Crystal Structures of Staphylococcus aureus Sortase A and Its Substrate Complex*

The cell wall envelope of staphylococci and other Gram-positive pathogens is coated with surface proteins that interact with human host tissues. Surface proteins of Staphylococcus aureus are covalently linked to the cell wall envelope by a mechanism requiring C-terminal sorting signals with an LP X TG motif. Sortase (SrtA) cleaves surface proteins between the threonine (T) and the glycine (G) of the LP X TG motif and catalyzes the formation of an amide bond between threonine at the C-terminal end of polypeptides and cell wall cross-bridges. The active site architecture and catalytic mechanism of sortase A has hitherto not been revealed. Here we present the crystal structures of native SrtA, of an active site mutant of SrtA, and of the mutant SrtA com-plexed with its substrate LPETG peptide and describe the substrate binding pocket of the enzyme. Highly conserved proline (P) and threonine (T) residues of the LP X TG motif are held in position by hydrophobic con-tacts, whereas the glutamic acid residue (E) at the X position points out into the solvent. The scissile T-G peptide bond is positioned between the active site Cys 184 and Arg 197 residues and at a greater distance from the imidazolium side chain of His 120 . All three residues, His 120 , Cys 184 , and Arg 197 , are conserved in sortase enzymes

Many bacteria adhere to host extracellular matrix proteins as an essential first step toward the pathogenesis of infection (1). In Gram-positive bacteria, the cell wall envelope serves as a surface organelle with immobilized surface proteins that are responsible for mediating adhesion to host tissues (2). Many of these surface proteins are covalently linked to the cell wall peptidoglycan by a mechanism requiring a C-terminal sorting signal with the conserved LPXTG motif (where X represents any amino acid) (3). Sortase A, a transpeptidase with an active site cysteine, cleaves the sorting signal between the threonine and the glycine of the LPXTG motif (4 -7). Sortase catalyzes the formation of an amide bond between the carboxyl group of threonine and the amino group of the cell wall cross-bridge, a pentaglycyl moiety in staphylococci (8,9). Several observations suggest that lipid II, a membrane anchored intermediate of cell wall synthesis, functions as the peptidoglycan substrate of sortase A (10,11). The product of the complete sorting reaction, surface protein tethered to lipid II, is presumed to be incorporated into the cell wall envelope via the penicillin-sensitive transpeptidation and transglycosylation reactions of peptidoglycan synthesis (10,12).
Sortase A (SrtA) is a polypeptide of 206 amino acids with an N-terminal membrane-spanning region and a C-terminal catalytic domain (4,14). Mutant staphylococci harboring a deletion of the srtA gene accumulate surface protein precursor molecules with C-terminal sorting signals in the membrane compartment (5). Although ⌬(srtA) staphylococci grow on laboratory media similar to the wild-type strain, the sortase mutants display severe defects in the pathogenesis of animal infections (16,17). Genes homologous to Staphylococcus aureus srtA are found in all Gram-positive bacterial genomes (18). Considerable evidence has now accumulated that the inactivation of sortase genes interferes with the anchoring and the surface display of distinct sets of proteins (defined by their sorting signals), thereby reducing the adhesiveness and virulence of bacterial pathogens (19 -26). Because of the central role of sortases in the functional assembly of the cell wall envelope and in bacterial pathogenicity, sortases have been acknowledged as a target for the development of therapeutic agents that may disrupt human infections caused by Grampositive bacteria (27).
A single conserved cysteine (Cys 184 ) in SrtA is absolutely essential for its activity (6). Substitution of Cys 184 with Ala or addition of cysteine reactive reagents such as methyl methanethiosulfonate and p-hydroxymercury benzoate abolish the activity of the enzyme, suggesting that SrtA is a cysteine protease (6, 28 -30). The cleavage of the LPXTG motif between the threonine (T) and glycine (G) residues by SrtA leads to the formation of a covalent thioester bond between the thiol group of the enzyme cysteine and the carboxyl group of the substrate threonine residue, and this transient acyl-enzyme intermediate is subsequently resolved by the nucleophilic attack of the amino group of pentaglycyl (6,13). A new amide bond is then formed between the threonine and the N-terminal glycine residue of the peptidoglycan, resulting in a substrate protein covalently linked to the peptidoglycan (8).
