Conformationally organized lysine isosteres in Streptococcus pyogenes M protein mediate direct high-affinity binding to human plasminogen

The binding of human plasminogen (hPg) to the surface of the human pathogen group A Streptococcus pyogenes (GAS) and subsequent hPg activation to the protease plasmin generate a proteolytic surface that GAS employs to circumvent host innate immunity. Direct high-affinity binding of hPg/plasmin to pattern D GAS is fully recapitulated by the hPg kringle 2 domain (K2hPg) and a short internal peptide region (a1a2) of a specific subtype of bacterial surface M protein, present in all GAS pattern D strains. To better understand the nature of this binding, critical to the virulence of many GAS skin-tropic strains, we used high-resolution NMR to define the interaction of recombinant K2hPg with recombinant a1a2 (VKK38) of the M protein from GAS isolate NS455. We found a 2:1 (m/m) binding stoichiometry of K2hPg/VKK38, with the lysine-binding sites of two K2hPg domains anchored to two regions of monomeric VKK38. The K2hPg/VKK38 binding altered the VKK38 secondary structure from a helical apo-peptide with a flexible center to an end-to-end K2hPg-bound α-helix. The K2hPg residues occupied opposite faces of this helix, an arrangement that minimized steric clashing of K2hPg. We conclude that VKK38 provides two conformational lysine isosteres that each interact with the lysine-binding sites in K2hPg. Further, the adoption of an α-helix by VKK38 upon binding to K2hPg sterically optimizes the side chains of VKK38 for maximal binding to K2hPg and minimizes steric overlap between the K2hPg domains. The mechanism for hPg/M protein binding uncovered here may facilitate targeting of GAS virulence factors for disease management.

Group A Streptococcus pyogenes (GAS) 2 is a human-selective pathogen that causes both superficial self-limiting infections (e.g. pyoderma) as well as morbid and lethal maladies (e.g. toxic shock syndrome and necrotizing fasciitis). Although penicillintype antibiotics are effective drugs against GAS diseases, more virulent pathovars have emerged because of the ability of this microbe to remodel its genome to adapt to specialized environmental niches within the host. These changes primarily occur through intra-and interspecies horizontal transfer and recombination of genetic materials, including integration of prophages and their associated virulence genes (e.g. superantigens) (1,2).
GAS employs various virulence strategies to perpetuate in its host and cause disease. These include a variety of proteins that serve as adhesins to attach to host epithelial cell surfaces (3) as well as an outer nonimmunogenic capsule to resist antibody recognition (4) and secreted and surface-bound proteases to encourage invasion of the bacteria through natural cellular barriers (5). Further, the ability of the bacteria to form protective biofilms (6) allows colonization of the bacteria in distal tissues within the host. Among these virulence determinants, the single-chain, surface-resident multifunctional M protein, present as a single type on all GAS strains, and utilized for GAS serotyping, is of particular relevance because it is used by GAS for a variety of survival benefits (e.g. adhesion to host cells (7), evasion of phagocytosis (8), and neutralization of antimicrobial peptides (9)).
One subtype of M protein, plasminogen-binding group A streptococcal M-like protein (PAM), has an additional wellknown function of coopting components of the host to evade innate immune responses. Important among these properties is the fibrinolytic system, which GAS employs to potently evade innate immunity (e.g. attenuate complement-based opsonization of GAS) (10,11) and to disseminate into deep tissue. To accomplish this function, PAM first interacts with very high affinity to host human plasma plasminogen (hPg) (12), a step that facilitates its activation to plasmin by GAS-secreted streptokinase (13,14). The proteolytic ability of plasmin that is localized on the GAS surface is resistant to natural plasmin inhibitors (13) and is thus utilized by the bacteria to disrupt barriers to its dissemination (e.g. fibrin that encapsulates GAS) and then This work was supported by National Institutes of Health Grant HL013423.
The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains supplemental Tables 1S and 2S and Fig. 1S cro ARTICLE to invade into deeper distal tissues by disrupting the extracellular matrix and tight cellular junctions (15,16). PAM is a ϳ42-kDa single-chain protein, assembled as a dimer in the mature protein, that has evolved in a modular fashion. The hypervariable N terminus of PAM-type M proteins contains two short A domain repeats (a1a2) that have been identified as the binding determinants for hPg (17,18).
Within hPg, the kringle 2 (K2 hPg ) domain has been identified as the receptor for PAM binding (19). Kringle modules in hPg generally interact with their receptors via C-terminal lysine residues (20 -24). However, PAM does not possess a C-terminal lysine. Therefore, the X-ray crystal structure (25) and NMR solution structure (26,27) of the complex of a truncated form of the a1a2 domain of PAM53 from GAS isolate AP53 (VEK30; Table 1) was solved to propose a mechanism for this tight binding interaction. The resulting structural model that was developed involved the formation of an internal conformational isostere of lysine from side chains of VEK30 (His 17 , Arg 18 , and Glu 20 ). Their proper orientation depended on the adoption of a VEK30 ␣-helix upon binding. However, the K D of this interaction was ϳ30 -50-fold weaker than that of the PAM53-K2 hPg complex. A redesign of VEK30 resulting from an addition of two C-terminal amino acids ( 30 RH 31 ) from PAM53 or five C-terminal amino acids ( 30 RHDHD 34 ) from PAM53 (Table 1) strengthened the binding affinity of the resulting peptides (VEK32 and VEK35, respectively) to that of full-length PAM53 (12). We then decided to assess the role of these additional amino acids in a1a2 binding to K2 hPg via functional and highresolution structural analyses.
The results presented herein demonstrate for the first time that a peptide, VKK38, from a PAM-containing isolate, NS455, harbors two binding sites for K2 hPg. Upon binding to K2 hPg , VKK38 undergoes a conformational transition from a predominantly random coil to an end-to-end ␣-helix. Overall, the data obtained reveal a mechanism for binding of PAM to hPg that will facilitate the targeting of these virulence factors of GAS in disease management.

