Specificity and affinity of the N-terminal residues in staphylocoagulase in binding to prothrombin

In Staphylococcus aureus–caused endocarditis, the pathogen secretes staphylocoagulase (SC), thereby activating human prothrombin (ProT) and evading immune clearance. A previous structural comparison of the SC(1–325) fragment bound to thrombin and its inactive precursor prethrombin 2 has indicated that SC activates ProT by inserting its N-terminal dipeptide Ile1-Val2 into the ProT Ile16 pocket, forming a salt bridge with ProT's Asp194, thereby stabilizing the active conformation. We hypothesized that these N-terminal SC residues modulate ProT binding and activation. Here, we generated labeled SC(1–246) as a probe for competitively defining the affinities of N-terminal SC(1–246) variants preselected by modeling. Using ProT(R155Q,R271Q,R284Q) (ProTQQQ), a variant refractory to prothrombinase- or thrombin-mediated cleavage, we observed variant affinities between ∼1 and 650 nm and activation potencies ranging from 1.8-fold that of WT SC(1–246) to complete loss of function. Substrate binding to ProTQQQ caused allosteric tightening of the affinity of most SC(1–246) variants, consistent with zymogen activation through occupation of the specificity pocket. Conservative changes at positions 1 and 2 were well-tolerated, with Val1-Val2, Ile1-Ala2, and Leu1-Val2 variants exhibiting ProTQQQ affinity and activation potency comparable with WT SC(1–246). Weaker binding variants typically had reduced activation rates, although at near-saturating ProTQQQ levels, several variants exhibited limiting rates similar to or higher than that of WT SC(1–246). The Ile16 pocket in ProTQQQ appears to favor nonpolar, nonaromatic residues at SC positions 1 and 2. Our results suggest that SC variants other than WT Ile1-Val2-Thr3 might emerge with similar ProT-activating efficiency.

higher than that of WT SC(1-246). The Ile 16 pocket in Pro-T QQQ appears to favor nonpolar, nonaromatic residues at SC positions 1 and 2. Our results suggest that SC variants other than WT Ile 1 -Val 2 -Thr 3 might emerge with similar ProT-activating efficiency.
Blood clot formation by Staphylococcus aureus can be attributed to the combined effects of pathogen clumping and the generation of fibrin (Fbn). 6 The latter is initiated by the secreted virulence factor, staphylocoagulase (SC). Based on the bacteria's ability to promote clot formation in rabbit plasma, S. aureus is divided into coagulase-positive and -negative subgroups. Typing of bacterial isolates for SC is still performed today in clinical diagnosis. Coagulase-positive S. aureus is a potent human pathogen that causes conditions ranging from minor skin infections to life-threatening diseases, such as severe pneumonia, meningitis, and bone, joint, and heart infections. Each year ϳ500,000 patients in American hospitals contract staphylococcal infections that lead to ϳ30,000 deaths (1,2). Turbulent blood flow can cause endothelial damage to heart valves, exposing subendothelium that leads to deposition of platelets and Fbn. The Fbn-platelet matrix deposited on damaged valves serves as a focus for adhering S. aureus bacteria circulating in the blood (3). The S. aureus-platelet interaction is facilitated by fibrinogen (Fbg), fibronectin, thrombospondin, and MSCRAMMs (microbial surface components recognizing adhesive matrix molecules), such as protein A and Fbg-binding ClfA (clumping factor A) (4 -11). In acute bacterial endocarditis, Fbn formation is mediated by the thrombin precursor prothrombin (ProT), conformationally activated by bacterial SC. This furthers aggregation of platelets and enlargement of platelet-Fbn-bacteria vegetations on the valves (3). These friable vegetations can break up and cause pulmonary embolism and stroke. The pathogens also utilize these vegetations to disseminate and avert clearance by the host immune system (12). Acute bacterial endocarditis caused by S. aureus leads to 20 -40% mortality despite antibiotic therapy (13).
Blood clotting is a highly regulated process, with a delicate balance between clotting and fibrinolysis. SC bypasses the clotting cascade by directly and nonproteolytically activating ProT, thereby shifting the balance to a procoagulant state. SC binds and conformationally activates ProT, forming a complex of catalytically active zymogen and activator, SC⅐ProT* (with the asterisk denoting a functional catalytic site). The activated complex cleaves Fbg to form Fbn clots involved in enlarging the vegetations. No physiological inhibitors of the SC⅐ProT* complex have been reported to date, and it is resistant to the plasma serpins, antithrombin-heparin and heparin cofactor II, 7 ␣2-macroglobulin (14), and the leech inhibitor, hirudin (15).
SC is a bifunctional protein with a molecular mass of ϳ75,000 Da. Its N-terminal region binds ProT (16), whereas the C-terminal region contains seven 27-amino acid-repeat sequences that bind Fbg ( Fig. 1) (17,18). We previously described that the SC fragment consisting of N-terminal residues 1-325, SC(1-325), binds ProT extremely tightly (ϳ17-72 pM) and noncovalently in a 1:1 stoichiometry to form the active SC(1-325)⅐ProT* complex (19). A comparison of the crystal structures of SC(1-325) with thrombin and the inactive zymogen prethrombin 2 (Pre2, prothrombin without the fragments 1 and 2) showed that the first six residues of the SC fragment were fully defined by electron density in the complex with Pre2 but not with thrombin. SC activates the zymogen by inserting its N-terminal Ile 1 -Val 2 (I1V2 amino acid single-letter code) residues into the Ile 16 pocket of Pre2, forming a salt bridge with Asp 194 and inducing a functional active site in the zymogen (19). Formation of the SC(1-325)⅐ProT* and SC(1-325)⅐Pre2* complexes partially blocks exosite 1, the Fbg-binding site, but expresses a new Fbg substrate recognition site that facilitates Fbg binding and cleavage (Fig. 2). SC  consists of two three-helix-bundle domains (D1 (residues 1-146) and D2 (residues 147-325)) with a boomerang-like structure. The D1 domain interacts with the 148-loop of thrombin or Pre2 and the south rim of the catalytic site, and the D2 domain binds (pro)exosite 1 on Pre2 and thrombin. We report here that SC(1-246), with a partially truncated D2 domain, is capable of ProT activation and binds ProT with a K D of ϳ1 nM, which made it a suitable probe for determining the affinities of mutant SC constructs by competitive equilibrium binding. The I1V2 residues are critically involved in SC-mediated ProT activation, and a comparison of the SC N-terminal residues from 12 different S. aureus strains showed strict conservation of I1V2T3. This raised the question of whether S. aureus could possibly make SC mutants that contain N-terminal residues different from I1V2T3 and what the effect of these other N-terminal residues on the affinity and activation of ProT might be. Because 8,000 different combinations are possible, our selection was guided by in silico protein modeling with Rosetta and virtual screening to prioritize combinations with high binding free energy and to identify a range of affinities and activation potencies. In this study, 46 different N-terminal mutants of SC(1-246) were generated through site-directed mutagenesis and characterized for binding and activation of a ProT(R155Q,R271Q,R284Q) mutant (ProT QQQ ) that can be proteolytically activated to a meizothrombin form refractory to cleavage by prothrombinase and thrombin (20) but was used in this study to monitor conformational activation of prothrombin instead of proteolytic activation. The panel displayed a wide range of activation potencies and K D values both by equilibrium binding and Pro-T QQQ activation. The V1V2, I1A2, and L1V2 mutants bound   domain colored yellow and the D2 domain colored gold. B, the complex above is rotated ϳ90º from the standard orientation to show the insertion of the N-terminal SC peptide Ile 1 -Val 2 -Thr 3 into the Ile 16 -binding pocket of Pre2, triggering activation. This figure was constructed with UCSF Chimera using the X-ray crystal structure 1NU9.pdb (16).

