Staphylococcus aureus Sortase A Transpeptidase

Many virulence factors in Gram-positive bacteria are covalently anchored to the cell-wall peptidoglycan by sortase enzymes, a group of widely distributed cysteine transpeptidases. The Staphylococcus aureus Sortase A protein (SrtA) is the archetypal member of the Sortase family and is activated by Ca2+, an adaptation that may facilitate host colonization as elevated concentrations of this ion are encountered in human tissue. Here we show that a single Ca2+ ion bound to an ordered pocket on SrtA allosterically activates catalysis by modulating both the structure and dynamics of a large active site loop. Detailed nitrogen-15 relaxation measurements indicate that Ca2+ may facilitate the adaptive recognition of the substrate by inducing slow micro- to millisecond time-scale dynamics in the active site. Interestingly, relaxation compensated Carr-Purcell-Meiboom-Gill experiments suggest that the time scale of these motions is directly correlated with ion binding. The results of site-directed mutagenesis indicate that this motional coupling is mediated by the side chain of Glu-171, which is positioned within the β6/β7 loop and shown to contribute to Ca2+ binding. The available structural and dynamics data are compatible with a loop closure model of Ca2+ activation, in which the β6/β7 loop fluctuates between a binding competent closed form that is stabilized by Ca2+, and an open, highly flexible state that removes key substrate contacting residues from the active site.

Surface proteins on bacteria are frequently virulence factors, promoting bacterial adhesion, resistance to phagocytic killing, and host cell invasion during infection. In Gram-positive bacteria these proteins are often covalently anchored to the cell wall by sortase enzymes, a family of novel cysteine transpeptidases (1)(2)(3). The sortase A protein (SrtA) 2 from Staphylococcus aureus has been characterized extensively (4) and anchors proteins bearing a cell wall sorting signal that consists of a conserved LPXTG motif (where X is any amino acid), a hydrophobic domain, and a tail of mostly positively charged residues (4 -6). SrtA cleaves in between the threonine and glycine of the LPXTG motif (7) and catalyzes the formation of a peptide bond between the carboxylgroup of the threonine and the amine-group of the cell-wall precursor lipid II (7)(8)(9). The lipid II-linked protein is then incorporated into the peptidoglycan of the cell wall via the transglycosylation and transpeptidation reactions of bacterial cell-wall synthesis. Sortases represent an attractive target for new anti-infective agents, because they are widely distributed among a variety of bacterial pathogens (10, 11) (e.g. Bacillus anthracis, Listeria monocytogenes, Streptococcus pneumoniae, and Streptococcus pyogenes), and have been shown to be required for virulence (12)(13)(14)(15)(16).
The catalytic domain of SrtA (SrtA ⌬N59 , residues 60 -206) adopts a conserved eight-stranded ␤-barrel fold (17,18). The active site is organized around the catalytically essential side chain of Cys-184, whose thiolate nucleophilically attacks the threonine carbonyl carbon within the LPXTG sorting signal, forming a thioester linkage between the enzyme and substrate (19). In addition to Cys-184, the hydrophilic side chains of His-120 and Arg-197 are absolutely required for catalysis (20 -22). These residues likely participate in general acid/base catalysis, and one of them must activate the thiol for nucleophilic attack, because it is protonated at neutral pH (23). The indole ring of Trp-194 partially shields the cysteine thiol from the solvent, and its mutation to alanine reduces enzyme activity 4-fold through an unknown mechanism (20). Using NMR and crystallography, the LPXTG sorting signal binding site has recently been localized to a surface formed by strands ␤4 and ␤7, and to a proximal loop that connects strands ␤6 to ␤7(the ␤6/␤7 loop) (18,22). Substrate binding may occur through an induced-fit mechanism involving conformational changes in the ␤6/␤7 loop, because it is disordered in the absence of the sorting signal substrate (17,18).
Ca 2ϩ stimulates the activity of SrtA ⌬N59 in vitro (17) and may enable S. aureus to increase the rate of surface protein anchoring as it encounters elevated concentrations of this ion at sites of infection. Because many surface proteins function as virulence factors, the stimulatory effect of Ca 2ϩ likely plays an important role in the infection process. Previously we showed that Ca 2ϩ bound to an ordered pocket positioned distal to the active site (hereafter referred to as the ␤3/␤4 pocket) (17). However, this work did not reveal how Ca 2ϩ stimulated enzyme activity. Here we show using enzyme kinetic and detailed NMR nitrogen-15 measurements that a single Ca 2ϩ ion bound to the ␤3/␤4 pocket promotes substrate binding. We provide evidence that ion binding to this distal site allosterically controls enzyme activity by altering motions in the active site ␤6/␤7 loop. In particular, we show that Ca 2ϩ retards motions and induces slow micro-to millisecond time-scale dynamics within the loop that may promote the adaptive recognition of the substrate. The results of relaxation compensated Carr-Purcell-Meiboom-Gill (CPMG) experiments suggest that the active site motions are correlated with the rate of ion binding. The results of site-directed mutagenesis suggest that this motional coupling is mediated by the side chain of Glu-171, which is located at the C-terminal end of the ␤6/␤7 loop and required for high affinity ion binding. This work represents the first NMR dynamics study of a sortase enzyme and reveals that complex conformational dynamics contribute to the function of these enzymes. 15 N-enriched SrtA ⌬N59 (residues 60 -206) was obtained as previously described (17). Three separate samples of 3 mM 15 N SrtA ⌬N59 were prepared for relaxation studies by dissolving weighed lyophilized protein in 500 l of 50 mM Tris-HCl, 100 mM NaCl, 3 mM dithiothreitol, 7% D 2 O, and 0.01% NaN 3 . The pH of the sample was adjusted to 6.20 (uncorrected for the deuterium effect). Each sample differed in the amount of Ca 2ϩ present: 1) no Ca 2ϩ (apo-Ca 2ϩ SrtA ⌬N59 ), 2) 20 M of CaCl 2 , or 3) 20 mM Ca 2ϩ (Ca 2ϩ -bound SrtA ⌬N59 ). Where needed, the water used for the samples was preconditioned with Chelex resin. The Mn 2ϩ and Ca 2ϩ titration experiments were conducted on similar samples of 15 N SrtA ⌬N59 using defined amounts of these ions. The histidine-tagged single amino acid mutants of sortase were purified using a nickel column and exchanged into the appropriate buffer for enzymatic and NMR studies.

