sedimented by centrifugation (8000 ϫ g, 15 min), suspended in 10 ml of Buffer A, and dialyzed overnight at 4°C against the same buffer. Dialyzed material was centrifuged, and the supernatants were loaded by gravity flow onto a Strep-Tactin column (IBA). The column was washed with Buffer A and eluted with 5 mM desthiobiotin (IBA), and the eluate containing ClfA SD5 proteins was desalted via HiTrap chromatography (GE Healthcare). To cleave ClfA SD5 , 50 g of purified protein was treated with 5 g of TEV protease (Sigma) for 1 h at 30°C. Samples were analyzed by SDS-PAGE and immunoblotting or subjected to mass spectrometry.
Combined Liquid Chromatography-Electrospray Ionization Mass Spectrometry with Fraction Collection (LC-ESI-MS ϩ )-Sample aliquots containing ClfA SD5 proteins cleaved or not with TEV protease were dried in a vacuum concentrator, dissolved in water (250 l), and loaded onto a polymeric reverse phase column (Agilent PLRP-S, 300-Å pore size, 5-m particle size, 2.1 ϫ 150 mm) equilibrated in solvent B (100:0.1 (v/v) water/formic acid) and eluted (100 l/min) with a linearly increasing concentration of solvent C (100:0.1 (v/v) acetonitrile/formic acid, 0 min/5%, 5 min/5%, 45 min/90%, 50 min/5%, and 60 min/5%). The column effluent was passed through a stream splitter; a proportion (ϳ50%) was directed to a fraction collector (1 min/fraction), and the remainder was directed to an Ionspray TM source connected to a triple quadrupole mass spectrometer (ABI Sciex API III ϩ ) scanning in the positive ion MS mode from m/z 500 -1850 (step size of 0.3 Da, dwell of 1 ms, 4.82 s/scan, orifice of 90 V). Data were interrogated using Mac-Spec TM (version 3.3) for molecular mass calculations from multiply charged ion clusters and BioMultiView© (version 1.3.1) for display of the deconvoluted spectra.
MALDI-MS-Select fractions from the LC-ESI-MS ϩ experiments were pooled, dried in a vacuum concentrator, and dissolved in water/acetonitrile/formic acid (50:50:0.1), and aliquots (2 l) were spotted with either ␣-cyanocinnamic or sinapinic acid matrices (both 5 mg/ml in 40:60:0.1 water/acetonitrile/trifluoroacetic acid) on a stainless steel target, and laser desorption spectra were recorded with a Bruker-Daltonics ultrafleXtreme tandem time-of-flight mass spectrometer (in the MS mode, 10,000 laser shots/spectrum; in the collisionally activated dissociation MS/MS mode, 20,000 laser shots/spectrum with a 12-Da window for precursor ion selection). Data were interrogated with mMass software (version 5.5.0). To compensate for m/z drift during the MS/MS data acquisition process, paired m/z and intensity values from individual spectra were loaded into Microsoft Excel© (version 14.0.7106.5003), and the m/z values were globally shifted (0 -10 Da) to align a preselected ion in each spectrum. The shifted spectra were then reloaded into mMass for further interpretation and presentation. Fragmentation predictions from theoretical sequences were obtained using the MS-Product tool within ProteinProspector (version 5.10.13).
Mouse Bloodstream Infection-Overnight cultures of S. aureus Newman (wild-type) and clfA, aggB (sdgA), and aggC (sdgB) mutants were diluted 1:100 into fresh TSB and grown for 2 h at 37°C. Cultures were centrifuged (4000 ϫ g, 10 min), and the pellets were washed and resuspended in PBS to the desired bacterial concentration. Inocula were quantified by spreading sample aliquots on tryptic soy agar and enumerating cfu. BALB/c mice (groups of 30, n ϭ 30) were anesthetized via intraperitoneal injection with 65 mg⅐ml Ϫ1 ketamine and 6 mg⅐ml Ϫ1 xylazine per kg of body weight. Mice were infected by injection with 1-2 ϫ 10 8 cfu of S. aureus Newman or mutants into the periorbital venous sinus of the right eye. At 12 h following infection, cohorts of mice (n ϭ 10) were killed by CO 2 inhalation and necropsied. Hearts were removed, and staphylococcal loads were analyzed by homogenizing tissue in PBS containing 0.1% Triton X-100. Serial dilutions of homogenates were spread on tryptic soy agar and incubated for enumeration of cfu. Additional cohorts of animals (n ϭ 10) were killed, and hearts were examined by histopathology. Briefly, hearts were fixed in 10% formalin for 24 h at room temperature, embedded in paraffin, thin-sectioned, stained with hematoxylin/eosin, and inspected by light microscopy to enumerate pathological lesions. The remaining animals (n ϭ 10) were monitored for survival over 10 days. All mouse experiments were performed at least twice and conducted in accordance with the institutional guidelines following experimental protocol review and approval by the Institutional Biosafety Committee and the Institutional Animal Care and Use Committee at The University of Chicago. The statistical significance of the mouse sepsis model was analyzed by two-tailed log-rank analysis. Bacterial loads in the experimental animal infection model were analyzed by one-way ANOVA to measure statistical significance. Two-way ANOVA was used to analyze the statistical significance of histopathological analysis in the experimental animal infection model. All data were analyzed using Prism. p values Ͻ0.05 were deemed significant.