Amino acid sequence alignment for PAM53 and PAM455
The PAM-subtype M protein is present in all pattern D GAS serotypes, and its sequence variability in different GAS isolates provides natural mutations in which to assess the nature of its various functions. The most well-studied PAM was originally extracted from GAS isolate AP53 (18). GAS NS455 is another pattern D direct hPg-binding isolate (28), and we aligned PAMs from AP53 (PAM53) and NS455 (PAM455) to determine the extent to which these two proteins from different pattern D strains were similar ( Fig. 1).
At the N termini, the signal peptides (residues 1-41) of the two PAMs exhibit very high homology; however, low homology was observed in the immediate downstream hypervariable domains, thus accounting for their different serotypes (emm53 for AP53 and emm52 for NS455). Following the hypervariable region are the a1 (residues Val 97 -Glu 115 ) and a2 (residues Glu 116 -Asp 130 ) repeats in PAM53 and, correspondingly, Val 103 -Glu 124 and Glu 125 -Asp 139 in PAM455. Both proteins have two RH dipeptides, one in the a1 region ( 113 RH 114 in PAM53 and 119 RH 120 in PAM455) and another in the a2 segment ( 126 RH 127 in PAM53 and 135 RH 136 in PAM455). These RH residues, supported by Glu 125 in PAM53, are most relevant for formation of the lysine isostere needed for hPg binding. hPg binding occurs at RH1 and/or RH2. It is noteworthy that at the end of the a1 repeat in PAM455, a tripeptide insertion, 121 VHD 123 , was found that interrupted the relative orientation of Glu 125 with RH1. One central issue in this study is whether this VHD natural insert would make a significant difference in the hPg-binding pattern and structure of PAM455 compared with PAM53.
Other important domains are present in all M proteins. In PAM-type M proteins, downstream of the a2 region, somewhat higher homology is seen in the B-like repeats. The C1-C3 repeats, the D-repeats, and the Pro-Gly (PG) region, used for pattern mapping, have high homology between the PAM proteins. Both PAMs have been identified as pattern D subtypes with high tropism for skin infections (29). In addition, the downstream sortase A motifs were identical, thus ensuring that both mature PAMs are covalently anchored to the cell wall.

Binding affinities of PAMs to hPg and VEK/VKK peptides to K2 hPg
The binding affinities of full-length PAMs to hPg or of VEK/ VKK peptides (listed in Table 1) to K2 hPg were assessed by surface plasmon resonance (SPR). The prototype PAM53 has strong binding ability to hPg, with a K D of ϳ1 nM, consistent with previous results (12). Despite PAM455 containing a naturally occurring 121 VHD 123 motif immediately following the RH dipeptide in the a1 repeat (RH1), no effects on the binding affinity of PAM455 to hPg were found ( Table 2). This is also true for the mutation D117A in PAM53 as well as the insertion () of 115 VHD 117 or 115 VHA 117 into PAM53 at the end of the a1 domain. In addition, mutation of PAM455 to a form without the VHD tripepetide (PAM455/⌬ 121 VHD 123 ) also did not show The GS dipeptide outside the parentheses is an exogenous product of the expression cassette and is not used in the numbering system. The C-terminal Tyr residue is included in all peptides to determine concentrations by A 280 nm . b The residues of the binding sites are italicized, and the VHD insertions are shown in bold. c Truncated peptides published previously (12) are listed for comparison. a significant difference in hPg binding. Therefore, the presence of VHD at the end of the a1 region does not affect the binding of PAM to hPg.
The truncated peptides, including a1a2 repeats from AP53 and NS455, have been constructed and expressed in Escherichia coli BL21 cells. The expressed peptides are numbered from the Val 1 at the beginning of the a1 domains; VEK peptides are from PAM53, and VKK peptides are from PAM455. Insertion of 19 VHD 21 into VEK peptides, VEK30 (VEK30/ 19 VHD 21 ), VEK32 (VEK32/ 19 VHD 21 ), and VEK35 (VEK35/ 19 VHD 21 ), does not change their binding affinities to hPg (Table 3). VEK30/ 19 VHD 21 exhibits a high dissociation rate; thus, we confirmed its K D to K2 hPg using the SPR affinity method. A value of 59 Ϯ 3 nM was determined for the K D of this interaction, which is comparable with that obtained from kinetic analysis of the binding (Fig. 2). It is noted that the binding affinity observed from VEK30 and VEK30/ 19 VHD 21 is about 50-fold weaker compared with that of VEK32 and VEK35, because VEK30 only harbors a single RH motif. Furthermore, VKK38, as the counterpart of VEK35, but with the 19 VHD 21 motif, binds to K2 hPg with a K D of ϳ1 nM, a value very similar to that of VEK35 (1.9 nM). Thus, both sets of binding data, from full-length PAMs to hPg and from VEK/VKK peptides to K2 hPg , indicate that the VHD tripeptide that naturally occurs in some PAMs does not significantly affect the binding of these ligands to hPg and K2 hPg .