Protein Preparation and Purification-Uniformly
Modification of SrtA ⌬N59 by the Peptidyl-Sulfhydryl Compound-A peptidyl-sulfhydryl compound (benzoyloxycarbonyl-Leu-Pro-Ala-Thr with a C-terminal -CH 2 SH group) was used to modify SrtA ⌬N59 (the synthesis of the compound will be reported elsewhere). A 5-fold molar excess of the compound was added to two samples of SrtA ⌬N59 (500 l of a 20 M protein solution) in buffer I (pH 8.0, 50 mM Tris-HCl, and 100 mM NaCl), one without Ca 2ϩ and one containing 20 mM CaCl 2 . The mixture was incubated on a rotating wheel at room temperature, and samples were removed periodically and analyzed using reverse phase high-performance liquid chromatography on a C18 column (Waters, Mildford, MA). The areas under the peaks corresponding to modified and unmodified SrtA in the chromatogram were integrated to calculate the percentage of modification at each time point.
Site-directed Mutagenesis-Wild-type SrtA ⌬N59 plasmid containing a C-terminal 6-His tag was generated as described previously (24). Single amino acid mutations of SrtA ⌬N59 were generated by PCR using the SrtA ⌬N59 plasmid template and Pfu Turbo DNA polymerase (Stratagene). The Glu-108 codon was mutated to encode Ala-108 using the primers E108A-F (TAAGCTTTGCAGAAGAAAATGCATCACTA-GATGATCAAAATATTT) and E108A-R (AAATATTTTGATCAT-CTAGTGATGCATTTTCTTCTGCAAAGCTTA). The Asp-170 codon was mutated to encode Ala-170 using the primers D170A-F (A-CAGATGTAGGAGTTCTAGCAGAACAAAAAGGTAAAGATAA) and D170A-R (TTATCTTTACCTTTTTGTTCTGCTAGAACTCC-TACATCTGT). The Glu-171 codon was mutated to encode Ala-171 using the primers E171A-F (CAGATGTAGGAGTTCTAGATGCAC-AAAAAGGTAAAGATAAACAATT) and E171A-R (AATTGTTTA-TCTTTACCTTTTTGTGCATCTAGAACTCCTACATCTG) (underlined nucleotides mark mutational changes). The identity of the mutants was confirmed by DNA sequencing. Purification of the wild-type and mutant enzymes was performed as previously described (24). Enzyme Kinetics Measurements-A self-quenched fluorescent peptide, o-aminobenzoyl-LPETG-2,4-dinitrophenyl, was used as a substrate in the cleavage reaction as previously described (25). Reactions contained 1.5 M SrtA ⌬N59 enzyme in the assay buffer (20 mM HEPES, pH 7.5, with various concentrations of CaCl 2 ). The aminobenzoyl-LPETG-dinitrophenyl substrate was dissolved in Me 2 SO and added to the reaction to a final concentration between 6.25 and 25 M, for a total reaction volume of 200 l. The increase in fluorescence intensity was monitored at room temperature using excitation at 335 nm and recording the emission maximum at 420 nm on a SPEX spectrofluorometer (Photon Technology International, Lawrenceville, NJ). The steady-state velocities (V s ) from the biphasic progress curves were calculated as described previously (25). The progress curves were fit to the following equation, where is the amplitude of the burst phase. The kinetic parameters were calculated from the substrate dependence of V s as described previously (25).
Relaxation Data Acquisition and Processing-NMR data were acquired on Bruker Avance 500-and 600-MHz spectrometers equipped with 5-mm triple resonance cryo-probes and single axis pulsed field gradients. The chemical shift assignments have been reported previously for SrtA in the presence of 20 mM CaCl 2 (17) (BioMagResBank Code 4879), and near complete backbone assignments for the apo-Ca 2ϩ form of the protein were obtained using standard triple resonance techniques. 3 The 15 N spin-lattice/longitudinal (R 1 ), and spin-spin/transverse (R 2 ) relaxation rates, and the steady-state heteronuclear { 1 H}- 15 N NOE values were measured as previously described (26 -28) , and 138 ms. The NOE experiment was carried out in an interleaved manner, with and without proton saturation and repeated thrice under identical conditions. All experiments were acquired with 2048 ϫ 256 complex points in the F2 and F1 dimensions with corresponding spectral widths of 10,000 and 2,027 Hz, and the proton carrier frequency was set to the water resonance. The data sets were processed using nmrPipe (29), and peak heights were determined using Sparky (30) operating on a Linux work station.
