All simulations were performed with the GPU accelerated CHARMM/OpenMM interface (Brooks et al., 2009; Lee et al., 2016; Eastman et al., 2017) in CHARMM36m force field (Huang et al., 2017), which is one of the latest general-purpose protein force field specifically optimized for both IDPs and structured proteins. The initial structures of SPIN-aureus and SPIN-delphini in complex with human MPO were taken from the crystal structures [PDB 5UZU SPIN-aureus (de Jong et al., 2018) and 6BMT for SPIN-delphini (Ploscariu et al., 2018)] (see Figure 1A). To reduce the computational cost, only segments of MPO that are within 12 Å of SPIN are included in the current simulations, which consist of residues 167–200, 255–444, 490–506, 526–540, and 566–596 for MPO (Supplemetary Figure S1). To prevent the unfolding of MPO, all backbone heavy atoms of structured MPO segments (excluding loop residues 268–288, 380–395, and 317–328) and the bound heme group were restrained by harmonic potentials with a force constant of 1.0 kcal/(mol Ų) in all simulations. Proper amount of Na⁺ and Cl⁻ ions were added to neutralize the systems and to reach a NaCl concentration of 50 mM in accord with the experimental conditions (Ploscariu et al., 2018). The final solvated box contains about ∼30,000 TIP3P water molecules and has a dimension of ∼9.2nm³ × 9.4nm³ × 11.3 nm³.

Each solvated system was first energetically minimized for 500 steps using steepest decent and another 500 steps using the adopted basis Newton-Raphson algorithm. The system was then slowly heated up from 100 K to 300 K in 10 ps under the constant volume condition. Equilibration simulations were then performed at 300 K and 1 atm for a total of 1 ns, during which all protein heavy atoms were additionally restrained using harmonic potentials with force constants slowly decreasing from 5.0 kcal/(mol Ų) to 0.0 kcal/(mol Ų). Langevin thermostat was used to control the temperature and Monte Carlo barostat with volume move attempt every 25 steps was used to control the pressure. Lengths of all bonds involving hydrogen atoms were constrained using the SHAKE algorithm (Ryckaert et al., 1977) to allow for an integration time step of 2 fs. Long-range electrostatic interactions were treated using the particle mesh Ewald method (Darden et al., 1993), and the short-range van der Waals (vdW) interactions were treated with the twin-range cutoff at 12 and 14 Å.

To identify the optimal temperatures for unbinding/unfolding simulations, a series of pilot simulations were performed at temperatures ranging from 400 K to 500 K at 1 atm. Note that the complex was found to be highly stable below 400 K. Once an optimal temperature was identified (450 K), two sets of simulations were performed for each complex to probe temperature-induced SPIN unfolding and unbinding process. In one set, three additional simulations were performed at 450 K with different initial velocities to better characterize the dissociation of SPIN from MPO. These simulations were run until the NTD dissociated from the active pocket (i.e., with the fraction of native contacts between two molecules Q _inter < 0.3), which all occurred within 400 ns. In the second set, 40 independent replicas were performed for each complex at 450 K for 250 ns each, with the helical region of SPIN CTD (Figure 1B) harmonically restrained with a force constant of 1.0 kcal/(mol Ų). The purpose of the second set is to directly examine the unfolding and unbinding of the NTDs.

The initial folded hairpin structures of SPIN-NTDs were taken from the same complex structures (Figure 1A). Both systems contain a 13-residue fragment (SPIN-aureus residues 33–45 and SPIN-delphini residues 28–40; see Figure 1B for sequences). 20 replicas were used to simulate the unfolding events for two SPIN-NTDs at 300 K. The solvated systems contain ∼3,500 TIP3P waters and have dimensions of ∼4.2nm³ × 4.3nm³ × 5.4 nm³. Similar protocols as described above were applied to minimize and equilibrate the system. For each system, 20 independent production simulations were performed for 50 ns each at 300 K, which was sufficient to observe spontaneous unfolding of the β-hairpin structure.

