Activation of the SspA Serine Protease Zymogen of Staphylococcus aureus Proceeds through Unique Variations of a Trypsinogen-like Mechanism and Is Dependent on Both Autocatalytic and Metalloprotease-specific Processing*

The serine and cysteine proteases SspA and SspB of Staphylococcus aureus are secreted as inactive zymogens, zSspA and zSspB. Mature SspA is a trypsin-like glutamyl endopeptidase and is required to activate zSspB. Although a metalloprotease Aureolysin (Aur) is in turn thought to contribute to activation of zSspA, a specific role has not been demonstrated. We found that pre-zSspA is processed by signal peptidase at ANA29 ↓, releasing a Leu30 isoform that is first processed exclusively through autocatalytic intramolecular cleavage within a glutamine-rich propeptide segment, 40QQTQSSKQQTPKIQ53. The preferred site is Gln43 with secondary processing at Gln47 and Gln53. This initial processing is necessary for optimal and subsequent Aur-dependent processing at Leu58 and then Val69 to release mature SspA. Although processing by Aur is rate-limiting in zSspA activation, the first active molecules of Val69SspA promote rapid intermolecular processing of remaining zSspA at Glu65, producing an N-terminal 66HANVILP isoform that is inactive until removal of the HAN tripeptide by Aur. Modeling indicated that His66 of this penultimate isoform blocks the active site by hydrogen bonding to Ser237 and occlusion of substrate. Binding of glutamate within the active site of zSspA is energetically unfavorable, but glutamine fits into the primary specificity pocket and is predicted to hydrogen bond to Thr232 proximal to Ser237, permitting autocatalytic cleavage of the glutamine-rich propeptide segment. These and other observations suggest that zSspA is activated through a trypsinogen-like mechanism where supplementary features of the propeptide must be sequentially processed in the correct order to allow efficient activation.

The serine and cysteine proteases SspA and SspB of Staphylococcus aureus are secreted as inactive zymogens, zSspA and zSspB. Mature SspA is a trypsin-like glutamyl endopeptidase and is required to activate zSspB. Although a metalloprotease Aureolysin (Aur) is in turn thought to contribute to activation of zSspA, a specific role has not been demonstrated. We found that pre-zSspA is processed by signal peptidase at ANA 29 2, releasing a Leu 30 isoform that is first processed exclusively through autocatalytic intramolecular cleavage within a glutamine-rich propeptide segment, 40 QQTQSSKQQTPKIQ 53 . The preferred site is Gln 43 with secondary processing at Gln 47 and Gln 53 . This initial processing is necessary for optimal and subsequent Aurdependent processing at Leu 58 and then Val 69 to release mature SspA. Although processing by Aur is rate-limiting in zSspA activation, the first active molecules of Val 69 SspA promote rapid intermolecular processing of remaining zSspA at Glu 65 , producing an N-terminal 66 HANVILP isoform that is inactive until removal of the HAN tripeptide by Aur. Modeling indicated that His 66 of this penultimate isoform blocks the active site by hydrogen bonding to Ser 237 and occlusion of substrate. Binding of glutamate within the active site of zSspA is energetically unfavorable, but glutamine fits into the primary specificity pocket and is predicted to hydrogen bond to Thr 232 proximal to Ser 237 , permitting autocatalytic cleavage of the glutamine-rich propeptide segment. These and other observations suggest that zSspA is activated through a trypsinogen-like mechanism where supplementary features of the propeptide must be sequentially processed in the correct order to allow efficient activation.
Staphylococcus aureus can colonize and infect virtually every tissue and organ system of the body (1,2). A key factor in these defining traits is its ability to sustain bacteremia and adhere to tissues. Consequently it has adapted to growth in blood by sub-version of plasma proteins (3) and the tissue extracellular matrix (4). Members of the MSCRAMM (microbial extracellular matrix-binding protein) family of adhesion proteins bind extracellular matrix ligands such as collagen, fibronectin, and fibrinogen (4) of which the latter two are also abundant in plasma. Manipulation of other plasma proteins such as IgG and von Willebrand factor is facilitated by staphylococcal Protein A (5), whereas coagulase promotes fibrin clot formation by activation of prothrombin (6), and staphylokinase activates plasminogen (7) to facilitate fibrin clot dissolution. Our studies on secreted proteases are also supportive of S. aureus as a paradigm for the manipulation of plasma and coagulation proteins.
The SspA 4 glutamyl endopeptidase, also known as V8 protease, moderates adhesion of S. aureus to fibronectin by degrading cell surface fibronectin-binding proteins in which there is a high glutamic acid content, including a conserved motif that is essential for ligand binding (8,9). SspA is expressed in an operon with a cysteine protease, SspB, and activates the zSspB zymogen by removing its N-terminal propeptide (10). Mature 20-kDa SspB mimics plasma serine proteases as noted by its ability to (i) convert high molecular weight kininogen into heavy and light chains, (ii) hydrolyze the Bz-Pro-Phe-Arg substrate of kallikrein, a serine protease that processes single chain kininogen by excision of the vasoactive peptide bradykinin (RPPGFSPFR), (iii) process the N terminus of the fibrinogen ␤-chain at the same site as plasmin, and (iv) remove the N-terminal domain of fibronectin with a specificity equivalent to plasminogen activator (10). We have also found that antibodies to SspB are produced when mice are challenged with hypervirulent strains of community-acquired methicillin-resistant S. aureus and that SspA and SspB are rapidly expressed in neutrophil vacuoles following phagocytosis (11), alluding to a role in modifying vacuolar proteins.
