Comprehensive Characterization of Methicillin-resistant Staphylococcus aureus subsp. aureus COL Secretome by Two-dimensional Liquid Chromatography and Mass Spectrometry*

Two-dimensional LC combined with whole protein and peptide mass spectrometry is used to characterize proteins secreted by methicillin-resistant Staphylococcus aureus COL. Protein identifications were accomplished via off-line protein fractionation followed by digestion and subsequent peptide analysis by reverse phase LC-ESI-LTQ-FT-MS/MS. Peptide MS/MS analysis identified 127 proteins comprising 59 secreted proteins, seven cell wall-anchored proteins, four lipoproteins, four membrane proteins, and 53 cytoplasmic proteins. The identified secreted proteins included various virulence factors of known functions (cytotoxins, enterotoxins, proteases, lipolytic enzymes, peptidoglycan hydrolases, etc.). Accurate whole protein mass measurement (±1.5 Da) of the secreted proteins combined with peptide analysis enabled identification of signal peptide cleavage sites and various post-translational modifications. In addition, new observations were possible using the present approach. Although signal peptide cleavage is highly specific, signal peptide processing can occur at more than one site. Surprisingly, cleaved signal peptides and their fragments can be observed in the extracellular medium. The prediction accuracies of several signal peptide prediction programs were also evaluated.

Staphylococcus aureus, a Gram-positive human pathogen, is the leading cause of nosocomial infections, imposing tremendous economic burden on patients and hospitals throughout the world (1)(2)(3). The spectrum of staphylococcal infections is very wide, ranging from minor skin lesions to life-threatening conditions such as bacteremia, pneumonia, endocarditis, osteomyelitis, toxic shock syndrome, and septicemia (2)(3)(4). The treatment of staphylococcal infections has become extremely challenging because of its propensity to rapidly evolve antibiotic-resistant strains. The methicillin-resistant staphylococci (MRSA) 1 is the most notorious in that it causes an estimated 94,000 life-threatening infections and 19,000 deaths a year in the United States (4,5). Hospitalization costs associated with MRSA infections are also significant with a mean attributable cost of $35,000 per infection (6). Furthermore, the recent emergence of strains resistant to vancomycin, a glycopeptide antibiotic that is often considered as the last resort drug in treating MRSA infections, has compounded the problem (7)(8)(9). Needless to say, it is of paramount importance to discover effective vaccines and to develop new strategies to treat S. aureus infections. This urgency motivated the scientific community to direct significant research effort toward whole genome sequencing of S. aureus strains in the past few years. The wealth of information available from the nine fully annotated and sequenced genomes of S. aureus has provided us with an excellent opportunity to apply powerful technologies including proteomics to gain a comprehensive understanding of the biology of this organism (10 -13).
Pathogenesis of S. aureus is complex and involves the synthesis of an array of virulence factors followed by their transport across the cytoplasmic membrane to destinations outside the cell. A majority of the exported proteins in S. aureus are predicted to be secreted via secretory (Sec) pathway, which requires an N-terminal signal peptide (14) at the N terminus of the protein and a cleavage site that is recognized by type I signal peptidases (SPases I) (15). During translocation, or shortly thereafter, the signal peptide of the preprotein is removed by SPase I resulting in the release of the mature protein from the membrane (16,17). The mature protein may be further modified and is either retained on the cell surface or secreted into the extracellular host milieu. The secreted proteins of S. aureus are postulated to play a prominent role in host infection and are believed to be engaged in tissue damage, invasion and evasion of host immune responses. There-fore, a comprehensive description of secretory proteins (the secretome) of different S. aureus strains is vital to gain insights into its pathogenesis. This information will be valuable in identifying novel virulence factors and should ultimately help in the development of new diagnostic tools and vaccines.
To this end, S. aureus secretory proteins have been identified using a variety of classical techniques including Western blot, ELISA, and one-and two-dimensional gel electrophoresis (1DE/2DE) with N-terminal sequencing (18 -20). Although gel-based techniques are well established for separating proteins mixtures, they have several drawbacks including poor reproducibility and sensitivity and limited dynamic range, and they are tedious, labor-intensive, and technically challenging (21)(22)(23). Recent proteomics strategies have coupled gel electrophoresis with mass spectrometry (24 -27). In these approaches, the peptides resulting from in-gel digestion of excised spots were analyzed by MALDI-TOF-MS or by LC-ESI-MS/MS. A critical drawback of peptide analysis by mass spectrometry is that it provides very limited molecular information about intact secreted proteins. Valuable information regarding loss of signal peptides, signal peptide cleavage sites, post-translational modifications, and protein degradation is usually completely lost. This is particularly true when peptide mass fingerprinting is used. To overcome the shortcomings of the current techniques, a sophisticated gel-free approach combining whole protein and peptide mass spectrometric approaches was used in the present work. To the best of our knowledge, there is only one report in the literature pertaining to whole protein mass analysis of S. aureus secretory proteins. Kawano et al. (18) attempted to identify S. aureus secreted proteins via one-dimensional reverse phase (RP) LC-ESI-MS and N-terminal Edman degradation. Using this approach, they were able to tentatively identify three and four secreted proteins in NCTC 8325 and MRSA 3543 strains, respectively.
In the present study, the secretome of methicillin-resistant Staphylococcus aureus subsp. aureus COL (S. aureus COL) was more comprehensively characterized. Secreted proteins were separated with an in-house constructed automated twodimensional LC system (28) with on-line fractionation followed by whole protein mass measurement by ESI-Q-TOF-MS. Definitive protein identifications were accomplished via off-line collection of protein fractions followed by protein digestion and subsequent peptide analysis by RPLC-ESI-LTQ-FT-MS/ MS. Genome-based signal peptide algorithms predict 71 secretory proteins for S. aureus COL (29). Our peptide analysis successfully identified 59 of these secreted proteins from the culture supernatants of S. aureus COL with average sequence coverage of 79%. In addition, combined information from the two mass spectrometric approaches allowed detailed characterization of 53 of these secreted proteins. The accurate whole protein mass measurement of the secreted proteins allowed verification of signal peptide cleavage sites. Also, the current study provided us with an opportunity to compare the accuracy of several computational tools available for predicting signal peptide cleavage sites. Additionally, we report surprising findings on the presence of cleaved signal peptides and signal peptide fragments in the extracellular medium.

EXPERIMENTAL PROCEDURES
Materials-HPLC grade ACN and urea were purchased from EMD Chemicals (Gibbstown, NJ). Water was purified using a Barnstead/ Thermolyne E-Pure system (Barnstead/Thermolyne, Dubuque, IA). 40% aqueous methylamine was obtained from Aldrich. Sodium chloride, formic acid, TFA, and TCA were procured from Mallinckrodt Baker. Ammonium bicarbonate was obtained from Mallinckrodt Baker. Proteomics grade trypsin (T-6567) was purchased from Sigma, and endoproteinase Glu-C was obtained from New England Biolabs (Beverly, MA).