The following programs used to analyze the relaxation data were kindly provided by Prof. Arthur G. Palmer 3rd at Columbia University: Curvefit, Pbdinertia, R2R1_Diffusion, Quadric Diffusion, and Modelfree 4.01. The rate constants were evaluated using the program Curvefit, assuming mono-exponential decay of the peak intensities. The errors in peak intensities were calculated from the two duplicate experiments using a Perl language script kindly provided by Prof. Patrick Loria at Yale University. The error in the Heteronuclear NOEs was calculated by propagating the base-plane noise as calculated from the signal-to-noise ratio and is the average derived from three experiments. The 1 H-15 N HSQC spectra of SrtA ⌬N59 in both the absence and presence of Ca 2ϩ are well resolved, enabling the reliable measurement of relaxation parameters for 92 and 98 residues respectively, out of a total of 148. Optimization of Diffusion Tensor-Multiple approaches were used to assess the overall motion of SrtA ⌬N59 , because motional anisotropy also contributes to the R 2 values, and if miss-identified can cause erroneous internal correlation times ( e ) and R ex values (31,32). The principal moments of the inertia tensor were calculated using the program Pdbinertia, and the values are 1.00:0.90:0.80 and 1.00:0.90:0.78, for the NMR solution structure of Ca 2ϩ -bound SrtA ⌬N59 (PDB accession code: 1IJA) and the Ca 2ϩ -free crystal structure (PDB accession code: 1T2P) after the addition of hydrogen atoms using the program MolMol (33), respectively. Similar moments were obtained using the program HydroNMR version 5a (34), where the proteins were assumed to be hydrated with a 3.1 Å water shell (1.00:0.88:0.84 for Ca 2ϩ -bound SrtA ⌬N59 and 1.00:0.88: 0.86 for apo-Ca 2ϩ SrtA ⌬N59 ). The latter analysis also provided an estimate of the molecular correlation time ( m ) of each form of the protein: 11.03 and 9.89 ns for apo-Ca 2ϩ and Ca 2ϩ -bound SrtA ⌬N59 , respectively. Because the relative moments for both apo-Ca 2ϩ and Ca 2ϩ -bound SrtA ⌬N59 vary significantly from a perfect sphere, the approach outlined by Tjandra et al. (31) was used to check the statistical significance of fitting the relaxation data to an axially symmetric model versus an isotropic model (R2R1_Diffusion program). These calculations were performed using data from residues whose R 2 /R 1 ratios were within one standard deviation from the average R 2 /R 1 ratio, and which had NOE ratios Ͼ 0.65 (35). The results from 500 MHz (and 600 MHz) data suggested a m of 9.69 ns (9.54 ns) for Ca 2ϩ -bound SrtA ⌬N59 and a m of 10.35 ns (9.94 ns) for apo-Ca 2ϩ SrtA ⌬N59 . The data also showed a statistically significant improvement when using the axial symmetric model over the simple isotropic diffusion model. Interestingly, the Ca 2ϩ -bound SrtA ⌬N59 data fit best to a prolate ellipsoid, whereas the apo-Ca 2ϩ SrtA ⌬N59 data fit best assuming an oblate ellipsoid. This variation is mostly due to subtle differences in the ␤6/␤7 loop orientation in the two structures. The tensor parameters were also calculated using the approach outlined by Bruschweiler et al. (36 -38) using the program Quadric_Diffusion, which gave slightly elevated correlation times of 9.86 ns (9.80 ns) for Ca 2ϩ -bound SrtA ⌬N59 and 10.62 ns (10.17 ns) for apo-Ca 2ϩ SrtA ⌬N59 from 500 MHz (and 600 MHz) data, but the axial symmetric model was preferred over the isotropic or fully anisotropic models. The results from this final calculation were used as an initial guess for Modelfree analysis described below.
Modelfree Analysis-The amplitudes and effective correlation times for a protein's internal motions can be extracted from relaxation data using the Lipari-Szabo Modelfree formalism (39,40). The analysis considers five semi-empirical forms of the spectral density function with each form composed of terms describing the motion of the N-H bond vector. This internal motion can be assumed to occur on two different, fast and slow time scales and can be characterized by effective correlation times, f and s (where f Ͻ Ͻ s Ͻ Ͻ m ) and the square of order parameters, S f 2 and S s 2 . The square of the generalized order parameters is defined as S 2 ϭ S f 2 S s 2 and corresponds to the spatial restriction of the N-H bond vector (where 0 Յ S 2 Յ 1). The analysis also accounts for line broadening due to chemical exchange, R ex . All these motional parameters were fit to the spin-relaxation data using the program Modelfree 4.01 (41,42). For each model, 500 randomly distributed data sets were generated, and model selection was done using a statistical testing protocol described by Mandel et al. (42). Initially, models were selected at fixed diffusion tensor parameters by comparing the sum-squared error of optimal fit with the 0.05 critical value of the distribution and wherever applicable, by F-test comparisons to the 0.20 critical value of the distribution. In the next step, after a model had been assigned to each spin, both the diffusion tensor and model parameters were optimized simultaneously. We used an N-H bond length of 1.02 Å and 15 N chemical shift anisotropy values of Ϫ160 ppm in our backbone spin calculations. For the sole tryptophan (Trp-136) side-chain spin, a chemical shift anisotropy value of Ϫ126 ppm was used (43). A minimum 3% base error is assumed in all parameters in the Modelfree analysis (44,45). Out of 92 quantifiable residues in apo-Ca 2ϩ SrtA ⌬N59 , 89 could be satisfactorily fit using Modelfree analysis (the relaxation data of Ser-70, Glu-95, and Ile-123 could not be fit). Model 1 (S 2 -only) was an appropriate fit for 63 residues, 3 residues fit to model 2 (S 2 and e ), 13 residues fit to model 3 (S 2 and R ex ), 2 fit to model 4 (S 2 , e , and R ex ), and 11 residues fit to model 5 (S f 2 , S s 2 , and e ). In the Ca 2ϩ -bound SrtA ⌬N59 data, 94 out of a total of 98 quantifiable spins could be fit satisfactorily (the exceptions were Glu-95, Arg-99, Ile-123, and Gln-172). In the Ca 2ϩ -bound SrtA ⌬N59 analysis, Model 1 (S 2 ) was selected for 75 residues, 3 residues fit to model 2 (S 2 and e ), 12 residues fit to model 3 (S 2 and R ex ), 2 fit to model 4 (S 2 , e , and R ex ), and 6 residues fit to model 5 (S f 2 , S s 2 , and e ). Interestingly, in both Ca 2ϩ -bound and apo-Ca 2ϩ SrtA ⌬N59 , the relaxation data from Glu-95 and Ile-123 could not be fit to any model, suggesting that they undergo more complicated motions. The results of the model-free analysis for both conditions are provided in supplemental Tables S1 and S2. The average order parameters for apo-Ca 2ϩ and Ca 2ϩ -bound SrtA ⌬N59 are 0.88 Ϯ 0.14 and 0.89 Ϯ 0.10, respectively. If only residues located in regions of regular secondary structure are considered, the order parameters are 0.943 Ϯ 0.030 and 0.935 Ϯ 0.016, for apo-Ca 2ϩ and Ca 2ϩbound SrtA ⌬N59 , respectively.