A third gene in the ssp operon encodes a small protein, SspC, now commonly referred to as Staphostatin B, which we identified as an inhibitor of SspB (10) and for which others defined the structural basis of inhibition (12). The need for an inhibitor is supported by the finding that expression of SspB in Escherichia coli is lethal unless accompanied by Staphostatin B or a Cys 3 Ser active site mutant in SspB. Moreover inactivation of sspC in S. aureus promotes a gross disturbance in growth and protein export, supporting a role for Staphostatin B in protecting against activation of zSspB during protein export (13)(14)(15). One means by which this could occur would be if zSspA were also activated prematurely. However, although it has long been thought that the metalloprotease Aureolysin is in turn essential for activation of zSspA (16), one study found that zSspA was correctly activated albeit with reduced efficiency in an aur-deficient strain (17). In the present study, we elucidated a novel stepwise process for activation of zSspA that minimizes the possibility of its untimely activation. We found that although Aureolysin is essential for activation of zSspA the first step in processing of the N-terminal propeptide requires autocatalytic intramolecular cleavage at glutamine. Aureolysin then processes at Leu 58 and then Val 69 to produce the first active molecules of mature SspA, which then feed back to promote efficient autocatalytic intermolecular processing of remaining zSspA at Glu 65 . This exposes an N-terminal 66 HAN tripeptide that blocks the active site until final Aur-dependent processing exposes the hydrophobic N terminus ( 69 VIL) that is common in the trypsin family proteases. These events must occur in the proper sequence for efficient activation.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-Bacterial strains and recombinant plasmids used in this study are listed in Table  1. S. aureus and E. coli were cultured at 37°C in tryptic soy broth (Becton Dickinson and Company) and LB medium (Invitrogen), respectively. Medium was supplemented with agar (15 g⅐liter Ϫ1 ; Difco) or antibiotic (ampicillin, 100 g⅐ml Ϫ1 ; erythromycin, 10 g⅐ml Ϫ1 ; chloramphenicol, 10 g⅐ml Ϫ1 ; or kanamycin, 25 g⅐ml Ϫ1 ) as required. For production of secreted proteins, cultures were grown for 18 h in tryptic soy broth with vigorous aeration. E. coli DH5␣ (Invitrogen) was used as a host for construction of shuttle vectors that were transferred to S. aureus RN4220 by electroporation (18) before transfer into other strains as indicated in Table 1. The aur::lacZ and sarA::km alleles were transferred from the original host strains into S. aureus RN6390 via transduction with 85 (19) as detailed in Table 1.
PCR and Recombinant DNA Procedures-S. aureus genomic DNA was isolated using the DNeasy Tissue kit (Qiagen) following the manufacturer's protocol for Gram-positive bacteria. Plasmid DNA was isolated using the GenElute Plasmid Miniprep kit (Sigma) following recommended protocols. Polymerase chain reaction was performed with Biotools DNA polymerase (Interscience) in the buffer supplied or with the Expand Long Template PCR System (Roche Applied Science) for products larger than 2 kb. PCR products used in cloning were purified using the QIAquick PCR Purification kit (Qiagen) or Concert Rapid Gel Extraction System (Invitrogen). All plasmids harboring cloned DNA fragments were submitted to the Center for Applied Genomics at the Toronto Hospital for Sick Children for verification of the expected nucleotide sequences.
Purification of Proteases from S. aureus-Mature SspA and SspB were purified from culture supernatant of S. aureus RN6390sarA::km, which expresses high levels of secreted protease due to a defect in SarA, which is a repressor of protease gene expression (53). The 40-kDa SspB zymogen zSspB was purified from culture supernatant of S. aureus RN6390sspA::erm, which cannot activate zSspB due to a defect in production of the SspA serine protease. The purification schemes were as described previously (10). Aureolysin was purified from a sarA::km derivative of RN6390. Protein in stationary phase culture supernatant was precipitated with 80% saturation of ammonium sulfate followed by dialysis into 20 mM Tris-Cl, pH 7.8, containing 5 mM CaCl 2 . Aur was then purified by successive anion exchange steps on HiLoad 16/10 Q Sepharose Fast Flow (Amersham Biosciences) and Tricorn Mono Q medium (Amersham Biosciences). Protein was eluted in linear ascending NaCl gradients. The concentration of purified proteins was determined using the absorbance at 280 nm (A 280 ) combined with extension coefficients calculated using the ExPASy ProtParam tool (ca.expasy.org/tools/protparam.html).