S. aureus agg Mutants Defective in Agglutination-
The wild type and insertional mutants were screened with human plasma for factors that contribute to defects in staphylococcal agglutination (4,29). This search identified several mutants with bursa aurealis insertions in two genes, designated aggA and aggC (Fig. 1A). Unlike clfA or coa/vwb mutants, which are completely defective for agglutination (4), variants with lesions in aggA or aggC displayed reduced staphylococcal agglutination (Fig. 1, B and C). The aggA and aggC genes flank the sdr locus, which is composed of tandem repeats of the genes encoding SD repeat proteins SdrC, SdrD, and SdrE (14). Insertional lesions in these neighboring genes, including sdr genes, did not display a significant defect in agglutination compared with the wild type; however, the aggB variant was examined further because of the homology between the aggB and aggC gene products. Defects associated with clfA, aggA, and aggC were complemented by expressing the wild-type genes on a plasmid (Fig. 1, B and C). Recently, the products of the aggB and aggC genes were characterized as UDP-GlcNAc transferases that modify SD repeat proteins and have been designated SdgA and SdgB, respectively (25). SdgA and SdgB are 490 and 496 amino acids long and share 44% identity (E-value of 3eϪ144) with a conserved GtfAlike domain, named after Streptococcus gordonii GtfA, a protein that catalyzes the O-linked glycosylation of the cell surface protein GspB (33).
S. aureus aggB (sdgA) and aggC (sdgB) Mutants Affect Mobility of ClfA and ClfB on SDS-PAGE-To test whether agg mutants affect the production or secretion of known agglutination factors, staphylococcal cultures treated with the murein hydrolase lysostaphin were analyzed by immunoblotting with Coa-, vWbp-, or ClfA-specific antibodies. These experiments revealed that aggA, aggB (sgdA), and aggC (sdgB) mutants produced similar amounts of Coa and vWbp (Fig. 2). Immunoblotting for the sortase A enzyme (SrtA) was used as a loading control (Fig. 2). Although the aggB (sgdA) and aggC (sdgB) mutants harbored similar amounts of ClfA compared with wild-type and aggA mutant staphylococci, their immunoreactive material migrated with increased mobility on SDS-PAGE (Fig. 2). Of note, aggC (sdgB)-mutant ClfA appeared as the fastest migrating species. sdgA-mutant ClfA assumed intermediate mobility, and the mobility of aggA-mutant ClfA was unaffected (Fig. 2). All of these species were specific for ClfA, as the clfA mutant extract did not react with the ClfA-specific antibodies (Fig. 2). Plasmids providing for the expression of wild-type aggB (sgdA) or aggC (sdgB) restored the SDS-PAGE mobility of ClfA in the corresponding mutants to the same level as ClfA in wildtype staphylococci. We wondered whether the SDS-PAGE mobility of other proteins with SD repeats is also affected in the agg mutants. Immunoblotting of staphylococcal cell wall prep-arations with ClfB-specific antibodies revealed altered mobility of ClfB in aggB (sgdA) or aggC (sdgB) mutant strains (Fig. 2). SdrC, SdrD, and SdrE also displayed differential mobility in these mutants (data not shown). Hereafter, we refer to aggB and aggC as sgdA and sdgB (25).