VKK38 forms a complex with K2 hPg with a molar ratio of 1:2
Because of the two RH groups in VKK38, it is important to demonstrate whether each RH motif binds to K2 hPg separately. The binding properties of VKK38 to K2 hPg were measured by combining isothermal titration calorimetry (ITC), analytical ultracentrifugation (AUC), and NMR titrations.
ITC measurements were carried out by titrating a constant level of K2 hPg in the cell with variable amounts of VKK38 and measuring the incremental liberation of heat (Fig. 3A). The exo- Figure 1. Amino acid sequence alignment of full-length PAMs from AP53 and NS455. The translated sequences were aligned by ClustalX. Despite differences in the number of residues, both sequences contain a 41-residue signal peptide at their N termini, followed by a hypervariable (HV) region, PAM-specific a1a2 repeats, a B-like domain, C1-C3 repeats, a D domain, a Pro-Gly (PG) region, a sortase A reactive motif, and last a membrane anchor region at C termini with a small intracellular region. The tripeptide sequence included in VKK38 at the end of the a1 domain is marked by a black rectangle. *, translation stop codon.

PAM binds plasminogen via two lysine isosteres
thermic reaction leveled after ϳ40 M of VKK38, implying saturation of all binding sites on VKK38 by K2 hPg . At this stage, the concentration of K2 hPg was ϳ95 M, and the molar ratio of VKK38 -K2 hPg was calculated as 0.42, suggesting a binding stoichiometry of 2:1 (m/m) for K2 hPg -VKK38.
These results were confirmed by determining the molecular weight of the VKK38 -K2 hPg complex utilizing AUC (Fig. 3B).
To obtain the needed sample, VKK38 was mixed with excess K2 hPg , and the tightly bound complex of VKK38 and K2 hPg was separated from free components by passing the solution over a gel filtration column. The sequence-based calculated molecular masses of VKK38 -K2 hPg at a 1:1 (m/m) ratio and 1:2 (m/m) ratio were 14,451 and 25,170 Da, respectively. From AUC analysis, the apparent molecular mass of the complex was 22,890 Ϯ 230 Da, suggesting that VKK38 possesses two binding sites for K2 hPg . 1 H- 15 N HSQC experiments were recorded on [ 15 N]VKK38 samples titrated with increasing concentrations of unlabeled K2 hPg . A tight VKK38 -kringle 2 complex is suggested by the observation of two cross-peaks for a subset of the residues during the titration. Amide resonances of the residues at N and C termini of VKK38 were continually shifted until a plateau was reached at an ϳ1:2 (m/m) ratio of VKK38 to K2 hPg (Fig. 4A), again suggesting that two binding sites for K2 hPg are present on VKK38 at the 1:2 VKK38/kringle 2 ratio.
Two mutant peptides, VKK38/R17A/H18A and VKK38/ R34A/H35A, were separately designed by replacing RH1 and The experimental data were fitted using a 1:1 Langmuir binding model. The binding isotherms of full-length PAMs to hPg were assayed under similar conditions. The corresponding binding constants for these curves are provided in Table 2. Table 2 Binding isotherms of full-length PAMs to hPg k on , k off , and k off /k on (K D ) values were determined separately for three independent experiments for each M-protein. The values of k on , k off , and K D were separately averaged, and the mean Ϯ the S.E. was determined for each parameter.

PAM binds plasminogen via two lysine isosteres
RH2 with alanines to confirm the nature of the binding sites for K2 hPg on VKK38. AUC analysis was then performed on the complex of VKK38/R17A/H18A-K2 hPg and VKK38/R33A/ H34A-K2 hPg . Molecular masses of 12,500 -15,500 Da were obtained for these complexes (supplemental Fig. 1S). These results further confirm that VKK38, comprising the a1a2 domain of PAM, contains two binding sites for K2 hPg , and the binding is mediated by RH motifs in each repeat of the a1a2 domain.