Construction of His 6 -SspA and Mutants in S. aureus-The sspA gene with its native promoter was amplified by PCR of genomic DNA from S. aureus RN6390 using primer aagctt CCATTCGCTCTCAATTCCTTTC and reverse primer ctgcag-TTAGTGATGGTGATGGTGATGAGCTGCATCTGGATT-GTCTG. The primers incorporate HindIII and PstI sites (in italic lowercase), respectively, and the underlined sequence in the reverse primer incorporates a His6 tag prior to the stop codon at the 3Ј-end of sspA. The PCR products were cloned into the HindIII and PstI sites of pRN5548a (10), a variant of pRN5548 (10,20) that lacks the blaZ promoter. The resulting plasmid, pV8H, was then transferred into S. aureus SP6391sspA::erm (21) and 6390aur::lacZ. To substitute the active site serine of SspA, we first excised sspA from pV8H and ligated it into the complementary HindIII and PstI sites of pUC18. PCR-directed mutagenesis was performed on the ligation mixture using the QuikChange II site-directed mutagenesis strategy (Stratagene) together with the complementary primers GTACAACTGGTGGTAATGCAGGTTCACCT-GTATTTAATG and CATTAAATACAGGTGAACCTGCA-TTACCACCAGTTGTAC, which alters (TCA)Ser 237 to (GCA)Ala 237 . The mutagenized plasmid was transformed into E. coli, and after sequencing to confirm the desired substitution, the 3Ј-end of sspA containing the Ser 237 3 Ala substitution was excised by digestion with PspGI and PstI and then used to substitute for the same fragment of the wild type pV8H plasmid from S. aureus SP6391. The resulting pV8H-S237A was then transformed into SP6391 and 6390aur::lacZ genetic backgrounds. The expected nucleotide changes and integrity of the sspA gene was confirmed by nucleotide sequence analyses.
Purification of His 6 -SspA Variants-His-tagged SspA variants were isolated from culture supernatant using magnetic Ni-NTA (Fe-Ni 2ϩ ) beads (Qiagen) for SDS-PAGE and N-terminal sequence analysis or Ni-NTA-agarose slurry (Qiagen) for purification of larger quantities of protein. Briefly the supernatant from a late exponential phase culture (5 h) was mixed with an equal volume of ice-cold 40 mM sodium phosphate buffer, pH 7.4, containing 1 M NaCl and 40 mM imidazole, and rocked with Ni-NTA beads for 30 min at 4°C. After washing the magnetic beads three times in phosphate buffer containing 20 mM imidazole and 500 mM NaCl, bound protein was eluted by boiling in 1ϫ SDS-PAGE reducing buffer. Alternately Ni-NTA-agarose slurry was mixed with culture supernatant and packed into a disposable 5-ml polypropylene column (Pierce) followed by application of 50 ml of wash buffer and elution with 500 mM imidazole elution buffer. Samples were desalted into 20 mM Tris-Cl, pH 7.4, using either PD-10 columns (Amersham Biosciences) or Slide-A-Lyzer dialysis cassettes (Pierce).
Purification of zSspA-The zSspA zymogen was purified from 4-h culture supernatant of RN6390aur::lacZ harboring pV8H. Protein was precipitated from culture supernatant with 80% ammonium sulfate saturation and purified using a 1-ml HiTrap chelating column (Amersham Biosciences) in 20 mM sodium phosphate, pH 7.4, 500 mM NaCl buffer with a linear gradient from 10 to 500 mM imidazole. Eluted fractions were exchanged into 20 mM Tris-Cl, pH 7.8, containing 150 mM NaCl using the HiPrep 26/10 desalting column (Amersham Biosciences). Where indicated, zSspA was processed with either Aureolysin (1:1 zSspA:Aur) or native SspA (50:1 zSspA:native SspA) by incubation for 1 h at 37°C in 40 mM Tris-Cl, pH 7.8, with 5 mM CaCl 2 for Aur or without calcium for SspA. The in vitro processed zSspA was then separated from the non-Histagged proteases by capture on Ni-NTA magnetic beads and eluted with 500 mM imidazole. Slide-A-Lyzer dialysis cassettes (Pierce) were used to dialyze samples into 20 mM Tris-Cl, pH 7.8, 25 mM NaCl.
Enzyme Assays and Substrates-SspA glutamyl endopeptidase activity was measured by monitoring the release of pnitroaniline (pNA) from Z-Phe-Leu-Glu-pNA (Bachem) at 37°C. Assays were conducted in triplicate wells of a microtiter plate in a 100-l volume containing 40 mM Tris-HCl, pH 7.4, and 0.4 mM substrate. Absorbances were read at defined time points on a Bio-Rad model 3550 microplate reader equipped with a 405-nm filter. Stock solutions of 4-nitroaniline (Sigma) were used to create a standard of absorbance versus nmol of pNA to allow calculation of specific activity.
Production of Antibodies-Purified SspB zymogen, native SspA, or Aureolysin were emulsified in Freund's complete adjuvant (Sigma) and administered by subcutaneous injection of 100 g of protein in New Zealand White rabbits. Booster injections administered at 2-week intervals consisted of 100 g of protein emulsified in Freund's incomplete adjuvant (Sigma). Antibodies were affinity-purified from the respective antisera obtained after the second boost using the reagents and protocols provided with the AminoLink Plus Immobilization kit (Pierce).

SDS-PAGE, Western Blotting, and N-terminal
Sequencing-Proteins in cell-free culture supernatant were precipitated with an equal volume of ice-cold 20% (w/v) trichloroacetic acid. After washing with 70% ethanol, the pellets were air-dried and then solubilized in 1ϫ SDS-PAGE sample buffer. After SDS-PAGE, proteins were either visualized by staining with Coomassie Brilliant Blue R-250 or transferred to polyvinylidene difluoride membrane (Millipore) using standard buffer and transfer conditions for Western immunoblots or CAPS transfer buffer using modified conditions (22) when required for N-terminal sequence analysis. Western blot procedures followed standard blocking and washing protocols. Primary antibodies were antigen-specific, and secondary antibody was alkaline phosphatase-conjugated AffiniPure goat anti-rabbit (heavy and light chains) IgG (Jackson ImmunoResearch Laboratories). For N-terminal sequence analysis, blotted proteins were visualized by brief staining with 0.1% Coomassie Blue in 40% methanol. The appropriate bands were excised and submitted to the Advanced Protein Technology Center at the Hospital for Sick Children, Toronto, Ontario, Canada for N-terminal sequence determination.