RESULTS AND DISCUSSION
Ca 2ϩ Promotes Substrate Binding-To identify the step(s) at which Ca 2ϩ stimulates catalysis, we monitored the in vitro activity of SrtA ⌬N59 using an internally quenched fluorescent substrate analogue (o-aminobenzoyl-L-P-E-T-G-2,4-dinitrophenyl) (25). The SrtA ⌬N59 mediated cleavage of the peptide bond between the threonine and glycine residues in this substrate results in an increase in fluorescence and can be used to monitor both hydrolysis and transpeptidation. The presence of Ca 2ϩ stimulates both the hydrolytic and transpeptidation activities of SrtA ⌬N59 (data not shown). Because both reactions differ only in the nucleophile used to resolve the acyl-intermediate, these results indicate that ion binding activates an early step in catalysis (e.g. sorting signal binding, activation of the Cys-184 thiol, or resolution of the enzymesorting signal thioacyl intermediate). To further pinpoint the step at which it acts, we monitored the hydrolytic activity of SrtA ⌬N59 in the presence of varying amounts of Ca 2ϩ . The hydrolysis reaction can be represented as in Reaction 1, where E, E⅐RCO-X, and RCO-E represent the free enzyme, the enzymebound to the sorting signal, and the acyl enzyme⅐substrate complex, respectively; RCO 2 H and XH are the cleaved peptide and the free glycine, respectively; and k 1 , k Ϫ1 , k 2 , and k 4 are the rate constants describing their interconversion (25). Lineweaver-Burk plots of the hydrolysis data recorded at varying Ca 2ϩ concentrations reveal a common 1/V s intercept, indicating that the k cat of the reaction is unaffected (Fig. 1A). It can also be concluded from this data that the values of k 2 and k 4 are independent of Ca 2ϩ , because k cat is defined as k 2 * k 4 /[k 2 ϩ k 4 ]. In contrast, the presence of Ca 2ϩ clearly alters the Michaelis-Menten constant (K m ) of the reaction, because the slopes of the data are inversely proportional to K m * k cat , and the value of k cat is independent of Ca 2ϩ . Because K m is defined as [k 2 ϩ k Ϫ1 ]/k 1 , and k 2 is unchanged by Ca 2ϩ , we conclude that Ca 2ϩ stimulates SrtA ⌬N59 activity by promoting substrate binding; it either increases k 1 or decreases k Ϫ1 .
The ␤3/␤4 Pocket of SrtA ⌬N59 Binds One Ca 2ϩ Ion with Millimolar Affinity-Previously we used NMR to localize a Ca 2ϩ binding site on SrtA ⌬N59 to residues within the loop connecting strands ␤3 to ␤4 (the ␤3/␤4 pocket) (17). However, the effects of Ca 2ϩ are extensive, and, in addition to the ␤3/␤4 pocket, dramatic chemical shift changes were observed in a large adjacent loop that connects strands ␤6 to ␤7 (the REACTION 1

Dynamics of Calcium Modulated Loop Closure of Sortase A
␤6/␤7 loop). Because many of the perturbed residues in the ␤6/␤7 loop are far away from the ␤3/␤4 pocket, their shift changes cannot be caused by direct effects (see Fig. 4B of Ref. 17). Instead, either additional ions bind to the ␤6/␤7 loop, and/or ion binding to the ␤3/␤4 pocket triggers a conformational rearrangement that perturbs the chemical shifts of residues within the ␤6/␤7 loop. Because our previous work did not distinguish between these two possibilities, we precisely localized the divalent ion binding site(s) on SrtA ⌬N59 by titrating the enzyme with Mn 2ϩ and monitoring its 1 H-15 N HSQC NMR spectrum. It is expected that Mn 2ϩ will bind to the same site(s) as Ca 2ϩ , because they are both divalent ions and because Mn 2ϩ also stimulates enzyme activity (albeit to a lesser extent) (17). Importantly, titrations with Mn 2ϩ enable precise definition of the ion-binding pocket, because it is paramagnetic, such that even at sub-saturating levels it selectively broadens only nearby protein resonances (46). A series of 1 H-15 N HSQC spectra of apo-Ca 2ϩ SrtA ⌬N59 were recorded in the presence of varying concentrations of Mn 2ϩ (0, 25, 50, 100, 200, 500, and 1000 M). When 50 M Mn 2ϩ is added, several 1 H-15 N resonances within the ␤3/␤4 pocket selectively disappear (Fig. 2B), whereas at higher Mn 2ϩ concentrations, residues adjacent to surface-exposed acidic side chains exhibit nonspecific line broadening ( Fig. 2A). Only a single Ca 2ϩ ion binds to the ␤3/␤4 pocket as a superposition of a series of 1 H-15 N HSQC spectra recorded with varying amounts of Ca 2ϩ reveals linear changes in the protein's chemical shifts (Fig. 2C), whereas non-linear changes are expected if multiple ions were to bind (47). Moreover, an analysis of a plot of the Ca 2ϩ chemical shift changes versus [Ca 2ϩ ]/[SrtA ⌬N59 ] reveals a binding ratio of 1:1 (data not shown). Taken together, these data show that the aforementioned ␤3/␤4 pocket binds to a single Ca 2ϩ ion and that it is the only high affinity ion binding site on sortase. The extensive Ca 2ϩ -dependent chemical shift changes in the ␤6/␤7 loop can therefore be attributed to a structural rearrangement within or nearby this region (Fig. 3A).
The affinity of the ␤3/␤4 pocket for Ca 2ϩ has not been determined quantitatively and is needed for NMR dynamics and kinetic studies. Values for the dissociation constant were therefore estimated from the Ca 2ϩ dependence of the apparent K m (Fig. 1B) where K m max and K m min are the K m values of the enzyme in the absence and presence of Ca 2ϩ , respectively, and K d is the dissociation constant for ion binding (48). Curve fitting of the data reveal that saturating amounts of Ca 2ϩ cause a 4.2-fold decrease in the K m from 121 M to 33 M and that the ion binds with a K d of 1.6 Ϯ 0.3 mM. A similar binding constant of 2.2 Ϯ 0.5 mM is obtained by directly curve fitting the Ca 2ϩ dependence of the chemical shift data (Fig. 2D). Because the concentration of Ca 2ϩ is ϳ2.5 mM in human serum, these results suggest that about half of the SrtA molecules on the surface of S. aureus are Ca 2ϩbound during bacteremia. However, this is likely an underestimate because the binding of the substrate and ion to the protein presumably forms a closed thermodynamic cycle, which necessitates that the substrate-bound enzyme exhibit ϳ4-fold higher affinity for Ca 2ϩ as compared with the substrate-free enzyme (49).
