N-terminal processing unlocks global dynamics for substrate engagement in Spl proteases

Abstract

Staphylococcus aureus secretes a family of serine protease–like enzymes (SplA to SplF) that resemble eukaryotic granzymes, yet the mechanism by which amino-terminal processing activates this subclass has remained unresolved. Structural studies show insertion of the processed amino terminus without detectable changes in active-site geometry, creating a longstanding paradox as to how catalytic competence is achieved. Here, we identify SplB as the most highly expressed member of this family in a pathogenic methicillin-resistant S. aureus strain and use it to define the basis of activation. Solution nuclear magnetic resonance spectroscopy shows that precise amino-terminal processing triggers a long-range allosteric network coupling the amino terminus to the active site ∼20 angstroms away, unlocking global microsecond-to-millisecond dynamics that enable substrate engagement. Molecular dynamics simulations reveal the conformational ensembles underlying these motions. Last, mutational perturbation of this dynamic network modulates substrate engagement and catalytic activity in a manner consistent with dynamic control of binding competence. Together, these findings establish dynamic allostery as the mechanism of N-terminal activation in this subclass of serine proteases.

N-terminal processing expands conformational dynamics to enable substrate engagement in Spl proteases.

INTRODUCTION

Serine proteases comprise one of the largest enzyme families in biology (1), functioning in processes ranging from microbial pathogenesis to host immunity (2). Many are produced as inactive precursors that require proteolytic maturation to achieve catalytic competence, making the mechanisms by which proteolysis activates these enzymes central to their biological regulation. In many cases, activation involves removal of propeptides that either directly occlude the active site or induce structurally observable conformational changes that facilitate substrate binding. For example, in thrombin, residues 215 to 217 undergo a localized conformational rearrangement that promotes substrate engagement (3).

In contrast, an important subclass, which includes mammalian granzymes and bacterial serine protease–like (Spl) enzymes, requires precise N-terminal processing by extracellular proteases despite lacking steric blockage of the catalytic cleft (Fig. 1A) (4, 5). Structural comparisons of these enzymes reveal a conserved fold with minimal changes in the catalytic triad upon activation (Fig. 1, B and C), offering little explanation for how distal N-terminal processing enables catalytic competence. For example, crystallographic comparisons of SplB with and without an inactivating N-terminal overhang show no repositioning of catalytic residues (Fig. 1D) (4). Instead, activation correlates with insertion of the processed N terminus and formation of a hydrogen-bonding network involving Glu1 and Thr135 (Fig. 1E), raising the central question of how this structural rearrangement confers catalytic competence.

Fig. 1. Paradox of Spl protease activation.

(A) Canonical zymogen activation mechanism of serine proteases. (B) Representative Spl family structures. Catalytically active proteases: SplA (red; 2W7S), SplB (white; 2VID), and SplE (green; 5MM8). Inactive proteases: SplC (blue; 2AS9), SplD (cyan; 4INK), and SplF (yellow; 6SF7). (C) Representative catalytic triad illustrated with the active site of SplB (H39-D77-S157). (D) X-ray crystal structures of SplB with (magenta; 4K1S) and without (white; 2VID) an N-terminal overhang show no difference in the catalytic triad. (E) Precise N-terminal processing enables formation of a hydrogen bond between E1 and T135 (dashed line), locking the N terminus in place and establishing the hydrogen-bond relay that activates the catalytic triad.

Among the Spl family, SplB has emerged as a particularly relevant virulence factor. The Spl protease cluster contributes to S. aureus pathogenesis (6), and recent work demonstrates that SplB modulates host immunity and can be neutralized in vivo by monoclonal antibodies (7). To assess relative expression of Spl family members, we performed quantitative proteomic analysis of representative methicillin-resistant S. aureus (MRSA) and methicillin-sensitive S. aureus (MSSA) clinical isolates. While no Spl family member was detected in the MSSA strain under these conditions, several Spl proteins were highly abundant in the MRSA isolate, with SplB emerging as the most highly expressed family member (fig. S1, Supplementary Text, and data S1 and S2). These findings position SplB as the dominant Spl protease produced by this pathogenic strain. Despite this biological importance, the molecular mechanism by which N-terminal processing activates SplB and the entire serine protease subfamily remains unknown.

To resolve this long-standing mechanistic question, we directly compared mature SplB to a zymogen mimic retaining four N-terminal residues (SplB-4). Nuclear magnetic resonance (NMR) solution studies revealed that activation unlocks global dynamics that facilitates substrate engagement, thereby identifying the underlying dynamic basis of activation. Specifically, a structural “switch” that couples the N terminus to the active site simultaneously unlocks global microsecond-to-millisecond motions and substrate binding. While molecular dynamics (MD) simulations revealed the structural basis of this allosterically coupled network, mutations within this network rewired both movements and function, establishing the mechanistic link between N-terminal processing, conformational dynamics, and catalytic activity. Together, these findings resolve the longstanding structural paradox of Spl activation by demonstrating that N-terminal maturation activates this subclass of serine proteases through dynamic allostery rather than static structural rearrangement. We show that N-terminal activation unlocks global motions on the microsecond-to-millisecond timescale to facilitate substrate binding and that this dynamic network acts to both suppress global dynamics and substrate engagement.

