Engineering the Substrate Specificity of Staphylococcus aureus Sortase A

The Staphylococcus aureus transpeptidase Sortase A (SrtA) anchors virulence and colonization-associated surface proteins to the cell wall. SrtA selectively recognizes a C-terminal LPXTG motif, whereas the related transpeptidase Sortase B (SrtB) recognizes a C-terminal NPQTN motif. In both enzymes, cleavage occurs after the conserved threonine, followed by amide bond formation between threonine and the pentaglycine cross-bridge of cell wall peptidoglycan. Genetic and biochemical studies strongly suggest that SrtA and SrtB exhibit exquisite specificity for their recognition motifs. To better understand the origins of substrate specificity within these two isoforms, we used sequence and structural analysis to predict residues and domains likely to be involved in conferring substrate specificity. Mutational analyses and domain swapping experiments were conducted to test their function in substrate recognition and specificity. Marked changes in the specificity profile of SrtA were obtained by replacing the β6/β7 loop in SrtA with the corresponding domain from SrtB. The chimeric β6/β7 loop swap enzyme (SrtLS) conferred the ability to acylate NPQTN-containing substrates, with a \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(k_{\mathrm{cat}}{/}K_{m}^{\mathrm{app}}\) \end{document} of 0.0062 ± 0.003 m-1 s-1. This enzyme was unable to perform the transpeptidation stage of the reaction, suggesting that additional domains are required for transpeptidation to occur. The overall catalytic specificity profile (\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(k_{\mathrm{cat}}{/}K_{m}^{\mathrm{app}}(\mathrm{NPQTN}){/}k_{\mathrm{cat}}{/}K_{m}^{\mathrm{app}}(\mathrm{LPETG})\) \end{document}) of SrtLS was altered 700,000-fold from SrtA. These results indicate that the β6/β7 loop is an important site for substrate recognition in sortases.

Recently, much attention has been focused on the role of surface protein virulence factors in the pathogenesis of Grampositive bacteria (1)(2)(3). These covalently anchored proteins mediate such diverse processes as adherence to and invasion of host endothelial tissues, immune system evasion, and iron acquisition (4,5). Inhibition of surface protein anchoring has thus emerged as a possible new therapeutic strategy for treatment of Gram-positive infections (6).
Sortases are a family of transpeptidases that catalyze the covalent attachment of these surface proteins to the peptidoglycan layer of the Gram-positive cell wall (4). Several lines of evidence support a role for sortases in the pathogenicity of infections. Strains of Staphylococcus aureus in which srtA was genetically deleted are unable to anchor surface proteins in vivo and show a reduced ability to form renal abscesses in mice (7). The srtA knock-out in Listeria monocytogenes has been shown to be less invasive in vitro, fails to anchor internalin to the cell wall in vivo, and is reduced in its ability to colonize the liver and spleen in a mouse model of infection (8,9). In Streptococcus gordonii, srtA inactivation leads to decreased surface protein display, reduced fibronectin binding in vitro, and a reduced ability to colonize the oral mucosa of mice (10). Likewise, srtA knockouts from Streptococcus family members S. mutans (11), S. pneumonia (12), S. pyogenes (13), S. suis (14), S. sanguinis (15), and S. agalactiae (16) have been shown to have attenuated pathogenicity.
Most Gram-positive bacterial species whose genomes have been sequenced to date contain at least one sortase homolog, and many species contain multiple sortase genes (17). In many cases, putative sortase genes are located within operons that contain genes encoding potential sortase substrates (proteins with C-terminal LPXTG motifs). This raises the possibility that many of these additional sortase genes are co-expressed along with their substrates, suggesting that they might play discrete functional roles by anchoring different subsets of surface proteins (17).
Within the genome of S. aureus, there are two sortase isoforms, termed Sortase A (SrtA) and Sortase B (SrtB). SrtA is responsible for anchoring proteins containing a C-terminal tripartite sorting signal, which consists of 1) an LPXTG pentapeptide followed by 2) a less well conserved hydrophobic domain and 3) a basic charged tail (18,19). Both the hydrophobic domain and the charged tail help to retain the putative surface protein in the membrane prior to sortase-catalyzed anchoring. SrtA cleaves the LPXTG sequence between the threonine and glycine residues by use of a nucleophilic active site cysteine (Cys 184 ) (20). An acyl-enzyme intermediate is formed, which is then resolved by nucleophilic attack from an amino group on the Gly 5 cross-bridge of branched Lipid II. This leads to covalent attachment of the N-terminal portion of the surface protein to Lipid II, from which it is incorporated into the bacterial cell wall (21).
SrtB functions analogously to SrtA with two key differences. First, SrtB has a specificity profile different from that of SrtA. It selectively anchors the NPQTN-containing protein IsdC, and it does not anchor LPXTG-containing proteins (22). Second, the anchor structure for SrtB is not Lipid II (a peptidoglycan biosynthetic precursor) but is the Gly 5 cross-bridge of mature peptidoglycan from the staphylococcal cell wall (23). In addition, the gene for SrtB is located along with its substrate in an ironresponsive operon, which is only transcribed under conditions of iron starvation (22). Since the acquisition of iron is an essential process for bacterial colonization of mammalian host tissues (24), it is likely that SrtB carries out an essential function for virulence by anchoring a protein with a specific role in iron acquisition to a unique anchor motif in the cell wall (25). Although SrtA and SrtB are highly specific for their unique recognition motifs, to date little is known about how these enzymes discriminate between their substrates. Herein we report the application of mutagenesis, domain swapping, and kinetic analysis to identify specificity-determining regions of SrtA and SrtB.

