Structure of the Bacillus anthracis Sortase A Enzyme Bound to Its Sorting Signal

Background: The Bacillus anthracis sortase A (BaSrtA) enzyme attaches virulence factors to the cell surface. Results: The structure of BaSrtA bound to a substrate analog reveals key amino acids involved in substrate recognition and catalysis. Conclusion: BaSrtA modulates substrate access using a unique N-terminal appendage. Significance: This research could facilitate the design of new anti-infective agents that work by disrupting surface protein display. The endospore forming bacterium Bacillus anthracis causes lethal anthrax disease in humans and animals. The ability of this pathogen to replicate within macrophages is dependent upon the display of bacterial surface proteins attached to the cell wall by the B. anthracis Sortase A (BaSrtA) enzyme. Previously, we discovered that the class A BaSrtA sortase contains a unique N-terminal appendage that wraps around the body of the protein to contact the active site of the enzyme. To gain insight into its function, we determined the NMR structure of BaSrtA bound to a LPXTG sorting signal analog. The structure, combined with dynamics, kinetics, and whole cell protein display data suggest that the N terminus modulates substrate access to the enzyme. We propose that it may increase the efficiency of protein display by reducing the unproductive hydrolytic cleavage of enzyme-protein covalent intermediates that form during the cell wall anchoring reaction. Notably, a key active site loop (β7/β8 loop) undergoes a disordered to ordered transition upon binding the sorting signal, potentially facilitating recognition of lipid II.

Bacillus anthracis is a Gram-positive endospore forming bacterium that causes lethal anthrax disease in humans and animals. Bacterial spores are an infectious form of the pathogen, and predominantly enter the human body through inhalation, ingestion, or cuts in the skin (1). Gastrointestinal (inges-tion) and pulmonary (inhalation) forms of anthrax are often fatal if the illness progresses to the fulminant phase (2). In pulmonary anthrax, inhaled spores are first phagocytized by alveolar macrophages in the lungs, and then transported to the lymph nodes. The spores are resistant to killing and germinate into vegetative cells that enter the bloodstream to cause host damage via secreted toxins. The ability of the microbe to thrive and replicate within macrophages is dependent upon the display of bacterial surface proteins that are attached to the cell wall by the B. anthracis Sortase A ( Ba SrtA) 2 enzyme (3). Understanding the mechanism through which this enzyme displays surface proteins is of fundamental importance and could facilitate the design of new anti-infective agents that work by disrupting surface protein display. Ba SrtA is a member of the sortase superfamily, a large group of extracellular membrane-associated cysteine transpeptidase enzymes that are predominantly found in Gram-positive bacteria (4 -11). At present, over 800 genes encoding sortase proteins, and a much larger number of sortase protein substrates have been identified in ϳ260 species of bacteria (12). The sortase A enzyme from Saccharomyces aureus ( Sa SrtA) has been studied in detail. It attaches proteins to the cell envelop by catalyzing a transpeptidation reaction that joins a C-terminal cell wall sorting signal within its protein substrate to the crossbridge peptide. In this process, a full-length precursor protein containing an amino-terminal leader peptide is exported from the cytoplasm through the secretory (Sec) pathway. The cell wall sorting signal in sortase substrates consists of a LPXTGmotif (where X denotes any amino acid), followed by a segment of hydrophobic amino acids, and a tail composed primarily of positively charged residues. The C-terminal charged tail presumably retards export, positioning the protein for processing by the extracellular membrane-associated Sa SrtA (13). Catalysis occurs through a ping-pong mechanism that is initiated when the active site cysteine residue in Sa SrtA nucleophilically attacks the backbone carbonyl carbon of the threonine residue within the LPXTG motif, breaking the threonine-glycine peptide bond to create a sortase-protein complex in which the components are linked via a thioacyl bond (14,15). The protein is then transferred by Sa SrtA to the cell wall precursor lipid II, when the amino group in this molecule nucleophilically attacks the thioacyl linkage to create a peptide bond-linked protein-lipid II product. Transglycosylation and transpeptidation reactions that synthesize the cell wall then incorporate this product into the peptidoglycan. Other members of the sortase superfamily are believed to operate through a similar enzymatic mechanism as Sa SrtA, but in some microbes (primarily enterococal and corynebacterial species), sortases perform equivalent transpeptidation reactions that assemble pili on the bacterial cell surface (16).
Based on their primary sequences members of the sortase superfamily can be partitioned into at least six subfamilies, called class A to F enzymes (6,17,18). The genome of B. anthracis encodes class A, B, and D sortases, known as Ba SrtA, Ba SrtB, and Ba SrtC, respectively (19). Ba SrtA, the focus of this study, anchors seven proteins to the cell wall by joining the threonine of the LPXTG sorting signal at the C terminus of the protein to the amine group of meso-diaminopimelic acid (DAP) within lipid II (20). Ba SrtB anchors only one protein to the cell wall, IsdC, which is important for heme binding (21), whereas Ba SrtC attaches two proteins involved in sporulation (BasH and BasI) (22).