Solution structures of VKK38 in the nonbound and K2 hPgbound forms
The structure prediction for VKK38 in its nonbound form was performed by use of the TALOS-N program on the basis of chemical shift data (Fig. 5A). NOESY spectra were analyzed to define the secondary and tertiary structure of apo-VKK38. However, most residues of apo-VKK38 only have sequential connectivities, suggesting a lack of secondary structural elements. The N-and C-terminal regions, residues Val 1 -Asp 6 and Leu 29 -Tyr 38 , exist as random coils in VKK38. However, VKK38 has helical character formed by residues Glu 7 -Lys 14 and Glu 22 -Arg 28 . These two ␣-helical segments are connected by a loop from residue Asn 15 to Glu 22 , which contains RH1.
The conformation of VKK38 changes significantly when it forms a complex with K2 hPg , as has been demonstrated from chemical shift changes (Fig. 4B), secondary structure predictions using TALOS-N, and the solution structures generated by Xplor-NIH along with NOE restraints (Fig. 5C). Bound VKK38 shows an ␣-helix from Asn 5 to Asp 37 . It is noted that in the nonbound form, the 19 VHD 21 motif and the second RH group ( 33 RH 34 ) are located in flexible loop regions but become more rigid in the complex. Compared with RH2 ( 33 RH 34 ), RH1 ( 17 RH 18 ) remains in a less structured region, which bends the helix in this region.
As shown in Fig. 5C, the two RH groups are positioned on opposite faces of the VKK38 helix. However, RH2 is close to the C terminus, which leaves sufficient space for VKK38 to bind two K2 hPg molecules.
Importantly, we observe that when the bound forms of 17 RH 18 in VKK38 and VEK30 are overlaid, the position of Glu 20 in VEK30 is replaced by Asp 21 in VKK38 (Fig. 5D). On the same face with residues 17 RH 18 and Asp 21 , other charged residues are present (viz. Glu 7 , Lys 14 , Asp 21 , Glu 25 , Glu 32 , and Arg 28 ). Meanwhile, the 33 RH 34 group with Asp 37 exists on the opposite face with more uncharged residues (viz. Asn 5 , Val 8 , Val 19 , and Leu 26 ).

Structural model of the VKK38 -K2 hPg complex
A previous structural study of the VEK30 -K2 hPg complex has shown that in K2 hPg , a hydrophobic aromatic groove formed by Tyr 35 , Phe 40 , Trp 60 , Trp 70 , and Tyr 72 , combined with an anionic center formed by Asp 54 and Asp 56 and a cationic locus at Arg 69 , was predicted to serve as the binding site for RH1 of VEK30 (25). The interactions observed at RH1 from the VEK30 -K2 hPg complex were employed in Xplor-NIH to predict the binding model of the first K2 hPg with VKK38 at the RH1 lysine isostere. This structure model was then used as an initial starting point in HADDOCK (30) to generate a model for the complex of the second K2 hPg with VKK38 at RH2 (Fig. 6A). The simulated binding model shows that residues Arg 17 and His 18 of RH1 in VKK38 have hydrogen bond interactions with residues Tyr 35 , Asp 56 , Trp 70 , and Tyr 72 of one K2 hPg molecule, similar interactions predicted for those found between VEK30-RH1 and K2 hPg (Fig. 6, B and C). However, in VKK38, due to the insertion of VHD, Glu 20 of VEK30 has been displaced from the lysine isostere and cannot interact with Arg 69 of K2 hPg . In its place, a new hydrogen bond is formed between Asp 21 of VKK38 and Lys 68 of K2 hPg . Similarly, Arg 33 , His 34 , and Asp 37 have

PAM binds plasminogen via two lysine isosteres
hydrogen bond interactions with Asp 54 , Asp 56 , and Arg 69 of the second K2 hPg molecule. 15 N longitudinal (spin-lattice) and transverse (spin-spin) relaxation rates (R 1 ϭ 1/T 1 ; R 2 ϭ 1/T 2 ) of the VKK38 in its nonbound and K2 hPg -bound forms show a bell-like profile with highly mobile N and C termini. The R 2 /R 1 ratio along the sequence falls in the region of 2.6 -4.0 for the nonbound form of VKK38 (Fig. 7A). This indicates that 15 N relaxation is mostly dominated by higher-frequency motions. However, the R 2 /R 1 ratio is dramatically changed to a range of 50 -80 for most of the residues of VKK38 bound to K2 hPg due to the changes in 15 N backbone dynamics as well as the increase of molecular weight from the addition of two K2 hPg domains to VKK38. On the basis of the R 2 /R 1 ratio, global correlation

PAM binds plasminogen via two lysine isosteres
time ( m ) estimates of 3.8 and 10.7 ns were obtained for nonbound VKK38 and VKK38 in complex with K2 hPg , respectively.
Heteronuclear 15 N-NOE values for nonbound VKK38 decrease toward both the N and C termini (Fig. 7B), suggestive of fast motions and high flexibility at each end. Comparably higher 15 N NOE enhancements have been observed when VKK38 is bound to K2 hPg , especially in the areas of RH1 and RH2. The average 15 N NOE value is 0.78 for the region between Asp 6 and Asp 35 , indicating that a rigid structure is formed in VKK38 upon binding to K2 hPg .