EXPERIMENTAL PROCEDURES
Materials-Buffer salts were purchased from Sigma. Standard Fmoc 2 amino acids (Novabiochem), Fmoc-Dap-(DNP)-OH (Bachem), NH 2 -Gly 5 -OH (Bachem), and t-butoxycarbonyl-Abz-OH (Bachem) were purchased and were used with no further purification. Oligonucleotide primers were purchased from Integrated DNA Technologies (Coralville, IA) and used without further purification. Protein purification was performed on an AKTA fast protein liquid chromatography system with a fraction collector (Amersham Biosciences). Chelating Sepharose fast flow chromatography resin (Amersham Biosciences) and a HiPrep 26/60 Sephacryl S-200 high resolution gel filtration column (Amersham Biosciences) were used according to the manufacturer's recommendations. HPLC was performed using an Agilent 1200 series system equipped with an autosampler, fraction collector, diode array UV detector, fluorimeter, and either a semipreparative Jupiter TM octadecyl silica column (Phenomenex) or a fast analytical (4.6 ϫ 50 mm, 3 m) octadecyl silica column (Vydac). MALDI-MS was performed on a Voyager TM MALDI-time-of-flight mass spectrometer (Applied Biosystems). Electrospray ionization mass spectrometry (ESI-MS) was performed on an Agilent 1100 series LC/MSD ion trap electrospray mass spectrometer.
Cloning, Expression, and Purification of SrtB ⌬N21 -SrtB ⌬21 was amplified from the genomic DNA of S. aureus strain N315, using the primers SrtB ⌬21 fwd (5Ј-CCCGAATTCCATATGGGTTAC-AAAATTGTTCAAACATATATT-3Ј) and SrtB ⌬21 rev (5Ј-CAT-TAGCGTGGATCCCTCGAGTTAACTTACCTTAATTATT-TTTGC-3Ј). The resulting DNA fragment was doubly digested with EcoRI and BamHI and ligated with T4 DNA ligase into the subcloning vector pUC19, which had previously been digested with EcoRI and BamHI, to generate the plasmid pUC19SrtB ⌬N21 . This plasmid was transformed into DH5␣ cells and isolated with a QIAprep TM spin miniprep kit, and the desired sequence was confirmed by DNA sequencing. BamHI and NdeI were used to liberate the gene from the plasmid, and it was then cloned into a similarly treated pET15b expression vector to generate the plasmid pET15bSrtB ⌬N21 . This construct was transformed into BL21(DE3) cells, and cells were grown in LB medium with 100 g/ml ampicillin. Cells were grown to an A 600 of 0.7, and then 1 mM isopropyl 1-thio-␤-D-galactopyranoside was added to induce expression of SrtB ⌬N21 ; cells were grown overnight at 25°C and then harvested by centrifugation. Cells were resuspended in the same buffer, and protein was purified in the same manner as for SrtA ⌬N24 . Purified SrtB ⌬N21 was concentrated to 948 M using 10,000 molecular weight cut-off Centriplus centrifugal filters (Amicon). SrtB ⌬N21 concentration was determined spectrophotometrically, using the extinction coefficient ⑀ 280 ϭ 26,360 M Ϫ1 cm Ϫ1 . Protein molecular weight was confirmed by MALDI-MS.
Sequence and MEME Analysis of Sortase Family Members-The sequences of S. aureus SrtA and SrtB were used as starting points for a bioinformatic analysis. PSI-BLAST (NCBI) (27) searches were performed on each to identify similar sequences. Five iterations of PSI-BLAST identified 214 similar sequences to SrtA, whereas two iterations of PSI-BLAST produced 42 sequences with similarity to SrtB. Combining these and removing all proteins with E values of greater than 3.0 ϫ 10 Ϫ5 left 175 hits. Further analysis reduced this number to 135 sequences that showed conservative identities with either SrtA or SrtB.
ClustalW (EMBL-EBI) (28) was used to generate a large scale multiple sequence alignment of these 135 sequences. T-Coffee (EMBL-EBI) (29) was used to generate an alignment of a smaller subset of 25 sequences (supplemental Fig. S1).
Motifs were discovered using MEME (SDSC) (30). MEME analysis was performed on a small scale alignment of 25 sequences (17 SrtA homologs, 8 SrtB homologs). Discovered motifs were mapped onto a T-Coffee multiple sequence alignment using GeneDoc (31) and then onto the existing structures of SrtA (Protein Data Bank code 1T2W) and SrtB (Protein Data Bank code 1NG5) (Fig. 1). In light of previous sortase activity studies (26), this sequence-structural analysis helped identify putative sequence motifs and residues likely to be involved in substrate binding, recognition, and catalysis.
