Identifying and controlling inactive and active conformations of a serine protease

Abstract

Serine proteases have been proposed to dynamically sample inactive and active conformations, but direct evidence at atomic resolution has remained elusive. Using nuclear magnetic resonance (NMR), we identified a single residue, D164, in exfoliative toxin A (ETA) that acts as a molecular “switch” to regulate global dynamic sampling. Mutations at this site shift the balance between inactive and active states, correlating directly with catalytic activity. Beyond identifying this dynamic switch, we demonstrate how it works in concert with other allosterically coupled sites to rationally control enzyme movements and catalytic function. This study provides a framework for linking conformational dynamics to function and paves the way for engineering enzymes, in particular, proteases, with tailored activities for applications in medicine and biotechnology.

Identifying and controlling the inherent dynamics of a serine protease between inactive and active ensembles.

INTRODUCTION

Enzymes are not static entities but dynamic molecular machines that undergo conformational changes essential for their catalytic function. These motions, ranging from localized side-chain rearrangements to large-scale domain movements, enable enzymes to transition between multiple conformational states. The theory of conformational selection, which posits that enzymes exist in a dynamic equilibrium between inactive and active forms, has emerged as a unifying framework for understanding enzyme mechanisms (1). Experimental techniques such as nuclear magnetic resonance (NMR) spectroscopy, x-ray crystallography, and molecular dynamics (MD) simulations have revealed dynamic processes in a variety of enzymes, including kinases (2, 3), reductases (4–6), isomerases (7), and serine proteases (8).

Serine proteases provide particularly compelling examples of how dynamics govern function. Serine proteases represent the most highly expressed type of proteases (8) and have been proposed to sample both inactive and active conformations for over a decade (9–11), reflecting the principles of conformational selection (12). However, directly detecting these conformations in solution at atomic resolution has been challenging. Recent NMR studies have identified dynamic exchange processes within several serine proteases, but issues such as protein size, resonance dispersion, and the dependence on large activators have complicated the establishment of a causative link between dynamics, allostery, and catalysis (13–16). Considering the widespread use of serine proteases in everything from commercial products to medicine, understanding and controlling conformational sampling could have wide applications in engineering proteases with modified activities (17).

Here, we focus on exfoliative toxin A (ETA), the primary member of the serine protease subfamily of Glu-endopeptidases secreted by highly infectious strains of Staphylococcus aureus, which target epithelial junctions to increase infections (18). Using our recently developed mutagenesis/NMR approach, called relaxation and single-site multiple mutations (RASSMMs) (19, 20), we have identified a single site within ETA that controls the global sampling of inactive and active conformations. We further demonstrate how mutagenesis can be used to rationally engineer this dynamic process to predictively control catalysis. Thus, we move beyond characterizing the dynamics that control sampling of inactive and active conformations of a serine protease to also demonstrate the power of dynamic design to precisely engineer catalysis.

RESULTS

A single mutation controls global sampling of inactive and active conformations

The prediction that serine proteases sample both inactive and active conformations implies that the energy landscape comprises ensembles representative of both forms (Fig. 1A), which can be tested. Specifically, NMR resonance positions that report on the chemical environment are a weighted average over sampled conformations. Here, we have discovered that mutation of one residue within ETA, D164, facilitates chemical shift perturbations (CSPs) in either of two directions measured within ¹⁵N–heteronuclear single-quantum coherence (HSQC) spectra (Fig. 1B). To compare these shifts to the catalytic activities, we used a model spectroscopic substrate previously used to probe the ET family of serine proteases (21, 22). This model substrate is used in lieu of their human host targets that are large adhesion proteins and difficult to purify, noting that peptide mimics do not bind and illustrating the exquisite specificity of ETs (23). The ETA CSPs due to D164 mutations also corresponds to shifts in the observed catalytic activity (k_cat) using this substrate (Fig. 1C and Table 1). Even higher substrate contractions lead to substrate/product inhibition but still exhibit the same pattern of lower or higher apparent turnover (fig. S1A). These data indicate that resonance positions are largely predictive of the dynamic sampling of inactive and active conformations. We first explain how this critical ETA D164 site was identified and then elucidate both the dynamic and structural basis for this global sampling herein.

Fig. 1. Controlling dynamic sampling of inactive and active conformations within ETA.