Ca 2ϩ Does Not Directly Interact with the Sorting Signal-NMR, x-ray, and targeted mutagenesis studies have localized the sorting signal binding site to a large hydrophobic surface immediately adjacent to the ion-binding pocket (18,22), raising the possibility that direct ion-substrate interactions stimulate catalysis. An inspection of the recently determined x-ray structure of a Cys-184 3 Ala mutant of SrtA ⌬N59 ( C184A SrtA ⌬N59 ) bound to a LPETG peptide reveals that the side chain of the central glutamic acid in the peptide is nearest to the ion-binding pocket (18). However, this structure cannot reveal whether the ion directly contacts the substrate, because it was solved in the absence of Ca 2ϩ . Because the stimulatory effect of divalent cations has only been demonstrated using a fluorogenic substrate that also has glutamic acid at this position, we wondered whether acidic residues at the central position within the substrate were needed for Ca 2ϩ stimulation. To answer this question, the effect of Ca 2ϩ on the rate of enzyme modifi-  Table 1. C, graph depicting the effect of Ca 2ϩ (20 mM) on the rate of SrtA ⌬N59 modification by a peptidyl-sulfhydryl compound containing the sorting signal sequence LPAT. The compound modifies SrtA by forming a disulfide bond with Cys-184, changing the retention time of the protein on reverse phase-high-performance liquid chromatography. The percentage of SrtA ⌬N59 modified was calculated by integrating the peaks corresponding to modified and unmodified SrtA ⌬N59 in the chromatogram in the presence (F) and abscence (E) of Ca 2ϩ . FIGURE 3. A, schematic showing Ca 2ϩ ion coordination by SrtA ⌬N59 . Ca 2ϩ ion binding site(s) predicted previously (17) was precisely localized by Mn 2ϩ titration experiments monitoring its 1 H-15 N HSQC NMR spectra and by inspection of the NMR solution structure of the Ca 2ϩ -bound form of the protein (PDB accession code: 1IJA). These new data indicate that the previously predicted ␤3/␤4 pocket binds to a single Ca 2ϩ ion, and it is the only high affinity Ca 2ϩ binding site on SrtA ⌬N59 . B, overlay of the NMR (pink) and crystal (gray) structures solved in the presence and absence of Ca 2ϩ , respectively. This figure shows that only the structure of the loop connecting strands ␤6 to ␤7(the ␤6/␤7 loop) is dramatically affected upon Ca 2ϩ binding. The side chains of Val-168 and Leu-169 (within the ␤6/␤7 loop) are colored pink and gray in the NMR and crystal structures, respectively. The side chains of active site residues and Ca 2ϩ -ligating residues are shown in blue and green, respectively. C, a ribbon diagram of the crystal structure of the complex between C184A SrtA ⌬N59 and a peptide containing the sequence LPETG. Regions 1-3 of the ␤6/␤7 loop are colored yellow, red, and green, respectively. The ␤3/␤4 pocket, which is responsible for chelating the Ca 2ϩ ion, is shown in light brown. The LPETG peptide is shown in purple, and the side chains of Pro-163 and the catalytically important Arg-197 are shown in yellow and pink, respectively. cation by a peptidyl-sulfhydryl compound containing the sorting signal sequence LPAT was tested (benzoyloxycarbonyl-Leu-Pro-Ala-Thr, with the C-terminal carbonyl group of Thr replaced with -CH 2 -SH). This compound modifies SrtA ⌬N59 by forming a disulfide bond with Cys-184 within the active site (22). As shown in Fig. 1C, the presence of Ca 2ϩ increases the rate at which it modifies the enzyme, presumably by promoting its binding similar to the substrate. Because the hydrophobic alanine side chain cannot serve as a ligand for Ca 2ϩ , we conclude that interactions from the central position in the substrate are not required for the stimulatory effect of Ca 2ϩ . This finding is substantiated by recent studies by the McCafferty group that have demonstrated that peptide substrates in which the glutamic acid is replaced by a range of amino acids are still effectively cleaved by SrtA (50). Moreover, we have recently determined the solution structure of the sortase⅐peptide com-plex in the presence of Ca 2ϩ , which shows that the substrate and ion to do not directly interact with one another. 4 Nitrogen-15 Relaxation Measurements-The proximity of the ␤6/␤7 loop to the active site and its extensive Ca 2ϩ -dependent chemical shift changes argue that it undergoes an ion-induced structural change that promotes substrate binding. Interestingly, structural data also suggest that the loop is flexible, because in both the NMR and x-ray structures, its residues exhibit elevated root mean square deviations (Fig. 4E) and B-factors (Fig. 4F), respectively (17,18). This raises the interesting possibility that Ca 2ϩ also modulates the flexibility of the loop to stimulate substrate binding. Because nothing is known about the conformational The average order parameter for apo-Ca 2ϩ SrtA ⌬N59 is 0.88 Ϯ 0.14, whereas for Ca 2ϩ -bound SrtA ⌬N59 , it is 0.89 Ϯ 0.10. The order parameters for each form are mapped onto ribbon diagrams of SrtA ⌬N59 in panels C (apo-Ca 2ϩ ) and D (Ca 2ϩ -bound). The width of each tube connecting the secondary structural elements is inversely proportional to the size of the S 2 value (i.e. wider regions indicate more flexible parts of the protein). Because residues in regular secondary structure exhibited uniformly high S 2 values, they are represented in blue by arrows (␤ sheets) and coils (␣ helices) of uniform size. The catalytic (His-120, Cys-184, and Arg-197) and Ca 2ϩ binding side chains are displayed in green. The relaxation data are mapped onto the coordinates of the NMR structure of Ca 2ϩ -bound protein. E, representative histogram of the root mean square deviations of the local backbone assignments of the SrtA ⌬N59 solution structure (17). F, representative histogram of B-factors of the SrtA ⌬N59 crystal structure (18). dynamics of any sortase enzyme and there is no quantitative relationship between structural parameters and mobility, we rigorously defined the Ca 2ϩ dependence of motions in SrtA ⌬N59 by measuring the rates of longitudinal (R 1 ) and transverse (R 2 ) relaxation, as well as { 1 H}- 15 N NOE values of the backbone nitrogen-15 atoms. Two samples of SrtA ⌬N59 in different states of ligation were investigated: (i) an apo-Ca 2ϩ form (50 mM Tris-HCl, 100 mM NaCl, 3 mM dithiothreitol, 7% D 2 O, and 0.01% NaN 3 ; pH 6.2) and (ii) a Ca 2ϩ -bound form (conditions identical to the apo-Ca 2ϩ SrtA form, but with 20 mM Ca 2ϩ present). The relaxation data were then interpreted using the Modelfree formalism to gain insights into the magnitudes and time scales of motion. Graphs showing the relaxation data as a function of residue number are shown in supplemental Fig. S1 and tables listing the values of the Modelfree parameters (S 2 , e , and R ex ) are provided as supplemental Tables S1 and S2.