Atomic structures of class A to D enzymes have now been determined revealing a common eight-stranded ␤-barrel-fold that contains conserved arginine, cysteine, and histidine active site residues (4 -11). However, it is not well understood how sortases recognize their sorting signal and nucleophile substrates, because nearly all sortase structures reported to date have been determined in the apo-state. Previously we discovered that the class A Ba SrtA sortase contains a unique N-terminal appendage that wraps around the body of the protein to contact the active site of the enzyme (23). However, the function of this appendage in catalysis remained unknown. In this article, we report the NMR structure of Ba SrtA bound to a LPXTG sorting signal analog that reveals for the first time the mechanism of substrate recognition by this enzyme. In the structure of the complex, the appendage partially encapsulates the bound sorting signal and we show that these interactions play a role in the in vivo and in vitro function of the enzyme. A model of the thioacyl intermediate constructed from the coordinates of the complex suggests that the N-terminal appendage may obstruct nucleophile access to the active site. We propose that on the cell surface, the N-terminal appendage may increase the efficiency of protein display by reducing the unproductive hydrolytic cleavage of enzyme-protein covalent intermediates that form during the cell wall anchoring reaction.
NOE assignments were obtained automatically using the ATNOS and CANDID algorithms in UNIO' 10 (36, 37), and structure calculations were performed using XPLOR-NIH (38). The three-dimensional 15 N-edited NOESY-HSQC spectrum of the complex dissolved in H 2 O, and three-dimensional 13 C-edited NOESY-HSQC spectrum of the complex dissolved in D 2 O were used as input for UNIO'10. These data were supplemented with dihedral angle restraints and carbon chemical shifts for secondary structure predictions. Seven cycles of ATNOS/CANDID and XPLOR-NIH calculations yielded a converged ensemble of the protein in the complex. All of the NOE assignments made by CANDID were subsequently veri-fied by manually inspecting the NOESY data. In the structure refinement process, additional intra-and intermolecular distance restraints were identified manually, 3 J HN␣ couplings and hydrogen bond restraints were included in structure calculations. Intermolecular NOEs were assigned in an iterative manner. Initially, the positioning of the peptide in the active site was determined using only unambiguous NOEs (intermolecular NOEs that could be readily identified in the filtered and edited NMR spectra of the complex because they were not overlapped with intramolecular protein NOEs). The initial model of the complex was then was used to guide the assignment of more ambiguous intermolecular NOEs. In all cases, only NOEs present in both the filtered and edited NOESY spectra were employed as distance restraints. Hydrogen bonds were identified by inspecting the NOESY data for characteristic patterns along with deuterium exchange experiments, and restraints for hydrogen bonds were implemented with the HBDB algorithm (39). At the final refinement stage, 1 D NH and 1 D NCO residual dipolar couplings were included, which were measured using protein samples partially aligned in PEG C12E5/hexanol (40), using two-dimensional 15 N-coupled IPAP 1 H-15 N HSQC and two-dimensional carbonyl-coupled 1 H-15 N HSQC experiments, respectively (41,42). Axial and rhombic components of the alignment tensor were fitted using Module (43). In the final set of calculations, a total of 200 structures were generated, of which 94 had no NOE, dihedral angle, or scalar coupling violations greater than 0.5 Å, 5°, or 2 Hz, respectively. Of these, 20 structures with the lowest overall energy were chosen to represent the structure of Ba SrtA-LPAT* and have been deposited in the Protein Data Bank (PDB code 2RUI). The programs MOL-MOL (44) and PyMOL (45) were used to generate figures.
Backbone Dynamics of the Ba SrtA-LPAT* Complex Determined from 15 N Relaxation Data-The 15 N relaxation data were collected using a diluted (1.2 mM) 15 N-and 13 C-labeled sample of the complex dissolved in H 2 O on a Bruker Avance 600-MHz NMR spectrometer equipped with a triple resonance cryogenic probe. Data were analyzed using SPARKY (46) and included: 15 N longitudinal relaxation rates (R 1 ), transverse relaxation rates (R 2 ), and { 1 H}-15 N heteronuclear NOEs. Complete quantifiable relaxation information (R 1 , R 2 , and NOE) could be measured for 119 of 151 backbone amides. The average R 1 , R 2 , and NOE values for the complex are 1.47 Ϯ 0.16, 13.14 Ϯ 3.97, and 0.71 Ϯ 0.25 s Ϫ1 , respectively. The relaxation data were analyzed using the suite of analysis programs kindly provided by Prof. Arthur G. Palmer III, and the strategy used to analyze the relaxation data has been described previously (47). Briefly, Pdbinertia was used to calculate the principal moments of inertia and yielded relative moments of 1.00:0.91:0.68. R2R1_tm was used to calculate an approximate correlation time of 9.3 Ϯ 0.4 ns for rigid residues using R 2 /R 1 ratios. To ensure only rigid residues were used in the correlation time analysis, only R 2 /R 1 ratios that met the following criteria were used in the analysis: 1) the residue had a { 1 H}-15 N NOE value Ͼ0.65, and 2) they were within 1 S.D. of the average. The tensor parameters were then calculated using Quadric_Diffusion (48,49), which indicated that the isotropic model is statistically preferred for the complex over the axially symmetric or anisotropic models of tumbling. The relaxation data were then interpreted using the Lipari-Szabo Model-free formalism (50,51) using the program FAST-Modelfree (52) to iteratively run the program Modelfree 4.20 (53). Of the 119 amino acids that gave complete quantifiable relaxation information, data from 98 residues could be fit satisfactorily. Forty-five residues were fit to Model 1 (S 2 only), 3 residues were fit to Model 2 (S 2 and e ), 29 residues were fit to Model 3 (S 2 and R ex ), 5 residues were fit to Model 4 (S 2 , e , and R ex ), and 16 residues were fit to Model 5 (S f 2 , S 2 s , and e ). S 2 parameters could be measured for backbone amides within all regions of the protein, with the exception of residues in the loop that connects helix H1 and strand ␤1 (residues 70 -79), and the latter half of the ␤7/␤8 loop (residues 192-195).