Structural changes in VKK38 upon its binding to K2 hPg
In the current study, we provide a high-resolution characterization of the hPg-binding a1a2 domain (VKK38) of PAM455 in complex with its binding partner in hPg, K2 hPg , using NMR and molecular dynamics approaches. Two RH motifs (RH1 and RH2) in VKK38, which are indispensable for interacting with hPg, were found in flexible peptide regions in the absence of K2 hPg , suggesting that the binding sites for K2 hPg are exposed and readily available to interact with K2 hPg .
For comparison, the structural model of VEK30 in the nonbound form was predicted on the basis of the chemical shift data of this peptide using TALOS-N and CS-Rosetta programs (31). It is noted that in the nonbound form, the RH1 motif in VEK30 is also located in a flexible region of this peptide (Fig.  5B). In VKK38, the random secondary structures between Ala 15 and Asp 21 and between Leu 29 and Tyr 38 include the critical RH1 and RH2 dipeptides. However, upon binding to K2 hPg , both RH1 and RH2 are structured as part of an end-to-end helix in the bound conformation of VKK38. These data imply that a1a2 repeats may adapt better to more flexible and exposed conformations that better fit the binding sites for K2 hPg . Subsequently, the binding of VEK and VKK peptides to their receptor, K2 hPg , induces a conformational transition allowing both VEK30 and VKK38 to adopt a fully helical form.

The role of VHD motif
On the basis of the structure of VEK30/K2 hPg determined previously (25)(26)(27), a model was proposed in which the lysinebinding site of K2 hPg interacts with residues Arg 17 , His 18 , and Glu 20 of VEK30 (Arg 113 , His 114 , and Glu 116 in PAM numbering). The charged side chains of these residues, contained

PAM binds plasminogen via two lysine isosteres
within a single turn of the VEK30 helix, assume a spatial arrangement that is isosteric with the carboxylate and ⑀-amino groups of a C-terminal lysine residue that is necessary for binding of hPg to its known receptors. Thus, because neither VEK30 nor PAM contain a C-terminal lysine, these three residues could be critical for VEK30 binding to K2 hPg . Our present study shows that the insertion of the naturally occurring VHD motif between 17 RH 18 and Glu 20 does not significantly affect the binding of PAM to hPg, as seen from the measured binding constants for full-length PAM, the mutant PAMs, and the truncated peptide variants.
Comparisons of the solution structures of VKK38 and VEK30 (Fig. 5D) suggest that the VHD tripeptide insertion in VKK38 (and in PAM) might not significantly change the orientation of the RH dipeptide (i.e. 33 RH 34 in VKK38 and 135 RH 136 in PAM455), which is needed to maintain strong binding to hPg. Meanwhile, hydrogen bond interactions between RH1, at residues Arg 17 , His 18 , and Asp 21 in VKK38, and K2 hPg have been observed from the binding model (Fig. 6B).
The residues 17 RH 18 and Glu 20 in VEK30 ( 113 RH 114 and Glu 116 in wild-type PAM53) are spaced by two amino acid residues and juxtaposed to one another, an arrangement that provides the bipolar ligand required of the lysine-binding site in K2 hPg . Instead of Glu 20 in VEK30 (or Glu 116 in PAM53), residue Asp 21 in VKK38 (Asp 123 in PAM455) provides the negatively charged environment for interacting with K2 hPg . The corresponding bipolar ligand could also be composed by 17 RH 18 and Asp 21 in VKK38. Thus, we propose that in PAM455 and other strains of GAS containing PAM with the VHD inclusion, 119 RH 120 and Asp 123 , which occupy the positive and negative ends of the lysine-binding site of K2 hPg , provide the pseudolysine for the binding to K2 hPg .
Based on the same mechanism, and in confirmation of this conclusion, the insertion of the VHD tripeptide in VKK30, VEK32, or VEK35 does not significantly affect the binding affinities of the peptides or their full-length counterpart PAMs with hPg or K2 hPg ( Table 1).

The role of the RH motifs in the a1a2 domain of PAM
PAM is a virulence factor expressed by pattern D strains of GAS. However, no structure for this protein is available despite its pathological significance. PAM was previously shown to be a dimer (12,32). Two RH motifs, located in flexible region of the a1a2 domain of PAM, play important roles in its binding to host hPg. Residues Arg 113 and His 114 of the a1 repeat of PAM have been shown to make numerous salt bridges and hydrophobictype electrostatic interactions with recombinant K2 hPg , forming a pseudoligand similar to the lysine analogue, ⑀-aminocaproic acid (25). It is thus likely that the corresponding residues Arg 126 and His 127 in the a2 repeat of PAM interact with hPg in a similar fashion. Our current study for a truncated PAM clearly shows that each VKK38 peptide contains two binding sites for K2 hPg , which raises the question as to whether each PAM molecule could bind two hPg molecules. In the dimeric forms of PAM and these peptides, it is possible that one of the RH sites is masked. Thus, further structural studies of dimeric forms of extended a1a2 domains are required, and these are best studied by X-ray crystallography.

