A local α-helix drives structural evolution of streptococcal M-protein affinity for host human plasminogen.

Plasminogen-binding group A streptococcal M-protein (PAM) is a signature surface virulence factor of specific strains of Group A Streptococcuspyogenes (GAS) and is an important tight binding protein for human plasminogen (hPg).  After activation of PAM-bound hPg to the protease, plasmin (hPm), GAS cells develop invasive surfaces that are critical for their pathogenicity.  PAMs are helical dimers in solution, which are sensitive to temperature changes over a physiological temperature range.  We previously categorized PAMs into three classes (I-III) based on the number and nature of short tandem a-helical repeats (a1 and a2) in their NH2-terminal A-domains that dictate interactions with hPg/hPm.  Class II PAMs are special cases since they only contain the a2-repeat, while Class I and Class III PAMs encompass complete a1a2-repeats.  All dimeric PAMs tightly associate with hPg, regardless of their categories, but monomeric Class II PAMs bind to hPg much weaker than their Class I and Class III monomeric counterparts.  Additionally, since the A-domains of Class II PAMs comprise different residues from other PAMs, the issue emerges as to whether Class II PAMs utilize different amino acid side-chains for interactions with hPg.  Herein, through NMR-refined structural analyses, we elucidate the atomic level hPg-binding mechanisms adopted by two representative Class II PAMs.  Further, we develop an evolutionary model that explains from unique structural perspectives why PAMs develop variable A-domains with regard to hPg-binding affinity.

The [ 13 C, 15 N]-filtered/edited 3D NOESY-[ 1 H, 15 N]-HSQC spectra were recorded under the same conditions [34]. Given that both unlabeled K2hPg-bound [ 13 C, 15 N]-AGL55NS88.2 and K2hPg-bound [ 13 C, 15 N]-KTI55SS1448 displayed some overlapping peaks, thus causing uncertainties for the assignments of cross peaks in the NOESY spectra, intermolecular NOEs collected from these two samples were not included as distance restraints in the docking.

High ambiguity-driven protein-protein docking (HADDOCK)
The structures of AGL55NS88.2/K2hPg and KTI55SS1448/K2hPg were calculated through the HADDOCK platform on the basis of multiple restraints [43]. In each complex, the residue numbers in the linear sequence of K2hPg, AGL55NS88.2, and KTI55SS1448 are designated as 166-243, 95-149, and 85-139, respectively ( Table 1). Two short terminal fragments in recombinant K2hPg, i.e., the exogenous residues in Y -7 VEFSEE -1 and the A 1001 A 1002 , are not derived from the linear sequence of K2hPg, but from the cloning steps. Likewise, the NH2-terminal G -2 S -1 dipeptide in both AGL55NS88.2 and KTI55SS1448 originates from the expression plasmid after the cleavage by thrombin. All of these exogenous sequences are not counted in the numbering of the corresponding peptide.
Downloaded from https://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20200197/871532/bcj-2020-0197.pdf by guest on 13 April 2020 The structure of K2hPg applied here is derived from the complex structure of VEK75AP53/K2hPg [20]. energy, the smallest RMSD, and accordingly the most negative HADDOCK score, were selected [43]. These structures were examined through MolProbity in terms of all-atom contacts and protein geometry [41,42]. For each complex, 20 of the best structures were overlaid on the K2hPg moiety for the subsequent deposition. Detailed statistics with respect to structural calculations are listed in Table S1.

Surface plasmon resonance (SPR)
Binding kinetics of recombinant PAM-derived peptides to K2hPg were measured in real-time in a BIAcore X100 Biosensor system (GE Healthcare   Figure S1.  (Figure 1A, Figure S3). Three residues at the COOH-terminal end of the upstream hypervariable region (HVR) of PAMNS88.2, specifically, E 105 , L 107 , and K 108 , also demonstrate relatively large CSPs ( Figure 1A). Considering that the a-repeat is responsible for binding to K2hPg, it is expected that residues in this binding region are influenced most when they interact with K2hPg.