RESULTS

SplB activation requires precise N-terminal processing for substrate engagement

To define how N-terminal processing activates SplB, we compared the mature enzyme, beginning with its N-terminal glutamate (E1), to a zymogen mimic (SplB-4) that retains four residues (TAKA) from the 36-residue propeptide. Even minimal N-terminal overhangs have previously been shown to abolish catalytic activity in Spl proteases, including retention of a Gly-Ser dipeptide following tag removal (4, 8). To precisely control the N-terminal sequence and eliminate unintended residual residues from affinity tags, we used a SUMO fusion strategy that leaves no overhang upon cleavage (9). Thus, mature SplB contains no N-terminal overhang, whereas SplB-4 serves as a mechanistic surrogate for incomplete processing, allowing us to directly test how even a short residual extension affects activation.

As expected, only mature SplB was competent in cleaving a recombinant SUMO-WELQ-GB1 substrate (Fig. 2A), which incorporates the preferred WELQ cleavage motif previously defined for SplB (10). Retention of even four N-terminal residues from the propeptide is therefore sufficient to prevent functional activation.

Fig. 2. Catalysis mature SplB versus SplB-4 and the N-terminal allosteric switch.

(A) Cleavage of a recombinant SUMO-WELQ-GB1 substrate is observed for mature SplB but absent for SplB-4. Cartoon illustrates substrate cleavage by SplB (scissors). (B) Selected regions of ¹⁵N–HSQC (heteronuclear single-quantum coherence) titrations with WELQ peptide for mature SplB (left) and SplB-4 (right). Protein concentrations were 0.5 mM, and titrations were performed at 0, 0.25, 0.5, 1, 2, and 4 mM WELQ. A dissociation constant of K_d = 950 ± 80 μM (fig. S2C) was obtained for mature SplB (fig. S2C). For clarity, only the initial (black) and final (red) spectra are shown for SplB-4. (C) Amide CSPs upon addition of 4 mM WELQ peptide for mature SplB (top) and SplB-4 (bottom). Dashed lines indicate the threshold defined as the mean CSP plus an SD [SplB: 0.38 parts per million (ppm); SplB-4: 0.14 ppm]. (D) Mapping of amide CSPs greater than 0.38 ppm for mature SplB (0 mM versus 4 mM WELQ) onto the x-ray structure [Protein Data Bank (PDB): 2VID]. Residues with CSPs > 0.38 ppm are highlighted. (E) Cα CSPs between mature SplB and SplB-4. The upper dashed line indicates the threshold defined as the mean plus one SD (magenta arrow at 0.43 ppm), and the lower dashed line indicates the mean plus half an SD (violet arrow at 0.28 ppm). (F) Mapping of Cα CSPs onto the SplB structure (PDB: 2VID). Residues with CSPs greater than the mean plus one SD (magenta) and between half and one SD (violet) are highlighted. Titration data were collected at 900 MHz and three-dimensional (3D) assignment data at 600 MHz. All data were collected at 25°C.

We next sought to identify the step in the catalytic cycle disrupted by incomplete processing. NMR titrations revealed clear, saturable chemical shift perturbations (CSPs) upon addition of WELQ peptide to mature SplB, whereas SplB-4 exhibited minimal perturbations even at high peptide concentrations (Fig. 2, B and C, and fig. S2A). Thus, incomplete N-terminal processing blocks substrate engagement itself, indicating that activation governs access to a binding-competent state rather than the chemistry of catalysis.

WELQ binding to mature SplB induces widespread CSPs spanning both the active site and distal N-terminal regions (fig. S2, A and B), indicating long-range allosteric coupling across the enzyme. Although CSPs enable determination of binding affinity [dissociation constant (K_d) = 950 ± 80 μM; fig. S2A], many residues within the catalytic region exhibit pronounced line-broadening upon peptide addition (fig. S2B). Notably, catalytic residues H39 and S157 lose intensity or disappear during titration, indicating increased conformational exchange within the catalytic machinery. While CSPs for these residues cannot be quantified because of exchange broadening and overlap of D77, their behavior reveals inherent microsecond-to-millisecond dynamics that become more pronounced upon substrate engagement.

To define the allosteric network underlying activation, we compared backbone resonances of SplB-4 and mature SplB. In contrast to x-ray crystal structures, which show no detectable differences between these forms (4), widespread changes are observed in solution, including residues within the catalytic region as reported by ¹⁵N–HSQC (heteronuclear single-quantum coherence) spectra (fig. S2, C and D). Extensive Cα CSPs, which are sensitive to backbone conformation, define an allosteric network connecting the N terminus to the active site (Fig. 2, E and F), and amide CSPs further delineate this network (fig. S2, F and G). These data indicate that, while the ground-state structures appear similar, the conformational ensembles sampled by SplB-4 and mature SplB in solution differ substantially.