Bacterial Strain and Growth Conditions-The methicillin-resistant S. aureus COL was obtained through the Network on Antimicrobial Resistance in S. aureus (NARSA) program supported under NIAID, National Institutes of Health Contract N01-AI-95359. To prepare stock cultures, a single colony of S. aureus COL was inoculated into 50 ml of sterile brain-heart infusion (BHI) broth (Difco/BD Biosciences) and incubated overnight at 37°C with rotary aeration (200 rpm). 1 ml of the overnight culture was used to inoculate 200 ml of fresh sterile BHI broth and incubated for 10 h. The resulting culture was aliquoted into 2-ml stocks with 20% by volume glycerol as the cryoprotective agent and stored at Ϫ80°C until required. In a typical experiment, the stock culture was inoculated into 100 ml of sterile BHI broth and grown overnight at 37°C with rotary aeration (200 rpm). A 2-ml portion of this overnight preculture was then diluted 1:100 in fresh sterile BHI broth and incubated at 37°C with shaking at 200 rpm. The growth was monitored spectrophotometrically by measuring A 600 . Bacteria were cultured to stationary growth phase (A 600 ϭ 4), which was generally attained in about 8 h.
Precipitation and Preparation of Extracellular Protein Fraction-Bacterial cells were separated from the stationary phase culture by centrifugation at 8500 ϫ g (Sorvall RC-5B centrifuge, DuPont) for 30 min at 4°C. To remove residual bacteria, the supernatant was filtered using a Stericupா/Steritop TM filtration device with 0.22-m pore size polyethersulfone membrane (Millipore Corp., Billerica, MA). Soluble proteins in the filtered supernatant were precipitated overnight with 10% (w/v) TCA at 4°C. The resulting precipitate was pelleted by centrifugation at 8500 ϫ g for 70 min at 4°C, washed several times with ice-cold acetone, and dried in a SpeedVac vacuum centrifuge (Jouan/Thermo Electron). The protein extract was dissolved in an appropriate amount of 8 M urea, 2 M thiourea solution and centrifuged at 14,100 ϫ g (MiniSpin plus, Eppendorf, Westbury, NY) for 2 min to remove insoluble materials and was immediately used for proteomics analysis. Protein concentration was determined by Bradford assay (30) using bovine serum albumin as a standard.
Live/Dead Staining-500 l of S. aureus COL stationary phase culture was centrifuged at 16,000 ϫ g for 2 min, and the resulting cell pellet was stained with the LIVE/DEADா BacLight TM bacterial viability kit (Molecular Probes, Carlsbad, CA) following the protocol provided by the manufacturer and visualized by fluorescence microscopy to detect cell lysis.
Isolation and Preparation of Extracellular Peptide Fraction-To isolate cleaved signal peptides present in the extracellular medium, S. aureus COL was cultured to stationary phase as described above. After removal of bacterial cells by centrifugation, the supernatant was filtered through Microcon Ultracel YM-10 to remove all the high molecular weight proteins (Millipore Corp.). The resulting filtrate was lyophilized and reconstituted in water, and the peptides were isolated, concentrated, and desalted using PepClean C 18 spin columns (Pierce). The eluted peptides were dried in a SpeedVac vacuum centrifuge, resuspended in 40 l of 0.1% TFA in H 2 O, and analyzed directly without any enzyme pretreatment by nano-RPLC-nano-ESI-LTQ-FT-MS/MS as described below.
Protein Sequence Data-S. aureus COL proteome was obtained from The J. Craig Venter Institute (GenBank TM accession number CP000046.1) (31). Theoretical protein masses were calculated from protein sequences using the Protein Analysis Worksheet Software (PAWS) program (freeware edition for Windows 95/98/NT/2000, Pro-teoMetrics, LLC, New York, NY). Additional sequence information was obtained from the Swiss-Prot database at ExPASy (32).
Whole Protein Two-dimensional Liquid Chromatography and Mass Spectrometry-S. aureus COL secreted proteins were separated using an automated two-dimensional LC system that has been described previously (28). In a typical analysis, S. aureus COL extracellular protein extract was first separated by strong cation exchange chromatography and fractionated on line using 20 trapping columns. The contents of the trap were then separated by reverse phase chromatography followed by measurement of protein masses by a quadrupole time-of-flight mass spectrometer. A more detailed description of the method is provided in the supplemental data. Mobile phases along with the gradients used are shown in supplemental Tables 1 and 2.
Peptide Nanoliquid Chromatography-Tandem Mass Spectrometry and Data Analysis-A detailed description of peptide analysis and information on MS/MS database search parameters are provided in the supplemental data. Briefly, proteins trapped on each trapping column were eluted with organic mobile phase and collected off line for trypsin or endoprotease Glu-C digestion. Each trap digest was subsequently analyzed by nano-RPLC-nano-ESI-LTQ-FT-MS/MS.
Signal Peptide and Protein Localization Predictions-To predict the presence of signal peptides and signal peptidase I cleavage sites for the proteins identified in the present work, the following prediction tools were used: SignalP-NN (neural network model) and SignalP-HMM (hidden Markov model) version 2.0 (33) and version 3.0 (34), PrediSi (35) (position weight matrix method), and SigCleave (36, 37) (weight matrix method). When required, LipoP version 1.0 (38) and TMHMM program version 2.0 (39) were used to predict lipoproteins and membrane proteins, respectively. PSORTb version 2.0 (40) was used to predict protein subcellular localization.

RESULTS
Protein Identification Strategy-Sibbald et al. (29) used a rigorous approach utilizing a combination of computational tools and an optimized type I signal peptidase (SpsB) recognition search pattern to estimate that 71 extracellular proteins are produced by S. aureus COL.
To identify proteins present in S. aureus COL extracellular medium, proteins were extracted by TCA precipitation from cultures at stationary growth phase, a phase during which extracellular proteins are preferentially expressed (27,41). The first step in the analysis was to identify the S. aureus COL secretome through peptide analysis. The proteins from each of the C 4 trapping columns were digested and analyzed as discussed above. From these peptide data, a total of 127 proteins (supplemental Table 3) were identified in the extracellular medium, and using bioinformatics tools, we classified these proteins into five categories based on their predicted subcellular localization. We classified 59 of the identified proteins as secreted proteins (Table I), and all of these proteins   except two contained potential Sec-type signal peptides with  SPase I cleavage sites and lacked any cell wall or membrane  retention signals. The remaining 68 proteins included seven  cell wall-anchored proteins, four lipoproteins, four membrane  proteins (supplemental Table 4), and 53 cytoplasmic proteins (supplemental Table 5); these proteins were not predicted to be secreted into the growth medium.