Modeling of the Thioacyl Intermediate-The procedure used to generate the energy minimized model of the thioacyl intermediate has been described previously (54). Briefly, the methylene group and sulfur of the modified threonine were removed in PyMOL (45), and a carbonyl group was placed between the C␣ of the threonine and the sulfur of Cys 187 , with the oxygen on the carbonyl pointed toward or away from Arg 196 . The rough models of the thioacyl intermediate were solvated in a periodic water box with a solvent distance of 10 Å, and parameterized in tLEAP (55). Restraints were placed on the initial atom positions of heavy atoms within the protein and the LPAT peptide, except for the side chain atoms of Arg 196 . Restraints were not placed on the initial coordinates of Arg 196 because the position of its side chain was not well defined in the ensemble due to peak broadening. Models were then energy minimized and equilibrated in NAMD (56) by slowly removing restraints over 1 ns using 2-fs steps. Details about NAMD set up can be found in Jacobitz et al. (54). In the ensemble of NMR structures of the Ba SrtA-LPAT* complex, the side chain of Arg 196 adopts multiple conformations and in some conformers its side chain points away from the bound peptide. Importantly, the starting conformation of the arginine side chain prior to energy minimization does not affect its positioning in the model of the thioacyl intermediate, because when test calculations were performed using different starting structures in which the side chain pointed either away or toward the substrate similar results were obtained. Starting models in which the orientation of the oxygen atom within the thioacyl bond pointed toward, or away from, the active site arginine residue (Arg 196 ) also yielded a similar final model of the intermediate after energy minimization.
Site-directed Mutagenesis and Enzyme Kinetics Measurements-Three single amino acid mutants (S59G, I61A, and C187A) of Ba SrtA were produced for kinetic measurements. Mutations were made using the QuikChange method (Agilent, San Diego, CA) and confirmed by DNA sequencing. Mutants, wild-type Ba SrtA (amino acid residues Asp 57 -Lys 210 ), and wild-type Ba SrtA ⌬64 (amino acid residues Asp 65 -Lys 210 ) were purified as described previously (23). In vitro substrate hydrolysis assay was performed as previously described (57). Briefly, the cleavage of a self-quenched fluorescent peptide, Abz-LPETG-Dap(Dnp)-NH 2 (Peptide 2.0 Inc., Chantilly, VA), was monitored by excitation at 335 nm and recording emission at 420 nm on an Infinite M1000 PRO spectrofluorometer (Tecan, San Jose, CA). Assay conditions consisted of 20 mM HEPES, pH 7.5, and 10 M enzyme. Abz-LPETG-Dap(Dnp)-NH 2 concentrations between 5 and 320 M were added to the reaction for a total reaction volume of 50 l. Fluorescence was recorded for 12 h at 10-min intervals. Initial velocities were calculated in terms of relative fluorescence units/s. Standard curves were used to account for inner filter effects and convert relative fluorescence units to product concentration (14). K m and k cat values were calculated by direct nonlinear fitting three independent sets of data to the Michaelis-Menten equation using SigmaPlot 6.0 (SPSS, Chicago, IL). Cellular SrtA Activity Measured by BasC FLAG/MH6 Display-pAHG322 plasmids (20) with basC FLAG/MH6 and the WT, S59G, or I61A mutant srtA gene were transformed into the B. anthracis AHG188 strain (20) lacking srtA and transformants selected by plating on LB-chloramphenicol (10 g/ml). BasC FLAG/MH6 was collected by growing 200 ml of each isolate to A 600 of 1 in LB (AHG188) or LB-chloramphenicol (10 g/ml, complemented strains). Cells were harvested by centrifugation and resuspended in 5 ml of TSM-lysozyme buffer (50 mM Tris-HCl, pH 7.5, 0.5 M glucose, 10 mM MgCl 2 , 10 mg/ml of lysozyme), shaken at 37°C for 4 h, then sedimented by centrifugation at 15,000 ϫ g for 30 min. Purification of BasC FLAG/MH6 was achieved by applying cleared lysates to 1 ml of Ni-Sepharose (GE Healthcare, Pittsburgh, PA) and eluting with imidazole as described in Gaspar et al. (20). Eluted proteins were precipitated with acetone, subjected to SDS-PAGE, and transferred to a polyvinylidene difluoride membrane. BasC FLAG/MH6 was detected by chemiluminescence using an anti-FLAG polyclonal rabbit antibody (F7425, Sigma) and an anti-rabbit HRP secondary antibody (A0545, Sigma). Immunoblot images were quantified by measuring integrated densities of the BasC bands in ImageJ. Normalized values for each biological replicate were used for statistical analysis, and significance determined by a two-tailed Student's t test.