Binding of the staphylocoagulase N terminus to prothrombin
ProT QQQ with affinities similar to that of WT SC(1-246) and with ProT QQQ activation potencies that were similar or up to 1.8-fold greater. Activation potencies of mutants with both weak equilibrium binding and activation-based affinities were typically reduced. Select mutants carrying nonpolar residues at position 1 bound moderately to ProT QQQ when measured by equilibrium binding but exhibited tight (approximately nanomolar) binding when measured by ProT QQQ activation, indicating a difference in affinity when binding to a disordered Ile 16 pocket (equilibrium binding) and an Ile 16 pocket conformationally stabilized by substrate occupation of the specificity site. These mutants had limiting activation rates similar to or exceeding that of WT SC . Overall, the Ile 16 pocket of ProT QQQ favors nonpolar, nonaromatic residues at positions 1 and 2, however with specific restrictions governed by steric complementarity between the N terminus of SC and the Ile 16 pocket of ProT. Our results, including efficient activation of ProT by weaker binding mutants at saturating concentrations, suggest that SC variants might emerge with similar or higher efficiency to activate ProT.

Characterization of SC(1-246)-BODIPY and equilibrium binding of labeled and unlabeled SC(1-246) to ProT QQQ
SC(1-246)-BODIPY had a labeling ratio of 0.87 BODIPY-FL thiol-sensitive probe to SC(1-246) with a S7C substitution for covalent probe binding (Fig. 3A). Incubation of ProT QQQ with SC(1-246)-BODIPY showed binding in an approximately 1:1 ratio as observed by native PAGE (Fig. 3B). Competitive equilibrium binding studies were performed to determine the affinity and stoichiometry of four separate ProT QQQ preparations for unlabeled SC(1-246) and SC(1-246)-BODIPY. Unlabeled SC(1-246) bound very tightly to the ProT QQQ preparations, with K D of 0.6 -1.0 nM, and a 1:1 stoichiometry ( Fig. 4 and Table  1). SC(1-246)-BODIPY bound ProT QQQ with K D of 2.9 -5.9 nM and a stoichiometry of 1:1. This weaker affinity is attributed to the BODIPY-FL label. Batch-to-batch variability of ProT QQQ was modest, as reflected by the consistent affinity values for unlabeled and labeled SC(1-246). The maximum fluorescence intensity was 0.6 Ϯ 0.1 for all of the ProT QQQ preparations.   Table 1).