The second step in the analysis of extracellular proteins involved mass measurements of whole proteins captured on each trap by ESI-Q-TOF-MS. Whole protein masses from a particular trap were assigned to molecules that had been identified by peptide analysis of the same trap. The general experimental approach used to identify S. aureus COL extracellular proteins is demonstrated below by using Trap 7 as an example. Peptides corresponding to thermonuclease (Nuc) (Fig. 1A), ␤-hemolysin (Hlb), SACOL0755, SACOL0859, serine protease SplA (SplA), serine protease SplE (SplE), penicillinbinding protein 2 (MecA), and ribosomal protein L9 (RplI) were heavily populated in Trap 7. Because Nuc, Hlb, SACOL0755, SACOL0859, SplA, and SplE are predicted to undergo posttranslational cleavage of the signal peptide, predicted masses for the mature secreted proteins were calculated by obtaining predicted cleavage site information from six signal peptide prediction programs; the resulting list of predicted masses was then matched against the experimental masses derived from whole protein MS analysis of Trap 7 proteins. For example, signal peptide cleavage site prediction of Nuc (theoretical mass of 25119.9 Da) yielded five different predictions (cleavage at residue 23, 25, 30, 57, and 60) resulting in five possible predicted masses for the mature Nuc (Fig. 1B). It is important to point out here that the computational methods used to predict signal peptide cleavage sites frequently provide conflicting predictions as exemplified by this case. Fig. 1C displays the total ion chromatogram of the Trap 7 fraction. Deconvolution of the raw spectrum (Fig. 1D) yields a protein peak at a retention time of 47 min with a mass of 18,782.0 Da (Fig. 1E). This observed mass matches very closely with only one of the possible predicted masses of Nuc (18,782.3 Da) and confirms the signal peptide cleavage site position as Ala 60 . The cleavage site position of Nuc was further corroborated by identification of the N-terminal peptide of the mature protein by MS/MS analysis as shown in Fig. 1F. Whole protein identification and confirmation of signal peptide cleavage site positions of the other Trap 7 secreted proteins were accomplished similarly. Of the six programs used in the present study, only two (SignalP 3.0-NN and PrediSi) yielded the correct cleavage site for Nuc. Capitalizing on the ability to accurately determine the whole protein mass and hence the cleavage site position of the secreted proteins, we took this opportunity to evaluate the prediction accuracy of different signal peptide prediction programs, and the results are discussed below.
Whole protein identification of membrane proteins and cytoplasmic proteins was achieved by directly matching the observed masses with theoretical masses calculated from the  protein sequences. In the case of cell wall-anchored proteins and lipoproteins that contain cleavable signal peptides, the strategy used for secreted proteins was applied. For any protein that could not be identified by matching theoretical or predicted masses (after loss of signal peptide) to the observed masses, additional modifications including methylation (ϩ14.03 Da), acetylation (ϩ42.04 Da), oxidation (ϩ15.99 Da), formylation (ϩ27.99 Da), and protein truncation were considered. The PAWS program was used to assign unmatched protein masses derived from whole protein MS analysis to the corresponding truncated secreted proteins. In this approach, an unassigned protein mass from a particular trap was searched against the entire sequence of the suspected protein to determine whether any part of the sequence has a mass that matches the input mass within the experimental error (Ϯ1.5 Da). cessfully confirmed 53 of the 59 secreted proteins that had been identified by peptide analysis (Table I). Of these, 39 proteins were identified directly by matching the observed masses to the predicted masses calculated by removing a signal peptide. Remaining proteins were identified by considering additional post-translational modifications. Six proteins could not be identified by whole protein analysis probably because of their low abundance or extensive degradation during culture or sample preparation by secreted proteases (20,26,42). Only notable secreted proteins exhibiting modifications other than routine signal peptide loss will be discussed in detail below.

Post-translational Modifications of Secreted Proteins-Whole protein mass measurements of 20 trap fractions suc-
Cytotoxins-␦-Toxin is a 45 residue (Hld-45; 5009.1 Da) protein that does not contain a classical N-terminal signal peptide. However, the mature form is 26 residues (2978.5 Da) in length, indicating that the first 19 residues constitute the signal peptide (43). Because Hld was identified predominantly as an N-terminal methionine-formylated species, some studies have suggested that the translational start codon has been misassigned such that the native form of Hld is only 26 (Hld-26) residues long and is secreted without a signal peptide (44,45). In the present study, two forms of Hld (Fig. 2) were identified by peptide MS/MS analysis of extracellular peptide fraction (see "Experimental Procedures"); Hld 1 with a monoisotopic mass of 3020.6 Da corresponds to Hld-26 with oxidized N-terminal formylated methionine, and Hld 2 (monoisotopic mass of 2976.6 Da) corresponds to Hld-26 with unformylated N-terminal methionine. It is not totally clear whether Hld 2 is formed as a result of post-translational deformylation of N-terminal formylated Hld-26 or whether it results from Hld-45 by signal peptide cleavage. The ambiguity arises from the fact that deformylation by peptide deformylase is usually followed by removal of N-terminal methionine by methionine aminopeptidase if the succeeding residue has a small side chain; although Ala is the second residue in Hld-26, we did not find any evidence of methionine removal.
The theoretical mass of leukotoxin LukE (LukE) is 34,819.1 Da. Signal peptide cleavage at Ala 27 was predicted by all of the programs, leading to an expected mass of 31,864.5 Da. Several peptides corresponding to LukE (sequence coverage, 68%) including the N-terminal peptide of the mature protein (NTNIEN-IGDGAEVIKR, residues 28 -43) were identified in the Trap 8 tryptic digest. Whole protein mass spectra from Trap 8 did not contain any peak that matched the predicted mass. However, as seen in Fig. 3, a mass of 31,751.5 Da that is 114 Da (Ϯ1.5 Da) lower than the predicted mass of LukE was observed in Trap 8. This was identified as LukE with a single C-terminal residue (Asn) truncation. It is most likely that the truncated form resulted from C-terminal degradation by extracellular proteases.