The catalytic triad itself exhibits only small Cα CSPs between SplB-4 and mature SplB (fig. S2H), indicating minimal changes in ground-state structure. In contrast, catalytic residues display pronounced line-broadening upon activation (fig. S2D), consistent with enhanced microsecond-to-millisecond conformational exchange. As shown below, these effects arise from changes in conformational dynamics rather than structural rearrangement. Together, these findings demonstrate that N-terminal processing activates a long-range allosteric network that enables substrate engagement.

N-terminal processing unlocks global microsecond-to-millisecond dynamics

If N-terminal processing enables substrate engagement without altering the static ground-state structure, we next asked whether activation instead modulates conformational dynamics. To test this, we compared ¹⁵N R₂-CPMG dispersion profiles for mature SplB and the zymogen mimic SplB-4. R₂-CPMG experiments detect conformational exchange processes occurring on the microsecond-to-millisecond timescale, reporting on transitions between distinct chemical environments within this kinetic regime. When motions occur within this window, the effective transverse relaxation rate (R_2,eff) becomes dependent on the applied CPMG field (v_CPMG).

Notably, while SplB-4 exhibits limited R₂-CPMG dispersion at a small subset of residues, activation to mature SplB is accompanied by a substantial expansion of microsecond-to-millisecond conformational dynamics, consistent with the unlocking of a global dynamic network (Fig. 3, A and B, and fig. S3). Residues that exhibit increased exchange upon activation include catalytic triad residues (H39 and S157), positions within the allosteric network identified by CSPs (K6 and Y121), and distal sites throughout the fold (N55 and K58), each of which displays pronounced dispersion in the mature enzyme relative to SplB-4 (Fig. 3A and fig. S3A). In contrast, a subset of residues exhibits comparable dispersion profiles in both SplB-4 and mature SplB (Fig. 3B and fig. S3B), indicating that microsecond-to-millisecond conformational exchange is present before activation but becomes more broadly distributed upon N-terminal processing. Mapping residues that exhibit dispersion onto the x-ray structure reveals a spatial segregation of these behaviors (Fig. 3C). Specifically, residues exhibiting similar exchange in both states localize to one region, while those that increase in exchange upon activation are distributed throughout the enzyme. Together, these findings demonstrate that N-terminal processing of SplB markedly expands microsecond-to-millisecond conformational dynamics across a globally coupled network.

Fig. 3. R₂-CPMG dispersions of mature SplB and SplB-4.

(A) R₂-CPMG dispersions are shown for residues that increase exchange from SplB-4 (magenta) to mature SplB (green). Additional examples are shown in fig. S3A. (B) R₂-CPMG dispersions are shown for residues that are similar in exchange between SplB-4 (magenta) and mature SplB (green). Additional examples are shown in fig. S3B. (C) Amides exhibiting exchange (>1.5 s⁻¹) are mapped onto the x-ray structure (green spheres). Data were collected at 900 MHz at 25^oC, and full fits that included 600-MHz R₂-CPMG dispersions are shown in fig. S2.

To quantitatively define the exchange processes underlying these dispersions, we collected data for mature SplB at two magnetic field strengths (600 and 900 MHz) and independently fit using the Carver-Richards formalism (fig. S4 and table S1) (11). Extracted parameters reveal exchange rates on the order of ∼2000 s⁻¹ with minor-state populations of ∼1%. These values indicate rapid interconversion between a dominant ground state and a sparsely populated higher-energy conformational ensemble.

Together, these findings demonstrate that N-terminal processing does not simply stabilize an active-site geometry but instead unlocks a global microsecond-to-millisecond dynamic regime that facilitates access to higher-energy conformations. How these unlocked dynamics directly govern catalytic efficiency was next addressed by both MD simulations and mutational perturbations.

Identifying the structural basis of unlocking global dynamics through MD simulations

MD simulations (>110 μs in total) show that N-terminal processing activates SplB by redistributing a globally stable conformational ensemble toward more dynamic states, consistent with the microsecond-to-millisecond exchange detected by NMR (fig. S5 and Supplementary Text). Projections of the backbone Cartesian coordinates and residue-residue contact dynamics onto principal components (PCs) show that SplB-4 remained confined to a narrow basin, whereas mature SplB sampled a broader range of substates (Fig. 4, A and B; fig. S5, C to F; and Supplementary Text). Differences in residue-residue contacts calculated as previously described (12) and backbone fluctuations localized to the N terminus and two distal loops (residues 149 to 157 and 172 to 186), defining an allosteric network that extends to the loop containing residues 55 to 58 and links the site of activation to distal elements near the catalytic region (Fig. 4, C and D). Representative low-energy structures preserved the same overall fold in both states but revealed reorientation of the catalytic triad and repositioning of loops toward the active site in mature SplB relative to SplB-4 (Fig. 4E). These findings support a model in which removal of the extra N-terminal residues relieves conformational constraint and promotes SplB activation through long-range dynamic coupling. Notably, calculating the difference between the contact networks revealed by MD, as previously described (13), closely mirrors the experimental CSPs observed between mature SplB and SplB-4 (Fig. 2D). In particular, residues within the 55-to-58 loop and the catalytic-proximal loops (149 to 157 and 172 to 186) appear in both computational datasets, indicating that the conformational changes detected by NMR correspond to the same long-range network identified in the simulations. This agreement between experiment and simulation suggests that N-terminal processing redistributes the conformational ensemble along a defined allosteric pathway connecting the N terminus to the catalytic machinery.