(A) A schematic of the conformational coordinate for ETA sampling of inactive (orange) and active (green) conformations, which are shifted by ETA D164E and D164A, respectively. (B) ETA CSPs are induced in either of two directions with D164E and D164A mutations, which include N65 and H72, as shown from ¹⁵N-HSQC spectra. (C) D164 mutations also shift catalysis of the Glu-Ophenyl substrate in either of two directions. Activity was monitored at 270 nm at 25°C.

Table 1. Catalytic parameters of ETA and mutants.

Data were collected at 25°C in triplicate in a 1 mm cuvette pathlength using 20 μM ETA and the Glu-Ophenyl substrate. All data were fit to a standard Michaelis-Menten equation.

Mutant	Vmax (μM s−1)	kcat (s−1)	KM (mM−1)
WT	25.9 ± 4.2	1.29 ± 0.21	4.13 ± 1.24
D164A	93.5 ± 10.8	4.67 ± 0.54	3.46 ± 0.80
D164E	ND	ND	ND
D164N	55.6 ± 17.0	2.78 ± 0.85	7.21 ± 2.65
D164R	53.0 ± 16.8	2.65 ± 0.70	7.84 ± 2.50
D164S	25.1 ± 2.9	1.26 ± 0.14	1.42 ± 0.49
W14A	4.74 ± 0.7	0.24 ± 0.04	3.64 ± 1.09
W14Y	7.43 ± 0.8	0.37 ± 0.04	2.13 ± 0.57
W14F	7.97 ± 0.9	0.40 ± 0.05	1.82 ± 0.58
W14A,D164A	19.2 ± 0.4	0.96 ± 0.03	1.5 ± 0.80

RASSMM identifies a global process coupled to ETA D164

We used the RASSMM approach, a method designed to elucidate the function of dynamic sites (“hot spots”) outside of an active site through systematic mutagenesis and NMR (19, 20). For ETA, all backbone amides resonances were assigned except D164 and G193, due to severe micromillisecond dynamics causing line broadening confirmed via mutations described below. D164 lies outside of the active site and its side chain forms a hydrogen bond to the G193 amide that is part of the active site (Fig. 2A). Notably, D164 is not directly involved in the catalytic triad of ETA (H72, D120, and S195) and its absence is consistent with an inherent dynamic. Thus, on the basis of our RASSMM approach that seeks inherently dynamic sites to mutate in order to modulate global dynamics and determine the site-specific roles in function, we hypothesized that D164 could act as a dynamic “hot spot” coupled to function.

Fig. 2. Identification of the global dynamic control point, ETA D164.

(A) The x-ray crystal structure of ETA with its catalytic triad shown (H72, D120, and S195) along with the hydrogen bond between D164 and G193 (PDB 1AGJ). (B) ¹⁵N-HSQC spectra reveal missing amides of D164/G193 in ETA WT that are observable only within the inactive D164E mutant. (C) Amides that exhibit collective CSPs in one of two directions due to D164E and D164A mutations are illustrated (red spheres) with D164 shown (yellow bonds). These include W14, K16, V20, N24, F30, K32, V33, N48, F50, V51, G53, T55, S56, A57, T58, G59, N65, V67, R71, H72, I73, K75, G83, E113, F115, D120, I124, G153, L158, G160, K166, V167, N168, R172, E174, N194, S197, G198, I199, H210, S212, K223, G226, and I237. (D) ETA D164N resonance positions are shown for N65 and H72 in the context of the inactivating D164E and activating D164A mutation within ¹⁵N-HSQC spectra along with their comparisons to catalytic activity of Glu-Ophenyl. (E) ETA D164R is shown as in (D). (F) ETA D164S is shown as in (D) and (E).

To test this, we introduced mutations at D164, revealing substantial NMR and functional changes. The D164A mutation globally increased line broadening across the enzyme, while D164E globally reduced it, facilitating the observation of amide resonances for the D164E mutation itself and G193 (Fig. 2B). CSP analysis (CHESPA) was used to identify coupled networks through their coupled CSPs (24). CHESPA revealed 40 residues with opposing CSPs between D164A and D164E, indicating a global conformational shift along two trajectories (Fig. 2C and fig. S1). This suggests that D164 acts as a “switch,” biasing the dynamic sampling of inactive (D164E) versus active (D164A) conformations, with CSPs predictive of catalytic activity.