All the analyses were carried out using CHARMM and additional in-house scripts. All molecular visualizations were prepared using VMD (Humphrey et al., 1996). The fractions of intermolecular and intramolecular native contacts, Q _inter and Q _intra, are calculated to monitor the unfolding and unbinding process. The native contacts are first identified from the crystal structure of two complexes if the minimum heavy atom distance between two residues is no greater than 4.2 Å (Supplemetary Tables S1, S2). Note that for intramolecular native contacts, we exclude residue pairs that are close in sequence space and only consider those whose residue IDs are different by at least three. The contacts in simulation trajectories were then calculated using the same criterion. Based on protein folding funnel theory, native interactions dominate the overall pathway (Onuchic et al., 1996; Tsai et al., 1999; Shoemaker et al., 2000; Best et al., 2013). Therefore, only native contacts were considered here. The unbinding and unfolding kinetics were analyzed using a double exponential approximation of the decay of Q _inter and Q _intra averaged over all replica runs (40 for the complexes and 20 for free NTDs). The first 50 ns trajectories were considered in unfolding and unbinding kinetic analysis, which were sufficient to capture the dissociation events. Pseudo free energy surfaces were also calculated to better characterize the baseline mechanisms of coupled binding and folding, derived directly from two-dimensional (2D) probability distributions along Q _inter and Q _intra. For the data used to construct contact probabilities, we specifically focused on short segments of the trajectories where actual dissociation transitions occurred. For example, only the first 15 ns trajectory in replica one of SPIN-aureus/MPO simulation was considered, which included the entire unbinding and unfolding transition (see Supplemetary Figure S2. By doing this, the results will not be interfered by the transient refolding events observed after complete dissociation (see Supplemetary Figure S2 replica 40 at 200 ns for example). The segments for each trajectory that were selected to calculate the contact maps can be found in Supplemetary Table S3. Note that for replicas where NTD remains bound and folded at the end of the 250 ns-simulation, we only selected the first 50 ns of trajectories to compute contact maps. In this way, we could avoid masking important details about the transition pathways by over-representing data of the bound and folded state.

High temperature simulations have been shown to be capable of providing reliable mechanistic insights in to folding of structured proteins as well as coupled binding and folding of IDPs (Verkhivker et al., 2003; Daggett, 2006; Chen and Luo, 2007; Schaeffer et al., 2008; Zhang et al., 2012). The assumption here is that unfolding and unbinding is largely a reverse of coupled binding and folding. While many mechanistic details derived from high-temperature simulations have compared well to experiments (Verkhivker et al., 2003; Daggett, 2006; Chen and Luo, 2007; Schaeffer et al., 2008; Zhang et al., 2012), it is also known that the most probable transition pathways may depend on the temperature (Dinner and Karplus, 1999). Therefore, it is important to find the lowest temperature to drive the unfolding and unbinding process within a given simulation timeframe. The pilot simulations suggest that the NTD of SPIN-aureus only starts to dissociate from the active pocket of MPO at 450 K within ∼100 ns timescale, which becomes much faster at higher temperatures (Supplemetary Figure S1B). Note that rapid dissociation (e.g., at 475 K) is not always preferred due to the risks of missing important details under non-physiological conditions and activating pathways not generally accessible under the physiological conditions. For example, the three-helix bundle of SPIN CTD would melt rapidly at 475 K and above, leading to premature disassociation from MPO within 10 s of ns. This is consistent with the experimental observation that SPIN CTD largely dictates MPO binding (de Jong et al., 2018). Instead, simulations at 450 K seem to depict a more realistic dissociation process, where NTD unbinds first while the CTD remains largely fold and bound (Figure 2). The apparent decoupling and step-wise nature of the binding of SPIN CTD and NTD could explain why there is little correlation between the inhibition strength and binding affinity for different SPIN homologs. It’s likely that two domains of SPIN bind and function almost independently when interacting with MPO. As such, some SPIN homologs, e.g., SPIN-agnetis, show comparable nanomolar binding affinity as SPIN-aureus, but have no detectable inhibitory ability to human MPO (Ploscariu et al., 2018). Based on these observations, we will focus on the coupled binding and folding of SPIN NTD while the CTD is harmonically restrained to the bound state in subsequent simulation and analysis.