Residues in the ␤6/␤7 Loop Transiently Participate in Ion Binding-The coordinates of the ␤6/␤7 loop are poorly defined in both the NMR and x-ray structures of the enzyme (Fig. 4, E and F). However, its residues can be divided into three sections based on their positioning relative to the active and ion binding sites, and their dynamics properties were revealed by NMR (Fig. 3C). At its N-terminal end, residues Lys-162 to Val-166 (region 1) ascend from the body of the protein so as to position the ring of Pro-163 immediately adjacent to the catalytic side chain of Arg-197. Residues Gly-167 to Asp-170 (region 2) then form the substrate-contacting surface and are followed by residues Asp-170 to Asp-176 (region 3), which are positioned proximal to the ion-binding site in the ␤3/␤4 pocket. From the relaxation data analysis, the S 2 parameter is calculated, which gives a concise account of each the mobility of N-H bond vector on the picosecond time scale; it ranges from 0 to 1, with a value of 1 indicating that the amide is completely immobilized. Inspection reveals that only the ␤6/␤7 loop exhibits significant Ca 2ϩ -dependent changes in its dynamics (Fig. 4, compare A and  B). When the S 2 data are mapped onto the structure, it is apparent that residues in regions 2 and 3 of the ␤6/␤7 loop become partially immobilized when the ion is bound (in Fig. 4, C and D, the thickness of the chain is correlated with increased mobility in the apo-Ca 2ϩ and Ca 2ϩ -bound forms, respectively). Interestingly, residual picosecond motions in the ␤6/␤7 loop persist even in the presence of Ca 2ϩ , implying that it transiently binds Ca 2ϩ . This may explain the weak affinity of the protein for the ion (Figs. 1B and 2D) and the observed disorder in the ␤6/␤7 loop in the NMR structure of the Ca 2ϩ -bound enzyme.
Glu-171 in the ␤6/␤7 Loop Is Important for Ca 2ϩ Binding-Because the NMR data indicate that the ␤6/␤7 loop becomes immobilized when the ion is present, it seems likely that it contains one or more residues that directly contact the ion in the ␤3/␤4 pocket. Because the coordinates of the ␤6/␤7 loop are poorly defined in both the NMR and crystal structures, it is not possible to unambiguously identify contacts from it to the ion (Fig. 4, E and F). However, as previously noted, the side chain of Glu-171 within the ␤6/␤7 loop is a likely candidate for ion binding, because in several of the conformers of the NMR structure of Ca 2ϩbound SrtA ⌬N59 it is poised to interact with the ion. This is also consistent with the finding that the adjacent backbone amides of the nearby residues Leu-169 and Asp-170 experience the largest ion-dependent changes in their S 2 parameters and chemical shifts, respectively.
To determine if ion contacts from Glu-171 act to immobilize the ␤6/␤7 loop, the Ca 2ϩ binding properties of three single amino acid mutants of SrtA ⌬N59 were tested. Each mutant replaces potential ion binding acidic side chains with alanine and target residues that are located in the ␤6/␤7 loop (Asp-170 3 Ala and Glu-171 3 Ala) or the ␤3/␤4 ion binding pocket (Glu-108 3 Ala). All of the mutant proteins remain folded as judged by their NMR spectrum (data not shown) and retain enzymatic activity ( Table 1). The Ca 2ϩ binding properties of the mutants were assessed by measuring the Ca 2ϩ dependence of their enzymatic activity as previously described for the wild-type protein. As shown in Fig. 1B, the K m values for the wild-type and D170A mutant proteins show a similar dependence on Ca 2ϩ , indicating that Asp-170 within the ␤6/␤7 loop does not bind Ca 2ϩ . In contrast, both the E108A and E171A mutants show reduced Ca 2ϩ sensitivity (Fig. 1B). The E108A mutant serves as a positive control, because it has been shown to unambiguously interact with the ion based on Mn 2ϩ titration data ( Fig. 2A). The finding that it and E171A have similar effects on catalysis suggests that Glu-171 within the ␤6/␤7 loop contacts the ion. Most importantly, fits of the Ca 2ϩ dependence of the K m values to Equation 2 reveal that the E108A and E171A mutants bind Ca 2ϩ with 10-fold lower affinity than the wild-type or D170A proteins. To verify the relative Ca 2ϩ binding affinities of the mutants, NMR was also used to directly monitor ion binding. As shown in Table 1, affinity measurements by NMR gave similar results as the kinetic analysis and indicate that Glu-108 in the ␤3/␤4 pocket and Glu-171 in the ␤6/␤7 loop are involved in binding the ion, whereas Asp-170 is not. These data are consistent with ion contacts from Glu-171 acting to stabilize ion binding.
Ion Binding Redistributes Slow Motions toward the Active-and Substrate-binding Sites-Protein motions on micro-to millisecond time scales are important for enzymatic catalysis and ligand recognition (51,52). Because the Modelfree approach only characterizes these processes indirectly as a contribution to transverse relaxation in the form of the R ex term, we performed relaxation-compensated CPMG (rc-CPMG) experiments to directly investigate these motions (53,54). Slow timescale conformational rearrangements in SrtA ⌬N59 are revealed by the data shown in Fig. 5 (A and B), which displays plots of the R ex terms derived from the Modelfree analysis (open bars) and the difference in transverse relaxation rates when long ( cp ϭ 4 ms) and short ( cp ϭ 1 ms) inter-pulse delays are used in the CPMG sequence (filled circles). Positive differences in the transverse relaxation rates are indicative of slow chemical exchange events and are in good qualitative agreement with the Modelfree data.