Fig. 4. MD analysis reveals N-terminally controlled dynamic allostery in SplB.

(A) Time evolution of the first principal component (PC1) from principal components analysis (PCA) of the combined trajectories for the mature SplB (green) and SplB-4 (magenta). (B) Projection of the simulation trajectory onto the PC1 and PC2 conformational space of for mature SplB (green) and SplB-4 (magenta). (C) Mapping of residue-residue contact changes between mature SplB and SplB-4 states onto the SplB structure. Red depicts residue-residue contacts that are less formed from going from SplB-4 to mature SplB, and blue depicts contacts that are formed more from the same transition. The thickness of the cylinder depicts the magnitude of the change. (D) Absolute difference in residue flexibility (|ΔRMSF|) of each protein residues between the mature SplB and SplB-4 states with the average of 0.21 ± 0.40 Å delineated (dashed line). (E) Comparison of sampled structures of mature SplB (green) and SplB-4 (magenta) at 30 ms illustrating then conformational changes during MD simulations (left), along with the catalytic triad (right).

Dynamics-based mutagenesis alters conformational dynamics and controls substrate engagement

To determine whether the microsecond-to-millisecond dynamics unlocked by N-terminal processing directly govern SplB function, we applied a dynamics-guided mutagenesis strategy previously used to relate conformational exchange to catalytic output (14, 15). This approach selectively mutates dynamic residues identified via their intrinsic R₂-CPMG dispersion and evaluates whether functional changes correspond to altered dynamics. Furthermore, mutations are guided by evolutionary substitutions that we have shown often induce functional shifts (16). Here, we describe the selection of SplB mutations and then both their functional and dynamic effects, which reveal the underlying dynamic basis of SplB substrate engagement.

For the selection of SplB mutations, we focused on a flexible loop spanning residues 52 to 61 that displayed pronounced microsecond-to-millisecond exchange upon activation and was distal from the catalytic triad (Fig. 3A), which was also coupled to the N-terminal processing in the MD simulations (Fig. 4C). From this region, we selected N55 and K58, both of which exhibit clear R₂-CPMG dispersion, to test whether modulating intrinsic dynamics affects substrate engagement and turnover. K58 was mutated to tyrosine (K58Y), a residue present in other Spl family members, and N55 to serine (N55S), corresponding to SplA (fig. S6, A and B). Both SplB K58Y and N55S variants were well folded, and their backbone atoms were assignable by three-dimensional (3D) HNCA spectra. As expected, CSPs were strongest near the mutation site, but additional distributed CSPs extended beyond the loop (fig. S6, C and D), consistent with long-range structural coupling rather than global unfolding.

To probe the functional effects of these dynamics-based mutations and compare these functional effects to dynamics effects below, we assessed both their catalytic activities using ultraviolet (UV)–based assays and binding using NMR. For catalysis, we used the commercially available Boc-Glu-phenyl ester substrate commonly used for similar serine proteases (17). As expected, mature SplB catalyzed turnover for this substrate, whereas the SplB-4 zymogen mimic was catalytically inactive (Fig. 5A), consistent with its inability to process the SUMO-WELQ-GB1 (Fig. 2A). The dynamics-based mutations modulated catalytic turnover. While SplB K58Y showed a modest twofold increase in k_cat relative to wild type (WT), N55S exhibited an approximately threefold reduction. Thus, perturbing distal dynamic residues modulates catalytic turnover, with N55S placing functionally between the inactive SplB-4 and the fully active mature enzyme. For binding, we compared affinities of the preferred WELQ peptide motif (10). SplB WT and K58Y exhibited comparable global CSPs and slow-exchange behavior for residues such as Y148, whereas N55S displayed markedly attenuated CSPs (Fig. 5B and fig. S7A). Although apparent affinities were weakened for both mutants, N55S titrations remained linear within the accessible concentration range (fig. S7B), consistent with severely compromised engagement similar to the SplB-4 mimic. Thus, SplB N55S and the SplB-4 zymogen mimic engage the WELQ peptide much weaker than the mature SplB, raising the question of whether their dynamics were similar to each other and distinct from the mature SplB.

Fig. 5. Dynamics-based mutations and their functional outcomes.