SrtA ⌬N59 , a recombinant sortase A lacking the N-terminal 59 residues, is fully functional in vitro and catalyzes the cleavage of peptides bearing an LPXTG motif as well as the transpeptidation reaction with pentaglycine substrate (29,31). Previous determination of the three-dimensional structure of SrtA ⌬N59 using NMR revealed three residues, Cys 184 together with His 120 and Asn 98 , arranged in a disposition similar to that of the catalytic triad observed in classic cysteine proteases such as papain (31). Because a thiolate-imidazolium ion pair is structurally and functionally conserved among the superfamily of cysteine proteases, proximally positioned Cys 184 and His 120 residues in SrtA were thought to act as the catalytic ion pair (30,31). However, the thiol group of Cys 184 was seen pointing away from the imidazole ring of His 120 and the distance between the two side chains was much farther than that observed in a typical catalytic cysteine-histidine ion pair. In addition, the side chain of His 120 , deeply buried under a hydrophobic cluster, may not be accessible to a potential substrate. A recent study estimated the pK a of the active site Cys 184 thiol and His 120 imidazole ring to be around 9.4 and 7.0, respectively, suggesting that SrtA does not form a thiolate-imidazolium ion pair in its active site (32). Hence, neither the actual disposition of the active site nor the catalytic apparatus of SrtA is perfectly understood. In addition, the crystal structure of S. aureus sortase B, a cysteine transpeptidase responsible for anchoring surface proteins with C-terminal NPQTN sorting motifs to the cell wall, had raised the possibility of sortases utilizing a Cys-Arg catalytic dyad for catalysis. However, these studies left unresolved the details of the substrate binding pocket that may enable S. aureus sortases to distinguish between LPETG and NPQTN sorting signals (33). Herein we sought to characterize the substrate binding pocket and the catalytic mechanism of sortase A by determining the three-dimensional crystal structures of SrtA ⌬N59 , its active site mutant (C184A) SrtA ⌬N59 , where we mutated the active nucleophilic residue Cys 184 to an alanine, and a complex of the mutant and its peptide substrate (SrtA ⌬N59Cys184Ala ϩ LPETG peptide). Furthermore, by comparing S. aureus SrtA ⌬N59 and sortase B crystal structures (33), we sought to define the substrate binding pocket of both sortases and their catalytic apparatus, and propose a catalytic mechanism that may be universal for sortases.

EXPERIMENTAL PROCEDURES
Protein Purification and Crystallization-The recombinant enzyme SrtA ⌬N59 and its variant C184A were cloned and expressed in Escherichia coli as described previously (6). The truncation of the N-terminal 59 amino acids did not affect the activity of the enzyme (29, 31), but enhanced the solubility and helped the crystallization. All crystals were grown by the hanging-drop technique with a protein concentration of 50 mg/ml in 25 mM MES, 1 pH 6.35. The crystallization condition includes 3.2 M ammonium sulfate, 0.1 M NaCl, and trace amounts of ethylene glycol. A single SrtA ⌬N59 crystal was obtained after 4 weeks. The recombinant SrtA ⌬N59 enzyme was confirmed to be active at crystallization, pH 6.35 (data not shown). The crystal was harvested and ground in its crystallization solution to be used for later seeding. All other crystals, including SrtA ⌬N59Cys184Ala , were obtained by the micro-seeding technique. Selenomethoinine-substituted SrtA ⌬N59Cys184Ala was expressed using an established protocol (35), and the corresponding crystals were obtained by microseeding with unlabeled SrtA ⌬N59 crystals to perform a selenium MAD phasing experiment. High performance liquid chromatography purified LPETG peptide was obtained from the University of Alabama core peptide synthesis facility. The SrtA ⌬N59Cys184Ala -peptide complex crystals were obtained by soaking the peptide into the crystals under similar crystallization conditions. X-ray Diffraction Data Collection-Some native SrtA ⌬N59 diffraction data were recorded using in-house x-ray radiation, equipped with a RAXIS-IV detector. Diffraction data to 2.3-Å resolution were collected using Seleno-Met crystals of SrtA ⌬N59Cys184Ala at 100 K on the X4-A beamline of NSLS. Standard three wavelength selenium MAD data were collected and processed with HKL2000 and the SCALEPACK suit of programs (36). A complete diffraction data set for the SrtA ⌬N59Cys184Ala -LPETG complex was also collected at 100 K using the APS 19BM beamline. However, the crystal structure of SrtA ⌬N59Cys184Ala was determined by the single wavelength anomalous dispersion method using only the "peak data" collected at 0.9792 Å. Identification of selenium atom positions, refinement, and subsequent phasing were done with the help of CNS and SOLVE programs (37). The crystal structure of SrtA ⌬N59 was determined by molecular replacement methods. Model building was performed with the help of the "O" program (38) and refined with the help of CNS. The data collection and structural refinement statistics are shown in Table I.