Generation of a Chimeric ␤6/␤7 Loop Swap (SrtLS)-To generate the novel sequence of SrtLS ⌬N24 , the technique of splicing by overlap extension was used (32). The plasmids pET15bSrtA ⌬N24 (26) and pET15bSrtB ⌬N21 were used as starting templates for PCR. Two rounds of this technique were needed to generate SrtLS ⌬N24 , which contains the sequence of SrtA ⌬N24 from the 5Ј end to the ␤6/␤7 loop, the sequence of SrtB ⌬N21 for the loop, and then the sequence of SrtA ⌬N24 from the loop end to the 3Ј end of the gene. Splicing by overlap extension-PCR was performed according to the method of Horton et al. (32). Reaction products were gel-purified following each successful amplification and used as templates for the next rounds of PCR. The following primers were used (underlined sequences are those derived from srtB): Primer a, 5Ј-CCCGCGAAATTAATACGAC-TCACTATAGGG-3Ј; Primer b, 5Ј-CTTTAGTAGTAGTTTTA-TCTCTTATACTTG-3Ј; Primer c, 5Ј-CAAGTATAAGAGAT-AAAACTACTACTAAAG-3Ј; Primer d/g, 5Ј-CCGTTTAGAG-GCCCCAAGGGGTTATGC-3Ј; Primer e, 5Ј-GTAATTAATG-TTAATTGTTTATCTTTTACCG-3Ј; Primer f, 5Ј-CGGTAAA-AGATAAACAATTAACATTAATTAC-3Ј.
The final PCR product was digested with NdeI and BamHI, ligated into pET15b vector that had previously been digested with NdeI and BamHI, and then transformed into DH5␣ cells. DNA sequencing confirmed generation of the vector pET15bSrtLS ⌬N24 .
Mutagenesis of SrtLS-Single point mutations A118Y and I208S and the double mutant A118Y/I208S were introduced into the plasmid pET15bSrtLS ⌬N24 by PCR, using the same methods and primers used for mutagenesis of SrtA. Mutations were confirmed by DNA sequencing.
Circular Dichroism Spectroscopy-Aliquots (300 l) of SrtA ⌬N24 , SrtB ⌬N21 , and SrtLS ⌬N24 were dialyzed overnight at 4°C against buffer containing 10 mM Tris (pH 7.5), 30 mM NaCl, and 5 mM CaCl 2 . Samples were diluted in the same buffer until final protein concentrations were ϳ10 M. Protein concentrations of the final solutions were obtained spectrophotometri- For each protein, 400 l was transferred to a quartz cuvette with a 0.1-cm path length for CD data collection. CD spectra were measured on an Aviv Instruments model 202 circular dichroism spectrometer. Data were collected at 25°C, from 260 to 190 nm in 1-nm steps, with a scanning speed of 4 s/nm. Three scans were taken for each sample and were averaged and corrected for background absorbance. The observed CD signal (millidegrees) was converted to mean residue ellipticity ( m , degrees cm 2 dmol Ϫ1 residue Ϫ1 ) using Equation 1, where obs represents the observed ellipticity in degrees, c is the protein concentration (M), d is the optical path length in cm, and n is the number of amino acid residues in the protein (202 for SrtA ⌬N24 , 244 for SrtB ⌬N21 , and 228 for SrtLS ⌬N24 ).
Solid Phase Synthesis of Peptide Substrates-Substrates Abz-LPETG-Dap(DNP)-NH 2 and Abz-KVENPQTNAGT-Dap-(DNP)-NH 2 were synthesized by the Fmoc/piperidine strategy on PAL resin on a 0.25-mmol scale using an Applied Biosystems 433A synthesizer. Peptides were cleaved from the resin via incubation with a 95:2.5:2.5 trifluoroacetic acid/water/triisopropylsilane mixture for 2.5 h. The peptides were precipitated using cold diethyl ether following the removal of excess trifluoroacetic acid via rotary evaporation. After filtration, the precipitate was dissolved in a 50:50 water/acetonitrile mixture and lyophilized to yield crude peptide. A semipreparative C 18 Jupiter TM column was used to purify the peptides to Ն98% purity via reverse phase HPLC. MALDI-time-of-flight MS was used to verify the identities of the purified products (Abz-LPETG-Dap- HPLC Assays for Sortase Activity-The assay procedure of Kruger et al. (33) for SrtA was employed to measure SrtB activity, with the modifications indicated (see Fig. 3). For SrtB, initial activity assays were performed with varying concentrations of enzyme (1, 10, 50, or 94.8 M) and the peptide substrates Abz-KVENPQTNAGT-Dap(DNP)-NH 2 (2.4 mM) and NH 2 -Gly 5 -OH (2 mM). Reactions were performed in a 100-l volume in standard assay buffer (150 mM NaCl, 5 mM CaCl 2 , 300 mM Tris-Cl, pH 7.5) at 37°C, and the enzyme volume added to each assay was kept under one-tenth of the final assay volume. Assays were initiated by the addition of the enzyme; were incu-Engineering the Substrate Specificity of S. aureus SrtA bated for 0.5, 1, 2, or 4 h; and were quenched by the addition of 50 l of 1.2 N HCl. 50 l of the quenched reaction mixture was then injected onto a Vydac reverse phase fast analytical C 18 column (4 ml/min) and separated using a linear gradient from 0 to 45% CH 3 CN, 0.1% trifluoroacetic acid over 5 min. Peaks corresponding to elution of the substrate (Abz-KVENPQTNAGT-Dap(DNP)-NH 2 ) and product (NH 2 -G-Dap(DNP)-NH 2 ) were monitored at 355 nm. At the same time, peaks corresponding to elution of the substrate and the transpeptidation product (Abz-KVENPQTGGGGG-OH) were monitored by fluorescence, with excitation at 318 nm and emission at 420 nm. Integration of the areas of the substrate and product peaks at 355 nm allowed for determination of the percentage of substrate converted to product. Linearity of SrtB was confirmed by plotting product formation as a function of time over the 4-h assay period.