A total of 40 independent 250-ns simulations were performed at 450 K to explore the conformational fluctuations, dynamic interactions and dissociation pathways of SPIN-aureus NTD with human MPO. As summarized in Supplemetary Figure S2, SPIN-aureus NTD tends to dissociate rapidly and its unfolding and unbinding often happen simultaneously. For example, in 36 out of 40 replicas (except for replicas 6, 8, 16 and 25), NTD fully dissociated (with Q _inter < 0.2) within 200 ns. Particularly, among 30 out of the 36 runs (except for replicas 10, 11, 17, 23, 29 and 35) unbind/unfold occurred within the first 50 ns, or sometimes even more rapidly within 15 ns. To quantitively describe the dissociation process and probe the mechanisms of coupled binding and folding, we calculated the average fractions of intermolecular and intramolecular native contacts formed by NTD, denoted Q _inter and Q _intra, respectively, from all replicas. The results were then fitted with a double exponential function (Figure 3A). Not surprisingly, the unbinding and unfolding kinetics of SPIN-aureus NTD are similar, consistent with the observation that they appear highly correlated. As shown in Figure 3A, the initial fast phase τ 1 for unbinding and unfolding are 0.12 and 0.28 ns, respectively, followed by a slow phase unbinding ( τ 2 of 11.40 ns) and unfolding ( τ 2 of 13.45 ns). We further constructed the pseudo 2D free energy surface as a function of NTD Q _inter and Q _intra, derived from the dissociation transition segments (see Methods for details). The result, shown in Figure 3B, confirms a highly cooperative mechanism of SPIN-aureus NTD coupled binding and folding with NTD Q _inter and Q _intra increasing simultaneously in a highly correlated fashion. The minimum free energy path (dashed line) largely follows the diagonal line expected for an ideally cooperative mechanism.

The free energy surface also reveals three major conformational states of NTD folding and binding to MPO. State B (bound), with both Q _inter and Q _intra above 0.8, is the fully folded and bound state, and State U (unbound), with both Q _inter and Q _intra below 0.2, is the fully unfolded and unbound state. In addition, there is a partially bound and folded substate, I (intermediate state), where the values of Q _inter and Q _intra are around 0.4. Representative conformations for the three states of the complex (Figure 3C) illustrate that SPIN-aureus NTD does not tend to pre-fold into some “native-like” β-hairpins conformations prior to binding to the active site of MPO, and vice versa. The cooperative nature of SPIN-aureus NTD is in contrast to previous experimental and computational studies of coupled binding and folding of IDPs into non-β-hairpin structures (Chen, 2012), where induced folding-like mechanisms are prevalent. However, this may not be surprising given the cooperative nature of folding of isolated β-hairpins (Munoz et al., 1997; Klimov and Thirumalai, 2000; Chen et al., 2018). In particular, the “speed-limit” of β-hairpin folding usually is ∼μs⁻¹, much slower compared to helix-coil transitions (∼100 ns), due to the requirement of forming long-range interactions and the presence of entropy-dominant barriers. Therefore, once SPIN-aureus CTD is tightly bound, native-like interactions with the MPO surface play a direct role to facilitate the rapid folding of NTD and achieve a facile blockage of the MPO active site for inhibition.

Compared to SPIN-aureus, which is secreted by S. aureus that appears to be particularly adapted to survive the neutrophil-mediated immunity with the highest binding affinity (K_D = 15.9 nM) and inhibition strength (IC₅₀ = 4.6 nM) to human MPO, SPIN-delphini has a moderate binding affinity (K_D = 310 nM) but the second strong inhibitory ability (IC₅₀ = 29.7 nM) among nine SPIN homologs previously analyzed (Ploscariu et al., 2018). A possible explanation is that SPIN-delphini NTD may have evolved to be less dependent on the tight binding of CTD. Interestingly, high-temperature simulations indeed reveal significant differences between coupled binding and folding of NTDs from SPIN-aureus and SPIN-delphini. As summarized in Supplemetary Figure S3, ∼40% of the 40 (17/40) replicas failed to observe full dissociation of SPIN-delphini NTD’s during the 250 ns simulations, which is about 3-fold of ∼10% for SPIN-aureus. The implication is that SPIN-delphini NTD fits the active site of MPO tighter than SPIN-aureus NTD, which would be consistent with disproportionally strong inhibitory function of SPIN-delphini despite weakened overall binding affinity.