RESULTS
Structure of SrtA ⌬N59 -The crystals of SrtA ⌬N59 and SrtA ⌬N59Cys184Ala belong to the P2 1 space group and there are 3 molecules (A, B, and C) in the crystallographic asymmetric unit. We were unsuccessful in solving the SrtA ⌬N59 crystal structure by molecular replacement methods using the NMR structures (Protein Data Bank code 1IJA) as starting models. Although 3-wavelength selenium MAD phasing data on a seleno-Met SrtA ⌬N59Cys184Ala crystal were collected, because of unexpected scaling problems between the three data sets, the single wavelength anomalous dispersion phasing technique was utilized to solve the crystal structure by using the peak wavelength data alone. A total of 6 sites of selenium were found using the CNS program (2 sites for each molecule) (37). The electron density map, calculated to 2.3 Å, was readily interpretable once density modification procedures were completed. In addition to the SrtA ⌬N59Cys184Ala crystal structure, the crystal structures of native SrtA ⌬N59 and the SrtA ⌬N59Cys184Ala ϩ LPETG peptide complex were determined by molecular replacement methods and refined to 2.0-and 1.7-Å resolutions, respectively. The core of the SrtA ⌬N59 crystal structure is an 8-strand ␤-barrel. One ␣ helix and two 3-turn helices connect the ␤ strands, forming a novel -fold ( Fig. 1). A DALI search confirmed the absence of homologues proteins in the protein data bank. The three molecules (A, B, and C) present in the asymmetric unit are not related by any identifiable non-crystallographic symmetry. Two N-terminal residues in molecule A, 4 in B, and 3 in C have no electron density in 2F o Ϫ F c maps, indicating N-terminal flexibility. The r.m.s. deviations between molecules A, B, and C are less than 0.57 Å. The major difference among the three molecules is in the conformations of the loop connecting the ␤6 and ␤7 strands (depicted in red color, Fig. 1). One side of the ␤-barrel, formed by the ␤4, ␤7, and ␤8 strands, is concave in appearance and, along with three of the surrounding loops, forms a tunnel-like hydrophobic pocket in the center of which the catalytic Cys 184 is located. The side chain of Cys 184 is pointed out into the open region. However, access to this pocket is blocked in the case of molecules B and C by the loop regions of symmetry related molecules, whereas the hydrophobic pocket of the active site Cys residue in molecule A has clear access from the solvent.
Superimposition of the SrtA ⌬N59 crystal structure with one of its NMR counterparts indicates that the core ␤-barrel structure seems to be subtly different, with an r.m.s. deviation of ϳ1.95 Å for all the C␣ atoms. These observations may explain our failure to phase the diffraction data by molecular replacement methods using NMR structures as search models. Not surprisingly, many differences are also seen in the side chain conformations between the crystal and NMR structures in the core region as well as in peripheral regions of the molecule, and these differences may be an indication of the low resolution nature of NMR structures.