Bacterial strains
Pattern D GAS isolates AP53 and NS455 were provided by G. Lindahl (Lund, Sweden) and M.J. Walker (Queensland, Australia), respectively. Both strains were cultured on sheep blood agar plates or in Todd-Hewitt broth (BD Bacto, Franklin Lakes, NJ) supplemented with 1% (w/v) yeast extract (THY) at 37°C in an atmosphere of 5% CO 2 .

PAM binds plasminogen via two lysine isosteres Protein expression plasmids
To construct the PAM expression plasmids, the coding sequence for PAM from AP53 (PAM53), comprising amino acid residues 42-392, which did not contain the N-terminal signal peptide and C-terminal LPXTG cell membrane anchoring domain, was amplified from GAS AP53 genomic DNA with primers pamF and pamR. The latter primer introduced a coding sequence for a His 6 tag at the 3Ј-end of the pam gene to aid purification. After double digestion with NcoI and EcoRI, the PCR product was ligated into NcoI/EcoRI-digested pET28a.

PAM-derived peptide expression plasmids
The construction of peptide expression plasmids was performed as described previously (27). PAM-derived VEK/VKK peptides from GASAP53 and NS455, respectively, were expressed in E. coli BL21 (DE3) employing the His 6 -tagged streptococcal protein GB1 domain fusion expression system. The final constructs contained, sequentially from the 5Ј-end, an ATG initiation codon, a His 6 tag for purification purposes, the GB1 domain for enhanced solubility, a 9-residue linker, and a thrombin cleavage site, LVPR2GS. DNA fragments containing the VEK/VKK coding sequences were inserted downstream of this cassette into the bacterial expression vector, pET-15b (Novagen, Gibbstown, NJ). A translation stop codon was placed immediately downstream of the open reading frame. Thus, all peptides cleaved with thrombin possessed a GS dipeptide at their N termini. In addition, a Tyr residue was intentionally placed at the C termini of each peptide for 280-nm absorption properties (12).
E. coli Top10 (Invitrogen) cells were transformed by electroporation. Clones that harbor plasmids inserted with correct genes were screened by DNA sequencing.

Full-length PAM gene expression and protein purification
To express PAM53 and PAM455, E. coli BL21 (DE3) cells (New England Biolabs) were transformed with the corresponding expression plasmids. The overnight cultures of the transformed cells were inoculated at 1% (v/v) in 1 liter of LB broth plus 40 g/ml kanamycin and incubated at 37°C until an A 600 nm of 0.6 -0.8 was reached. Protein expression was induced by the addition of 0.8 mM isopropyl-1-thio-␤-D-galactopyranoside. The resulting cultures were grown for 5 h at 37°C and then centrifuged. The resulting cell pellets were resuspended in 40 ml of binding buffer, which contained 50 mM Tris, 300 mM NaCl, 10 mM imidazole, pH 8.0, 500 g/ml lysozyme, 1 mM PMSF. The cells were then disrupted by sonication, and supernatants were collected after centrifugation at 10,000 rpm for 30 min at 4°C (14).
The supernates containing the tagged proteins were loaded onto a Ni 2ϩ -Sepharose affinity chromatography column (His-Trap HP, GE Healthcare) at 4°C. After the column was washed with binding/wash buffer (50 mM Tris, 300 mM NaCl, 40 mM imidazole, pH 8.0), the proteins were eluted with a solution containing 50 mM Tris, 300 mM NaCl, 250 mM imidazole, pH 8.0. The eluates contained purified PAM53 or PAM455, and their molecular weights were confirmed by MALDI-TOF mass spectrometry on an Autoflex III system (Bruker Daltonics, Bremen, Germany).

PAM-derived peptide expression and purification
Unlabeled truncated VEK and VKK peptides (Table 1)  C]VKK38 was selected and grown in 2 ml of LB broth for ϳ4 h. The cells were collected by centrifugation, resuspended in 100 ml of prefiltered M9 medium, and grown at 30°C overnight. The overnight culture was then inoculated into 1 liter of prefiltered M9 medium with 100 g/ml ampicillin and grown at 37°C to an A 600 nm of ϳ2.0. After induction with 0.2 mM isopropyl-1-thio-␤-D-galactopyranoside for 5 h at 37°C, cell pellets were collected and frozen at Ϫ80°C. After resuspension in 40 ml of binding buffer with 1 mM PMSF, the cells were then disrupted by sonication, and the supernates were collected after centrifugation at 10,000 rpm for 30 min at 4°C (26).
The VEK and VKK peptides were purified using a Ni 2ϩ -Sepharose affinity chromatography column, as above, and concentrated by ultrafiltration. Thrombin (1,000 units; ERL, South Bend, IN) was then added. The cleaved fragments were further separated using a HiTrap HP affinity column (GE Healthcare). The flow-through fractions contained the desired peptides, which were then applied to a p-aminobenzamidine-agarose affinity column (Sigma) to remove thrombin. The integrity of all proteins and peptides was determined by MALDI-TOF mass spectrometry (supplemental Table 1S). For all peptides, single mass peaks were obtained at the correct molecular weights, indicating, for the labeled peptides, that complete incorporation of the heavy isotopes occurred.