Chemical shift changes in the [ 1 H, 15 N]-HSQC spectra indicate the active binding region
A similar phenomenon was observed in the correlation between the KTI55SS1448 sequence ( Table   1, Figure S2) and the CSP of each residue.  Figure S5A).

The secondary structures of AGL55NS88.2 and KTI55SS1448 are altered upon binding to K2hPg
Based on the chemical shift assignments for each peptide, we next applied TALOS-N [37] to calculate dihedral angles and to predict the secondary structures of each peptide. The results show a lower probability of α-helix in the RH-motif of apo-AGL55NS88.2 (R 119 H 120 ; Figure 2A) and apo-KTI55SS1448 (R 110 H 111 ; Figure 2B) than in the residues flanking this motif. These findings support the previously reported mobile feature of this motif in the absence of binding to K2hPg [17]. Notably, a long fragment within KTI55SS1448, K 107 NERHDHD 114 , is not α-helical

AGL55NS88.2 utilizes multiple sites in the A-domain to bind to K2hPg
A total of 13 and 9 intermolecular NOEs were observed for AGL55NS88.2/K2hPg and KTI55SS1448/K2hPg complexes, respectively ( We next applied the high ambiguity-driven protein-protein docking (HADDOCK) program to conduct modeling of complex structures, mainly based on the intermolecular NOE and RDC restraints. We further used both HADDOCK scores and RDC restraints to evaluate the quality of the calculated complex structures ( Figure S6). Given that a loop consisting of L 115 N 116 E 117 disrupts the α-helix in the A-domain, the overall structure of AGL55NS88.2 in complex with K2hPg appears in a "V"-shape ( Figure 3A). Three sets of interactions were found between AGL55NS88.2 and K2hPg, among which two are related to the D 219 R 220 D 221 -motif in the LBS of K2hPg. First, the anionic carboxyl group of E 111 in AGL55NS88.2 inserts into the cationic center formed by K 208 and hydrogen bonds in addition to electrostatic attractions ( Figure 3B). However, partially attributable to movement of the long side chains of K 208 and R 220 in K2hPg, this proximity is variable in different low energy complex structures, even rising to ~4.0 Å, which disfavors formation of a H-bond between E 111 in AGL55NS88.2 and K 208 of K2hPg as illustrated in another low energy structure ( Figure 3C). ( Figure 3D).

The second interaction occurs
It was previously reported that the side chain of R 234 in K2hPg interacts with the negatively charged side chains of E 112 and E 116 in the various VEK-peptides from PAMAP53, through electrostatic attractions and/or hydrogen bonds [13,20]. But such interactions disappear in the case of AGL55NS88.2, which is ascribed to the DHD tripeptide following the RH-motif ( Figure 3E).
Unlike PAMAP53 (Class I) that features an E 112 RHEE (Table 1, Figure S2) sequence at the end of the a1-repeat, PAMNS88.2 is naturally devoid of the a1-repeat, and thus only contains the a2-repeat that interacts with hPg. Therefore, this sequence, E 112 RHEE, is absent from PAMNS88.2 and its derived peptide, AGL55NS88.2. On the other hand, the other highly conserved sequence, ERHDHD, which exists in the a2-repeat of different PAMs, is also present in AGL55NS88.2. As a consequence, by means of steric hindrance, the imidazole ring of H 122 in this DHD tripeptide maintains the cationic side chain of R 234 in K2hPg distal from the anionic E 118 and D 121 in AGL55NS88.2 ( Figure 3E). Therefore, the positively charged head group of R 234 in K2hPg is not