For determination of the kinetic parameters of SrtB, 94.8 M enzyme was used in a 3-h assay, since these conditions were expected to give product conversions of 1-5% during the assay. The substrate Abz-KVENPQTNAGT-Dap(DNP)-NH 2 was varied from 74.8 M to 21.9 mM, whereas NH 2 -Gly 5 -OH was held constant at 2 mM. GraFit version 4.0 (Erithacus Software) was used to fit the raw data to the Michaelis-Menten equation, producing estimates of the kinetic parameters k cat and K m .
Initial assays for comparison of SrtA, SrtB, and SrtLS activity with Abz-KVENPQTNAGT-Dap(DNP)-NH 2 were conducted using a 94.8 M concentration of each enzyme and 2.7 mM substrate. Assays were incubated for 3, 6, 9, and 12 h, and the amount of product was quantitated as described above for SrtB. Estimates of the second order parameter k cat /K m app were obtained using the relation, where v o is the initial velocity, [E] is the concentration of enzyme, and [S] is the concentration of substrate. Initial activity assays for each enzyme with Abz-LPETG-Dap(DNP)-NH 2 were performed and analyzed similarly.
Mass Spectrometry of Reaction Products-To confirm the identity of reaction products, individual product peaks from the HPLC assay were collected with a fraction collector and concentrated using a speed vacuum system. Dried, concentrated products were resuspended in 100 l of 50% CH 3 CN in H 2 O and injected onto an Agilent 1100 series LC/MSD ion trap electrospray mass spectrometer. MS data were collected in positive ion mode, and ions corresponding to putative reaction products were selected, fractionated, and analyzed by MS/MS.

RESULTS AND DISCUSSION
Despite carrying out similar biological functions, SrtA and SrtB have different specificity profiles. SrtA anchors LPXTGcontaining proteins, whereas SrtB is specific for NPQTN-containing proteins (22,26). In order to better understand the basis for this specific recognition, we used sequence analysis to look at SrtA and SrtB isoforms from various Gram-positive bacteria. We looked for sequence positions that were highly conserved among SrtA and SrtB sequences but were isoform-specific. We then performed global analysis of the structures of S. aureus SrtA (Protein Data Bank code 1T2W) and SrtB (Protein Data Bank code 1NG5) to identify larger domains that differed structurally between the two enzymes (34,35). Mutagenesis and domain-swapping were then used to explore the effect of each identified region on SrtA specificity.
Bioinformatic and MEME Analysis of Sortase A and B Isoforms-A large scale multiple sequence alignment of 135 sequences of Sortase A and B isoforms from different Grampositive species was performed using the program ClustalW. Analysis of this alignment highlighted the presence of several conserved residues and domains. The active site cysteine (Cys 184 in SrtA) was conserved among all sortase homologs, as were the nearby histidine and arginine residues (His 120 and Arg 197 in SrtA). These three conserved residues are located within two conserved domains along the ␤4, ␤7, and ␤8 strands, which make up the sortase active site. MEME analysis of a smaller alignment of 25 sequences revealed the presence of seven conserved motifs ( Table 1). Four of these motifs were common to both SrtA and SrtB isoforms. When these motifs were mapped onto the existing crystal structures of SrtA and SrtB from S. aureus, they corresponded to distinct structural features. Motifs 1 and 3 corresponded to the ␤4 and ␤7 strands, respectively, which make up the floor of the sortase active site. Motif 2 maps onto the ␤1 and ␤2 strands of both proteins, where it forms part of the underlying structure. The N terminus of both proteins is membrane-associated, and this sequence region also corresponds to a conserved motif (Motif 5).
In addition to these common sortase motifs, there was a single motif found only in SrtA sequences, and two motifs that were unique to SrtB sequences. Motif 6 was found only in SrtB TABLE 1 MEME-discovered motifs in SrtA and SrtB sequences MEME was used to search for conserved motifs in 25 sequences of SrtA and SrtB isoforms from Gram-positive bacteria, as described under "Experimental Procedures." Four motifs common to both SrtA and SrtB were observed (motifs 1, 2, 3, and 5). Each of these maps onto conserved structural elements. Motifs 1 and 3 correspond to the ␤4 and ␤7 strands, which make up the basis for the sortase active site, and contain the catalytic His and Cys residues, respectively. Motif 6 was found only among SrtB isoforms and maps onto a structural feature (␤2/␤3 hairpin loop) not found in SrtA. Motifs 4 and 7 are unique to SrtB and SrtA, respectively, but map onto the same structural feature (␤6 strand, ␤6/␤7 loop).

Motif
No  (Fig. 1). Motifs 4 and 7 were unique to SrtB and SrtA, respectively. However, they correspond to roughly the same structural features: the ␤5 and ␤6 strands and the ␤6/␤7 loop. This was the only case in our analysis where unique motifs in SrtA and SrtB matched up to an equivalent structural feature. Given the high degree of conservation of these sequences within SrtA and SrtB isoforms and their difference between isoforms, it seems plausible that this conservation is functionally important. Summaries of the MEME-determined motifs, their consensus sequences, and their E values are listed in Table 1.