Binding of the staphylocoagulase N terminus to prothrombin Binding prediction and competitive equilibrium binding of SC(1-246) N-terminal mutants to ProT QQQ
Computational modeling used Rosetta's ⌬⌬G functionality to calculate the predicted change in binding energy for each mutation (21), sampling all 20 amino acids at positions 1-3. The fixed backbone design strategy allowed discrimination between steric clashing and nonclashing sequences, and based on the resulting energies in Rosetta energy units (REU), a representative double and triple N-terminal mutant panel was selected for experimental binding studies, with expected K D values between 1 and ϳ1000 nM. Equilibrium binding showed that the double mutants V1V2, I1A2, and L1V2 bound to Pro-T QQQ very tightly, with affinities of ϳ1 nM, similar to that of WT SC(1-246) (Fig. 5, A-C). Their excellent ProT QQQ activation potency also indicated efficient salt bridge formation with ProT QQQ Asp 194 and formation of the ProT QQQ active site. The stoichiometric factor was ϳ1, indicating that 1 mol of ProT QQQ binds to 1 mol of SC(1-246) mutant. The K D values from equilibrium binding were measured in the absence of a conformationally stabilizing tripeptide thrombin substrate, and they reflect global binding to ProT QQQ with a disordered Ile 16 pocket and specificity subsite. The V1V2, I1A2, and L1V2 combinations of nonpolar residues had comparable equilibrium binding and activation K D values, suggesting a favorable steric complementarity to the Ile 16 pocket. Mutants with weaker equilibrium K D values may be governed mainly by the D1 and truncated D2 domain interactions with ProT QQQ , although double mutants with Ile, Val, Leu, and Thr in combination with polar, nonpolar aromatic, bulky, or small residues at position 2 exhibited considerable tightening of the binding to ProT QQQ when an Ile 16 pocket-stabilizing substrate was present ( Fig. 5 and Tables 2 and 3). The A1S2 and A1T2 mutants exhibited weak equilibrium binding affinity; however, the presence of a chromogenic substrate during ProT QQQ activation caused these mutants to bind with ϳ30 -100-fold tighter K D , respectively, and exhibit WT-like activation potency. This suggests some conformational flexibility of the ordered Ile 16 pocket. Mutants with charged residues (Lys, Arg, Asp, Glu, His) in positions 1 and/or 2; Gly or Trp in position 1; or Pro in position 2 typically bound weakly and were poor activators, with G1G2 the weakest binder ( Fig. 6 and Tables 2 and 3). Although disrupted binding for Pro was expected, even some pairs with small, hydrophobic residues in positions 1 and 2 may not optimally fit, illustrating the steric specificity of the Ile 16 -binding pocket.
Equilibrium binding K D values of triple mutants varied from ϳ4 to ϳ55 nM ( Table 4). The mutant I1I2V3 bound ProT QQQ with a K D of 4 Ϯ 3 nM (Fig. 7A) and activated ProT QQQ at an appreciable rate, suggesting that the ProT QQQ Ile 16 pocket accommodates these nonpolar residues with reasonable fit, conducive to forming a salt bridge with ProT QQQ Asp 194 . The mutants R1H2W3, F1L2Q3, E1S2W3, D1D2Y3, G1G2G3, and E1L2K3 had K D values of 36 -55 nM (Fig. 7, B-G) but activated ProT QQQ poorly. Interestingly, substituting Gly 3 for WT Thr 3 rescued equilibrium binding ϳ10-fold, compared with G1G2, but with no improvement in activation potency.

Table 1 Parameters for SC(1-246)-BODIPY and unlabeled WT SC(1-246) binding to four separate ProT QQQ preparations
Reference titrations of SC(1-246)-BODIPY with four separate ProT QQQ batches were obtained at two fixed probe concentrations. The competitive binding data for the ProT QQQ preparations were obtained by titrations of fixed concentrations of SC(1-246)-BODIPY probe, and SC(1-246) as competitor, with ProT QQQ . Data were fit simultaneously by the cubic equation to obtain the dissociation constant for ProT QQQ and SC(1-246)-BODIPY (K O , probe) and the competitor SC(1-246) (K C , competitor); the stoichiometric factor for SC(1-246)-BODIPY (n) and SC(1-246) (m); and the maximum fluorescence intensity (⌬F max /F o ). Experimental error represents Ϯ 2 S.D. Competitive equilibrium binding studies and data analysis were performed as described under "Experimental procedures." SF, stoichiometric factor.

