Staphylokinase has distinct modes of interaction with antimicrobial peptides, modulating its plasminogen-activation properties

Staphylokinase (Sak) is a plasminogen activator protein that is secreted by many Staphylococcus aureus strains. Sak also offers protection by binding and inhibiting specific antimicrobial peptides (AMPs). Here, we evaluate Sak as a more general interaction partner for AMPs. Studies with melittin, mCRAMP, tritrpticin and bovine lactoferricin indicate that the truncation of the first ten residues of Sak (SakΔN10), which occurs in vivo and uncovers important residues in a bulge region, improves its affinity for AMPs. Melittin and mCRAMP have a lower affinity for SakΔN10, and in docking studies, they bind to the N-terminal segment and bulge region of SakΔN10. By comparison, lactoferricin and tritrpticin form moderately high affinity 1:1 complexes with SakΔN10 and their cationic residues form several electrostatic interactions with the protein’s α-helix. Overall, our work identifies two distinct AMP binding surfaces on SakΔN10 whose occupation would lead to either inhibition or promotion of its plasminogen activating properties.


ITC: selectivity and thermodynamics of Sak-AMP interactions.
In order to determine the interaction strength of different antimicrobial peptide-Sak pairings, a series of ITC experiments were performed (Table 1). LfcinB and Tritrp are moderate-affinity binding partners for Sak, with respective dissociation constants of 27.2 μ M and 11.2 μ M. Their isotherms show that they interact with Sak at 1:1 ratios and with exothermic heats of binding. The cathelicidin mCRAMP shows no binding in our tested conditions. This result is unexpected because mCRAMP has previously been found to associate with Sak using polyclonal antibody detection methods 17 . mCRAMP has a net + 6 charge at neutral pH compared to + 7 and + 4 for LfcinB and Tritrp, respectively, therefore there should be a similar potential for electrostatic interactions between these AMPs and a negatively charged surface on Sak. Due to the temperature dependence on binding enthalpy 43 , it is possible that there is a high affinity between the two molecules and that the titration with mCRAMP may have been set at a temperature where enthalpy is minimal. For this reason, we repeated the experiment at 37 °C. This result, as well as the titration data from our NMR experiments, confirms the lack of strong binding (K D < 10 −4 M) between mCRAMP and fulllength Sak (not shown). In the titration of melittin to Sak, the thermogram shows very small endothermic heats of interaction which presented difficulties in curve fitting and determining an accurate dissociation constant.
In a second set of ITC experiments, we used SakΔ N10, an in vivo Sak variant with its first ten residues truncated. Overall, SakΔ N10 was found to be a better binding partner for all AMPs tested (Fig. 1). The dissociation constants for LfcinB and Tritrp are in the low micromolar range. In particular, LfcinB has a K D of 0.37 μ M,  Table 1. Dissociation constants, stoichiometries and thermodynamic parameters determined for antimicrobial peptide binding to staphylokinase as determined by isothermal titration calorimetry. * p-values < 0.05 against corresponding K D 's with Sak using the t-test at the 95% confidence level.
which is two orders of magnitude lower compared to full length Sak. The titration with mCRAMP now shows small endothermic heats approaching the profile observed with melittin. This suggests that the weak association between SakΔ N10 and these peptides may be driven in part by hydrophobic interactions. The curves become mostly flat after 2:1 peptide:SakΔ N10, therefore the stoichiometry value N for mCRAMP and melittin is likely to be 1, in agreement with the stoichiometry observed for LfcinB and Tritrp.   Fig. 1) 37 . Most of the non-proline residues could be assigned except for the first few N-terminal residues and the following residues: Lys57, Lys74, and Ile108. The peaks for these residues have likely undergone broadening due to chemical exchange. While the amide peak for Tyr44 is not present in the Sak spectrum presumably due to chemical exchange with the solvent 37 , it appears in the SakΔ N10 spectrum. For both forms of the protein, the histidines of the C-terminal His 6 -tag were left unassigned. The positions of the vast majority of the amide peaks remain largely unchanged between Sak and SakΔ N10, which indicates that the truncation of the first ten residues does not affect the overall fold of the protein.