Mass Spectrometry-Mass spectrometry was conducted at the Toronto Angiogenesis Research Center Proteomics Core Facility in the Sunnybrook Research Institute. For mass determination, purified protein samples were exchanged into 5% formic acid by ultrafiltration, then diluted 20-fold in 50% acetonitrile and 2% formic acid, and analyzed by electrospray ionization-hybrid quadrupole time of flight mass spectrometry (QStar-XL, Applied Biosystems/MDS Sciex) through infusion at 6 l⅐min Ϫ1 flow rate. The machine was cleaned by infusing 100 l of 50% trifluoroethanol to eliminate carryover between samples and was calibrated immediately before each sample. Spectra obtained over 3 min of steady spray were accumulated and used for molecular weight reconstruction with "Bayesian Protein Reconstruct" in the BioAnalyst 1.1.5 software package (Applied Biosystems/MDS Sciex) using default settings unless otherwise indicated. For protein identification, proteins were digested with trypsin (Promega) using the in-gel digestion pro-tocol (23). Peptides were analyzed by liquid chromatographytandem mass spectrometry using the Agilent 1100 nanoflow high pressure liquid chromatography system with Agilent XCT-Plus ion trap (Agilent Technologies), and the data base search against NCBInr was performed by Spectrum Mill software (Agilent Technologies).
Bioinformatics and Protein Modeling-Domain identity and organization of the preprozymogen forms of the Aur and SspA proteases were obtained from the Simple Modular Architecture Research Tool web server interface provided by the European Molecular Biology Laboratory at smart.embl-heidelberg.de (24). To model the penultimate intermediate in activation of zSspA, the amino acid residues HAN were first built onto the N terminus of mature V8 protease with the modeling program TURBO-FRODO (25). The residues were positioned visually, and modeled protein was subjected to energy minimization using the program MacroModel (26). The Merck molecular force field was used with water as the solvent medium, and the minimization was carried out by the molecular dynamics simulated annealing method.

RESULTS
Temporal Analysis of Zymogen Activation-The 33-kDa zSspA has a predicted 29-amino acid signal peptide followed by a small N-terminal propeptide harboring a glutamine-rich segment that is predicted to be a region of intrinsic disorder (Fig.  1). The mature protease begins at Val 69 and has a trypsin-like serine protease domain where His 119 , Asp 161 , and Ser 237 form the catalytic triad. This is followed at the C terminus by another predicted disordered segment of unknown function consisting mainly of Pro, Asp, and Asn. Because the 53-kDa zAur and 40-kDa zSspB proteins have larger 20-kDa propeptides, we could determine the activation status of each protease in culture supernatant during growth of S. aureus RN6390 (Fig. 2). There was no evidence of zAur at any time point. Mature 40-kDa Aur appeared at 3 h, reached a maximum at 5-6 h, and then began to diminish. In contrast zSspA and zSspB were prominent at 5 h and were each converted to mature protease over a span of more than 3 h. Activation of zSspA proceeded through at least two intermediates, whereas zSspB was directly converted to mature protease, confirming our previous finding that in vitro processing by SspA at Glu 177 alone was sufficient for activation (10). Therefore, secretion and maturation of Aur precedes the appearance of zSspA, which is activated by a mechanism that is distinct from zAur or zSspB.
Identification of Intermediates in zSspA Activation-The N-terminal Val 69 of mature SspA suggests that maturation is synchronous with processing at REHAN2 69 VILP. This is consistent with a report that a metalloprotease was needed for activation of V8 protease (16) and that the metalloprotease Aureolysin cleaves peptide bonds where the amide nitrogen atom is contributed by a hydrophobic amino acid (27). As there are only a few such sites within the propeptide (Fig. 1), we sought to confirm these as activation intermediates by expressing zSspA with a C-terminal His 6 tag in an sspA::erm derivative of S. aureus RN6390. When secreted proteins from a 5-h culture were immediately subjected to trichloroacetic acid precipitation followed by SDS-PAGE and Western blot, four immunoreactive polypeptides were evident (Fig. 3). Subsequent N-terminal sequencing established that the largest polypeptide corresponded to unprocessed zSspA, having been cleaved only at the predicted signal peptidase site ANA2 30 LSSKA. The two intermediate polypeptides had N-terminal sequences consistent with processing at QQTQ2 44 SSKQQ in the predicted disordered segment and at KGGN2 58 LKPLE, which complies with the specificity of Aur. The smallest isoform was processed at QRE2 66 HANV just proximal to Val 69 of mature SspA. As this was unexpected, we performed mass spectrometry analysis on native SspA purified from S. aureus culture supernatant ( Table 2). Three mass isoforms were evident in the purified FIGURE 1. Domain organization of zSspA. The signal peptide is labeled SP, and the N-terminal propeptide is shaded in a gray gradient with an internal segment of predicted intrinsic disorder shaded light gray. The amino acid sequence of the N-terminal propeptide beginning at Leu 30 is provided above the diagram with the glutamine-rich segment of predicted intrinsic disorder also shaded in light gray. The 2 69 VILP segment indicates the experimentally determined N terminus of mature SspA, which contains a trypsin-like serine protease protein family domain (Tryp_Spt; Pfam PF00089). A second region of predicted intrinsic disorder and its corresponding amino acid sequence is shown at the C terminus of the protein. Additional details are provided under "Results" and "Discussion."  SspA protein that by integration of peak areas in the mass spectrum were present in a 100:48:21 ratio. The most abundant corresponded to the expected mass of native SspA with an N-terminal Val 69 . The next most abundant appeared to be a variant of SspA that lacked two C-terminal alanine residues. The least abundant species corresponded to the expected mass of the N-terminal 66 HANV isoform that was identified by N-terminal sequencing of the His 6 -polypeptides. Therefore, this is not an artifact of expressing recombinant His 6 -tagged proteins. As His 66 is preceded by a glutamic acid, this finding alludes to the possibility of a role for autocatalytic processing.