(A) UV kinetics of mature SplB WT (green), K58Y (orange), N55S (blue), and SplB-4 (magenta) using the model Boc-Glu-phenyl substrate. Extracted parameters from Michaelis-Menten kinetics fits were WT: K_M = 5.7 ± 1.3 mM, k_cat = 2.3 ± 0.3 s⁻¹; K58Y: K_M = 5.9 ± 0.7 mM, k_cat = 5.7 ± 0.7 s⁻¹; N55S: K_M = 3.0 ± 0.9 mM, k_cat = 0.7 ± 0.1 s⁻¹. (B) Selected region of ¹⁵N-HSQC titrations of the WELQ product for WT (top panel), K58Y (middle panel), and N55S (bottom panel) with WELQ at 0 mM (black), 0.25 mM (skyblue), 0.50 mM (orange), 1 mM (green), 2 mM (purple), and 4 mM (red). The amide of Y148 is in slow exchange (red arrow). (C) R₂-CPMG dispersions are shown for SplB WT (green), K58Y (orange), and N55S (blue) at 900 MHz of representative residues from two groups based on their dynamic response to mutations. All other exchanging residues are shown in fig. S7C. (D) Group I (cyan) and group II (magenta) are mapped onto the x-ray crystal structure, along with residues within group III that displayed different responses (gray). R₂-CPMG data and titration data were collected at 900 MHz at 25^oC.

Comparative R₂-CPMG analyses of SplB WT, K58Y, and N55S revealed two distinct dynamic responses, referred to as group I and group II (Fig. 5, C and D, and fig. S7C), which were nearly identical to those residues that increase exchange upon activation and those that remain similar, respectively. Group I residues showed modest attenuation of exchange in K58Y relative to WT but near-complete suppression in N55S, closely resembling the dynamically restricted SplB-4 zymogen. In contrast, group II residues showed enhanced dispersion in K58Y while retaining exchange in N55S. Thus, the increase in R₂-CPMG dispersion observed upon N-terminal activation from SplB-4 to mature SplB is selectively reversed by N55S at group I residues. In contrast, group II dynamics change within both mutations but remain dynamic, indicating coupling to both mutation sites.

Together, these data establish a direct correspondence between conformational dynamics and function. Activation from SplB-4 to mature SplB is accompanied by an expansion of microsecond-to-millisecond dynamics within group I residues and a corresponding increase in substrate engagement. The N55S mutation largely reverses these activation-dependent dynamics for group I residues while maintaining differential dynamics within group II residues and retaining partial catalytic activity. These observations suggest that the conformational ensemble of N55S is shifted toward that of the SplB-4 zymogen mimic, but not identical, with residual sampling of catalytically competent ensembles. These findings demonstrate that microsecond-to-millisecond dynamics selectively activated upon N-terminal processing govern access to binding-competent states.

DISCUSSION

Proteolytic activation by N-terminal processing is essential for many serine proteases, yet for subclasses such as the Spl enzymes and mammalian granzymes, the underlying mechanism has remained unresolved (4, 5). Structural studies established that processing enables insertion of the newly generated N terminus into the protease core, but crystallographic comparisons revealed no change in catalytic triad geometry (4). This apparent repositioning of the N terminus without detectable active-site rearrangement has created a longstanding mechanistic paradox.

Here, we resolve this paradox by showing that N-terminal processing remodels the solution ensemble rather than the static ground-state structure. N-terminal insertion reorganizes an allosteric network that is coupled to a global expansion of microsecond-to-millisecond conformational dynamics, enabling substrate engagement (Fig. 6A). The SplB-4 zymogen mimic is properly folded yet exhibits only limited, localized dynamics and is incompetent in substrate binding, whereas the mature enzyme displays more widespread conformational exchange that connects the N terminus to the catalytic core. These two groups of residues, localized dynamics that are comparable between SplB-4 and mature SplB and those that increase within mature SplB, also fall within two groups based on their dynamic responses to mutations, referred to as group II and group I, respectively. MD simulations further support this redistribution of conformational states upon activation, including changes in the relative orientations sampled by the catalytic machinery. Thus, activation is associated with access to conformational states required for productive substrate engagement (Fig. 6B).

Fig. 6. Model of N-terminally driven allosteric activation in SplB.

(A) N-terminal processing triggers an allosterically coupled switch (magenta) that unlocks a global dynamic switch (green spheres), creating a communication pathway between the N terminus and the catalytic region. (B) Schematic of the activation landscape. The zymogen is catalytically “locked” despite being properly folded and exhibiting limited, localized dynamics. N-terminal processing expands microsecond-to-millisecond conformational dynamics, enabling the enzyme to access a broader ensemble that includes states competent for substrate binding, depicted as binding to the WELQ peptide.

Potentially one of the most telling findings regarding the SplB dependence on dynamics is the comparisons of SplB-4, mature SplB, and the SplB N55S mutation. Namely, activation, monitored as the transition from SplB-4 to mature SplB, increases exchange and peptide binding for a largesubset of residues, whereas the N55S mutation largely suppresses these activation-dependent dynamics for the same residues (group I) and is accompanied by reduced peptide binding. These results establish a direct correspondence between conformational dynamics and function, where activation-dependent expansion of dynamics enables substrate engagement, while suppression of these motions impairs access to binding-competent states.