The topology of staphylokinase consists of a central α -helix (Lys57-Asp69) adjacent to a short two-stranded β -sheet, all of which lie perpendicular on a five-stranded β -sheet ( Fig. 2A) 37,38 . The N-terminal segment, which encompasses the first ten residues and extends to Glu19, lacks regular secondary structure and folds back to make several contacts to different parts of the protein. However, heteronuclear 15 N-{ 1 H} NOE data indicated that this region has higher conformational variability than the remainder of the protein 37 . In the X-ray crystal structure of Sak, residues 1-15 are in fact missing due to low electron density observed 38 . Using the mean NMR solution structure as a starting point (PDB 1SSN) 37 , residues 1-10 (SSSFDKGKYK) were removed to create a model structure for SakΔ N10 and evaluate its electrostatic surface potential (Fig. 2B). With its missing electron density for the N-terminal residues, the X-ray crystal structure also presents a potential model for SakΔ N10, however there are subtle differences between the structures due to the different experimental approaches (NMR vs. X-ray), which would make it difficult to make a direct comparison. Overall, there are several negatively charged surfaces on Sak interrupted by scattered positive charges. The N-terminal deletion of residues 1-10 lowers the pI of SakΔ N10 from 6.15 to 5.47 and it removes a small cluster of positively charged residues. Lys11 is likely to be very mobile as it is now the new N-terminus of the SakΔ N10 protein. The solvent exposure for all residues of Sak and SakΔ N10 were calculated by MOLMOL 44 and plotted in Fig. 2C. The biggest increases in exposure are found for the new N-terminus Lys11 (38.4% to 65.8%), His43 (29.7% to 39.5%), Tyr44 (25.2% to 39.2%), Ala72 (13.0% to 33.1%), and Glu75 (20.4% to 28.9%). The newly exposed areas are mostly uncharged, which creates a larger uninterrupted anionic surface on the protein.
NMR titration experiments: mapping a surface for AMP interaction. 1 H, 15 N HSQC spectra obtained for titrations of the peptide into 15 N-labeled SakΔ N10 samples are shown in Fig. 3. In these spectra, and with Sak to a lesser degree, some N-H peaks can be observed to undergo varied amounts of line broadening. This indicates that the interacting residues are in intermediate exchange on the NMR timescale. The magnitude of peak intensity losses for the affected residues is in general agreement with the different affinities for each peptide as determined from the ITC data. For mellitin and LfcinB, there are a few peaks that also undergo fast exchange between the unbound and bound conformations, allowing for the chemical shift perturbations to be followed.
The spectra for full-length Sak remain mostly unchanged in titrations with the α -helical peptides mCRAMP and melittin, and there are no clear regions in the protein sequence with significant peak intensity loss (Fig. 4A,B). This result also occurs when mCRAMP is added to SakΔ N10, however there is slight peak movement for Phe18, His43, Tyr44, Val45 and Tyr73. Residues 43 to 45 are part of a bulge region in staphylokinase (Leu40 to Glu46) of irregular secondary structure that follows β 1 and precedes the short β 2 strand (Phe47 to Ile49). The chemical shift perturbations seen here are very low in value compared to usual studies, which likely reflects the weak nature of these interactions.
In the titration of melittin to SakΔ N10, His43 to Glu46 of the bulge region undergo both intermediate and fast exchange. There are more residues involved in the interaction than for mCRAMP. The peaks for Asp14 to Glu19 (excluding Ser16) from the N-terminal region are shifted during the titration, and there are additional peaks that undergo significant intermediate exchange. These new peaks, having lost > 0.5 signal intensity at 1:1 melittin:SakΔ N10, correspond to Ala70, Tyr73, and Lys74. Residues 69 to 76 form a loop region that joins the central α -helix to strand β 3. Unsurprisingly, the residues that have increased solvent exposure after the deletion of residues 1-10 are 43 to 46 and 71 to 75 (Fig. 2C). These neighbouring segments in SakΔ N10 can provide an interface for binding melittin, while mCRAMP seems to interact with residues 43 to 46 exclusively. In the bulge region, the side chains of Tyr44, Val45 and Phe47 offer sites for hydrophobic interactions, and Glu46 for electrostatic interactions with cationic AMPs. The low 70's loop presents a less ideal area for AMP interaction because it contains a somewhat hydrophobic Tyr73 and the negative charge of Glu75 is offset by Lys74. In addition, there are peaks in residues corresponding to the end of the N-terminal fragment (Asp14 to Glu19) that are in intermediate and fast exchange. This region is well suited to extend the binding surface to accommodate a longer α -helical peptide such as melittin.