␥-Hemolysin, component B (HlgB) is a 325-residue protein with a theoretical mass of 36,711.0 Da. The predicted signal peptide cleavage site is Ala 26 , yielding an expected mass of 34,048.7 Da. Peptides corresponding to HlgB were observed in Trap 4 (sequence coverage, 69%); however, whole protein MS analysis of the same trap did not show a mass that matched with the expected mass of HlgB. Instead, three co-eluting protein masses were observed that differed in mass by less than 150 Da, suggesting modified forms of the same protein. The PAWS program revealed that these three protein masses matched well with the three forms of mature HlgB (HlgB 1 , HlgB 2 , and HlgB 3 ). As shown in Fig. 4, the observed mass of 33,392.6 Da corresponds to Pro 32 -Asn 324 (HlgB 1 ), the observed mass of 33,463.6 Da corresponds to Glu 27 -Glu 320 (HlgB 2 ), and the observed mass of 33,507.4 Da corresponds to Lys 29 -Glu 322 (HlgB 3 ). The most plausible explanation for the presence of three HlgB forms is truncation of both N-terminal and C-terminal residues due to proteolytic degradation as suggested above for LukE. Aerolysin/leukocidin family protein (SACOL2006) has a theoretical mass of 40,434.0 Da, and signal peptide cleavages at Ser 29 and Ala 27 were predicted for the protein. Peptide analysis of the Trap 10 digest identified SACOL2006 with a sequence coverage of 73%. Interestingly, Trap 10 whole protein MS analysis identified two co-eluting forms of the protein, SACOL2006 1 (37,419.3 Da) and SACOL2006 2 (37,620.8 Da), that appear to have formed as a result of signal peptide processing at both Ser 29 and Ala 27 , respectively (Fig. 5). This observation of signal peptide processing at two different sites is quite unusual because signal peptidases generally cleave signal peptides with high fidelity (46). This observation along with another example will be discussed below.
Superantigenic Toxins-Staphylococcal enterotoxin (Sek) has a theoretical mass of 27,721.1 Da. Signal peptide cleavage at Ala 23 was predicted by all the programs, leading to an expected mass of 25,333.1 Da. Peptide analysis of the Trap 12 digest identified Sek with a sequence coverage of 82%. Trap 12 whole protein MS analysis identified Sek as a Cterminal truncated protein because the observed mass of 24,699.6 Da coincided with the mass of the mature Sek protein formed by signal peptide cleavage at position Ala 23 and removal of residues YKTI from the C terminus (24,698.4 Da). The signal peptide cleavage at Ala 23 was further confirmed by observation of the N-terminal peptide QGDIGIDNLR (residues 24 -33). The presence of truncated Sek is most probably due to the degradation of the protein.
The theoretical mass of staphylococcal enterotoxin type I (Sei) is 28,184.6 Da. Signal peptide cleavage at Ala 26 was predicted by all the programs, leading to an expected mass of 25,076.9 Da. Sei was identified from the Trap 20 digest with a sequence coverage of 77%. Trap 20 whole protein MS analysis identified Sei as a C-terminal truncated protein with an observed mass of 24,848.0 Da corresponding to signal peptide cleavage at position Ala 26 and removal of C-terminal residues TE. The signal peptide cleavage at Ala 26 was further confirmed by observation of N-terminal peptide DVGVINLRN-FYANYEPE (residues 27-43). C-terminal truncated Sei may be formed by proteolytic degradation.
Proteases-Serine proteases SplA, SplB, SplC, SplE, and SplF and cysteine protease precursor SspB were identified in S. aureus COL extracellular medium by peptide MS/MS as well as whole protein MS analysis, and signal peptide cleavage was the only modification observed. Serine protease SplD was somewhat more interesting. Its theoretical mass is 25,669.1 Da. Signal peptide cleavage at Ala 26 was predicted by the prediction programs, yielding an expected mass of 22,001.8 Da. Peptide analysis of the Trap 5 protein digest identified SplD with a high sequence coverage of 77%. Nevertheless, Trap 5 did not yield a whole protein mass that matched the predicted mass. Instead, a recurring mass peak at 22,012.2 Da that is 10.4 Da higher than the predicted mass was observed co-eluting with SplF (Fig. 6A). Because SplD shares 96% sequence identity with SplF (47), we expected it to co-elute with SplF. The observed 10.4 Da mass difference was suspected to be a sequencing error. Examination of single nucleotide substitutions that could account for the observed mass discrepancy yielded two potential candidates for a sequencing error, Ser 3 Pro with a mass difference of 10 Da and Gln 3 His with a mass difference of 9 Da. Phylogenetic analysis combined with mass spectrometry has been used previously in our laboratory to identify a large number of sequencing errors in Bacillus subtilis strain 168 (48). A similar approach was utilized in the present study. Multiple amino acid sequence alignment of SplD from different S. aureus strains using the ClustalW program indicated high sequence homology between S. aureus COL and other strains (99.6% sequence identity) and revealed Gln 68 3 His 68 as the plausible sequencing error (Fig. 6B). This was in fact confirmed by MS/MS analysis of the peptide LITNTNVAPYSGVTWMGAGT-GFVVGNHTIITNK (residues 42-74) (Fig. 6C).
A comparison of SplD nucleotide sequences from various S. aureus strains revealed that the CAT codon at position 68 is highly conserved in all strains, suggesting that the observed discrepancy is not a single nucleotide polymorphism. Only SplD of S. aureus COL has a CAA codon at position 68 in the reference genome, and we believe that an A3 T nucleotide sequencing error at this position resulted in the observed Gln 68 3 His 68 sequencing error.
Lipolytic Enzymes- Rollof and Normark (49) have reported that lipase (76 kDa) in S. aureus strain TEN 5 is secreted into the culture medium as a prolipase (82 kDa) after cleavage of the signal peptide; subsequent processing (removal of propeptide) of the prolipase results in a mature lipase (44)(45). It is important to point out that the molecular mass of the observed prolipase is significantly higher than the predicted mass (73 kDa); the reason for the observed mass difference, however, was not explained. In the present study, Lipase1 (Lip1) was identified by peptide analysis of the Trap 8 digest (96% sequence coverage), and Lipase2 (Lip2) was identified from the Trap 9 digest with a sequence coverage of 71%. Lip1 is a 680-residue protein (76, (67,152.0 Da), respectively, and revealed that after 8 h of growth both Lip1 and Lip2 were present in the extracellular medium as unprocessed proenzymes (Fig. 7). We did not find any evidence of mature lipase forms in the extracellular medium. In contrast to previous studies based on SDS-PAGE and Western blot analysis where the observed masses of staphylococcal prolipases were inexplicably higher than predicted masses (49 -51), our study provided accurate mass determination of prolipases, allowing confident identification of the proteins.