Large scale multiple sequence alignments and MEME analysis are useful tools for looking at motif-wide differences between related sequences. However, these tools may obscure subtle differences between isoforms, since the much larger number of SrtA sequences skews the results. To further examine positional differences between SrtA and SrtB isoforms, we carried out a smaller scale alignment of a dozen SrtA and SrtB sequences. Within the highly conserved active site motifs determined earlier, we looked for individual residues that differed between SrtA and SrtB isoforms. Two residues in particular stood out (Fig. 2). The position 2 residues N-terminal to the catalytic cysteine (Cys 184 in S. aureus SrtA, Cys 233 in S. aureus SrtB) is conserved as an isoleucine in all of the SrtA isoforms but as a serine in the SrtB isoforms. This yields an active site motif of ITC for SrtA isoforms but STC for SrtB.
Similarly, the position 2 residues N-terminal to the catalytic histidine (His 120 in S. aureus SrtA, His 130 in S. aureus SrtB) is differentially conserved, yielding an AXH motif in SrtA isoforms and YXH in SrtB. Since the active sites of sortases are composed of extended ␤-sheets, these residues (2 residues away from the catalytic Cys-His pair) would be expected to have their side chains oriented nearly parallel to the catalytic residues and would be positioned to contact and orient the peptide substrate. The presence of the more polar Tyr 128 -Ser 221 pair in SrtB, as compared with the Ala 118 -Ile 182 pair in SrtA, might help to stabilize the binding of the more polar NPQTN sequence versus the LPXTG of SrtA substrates. In addition, the Tyr 128 and Ser 221 residues in the SrtB active site are both capable of forming side chain hydrogen bonds, which might better stabilize the substrate for catalysis by Cys 223 . To test these hypotheses, we selected the residues Ala 118 and Ile 182 from SrtA for mutagenesis.
Activity Assays of SrtA Point Mutants-To determine if the SrtA A/Y and I/S mutations affected substrate specificity, we prepared the point mutants SrtA A118Y and I182S as well as the double mutant SrtA A118Y/I182S. Initial activity assays showed that the mutants I182S and A118Y were inactive against an NPQTN peptide (results not shown). The double mutant, SrtA A118Y/I182S, was inactive against both LPETG and NPQTN peptides. Thus, the effect of the point mutations was to reduce native SrtA activity and not to alter substrate preference.
Identification of the ␤6/␤7 Loop as a Potential Specificity Determinant-Since point mutations had little effect on the specificity of SrtA, we sought to identify larger domains that differed between SrtA and SrtB and thus might be expected to be functionally important for specificity. A comparison of the crystal structures of SrtA and SrtB was performed to identify domains that differed in structure between the two proteins. Although sequence alignments have shown that S. aureus SrtA and SrtB have very low sequence homology (40% homology, 23% identity over 184 equivalent residues (36)), their structures show similar overall folds (Fig. 1). The overall structures align with a root mean square deviation of 3.30 Å over 127 equivalent residues. However, within the core ␤-barrel, the two proteins are very similar, aligning with a root mean square deviation of only 1.31 Å. Thus, most of the structural differences between SrtA and SrtB lie in surface-exposed loops and helices, not in the ␤-sheet core.
Most notable among these structural differences are the presence of an extra domain in SrtB (the ␤2/␤3 hairpin loop) and the vastly different sequences and conformations of the ␤6/␤7 loops in the two enzymes. In SrtA, this region is a short, 15-amino acid unstructured loop (Val 161 -Asp 176 ), whereas in SrtB, it is 41 amino acids long (Lys 174 -Asp 215 ) and consists of a short loop, an ␣ helix, and another loop. In SrtA, the ␤6/␤7 loop contains several highly conserved hydrophobic residues, such as Val 168 and Leu 169 , which are replaced by more polar residues in SrtB, such as Asn 180 , Tyr 181 , and Arg 183 . Although the size

Engineering the Substrate Specificity of S. aureus SrtA
and composition of the ␤6/␤7 loops are different in SrtA and SrtB, they occupy equivalent structural positions, with the points of connectivity of the loops being the same in both structures (Asp 160 /Tyr 173 at the N termini in SrtA/SrtB, respectively, Lys 177 /Lys 216 at the C termini). The loop in SrtA and the first loop of the domain in SrtB are both positioned at the back of the active site cavity and appear to be ideally located to contact the N-terminal residues of the recognition motif (Leu-Pro in SrtA substrates, Asn-Pro in SrtB substrates) (Fig. 2).
Calcium binding is known to be required for efficient SrtA catalysis (37). Recent NMR analyses by Naik et al. (38) revealed that the backbone dynamics of the ␤6/␤7 loop in SrtA change significantly upon calcium binding. 15 N relaxation measurements showed that the loop is highly disordered in the absence of bound Ca 2ϩ . In the apoenzyme, the loop rapidly fluctuates between a substrate-binding closed state, in which the loop is near the active site, and an open state, in which the loop is far from the active site. The calcium-bound form of SrtA contains a single Ca 2ϩ binding site, made up of 4 residues (Glu 105 , Glu 108 , Asp 112 , and Asn 114 ) on the ␤3/␤4 loop and a single conserved glutamate (Glu 171 ) on the ␤6/␤7 loop. Calcium binding thus constrains the relative motions of the ␤3/␤4 and ␤6/␤7 loops, locking the ␤6/␤7 loop into the "closed" state (38).