Binding of the staphylocoagulase N terminus to prothrombin
The Gibbs free energy ⌬G for binding of the mutants varied from Ϫ12.45 to Ϫ8.44 kcal/mol (Tables 2-4 and Fig. 8), with the lowest value for WT SC(1-246), calculated from the averaged K D for binding to the four ProT QQQ batches. V1V2, I1A2, L1V2, I1T2, I1W2, and I1I2V3 had ⌬G values similar to the WT protein, consistent with K D values in the nanomolar range. A good correlation was observed between the predicted Rosetta energies and the ⌬G values calculated from equilibrium bind-ing ( Fig. 9), except for a few outliers (T1P2, W1A2, M1W2, A1W2, W1E2, L1P2, and G1P2) that gave inconsistent Rosetta energies but also exhibited reduced activation potency (Fig. S1). These outliers occur in the presence of Pro 2 and of Trp 1 or Trp 2 and display an off-scale energetic prediction because Rosetta is unable to fit them into the structure of the complex. This is not unexpected behavior; the introduction of a Pro or Trp residue may require such major conformational rearrangement, due to

Binding of the staphylocoagulase N terminus to prothrombin
steric clash or backbone geometry restriction, that the score penalty increases beyond the ability of the sampling protocol to incorporate it. Additional sampling might be necessary to create accurate models for these large or conformationally restricted amino acids. The only unexpected discrepancy was I1W2 with an REU score of 481 but exhibiting WT-like binding and ProT QQQ activation properties. The behavior of I1W2 is difficult to rationalize; the high Rosetta score reflects our expectation that inserting the steric bulk of a Trp residue at position 2 would be unfavorable. This prediction is consistent with the predictions and measurements of other mutants containing Trp at position 2. We can only conjecture that the I1W2 combination permits binding and activation via an unknown mechanism.

Prothrombin activation by SC(1-246) N-terminal double and triple mutants
Initial rates v 0 of p-nitroanilide formation upon chromogenic substrate cleavage by the ProT QQQ ⅐SC complexes were linear for WT SC(1-246) and tight-binding mutants, and the titrations showed saturation around 20 nM SC(1-246) variant. A few mutants caused hysteresis-like lag phases in substrate hydrolysis by their complexes with ProT QQQ , and post-lag, linear v 0 rates were used for analysis of these mutants. The limiting velocity (V lim ) of WT SC(1-246) was 20 Ϯ 5 mAbs/min, in good agreement for all four ProT QQQ batches, and used as the 100% value, or 1.00, for normalizing assays of slow-activating mutants to 1 nM ProT QQQ . K D and V lim values derived from the ProT QQQ activation analysis are given in Table 3. The V1V2, I1A2, and L1V2 mutants with affinities similar to WT SC(1-246) activated ProT QQQ with similar or higher potency than WT SC(1-246). I1A2 and L1V2 showed ϳ81 and ϳ41% increase in ProT QQQ activity, suggesting that Ala 2 and Leu 1 nonpolar residues bound tightly and fit optimally in the Ile 16 pocket of ProT QQQ for Ile 1 and Leu 1 bonding with Asp 194 , resulting in increased ProT QQQ activity. Apparent K D values for these tight-binding mutants, derived from the activation kinetics, were in good agreement with those measured by equilibrium binding (Fig. 10). Most mutants with a weaker equilibrium binding K D for ProT QQQ also activated ProT QQQ weakly, with relative V lim Ͻ Ͻ 1, although some, like M1L2, A1T2, A1S2, M1K2, T1A2, L1Q2, I1L2, L1T2, L1K2, V1G2, and T1V2 activated ProT QQQ similarly or up to 1.7-fold better than WT SC(1-246) at mutant concentrations approaching ProT QQQ saturation with regard to K D calculated from the activation profiles (Tables 2-4). The affinities of these mutants, defined by ProT QQQ activation, were typically higher than their counterparts defined by equilibrium binding, due to allosteric modulation of the binding by the presence of the chromogenic substrate occupying the specificity pocket of the zymogen. This was previously also reported for vWbp binding to FPR derivatives of prothrombin, prethrombin 1 and 2 (22), and the binding of oligopeptides to trypsinogen with the specificity site occupied by a covalent ligand or a tight-binding inhibitor (23). Affinities defined for triple mutants that activated ProT QQQ poorly were not well-defined due to large experimental error. Overall, nonpolar hydrophobic residues were well-tolerated in position 1, whereas polar, aromatic, or charged residues generally diminished the ProT QQQ activation potency. Due to its conformational rigidity and unusual configuration, Pro in position 2 is thought to hamper efficient salt bridge formation of any residue at position 1, with a greatly diminished activation potency as a result.

Discussion
In physiological blood coagulation, ProT is proteolytically cleaved in a multistep process to form the central clotting protease, thrombin (24). Proteolytic activation of ProT and Pre2 is initiated by cleavage of the peptide bond between Arg 15 and Ile 16 (chymotrypsinogen numbering). The newly formed I1V2 N terminus inserts into the Ile 16 pocket of the zymogen, triggering the folding of zymogen activation domain residues 142-152, 186 -194, and 216 -226, and the ␣-ammonium of group of Ile 1 forms a salt bridge with the carboxylate of Asp 194 . This generates the substrate recognition subsites and the oxyanion hole (25,26). In contrast, SC, a virulence factor secreted by S. aureus, is a potent nonproteolytic ProT activator. Our structure-function studies on SC constructs in complex with ProT, Pre2, and thrombin demonstrate that the SC I1V2 residues are critical for ProT and Pre2 activation. As shown in the crystal structure of the SC(1-325)⅐Pre2 complex, these residues insert into the Ile 16 pocket of Pre2, with SC Ile 1 forming the salt bridge with Asp 194 , and conformational changes resulting in the gen-