Serine proteases have been proposed to dynamically sample both inactive (E*) and active (E) conformations (1), which we have directly visualized in our recent study of exfoliative toxin A (15). The SplB studies presented here are consistent with the E* ⇌ E framework with one important distinction. Namely, N-terminal processing acts as an upstream regulatory event that enables this equilibrium by expanding the accessible conformational landscape. Rather than stabilizing a single active conformation, N-terminal insertion promotes transitions into a substrate-competent ensemble. In this context, activation reflects a shift in the population and accessibility of conformational states rather than a discrete structural transition. Dynamic conformational gating therefore provides the mechanistic link between zymogen maturation and catalytic competence.

More broadly, this mechanism likely extends across the Spl family and related granzymes, which share conserved modes of N-terminal activation despite minimal differences in their crystallographic ground states. For SplB, a protease highly expressed in pathogenic MRSA strains (fig. S1) and recently implicated in immune modulation (7), this dynamic regulatory mechanism directly connects maturation to virulence-associated function. These findings establish dynamic allostery as a central feature of N-terminal activation in this subclass of serine proteases and provide direct evidence that conformational sampling, rather than static structure alone, governs catalytic competence.

MATERIALS AND METHODS

Proteomic analysis

To assess global differences in protein abundance between Staphylococcus aureus strains, representative methicillin-sensitive (MSSA) and methicillin-resistant (MRSA) isolates were cultured overnight in 5 ml of rich growth medium (TSB), then subcultured into 10 ml of TSB, and grown at 37°C with shaking (225 rpm) until they reached early stationary phase [optical density at 600 nm (OD₆₀₀) ∼ 2.0]. Cells were harvested by centrifugation at ×3200g for 10 min at 4°C, washed twice with phosphate-buffered saline, and processed for mass spectrometry (MS) using S-Trap micro filters (Protifi, Huntington, NY) according to the manufacturer’s instructions.

Proteins were digested on-column, and resulting peptides were loaded onto individual Evotips (Evosep, Odense, Denmark). Peptides were separated using an Evosep One liquid chromatography system equipped with a PepSep C18 column (150-μm inner diameter, 15-cm length) packed with ReproSil C18 resin (1.9-μm particle size, 120-Å pore size). The chromatography system was coupled via a nanoelectrospray ion source (CaptiveSpray, Bruker Daltonics) to a timsTOF Pro mass spectrometer (Bruker Daltonics, Bremen, Germany). MS data were acquired in parallel accumulation–serial fragmentation (PASEF) mode. The ion mobility ramp time was set to 100 ms, and 10 PASEF tandem MS (MS/MS) scans were acquired per acquisition cycle. MS and MS/MS spectra were collected over a mass/charge ratio of 100 to 1700, with ion mobility scanned from 0.7 to 1.50 Vs/cm². Precursor ions were isolated within ±1 Thompson units and fragmented using ion mobility–dependent collision energies, linearly increased from 20 to 59 eV in positive ion mode. Low-abundance precursor ions with intensities above 500 counts but below a target value of 20,000 counts were repeatedly scheduled for fragmentation, while higher-abundance precursors were dynamically excluded for 0.4 min. Raw MS data were processed using MSFragger within the FragPipe software suite (v19) and searched against the appropriate S. aureus proteome database. Peptide-spectrum matches were validated using Percolator, applying a false discovery rate (FDR) threshold of 1% at the protein level. Label-free protein quantification was performed using IonQuant. Differential protein abundance analyses were carried out using MetaboAnalyst v5.0, with statistical significance assessed using two-sample t tests or analysis of variance (ANOVA), as appropriate, and adjusted for multiple hypothesis testing using FDR correction (adjusted P < 0.05).

SplB protein expression and purification

All S. aureus SplB constructs were cloned on the basis of the UniProt accession Q2FXC3, with residue numbering corresponding to the mature enzyme following removal of the N-terminal 37-residue zymogen segment, consistent with previously reported x-ray crystal structures [Protein Data Bank (PDB) IDs 2VID and 4KIS]. All SplB constructs were expressed from pET21 vectors encoding an N-terminal 6×His tag, followed by human SUMO1 (hSUMO1; UniProt P63165). The affinity tag was removed by cleavage with ULP1 protease, which was recombinantly produced and purified in house from pFGET19_ULP1 (Addgene), as described previously (9, 18, 19).

The initial SplB WT insert was polymerase chain reaction (PCR) amplified from S. aureus genomic DNA and ligated into a commercially synthesized pET21-6xHis-hSUMO1 vector containing an XhoI cloning site (Twist Bioscience). All additional constructs reported in this study (SplB-4, N55S, and K58Y) were generated using a nested PCR strategy, in which the N-terminal extension or point mutations were introduced via mutagenic primers, followed by reamplification of the full-length SplB and reinsertion into the same pET21-6xHis-hSUMO1 expression vector.

All pET21-6xHis-hSUMO1-SplB constructs were expressed using previously established protocols (9, 18, 19). Briefly, proteins were expressed in Escherichia coli BL21(DE3) cells. For unlabeled samples, 4-liter cultures were grown in LB medium supplemented with ampicillin (100 mg/liter) and induced with isopropyl-β-d-thiogalactopyranoside at an OD₆₀₀ of ∼0.6, followed by incubation for 3 hours at 37°C. For ¹⁵N- or ¹⁵N/¹³C-labeled samples, cells were grown in LB medium to an OD₆₀₀ of ∼0.5, pelleted, resuspended in M9 minimal medium containing ¹⁵N–ammonium chloride or ¹⁵N–ammonium chloride plus ³C-glucose, respectively, and allowed to express for an additional 5 hours at 37°C. Cells were harvested by centrifugation and stored at −80°C until purification.