Peptidoglycan Hydrolases-Bifunctional autolysin (Atl; 138 kDa) is a bacteriolytic enzyme capable of causing cell lysis. It consists of two functionally distinct domains. Several studies have reported that Atl undergoes proteolytic processing to generate 62-kDa (amidase) and 51-kDa (glucosaminidase) extracellular lytic enzymes (52,53). In S. aureus COL, Atl is a 1256-amino acid protein (137,334.9 Da) and is predicted to contain a signal peptide domain (residues 1-29), a propeptide domain (residues 30 -198), an N-acetylmuramoyl-L-alanine amidase domain (AM; residues 199 -775), and an endo-␤-Nacetylglucosaminidase domain (GL; residues 776 -1256). The predicted masses of mature AM and GL are 63,008.7 and 53,479.5 Da, respectively. Atl was identified in Traps 8 and 9, and peptide analysis yielded a high sequence coverage of 99%. Whole protein MS analysis of the same traps resulted in the identification of three gene products of Atl (Fig. 8)  Miscellaneous Enzymes-For SACOL1071, signal peptide processing was observed at multiple sites (Fig. 9A). The theoretical mass of SACOL1071 is 11,344.8 Da, and signal peptide cleavage at Ala 24 and Ala 26 was predicted, leading to Peptide analysis of the Trap 3 digest identified SACOL1071 with a high sequence coverage of 91%. Trap 3 whole protein MS analysis identified two co-eluting forms of the protein that represented signal peptide processing at Ala 26 as well as Ala 24 , SACOL1071 1 with an observed mass of 8720.1 Da (Thr 27 -Lys 105 ) and SACOL1071 2 with an observed mass of 8907.1 Da (Asp 25 -Lys 105 ). A comparison of the protein peak intensities suggested that SACOL1071 1 is the major form present in the extracellular medium. Furthermore, a whole protein mass of 9901.6 Da (SACOL1071 3 ) was also observed co-eluting with SACOL1071 1 . This matched the predicted mass of SACOL1071 (9902.0 Da) after removal of N-terminal residues 1-14. This observation implicated another signal peptide processing site for this protein. Identification of the N-terminal non-tryptic peptide ATLVTPNLNADATTNTTPQIK (residues 15-35), which represents the N-terminal peptide of the mature protein via peptide MS/MS analysis, confirmed signal peptide processing at position 14 (Fig. 9B). The general attributes of the 14-residue signal sequence of SACOL1071 do not conform to those typical for a Sec-type signal peptide. The implications of observing multiple cleavage sites in SACOL1071 will be discussed later.
Surface Adhesins-The theoretical mass of secretory extracellular matrix and plasma-binding protein (Empbp) is 38,484.9 Da (residues 1-340). Signal peptide cleavage is predicted at Ala 24 32 . SdrH was identified with high sequence coverage (81%) in the digest of Trap 3, but whole protein MS analysis of the same trap did not show any mass that was close to the predicted mass. Instead, we detected a C-terminally truncated protein with an observed mass of 38,085.1 Da (Lys 33 -Lys 376 ) formed by the removal of residues 377-419. The signal peptide cleavage at position Ala 32 was further confirmed via peptide MS/MS analysis by identification of N-terminal peptide KDNLNGEKPTTNLNHNITSPSVN-SEMNNNETGTPHESNQTGNEGTGSNSR (residues 33-82). Peptide analysis also indicated that the C-terminal peptides corresponding to the last 44 residues were missing.
Proteins with Unknown Functions-Staphylococcal secretory antigen SsaA2 (SsaA2) has a theoretical mass of 29,327.1 Da (267 residues) and a predicted mass of 26,715.1 Da after signal peptide cleavage at Ala 27 . Peptides corresponding to SsaA2 were found in Trap 3. However, whole protein MS analysis of the same trap did not provide a mass that corresponded to full-length mature protein. Instead, SsaA2 was identified as two protein fragments, an N-terminal protein fragment with an observed mass of 12,210.3 Da (Ser 28 -Gly 127 , SsaA2 1 ) and a C-terminal fragment with an observed mass of 14,294.9 Da (Ala 131 -His 267 , SsaA2 2 ). Furthermore, observation of the peptide ASYSTSSNNVQVTTTMAPSSNGR (residues 131-153) that resulted from non-tryptic cleavage at the N terminus of the peptide confirmed degradation of the protein into two fragments.
Staphyloxanthin biosynthesis protein (SACOL2295) has a theoretical mass of 17,424.8 Da. Signal peptide cleavage is predicted at Ala 22  the Trap 3 digest identified this protein with a high sequence coverage of 71%. The observed mass (14,756.7 Da) of SACOL2295 obtained from Trap 3 whole protein MS analysis was 17 Da higher than the predicted mass (14,739.7 Da). This led us to propose two modifications for SACOL2295: loss of signal peptide and oxidation of the mature protein. Thiol groups of cysteine residues are known to be sensitive toward oxidation; Wolf et al. (54) have already reported the oxidation of the Cys 69 residue in SACOL2295. In this study, it is highly possible that Cys 69 is the site of the proposed modification; however, we could not confirm the modification site by peptide MS/MS experiments as the residue was not mapped.
Virulence factor EsxA (EsxA) has a theoretical mass of 11,036.2 Da, and no Sec-type signal sequence is predicted at its N terminus. In the present study, peptide analysis identified EsxA in the Trap 3 digest with high sequence coverage (99%), and there was no evidence of loss of signal peptide from the protein. Burts et al. (55) identified EsxA along with EsxB in S. aureus strain Newman and have shown that they are exported via an ESAT-6 secretion pathway (type VII pathway). Because EsxA of S. aureus COL shares 100% sequence identity with S. aureus Newman, we expect that it is similarly exported. Whole protein MS analysis of Trap 3 did not uncover any mass that was close to the theoretical mass (11,036.2 Da). Instead, we observed a very intense peak at 10,905.2 Da that was 131 Da lower than the theoretical mass, suggesting the removal of N-terminal methionine by methionine aminopeptidase. Identification of the N-terminal methionine truncated peptide AMIKMSPEEIRAKSQSYGQGSDQIRQILSDLTRAQGE (residues 2-38) corroborated this hypothesis. The whole protein MS analysis in combination with peptide analysis definitively confirmed the absence of signal peptide processing in EsxA. In mycobacterium tuberculosis, proteins EsxA and EsxB form a tight 1:1 dimer (56) that is required for stability of the proteins, and this interaction is thought to take place in the cytosol prior to protein export. Burts et al. (55) reported that in S. aureus Newman EsxB is required for the synthesis and secretion of EsxA and vice versa. This led them to believe that EsxA and EsxB also form a heterodimer in S. aureus. However, in the present study, EsxB was not identified despite the fact that EsxA yielded a rather intense signal, suggesting that EsxB may not be required for secretion of EsxA in S. aureus COL. Sundaramoorthy et al. (57) also did not observe heterodimer formation following incubation of S. aureus EsxA and EsxB proteins; instead, EsxA crystallized as a homodimer.
The predicted mass of SACOL0270 is 30,421.1 Da following signal peptide cleavage at Ala 24 . Peptides corresponding to this protein were found in Trap 12; however, whole protein MS analysis of the same trap did not show any mass that matched the predicted value within the experimental error. Instead, an unmatched mass of 30,377.9 Da was observed that we believe is SACOL0270. The observed mass discrepancy of Ϫ42 Da is most probably due to arginine modification.