Results
Structure of the B. anthracis SrtA Enzyme Bound to a Substrate Analog-To investigate how the B. anthracis SrtA sortase recognizes its cognate LPXTG sorting signal, we synthesized a peptide analog of this substrate that covalently modifies the enzyme. This peptide, Boc-LPAT*, contains an N-terminal protecting group t-butyloxycarbonyl (Boc) and a modified threonine (T*) that replaces the carbonyl group with -CH 2 -SH (Fig.  1A). The thiol group on T* promotes the formation of a disulfide bond between the peptide and the active site Cys 187 of the enzyme to generate a covalent complex that structurally mimics the thioacyl intermediate of catalysis (58 -60). In this work, we studied a B. anthracis SrtA protein construct containing residues Asp 57 -Lys 210 ( Ba SrtA). Ba SrtA contains the catalytic core of the enzyme as well as its N-terminal appendage. The solution structure of the Ba SrtA-LPAT* complex was determined using multidimensional heteronuclear NMR and simulated annealing methods. A total of 2,689 experimental restraints define the structure, including: 1,802 intra-protein NOE distance, 59 intermolecular NOE protein-peptide distance, 318 dihedral angle, 77 3 J HN␣ coupling, 282 13 C secondary shift, and 151 residual dipolar coupling restraints. Because the peptide is unlabeled, to generate unambiguous intermolecular NOE protein-peptide distance restraints, we first identified intermolecular NOEs from three-dimensional (F1) 13 C, 15 N-filtered (F2) 13 C-edited NOESY-HSQC and three-dimensional (F1) 13 C, 15 N-filtered (F2) 15 N-edited NOESY-HSQC spectra, which show only intermolecular NOEs, and then we confirmed the intermolecular NOE assignments with two-dimensional (F1) 13 C-filtered NOESY spectrum, which shows both intermolecular and intrapeptide NOEs. An ensemble containing 20 conformers of the structure of the complex is shown in Fig. 2A. All conformers possess good covalent geometries and have no NOE, dihedral angle, or scalar coupling violations greater than 0.5 Å, 5°, or 2 Hz, respectively. In the ensemble, backbone atoms of amino acids Asp 57 -Lys 210 in the protein, and all residues in the bound peptide are well defined; the backbone and heavy atom coordinate root mean square deviation (r.m.s. deviation) to the average structure are 0.42 Ϯ 0.06 and 0.89 Ϯ 0.06 Å, FIGURE 1. The Boc-LPAT* modifier and NMR data of its complex with Ba SrtA. A, the chemical structure of the Boc-LPAT* peptide analog, where T* is (2R,3S)-3-amino-4-mercapto-2-butanol, and Boc is a tert-butyloxycarbonyl protecting group. B, selected panels showing intermolecular NOEs between Ba SrtA and the Boc-LPAT* peptide. Panels are taken from a three-dimensional (F1) 13 C, 15 N-filtered, (F2) 13 C-edited NOESY-HSQC spectrum of the Ba SrtA-LPAT* complex dissolved in deuterated buffer. The identity of the proton from Ba SrtA that gives rise to the set of NOEs and its chemical shift are shown at the top and bottom of each panel, respectively. On the right side of each cross-peak the sorting signal peptide proton that is proximal to the protein is indicated. respectively. Complete structure and restraint statistics are presented in Table 1.
In the structure of the Ba SrtA-LPAT* complex, the catalytic core of the enzyme recognizes the LPAT portion of the sorting signal via a large groove that leads into the active site, and the N-terminal appendage that precedes the catalytic core partially shields the peptide from the solvent (Fig. 2B). On the catalytic domain, the floor of the LPAT binding groove is formed by residues in strands ␤4 and ␤7, and its walls are formed by surface loops that connect strand ␤6 to ␤7 (␤6/␤7 loop), strand ␤3 to strand ␤4 (␤3/␤4 loop), and strand ␤2 to helix H2 (␤2/H2 loop). Within the groove the peptide adopts an "L"-shaped conformation, and it is positioned so as to join the C-terminal end of the signal via a disulfide bond to the active site cysteine residue.
The leucine, proline, and threonine residues within the substrate are recognized by the enzyme. Following convention, residues N-terminal to the Thr-Gly scissile peptide bond in the bound peptide are referred to as P1 to P4, and the corresponding subsites that they interact with on the enzyme are referred to as S1 to S4, respectively (61). The S4 binding site is formed by a pocket created by residues at the N-terminal end of the ␤6/␤7 loop, and strands ␤6 and ␤8. The leucyl side chain within the pocket is contacted by the non-polar side chains of Val 173 and Val 174 in the 3 10 -helix H4, as well as by the ␣-protons of nearby residues in the loop (Pro 168 , Asp 169 , Lys 170 , and Trp 171 ) (Fig.  2C). The bottom portion of the S4 pocket is formed by residues Val 166 , Ile 185 , and Val 198 , which together with the ␤6/␤7 loop encapsulate the leucine so as to bury ϳ90% of the accessible surface area of the side chain. The proline residue forms a ϳ90°k ink, redirecting the trajectory of the peptide toward Cys 187 within the active site. It is recognized by the S3 site, which is formed by side chains projecting from the floor of the groove (Ala 124 and Ile 185 in strands ␤4 and ␤7, respectively), and by residues within the surrounding walls (Val 174 , Val 110 , and Ser 98 in ␤6/␤7 loop, ␤3/␤4 loop, and ␤2/H2 loop, respectively). The proline also makes contacts with the side chain of Ser 59 within the N-terminal appendage, which is poised to interact with its carbonyl group. The alanine residue in the peptide is located at the X position of the LPXTG motif. Notably, its side chain methyl does not appear to be recognized as it is positioned on the surface of the enzyme in between Leu 103 in helix H2 and Ile 61 in the N-terminal appendage. This surface appears to be large enough to accommodate a range of amino acid side chains, consistent with the documented promiscuity of Ba SrtA for this site within the sorting signal (62). The side chain of the threonine residue in the substrate is buried in the S1 pocket formed by the side chains of Ile 185 in strand ␤7, Ala 124 in strand ␤4, and Leu 103 in helix H2 (Fig. 2, C and D). The threonine ␥-methyl group is also packed underneath the imidazole ring of the active site His 126 residue. This interaction is substantiated by the intermolecular NOE cross-peaks between the ␥-methyl group and the ⑀1and ␣-protons of His 126 (Fig. 1B). For a complete list of intermolecular NOEs identified and used in structure calculations, see supplemental Table S1. Extensive chemical shift assignments for apo-Ba SrtA have been reported previously (23) enabling a comparison to the chemical shifts of the Ba SrtA-LPAT* complex. The chemical shifts of both forms of the protein are presented in supplemental Table S2. Compatible with the structure, the most significant chemical shift changes in the protein that result from peptide binding localize to residues within the binding groove and N-terminal appendage that contact the substrate (Fig. 2, E and F).