Binding of the staphylocoagulase N terminus to prothrombin
eration of the Pre2 active site in the complex (19). Based on the ability to clot plasma, SC variants from different S. aureus strains are classified into 12 different serotypes (27,28). The SC D1 domain (except for the first seven residues) and the D2 domain of 10 different S. aureus serotypes showed identities ranging from 53 to 89% (D1) and from 57 to 92% (D2), with the conservation of predicted ProT residues (29). A new classification scheme, based on the D1 domain of SC from 103 S. aureus strains, showed an average of 67.1% D1 domain identity among the 12 SC serotypes, with the first seven N-terminal residues being highly conserved (28). Our studies showed that the conserved I1V2 residues in the natural N terminus of SC(1-325) were required for ProT activation, with SC(2-325) being 6-fold less efficient and SC(3-325) only exhibiting Ͻ2% activity. However, a Met-SC(1-325) construct still containing the initiating Met residue (Met 1 -Ile 2 -Val 3 ) also showed ProT activation potency and binding with K D ϭ 17 Ϯ 2 nM, suggesting some degree of SC promiscuity and also flexibility of the ProT Ile 16 pocket in accommodating different residues (19). The first six N-terminal SC residues were fully resolved in the structure of the SC(1-325)⅐Pre2 complex, and modeling showed that extending or shortening this hexapeptide by one residue still allowed for interaction with the prothrombin 2 activation pocket. Also, von Willebrand factor-binding protein (vWbp), a nonproteolytic ProT activator from S. aureus Tager 104 Newman D2, and streptokinase, a nonproteolytic plasminogen activator from Streptococcus equisimilis and Streptococcus pyogenes, have N-terminal I1V2 and I1A2 sequences, respectively, suggesting that the Ile 16 pocket of serine protease zymogens is highly suited to accommodate small nonpolar residues. Hence, we wanted to examine the tolerance of the Ile 16 pocket of ProT for a panel of N-terminal residues. This could help decipher why SC has this unique N-terminal conservation and predict whether under selective pressure, S. aureus may be capable of producing SC variants with other N-terminal residues that could have affinity and potency similar or superior to that of SC with the canonical I1V2 residues.
Our previous studies reported that SC(1-325) binds to ProT with K D 0.3 Ϯ 0.2 nM (19). The extremely high affinity makes this construct less suitable for use as a probe in measuring competitive binding of SC mutants with weaker affinity. Therefore, we used truncated SC(1-246) with a S7C substitution for BODIPY labeling, which we characterized to have a K D of ϳ3 nM for ProT QQQ binding, a minimally weaker interaction than what we measured for competitive binding of WT SC(1-246) with K D of ϳ1 nM, due to introduction of the fluorescence label. This probe allowed measurement of affinities of competitive N-terminal mutants up to K D ϳ650 nM. SC(1-246) contains an

binding and activation by SC(1-246) N-terminal triple mutants
Reference titrations of SC(1-246)-BODIPY with ProT QQQ were obtained at two fixed probe concentrations. The competitive binding data were obtained by titrations of fixed concentrations of SC(1-246)-BODIPY probe, and triple mutant SC(1-246) as competitor, with ProT QQQ . Data were fit simultaneously by the cubic equation to obtain the dissociation constant for ProT QQQ and SC(1-246)-BODIPY (K O , probe) and mutant SC(1-246) (K C , competitor); the stoichiometric factor for SC(1-246)-BODIPY (n) and mutant SC(1-246) (m); and the maximum fluorescence intensity (⌬F max /F o ). Experimental error represents Ϯ 2 S.D. Competitive equilibrium binding studies and data analysis were performed as described under "Experimental procedures."   Table 4).