Additional mutations at D164, designed to preserve its hydrogen bond to G193, further supported this model. Mutations such as D164N, D164R, and D164S exhibited intermediate CSPs and catalytic activities similar to the wild-type (WT) enzyme (Fig. 2, D to F, and Table 1). The conservation of similar residues at this position in related serine proteases supports its evolutionary significance (e.g., ETB N158 and ETD N189) along with more distantly related serine proteases like thrombin (N143) (25).

Attempts to probe ETA-substrate interactions through CSPs, using an inactivated mutant (S194A), were limited by the weak binding affinity of the spectroscopic substrate mimic used in these studies. Unfortunately, substrate binding was too transient to induce measurable CSPs, even at high substrate concentrations. In addition, no CSPs were observed in response to high concentrations of Glu alone, indicating minimal perturbations even for product interactions. These limitations underscore the challenges of directly linking substrate binding to the dynamic sampling of inactive and active ensembles using CSPs alone. To overcome this limitation, we turned to NMR relaxation studies of the free ETA enzyme to specifically address global dynamic sampling and directly quantify the dynamic effects imparted by D164 mutations.

NMR relaxation quantifies the shift in sampling inactive and active conformations

The CSPs observed in WT and mutant ETA provide direct evidence of altered dynamic sampling between inactive and active conformations. The shifts are in either of two directions, consistent with changes in the equilibrium between low-energy inactive and high-energy active ensembles induced by the D164 mutations. Specifically, D164E (inactive) decreases sampling of the high-energy active ensemble, while D164A (hyperactive) increases this sampling. Because NMR resonances represent weighted averages over all conformations sampled on the micromillisecond timescale, these bidirectional CSPs reflect shifts in ensemble populations.

To confirm this hypothesis and quantify the population shifts, we used R2-Car-Purcell-Meiboom-Gill (R2-CPMG) relaxation experiments. These experiments measure exchange contributions to relaxation (Rex), providing direct insight into the rate, populations, and chemical shift differences between sampled states (26, 27). The results revealed that D164E suppresses micromillisecond dynamics, while D164A enhances these exchange contributions, consistent with a global shift in the dynamic equilibrium toward inactive or active ensembles, respectively.

R2-CPMG dispersions collected for ETA WT, D164A, and D164E mutations highlight the dramatic influence of the D164 site on enzyme dynamics. The global nature of these motions is evident from exchange contributions plotted per residue at 900 MHz (Fig. 3A), where the D164E mutation eliminates micromillisecond dynamics, while the D164A mutation amplifies these contributions. This pattern is consistent across measurable dispersions for 53 amides in WT and D164A variants, except for F163, which lies adjacent to the mutation site (Fig. 3B). The D164A mutation increases the local sampling of minor (active) conformations, providing a physical basis for enhanced catalytic activity (Fig. 2, B and C, and fig. S2, A and B).

Fig. 3. ETA activity is increased by increased sampling of the active conformations.

(A) Rex extracted from R2-CPMG dispersion amplitudes at 900 MHz. Rex values were extracted as the experimental differences between the lowest field (80 Hz) and highest field (1000 Hz) are shown for WT (black), D164A (green), and D164E (orange). (B) Full R2-CPMG dispersions are shown for selected residues with all amides that exhibit a measurable increase of exchange within D164A mapped onto the x-ray crystal structure of ETA (green spheres). Diminished exchange is also shown for the one single D163A residue (blue sphere). All data were collected at 900 MHz at 25°C.

Despite this global increase in sampling of minor states, the dynamics remain localized. For instance, while D164A broadly increases exchange contributions (Rex), the rates of motion for individual residues show localized variations, with some decreasing and others remaining unchanged (fig. S2C). This localized behavior, despite large-scale coupling, is a hallmark of many enzymatic systems (28–30). The measured amide nitrogen CSPs between D164E and D164A are much smaller than the extracted chemical shift changes from R2-CPMG dispersions (fig. S2D). These smaller experimental CSPs likely reflect an “on-path” portion of the shifts to the excited ensembles associated with catalytic activity, while the extracted values from R2-CPMG dispersion capture a more comprehensive difference between ground and excited ensembles. This interpretation is consistent with the role of D164 as a key modulator of dynamic sampling and catalytic activity, wherein D164A shifts the equilibrium toward a more catalytically competent high-energy ensemble. Furthermore, both the ETA WT and D164A mutant sample largely identical “high-energy” conformations, as their extracted CSPs are largely identical (fig. S2E). This implies that the primary effect of the D164A mutation is to increase the dynamic sampling of the higher energy ensembles that are active, rather than drastically altering any particular conformation.