Comparison with Other Cysteine Proteases-The SrtA from S. aureus was the first cysteine transpeptidase to be identified (6). The initial steps in the transpeptidation reaction include the cleavage of the scissile bond of a substrate and the formation of an acyl-enzyme intermediate, steps that are very similar to those seen in reactions catalyzed by all cysteine proteases (6,30). Comparison of SrtA with other cysteine proteases would benefit our understanding of the mechanism of this enzyme. Cys 184 and His 120 are absolutely conserved in all sortase enzymes and were shown to be essential for SrtA catalysis (30,31). These observations provide the basis for a mechanistic model whereby SrtA may utilize a thiolate-imidazolium ion pair for function (30). A third residue, Trp 194 , was also implicated in assisting in the formation of a thiolate-imidazolium ion pair (30).
When examined for the papain superfamily proteases, the catalytic cysteine and histidine residues are typically positioned on loops at the ends of a ␤-strand and an ␣ helix, respectively, and are located in two separate domains (40,41). The substrate-binding site is a cleft between the two domains. Both catalytic residues, Cys 25 and His 295 , of papain are well exposed into the cleft and capable of accessing the scissile amide bond, whereas the amide nitrogen of the catalytic Cys 25 and a neighboring residue Gln 19 side chain cooperate to function as the "oxyanion hole" and stabilize the acylated intermediate ( Fig. 2A). In the crystal structure of SrtA ⌬N59 , Cys 184 and His 120 are anchored on two neighboring ␤-strands of a concave ␤-sheet, which inherently lacks flexibility (Fig. 2B). However, the geometric disposition of the putative catalytic residues, Cys, His, and Asn/Asp residues between papain, sortase A, and sortase B (Fig. 2C), is similar (31,33). Unlike papain, the thiol group of Cys 184 in SrtA ⌬N59 is positioned 7 Å away from the imidazole ring of His 120 , and this distance could not be shortened to less than 4.8 Å even if the closest rotamers for both residues were chosen. Short of a substantial structural rearrangement, Cys 184 and His 120 could not be positioned in close proximity to form a Cys-His ion pair. Additional support for the notion that a Cys-His ion pair may not be formed in the active site was obtained by the recent pK a measurements of the active site Cys 184 thiol and His 120 imidazole ring to be around 9.4 and 7.0, respectively (32).
In conventional cysteine and serine proteases, the catalytic histidine residue plays an important role in facilitating the highly reactive nucleophilic Cys/Ser attack by protonating the amide atom of the substrate scissile bond (42,43). However, in some special cases, other ionizable residues such as lysine and tyrosine are known to assume this role (44,45). Close inspection of the active site Cys 184 residue surroundings in the SrtA ⌬N59 crystal structure revealed several ionizable groups in addition to His 120 . For example, in a radius of 12 Å, polar and charged residues Thr 183 , Asp 185 , Lys 196 , Arg 197 , Glu 105 , Ser 116 , Thr 180 , and Try 187 can be observed. Among them, the residues Arg 197 (6.7 Å), Glu 105 (11.6 Å), Ser 116 (10.0 Å), Thr 180 (10.5 Å), and Tyr 187 (10.2 Å) are on the Cys 184 side of the ␤ sheet, whereas others are pointing either into the ␤-sandwich core or away to the solvent region. Under conditions of substrate binding and within a scenario where only residue side chain movement was allowed (considering that the putative catalytic residues are anchored onto ␤-strands within a rigid ␤-sheet), we could identify a single ionizable group, i.e. the side chain guanidino group of Arg 197 , that could be moved in proximity of Cys 184 (ϳ3.5 Å) to form a catalytic ion pair (42).