Binding of the staphylocoagulase N terminus to prothrombin
intact D1 domain but lacks residues 247-282 of the (pro)thrombin (pro)exosite I-binding D2 domain, which is thought to result in weakened affinity compared with a construct with an intact D2 domain. The mutants in this study can be categorized in four groups with regard to ProT QQQ binding and activation, compared with WT SC(1-246): (a) similar affinity, and similar or increased activation potency, in constructs with conserved or homologous nonpolar N-terminal residues; (b) modestly to significantly weaker equilibrium binding affinity, but substrate-induced, tight K D and induced fit to the Ile 16 pocket, and similar or increased activation potency; (c) modestly weaker affinity but significantly reduced activation potency due to poor fit of the N-terminal residues and mainly governed by partial D2 binding; and (d) significantly weaker affinity and significantly reduced or abolished activation potency in constructs with polar or proline-containing N termini, perhaps by triggering unfavorable long-range binding interactions through electrostatic or steric conformational changes. The overall affinity of our constructs for ProT QQQ is proposed to result from the combined effects of binding of N-terminal SC(1-246) residues to the ProT QQQ Ile 16 pocket, contact of the D1 domain with the ProT QQQ 148-loop, and binding of the truncated D2 domain to ProT proexosite I, whereas the activation potency is critically defined by the capacity of SC(1-246) residue 1 forming a salt bridge with ProT QQQ Asp 194 and an adequate and, if necessary, substrate-inducible fit of residue 2 in the ordered Ile 16 pocket. Our previous studies indicated that isolated D1 bound to ProT with modest affinity, K D ϳ780 nM when measured by ProT activation and K D ϳ3.5 M when measured by fluorescence equilibrium binding in the absence of a substrate. Isolated D2 binding to ProT proexosite I did not cause ProT activation due to the absence of the critical N-terminal residues; however, its affinity by equilibrium binding was ϳ30 nM (19,30).
Active-site ligands with high affinity for the proteinase are known to induce a proteinase-like conformation in the zymogen. Occupation of the ProT QQQ specificity site by a thrombin substrate caused a ϳ5-fold allosteric tightening of the D1 binding, which is facilitated by favorable steric complementarity of the first two N-terminal residues of SC and the Ile 16 -binding pocket (Fig. 11). A similar allosteric modulation is caused by binding of the S. aureus-secreted vWbp to prothrombin derivatives, with tighter binding to prothrombin forms that have their active site labeled with D-Phe-Pro-Arg-chloromethylketone (22). This tightening effect was previously also observed

Figure 10. Activation of ProT QQQ by SC(1-246) WT or double mutants.
ProT QQQ (1, 10, or 20 nM) and SC(1-246) WT (black circles) and a representative selection of mutants (navy blue, I1A2; red, L1T2; dark green, L1Q2; brown, V1V2; purple, A1K2; green, A1W2; cyan, K1A2; yellow, Q1K2; eggplant, E1S2) were incubated for 10 min at 25°C, and the reaction was initiated by adding S-2238. Activation of ProT QQQ was measured by the relative rates of increase in absorbance at 405 nm, and weak-binding mutants with low activation potency were titrated up to 1300 nM. The data were analyzed as described under "Experimental Procedures."

Binding of the staphylocoagulase N terminus to prothrombin
for binding of small peptides in the Ile 16 pocket of trypsinogen with its specificity site occupied by pancreatic trypsin inhibitor (PTI) or covalently bound p-guanidinobenzoate (pGB) (23,31). In the presence of PTI or pGB, the Ile 16 pocket is fully formed, in contrast with free trypsinogen, which shows a disordered specificity pocket and Ile 16 -binding pocket in the crystal structure. We found that the V1V2, I1A2, and L1V2 double mutants of SC(1-246) bind ProT QQQ with similar affinities and show similar or higher potency in activating ProT QQQ , and the I1I2V3 triple mutant binds and activates ProT QQQ only slightly more weakly. In a study with isolated di-and oligopeptides, Bode (23) reported that the more effective peptides I1V2 and I1V2G3 used in activation studies of trypsinogen carrying a ligand in the specificity pocket (pGB or pancreatic trypsin inhibitor) were identical to the newly formed N-terminal sequence after cleavage of the activation peptide. However, the peptides V1V2, I1A2, and L1V2 bound to pGB-trypsinogen with affinities 30-, 160-, and 190-fold weaker than I1V2, respectively. In our studies, additional interactions of the D1 and the truncated D2 domains were shown to contribute to enhanced binding affinity and ProT QQQ activation potential by SC(1-246) mutants with these N-terminal sequences. In the I1I2V3 triple mutant, Ile 2 is similar in size and hydrophobicity to Val 2 of the WT construct. Consequently, there was not much reduction in ProT QQQ affinity and activation potency of this mutant. The equilibrium binding and kinetic data presented here indicate that small and nonpolar residues are preferred over bulky and charged ones for sufficient ProT binding and activation (Table 2), due to a better fit in the Ile 16 pocket, even in the disordered state. The Val 1 , Leu 1 , and Ala 2 residues of the double mutants are as functional as Ile 1 and Val 2 of WT SC . The presence of these residues in the pocket favors the packing and alignment of the side chains triggering conformational activation in a similar fashion as seen in the SC(1-325)⅐Pre2 complex, with the ␣-ammonium group of Val 1 possibly connecting through a salt bridge with Asp 194 . Ala 2 may be stabilized through the formation of a hydrogen bond with Asp 189 in the ProT QQQ specificity pocket, as observed for Val 2 in the SC(1-325)⅐Pre2 crystal structure. In the crystal structure of SC(1-325)⅐Pre2, Ile 1 is completely buried in the hydrophobic Ile 16 pocket, whereas Val 2 partially contacts the outer solvent. Replacement of valine with the bulky amino acid leucine showed a 15-fold decrease in affinity for pGB-trypsinogen (23). Even though binding and insertion of the first two N-terminal are additive, the Ile 16 pocket can accommodate a less favorable residue at the second position.
Interestingly, the T1A2 mutant has the same T1A2T3 residues as those proteolytically generated in a ProT mutant upon cleavage at Arg 320 (32). Substitution of IVE to TAT following ProT Arg 320 did not prevent cleavage but ultimately generated a thrombin mutant IIa TAT with zymogen-like properties that bound the active site probe DAPA with ϳ32,000-fold weaker affinity than WT thrombin (K D ϳ1 nM) and only had 0.2% specific activity toward the thrombin-specific chromogenic substrate S2238. Our T1A2 mutant bound ProT QQQ with equilibrium K D of 100 Ϯ 27 nM, and 5.2 Ϯ 0.2 nM from ProT activation, in the presence of a chromogenic substrate and an activation potential 1.3 times that of WT SC . This functional mutant may attribute its potential of salt bridge formation with ProT QQQ Asp 194 and zymogen activation to cumulative N-terminal, D1 and D2 conformational interactions that shift the zymogen-protease equilibrium in IIa TAT toward the protease conformation.
Equilibrium binding of various mutants involving Pro and Trp was moderate to weak (T1P2, M1W2, A1W2, and L1P2, ϳ60 -200 nM), and their dramatically reduced activation potency suggests nonproductive interaction with the Ile 16 pocket and binding through the partial D2 domain that may be perturbed by electrostatic or steric effects introduced by the mutations. Variations in the structural orientation of the partial D2 domain may alter binding and lead to weaker overall mutant affinity for ProT QQQ . Typically, these mutants were outliers in the correlation of predicted REU scores and measured ⌬G values of equilibrium binding (Fig. S1), suggesting that the sampling protocol employed was not sufficient to overcome the major structural perturbation required to insert the steric bulk of a Trp residue or to accommodate the backbone angle restriction imposed by a Pro residue. The high energy penalty can be interpreted as a clear signal that the new sequence is incompatible with the native structure. Triple mutants containing bulky aromatic or charged residues as well as G1G2G3 bound ϳ10fold tighter than the G1G2 double mutant with K D of 48 Ϯ 33 nM, respectively, suggesting that Gly at position 3 instead of native Thr is more conducive to steric complementarity. However, the low activation potencies of both the double and triple mutant indicated impaired salt bridge formation with ProT QQQ Asp 194 .
In conclusion, we have determined the affinities of a panel of 46 different SC N-terminal mutants for ProT QQQ and showed that the Ile 16 pocket is specific for accommodating residues that are similar in size to I1V2, but improved fit for a variety of preferably noncharged position 2 residues except for proline can be induced by small substrate binding. This characterization of the SC N-terminal residues in the SC-ProT complex provides further information to better design antibodies as