The HSQC peaks for Sak and SakΔ N10 disappear very quickly upon addition of Tritrp such that the spectra are mostly blank after 1:1 Tritrp:SakΔ N10 (Fig. 3C). Signal intensity losses at the 0.5 molar equivalence point are plotted and corresponding residues that have mostly disappeared, i.e. those with normalized peak intensities < 0.1, are highlighted on the protein backbones in Fig. 5. Here, differences between Sak and SakΔ N10 are not as prevalent as with melittin and mCRAMP. Hence, the potential binding surface for Tritrp is sufficiently large and accessible so that the first ten residues of Sak do not block this interaction, although the ITC data indicate that the binding is weakened. Interacting residues come from the segments that have been previously mentioned: the end of the N-terminal arm (Tyr17 to Glu19), the bulge region (Leu40 to Glu46), and the loop following the α -helix (Asp69 to Phe76). The interaction surface grows larger with residues from the α -helix, which contains the sole Trp residue of Sak at position 66. Altogether, this presents a continuous surface on Sak that has three negatively charged side chains, two long chain aliphatic side chains, and seven aromatic rings. This large hydrophobic area is poised to accommodate Tritrp, especially with opportunities for ring stacking interactions with the five aromatic side chains of the peptide.
The interaction between LfcinB and staphylokinase illustrates another scenario for how the protein's structure accommodates AMPs (Fig. 6). As seen in the HSQC overlay of the titration, both intermediate and fast exchange occur (Fig. 3D, plotted in Fig. 6A). At 2.5:1 peptide:protein, at least half of the peaks retain significant intensity, which is less drastic reduction than the signal disappearance observed with Tritrp. The mixture of LfcinB with SakΔ N10 brings overall higher signal intensity losses compared to LfcinB with Sak, which reaffirms their difference in affinity, however the pattern of participating residues are mostly in agreement. A total of 75 out of the 115 assigned residues in SakΔ N10 undergo some kind of chemical shift perturbations, which suggests that most of the protein may be experiencing some conformational rearrangement. Similar to Tritrp, the interfacial residues involve the bulge region (Leu40 to Phe47) and secondary sites are at the end of the N-terminal arm (Tyr17 to Glu19), the central α -helix (Tyr62 to Leu68) and the following loop that leads to strand β 3 (Asp69 to Phe76). Additionally, the peaks for Val27, Val29 and Thr129 disappear almost completely at the first titration point. These residues belong to the adjacent anti-parallel strands β 1 (V27, V29) and β 5 (T129) and are buried in the interior of the protein, therefore it would be difficult for them to make direct interactions with LfcinB. Moreover, they serve as anchoring residues for the α -helix, therefore the loss in their signal intensities may be due to a movement or partial refolding of the α -helix as opposed to making direct contact with the peptide.
Docking models for SakΔN10 complexes reveal two locations to accommodate AMPs. Using the experimental NMR titration data, we generated data-driven docking models for the complexes of SakΔ N10 with three of the peptides studied here using structures available from the PDB and the HADDOCK web server 45 . The monomeric melittin template was taken from an X-ray crystal structure of the peptide in its aqueous tetrameric state 46 . It is mostly in an α -helical conformation, which is also expected for the peptide when it is membrane-bound. Tritrp requires a membrane-mimicking environment to fold into a converged conformation, and consequently the amphipathic micelle-bound structure of Tritrp 47 was used as an approximation of the peptide in complex with SakΔ N10. The amphipathic β -hairpin solution structure of LfcinB is conformationally restricted by a disulfide bond and was used in the docking calculations.