Cell pellets were resuspended in Ni-A buffer [50 mM Na₃PO₄ (pH 7.0), 500 mM NaCl, and 10 mM imidazole] and lysed by sonication. Lysates were clarified by centrifugation, and supernatants were applied to Ni²⁺-affinity resin (Sigma-Aldrich). Bound protein was eluted using Ni-B buffer (Ni-A supplemented with 400 mM imidazole) and dialyzed back into Ni-A buffer before overnight cleavage with ULP1 protease at room temperature. Cleaved samples were passed over a second Ni²⁺-affinity column to remove the 6×His-hSUMO1 tag and ULP1 protease. The flow-through containing untagged SplB was concentrated and further purified by size exclusion chromatography on a Superdex 75 preparative column (Cytiva) equilibrated in NMR buffer [50 mM Na₂PO₄ (pH 6.0) and 50 mM NaCl]. Final protein concentrations ranged from 0.5 to 1 mM. Purified proteins were aliquoted, flash frozen, and stored at −80°C until use.

SUMO-WELQ-GB1 protein expression, purification, and catalytic assay

For construction of SUMO-WELQ-GB1, a synthetic gene encoding an N-terminal 6×His-tagged yeast SUMO (UniProt accession Q12306), followed by the sequence WELQGS and the streptococcal GB1 domain (UniProt accession P06654), was cloned into pET21. The construct was synthesized as a Gibson assembly fragment (Twist Bioscience, San Francisco, CA). Protein expression was carried out under conditions analogous to those used for SplB constructs described above. Following expression, the protein was purified by Ni²⁺-affinity chromatography and further polished by size exclusion chromatography using a Superdex 75 column. Catalytic activity was assessed by SDS–polyacrylamide gel electrophoresis following incubation of 1 μM mature SplB or SplB-4 with 5 μM SUMO-WELQ-GB1 after a 2-hour incubation.

Enzyme kinetics

Catalytic activity of SplB was monitored using a chromogenic ester substrate assay adapted from previously reported protocols for serine proteases (17). Reactions were carried out in 300 μl of total volume using a 1-mm pathlength cuvette in a SpectraMax Plus 384 plate reader (Molecular Devices, San Jose, CA). Enzymatic cleavage of the Boc-Glu-phenyl ester substrate (Bachem, Bubendorf, Switzerland) was followed by monitoring the increase in absorbance at 270 nm, corresponding to release of the phenolic leaving group. Reactions were initiated by addition of enzyme to preequilibrated substrate solutions and monitored continuously at 25°C. Initial velocities were determined from the linear portion of the absorbance trace (typically the first 30 to 60 s) and converted to reaction rates using an extinction coefficient of 1500 M⁻¹ cm⁻¹.

NMR spectroscopy

All NMR samples of SplB were prepared in NMR buffer [50 mM Na₂PO₄ (pH 6.0) and 50 mM NaCl] supplemented with 5% D₂O. Backbone resonance assignments were obtained using uniformly ¹⁵N/¹³C-labeled SplB variants at concentrations of 0.5 mM. Assignment spectra were acquired at 25°C on a Bruker 600-MHz spectrometer equipped with a cryogenic probe. Standard triple-resonance experiments, including HNCACB, CBCA(CO)NH, HNCO, and HN(CA)CO, were collected to assign backbone amide resonances. While SplB-4 required all four spectra for complete backbone assignments due to the global changes of CSPs, the backbone of both mutants, N55S and K58Y, could be fully assigned through HNCA spectra. Full backbone assignments for both SplB-4 and mature SplB were deposited in the Biological Magnetic Resonance Data Bank (BMRB) as BMRB ID are 53758 and 53759, respectively. All spectra were processed using NMRPipe (20) and analyzed using CcpNMR Analysis (21).

CSP experiments were conducted on a Bruker 900-MHz spectrometer equipped with a TCI cryoprobe at 35°C. ¹⁵N-HSQC spectra were collected at each titration point with the indicated ligand concentrations. CSPs (Δδ) were calculated using the combined chemical shift difference formula: Δδ = sqrt [(5*ΔδH)² + (ΔδN)²], as described previously (22). The scaling factor of 5 accounts for the reduced chemical shift dispersion in the ¹H dimension. Binding curves were analyzed by fitting CSPs to a single-site binding model using GraphPad Prism. All spectra were processed with NMRPipe and visualized in CcpNMR Analysis.

For relaxation experiments, R₂-CPMG relaxation experiments were conducted on both the Rocky Mountain Bruker 900-MHz spectrometer and the Bruker 600-MHz spectrometer at 25°C using TROSY selection. Data were analyzed to extract exchange rates and populations using CPMG_fit, as previously described (15, 23). ZZ-exchange experiments were collected on the Rocky Mountain Bruker 900-MHz spectrometer as previously described (16).