Binding of the staphylocoagulase N terminus to prothrombin
drugs to target the SC N terminus. Our results suggest the distinct possibility that S. aureus may be capable of adapting to continuous use of antibiotics and selection pressure to escape the human immune response, by generating SC variants with similar or higher efficiency to activate ProT.

Experimental procedures
Expression, purification, and labeling of proteins SC(1-246) was cloned into a modified pET30b(ϩ) vector (Novagen) containing an N-terminal His 6 tag followed by a tobacco etch virus cleavage site (19,30). The SC(1-246) N-terminal mutants were prepared through site-directed mutagenesis using degenerate and specific primers (Table S1), and mutations were confirmed by DNA sequencing. The mutants were expressed in Rosetta 2 (DE3) pLysS Escherichia coli in the presence of 100 g/ml kanamycin, and expression was induced by 10 mg/ml lactose for 4 h. Mutants were purified from inclusion bodies, and the His 6 tag was removed as described (33,34). The proteins were stored in 50 mM HEPES, 125 mM NaCl, pH 7.4, at Ϫ80°C until use. The mutant concentrations were determined using the extinction coefficients and molecular mass calculated by the Expasy tool, RRID:SCR_018087 (Table S2). HEK293 cells expressing ProT QQQ , in which the prothrombinase cleavage site Arg 271 and the thrombin cleavage sites Arg 155 and Arg 284 were replaced by glutamine to prevent degradation, were a gift from Dr. Sriram Krishnaswamy (University of Pennsylvania School of Medicine) (20). ProT QQQ was expressed, purified, and stored as described (35,36). Four separate ProT QQQ batches were prepared, and the concentrations were determined using E 280 nm,0.1% 1.47 ml mg Ϫ1 cm Ϫ1 and M r 72,000.