The N-terminal Appendage Partially Encapsulates the Bound Sorting Signal-The catalytic domain of Ba SrtA is preceded by a structured N-terminal polypeptide segment that is formed by residues Asp 57 -Val 79 . The first eight residues, Asp 57 -Pro 64 , adopt an extended conformation and partially shield the bound peptide and the active site His 126 residue from the solvent. The remainder of the N-terminal appendage contains a short helix (H1, residues Leu 66 -Asn 71 ) and wraps around the surface of the catalytic domain to contact helix H2 and the ␤2/H2 loop. Two residues in the N-terminal appendage interact with the bound peptide, Ser 59 and Ile 61 (Fig. 2D). The non-polar side chain of Ile 61 interacts with the alanine side chain in the bound peptide and it also forms extensive contacts to His 126 in the active site. An inspection of the ensemble suggests that the hydroxyl group of Ser 59 is appropriately positioned to donate a hydrogen bond to the backbone carbonyl oxygen of the proline residue in the bound peptide (present in 30% of the conformers). Several intermolecular NOEs position the N-terminal appendage near  the peptide. These include NOE cross-peaks between: the ␣-proton of Ser 59 and ␤-methyl group of alanine in the peptide, the ␤-protons of Ser 59 and ␤and ␥-protons of the proline residue in the peptide, and the ␣-proton of Ile 61 and ␤-methyl group of alanine in the peptide (Fig. 1B). 15 N Relaxation Measurements: the N-terminal Appendage Transiently Detaches from the Enzyme-To investigate backbone motions within the Ba SrtA-LPAT* complex, we measured the { 1 H}-15 N NOEs, spin-lattice (R 1 ), and spin-spin (R 2 ) relaxation rates of the protein backbone nitrogen atoms, and interpreted the data using the Model-free formalism (63). Of the 151 non-proline amino acids in the protein, complete quantifiable relaxation information (NOE, R 1 , and R 2 ) can be obtained for 119 residues, of which data from 98 of these residues could be interpreted using the Model-free approach. This analysis yields three parameters that describe the motion of the protein backbone: the general order parameter S 2 , the effective correlation time for internal motion e , and R ex . S 2 reports the backbone mobility of the amide on the picosecond time scale and is characterized by the internal correlation time e . The value of S 2 ranges from 0 to 1, with a value of 1 indicating that the amide bond is completely immobilized. R ex is the chemical exchange contribution to R 2 and is diagnostic for the presence of slower micro-to millisecond time scale motions. The R 1 , R 2 , 1 H-15 N NOEs, and S 2 data are shown in Fig. 3, A-D. The complete Model-free analysis results are presented as supplemental Table S3). Notably, similar R 2 values are obtained when data are acquired using a more dilute protein sample (180 M Ba SrtA-LPAT* complex), indicating that the complex is not aggregating. The relaxation data indicate that the N-terminal appendage transiently detaches from the enzyme to adopt a disordered state. This is evident from 1 H-15 N NOE data that shows a gradient of increasing mobility toward the N terminus, presumably because the polypeptide progressively unlatches ("unzips") from the enzyme starting at its N terminus (Fig. 3C). Because the Model-free analysis only indirectly reveals the presence of R ex contributions to relaxation, we also performed constantrelaxation time Carr-Purcell-Meiboom-Gill (CPMG) experiments (64) that probe motions occurring on time scales ranging from 0.1 to 10 ms (63,65,66). However, analysis of 15 N-CPMG relaxation dispersion curves did not reveal the presence of R ex in the N-terminal appendage (data not shown). This indicates that the motions that detach the amino-terminal residues in the appendage from the enzyme occur on a time scale that is faster than 100 s (66).
In contrast to the N-terminal appendage that partially encapsulates the signal, the peptide binding surface on the catalytic domain adopts a rigid conformation. In particular, both the aforementioned ␤7/␤8 and ␤6/␤7 active site loops are rigid on the picosecond time scale according to their S 2 data (S 2 values for residues in these regions are Ͼ0.87). Overall, few residues exhibit significant R ex values in the signal binding pocket. A notable exception are backbone amide atoms of Asp 169 and Glu 172 within the ␤6/␤7 loop that are located next to the N-terminal Boc protecting group of the bound peptide that is likely in motion (Fig. 3E). In addition, several residues near the active site exhibit resonance line broadening that prevented their chemical shift assignment: Ser 97 within the ␤2/H2 loop, His 126 and Asn 127 in the ␤4/␤5 loop, Cys 187 in the end of strand ␤7, and residues Val 188 , Asn 193 , and Lys 195 in the ␤7/␤8 loop. These residues may be broadened as a result of localized protein motions in the active site, binding of the N-terminal appendage and/or changes in the tautomerization state of the imidazole side chain of His 126 .