The Complex Structure of SrtA ⌬N59Cys184Ala ϩ Peptide-Recombinant SrtA ⌬N59 and full-length SrtA were shown to cleave LPETG-bearing peptides with equal efficiency in vitro (6). The co-crystallization of wild type SrtA ⌬N59 and LPETG peptide was not entirely successful, only broken and patchy electron density around the thiol group of Cys 184 was found in 2F o Ϫ F c and F o Ϫ F c electron density maps. Our failure to observe the complete acyl-enzyme intermediate (covalently bound LPET peptide adduct) could be because of the high concentration of ammonium sulfate (3.25 M) used during crystallization and the possible existence of ammonium ions acting as a nucleophile to release the bound peptide adduct. It also could be because of the slow hydrolysis of the peptide adduct, as sortase has been shown to catalyze a hydrolysis reaction in vitro even in the absence of the peptidoglycan nucleophile albeit very slowly (29,46). LPETG peptide was diffused overnight into the crystals of SrtA ⌬N59Cys184Ala , an active site mutant whose crystals are isomorphous with those of SrtA ⌬N59 . Clear electron density was observed extending along the concave ␤ sheet in the proximity of C184A in the active site of molecule C in the asymmetric unit (Fig. 3A), but no such density was observed for molecules A and B. This discrepancy is mainly because of crystal-packing arrangements, as the peptide-binding sites of molecule A and B are blocked by the neighboring loop segments of the molecule. The peptide-binding site, seen in a surface plot (Fig. 3C), is in a deep cleft region flanked by two loops. The N-terminal part of the peptide is positioned toward the entrance of the cleft, whereas its C-terminal end resides close to the C184A (Fig.  3B). The conserved Leu and Pro residues of the ligand peptide (4) have settled in the highly hydrophobic surroundings formed by the residues present on the floor of the concave ␤ sheet and the two contributing loop segments. The side chain of the least conserved third residue in the LPXTG motif, the Glu residue of the substrate (4), is observed pointing outwards into the solvent. Most significantly, the scissile peptide bond between the C-terminal Thr and Gly residues of the ligand peptide is suitably positioned between C184A and Arg 197 , far from the His 120 side chain. Interestingly, the carbonyl oxygen of Thr is pointing toward the Arg 197 guanidino group N-⑀ atom, and the separation between them, much shorter for hydrogen bonding distance, is in the same range as observed for the scissile carbonyl oxygen and Gln 19 (a member of the transition state oxyanion hole) in the papain active site-substrate complex (42).
The binding of the peptide substrate did not re-arrange the ␤ structure of the enzyme. The His 120 side chain is at a distance of 11 Å from the substrate scissile peptide bond. A cluster of hydrophobic residues, including Trp 194 and Leu 97 , block His 120 access to the substrate (Fig. 4A). Based on this evidence, His 120 can be ruled out as a potential protonator/deprotonator during catalysis. However, the mutagenesis study indicated that His 120 was essential for enzymatic activity (30), suggesting that it may be playing a different role than that of a catalytic residue. Studies on cysteine protease papain suggest that the

FIG. 2. Comparison of Cys-His pairs in papain and SrtA ⌬N59 . A,
in SrtA ⌬N59 , Cys 184 and His 120 are anchored to neighboring strands of a ␤ sheet, which lacks flexibility. The imidazole ring of His 120 is buried in a hydrophobic cluster and the distance between it and the thiol group of Cys 184 is about 7 Å. The guanidino group of Arg 197 is the closest ionizable group to Cys 184 . B, in papain, the catalytic cysteine and histidine are from two domains and they are exposed to the interdomain cleft, where substrate binds. The distance between the two groups is about 4 Å. C, the geometric arrangement of putative catalytic residues in sortases and other cysteine proteases. Catalytic Cys-His-Asp/Asn residue triads from papain (green), staphopain (yellow), sortase A (cyan), and sortase B (magenta) structures are superposed revealing their identical geometrical arrangement and relative disposition. competence of the nucleophilic cysteine is not guaranteed by the existence of the Cys/His ion pair alone (48). The surrounding charged and polar groups that contribute to the resultant electric potential in the active site are critical for the catalytic competence of the nucleophile (48,49). Several charged and/or polar residues lying within a radius of 10 Å around the thiol group of Cys 184 , of which His 120 is one, may play a vital role in contributing to such an electrostatic environment. The possibility that the bulky and buried imidazole ring of His 120 may be important for the molecular folding and hydrophobic packing arrangement has been addressed as large structural rearrangements of the mutant enzyme could not be observed. However, mutation of His 120 to Ala may cause subtle changes in the architecture of the active site that could be responsible for the observed loss of the enzymatic activity.