The distance between the RH-motif and the LBS of K2hPg dictates the binding affinity
It has been shown that the hPg-binding of AGL55NS88.2 is >1,000 times weaker than that of VEK75AP53 [18], mainly due to the rapid dissociation rate of the AGL55NS88.2/hPg complex.
Likewise, AGL55NS88.2 binds to K2hPg with a ~300-fold lower affinity than VEK75AP53 ( Figure   4A-B; Table 3). Thus, the reasons that AGL55NS88.2 displays a high dissociation rate in the SPRbased measurements of its K2hPg or hPg binding affinity need to be assessed. When the K2hPg moieties from K2hPg/AGL55NS88.2 and K2hPg/VEK75AP53 complexes are overlaid, it can be seen that the RH-motif in AGL55NS88.2 is ~2.5Å farther from the active site of K2hPg than the counterpart residues in VEK75AP53 ( Figure 5A). This suggests that the RH-motif in AGL55NS88.2 would readily dissociate, thus accounting for the high off-rate found in the K2hPg-binding assays ( Figure   4B). contains the a2-repeat, are critical to its interaction with K2hPg, even though this peptide exhibits a more flexible overall structure ( Figure 6A). The anionic carboxyl groups of E 102 and E 104 , two residues at the beginning of the a2-repeat in KTI55SS1448, capture the guanidinium side chain of R 220 in K2hPg. Instead of a stable attraction, the distance between the charged heads of R 220 and E 102 /E 104 is variable, and could be as distant as ~4.0 Å (Figure 6D), or as close as 1.8 Å (Figure 6E), in different low energy structures. Only the latter proximity drives strong hydrogen bonds. This elastic feature also applies to the interaction of D 114 in KTI55SS1448 and R 234 in K2hPg (Figure 6F-G). Such highly variable proximities indicate that these two binding sites in KTI55SS1448 can associate with K2hPg, and easily dissociate, thereby explaining the high off-rate observed in the K2hPg-binding affinity determination of KTI55SS1448 ( Figure 4C, Table 3). hitherto been determined, either by NMR or X-ray crystallography ( Figure 8A)  Figure 8B). Nevertheless, this initial PAM probably exhibited a limited α-helical fraction in its A-domain, even lower than that of KTI55SS1448, which was insufficient for tight binding to hPg.

The most significant interaction between KTI55SS1448 and
PAM-positive strains subsequently adopt two major strategies to enhance the binding affinity of PAM to hPg, including optimization of the A-domain sequence and recombination with another arepeat, e.g., Class I and Class III PAMs (Figure 8B). Both strategies serve to increase the αhelical content in the A-domain. However, it is still not known why PAMGLS307 contains five a-Downloaded from https://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20200197/871532/bcj-2020-0197.pdf by guest on 13 April 2020 repeats in the A-domain (Table S2), since two a-repeats are sufficient for PAMs to achieve nMrange binding to hPg. One plausible explanation is that these five a-repeats sufficiently separate the A-domain and the HVR, making it possible for PAMGLS307 to capture two different host proteins through these two regions at the same time.

The evolutionary importance of the His residue in the PAM RH-motif
Concerning the RH-motif, the role of Arg in the interaction between PAM and hPg has been Class II PAM-derived peptides that contain only the a2-repeat. In addition, it is also biologically unique and relevant that the α-helical content in the A-domain determines the evolving state of PAMs, which is conducive for binding to hPg, a factor that enhances its virulence. More specifically, for a PAM that exhibits limited α-helices in the A-domain, we provide proof that His of the RH-motif, rather than Arg, inserts into the LBS of K2hPg. Special natural mutations and recombination events are present in some PAMs as well, e.g., substitution of RH to RY and appearance of five a-repeats, but why these PAMs contain such unusual changes needs further study.

Competing Interests
The authors declare that there are no competing interests associated with the manuscript.

Funding
The work was supported by a multi-PI grant (HL013423) to FJC, SWL, and VAP.

Data Availability
Backbone chemical shift assignments and experimental restraints used in the structure calculations             Phylogenetic tree of the SK β-domain This unrooted tree was constructed using the Neighbor-Joining method [57] in MEGA7 [58].