By analogy with SrtA, SrtB also contains a putative calcium-binding site on the ␤3/␤4 loop (Asp 113 , Asn 116 , Glu 117 , and Asn 122 ) and a conserved acidic residue (Asp 185 ) on the ␤6/␤7 loop. To date, a detailed analysis of the calciumbinding properties of SrtB has not been carried out. However, given the similarities between the structures and functions of SrtA and SrtB, it seems reasonable to assume that SrtB also utilizes bound Ca 2ϩ to stabilize the conformation of its ␤6/␤7 loop, allowing for productive substrate binding and catalysis.
Design of a ␤6/␤7 Loop Swap Chimera, SrtLS-Based on the above lines of evidence, we hypothesized that the ␤6/␤7 loops might be playing an important role in contacting substrate and determining the different specificities of SrtA and SrtB. In order to test this hypothesis, we designed a loop swap enzyme (SrtLS ⌬N24 ), wherein we exchanged the ␤6/␤7 loop of SrtA with the corresponding loop from SrtB. This chimeric protein contains the sequence of SrtA for residues Lys 26 -Asp 160 and then the sequence of SrtB for the next 42 residues (Lys 174 -Asp 215 in SrtB, renumbered Lys 161 -Asp 202 in this enzyme), and then the sequence of SrtA for the remainder (Lys 177 -Lys 206 in SrtA, renumbered Lys 203 -Lys 232 in this enzyme). Since the ␤6/␤7 loop is longer in SrtB than in SrtA by 26 residues, this causes the C-terminal residues in SrtLS to be renumbered ϩ26 from their corresponding residues in SrtA (i.e. Cys 184 from SrtA becomes Cys 210 in SrtLS).
Kinetics of SrtB with Abz-KVENPQTNAGT-Dap(DNP)-NH 2 -To serve as a benchmark for activity comparisons, we first characterized SrtB using an HPLC-based assay, similar to one developed by Kruger et al. to measure the activity of SrtA (33). In place of the SrtA substrate Abz-LPETG-Dap(DNP)-NH 2 , we  (Fig. 3). For comparison, previous studies have reported a k cat of 0.28 Ϯ 0.02 s Ϫ1 for SrtA ⌬N24 and a K m of 7.3 Ϯ 1 mM for SrtA ⌬N24 with Abz-LPETG-Dap(DNP)-NH 2 (39). These values show that SrtB has a K m for this substrate that is within the error of that measured for SrtA but a k cat that is ϳ500-fold lower.
Circular Dichroism Analysis of SrtA, SrtB, and SrtLS-Making large scale changes to the sequence of a protein may introduce changes in the secondary structure and folding patterns. In order to determine the effect of the loop swap on the folded structure of SrtA, we compared the CD spectra of SrtA, SrtB, and SrtLS. Since SrtLS is a hybrid of SrtA and SrtB, it was expected that the CD spectra of SrtLS would represent a combination of the spectra of SrtA and SrtB. CD data collected for SrtA, SrtB, and SrtLS are shown in Fig. 4. The spectra collected for SrtA shows a combination of random coil and ␤-sheet features, consistent with the three-dimensional structure of the enzyme (34,37). SrtA is fully enzymatically active under the conditions used for CD (data not shown) and is therefore assumed to be properly folded. SrtB shows a CD spectrum consistent with an equal mix of ␣-helical and ␤-sheet structure, which is in agreement with the crystal structure solved for this protein (36).
SrtLS has a CD spectrum that is a combination of the features of the spectra of SrtA and SrtB (Fig. 4). It has the same general shape as the SrtA spectrum, although it has more ␣-helical character in the far UV. The ␤6-␤7 domain from SrtB contains an ␣-helix that is not ordinarily found in SrtA. Thus, the presence of this swapped domain in SrtLS would be expected to add more ␣-helical character to the secondary structure, which is what was observed. Based on these results, it appears that SrtLS exhibits structural features common to both the folded structures of SrtA and SrtB, in a predicted manner. We therefore assayed SrtLS for enzymatic activity and altered substrate specificity.
Activity Comparisons of Sortase Mutants-To assess the role of the ␤6/␤7 loop swap on substrate recognition, we performed activity assays of SrtA, SrtB, and SrtLS with LPXTG and NPQTN substrates to directly compare rates and efficiencies. We also assayed the single point mutants SrtLS A118Y and SrtLS I208S, as well as the double mutant SrtLS A118Y/I208S. Mutants were compared based on their catalytic utilization Engineering the Substrate Specificity of S. aureus SrtA MARCH 2, 2007 • VOLUME 282 • NUMBER 9 ratios (k cat /K m app ), since their intrinsically low activities prevented individual determinations of k cat and K m . Slightly higher substrate concentrations (2.7 mM) were used than would ideally be the case; however, given the low levels of activity of several of the enzymes being measured, it was necessary to use low millimolar concentrations of substrate in order to obtain detectable levels of product using our HPLC-based assay. Estimates of k cat / K m app obtained for SrtA (with LPETG) and SrtB (with NPQTN) were found to closely match k cat /K m values obtained by individual measurements of k cat and K m , and Equation 2 was therefore determined to be valid for our assay conditions. SrtA readily reacts with its native substrate Abz-LPETG-Dap(DNP)-NH 2 , cleaving it with a k cat /K m app of 37 Ϯ 3 M Ϫ1 s Ϫ1 , in excellent agreement with values reported earlier (33,39). By contrast, SrtB had no detectable activity against the LPETG substrate, even after 12 h at 37°C and an enzyme concentration of 94.8 M. This is consistent with earlier studies in which no LPETG-cleaving activity was observed for SrtB in vitro (22). SrtLS was able to hydrolyze the LPETG substrate but with a vastly reduced k cat /K m app of 3.7 Ϯ 0.6 ϫ 10 Ϫ4 M Ϫ1 s Ϫ1 (Table 2). This amounts to a 100,000-fold decrease in LPXTG cleavage efficiency by the loop swap enzyme versus SrtA, suggesting that alteration of the ␤6/␤7 loop was sufficient to greatly reduce recognition and processing of the LPETG substrate.