Purification and Characterization of ClfA SD5 -To analyze the chemical structure of ClfA, we engineered a plasmid-expressed variant, ClfA SD5 , with insertion of a TEV protease cleavage site, Strep-tag, five tandem SD repeats, and a C-terminal hexahistidyl tag immediately downstream of the fibrinogenbinding A domain (Fig. 3A). When transformed into wild-type, sdgA mutant, and sdgB mutant staphylococci, pClfA SD5 provided for the expression of ClfA SD5 , which was partially purified from culture supernatants by ammonium sulfate precipitation and affinity chromatography. When subjected to SDS-PAGE and immunoblotting, ClfA SD5 from wild-type, sdgA mutant, and sdgB mutant staphylococci (ClfA SD5/WT , ClfA SD5/sdgA , and ClfA SD5/sdgB , respectively) migrated with similar mobility and reacted with antibodies specific for the A domain of ClfA (Fig.  3B, ␣ClfA), hexahistidyl (HIS probe), and Strep-tag (␣Strep). These data indicate that Strep-Tactin affinity chromatography of ClfA SD5 yielded a partially purified, full-length polypeptide.
These ClfA SD5 samples were subjected to LC-ESI-MS ϩ . That from wild-type staphylococci (ClfA SD5/WT ) generated a series of multiply charged ions that, upon deconvolution, revealed six distinct molecular species with measured masses of 59,857.86, 60,060.48, 60,264.44, 60,465.63, 60,671.95, and 60,875.89 Da (Fig. 3C). The average difference between these species is 203.6 Da, which corresponds to the residue mass of GlcNAc (203.193 Da calculated as the average (chemical) mass for C 8 H 13 NO 5 ). The sample from sdgB mutant staphylococci (ClfA SD5/sdgB ) generated multiply charged ion signals that were deconvoluted into a single species with a measured mass of 59,859.90 Da. This is similar to that of the lightest ClfA SD5/WT species, and both are in reasonable agreement with the predicted mass of unmodified ClfA SD5 protein with the 39-amino acid signal sequence   FEBRUARY 7, 2014 • VOLUME 289 • NUMBER 6 removed following secretion (calculated 59,848.04 Da) (Fig.  3C). The 0.0167% difference between the measured and calculated masses could be due to a single oxidation of one of the four methionine residues in the sequence. Accounting for a single oxidation reduces the difference to 0.009%, which is consistent with the accuracy of molecular mass measurements made by ESI on quadrupole mass spectrometers. The sample from sdgA mutant staphylococci (ClfA SD5/sdgA ) generated a series of multiply charged ions that were deconvoluted as three distinct species with measured masses of 59,873.89, 60,072.94, and 60,279.41 Da (Fig. 3C). The average mass difference between these (202.76 Da) is also similar to the chemical mass of GlcNAc (203.193 Da). These data suggest that ClfA SD5/WT is modified with at least five GlcNAc moieties, whereas ClfA SD5/sdgB is not modified, and ClfA SD5/sdgA displays an intermediate level of glycosylation.

Glycosylation of Staphylococcal SD Repeat Proteins
Structural Analysis of TEV Fragments of ClfA SD5 from Wildtype, sdgA (aggB) Mutant, and sdgB (aggC) Mutant Staphylococci-ClfA SD5/WT , ClfA SD5/sdgA , and ClfA SD5/sdgB were incubated with TEV protease and analyzed by Coomassie Blue-stained SDS-PAGE and immunoblotting (Fig. 4A). TEV protease treatment increased the mobility of ClfA SD5 , ClfA SD5/sdgA , and ClfA SD5/sdgB . The faster migrating species reacted with anti-ClfA antibody, but not with anti-Strep-tag or anti-hexahistidyl antibody, indicating that the protease had cleaved off the C-terminal SD5 repeat of ClfA SD5 (the small size of the C-terminal fragment precluded visualization on SDS-PAGE). The periodic acid-Schiff reaction failed to generate oxidized products with ClfA SD5/WT (data not shown). Nevertheless, succinylated wheat germ agglutinin, a GlcNAc-specific lectin (34), readily bound ClfA SD5/WT prior to TEV protease cleavage, but not after (Fig. 4A, fifth panel). These data suggest that ClfA SD5 in wildtype S. aureus is modified with GlcNAc in the C-terminal SD5 repeat in a manner that does not lead to periodate formation,  likely because the hemiacetal of the carbohydrate moiety is tethered to the polypeptide chain.
When analyzed by LC-ESI-MS ϩ , TEV protease-treated Clf-A SD5 samples from wild-type, sdgA mutant, or sdgB mutant staphylococci generated multiply charged ions that, upon deconvolution, yielded masses of 56,083.91, 56,091.42, and 56,083.73 Da, respectively (data not shown). These observations are in agreement with TEV protease cleavage at ClfA SD5 residue 565. These N-terminal fragments encompass 526 amino acids and lack both the signal sequence (first 39 amino acids) and the C-terminal SD5 peptide, and they are not modified with GlcNAc.