The ten top scoring clusters of SakΔ N10 in complex with each peptide are superimposed and shown in Fig. 7. Overall, SakΔ N10 does not change conformation drastically with backbone root mean squared deviations less than 0.8 Å between the complexed and starting forms of the protein. In the models, melittin is spatially confined by interactions with the bulge region and the N-terminal arm of SakΔ N10. Within the best scoring cluster of SakΔ N1:melittin complexes, there is one consistent intermolecular electrostatic interaction between Lys21melittin and Glu38SakΔ N10. The original structure of melittin consists of two α -helical segments connected by a slight hinge at Thr11-Gly12 46 . In the models with SakΔ N10, the bending of the peptide about these residues becomes more pronounced. For Tritrp and LfcinB, SakΔ N10 has several possible binding locations. However, the best scoring structure clusters of these complexes have the peptides resting on the α -helix of SakΔ N10 and the subsequent loop, with at least three prominent electrostatic contacts from Glu65 and Asp69 to cationic residues from the peptides.

Discussion
In the present study, our aim was to evaluate staphylokinase as a more general interacting partner for antimicrobial peptides of different structures and origin. Unexpectedly, the peptide that we had initially selected as a positive control, mCRAMP, did not show strong binding in our in vitro experiments. It is possible that additional factors may be involved in promoting its binding in vivo. During staphylococcal pneumonia in an intranasal infection mouse model, the expression of mCRAMP is upregulated and this in turn augments Sak-dependent plasminogen activation by an approximate two-fold factor 17 .
Regardless of the peptide, our data clearly show that SakΔ N10 is a better binding partner compared to the full-length protein. The missing N-terminal portion of Sak has the sequence SSSFDKGKYK. With two aromatic rings and three cationic side chains, this sequence somewhat resembles that of an AMP itself and therefore this portion of the native protein could block a region on the protein that an incoming AMP would have to outcompete. The first fifteen residues exhibit high flexibility in the Sak structure as determined by X-ray crystallography and 15 N NMR dynamics experiments 37,38 . However, the solution structure of Sak indicates that this section loosely folds back onto the protein with many observed NOEs from Phe4, Lys6, Lys8, Tyr9 and Lys10 to His43 and Tyr44 37 , two critical residues involved in AMP binding as identified by our NMR titration data. Although the cleavage at the Lys10-Lys11 peptide bond would normally occur during an activation step by plasmin while Sak is bound to it 1,48 , it seems that this N-terminal processing can be accomplished by a staphylococcal protease as well 42 . To date, very few proteins are known to interact directly with AMPs 49,50 and to our knowledge, we have shown here for the first time the preferential binding of certain AMPs to a proteolytically processed form of an extracellular protein. It is interesting to speculate that this phenomenon may be more widespread, for example proteolysis of various chemokines is known to be quite common, which markedly changes their interaction properties 51 . Clearly future studies should consider the effects of proteolytic or other posttranslational modifications on the binding of AMPs to proteins in the circulatory system. According to our findings, staphylokinase can interact with several AMPs using two distinct modes of binding. Tritrp and LfcinB are associated with a high affinity interaction type that includes several salt bridges with the α -helix of SakΔ N10 (Fig. 7B,C). The interface area is large enough that the peptides are still able to bind to full Sak even though residues 1-10 of the N-terminal arm could give rise to some steric hindrance. Moreover, these two peptides occupy the same space and interact with the same region as a plasminogen substrate during enzymatic action by the plasmin/Sak complex (Fig. 8) 52 . In this complex, the active site belongs to plasmin while Sak acts as a cofactor that restricts a binding subsite for the incoming substrate. According to