In the complex structure of SrtA ⌬N59Cys184ALa ϩ LPETG, the only residue suitable to replace His 120 as a catalytic protonator/ deprotonator is the Arg 197 residue. Arg 197 is the closest ionizable group to the T-G bond of the substrate (Fig. 4A). With a cumulative positive charge, the guanidino group of Arg 197 might also serve as a potential oxyanion hole in stabilizing the acylated adduct. Interestingly, the side chain of Arg 197 has a stronger and clearer electron density in the (SrtA ⌬N59Cys184ALa ϩ LPETG) crystal structure and is in hydrogen bonding distance of the substrate peptide threonine carbonyl oxygen. This conformation allows the guanidino group to interact with the peptide bond directly, and the flexibility and mobility of the guanidino group may be important for its potential role during catalysis. Moreover, the Arg 197 residue is absolutely conserved among all known sortases from various Gram-positive bacteria (33). On the other hand, the calculated pK a of the Arg residue is normally around 12.0, which is high compared with that of histidine (6.5) and lysine (10.0). The high pK a value of the Arg residue would pose a significant barrier to its transformation between ionization states, a mandatory step during catalysis. However, the physiological pK a of Arg 197 could be quite different from the calculated value for a free-standing residue (51), as neighboring hydrophobic environments can facilitate the lowering of pK a thereby facilitating a transition between both states. The suggestion that Arg 197 may be directly involved in the catalytic reaction of sortases is also based on the side chain mobility and proximity to the substrate scissile bond; however, further kinetic and mutational studies will be needed before its exact role in catalysis can be fully understood. Replacement of the Ala side chain with a Cys thiol group for the C184A residue reduced the distance between the thiol group and the carbonyl carbon of the scissile peptide bond in the substrate to about 3.8 Å (Fig. 4B), a configuration that is comparable with that seen in papain and other cysteine protease superfamily catalyzed reactions (42,52). In the wild type enzyme, the thiol group of Cys 184 and the guanidino group of Arg 197 might interact with the substrate from two different directions. In contrast, in the papain family of enzymes, the Cys-His ion pair is separated by about 3 to 4 Å and present on the same side of the scissile bond and the nucleophilic attack is thought to occur from only one direction. In sortases, the Cys-Arg ion pair brackets the scissile bond, analogous to the Cys-His ion pair bracketing the scissile bond in the cysteine protease family of caspases (38,57), implicating an identical role for Arg as a Cys catalytic partner in sortases (50). However, it is equally plausible that the nucleophilicity of the Cys 184 thiol group in sortases may exist independently of a Cys-Arg ion pair and, in light of its terminal guanidinium groups positive electrostatic potential, Arg 197 could play both roles of protonating the substrate amide bond N atom, facilitating the nucleophilic attack by the active Cys 184 thiol group, and function as the oxyanion hole in stabilizing the transition state or intermediate in both directions of the transpeptidase reaction.
Comparison of the SrtA and SrtB Active Sites-S. aureus encodes two sortases, SrtA and SrtB. SrtB recognizes the NPQTN sorting motif and is responsible for anchoring proteins related to iron acquisition (5,6). Crystal structures of native SrtB ⌬N39 (Protein Data Bank code 1NG5) along with those of SrtB ⌬N39 in complex with two inhibitors (Protein Data Bank codes 1QWZ and 1QX6) and one substrate (Protein Data Bank code 1QXA), have been elucidated (33). Although the sequence homology between the two enzymes is only about 40% (23% identity), their core structures are very similar (r.m.s. deviation 1.55 Å for the ␤ core structures) (Fig. 5A). The main differences come from loop regions connecting the ␤ strands and a number of extra helical segments observed in the SrtB structure. The loop connecting the ␤7 and ␤8 strands differs in length; however, the relative position of the active site Cys, His, and Arg are spatially conserved (28). Even though the main chain positions are identical, the putative catalytic Cys 223 and Arg 233 residues, along with the Glu 224 residue seen holding the Arg 233 residue side chain in the correct orientation, exhibit entirely different conformations in the native SrtB ⌬N39 and in the inhibitor-bound SrtB crystal structures. In the native SrtB ⌬N39 crystal structure, the Arg 233 side chain points away from Cys 223 , whereas the Glu 224 and Arg 233 pair point away from each other, completely different from that observed in inhibitor-bound SrtB ⌬N39 structures suggesting a residue reorientation following the acylation step. Similarly pronounced variations for Cys 184 and Arg 197 are not seen between native SrtA ⌬N59 and substrate-bound SrtA ⌬N59 crystal structures, which might be an indication of possible differences in catalytic rate and efficiencies between the two enzymes.