To distinguish between Aur-dependent and -independent processing, protein was recovered from the supernatant of isogenic sspA and aur strains using the engineered C-terminal His tag to capture the intermediates on Fe-Ni 2ϩ beads. From supernatant of the sspA strain, we recovered the same polypeptides as identified by direct trichloroacetic acid precipitation and N-terminal sequencing of proteins from culture supernatant except that the presumed Aur-dependent intermediate was not evident (Fig. 3, center), suggesting that it is a transient intermediate that was processed during manipulation of the beads. Just two major polypeptides were recovered from the aur strain consisting of the unprocessed Leu 30 isoform and the Q2 44 SSKQQ isoform. To further characterize this Aur-independent processing, we purified a larger quantity of zSspA from the aur strain using a procedure that involved ammonium sulfate precipitation followed by dialysis, binding and washing on Ni-NTA agarose slurry, elution in imidazole, and desalting. The Leu 30 isoform was no longer evident in this preparation (Fig. 3,  right), and there was a cluster of polypeptides centered on the Q2 44 SSKQ isoform. Mass spectrometry identified three isoforms with approximately equal integrated peak areas ( Table  2), one of which had a mass value equivalent to the expected Q2 44 SSKQ isoform that we identified by N-terminal sequencing. The other mass values corresponded to polypeptides that would result from processing at SKQ2 48 QTPK and QRE2 66 HANV, and a minor species was predicted to have been processed at KIQ2 54 KGG (Table 2). If protein was directly captured from the supernatant using Fe-Ni 2ϩ beads, a mass value corresponding to the expected Leu 30 zSspA isoform was again observed as well as mass values corresponding to Q2 44 SSKQ and a trace of the predicted KIQ2 54 KGG isoform, but the QRE2 66 HANV isoform was not present. Therefore, in the absence of Aur, zSspA is processed at a preferred site, Gln 43 ; more slowly at Gln 47 and Gln 53 ; and infrequently at Glu 65 . This establishes that the glutamine-rich segment illustrated in Fig. 1 is the site of initial and presumably autocatalytic processing of zSspA.
Confirmation of Autocatalytic and Aur-dependent Cleavage-To determine whether the Aur-independent processing was autocatalytic, we constructed a Ser 237 3 Ala active site substitution, zSspA(S237A). When processing of zSspA(S237A) was compared with wild type zSspA in the sspA background, the Q2 44 SSKQQ isoform was not evident (Fig. 4). There was also reduced processing of Leu 30 zSspA, accumulation of the Leu 58 isoform, and a limited amount of mature Val 69 SspA. In the aur strain (Fig. 4, right panel), zSspA(S237A) remained unprocessed. When this unprocessed N-terminal Leu 30 (S237A) isoform was incubated with a mixture of the Ser 44 , Gln 48 , Lys 54 , and His 66 self-processed isoforms purified from supernatant of the aur strain, there was no obvious conversion to smaller  a The C terminus of SspA is modified by inclusion of a His 6 (H 6 ) tag, which is reflected in the heavier mass values. b Culture supernatant protein was precipitated with ammonium sulfate followed by dialysis, binding on Hi-Trap Ni 2ϩ matrix, elution, and desalting. c Protein was captured directly from 5-h culture supernatant using Fe-Ni 2ϩ beads followed by elution and immediate processing for mass spectrometry. d It was necessary to reduce the signal to noise threshold value to detect this isoform. e Absolute area is the total intensity (counts/s) integrated over the m/Z range of the ions contributing to the mass information for each isoform collected over 3 min.
polypeptides over a 2-h incubation even at a direct 1:1 ratio (data not shown). Consequently the self-processed isoforms cannot promote reciprocal intermolecular cleavage of remaining zSspA at these same sites. The initial processing of zSspA at 43 Q2S is therefore exclusively intramolecular and autocatalytic.

Processing by Aur Is Essential and Rate-limiting in Activation-
The Leu 58 isoform could only be detected by trichloroacetic acid precipitation of culture supernatant protein or by expressing zSspA(S237A). To confirm that the Leu 58 isoform was generated by Aur and to determine whether its transient nature was due to subsequent (i) rapid Aur-dependent processing at Val 69 , (ii) intramolecular processing at Glu 65 , or (iii) feedback from the first molecules of Val 69 SspA to promote intermolecular processing at Glu 65 , we compared the ability of either native SspA or Aur to process the inactive monoisoform Leu 30 zSspA(S237A). When assayed at several dilutions in the amount of native SspA starting at a direct 1:1 ratio, the Leu 30 zSspA(S237A) was efficiently processed at Glu 65 (Fig. 5A), whereas the specificity of Aur processing was dictated by the ratio of zSspA:Aur (Fig. 5B). At a 1:1 ratio, zSspA was processed to the Val 69 isoform, whereas a 10:1 ratio produced mainly Leu 58 , and at 100:1 the Leu 30 zSspA was largely unprocessed. Therefore, intermolecular processing of zSspA by mature SspA at Glu 65 is ϳ10and 100-fold more efficient than Aur-dependent processing at Leu 58 and Val 69 , respectively. This explains the accumulation of the Leu 58 isoform and its slow conversion to mature Val 69 as observed in Fig. 4 when temporal processing of zSspA(S237A) was monitored.