an X-ray structure of the ternary complex, where smaller microplasmin proteins were used as substitutes for plasmin (PDB ID 1BUI) 52 , the plasminogen side chains His569, Pro570 and His571 extend into the hydrophobic groove consisting of Tyr62, Trp66, Ala70 and Tyr73 from the Sak α -helix-β 3 loop. Furthermore, mutational studies have identified residues in this so-called "activation epitope" that affect plasminogen activation without changing the affinity of the plasmin/Sak enzyme complex: Val45, Phe47, Tyr62, Tyr63, Trp66, and Ala70. Our experimental data also shows that interactions between LfcinB and the α -helix can affect the conformation of residues belonging to the In the case of melittin and mCRAMP, there is weaker affinity with more pronounced inhibition by the N-terminal segment in full Sak. These peptides belong to the canonical class of AMPs that gain α -helical structure in a membranous environment, however it is unknown if they fold into such conformations upon binding to SakΔ N10. The attachment location for these peptides is more restricted to the opposite face of the bulge region away from the activation epitope and has a limited interface composed of residues in the bulge area and Tyr17-Glu19 of the flexible N-terminal arm (Fig. 7A). This positions the peptide in proximity with the plasmin partner that attaches to Sak that forms the activating complex, however this may not present a direct interference in binding as plasmin has several contacts with the larger face of the β -sheet, specifically with strands β 3, β 5, β 1 and β 2 52 . In the plasmin/Sak enzyme complex, there are salt bridges from Lys769 and Arg719 of plasmin to Glu19 and Glu46 of Sak, respectively. The effect of plasmin binding causes residues 43 and 44 of Sak to protrude into the activation epitope and restricts the substrate binding pocket. Furthermore, this smaller space more closely resembles the binding pocket offered by urokinase and tissue plasminogen activator which are the native plasminogen activators in humans. Given the slight effects that mCRAMP and melittin have on His43 and Tyr44 of SakΔ N10 without influencing many more residues on the face of the activation epitope, we hypothesize that the peptides facilitate the plasmin-mediated fine-tuning of the substrate pocket. This would explain the doubled plasminogen activating activity of Sak in the presence of mCRAMP that promotes S. aureus invasion 17 .
The bulge residues 43 to 47 remain the focal point for both types of interaction with Sak, and the removal of the N-terminal arm from these residues facilitates either substrate or AMP binding. Previous molecular models have emphasized the importance of these residues, with His43 and Tyr44 making crucial cation-pi and pi-pi interactions with the Trp215 side chain of plasmin 2 . However, we did not observe any of the Trp indole rings of Tritrp or LfcinB to be oriented in such a manner in our data-driven docking models.
The structural analysis of staphylokinase presented here reveals versatility in its intermolecular interactions, and the two different modes of binding for AMPs may lead to different consequences for substrate binding and functional activation. The clinical development of Sak as a thrombolytic agent may be aided by an accompanying peptide that boosts its activity such as mCRAMP or melittin, however the latter is a strongly cytotoxic peptide, limiting its clinical usefulness. As well, we should re-emphasize the role of staphylokinase as an AMP sequestering agent. We observed a 1:1 binding ratio for the peptides studied here, however this may not be the case for all peptides. In fact, a molar ratio as low as 1:6 Sak:HNP was found to neutralize the peptide's bactericidal activity, suggesting that Sak can influence oligomerization of some AMPs 15 .
Altogether, our results reveal two major binding pockets that are present on the surface of staphylokinase to accommodate interactions with antimicrobial peptides, and large anionic surfaces on the protein allow for the binding of peptides of different sizes and structures. These results highlight the importance of SakΔ N10 as an important part of the resistance mechanism of S. aureus against host innate immune defenses.