Hydrolysis of arginine to form ornithine is a well known modification that results in a mass shift of Ϫ42 Da.
Post-translational Modifications of Non-secretory Proteins-Similar to secreted proteins, post-translational modifications of cell wall-anchored proteins, membrane proteins, lipoproteins, and cytoplasmic proteins were characterized, and a detailed discussion on the observed modifications is provided in the supplemental data.
Stable Cleaved Signal Peptides and Signal Peptide Fragments-Several reports have suggested that after cleavage of a signal peptide from a preprotein rapid removal and degradation of the signal peptide is important for proper functioning of the export machinery (58,59). Nevertheless, peptide analysis of some trap fractions indicated the presence of stable cleaved signal peptides and signal peptide fragments derived from a few secreted proteins. Because TCA does not precipitate peptides efficiently (60), we suspected that there may be more peptides in the extracellular medium than those identified in the TCA protein extract. Using the procedure outlined under "Experimental Procedures," we attempted to isolate the peptides present in the S. aureus COL stationary phase culture. Indeed, RPLC-ESI-LTQ-FT-MS/MS of the peptide extract revealed the presence of several peptides including stable cleaved signal peptides of five proteins, Sle1, SACOL0723, SceD, IsaA, and SACOL2295 (supplemental Table 6), and signal peptide fragments of 18 secreted proteins (Table II). As an example, MS/MS spectra of cleaved signal peptide of IsaA and signal peptide fragment of SACOL1164 identified in the present study are shown in Fig.  10. It is noteworthy that all of the observed signal peptide fragments are from C-terminal portions of respective signal peptides containing SPase I-cleaved sites and appear to have formed by cleavage in the hydrophobic region of the signal peptide. To account for the signal peptide fragments observed in the present study, we have proposed cleavage sites in the SPase I-processed signal peptides as shown in Table II. The implications of these observations will be discussed later.
Signal Peptide Prediction Accuracy-The prediction accuracy of computational programs commonly used to predict signal peptides and cleavage site position has been debated. Few studies have evaluated the performance of signal peptide prediction programs using experimentally verified signal peptide data from different organisms (61)(62)(63). To the best of our knowledge, the suitability of the commonly used prediction programs to predict secretory proteins and signal peptide cleavage sites in S. aureus has not been reported. Results from the present study are shown in Fig. 11. It is evident from the figure that   FAITATSGAAAFLTHHDAQA  MKKLA2FAITATSGAAAFLTHHDAQA  SACOL0860  VLTLVVVSSLSSSANA  MTEYLLSAGICMAIVSILLIGMAISNVSKGQYAKRFFYFATSCL2VLTLVVVSSLSSSANA  SACOL0908  ALVLTTVGSGFHSSSNYNGINNVAKA  MNKKLLTRTLIAS2ALVLTTVGSGFHSSSNYNGINNVAKA  SACOL1062  LTLVGSAVTAHQVQA  MAKKFNYKLPSMVA2LTLVGSAVTAHQVQA  SACOL1164  AISLTVSTFAGESHA  MKKNFIGKSILSIA2AISLTVSTFAGESHA  SACOL1864  TILTSITGVGTTMVEGIQQTAKA  MNKNIIIKSIAAL2TILTSITGVGTTMVEGIQQTAKA  SACOL1868  TILTSVTGIGTTLVEEVQQTAKA  MNKNVVIKSLAAL2TILTSVTGIGTTLVEEVQQTAKA  SACOL2003 ANLLLVGALTDNSAKA MVKKTKSNSLKKVATLAL2ANLLLVGALTDNSAKA SACOL2088 a SLAVGLGIVAGNAGHEAHA MKKTLLAS2SLAVGLGIVAGNAGHEAHA SACOL2197 LGLLSTVGAALPSHEASA Although previous studies confirmed the presence of a given protein in the S. aureus extracellular milieu, they failed to provide a detailed characterization of the proteins. This is a particular weakness of peptide-based analyses because only a fraction of the total theoretical peptide population of a given protein may be identified. The present study provides a comprehensive picture of the secretome of S. aureus COL by identification of the proteins and characterization of their post-translational modifications.
All but two of the 59 secreted proteins identified in the present study were predicted to possess Sec-type signal peptides, and we were able to verify the signal peptide loss in these proteins; this confirmed that they were exported via a Sec-dependent pathway. Also, we confirmed that EsxA, which is known to be exported via the ESAT-6 pathway, does not contain a cleavable signal peptide, and the only modification observed was the removal of N-terminal methionine.
In a majority of proteins, signal peptide loss was the only modification observed. Other observed modifications included proteolytic processing, N-terminal formylation, methionine removal, oxidation, formation of ornithine, and protein truncation. Degradation of a few secreted proteins observed in the present study indicated proteolytic activity in the culture supernatants. This observation has been reported by several investigators, and it has been suggested that the secreted proteins are degraded by the action of their secreted proteases during culture and sample preparation (20,26,42,(65)(66)(67). Degradation of proteins by extracellular proteases has also been reported in other microorganisms; in B. subtilis, it has been demonstrated that mutants lacking proteases exhibit a substantial increase in the abundance of various extracellular proteins compared with the wild type (65)(66)(67). Degradation of extracellular proteins may be due to slow or incorrect post-translational folding of the proteins or to the presence of exposed protease recognition sequences in the folded protein (52,54). It may also be a means of nutrient recycling for survival (20,66,68). However, in the present study, a majority of the proteins were refractory to nonspecific protease activity because they were identified as intact proteins. analysis from three or more peptides with a Mascot score above the threshold of significance. Seven predicted secretory proteins, putative uncharacterized protein (SACOL0129), exotoxin 3 (SACOL0468), exotoxin (SACOL0470), exotoxin 3 (SACOL0478), surface protein (SACOL0479), cell wall hydrolase (SACOL1264), and hypothetical protein (SACOL1870), were identified from one or two peptide sequences (Mascot score, p Ͻ 0.05) and were not included in the list of identified secreted proteins because of the stringent protein identification criteria used in the present study. These proteins are apparently present in the extracellular medium in low abundance. The remaining 12 predicted proteins not detected in the extracellular medium are probably not secreted by S. aureus COL under the conditions studied or are present in trace amounts. There is no evidence from the published literature to indicate that these proteins are indeed secreted by S. aureus COL. Recent proteomics data on membrane, cell wall, and extracellular proteins of B. subtilis revealed that a good number of proteins that are predicted to contain cleavable Sec-type signal peptide and an SPase I recognition site are not secreted into the medium but are in fact retained in the membrane (69); this could be the case for the predicted proteins not identified in the present study. Furthermore, five proteins that were not predicted to be secreted because of the presence of transmembrane domains (Nuc and SACOL0442), the presence of Thr in the Ϫ1 position (SACOL2179) or ϩ1 position (SACOL1071), or the presence of Tyr in the ϩ1 position (SACOL0270) relative to the cleavage site (29) were identified in the present study by whole protein and peptide MS analysis as being secreted proteins released into the medium by removal of Sec-type signal peptides. This indicates that Thr in the Ϫ1 and ϩ1 positions and Tyr in the ϩ1 position are accepted by S. aureus SPase I; a discussion of the amino acid residues accepted by SPase I at positions Ϫ3 to ϩ1 relative to the signal peptide cleavage site and the frequency of their occurrence is presented below.