The N-terminal Appendage Slows Peptide Hydrolysis and Contributes to Protein Display in Intact Cells-To understand the role of the N-terminal appendage in catalysis, a FRET-based assay was used to monitor the ability of wild-type and mutant Ba SrtA proteins to hydrolyze a self-quenched fluorescent peptide substrate (Abz-LPETG-Dap(Dnp)-NH 2 ). Notably, we were unable to reproduce the transpeptidation reaction in vitro using diaminopimelic acid as a nucleophile, which is the moiety on lipid II to which surface proteins are attached. This suggests that sortase recognizes additional features on lipid II. The kinetics of hydrolysis were determined for wild-type Ba SrtA, and three mutants: Ba SrtA ⌬64 , which removes Asp 57 -Pro 64 , and Ba SrtA S59G and Ba SrtA I61A single amino acid mutants that based on the structure should alter interactions between the appendage and the substrate (Fig. 4A). Wild-type Ba SrtA cleaves the peptide with k cat and K m values of 3.6 ϫ 10 Ϫ5 Ϯ 2 ϫ 10 Ϫ6 s Ϫ1 and 306 Ϯ 23 M, respectively (Table 2). In contrast, removal of residues Asp 57 -Pro 64 ( Ba SrtA ⌬64 ) causes a ϳ60% increase in k cat and ϳ43% decrease in K m . Based on the K m change it seems likely that part of the inhibitory effect of the appendage occurs because its residues collectively obstruct peptide binding to the active site when the isolated enzyme is studied. This is substantiated by data from Ba SrtA S59G and Ba SrtA I61A , because these proteins contain mutations that should disrupt appendagesubstrate interactions, but they have K m values similar to the wild-type protein. As described under "Discussion," the finding that the Ba SrtA I61A and Ba SrtA ⌬64 mutants have increased k cat values relative to the wild-type protein is compatible with the amino-terminal appendage acting to slow hydrolysis of the thioacyl intermediate.
To investigate the biological importance of the N-terminal appendage we determined how mutating this segment in Ba SrtA affected its ability to display proteins in intact B. anthracis cells. Protein display was monitored in a ⌬srtA strain of B. anthracis (strain AHG188) that was transformed with plasmid (pAHG322) that expresses wild-type or mutated Ba SrtA, as well as the BasC FLAG/MH6 reporter protein. BasC FLAG/MH6 has previously been shown to be anchored to the cell wall by sortase and contains the full-length BasC substrate of sortase harboring an active cell wall sorting signal, as well as His 6 and FLAG tags that facilitate its fractionation and immunoblot detection, respectively (20). The cell wall of B. anthracis was fractionated and BasC FLAG/MH6 was detected by immunoblotting with anti-FLAG polyclonal antibodies. Fig. 4B shows a representative Western blot, as well as the quantified results from three replicates of the experiment. S59G and I61A mutants of Ba SrtA decreased the amount of BasC FLAG/MH6 found in the cell wall by ϳ40 and ϳ50%, respectively, when compared with wild-type Ba SrtA. Although our results may be complicated by the presence of unprocessed BasC FLAG/MH6 trapped in the membrane, a significant decrease in BasC FLAG/MH6 present in the cell wall fraction indicates that the N-terminal appendage in Ba SrtA plays a role in sortase catalyzed attachment of surface proteins in B. anthracis.

Discussion
Sortase enzymes are used by Gram-positive bacteria to covalently attach protein virulence factors to the bacterial cell surface and therefore are promising drug targets (6,67). An understanding of the mechanism of substrate recognition could facilitate the rational development of sortase inhibitors. Previously, we reported the structure of the SrtA enzyme from B. anthracis ( Ba SrtA) that anchors surface proteins required for bacterial escape from macrophages, a key step in the progression of lethal anthrax disease (3). In contrast to previously studied sortases, Ba SrtA contains a unique N-terminal polypeptide appendage that directly interacts with its active site histidine residue (His 126 ). However, the function of the appendage and its role in signal recognition remained unclear. Here we report NMR studies that show that the appendage in Ba SrtA directly contacts the bound sorting signal, potentially increasing the efficiency of sortase-mediated cell wall protein anchoring by retarding the spurious hydrolysis of a key thioacyl reaction intermediate.

Ba
SrtA binds its sorting signal via a preformed binding groove located in the catalytic domain, which surprisingly triggers a disordered to ordered transition in the adjacent ␤7/␤8 active site loop. The vast majority of enzyme interactions with the signal are mediated by residues located in a structurally ordered binding groove that is formed by residues in ␤6/␤7 loop (Val 166 -Val 174 ), strand ␤4 (Ala 124 ), strand ␤7 (Ile 185 ), ␤3/␤4 loop (Val 110 ), ␤2/H2 loop (Ser 98 ), and helix H2 (Leu 103 ) ( Fig.  2A). In the apo-form of Ba SrtA, residues in the groove adopt a fixed structure and undergo only small positional changes upon ligand binding (the backbone atoms of the residues in the complex that form the pocket are aligned with an r.m.s. deviation of 0.76 Å to the structure of the apo-form of Ba SrtA). In contrast to this preordered binding groove, the ␤7/␤8 loop in the active site experiences large changes in its dynamics and structure when the signal binds (Fig. 5A). In the apo-form of Ba SrtA, the ␤7/␤8 loop is structurally disordered and its residues undergo motions on the micro to millisecond time scales; the backbone amide resonances of nearly all of its residues exhibited substantial motion-induced line broadening (only the backbone amide of Lys 196 could be assigned). In contrast, in the Ba SrtA-LPAT* complex, the coordinates of the ␤7/␤8 loop become well defined and more immobilized. The substrate dependent disorder-to-order transition is evidenced by the NMR relaxation data, our ability to assign the majority of backbone amide resonances of the loop in the NMR spectra of the complex (only the backbone amides of Val 188 , Asn 193 , and Lys 195 could not be assigned), and by the observation of numerous NOEs in the NOESY spectra that define the structure of the loop. It is also supported by NOESY spectra of the free and bound forms of the protein, because key long range and intra-loop NOE cross-peaks that are present in the spectra of the bound enzyme are either absent or significantly less intense (ϳ6-fold less intense) in the spectra of the apo-form. Collectively, these data strongly suggest that binding of the substrate causes the ␤7/␤8 loop to become more structurally ordered and immobilized.