Materials and Methods
Most peptides were purchased from Genscript (Piscataway, NJ) as custom-synthesized peptides synthesized using standard 9-fluorenylmethyoxycarbonyl (Fmoc) chemistry and they were purified by HPLC to > 95% purity. Melittin from honey bee venom (> 85% purity) was purchased from Sigma-Aldrich. The amino acid sequences of the peptides are as follows: LfcinB, FKCRRWQWRMKKLGAPSITCVRRAF with a disulfide bond between the underlined Cys residues; mCRAMP, GLLRKGGEKIGEKLKKIGQKIKNFFQKLVPQPEQ; melittin, GIGAVLKVLTTGLPALISWIKRKRQQ-NH 2 ; Tritrp, VRRFPWWWPFLRR-NH 2 . For the purpose of this study, Tritrp refers to the C-terminally amidated version of tritrpticin which has more potent antimicrobial properties than the original non-amidated peptide 47,53 . The concentrations for most of the peptides were determined by UV absorption at 280 nm using molar extinction coefficients determined by ProtParam 54 . In the case of mCRAMP which does not contain any Trp or Tyr residues, UV absorption at 214 nm was measured and a molar extinction coefficient was calculated from its amino acid content 55 . Protein expression and purification. A synthetic gene for staphylokinase with optimized codons for expression in Escherichia coli was purchased from GeneArt (Life Technologies). Staphylokinase genes have been cloned and sequenced from different origins: the bacteriophages sakϕC and sak42D, and the genomic DNA of a lysogenic S. aureus strain (sakSTAR) 56 . These indicate amino acid sequence differences at three positions in the coding sequences. In this study, we have worked with the sakSTAR protein sequence as it is the most commonly used variant 56 . The gene was subcloned into a pET29b (Novagen), giving the expressed protein a C-terminal His 6 -tag when expressed in E. coli BL21 (DE3) cells. Site-directed mutagenesis was performed with designed primers to make a gene construct for SakΔ N10, the more active form of Sak with its first ten residues deleted. 15 N-labeled or 13 C, 15 N-labeled Sak or SakΔ N10 were expressed in a minimal M9 medium containing 0.5 g/L 15 NH 4 Cl, and 3 g/L of either 13 C-labeled or unlabeled glucose. At an optical density at 600 nm of ~0.6, the cells were induced with 0.5 mM isopropyl β -D-1-thiogalactopyranoside for 4 h at 37 °C. The cells were collected by centrifugation and put through at least 3 passes through a French press in lysis buffer: 20 mM Tris, 100 mM NaCl, pH 7.4. The cell lysate was then clarified by high speed centrifugation, after which the supernatant was applied onto a column with chelating-Sepharose fast flow resin (GE Healthcare) loaded with NiCl 2 . The column was washed with a buffer containing 50 mM imidazole, and the protein of interest was eluted with 400 mM imidazole. The purified protein was dialyzed extensively in 8 mM NH 4 HCO 3 and lyophilized. SDS-PAGE indicated that the protein was > 95% pure.
ITC. All ITC runs were performed at 25 °C on a MicroCal VP-ITC microcalorimeter (Malvern, Westborough, MA). 200-400 μ M peptide was titrated into 8 μ M SakΔ N10 or Sak in 20 mM HEPES, pH 7.4. In preliminary trials, the inclusion of salt (100 mM NaCl) in the ITC buffer was found to promote peptide aggregation in the injection syringe as indicated by endothermic heats of dilution in peptide-to-buffer control experiments. Data analysis was performed using the MicroCal Origin software using the "One Set of Sites" binding model. At least two experiments were performed for each titration. Statistical analysis was performed to confirm the statistical validity of the differences in the K D values determined for LfcinB and Tritrp binding to SakΔ N10 or Sak. p-values were calculated from Student's t-tests at the 95% confidence level with a one-tailed hypothesis. NMR measurements. All NMR experiments were performed at 298 K on a Bruker Avance 700 MHz spectrometer equipped with a triple resonance inverse Cryoprobe with a single axis z-gradient. An initial 0.15-0.20 mM 13 C, 15 N-labeled Sak sample was dissolved in 20 mM sodium phosphate, 100 mM NaCl, 1 mM EDTA, and 0.5 mM 2,2-dimethyl-2-silapentane-5-sulfonate (DSS), pH 6.9, supplemented with 10% 2 H 2 O. To complete the assignment of the main chain NMR signals, a two-dimensional [ 1 H-, 15 N]-HSQC and three-dimensional HN(CA)CO, HNCO, HNCA, HN(CO)CA, HNCACB, and CBCA(CO)NH experiments were acquired with a non-uniform sampling scheme in the indirect dimensions. All spectra were processed using NMRPipe 57 and analyzed using the NMRView software package 58 . The three-dimensional experiments with non-uniform sampling were reconstructed with MDDNMR 59 interfaced with MDDGUI 60 . The backbone assignments for Sak were compared and verified with those reported in a previous publication 37 .