Further analysis of unfolding and unbinding kinetics and free energy surface reveal that SPIN-delphini NTD mainly follow a distinct conformational selection-like mechanism (Figure 4), where the NTD tends to gain substantial native β-hairpin structures prior to forming intermolecular interactions with MPO. This is well reflected in unbinding and unfolding kinetics. On average, the unbinding rates of SPIN-delphini NTD (( τ 1 = 0.06 ns, τ 2 = 18.4 ns, Figure 4A) are similar to those of SPIN-aureus NTD ( τ 1 = 0.12 ns, τ 2 = 11.4 ns, Figure 3A). However, the unfolding rates of SPIN-delphini NTD ( τ 1 = 1.31 ns, τ 2 = 196.97 ns) are over 10-fold slower than unbinding rates. In addition, SPIN-delphini NTD is considerably more folded at 15 ns, with Q _intra ∼ 0.6 compared to ∼0.3 for SPIN-aureus NTD. That is, while SPIN-aureus NTD unbinds and unfolds to similar levels at a given time (Figure 3A), SPIN-delphini NTD tends to retain much higher residual structures while it unbinds.

The minimal free energy paths, indicated by the dash lines in Figure 4B, demonstrate that SPIN-delphini NTD coupled binding and folding is not cooperative and follows two major routes with multiple intermediate states. Both routes go through an intermediate state I1, where there Q _inter drops below 0.7 while the β-hairpin structure is essentially intact with Q _intra ∼ 1.0. Overlay of representative structures from B (fully bound) and I1, shown in Supplemetary Figure S4, illustrates that how SPIN-delphini NTD becomes mobile within the active site pocket of MPO without unfolding. From state I1, the major pathway (path A) goes through another intermediate state I2a, which mainly has similar level of residual intermolecular native contacts (Q _inter ∼ 0.5) but the hairpin conformation becomes partially unfolded (Q _intra ∼ 0.5). From state I2a, SPIN-delphini NTD would further unbind and then unfold to reach the fully disassociated state (U). In the parallel pathway B, SPIN-delphini NTD would continue to become fully unbound from MPO without significant unfolding (I2b, Q _intra > 0.8, Q _inter < 0.2), before unfold outside of the MPO active site. The observed conformational selection-like mechanism of SPIN-delphini NTD interaction with MPO is summarized in Figure 4C. It shows that the disordered segment could become fully folded before inserting into the MPO active site (Path B), which is an ideally conformational selection mechanism. Such a process is best represented by high-temperature simulation run 24 (Supplemetary Figure S3, Rep24). Path A, which is more prevalent, involves multi-step conformational selections. In each step (U to I2a to B), the NTD first fold and then bind to MPO. The later pathway is best illustrated in Supplemetary Figure S3 Re33. The distinct mechanisms of SPIN-aureus and SPIN-delphini NTD coupled binding and folding may help explain why the inhibition strength doesn’t fully correlate with binding affinity among different SPIN homologs.

For conformational selection to be an efficient mechanism for coupled binding and folding, there should be high levels residual structures in unbound IDPs (Liu et al., 2019). Since SPIN-aureus NTD follows cooperative binding and folding while SPIN-delphini NTD prefers a conformational selection-like mechanism, we further characterized the stability of hairpin-like structures in their unbound states under the physiological conditions. As shown in Figure 5, although SPIN-delphini NTD showed slightly faster unfolding rates, it remained more structured than SPIN-aureus NTD. The limiting NTD Q _intra decayed to 0.57 and 0.40 for SPIN-delphini and SPIN-aureus, respectively (see Figure 5A). Importantly, the probability distributions of Q _intra show that there is a very high probability for SPIN-delphini NTD to remain partially folded (Q _intra > 0.5). Such an elevated residual β-hairpin structures in unbound SPIN-delphini NTD is consistent with the observation of conformational selection-like mechanism of its coupled binding and folding (see above). The more dynamic nature of SPIN-aureus NTD suggests that it depends on specific MPO interactions to facilitate its folding into the β-hairpin structure, thus following a cooperative binding and folding mechanism (Figure 3B).