Interestingly, substrate binding indirectly causes conformational ordering of the ␤7/␤8 loop by displacing the polypeptide backbone containing the active site Cys 187 residue. In the complex, the peptide forms a covalent bond with the thiol of Cys 187 located at the end of strand ␤7 immediately preceding the ␤7/␤8 loop. Peptide binding causes the backbone atoms Cys 187 to shift by ϳ7 Å away from the signal, which in turn displaces the ␤7/␤8 loop (Fig. 5A). As a result, the ␤7/␤8 forms stabilizing interactions with residues in the ␤4/␤5 loop; the side chain of Val 190 within the ␤7/␤8 loop packs against Met 128 and Ser 129 within the ␤4/␤5 loop. The structural change in the ␤7/␤8 loop may play an important role in catalysis as this loop has been postulated to form a surface that recognizes the second substrate of catalysis, lipid II (68,69). This may be a common feature of class A enzymes, as similar positional changes in the ␤7/␤8 loop have been observed in the S. aureus SrtA enzyme as a result of substrate binding.
Modeling studies suggest that the appendage may transiently mask the thioacyl intermediate. We modeled the thioacyl enzyme-substrate reaction intermediate using the NMR structure of the Ba SrtA-LPAT* complex. The coordinates of the methylene group and sulfur atom in the modified threonine residue in the sorting signal were replaced by a carbonyl group, and a thioacyl bond to the active site cysteine was established. The coordinates were then energy minimized to yield a model of the thioacyl reaction intermediate. In test calculations, a range of starting structures were energy minimized in which the active site arginine and carbonyl oxygen were positioned differently. In all cases, these calculations yielded the same basic model shown in Fig. 6 (see "Experimental Procedures" for   In addition to the structure of the Ba SrtA-LPAT* complex, two structures of sortase enzymes covalently bound to their sorting signal have been reported: the NMR and crystal structures of S. aureus SrtA and SrtB enzymes bound to LPAT* and NPQT* substrate analogs, respectively (54,68). A detailed comparison of these structures with the Ba SrtA-LPAT* complex is beyond the scope of this article. However, as in the Ba SrtA-LPAT* complex, the proline in these structures causes the sorting signal to adopt an L-shaped conformation that is critical for positioning of the P4 residue. Moreover, in both the Ba SrtA-LPAT* and Sa SrtB-NPQT* structures, the threonine residue in the bound signal points "in" toward the floor of the substrate binding site where it is poised to interact with active site arginine (Arg 196 and Arg 233 , in Ba SrtA and Sa SrtB, respectively). As described in detail in Jacobitz et al. (54) this suggests Ba SrtA catalysis is facilitated by the formation of a substrate-stabilized oxyanion hole that may be a common feature in sortase enzymes.
A unique feature of the Ba SrtA-LPAT* complex is the presence of an N-terminal appendage that directly interacts with the bound peptide. The N-terminal appendage contacts the active site histidine residue and is poised to hydrogen bond to the backbone of the bound signal via the side chain of Ser 59 . The appendage spans residues Asp 57 to Val 79 and contains helix H1. In the structure of the apo-form of the enzyme, the portion preceding helix H1 packs against the imidazole ring of His 126 via contacts from the side chains of Ile 61 and Pro 64 . These interactions, along with contacts from Leu 104 located in helix H2, shield the histidine side chain from the solvent (Fig. 5B). In the Ba SrtA-LPAT* complex, these interactions are largely preserved, except for contacts between His 126 and Ile 61 that are disrupted because the backbone of Ile 61 and residues that precede it in the primary sequence are slightly displaced (Fig. 5B). This movement positions the side chain of Ser 59 near the bound signal, potentiating hydrogen bonding interactions. The NMR relaxation data indicate that in both apo-and signal-bound forms of Ba SrtA the entire N-terminal appendage toggles between two distinct conformers, a closed conformation in which it contacts the catalytic core, and an open conformation in which residues including helix H1 of the appendage are detached from the body of the protein and become structurally disordered. The presence of the equilibrium is supported by the S 2 values of residues in the appendage, because in both the free and substrate-bound forms of the enzyme they exhibit small magnitudes (Fig. 3D). Detachment of the appendage presumably occurs on nanosecond or shorter time scales, because we were unable to detect significant R ex values. Moreover, as the S 2 values progressively become smaller toward the N terminus of the protein the appendage presumably unlatches in a "peeling" motion. As compared with the full-length protein, the construct studied here by NMR is N terminally truncated (it is missing residues 1-56). It is therefore possible that on the cell surface these missing residues could also influence the structure of the N-terminal appendage and/or substrate access to the active site.