Although their substrates are very similar, SrtA exhibits no activity toward peptides containing NPQTN sorting motifs in vitro and SrtB lacks the ability to process substrates with LPXTG sorting signals in vivo and in vitro (6). The two sorting motifs, LPETG and NPQTN, are homologous in size and both conserve the critical N-terminal Pro and C-terminal Thr residues. Superimposition of the apo-SrtB structure onto the SrtA ⌬N59Cys184Ala -LPETG complex revealed that the position of the N-terminal part of the LPETG peptide interfered with the position of the ␤7-␤8 loop of SrtB ⌬N39 (Fig. 5). In SrtA ⌬N59 , hydrophobic residues Val 168 and Leu 169 from the ␤6 -␤7 loop provide hydrophobic interactions with the N-terminal Leu residue of the LPETG peptide. However, the corresponding loop region in SrtB ⌬N39 is comprised of Asn 180 , Tyr 181 , Ile 182 , and Arg 183 , which might create polar interactions with the N-terminal Asn residue of NPQTN substrate. For the C-terminal end of the LPETG peptide, the active site of SrtA ⌬N59 apparently can only accommodate a small residue, in this case the glycine residue of the substrate peptide; in contrast, the active site of SrtB ⌬N39 around Cys 223 is less hydrophobic in nature and is surrounded by polar residues such as Tyr 128 and Asn 92 that might help stabilize the C-terminal end of NPQTN.
Molecular Modeling of the Second Substrate of SrtA-In the second step of the transpeptidation reaction catalyzed by sortases, the transient acyl-enzyme intermediate, formed in the first step, is resolved by the nucleophilic attack of (Gly) 3 , thereby releasing substrate peptide from both SrtA and SrtB (14). Based on the crystal structure of the SrtB ⌬N39 ϩ MTSET ϩ (Gly) 3 complex (33), we constructed a model for a SrtA ⌬N59 ϩ (Gly) 3 complex (Fig. 6). The N-terminal of (Gly) 3 can indeed be positioned in close proximity to catalytic residues Cys 184 and Arg 197 and the entire peptidogylcan peptide may be captured by the loop region between ␤7 and ␤8.
Interestingly, a conserved water molecule is found in close proximity to the putative catalytic Cys-Arg pair in all three molecules in the asymmetric unit of SrtA ⌬N59 , SrtA ⌬N59Cys184Ala , or the SrtA ⌬N59Cys184Ala -LPETG complex (Fig. 6). This water molecule, held by hydrogen bonds to the backbone atoms of the loop connecting strands ␤7 and ␤8, is about 5.6 Å from the thiol group of Cys 184 and 4.1 Å from the N-1 atom of the guanidino group of Arg 197 . We believe that this water molecule is appropriately positioned to be replaced by the second substrate (Gly) 3 peptide, during the second half of the transpeptidation reaction. A similar water molecule was also found in the active site of SrtB ⌬N39 crystal structures and is replaced by the second substrate (Gly) 3 in the crystal structure of SrtB ⌬N39 and the (Gly) 3 complex (33) . We can reasonably assume that this conserved water molecule might facilitate the hydrolysis of the peptide substrate in the absence of a polyglycine nucleophile in vitro.

DISCUSSION
In a recent report (33,50), we presented high resolution S. aureus SrtB inhibitor complex crystal structures and suggested the possibility that bacterial sortases, cysteine transpeptidases responsible for covalently linking surface proteins to the bacterial cell wall, utilize a novel and unique Cys-Arg catalytic dyad for transpeptidation. However, the crystal structures of covalently bound irreversible inhibitor enzyme complexes may not adequately define the catalytic apparatus or the substrate-binding site of the enzyme. With a novel but identical structural fold compared with all other cysteine proteases, it is interesting to note that S. aureus SrtA and SrtB recognize almost identical, but subtly different, sorting motifs LPETG and NPQTN, respectively. In this report, we have addressed the question how sortases may distinguish their substrates by analyzing the three-dimensional structures of enzyme substrate complexes.