LC-ESI-MS ϩ of the C-terminal SD5 wild-type sample following TEV protease cleavage yielded another series of ions that, upon deconvolution, showed six different compounds with measured masses of 3794.59, 3997. 39, 4200.49, 4404.26, 4607.50, and 4810.99 Da (Fig. 4B). The average mass difference between these species is 203.28 Da. LC-ESI-MS ϩ of the TEV protease-treated ClfA SD5/sdgA sample yielded a series of ions that, upon deconvolution, showed peptide species with masses of 3794.96 and 3999.46 Da (difference of 205.5 Da) (Fig. 4C). Finally, LC-ESI-MS ϩ of the TEV protease-treated ClfA SD5/sdgB sample gave rise to a single compound with an average mass of 3794.80 Da (Fig. 4D). Together, these data indicate that in wildtype S. aureus, the SD-containing region of ClfA SD5 is modified by GlcNAc in a manner requiring the glycosyltransferase encoded by sdgB and, to a lesser degree, the one encoded by sdgA.
SD Repeats of ClfA Are Modified with GlcNAc-To identify the amino acid residues that are modified with GlcNAc, the C-terminal SD peptides of ClfA SD5/WT and ClfA SD5/sdgB were analyzed off-line by MS/MS of the fractions collected during LC-ESI-MS ϩ . Collisionally activated dissociation MS/MS experiments during ESI were unsuccessful. However, collisionally activated dissociation MS/MS of the corresponding singly charged precursor ions produced during MALDI (MALDI-MS/ MS) yielded spectra that were interpreted for their y-and b-fragment ions. The non-glycosylated peptide (parent mass of 3793.7 Da) gave a virtually complete y-ion series, partially presented in Table 1 (ions y2-y31; ions y20, y24, and y29 were missing), but with a much less complete b-ion ladder (data not shown). This y-ion ladder confirms the predicted amino acid sequence of the unmodified C-terminal ClfA SD5 fragment ( Fig.  5A and Table 1). However, the MALDI-MS/MS spectra of the glycosylated isoforms were less populated with readily assignable fragment ions, and there were no clearly discernable y-or b-ion ladders, precluding unambiguous assignment of the sites of glycosylation. Furthermore, possible heterogeneity of the glycosylation pattern within each isoform would diminish signal intensity for any one fragment ion and would further complicate interpretation of the data. Overlaying the MALDI-MS/MS spectra of the mono-, di-, tri-, tetra-, penta-, and hexaglycosylated isoforms revealed an m/z drift of fragment ions, adding yet another complication to spectral interpretation. This interspectral drift is an unavoidable feature of the instrument with which the data were collected. Nonetheless, all spectra showed a strong signal at m/z 1072, assigned as y9 (Fig.  5A and Table 1). This signal was then used to align the spectra (Fig. 5B). Comparison of the aligned spectra in a stacked overlay format (Gel View in mMass software) showed diagonal ladders separated by 203 Da originating from y13 (single-step ladder), y17 (three-step ladder), y21 (five-step ladder), y23 (five-step ladder), and y25 (six-step ladder) (Fig. 5B). The regular step distances of 203 Da strongly suggest the presence of monoglycosylation at each site because a preponderance of double glycosylation would have resulted in ladders with 406-Da step sizes. Furthermore, ion y13 with sequence S 25 D and not y9 with sequence S 27 GS 29 was monoglycosylated ( Fig. 5A and Table 1). Additional glycosylation (ϩ203 Da) was observed in ions y17, y21, y23, and y25, encompassing sequences S 21 D, S 17 D, S 13 D, and S 9 D, respectively. The data also suggest the possibility of some diglycosylation at a single SD site. For example, the ions assigned as y17 and y21 in the hexa-GlcNac isoform showed additional rungs that are consistent with diglycosylation at   Fig. 5A. b The data were obtained from spectra aligned as described under "Experimental Procedures" and shown in Fig. 5B. Assignments for the heaviest glycosylated y-ions are shown. c Relative error (parts per thousand (ppt)) indicates the difference between the observed m/z value for each ion and the calculated value for that ion.