When assayed with Abz-KVENPQTNAGT-Dap(DNP)-NH 2 , SrtA generated a small amount of product, the first such detection of NPQTN cleavage activity in SrtA. However, the efficiency of cleavage was slow, with k cat /K m app equal to 8.7 Ϯ 0.6 ϫ 10 Ϫ4 M Ϫ1 s Ϫ1 . Compared with its LPETG cleavage efficiency, this represents an almost 45,000-fold preference for LPETG versus NPQTN-containing recognition motifs by SrtA. SrtB cleaved the NPQTN peptide as expected, with a k cat /K m app of 0.063 Ϯ 0.01 M Ϫ1 s Ϫ1 , in close agreement with our earlier estimate from the steady-state kinetics of SrtB (this work). The loop swap enzyme SrtLS also cleaved the NPQTN peptide, with an estimated k cat /K m app of 6.2 Ϯ 0.3 ϫ 10 Ϫ3 M Ϫ1 s Ϫ1 (Table 2). Although slow, this still reflects a 16-fold preference of SrtLS for NPQTN relative to LPETG. More importantly, the k cat /K m app of SrtLS is only 11 times lower than that of SrtB. Thus, the swapping of a single surface loop was sufficient to alter the specificity profile (k cat /K m app for NPQTN versus for LPETG) of SrtA by a factor of over 700,000. Taken together, the results of these kinetic assays strongly argue that the ␤6/␤7 loop is an important specificity-determining site in SrtA and SrtB.
Activity Comparisons of SrtLS Active Site Mutants-In order to assess the impact of the conserved isoleucine and alanine residues identified by our bioinformatic analysis, we generated point mutants of SrtLS, I208S and A118Y, as well as the double mutant A118Y/I208S. Each of these mutants was expressed and purified and then assayed for LPETG and NPQTN activity. The single point mutant SrtLS A118Y was able to cleave the LPETG peptide, at a k cat /K m app of 2.7 Ϯ 0.1 ϫ 10 Ϫ4 M Ϫ1 s Ϫ1 , about 70% of the efficiency of SrtLS. SrtLS I208S was also active against LPETG, with a k cat /K m app of 2.2 Ϯ 0.1 ϫ 10 Ϫ4 , about 60% of the efficiency of SrtLS ( Table 2). The double mutant SrtLS A118Y/ I208S was completely inactive for LPETG cleavage, to the limits of detection in our assay (Ͼ25 nmol of product formed). For the substrate Abz-KVENPQTNAGT-Dap(DNP)-NH 2 , the addition of the point mutations A118Y, I208S, and A118Y/I208S had no effect on the activity of SrtLS (Table 2). Although these point mutations did not increase the NPQTN cleavage activity of SrtLS, they did decrease the ability of SrtLS to recognize the LPETG substrate, leading to an overall increase in the specificity ratio of SrtLS ( Table 2). The effects of these mutations suggest that the residues Ala 118 and Ile 182 may play an important role in LPETG recognition by SrtA, whereas the corresponding residues from SrtB do not appear to be important for NPQTN recognition. . CD wavelength scans of SrtA (black), SrtB (red), and SrtLS (green). Spectra were collected at 25°C, and the observed signal was converted to mean residue ellipticity as described under "Experimental Procedures." SrtA shows a spectrum consistent with a secondary structure predominantly composed of ␤-sheet and random coil. The spectrum of SrtB is more consistent with a secondary structure of roughly equal parts ␣-helix and ␤-sheet. SrtLS has a spectrum that is similar overall to that of SrtA, with some influence from SrtB, consistent with its hybrid nature. Separate scans at 98°C (not shown) confirmed that the unfolded spectra for each protein differed from the 25°C spectra, suggesting that each protein is properly folded.

TABLE 2 Kinetic parameters measured for sortase mutants
Assays were performed at low ͓S͔ relative to the K m , allowing for estimates of k cat /K m app as described under "Experimental Procedures." Specificity ratio ϭ (k cat /K m app (NPQTN))/(k cat /K m app (LPETG)). ND, not detectable to the limit of detection (Ͼ25 nmol of product formation). Replacement of the ␤6/␤7 loop from SrtA with the loop from SrtB generated an enzyme (SrtLS) with a k cat /K m app for NPQTN only 11 times lower than SrtB and a specificity ratio shifted 700,000-fold from SrtA.