With the help of the SrtA ⌬N59Cys184Ala ϩ LPETG peptide complex crystal structure, we provide the supporting structural evidence for the recent observation regarding the absence of the thiolate-imidazolium ion pair S. aureus sortase A by Connollay et al. (32), and reveal that the conserved His residue in the active site of sortases is not suitably positioned to play the conventional protonator/deprotonator role in a peptide transfer reaction. In addition, the crystal structures of SrtA ⌬N59Cys184Ala ϩ LPETG confirmed the previous observation from the SrtB ⌬N30 ϩ E64 inhibitor complex (33, 50), a possible role for the conserved Arg residue in facilitating the stabilization of the acylated adduct. Kinetic data on S. aureus sortase A suggest that the enzyme functions as a hydrolase in the absence of nucleophilic peptidoglycan or its mimics (29,46). To facilitate such hydrolysis, a solvent molecule of moderate occupancy is conserved in the vicinity of the putative catalytic residues Cys and Arg in both the crystal structures of SrtA and SrtB. It has been suggested that sortases catalyze the transpeptidation reaction using a "ping-pong" mechanism, whereby the second substrate triglycine binds to the enzyme only after the acyl-enzyme adduct is formed (46). That supposition is corroborated by our observation that the above mentioned solvent molecule is replaced by the (Gly) 3 substrate when it is soaked into the crystals of the SrtB ⌬N30 ϩ MTSET inhibitor complex and by our inability to identify electron density for (Gly) 3 soaked into the native SrtA ⌬N59 and SrtB ⌬N30 crystals. The proposition that the amide bond between the first and second glycine in the (Gly) 3 substrate is essential for binding to the acylated enzyme active site received visual confirmation in the SrtB ϩ MTSET ϩ Gly 3 crystal structure, where the loop that joins the strands hosting catalytic Cys and Arg residues has only one or two polar interactions with (Gly) 3 peptide backbone atoms (33). The apparent K m value reported for the LPETG peptide binding to SrtA varies from 20 to 116 M, perhaps accounting for the relatively low occupancy of the peptide in our crystal structure.
S. aureus sortases SrtA and SrtB with identical ␤-sheet topology and conserved spatial disposition of putative Cys-Arg catalytic dyads exhibit significant differences in two regions around the catalytic sites. When superposed, the significantly longer ␤6 -␤7 loop of SrtB ⌬N30 , 41 residues corresponding to 17 residues in SrtA ⌬N59 , partially occupies the space designated for the peptide substrate in SrtA ⌬N59 . Similarly, the ␤7-␤8 loop region of SrtA ⌬N59 , longer by three residues than the corresponding segment of SrtB ⌬N30 , extends significantly upwards from the catalytic site of the enzyme and folds inwards, and thus occupies the (Gly) 3 peptide-binding site observed in the SrtB ⌬N30 ϩ MTSET ϩ (Gly) 3 crystal structure. From this comparison, we can also conclude that the substrate-binding site of SrtB ⌬N30 is much narrower because of the ␤6 -␤7 loop position and shifted laterally toward the ␤7-␤8 loop. The presence of Tyr 181 and Tyr 128 residues pointing into the substrate binding pocket also lessen the depth of the SrtB substrate binding pocket. Closer examination reveals the position and distribution of a number of hydrophobic residues, Ala 92 , Ala 104 , Pro 163 , Ile 182 , Ile 199 , and Trp 194 at the bottom of the substrate binding pocket of SrtA, which facilitate the sequestration of the hydrophobic Leu, Pro, and Thr residues of the LPETG peptide substrate. In addition to Tyr 181 and Tyr 128 residues, Ser 221 , Glu 224 , and Asn 92 residues of SrtB point into its substrate binding pocket, which by replacing hydrophobic Ile, Trp, and Pro residues observed in SrtA might be providing SrtB specificity toward relatively polar NPQTN sorting signal.