repeats S 17 D and S 21 D (Fig. 5B). In support of this assignment is the fact that y17 exhibited a maximum of three glycosylations C-terminal to repeat S 17 D, whereas the y21 ion contained only one additional SD repeat but harbored two additional glycosylations (Table 1). In addition, close inspection of the molecular ion region showed losses of 203 Da and weak but discernible signals with losses of 406 Da. Although it is impossible to definitively assign the origin of these signals, the 406-Da loss could arise from loss of a di-GlcNAc moiety. In conclusion, these experiments suggest that serine residues within SD repeats of ClfA are modified with O-linked GlcNAc in a manner requiring the SdgB glycosyltransferase. The majority of serine residues appear to be monoglycosylated, although some appear to be diglycosylated, and the interpretation cannot exclude the possibility of even higher order glycosylations at single residues. Because, ClfA SD5 is partially glycosylated in the sdgA mutant, it is possible that SdgA may favor additional glycosylation at sites already glycosylated by SdgB. Glycosylation of SD Repeat Proteins and Pathogenesis of Staphylococcal Infections-We sought to determine whether GlcNAc modification of serine residues in SD repeat proteins contributes to the pathogenesis of S. aureus infections. Equal doses of wild-type Newman or clfA, sdgA, and sdgB mutant S. aureus were injected into the bloodstream of BALB/c mice, and survival was monitored over a period of 10 days. Most animals infected with wild-type S. aureus succumbed to infection within 48 h (Fig. 6A). Mice infected with the clfA mutant displayed a significant delay in time to death as well as survival in some of the infected animals (Fig. 6A). Although all animals infected with the sdgB mutant succumbed to infection, these mice displayed a reproducible delay in time to death (Fig. 6A).
We did not detect a significant delay in time to death between animals that had been infected with wild-type S. aureus or the sdgA variant strain (Fig. 6A).
Heart tissues were examined for bacterial load and the histopathology of staphylococcal lesions. As reported earlier (4), the infectious load of the clfA mutant in heart tissue was reduced by 1.98 log 10 cfu (Fig. 6B). Histopathology revealed that, compared with wild-type S. aureus, the clfA mutant formed fewer lesions and did not agglutinate in the vasculature, which is characterized by intravascular thrombi of staphylococci and fibrin without immune cell infiltrates (Fig. 6C). The sdgB mutant displayed an intermediate phenotype with reduced bacterial load (⌬0.67 log 10 cfu⅐g Ϫ1 ) (Fig. 6B) and reduced frequency of infectious lesions in heart tissue (Fig. 6C). Staphylococcal agglutinations without immune cell infiltrate were rarely observed for the sdgB variant. Moreover, mixed histopathological features, including agglutinated bacteria and moderate immune cell infiltrates (staphylococcal agglutinations ϩ polymorphonuclear neutrophils (PMNs)) as well as massive neutrophil inflammation (PMNs), represented the most abundant infectious lesions of sdgB mutant S. aureus (Fig. 6, C and D).

DISCUSSION
Glycosylation of secreted proteins in bacteria has only recently been appreciated (35,36). In eukaryotic cells, protein oligosaccharyltransferase and differential processing of its N-linked carbohydrate side chains control the folding, traffick- mono-, di-, tri-, tetra-, penta-, and hexaglycosylated precursor ion isoforms, respectively. The spectra have been aligned as described under "Experimental Procedures." Important y-ion assignments with the number of glycosyl groups shown in superscript (G1-G6) are indicated. A more complete tabulation of the observed ions is presented in Table 1. . SdgB (AggC) glycosylation of SD repeats contributes to S. aureus sepsis. A, survival of BALB/c mice (n ϭ 10) following intravenous injection of S. aureus Newman or variants lacking clfA, sdgA (aggB), or sdgB (aggC). Statistical significance was measured by the two-tailed log-rank test. B, staphylococcal load (enumerated as cfu) in heart tissues of BALB/c mice (n ϭ 10) at 12 h post-infection with intravenous injection of S. aureus Newman or its variants (clfA, sdgA, and sdgB). Horizontal red lines represent mean cfu. Statistical significance was assessed by one-way ANOVA. *, p Ͻ 0.05. C, determination of histopathological features in thin-sectioned and hematoxylin/eosin-stained heart tissues from BALB/c mice (n ϭ 10) 12 h after intravenous injection of S. aureus Newman or its variants (clfA, sdgA, and sdgB). Error bars represent S.E. Statistical significance was determined by two-way ANOVA. **, p Ͻ 0.0001. D, representative images of lesions in heart tissues shown as staphylococcal agglutination without (SA) or with (SA ϩ PMNs) PMNs, PMN infiltration without staphylococcal agglutination (PMNs) following infection with Newman, and disruption of staphylococcal agglutination following infection with the sdgB (aggC) mutant (SA sdgB (aggC)). All data are representative of two independent experiments.