Computational analysis

MD simulations

To explore the conformational dynamics of the SplB protein, we generated time-resolved trajectories for both its mature SplB and SplB-4 systems using all-atom MD simulations conducted with the Amber 22 package (24). The initial atomic coordinates were obtained from the crystal structure available in the PDB (PDB ID 4K1S) (4). AlphaFold 3 was used to fill structural gaps in the crystallographic data (25). The protonation states of ionizable residues at physiological pH 7.0 were predicted using the PROPKA method (26). System preparation, including solvation and ion addition, was carried out using the tleap module of AmberTools (27). The proteins were placed in an octahedral box and solvated with TIP3P water molecules, ensuring a minimum clearance of 10 Å between the protein surface and the box boundaries. We kept the water molecules from the initial crystal structure. The Amber ff19SB force field was used to model the protein’s interatomic interactions (28). To neutralize the overall charge of each system, Na⁺ or Cl⁻ ions were added (29), followed by the addition of further ions to achieve a physiological ionic strength of 150 mM.

Energy minimization was first carried out for each system to relieve unfavorable steric contacts and stabilize the structure. Each system underwent 5000 minimization steps, including 3000 steps using the steepest descent method followed by 2000 steps of the conjugate gradient algorithm. Throughout the minimization process, harmonic positional restraints were imposed on solute atoms. These restraints were progressively relaxed across five consecutive stages, with the force constants gradually reduced from 500 to 0 kcal·mol⁻¹ Å⁻². After completion of the minimization stage, the systems were gradually heated from 100 to 300 K over a 500-ps period under constant-volume (NVT) conditions. Temperature regulation was achieved using a Langevin thermostat with a collision frequency of 1.0 ps⁻¹. During this heating stage, simulations were performed using a 1-fs time step, while positional restraints on the solute atoms remained in place. The strength of these restraints was systematically lowered in five stages, with force constants of 500, 300, 100, 50, and lastly 5 kcal·mol⁻¹ Å⁻². Once heating was completed, the systems were further equilibrated for 1 ns in the NPT ensemble at 300 K and 1 bar. During equilibration, a 2-fs integration time step was used, and positional restraints were removed. Pressure control was maintained using a Monte Carlo barostat with a pressure coupling constant of 1.0 ps, ensuring stable pressure conditions throughout the equilibration phase. For the treatment of electrostatics, long-range interactions were computed using the particle mesh Ewald method (30). Short-range nonbonded interactions were evaluated using a cutoff distance of 9 Å. To permit the use of the 2-fs time step, all bonds involving hydrogen atoms were constrained using the SHAKE algorithm (31). Coordinates from the simulations were saved every 10 ps during the trajectory for subsequent analysis. Production simulations were conducted for 60 μs for the mature SplB system and 50 μs for the SplB-4 system. The first 10 μs of the mature SplB system trajectory was discarded to allow for additional equilibration.

Principal components analysis

Principal components analysis (PCA) was performed to investigate dominant motions and to compare the conformational landscapes of the SplB WT and SplB-4 states of the protein. PCA was conducted on backbone atoms (N, Cα, C, and O) using the CPPTRAJ module of AmberTools. Before analysis, all conformations were aligned to the average SplB WT state structure to remove global translational and rotational motions. The variance-covariance matrix was diagonalized to yield eigenvectors and eigenvalues, which capture the magnitude of motion along each PC. For comparative analysis between the SplB WT and SplB-4 states, eigenvectors derived from the SplB WT state simulation were projected onto the trajectories of both states after alignment to a common reference frame. PCs were ranked by descending eigenvalues, reflecting their contribution to the total variance. A functionally relevant PC was identified as the one showing the highest correlation with a functional variable. To visualize structural changes along the first principal component (PC1), trajectories were rendered using Visual Molecular Dynamics software (32).

Residue-residue contact dynamics

Contact analysis was performed previously (12). Contact was defined when any two heavy atoms from different residues (with sequence separation ≥ 3) were within 4.5 Å. Contact probabilities were calculated over the simulation trajectories. The contact trajectory space for PCA of the contact trajectories was built from 455 residue-residue contacts, selected on the basis of contact probabilities between 0.1 and 0.9 to focus on interactions that form and break during the simulations while excluding contacts that are either consistently present or not present. Later PCA was applied to residue-residue contact trajectories to identify the dominant collective variations in contact patterns sampled during the simulations. Contact time series from both states were pooled and centered to construct a covariance matrix. This matrix can describe correlated fluctuations in contact formation. Eigenvalue decomposition of this covariance matrix generated orthogonal PCs ordered by their contribution to the total variance. Trajectories from both systems were subsequently projected onto the same PC subspace, allowing direct visualization of differences in contact dynamics between the states.

Supplementary Materials

The PDF file includes:

Supplementary Text

Figs. S1 to S7

Table S1

Legends for data S1 and S2

Other Supplementary Material for this manuscript includes the following:

Data S1 and S2