Comparison of His 66 and Val 69 Isoforms-For a stringent comparison of the His 66 and Val 69 isoforms of SspA, the heterogeneous mixture of self-processed polypeptides purified from aur culture supernatant was treated in vitro with native SspA, which results in their conversion to monoisoform His 66 SspA. Alternately the polypeptides were treated with native Aur for conversion to monoisoform Val 69 SspA. The in vitro processed proteins where then captured on Fe-Ni 2ϩ beads followed by washing to remove the native proteases and then elution and desalting. The resulting His 66 and Val 69 isoforms, which were indistinguishable from one another by SDS-PAGE (Fig. 5C), were then assayed relative to native SspA for their ability to convert the Leu 30 zSspA(S237A) into smaller isoforms (Fig. 6A). Native SspA and the in vitro activated Val 69 SspA both efficiently processed Leu 30 zSspA(S237A), but the in vitro processed His 66 isoform did not (Fig. 6A). The 40-kDa zSspB is also a very sensitive indicator of the status of SspA activity as we had demonstrated previously that SspA could convert zSspB to mature SspB at ratios that surpassed 500:1 zSspB:SspA (10). However, when zSspB was incubated with His 66 SspA at a 100:1 ratio, there was no obvious conversion to mature SspB, whereas the in vitro prepared Val 69 SspA facilitated its complete conversion to mature protease (Fig. 6B). The in vitro prepared His 66 SspA was also inactive when tested with the small synthetic substrate Z-Phe-Leu-Glu-pNA (Fig. 6C), whereas the in vitro activated Val 69 SspA was only 3-fold less active than native SspA.
A Model for Active Site Blocking by His 66 -The similarity of SspA to the trypsin family of proteases extends to the N-terminal Val 69 insofar as trypsin, chymotrypsin, and plasmin have hydrophobic residues at their N termini (28). In the structure of trypsin, the N-terminal Ile 16 is buried and stabilizes the active site by hydrogen bonding and salt bridges to the side chains of Asn 143 and Asp 194 , respectively (28 -30). Although the N-terminal Val 69 of SspA is located on one side of the primary specificity pocket, it was observed to form hydrogen bonds to the side chains of Thr 232 and Asn 261 (31). More importantly, the positively charged terminal NH 3 ϩ group of Val 69 was proposed  to have an important role in coordinating negatively charged glutamate-containing substrate peptides within the active site cleft. We therefore performed modeling experiments to predict how this would be influenced by the HAN tripeptide (Fig. 7). Our model predicts that the HAN motif extends toward the active site, placing His 66 adjacent to the catalytic His 119 . The OH group of the catalytic Ser 237 lies between these two His residues and forms hydrogen bonds with His 119 and His 66 . The model predicts that His 66 occludes the active site for all substrates.
Evolution of SspA Activation Mechanism-S. aureus harbors the coa gene encoding coagulase, whereas all other staphylococci belong to the coagulase-negative group typified by Staphylococcus epidermidis, a member of the commensal microflora. An sspA paralogue is also present in two S. epidermidis genomes that have been sequenced (32,33) that encodes a glutamyl endopeptidase that is 59% identical and 78% similar to SspA of S. aureus. However, in these situations, the gene is not associated with an operon. Therefore, there would be no requirement for the SspA paralogue to activate a downstream cysteine protease as in the case of the ssp operon of S. aureus. In this regard, although the mature SspA from S. epidermidis is well conserved with SspA of S. aureus (Fig. 8), this does not extend to the N-terminal propeptide. Strikingly the only conserved feature of the propeptide is the presumed metalloprotease-dependent processing site NIKP, similar to the N2 58 LKP site we identified in S. aureus. In addition to this distinct difference in the propeptide, there is also no predicted disordered segment at the C terminus. Similar observations are noted in the genomes of Staphylococcus haemolyticus and Staphylococcus saprophyticus, which also possess an sspA paralogue that is not associated with an operon (data not shown). One exception to date among the coagulase negative staphylococci is an equivalent of the ssp operon in Staphylococcus warneri (34) that duplicates the sspABC arrangement seen in S. aureus. In this situation, the SspA paralogue has an N-terminal propeptide and C-terminal disordered segment that closely mimics SspA of S. aureus (Fig. 8). An interesting difference is that the HAN motif is altered to RAN. When the 11 completed S. aureus genomes are compared with one another, 10 have HAN, but strain MRSA 252 has RAN. These observations suggest that the activation mechanism of SspA co-evolved with acquisition of sspB and sspC to comprise an operon. Specific features include a glutamine-rich disordered segment as the first site of autocatalytic processing and pressure to maintain a basic amino acid residue at position Ϫ3 of the propeptide, which would permit hydrogen bonding of an N ⑀ atom of the side chain to the side chain of the active site Ser 237 . We also note that after the first Aur-dependent processing step at Leu 58 , the remaining segment of the propeptide shows some resemblance to the trypsinogen activation peptide, which must be removed to activate trypsinogen precursor (29, 35-37).