ing, and stability of proteins in the secretory pathway (37). In contrast, O-linked carbohydrate modifications, e.g. O-linked GlcNAc at serine or threonine residues, occur predominantly on intracellular proteins and impact the biological activities of transcription factors, histones, and enzymes in eukaryotic cells (38). N-Linked glycosylation, the decoration of asparagine residues on proteins with lipid-linked oligosaccharide substrates, has been reported for some Gram-negative bacteria, but not for Gram-positive microbes (36). On the other hand, O-linked glycosylation occurs in Gram-negative and Gram-positive bacteria (36). In Gram-negative microbes, oligosaccharyltransferase appears to utilize undecaprenyl-linked carbohydrates that are transferred onto secreted proteins in the bacterial periplasm (36). In Gram-positive bacteria, O-linked glycosylation involves cytoplasmic glycosyltransferases utilizing nucleotide-linked monosaccharides for sequential decoration of serine or threonine residues (35). For several Gram-positive bacteria, O-linked glycosylation targets polypeptides that require their carbohydrate modification as a prerequisite for their entry into dedicated secretion pathways, which may culminate in assembly of adhesive molecules on microbial surfaces or in murein hydrolase function (39,40).
Recently, S. aureus proteins with SD repeats were reported to be glycosylated in a manner requiring two glycosyltransferases, SdgA and SdgB (25). In vitro glycosylation experiments revealed that SdgA and SgdB utilize UDP-GlcNAc for modification. Mass spectra analyses suggested that N-acetylglucosaminylation increased with the number of SDS repeats present in the protein substrate, but modified residues were not identified (25). Furthermore, N-acetylglucosaminylation by SdgA in vitro required prior GlcNAc modification with SdgB, suggesting that target residues may be sequentially modified with GlcNAc-GlcNAc moieties (25). Nevertheless, the chemical nature, site(s) of modification, and biological consequence of glycosylation for staphylococcal pathogenesis were not determined (25). In agreement with the work by Hazenbos et al. (25), we observed that S. aureus proteins with SD repeats are decorated with GlcNAc moieties in a manner requiring SdgA and SdgB glycosyltransferases. Using an engineered variant (Clf-A SD5 ) purified from S. aureus and analyzed by MS, we have shown here that the SD repeat appears to be modified at serine residues that are immediately adjacent to aspartyl. Whereas ClfA SD5 from wild-type S. aureus harbored GlcNAc modifications present as mono-or possibly disaccharides on seryl residues, these modifications were absent when ClfA SD5 was isolated from sdgB mutant staphylococci. ClfA SD5 from sdgA mutant staphylococci displayed an intermediate level of glycosylation.
Earlier work had implicated SD repeat proteins as important determinants of pathogenesis, and they coat S. aureus with fibrinogen and fibrin and provide for bacterial escape from innate host defenses (41,42). The important contributions of ClfA are readily apparent in a mouse model for S. aureus bloodstream infection. Compared with wild-type staphylococci, clfA mutants displayed a delayed time to death and increased survival. The sdgB mutant also displayed a delay in time to death during bloodstream infection in mice, which is in agreement with its contribution to S. aureus agglutination in human plasma and in mouse heart tissues. Of note, the sdgA mutant did not display a significant defect in the mouse bloodstream model, indicating that SdgA-mediated glycosylation of SD proteins is not a contributor to disease in mice. Interestingly, AggA, a predicted NADPH-dependent FMN reductase, contributed to S. aureus agglutination in human plasma but was dispensable during staphylococcal bloodstream infection in mice (data not shown). We presume that GlcNAc modification of ClfA is responsible for the observed SdgB-mediated agglutination and virulence defect, as S. aureus mutants defective in clfB or sdrCDE expression display wild-type agglutination phenotypes (4). It seems unlikely that the sdgB-mediated agglutination defect is due to the proposed mechanism of increased cathepsin B-mediated proteolysis of ClfA (25). First, mouse cathepsin does not cleave non-glycosylated ClfA, yet sdgB mutant staphylococci display virulence defects in mice (25). Second, neutrophil-mediated release of cathepsin does not occur during staphylococcal agglutination in human plasma (4). Future work will need to reveal whether the agglutination phenotype of non-glycosylated ClfA is caused by altered binding activities for fibrinogen and fibrin or whether they impact another, as of yet unknown host ligand.