Preparation and characterization of SC(1-246)-BODIPY
To create a labeled SC construct, Ser 7 of SC(1-246) was converted to cysteine through site-directed mutagenesis (Agilent Technologies) and confirmed by DNA sequencing. Purified SC(1-246)-S7C was reduced with 2 mM DTT and dialyzed against 5 mM MES, 125 mM NaCl, 2 mM DTT, pH 6.0. The reduced protein was run on Sephadex G-25 (1 ϫ 25 cm) in 50 mM HEPES, 125 mM NaCl, 1 mg/ml PEG, 10 mM EDTA, pH 7.4, buffer to remove free DTT. Approximately 5-10 mg of protein was incubated for 1 h at 25°C with a 10-fold molar excess of BODIPY-FL-iodoacetamide (Thermo Fisher Scientific) to label the free S7C thiol. The excess probe was removed by Sephadex G-25 chromatography in 50 mM HEPES, 125 mM NaCl, 0.1 mM EDTA, pH 7.4, buffer. Labeled SC(1-246)-S7C-BODIPY-FL (SC(1-246)-BODIPY) was dialyzed against storage buffer (50 mM HEPES, 125 mM NaCl, pH 7.4) and stored at Ϫ80°C. The concentration and labeling ratio were determined using E 280 nm, 0.1% 0.936 ml mg Ϫ1 cm Ϫ1 and M r 29,150 for WT SC(1-246) and SC(1-246)-S7C, and E 505 nm of 63,771 cm Ϫ1 M Ϫ1 for BODIPY-FL-iodoacetamide. An absorbance ratio (A 280 nm / A 505 nm ) of 0.03 was used to correct for the probe contribution to absorbance at 280 nm. The purity of SC(1-246)-BODIPY was established by 4 -20% polyacrylamide SDS-PAGE under reduced and nonreduced conditions. The fluorescence of the labeled protein was imaged under UV light, and proteins were then stained with colloidal Coomassie Blue G-250. To determine whether SC(1-246)-BODIPY forms a binary complex with ProT QQQ , a fixed concentration (2.5 M) of ProT QQQ was incubated with different concentrations of SC(1-246)-BODIPY (0, 0.5, 1.0, and 1.5 M) at 25°C for 15-30 min. The samples were run on a 6% polyacrylamide gel under native conditions (Tris-glycine buffer, pH 8.3, no SDS) at 4°C. The fluorescence was imaged, and the proteins were stained as described above.

Prothrombin activation assay
Activity titrations of ProT complexes with WT or mutant SC(1-246) were performed in 50 mM HEPES, 110 mM NaCl, 5 mM CaCl 2 , 1 mg/ml PEG 8000, pH 7.4, buffer, in PEG 20,000coated 96-well plates (Nunc). Varying concentrations of WT or mutant SC(1-246) were incubated with 1, 10, or 20 nM ProT for 10 min at 25°C. The reaction was initiated by the addition of 600 M chromogenic substrate S-2238 (Diapharma), and the rate was measured in a ThermoMax plate reader (Thermo Fisher Scientific) for 10 min at 405 nm until the absorbance reached 0.1. Initial rates (mAbs/min) for the mutants were normalized to that of WT SC(1-246). The normalized rate dependences as a function of the SC(1-246) concentration were fitted by the quadratic binding equation using SCIENTIST (MicroMath) to obtain the V lim and K D (19).

Direct and competitive fluorescence equilibrium binding
Fluorescence measurements were performed with a PTI QuantaMaster 30 spectrofluorometer at 25°C using acrylic cuvettes coated with PEG 20,000. Titrations were performed in 50 mM HEPES, 110 mM NaCl, 5 mM CaCl 2 , 1 mg/ml PEG 8000, pH 7.4, buffer with 1 mg/ml ovalbumin, and fluorescence was measured at ex ϭ 496 nm (3-6-nm band pass) and em ϭ 535 nm (4 -6-nm band pass). Two fixed concentrations of SC(1-246)-BODIPY were titrated with the ligand, ProT QQQ . The competitive binding assays were performed with one fixed concentration of SC(1-246)-BODIPY in the presence of a single fixed concentration of unlabeled WT or mutant SC(1-246) competitor, titrated with the ligand, ProT QQQ . The SC(1-246)-BODIPY control titrations with ProT QQQ were performed to obtain the stoichiometry and K O for SC(1-246)-BODIPY. The titrations of SC(1-246)-BODIPY in the presence of competing WT or mutant SC(1-246) were performed to obtain the stoichiometry and dissociation constant K C for the competitors. The fractional change in fluorescence was calculated as (F obs Ϫ F o )/F o ϭ ⌬F/F o , and the data were fit by the quadratic binding equation (37) using SCIENTIST (MicroMath) software. For the competition experiments, titrations in the absence and presence of competitor were fit simultaneously by the cubic binding equation (37). Nonlinear least-squares fitting was performed using SCIENTIST (MicroMath) either with or without fixed stoichiometry for the competitive data to obtain the dissociation constants K O and K C , maximum fluorescence intensities ((F max Ϫ F o )/F o ϭ ⌬F max /F o ), and stoichiometric factors n for SC(1-246)-BODIPY and m for unlabeled WT and mutant SC(1-246). The error estimates represent the 95% confidence interval. The Gibbs free energy (⌬G) values for WT and mutant SC(1-246) binding to ProT QQQ were calculated using the equation, ⌬G ϭ RT ln K D , where R ϭ 1.987 ϫ 10 Ϫ3 kcal mol Ϫ1 degree Ϫ1 and T ϭ 298.15 K (25°C), and K D is expressed in M.

Data availability
The structure of the SC(1-325)⅐Pre2 complex 1NU9 is available in the Protein Data Bank. All remaining data are contained within the article and supporting information.