SPR
The binding kinetics of full-length PAMs to hPg and VEK/ VKK peptides to K2 hPg were measured in real time by SPR using a BIAcore X100 Biosensor system (GE Healthcare). All binding experiments were conducted at 25°C employing HBS-EP (10 mM HEPES, 0.15 M NaCl, 3 mM EDTA, 0.005% polysorbate 20, pH 7.4) as the running buffer at a flow rate of 10 l/min. Both hPg and K2 hPg , diluted to 40 g/ml in 10 mM NaOAc, pH 4.5, were injected into flow cell 2 for immobilization on the CM5 sensor chip surface using the amine-coupling kit, to a level of ϳ900 response units. Nonbound sites on the sensor chip surface were blocked afterward by injection of 1 M ethanolamine, pH 8.5.
All binding experiments were conducted by injecting various concentrations of analytes in HBS-EP buffer over the hPg or K2 hPg -coupled CM5 chip surface. Each concentration was injected for an association time of 2 min followed by a 6-min dissociation time. The chip surface was regenerated between PAM binds plasminogen via two lysine isosteres cycles using 10 mM glycine, pH 1.5 (for hPg) or pH 2.0 (for K2 hPg ), which did not change the properties of ligands bound to the CM5 chip. The binding data from these sensorgrams were subtracted from those obtained using a reference flow cell prepared by the same method, but without immobilizing ligands on the chip. Sensorgrams were analyzed using BIA evaluation software version 2.0.1 (GE Healthcare). The apparent equilibrium K D values were calculated from the instrument software by the ratio of the dissociation (k off ) and association rates (k on ). Nonlinear fitting of the association and dissociation curves with a 1:1 binding model was employed (12,33).

AUC
AUC experiments were performed in a Beckman XL-I analytical ultracentrifuge (Beckman-Coulter, Fullerton, CA) at 25°C in the absorbance detection mode at 280 nm. Protein samples were diluted to a final concentration of 0.1-0.3 mg/ml in 100 mM sodium phosphate buffer, pH 7.4. A six-sector cell loaded with 110 l of samples and 125 l of reference buffer was used. The apparent molecular weight for the complex of VKK38 -K2 hPg was determined by sedimentation equilibrium at rotation speeds of 32,000 -48,000 rpm. The partial specific volumes of VKK38 and K2 hPg were calculated from their amino acid sequences using Sednterp and were determined to be 0.721 and 0.708 ml/g, respectively. The buffer density was also determined by Sednterp to be 1.0091 g/ml. AUC data were analyzed by the Optima XL-A/XL-I data analysis software (Beckman Coulter, Brea, CA), and the apparent molecular weight values were obtained. Sedimentation analysis software SEDPHAT was used to analyze the data (12).

ITC
ITC measurements were performed at 25°C on a VP-ITC 200 Microcal (Malvern, UK) calorimeter in 50 mM sodium phosphate, 50 mM NaCl, pH 7.4. A solution of K2 hPg (100 M) was stirred in a cell, and 750 M VKK38 was injected at a rate of 1 l/s with a 120-s equilibration time. Blank experiments were performed using buffer without proteins. The integrated heats of interaction were normalized as a function of the molar ratio of VKK38 to K2 hPg , and the data were fit with Origin-ITC version 7.0 software.