TABLE 1 Kinetic parameters of wild-type and mutant sortases
Triplicate data sets for each experiment were used to calculate the steady-state velocity (V s ) at different o-aminobenzoyl-LPETG-dinitrophenyl and Ca 2ϩ concentrations for each enzyme as described previously (25). K m and V max values were determined by double-reciprocal plot of the substrate dependence of V s . Dissociation constants (K d ) for Ca 2ϩ binding were calculated by curve fitting of the apparent K m data using Equation 2. Relative affinity is the K d ratio comparing to the wild-type enzyme.

Dynamics of Calcium Modulated Loop Closure of Sortase A
The rc-CPMG data indicate that Ca 2ϩ binding induces slow biologically relevant motions in the active site. In the absence of Ca 2ϩ , residues exhibiting slow motional dynamics are evenly distributed throughout the primary sequence (Fig. 5D, blue spheres, and supplemental Table  S1). The exception is the C-terminal portion of the ␤6/␤7 loop near the ion, which in addition to fast picosecond motions, is in flux on slower time scales prior to Ca 2ϩ binding (residues within (Gly-174 and Asp-176) and underneath (Ala-202, Thr-203, and Glu-204) this portion of the loop exhibit R ex values). In contrast, in the Ca 2ϩ -bound form, a spike in R ex values was observed in region 2 of the ␤6/␤7 loop (Fig. 5D, red spheres; residues Asp-165, Val-168, and Leu-169). These motions are likely triggered by an ion-induced movement of almost the entire loop, because all of it exhibits Ca 2ϩ -dependent chemical shift changes (Fig. 4B of Ref. 17). Interestingly, a general redistribution of slow motions toward the active site occurs upon ion binding, because significant R ex terms are also observed in residues Thr-121, Phe-122, and Asp-124, and in Ile-199, which immediately follow the catalytically essential His-120 and Arg-197 side chains (Fig. 5D). These motions may be correlated with the structural rearrangements that are required to catalyze hydrolysis. Interestingly, residues in region 2 form the substrate binding site in the crystal structure of the complex. This suggests that Ca 2ϩ -triggered slow motions in this part of the loop may facilitate the adaptive recognition of the sorting signal.
Ion Binding and Active Site Motions Occur on Similar Time Scales-Our results clearly illustrate that Ca 2ϩ acts to quench motions within the ␤6/␤7 loop. Although the mutagenesis data suggest that ion contacts from Glu-171 immobilize the loop, we sought further support for this model by rigorously assessing the rate of slow protein motions and ion binding. If active site dynamics were coupled to Ca 2ϩ binding via ion contacts from Glu-171, we reasoned that the rate of these conformational fluctuations would be similar. To investigate this issue, we quantified conformational exchange rates by performing rc-CPMG experiments on a new sample of SrtA ⌬N59 containing 20 M Ca 2ϩ (3.2 mM SrtA ⌬N59 , 20 M Ca 2ϩ , 50 mM Tris-HCl, 100 mM NaCl, 3 mM dithiothreitol, 7% D 2 O, and 0.01% NaN 3 , pH 6.20). This enables a more accurate interpretation of the NMR data, because the exact populations of the ion-free and -bound forms of SrtA can be determined (assuming a K d ϭ 2.2 mM, this sample is composed of 99.2% and 0.08% equilibrium populations of the apo-Ca 2ϩ and Ca 2ϩ -bound SrtA ⌬N59 , respectively). According to Ishima and Torchia (55), under conditions similar to ours, in which the populations of the interchanging conformers are dramatically skewed, the phenomenological transverse relaxation rate R 2 (1/ cp ) for all time scales is given by Equation 3, where ⌬ is the chemical shift difference between the two exchanging populations, k ex is the rate of exchange, cp is the delay between 180 degree pulses in the experiment, and p S and p C are the populations of the apo-Ca 2ϩ The panels also show the results of rc-CPMG experiments that serve to verify the existence of conformational exchange. Plotted as gray circles is the qualitative difference, ⌬R 2 (1/ cp ), in the measured transverse relaxation rates when cp ϭ 4 ms and cp ϭ 1ms. C, dispersion curves generated from the rc-CPMG experiment. Experimentally determined ⌬R 2 (1/ cp ) rate constants were fitted as a function of 1/ cp using Equation 3. Data are shown for two residues positioned immediately adjacent to the Ca 2ϩ -binding pocket: Glu-108 (square and solid line) and Ser-109 (circle and dashed line) where the symbols represent experimental data and the lines represent their fits to the equation. D, the redistribution of R ex terms toward active-and substratebinding sites when Ca 2ϩ is present. The data from panel A have been mapped onto the structure of SrtA ⌬N59 , with blue and red spheres denoting protein residues that exhibit R ex terms in the absence and presence of Ca 2ϩ , respectively. Some residues exhibit R ex terms in both the presence and absence of Ca 2ϩ and are also indicated by a red sphere (residues Asp-124, Ala-135, Val-168, Thr-180, Val-193, Ala-202, and Thr-203).
and Ca 2ϩ -bound forms of SrtA ⌬N59 , respectively. To extract k ex and ⌬, we measured dispersion curves in which R 2 was determined using different cp delays (representative data are provided in Fig. 5C). A total of ten residues exhibited adequate dispersion curves that enabled the extraction of their exchange parameters through Jackknife simulations (supplemental Table  S3). Eight of the ten residues serve as probes for the effects of Ca 2ϩ binding, because they are positioned in the ordered portion of the ion-binding pocket (Glu-108 and Ser-109), underneath the ion-binding pocket (Ala-202 and Glu-204), or in the ␤6/␤7 loop (Val-168 and Gly-174). The rate of dissociation of the ion from the protein (k off ) is calculated to be 1152 Ϯ 114 s Ϫ1 , if data solely from residues Glu-108 and Ser-109 are used. Because they reside in the rigid ion-binding ␤3/␤4 pocket, it can be assumed that their k ex values strictly reflect the Ca 2ϩ -binding event (k ex ϭ k off ϩ k off [Ca free 2ϩ ]) (56). Interestingly, similar k ex values are observed for residues Val-168 and Gly-174 within the ␤6/␤7 loop, and the Ca 2ϩ -dependent chemical shift changes predicted from the relaxation data are similar to their experimentally measured values (supplemental Table S3). Although this analysis assumes a simplified two-state mode for the dynamics, the results suggest that the time scale of motions occurring within the substrate binding surface are similar to the rate of ion exchange from the protein.