A comparison of secreted proteins identified in the present study with those identified in various S. aureus strains (29) showed an overlap between identified proteins and revealed potential vaccine and drug candidates. Of the 56 secreted proteins that have been identified in other S. aureus strains (29), 48 are encoded in S. aureus COL. 43 of these (Table I) were identified in the present study by three or more peptides, and three proteins (SACOL0478, SACOL0479, and SACOL1870) were identified by one or two peptides. Two proteins (SACOL0209 and SACOL2691), however, were not identified.
Signal Peptides and Cleavage Sites of S. aureus COL Secreted Proteins- Table III lists the signal sequences of 59 S. aureus COL proteins (secretory and cell wall-associated proteins) identified in the present study. In Gram-positive bacteria, SPase I recognizes residues at positions Ϫ3 and Ϫ1 with respect to the cleavage site, and Ala-X-Ala is the most common sequence preceding the signal peptide cleavage site (70). It is evident from Table III that the signal sequences of S. aureus COL proteins identified in the present study all contain the N-, H-, and C-domains of a typical Sec-type signal peptide. The length of the signal peptides varies from 23 to 60 amino acids with an average of 31 residues. Table IV lists the residues accepted at and around the verified SPase I cleavage sites of S. aureus COL proteins identified in the present study. In a majority of the proteins, Ala is predominantly preferred at Ϫ3 (77%) and Ϫ1 (97%) positions. Residues Val (10%) and Ser (10%) are also accepted at the Ϫ3 position and occur with a higher frequency than Thr, Leu, and Ile (2%). With respect to the Ϫ1 position, residues Ser and Thr are also accepted but with a markedly lower frequency (2%) than Ala. The residues found in Ϫ3 and Ϫ1 positions of S. aureus COL signal sequences are small and uncharged; this is in agreement with the assumption that side chains of residues at the Ϫ1 and Ϫ3 positions are bound in two shallow hydrophobic substrate-binding pockets (S1 and S3) of the active site of SPase I (71). In contrast, the side chain of the residue at position Ϫ2 is thought to be pointing outward from the enzyme. As a consequence, a variety of residues appear to be tolerated at the Ϫ2 position including Lys, Asn, Gln, His, Asp, Ser, Glu, Leu, Phe, Tyr, Gly, and Arg with a preference for Lys (23%). There appears to be a preference for Ala (31%) at the ϩ1 position. Other residues including Ser, Glu, Lys, Asp, Gln, Thr, Phe, Asn, Tyr, and Leu were also accepted in the ϩ1 position. Approximately 78% of S. aureus COL secreted proteins identified in the present study possess signal sequences that contain a helix-breaking residue (mostly glycine) in the middle of the H-domain, and about 50% contain a helixbreaking residue (proline or glycine) at position Ϫ7 to Ϫ4 relative to the predicted processing site for SPase I. Helixbreaking residues found at the end of the H-domain are thought to facilitate cleavage by SPase I (71).
Signal Peptide Processing at Two Cleavage Sites by SPase I-Whole protein MS analysis of SACOL1071 and SACOL2006 revealed processing of the signal peptide at more than one site. This observation is very interesting as signal peptide cleavage by SPase I is generally considered to be highly specific (72,73), and reports on the observation of signal peptide cleavage at multiple sites in wild-type proteins are very rare. The reasons for the high fidelity of SPase I are not clearly understood. An important requirement for cleavage by SPase I is the presence of amino acids with small neutral side chains at positions Ϫ1 and Ϫ3 in the C-region; the Ϫ3 position is less restrictive than the Ϫ1 position (33). This is also evident from Table IV. SACOL1071 and SACOL2006 each have more than one potential cleavage site that is in compliance with the substrate specificity of SPase I and may compete for recognition. The N-terminal signal sequence MNKLLQSLSALGVSATLVTPNLNA 24 2DA 26 2 of SACOL1071 contains the ubiquitous AXA motif (Ϫ3 to Ϫ1 position) as well as an LXA motif (Ϫ5 to Ϫ3 position) with observed cleavages (2) at the Ϫ3 and Ϫ1 positions. The weighted average signal intensities in the whole protein mass spectrometry experiments suggest that signal peptide processing at the Ϫ1 position (82%) is preferred over the Ϫ3 position (18%). Similarly, the N-terminal sequence of SACOL2006, MKNKKRVLIAS-SLSCAILLLSAATTQA 27 2NS 2 2 also contains two potential sites for signal peptide cleavage, the AXS sequence (Ϫ3 to Ϫ1 position) and the TXA sequence (Ϫ5 to Ϫ3 position) with observed cleavages (2) at the Ϫ1 and Ϫ3 position, respectively. The weighted average of signal intensities of the two mature proteins derived from whole protein MS experiments suggests that processing at the Ϫ3 position (64%) is preferred over the Ϫ1 position (36%). From our results, it appears that SPase I is capable of processing the signal peptide at more than one site depending on the availability of an alternate potential site close by (Ϫ3 and Ϫ1 positions in the two examples). The differences in the nature of residues at Ϫ1, Ϫ3, and Ϫ5 positions; the proximity of the cleavage site to the hydrophobic core; the ␤-turn structure; and the residues that are capable of breaking ␣-helix or ␤-strand structure probably play a role in the efficiency of processing at different sites (74). A close inspection of the signal sequences of the proteins identified in this study (Table III) indicates that there are 11 additional sequences that have SPase I-compatible residues in the Ϫ5 to Ϫ3 positions. However, signal peptide cleavage at only one site was detected in these cases. The observed cleavages at two sites for SACOL1071 and SACOL2006 may be due to some unique characteristics of their signal sequences. In general, prokaryotic signal peptides possess redundant information in that they contain more than one potential cleavage site, more than one basic residue in the N-region, and a longer H-region than required (75,76). It is not clear why alternate sites for signal peptide cleavage exist or whether there is some biological significance to this. Possible Secretion of SACOL1071 via Alternate Secretory Pathway-Chitin, a homopolymer of ␤-1,4-N-acetyl-D-glucosamine (GlcNAc), is one of the most abundant natural polymers. Several bacterial species secrete chitinolytic enzymes and chitin-binding proteins that are thought to degrade chitin (77,78). SACOL1071, a putative chitinase B protein, has a predicted Sec-type signal peptide and was identified in this study as a mature protein formed by removal of this peptide. In addition, another mature form of the protein, SACOL1071 3 , also was identified by whole protein and peptide analysis. It appears to have been secreted by cleavage of a 14-residue N-terminal segment (Fig. 9B). This 14-residue signal peptide, MNKLLQSLSALGVS, does not resemble any known signal peptide, suggesting secretion of SACOL1071 3 by an alternate pathway. In Gram-negative Pseudomonas aeruginosa bacteria, Folders et al. (79) identified a chitinase C (ChiC) protein in the extracellular medium following cleavage of an 11-residue signal peptide (MIRIDFSQLHQ). ChiC also does not contain a typical N-terminal sequence. Because SACOL1071 has a functional Sec-dependent export route, it would be intriguing if it were secreted by an alternate pathway.