Interestingly, binding of the sorting signal may slightly destabilize latching of the N-terminal appendage to the core of enzyme, because the S 2 values of residues preceding helix H1 in the primary sequence positioned near the active site (Ile 61 , Leu 66 , and Ala 67 ) become smaller when the sorting signal binds to the enzyme (Fig. 5C). Three lines of evidence further support this assertion. First, a comparison of the NOESY spectra of the apo-and complexed forms of the protein reveals that substrate binding causes a 3-fold drop in intensity in the NOE crosspeaks that occur between residues located in the N-terminal appendage and the rest of the protein. Second, an analysis of the random coil index (RCI), a method to measure protein flexibility by comparing backbone chemical shifts to reference random coil shifts (70,71), shows that residues in the N-terminal appendage have slightly higher RCI values when the LPAT* substrate analog is bound (Fig. 5D). Although the change in RCI is small, they occur in regions flanking helix H1 suggesting that when LPAT* binds there is an increased likelihood that the entire N-terminal appendage detaches from the protein. Last, a comparison of the structures of the apo-and complexed forms of the protein reveals that the substrate disrupts interactions between the side chains of Ile 61 in the appendage and His 126 in the active site, which presumably destabilize the closed form of the appendage. The biological significance of the destabilizing effects of signal binding on the attachment of the appendage is unclear, because the appendage is dynamic in both the free and bound states, and because the destabilizing effects are small.
Transient closure of the appendage over the sorting signal retards nucleophile access to the thioacyl reaction intermediate in vitro. On the cell surface, the first step in protein anchoring involves the formation of an enzyme-protein intermediate in which the threonine in the sorting signal of the protein substrate is connected to Cys 187 via a thioacyl bond. This intermediate is then nucleophilically attacked by lipid II to form the protein-lipid II linked product of the sortase-catalyzed reaction. Our kinetics data probes the first steps of catalysis and monitors the rate at which the thioacyl bond is hydrolyzed by a water molecule that functions as a surrogate for lipid II (Table 2). Previously reported enzyme kinetic studies of the Sa SrtA enzyme have shown that nucleophilic attack by the water molecule is rate-limiting in the hydrolytic reaction (14,15). Therefore, our observation that k cat increases when a portion of the N-terminal appendage is removed suggests that its presence hinders nucleophile attack of the thioacyl bond. This notion is compatible with the increased k cat of the I61A mutant as this alteration presumably disrupts contacts that stabilize positioning of the appendage over the active site. Importantly, the time scale of latching and unlatching of the appendage over the bound sorting signal is significantly shorter than the rate-limited hydrolysis step. Thus, in the isolated enzyme studied here, the N-terminal appendage repeatedly latches and unlatches over the thioacyl-linked sorting signal to hinder, but not completely prevent nucleophile access to the substrate.
The results of computational docking and whole cell display studies suggest that on the cell surface the N-terminal appendage may modulate lipid II access to the thioacyl intermediate. How sortases recognize lipid II remains unknown. To investigate the mechanism of binding we used the program FTSite to predict the binding site for meso-DAP, the amino group con-taining moiety in the B. anthracis lipid II molecule that is joined to the sorting signal by sortase (72,73). Two binding calculations were performed, one using the intact Ba SrtA-LPAT* complex, and a second in which the coordinates of the N-terminal appendage are removed. Only when a portion of the appendage is removed from Ba SrtA does FTSite predict the existence of a binding site for DAP that is located near the active site (compare Fig. 7, A and B). This site is located adjacent to His 126 , and the ␤7/␤8 and ␤4/H3 loops. Attractively, the site is near the electrophilic carbonyl carbon atom in the thioacyl bond, which could be accessed by the nucleophile with only minor rearrangements in the active site. Interestingly, only a few residues in the N-appendage (residues Asp 57 -Pro 64 ) need to be displaced from the enzyme to unmask the DAP binding site such that helix H1 in the appendage could in principle remain attached during the transpeptidation reaction. The notion that residues Asp 57 -Pro 64 in the N-terminal appendage modulate substrate access is consistent with our whole cell studies, which revealed that S59G and I61A mutations in this segment of the appendage reduced display by ϳ50 and 40% as compared with the wildtype enzyme, respectively (Fig. 4B).
Similar to Ba SrtA, many class C enzymes that polymerize pili contain an N-terminal appendage that interacts with the active site. This appendage, called a "lid," is structurally distinct from the Ba SrtA appendage, and it approaches the active site from a different direction. Moreover, unlike the appendage in Ba SrtA, the lid more fully occludes the sorting signal binding site (74). When Ba SrtA is located in its natural context on the extracellular membrane, it is conceivable that its N-terminal appendage forms more extensive interactions with the body of the enzyme similar to the lid structure, which could cause it to fully occlude the binding site for the sorting signal. In this scenario, the N-terminal appendage might have a function similar to the lid, holding the sortase in an inactive state until factors on the cell surface detach it to promote sorting signal binding. Our modeling and biochemical studies also suggest that closure of the N terminus over the bound sorting signal hinders nucleophile access to the thioacyl bond that joins sortase to its protein substrate during the anchoring reaction. Thus, on the cell surface, the appendage may prevent unproductive hydrolysis of the thioacyl-linked enzyme-protein intermediate by masking the thioacyl bond from the solvent until lipid II, or other factors become available to complete the transpeptidation reaction. A deeper understanding of how sortase enzymes are regulated and the mechanism of catalysis will require structure determination of additional reaction intermediates that contain the substrate nucleophile, as well as a systematic investigation of how the enzyme operates on the cell surface.