DISCUSSION
Our study established that zSspA is activated by a unique mechanism involving sequential processing of its small N-terminal propeptide. The intermediate and final steps are strictly dependent on Aur, but maturation is initiated by intramolecu-  lar autocatalytic cleavage at Gln 43 . Although the reason for this remains to be established, it is possible that this initial processing promotes better presentation of the remaining propeptide as a substrate for Aur such that subsequent cleavage at Leu 58 and Val 69 occurs in a processive manner through a protein complex to liberate the first active molecules of Val 69 SspA, which can then promote efficient intermolecular processing of remaining zSspA at Glu 65 . This produces a penultimate His 66 isoform that remains inactive until the HAN tripeptide is removed by Aur-dependent cleaving at Val 69 . We propose that this sequential series of intramolecular and intermolecular processing events involving both autocatalysis and Aurdependent cleavage has evolved to minimize the likelihood of untimely activation of zSspA during protein secretion.
Mechanistically a number of factors may contribute to inhibition of the penultimate intermediate by the HAN motif, for which we must consider the relationship of SspA to the trypsin family of serine proteases, and the activation mechanism of trypsinogen. Although the structure of zSspA has not been solved, the trypsinogen and trypsin structures are 85% identical with the differences being restricted to the S1 substrate binding site and the oxyanion hole, which are unformed in trypsinogen, rendering the zymogen inactive (28,29,36). On removal of the short activation peptide APFDDDDK by enterokinase (35), the newly exposed N-terminal Ile 16 stabilizes four previously disordered peptide segments known as the activation domain, defined by amino acids 16 -19, 142-152, 184 -194, and 216 -223 (28). Each segment contributes one amino acid to a salt bridge and hydrogen-bonding network that stabilizes the active site. The most significant of these is a salt bridge between the terminal NH 3 ϩ group of Ile 16 and the side chain of Asp 194 , which is adjacent to the active site serine. This is strengthened by hydrogen bonding of Asn 143 to the terminal NH 3 ϩ group of Ile 16 . Similar interactions occur in SspA where the terminal NH 3 ϩ group of Val 69 hydrogen bonds to side chains of Thr 232 and Asn 261 (31); hydrophobic interactions also occur between the Val 69 side chain and the hydrocarbon moiety of Pro 257 . These interactions have two important implications with respect to the maintenance of an inactive zymogen. First, it is apparent that addition of even a single amino acid to the N-terminal Val 69 could perturb these forces that stabilize the active site and constitute the primary specificity pocket. Second, when the HAN motif of the penultimate intermediate is exposed, it is likely that its conformational mobility is quickly restricted due to the proposed hydrogen bonding of His 66 to the active site Ser 237 and the observed hydrophobic interaction of the Val 69 side chain with Pro 257 . Potentially this limits the surface exposure of the 66 HAN motif, ensuring that Aur-dependent processing must take place slowly over a period of several hours as suggested in Fig. 4 where autocatalytic processing is negated by the S237A substitution.
Because zSspA undergoes autocatalytic cleavage at glutamine in the absence of Aur, its active site must be structured. Therefore, although a trypsinogen-like activation mechanism predicts that a newly exposed N-terminal hydrophobic amino acid (Val 69 ) would play an essential role in bringing order to the active site, our present data suggest that this is not so for SspA. Rather and as previously predicted (31), the N-terminal Val 69 seems to function primarily as a scaffold for placing a terminal NH 3 ϩ group in optimal proximity to the active site where it helps to neutralize the negative charge of glutamate-containing substrate peptides. This would not be a requirement for glutamine-containing peptides, and molecular modeling of the QTQSS propeptide segment showed that Gln 43 would fit nicely into the primary specificity pocket of zSspA and form a hydrogen bond with the side chain of Thr 232 and Asn 261 (Fig. 9). Our data can be explained by intramolecular cleavage where the propeptide threads through the active site. The glutamine-rich segment that is susceptible to autocatalytic processing is also a region of predicted intrinsic disorder, which would facilitate its ability to thread through the active site. Glutamate would not bind in the primary specificity pocket of zSspA because this would bury a negative charge and would be energetically unfavorable. This supports another mechanism for maintaining inactivity of zSspA, namely that additional amino acids beyond Val 69 will occlude all substrates from the active site.
A third explanation is supported by the evolutionary conservation of a basic amino acid at the Ϫ3 position of the propeptide sequence, either histidine or arginine. Our modeling data indicate that the side chain of His 66 will hydrogen bond to the catalytic OH group of Ser 237 and occlude the active site. A related observation was noted in high resolution structures of the extracellular lipase BsL of Bacillus subtilis that has the same catalytic triad as trypsin (38). Two different structural conformers were observed. In one, the O ␥ of the catalytic Ser 77 correctly formed a hydrogen bond with the N ⑀ atom of the catalytic His 155 , but in another, it formed a hydrogen bond to the sidechain atom of His 76 , which is adjacent to His 155 in the structure. It was proposed that this served to modulate the transition of the active site between active and inactive conformations. The His 66 isoform of zSspA as well as strains with the Arg 66 variant would have side chains containing an N ⑀ atom that hydrogen bonds to the hydroxyl group of the active Ser 237 side chain (Fig.  8), thus blocking the primary specificity pocket.