Engineering the Substrate Specificity of S. aureus SrtA
Mass Spectrometry of Reaction Products-To confirm the identity of our reaction products, we analyzed the product peaks from the SrtB and SrtLS assays with Abz-KVENPQT-NAGT-Dap(DNP)-NH 2 by ESI-MS. The major product peak of both the SrtB and SrtLS reactions (as monitored at 355 nm) had a retention time of 1.46 min. MS analysis of purified samples of this peak revealed the presence of an ion at m/z ϭ 613.2, corresponding to the expected cleavage product NAGT-Dap(DNP)-NH 2 . MS/MS analysis of this ion confirmed the identity of the product. At 420 nm, the major product peak of the SrtB reaction eluted at a retention time of 1.65 min. Subsequent ESI-MS analysis of this peak showed it to contain an ion with m/z ϭ 1219.5, corresponding to the expected transpeptidation product of Abz-KVENPQTGGGGG-OH (Fig. 5B). Also present in the MS spectra was an ion with m/z ϭ 934.4, which corresponds to the expected hydrolysis product Abz-KVENPQT-OH. MS/MS analysis of both of these ions confirmed their identities.
When the product peak of the SrtLS reaction (as monitored at 420 nm) was analyzed, only the ion at m/z ϭ 934.4 was observed (Fig. 5C). MS/MS analysis of this ion confirmed its identity as the hydrolysis product Abz-KVENPQT-OH. Ions corresponding to the predicted transpeptidation product were not found. Thus, it appears that whereas SrtLS is capable of acylating the NPQTN sorting signal, it is unable to perform transpeptidation of the cleaved sorting signal onto the Gly 5 substrate. Instead, the acyl-enzyme intermediate is resolved by hydrolysis from water.
The fact that SrtB, but not SrtLS, is competent to carry out the transpeptidation stage of the sorting reaction in our in vitro assays suggests that there are differences in the binding sites for the cell wall peptidoglycan substrates of SrtA and SrtB. A recent study by Marraffini and Schneewind (23) found evidence for a distinct anchor structure for the IsdC protein (a SrtB substrate) in vivo. They suggest that SrtB-anchored surface proteins are attached to peptidoglycan that has shorter glycan chains and fewer cross-links than is the case for SrtA-anchored proteins. This implies that there are different peptidoglycan substrates for SrtA and SrtB. If this is the case, then there must be unique features in the structure of SrtB that allow it to specifically recognize its less cross-linked peptidoglycan substrates.
In our study, pentaglycine was able to serve as an anchor for NPQTN peptide cleaved by SrtB but not by our chimera, SrtLS. Pentaglycine was also able to serve as an anchor for LPETG that had been cleaved by SrtA (data not shown). It is possible that the presence of the swapped loop in SrtLS somehow prevents pentaglycine from accessing the active site and thus is the cause of the lack of transpeptidation. Another possible explanation is that there may be additional domains or features on the surface of SrtB that are not included in SrtLS and that the presence of these domains is necessary for transpeptidation of the cleaved sorting signal to occur. One candidate domain for this is the conserved SrtB-only domain found in our MEME analysis (Fig.  1), which corresponds to a hairpin between strands ␤2 and ␤3 on SrtB. This unique fold is not observed in the structure of SrtA and could serve as a recognition site for the peptidoglycan substrate of SrtB. Studies are currently in progress to swap this loop into SrtLS, to see if its addition allows for successful transpeptidation onto pentaglycine.
Redesign of existing enzyme frameworks for altered specificity has long been a means to study molecular recognition and catalysis. In the case of malate dehydrogenase, a single point mutation was sufficient to convert it into a lactate dehydrogenase, with a k cat /K m shift of 10 7 (40). However, for many systems, larger numbers of mutations or more drastic changes are required to alter specificity (41,42). In some of these cases, domain swapping has been validated as a tool for changing the specificity of closely related enzymes for which crystal structures are available (43,44). Ma and Penning (43) used loop swaps as a tool to change mammalian 3␣-hydroxysteroid dehydrogenase into 20␣-hydroxysteroid dehydrogenase. In their study, point mutants had little effect on the specificity of 3␣-hydroxysteroid dehydrogenase, but replacement of surface loops with the corresponding loops from 20␣-hydroxysteroid dehydrogenase did change specificity, with a shift in k cat /K m of 2 ϫ 10 11 . Similarly, in their study on the conversion of trypsin to chymotrypsin, Hedstrom et al. (44) found that it was necessary to swap two of the trypsin surface loops in order to alter the specificity of the enzyme. A shift in k cat /K m of 10 9 was observed upon swapping of these loops. We observed a k cat /K m shift for SrtA that was slightly smaller, only 7 ϫ 10 5 . However, when one considers the very slow rates in vitro of SrtA and SrtB (k cat /K m on the order of 10 1 for SrtA and 10 Ϫ2 for SrtB), this is a significant change. Alteration of the ␤6/␤7 loop alone in SrtA was sufficient to convert it to an NPQTN-cleaving enzyme, with an efficiency of cleavage only 11 times lower than native SrtB.

CONCLUSIONS
Sequence analysis, mutagenesis, and domain swapping have been used as tools to explore specificity and molecular recognition in S. aureus sortase enzymes. Point mutants identified from analysis of SrtA and SrtB sequences did not affect the specificity profile of SrtA, suggesting that these residues play a role in stabilizing the folded structure of the SrtA active site, rather than contacting substrate. However, replacement of the ␤6/␤7 loop in SrtA with the corresponding loop from SrtB was sufficient to change the specificity profile of SrtA by over 700,000-fold. Additional point mutants were unable to boost the NPQTN-cleaving activity of SrtLS. Together, these results support the idea that the ␤6/␤7 loop is a primary substrate recognition site in sortase enzymes. Future studies are ongoing to identify additional sites that are involved in the molecular recognition and catalysis of sortase substrates and those that affect transpeptidation.