NMR spectroscopy
NMR spectra for resonance assignments and NOE identification were recorded at 298 K on a Bruker AVANCE II 800 spectrometer, equipped with a 5-mm triple resonance (TCI, 1 H/ 13 C/ 15 N) cryoprobe. All spectra were conducted on samples of uniformly 15 N-or 13 C/ 15 N-labeled recombinant VKK38 in 20 mM BisTris-d 19 , 2 mM DSS, 0.1% NaN 3 , 5% D 2 O, 95% H 2 O, pH 6.7, except for 2D 1 H/ 13 C HSQC, 3D HCCH-TOCSY (34), and 13 C-edited NOESY experiments, for which the sample was dissolved in 2 H 2 O. NMR data were processed with TopSpin version 3.5 software and analyzed by using Sparky (35). 1 H chemical shifts were referenced to internal DSS. 13 C and 15 N chemical shifts were referenced indirectly to DSS (36). The following spectra were collected: 15 N HSQC (37,38), 15 (34) for the backbone and aliphatic side chain resonance assignments, as well as 15 N NOESY-HSQC (43) and 13 C NOESY-HSQC (44) to collect intramolecular NOE distance constraints for use in the structure calculations.
To solve the structure of the VKK38 -K2 hPg complex, 15 Ndispersed NOESY (200-ms mixing time) and 13 C-dispersed NOESY (80-ms mixing time) spectra were collected on the mixture of [ 15 N/ 13 C]VKK38 and natural abundance K2 hPg at a ratio of 1:2 m/m. Intermolecular distance constraints were determined by 13  Backbone 15 N-relaxation parameters, including longitudinal relaxation rates (R 1 ), transverse relaxation rates (R 2 ), and steady-state heteronuclear 1 H-15 N NOEs, for free and complexed VKK38 were measured at 298 K using standard pulse sequences as described elsewhere (45). All 15 N R 1 /R 2 relaxation experiments were carried out in an interleave manner with a 2-s recycle delay between scans. The relaxation delays used for free and complexed VKK38 were 10, 50, 110 ϫ 2, 180, 300, 420, 570 ϫ 2, 650, 850, 1100, and 1500 ms for the R 1 experiments and 16, 32 ϫ 2, 48, 64, 80.1, 96 ϫ 2, 112, 128, 160.1, 176, and 192 ms for the R 2 experiments. Duplicate spectra were used to estimate experimental errors. The relaxation rates were determined by fitting the cross-peak intensities to a single exponential function using the nonlinear least-squares method. The error in the rate constants was assessed from Monte Carlo simulations. 1 H-15 N NOE experiments were carried out in the absence and presence of a 3-s proton saturation period before the 15 N excitation pulse, using recycle delays of 4 and 7 s. Heteronuclear NOE values were obtained from the ratios of the peak intensities measured with and without proton saturation. Peak intensity uncertainties were estimated from the noise level of the spectra.
NMR titration experiments were carried out by recording 1 H-15 N HSQC experiments on [ 15 N]VKK38 samples with increasing molar ratios of unlabeled K2 hPg . The amide nitrogen and hydrogen chemical shift perturbations were mapped for each amino acid according to ⌬␦ HN,N ϭ 1 ⁄4͌((␦ HN / 2 ) 2 ϩ (␦ N / 5 ) 2 ), where ⌬␦ HN and ⌬␦ N represent the chemical shift changes of 1 H and 15 N atoms between the apo and bound forms, respectively.

NMR structure calculations
Assignments for backbone atoms are 92 and 77% complete for VKK38 in the apo and bound forms, respectively, and Ͼ80% for side-chain resonances. Using Sparky, a total of 120 and 162 proton distance constraints for VKK38 in the apo and bound forms were obtained from analyzing the 15 N and 13 C NOESY spectra. Backbone torsion angles ( and ) were predicted using TALOS-N (46).
All restraint information was applied in a simulated annealing protocol using XPLOR-NIH version 2.36 (47,48). A summary of the experimental constraints as well as pertinent structural statistics is provided in supplemental Table 2S. Approximately 200 structures were calculated, from which 20

PAM binds plasminogen via two lysine isosteres
structures with the lowest restraint energy values were further refined with implicit water. The quality of the structures was analyzed with PROCHECK version 3.5.4 (49,50). For the 20 final conformers of apo-VKK38 and VKK38 bound to K2 hPg , 100 and 94.7% of all residues were found in the favored and allowed regions of the Ramachandran plot, respectively. Visualization of the structures was performed using PyMOL.

Molecular modeling
K2 hPg was modeled to the bound structure of VKK38 at a ratio of 1:2 (m/m). A total of 39 intermolecular NOEs from the NMR solution structure of VEK30/K2 hpg (BMRB entry ID 16311) were used as distance restraints for structure calculations for VKK38 -K2 hPg at residues 17 RH 18 . The lowest energy conformer of the 200 calculated structures was used as the starting template to bind with the second K2 hPg molecule at the 33 RH 34 binding site. This was as carried out using the HADDOCK web server combined with the chemical shift perturbation data from 15 N HSQC titration experiments for VKK38 -K2 hPg and previous results for VEK30 -K2 hPg . The result generated with the lowest HADDOCK score was selected to present the binding model of VKK38 -K2 hPg at a ratio of 1:2.

Data deposition
Backbone shift assignments and the experimental restraints used in the structure calculation for VKK38 in the nonbound and K2 hPg -bound forms have been deposited in the BioMagResBank with the accession numbers 30272 and 30271, respectively. The coordinates of the calculated structure ensembles have been deposited in the Protein Data Bank with the accession code 5V4U for VKK38 in the K2 hPg -bound form.
Author contributions-Y. Y., J. Z., C. Q., and V. C. performed the experiments and edited drafts of the manuscript; F. J. C. and V. A. P. edited drafts of the manuscript; F. J. C. consulted on experiments and prepared the final version of the manuscript.