Non-secretory Proteins in Extracellular Medium-The presence of non-secretory proteins in the extracellular medium has been reported by several investigators for a number of organisms including S. aureus (24,65,80). The possible explanation for the presence of cytoplasmic proteins in the extracellular medium is cell lysis during growth; this has been visually confirmed in the present study by fluorescence microscopy using the live/dead staining method. Although only 3.0 Ϯ 0.5% of the cells in the stationary phase culture were lysed, that is sufficient to detect the very abundant proteins listed in supplemental Table 5. The rest of the non-secretory proteins were most probably released into the extracellular medium by proteolytic processing, shedding, or cell wall turnover as suggested by other investigators (69,81).
Stable Cleaved Signal Peptides and Signal Peptide Fragments-We have reported our unexpected findings on the presence of cleaved stable signal peptides and signal peptide fragments in the extracellular medium. In Escherichia coli, two signal peptide peptidases, a membrane-bound protease IV and cytoplasmic oligopeptidase A, have been identified (83). It has been proposed that protease IV initially cleaves the SPase I-processed signal peptides by making endoproteolytic cuts, and the products of the initial cleavage may diffuse or be transported back into the cytoplasm and further degraded into amino acids by oligopeptidase A and other cytoplasmic enzymes. Similar proteins have also been identified in B. subtilis (84). However, homologous proteins have not been found in S. aureus (29). Because we have identified fragments of signal peptides, degradation of cleaved signal peptides must also happen in S. aureus by some unknown proteases. All the signal peptide fragments observed in the present study appear to have formed by endoproteolytic cleavage at hydrophobic residues including Leu, Thr, Ala, Val, Phe, and Met (Table II), suggesting that the protease responsible for processing signal peptide in S. aureus may have a preference for hydrophobic residues and that its substrate specificity is similar to protease IV of E. coli (83). As mentioned earlier, only C-terminal signal peptide fragments and not N-terminal signal peptide fragments were identified in the extracellular medium in the present study. This suggests that an endoproteolytic cut in the hydrophobic region of the signal peptides releases the C-terminal signal peptide fragments into the extracellular medium, whereas the N-terminal fragments are retained in the membrane or released into the cytoplasm or completely degraded. The observation of cleaved stable signal peptides in the extracellular medium raises some interesting questions. Because various reports have suggested that after cleavage from the preprotein rapid degradation of signal peptide is important for proper functioning of the export machinery (58,59), why are the observed cleaved signal peptides stable? Also, because their long hydrophobic cores should lead to peptide retention in the membrane, how are they released into the medium? What is the biological role of the released signal peptide and signal peptide fragments? The fact that the signal peptides of Sle1, SACOL0723, SceD, IsaA, and SACOL2295 were detected in the culture supernatant indicates that they have high solubility in water. Also the Zyggregator algorithm (85) (which analyzes aggregation propensities of proteins) predicts that the cleaved signal peptides identified in the present study should not have the propensity to aggregate. It is possible that certain unique biophysical properties allow the observed cleaved signal peptides to be readily released from the membrane into the extracellular medium after cleavage from the preprotein.
It is important to note here that the fate of cleaved signal peptides and their fragments is not clearly understood, particularly for prokaryotes. However, recent studies have suggested that cleaved signal sequences and their fragments may have important biological functions including signaling. For instance, bacteriocin release protein, an E. coli lipoprotein, is processed very slowly by signal peptidase II to yield a mature bacteriocin release protein and a stable signal peptide that mediate the translocation of cloacin DF13 (86). In eukaryotes, the N-terminal signal peptide fragments of preprolactin and human immunodeficiency virus type 1 p-gp160 were found to be released into the cytoplasm and bound to calmodulin/Ca 2ϩ , suggesting that the liberated signal peptide fragments may influence signal transduction pathways in the cell (87). As stated by Martoglio and Dobberstein (88), signal sequences are more than just simple greasy peptides possessing a wealth of functional information. It can be conjectured that cleaved signal peptides and signal peptide fragments of S. aureus COL released into the extracellular medium also have the potential to perform important roles. Tristan et al. (82) have proposed that the signal peptide of Panton-Valentine leukocidin LUKS (LukS-PV), a component of hetero-oligomeric pore-forming toxin (PVL) in S. aureus PVLpositive strains, mediates increased adhesion to extracellular matrix components. According to their hypothesis, LukS-PV signal peptide is released from the membrane after cleavage by SPase I and attaches to the cell wall through its unique C-terminal (TSFHESKA) sequence, thus exposing the positively charged N terminus to the extracellular matrix molecules. S. aureus COL does not encode PVL gene; therefore, we could not confirm the release of LukS-PV signal peptide. Also, none of the signal peptides identified in the present study share homology with LukS-PV signal peptide.
Conclusions-We have presented a new approach to study S. aureus extracellular proteins utilizing the combination of whole protein MS analysis and peptide MS/MS analysis that enabled us to provide the most comprehensive view of S. aureus extracellular proteins reported to date. We identified 59 secreted proteins in S. aureus COL and characterized their post-translational modifications. Accurate determination of signal peptide cleavage sites of S. aureus COL secreted proteins allowed us evaluate the prediction accuracies of several programs. Because S. aureus COL secreted proteins are potential drug targets or vaccine candidates, the signal peptide cleavage site information provided by this study will be useful in making constructs and recombinant proteins and in protein engineering. This information will also extend our current knowledge base on experimentally verified signal peptide cleavage sites and may assist in improving prediction accuracies of the programs. We also detected signal peptide processing at multiple sites and the release of cleaved signal peptides and signal peptide fragments into the extracellular medium. These observations are unusual, and their biological significance is yet to be understood. The current approach should be useful in secretome analysis of other S. aureus strains and other pathogenic organisms.