Interpretation of the Dynamics Data-The Mn 2ϩ , NMR relaxation, and mutagenesis data reported herein, and previous chemical shift mapping studies, indicate that ion binding to the ␤3/␤4 pocket causes the ␤6/␤7 loop to undergo a structural change that reduces its mobility. Because the coordinates of the ␤6/␤7 loop are poorly defined in both the crystal and NMR structures (Fig. 4, E and F) no detailed structural framework exists to interpret the dynamics data. However, gross features of these structures suggest that the loop toggles between two basic states that differ in their overall positioning relative to the body of the protein. In the NMR structure solved in the presence of Ca 2ϩ , the loop is partially immobilized and tethered to the body of the protein in a "closed" state. The NMR data strongly support this conformation, as the side chains of Val-168 and Leu-169 in the loop are unambiguously positioned adjacent to the active site (Fig. 3B) by several NOE cross-peaks between these residues and atoms in the underlying beta sheet (data not shown). A closed state for the loop in the presence of Ca 2ϩ is also compatible with the results of our mutagenesis studies, which show that an acidic side chain positioned adjacent to these residues (Glu-171) presumably closes the loop by contacting Ca 2ϩ (Figs. 1B and 3A). The disorder of the ␤6/␤7 loop observed in the NMR structure can be attributed to residual slow conformational dynamics of this region detected by the rc-CPMG experiments and to the prevalence of hydrophilic side chains whose degenerate chemical shifts make it notoriously difficult to identify distance restraints. In the absence of Ca 2ϩ NMR data indicate that the ␤6/␤7 loop is more flexible (Fig. 4, A and B). The available structural data are compatible with this finding, because in the crystal structure solved in the absence of Ca 2ϩ , regions 2 and 3 of the loop form a flap that adopts an "open" conformation in which the hydrophobic side chains of Val-168 and Leu-169 are rotated away from the body of the protein. Interestingly, in the three molecules of the asymmetric unit the flap adopts very distinct structures, or is missing electron electronic density. This variability and our observation of large amplitude picosecond time-scale motions in these residues suggest that in the absence of Ca 2ϩ regions 2 and 3 adopt a variety of rapidly interchanging conformers in which the flap is primarily untethered from the body of the protein. The NMR data indicate the ␤6/␤7 loop in the absence of substrate is mobile regardless of ion occupancy. In the spectra of the apo-Ca 2ϩ and Ca 2ϩ -bound forms, extensive line broadening is observed in the beta bulge structure at the end of strand ␤6, which is positioned below the flap (regions 2 and 3). Even after extensive attempts to assign them, only weak amide correlations for residues Arg-159 through Val-161 could be identified. Because these residues exhibit low temperature factors in the crystal structure and some of their amides are protected from deuterium exchange in solution, it is possible that they adopt a rigid conformation and that their broadening in the NMR spectra is indirectly caused by flap motions of regions 2 and 3. A similar, but less pronounced broadening is observed in residues in strand ␤8, consistent with their more distal positioning underneath the flap. Because broadening is observed in both the presence and absence of Ca 2ϩ , the flap must constantly be in motion, presumably sampling the open and closed states prior to binding the sorting signal. Interestingly, the broadened resonances reappear in the NMR spectra of Ca 2ϩ -bound SrtA complexed with a peptidyl-sulfhydryl compound that mimics the substrate (data not shown). Although the NMR sample of this complex is too dilute for detailed relaxation studies, the observation of a single set of NMR resonances strongly suggests that the substrate induces the loop to adopt a single conformation.
The precise mode of substrate recognition cannot be gleaned from the recently determined crystal structure of the C184A SrtA ⌬N59 -LPETG peptide complex because of the high temperature factors of the bound substrate, which indicate that it is structurally disordered. However, the ␤6/␤7 loop in the structure of the complex adopts a conformation more reminiscent of the closed state, which enables the side chains of Val-168 and Leu-169 to bracket the proline ring of the substrate. Because a generally similar conformation is observed in the NMR structure of the Ca 2ϩ -bound apo-substrate enzyme, the available data suggest that Ca 2ϩ bound to the ␤3/␤4 pocket promotes substrate binding by stabilizing a loop conformation better suited for binding by forming interactions with Glu-171 that help to tether the flap to the protein body.
The increased enzymatic activity of SrtA in response to Ca 2ϩ enhances the chances of S. aureus colonizing its host by up-regulating the number of proteins it displays on its surface. In other Ca 2ϩ -regulated proteins, Ca 2ϩ elicits its effects at a structural level, for example, facilitating the correct assembly of enzyme active sites or exposing protein surfaces required for function (57)(58)(59)(60). It can also act directly in catalysis by interacting with the substrate or by stabilizing key reaction intermediates through electrostatic effects (61). Here we have shown that, in SrtA, Ca 2ϩ promotes substrate binding by modulating both the structure and dynamics of a large substrate-contacting loop. Although The panel on the left shows SrtA in the absence of Ca 2ϩ , emphasizing that the substrate contacting side chains of Val-168 and Leu-169 are removed from the active site, and the loop is more flexible. The panel on the right shows how contacts from the ␤6/␤7 loop to the Ca 2ϩ ion bound in the ␤3/␤4 site act to immobilize and reposition the loop for favorable contacts to the LPXTG sorting signal. more complex models of motion are possible, the available structural and dynamics data are consistent with a model in which a portion of the loop is generally disordered and capable of toggling between two states (Fig. 6), a binding competent closed form, and a highly flexible open state that removes key substrate contacting residues from the active site. We propose that these conformers are in dynamic equilibrium and that a single Ca 2ϩ ion acts to bias the loop toward its binding competent closed form by transiently tethering the C-terminal end of the loop to the body of the protein by contacting the side chain of Glu-171. In other enzymes the dynamic behavior of active site loops has been shown to be important for catalysis (62)(63)(64). Interestingly, in SrtA, the magnitude and time scales of these motions are responsive to Ca 2ϩ , which when bound allosterically, triggers slow micro-to millisecond motions that may be ideally suited for substrate recognition. Bacteria becoming increasingly resistant to multiple antibiotics is an increasing health concern. The central role of sortases in bacterial virulence makes them an attractive target for new anti-infective agents, and an understanding of how Ca 2ϩ regulates their activity should aid in the ongoing development of small molecule inhibitors of this enzyme class (65)(66)(67).