Contrary to this reasoning, it was reported that the mature SspA paralogue of S. warneri was active as the RANVILP iso- form (34). However, this violates the principals of the trypsin family of serine proteases, which have a hydrophobic amino acid at the N terminus. The apparent activity of the RAN isoform in S. warneri could be due to the presence of the VILP isoform. Using mass spectrometry, we noted that native SspA purified from culture supernatant after 18 h of growth contains a significant amount of the HANVILP isoform (Table 2), which cannot be resolved from the mature Val 69 SspA by SDS-PAGE or chromatography. The N-terminal sequencing used to define the S. warneri isoforms may not have been sensitive enough to detect minor amounts of the active VILP isoform.
As it is evident that just three additional amino acids are sufficient to keep SspA inert, this brings into question the function of the remainder of the zSspA propeptide. Although not a recognized classification scheme, we may assign microbial serine proteases into three groups based on their propeptides: (i) those with large propeptides of known function, (ii) those with no N-terminal propeptides, and (iii) proteases with minimal propeptides. It appears that zSspA does not fit into any of these. In the first group, subtilisin of B. subtilis has an 83-amino acid propeptide that acts as an intramolecular chaperone to promote folding of the protease domain and to maintain the protease as an inactive precursor during protein export (39,40). The propeptide alone is intrinsically disordered but when mixed with mature protease adopts a flower-like structure with its stem inserted into the active site (41). This dual inhibitor/ chaperone paradigm is maintained by the larger 174-amino acid propeptide of ␣-lytic protease expressed by Lysobacter enzymogenes even though it has no obvious sequence or structural similarity to the subtilisin propeptide (42,43). This propeptide defines the Pro_Al_Protease protein family domain (Pfam 02983), which is also found in serine proteases of the Actinomycetales family of soil bacteria, including Thermomonospora, Rarobacter, and several proteases of Streptomyces (44). Although this latter group includes a glutamic acidspecific protease, SgpE (45), the small N-terminal propeptide of SspA has no obvious similarity to this or any other propeptide in this family.
Proteases with no propeptides include the exfoliative toxins of S. aureus and several serine protease-like (Spl) proteins also produced by S. aureus for which the structures of each can be superposed on the C ␣ trace of the SspA/V8 protease (31, 46 -48). The exfoliative toxins are glutamyl endopeptidases that primarily cleave a single host protein, desmoglein (49), whereas the Spl proteins are encoded by tandem gene repeats in a genomic island structure and are either very restricted in their substrate specificity or could not be demonstrated to have protease activity (46,50). Presumably because they are restricted in their substrate specificity, there is no need for an intramolecular inhibitor. For the SplC protein, structural determination revealed that the active site is well ordered, but an essential catalytic histidine residue is pushed away from the active site by conflict with glycine residues in a flexible loop that passes over the active site. It was proposed that contact with substrate containing peptides would trigger a conformational change of the flexible loop, allowing the catalytic histidine to rotate into the active site. Studies with recombinant Spl proteins also supported the contention that even a single additional amino acid at the N terminus would be sufficient to maintain zymogen status (46).
The third family of serine proteases in this scheme is also represented in the soil bacterium Streptomyces griseus, which in addition to the Pro_Al_Protease family expresses two similar trypsin proteases, SprT and SprU (51), that have only a fouramino acid propeptide known as the activation peptide that is removed by processing at APNP2VVG or APQP2IVG. It is notable that the underlined N terminus of these short propeptides is identical to the N terminus of the trypsinogen activation peptide APFDDDDK2IVG (35) and also that SprT, SprU, and SspA have a common property of two or three hydrophobic amino acids at the N terminus of the mature protease that is shared among the trypsin family (28). Intriguingly as shown in Fig. 8, the first Aur-dependent processing site in maturation of zSspA exposes a new N-terminal segment, LKPLEQREHAN2VI, where the underlined residues are similar to the trypsinogen activation peptide (Fig. 8).
Although trypsinogen activation strictly depends on enterokinase cleaving at Lys-2-Ile 16 , the first molecules of active trypsin then process the same site in remaining trypsinogen through intermolecular cleavage. In a variation of this theme, activation of zSspA strictly depends on Aur processing at Leu 58 and then Val 69 , but this occurs less effectively unless preceded by initial intramolecular processing at Gln 44 . Autocatalytic intermolecular processing at Glu 65 also occurs after the first molecules of Val 69 SspA are produced by Aur. Consequently as with trypsinogen, activation of zSspA is strictly dependent on another protease but is also assisted by autocatalytic processing.
It is apparent now that activation of zSspA most closely resembles trypsinogen with some variations. Notably although the zSspA propeptide has some similarity to the trypsinogen activation peptide, additional features place it in a class of its own. These supplementary features seem to function as a molecular fuse, which must be "burned" or processed in proper sequence for effective activation. These features together with a repetitive disordered segment of unknown function at the C terminus of zSspA are uniquely associated with S. aureus and one species of coagulase-negative Staphylococcus where SspA is in an operon and serves to activate an adjacently encoded cysteine protease. Others have found that expression and secretion of zSspB is toxic to S. aureus unless accompanied a third gene in the ssp operon encoding a cytoplasmic protein, SspC (Staphostatin B), that binds to and inhibits mature SspB but reportedly has no interaction with zSspB (12,14,15). Our present study has described additional safety features incorporated into the zSspA propeptide that appear to have evolved to minimize the likelihood of either protease being activated prematurely during protein export.