In addition, extracted chemical shift changes (Δv) from R2-CPMG dispersions for D164A and D164E mutants exhibit no clear correlation with experimental CSPs (fig. S2D), further supporting a complex interplay of local dynamics. However, both D164A and WT enzymes show similar Δv values, indicating that they sample the same minor (active) conformations (fig. S2E).

Collectively, these findings indicate that ETA activity is modulated by shifts in the dynamic equilibrium between inactive and active conformations, without altering the ground state. Crystallographic structures of WT, D164A, and D164E enzymes confirm that all three variants share the same ground-state conformation (fig. S3 and table S2). This strongly supports the conclusion that activity changes result from altered dynamic sampling rather than structural changes in the ground state.

MD and experimental validation reveal the structural basis of allostery in ETA

MD simulations identified the structural basis of allostery in ETA by elucidating conformational changes coupled between position 164 and the N-terminal helix. Specifically, principal components analysis (PCA) of a 30-μs MD trajectory for ETA highlighted a dominant ratcheting motion of the N-terminal helix (Fig. 4A and movie S1), which can be observed in the ensemble distributions in PC1 (fig. S4A). While the D164A mutation induced a less distinct separation between these ensembles (fig. S4B), the D164E mutation led to a confined distribution (fig. S4C). D164 mutations that exhibit similar activities to the WT (D164N, D164R, and D164S) also exhibit similar MD distributions as the WT (fig. S4, D to F). This ratcheting motion modulates the positions of S195, H72, and D120 between open and closed states, essential for catalytic function (fig. 4A). Either a truncation of the entire N-terminal helix or mutation of W14 eliminated this coupling in similar MD simulations (fig. S4, G and H), underscoring how position 164 modulates the ratcheting motion of the N-terminal helix, crucial for allosteric regulation.

Fig. 4. Identifying ETA sampled conformations and testing these through ETA W14 mutations.

(A) Representative structures from ETA MD simulations highlighting the ratcheting motion of the N-terminal alpha helix and the reorientation of the catalytic triad (S195, D120, and H72) along the PC1 axis (see fig. S4). (B) ETA W14A CSPs >0.24 parts per million (ppm) relative to ETA WT (average CSP plus ½ SD) are mapped onto the x-ray crystal structure (red spheres) with the electrostatic surface connecting W14 (yellow) to the active site. (C) Catalytic activity of ETA W14 mutants using Glu-Ophenyl was conducted under identical conditions to those described in Figs. 1 and 2. Each W14 mutant is shown as W14F (purple), W14Y (cyan), and W14A (brown). (D) Resonance positions for W14 mutants for H72, R71, and R185 are shown in the context of D164E and D164A mutations from ¹⁵N-HSQC spectra. The positions of ETA WT resonances are shown (gray vertical dashed line) and shifts of D164E, D164A mutations indicated with arrows (top of each spectra).

To experimentally confirm the role of the N-terminal helix in coupling dynamics to the active site, we combined mutagenesis with NMR analysis using ¹⁵N-HSQC spectra. W14, embedded within a pocket linking the helix to the active site, was chosen for targeted mutation. W14 was chosen on the basis of both on its likely coupling of the N-terminal helix to the catalytic core but also because it is one of the 40 residues that exhibits CSPs in either of two opposing directions induced by D164E and D164A, supporting its involvement in the coupled network. Mutations to W14 induced CSPs that propagated to the active site (Fig. 4B) and led to reduced activities (Fig. 4C). Unlike D164 mutations, W14 mutations exhibited varied directional CSPs, indicating a “short circuiting” of long-range coupling (Fig. 4D). For example, while some CSPs were in the direction of the hyperactive D164A mutation (e.g., H72), others remained unchanged (e.g., R71) or shifted toward the inactive D164E mutation (e.g., R185). In other words, the CSPs were no longer predictive of the catalytic outcome. This disruption of coordinated dynamics reduced catalytic efficiency, a phenomenon previously observed in other enzymes (29). These results illustrate that precise dynamic coupling is critical for optimal allosteric regulation.

By integrating MD simulations and experimental validation, we reveal the N-terminal helix as a key modulator of allostery in ETA. This helix orchestrates global dynamics by linking distal sites, such as W14, to the active site, enabling fine-tuned control over conformational sampling critical for catalysis.

Predictive “fine-tuning” of catalysis

The ability to manipulate conformations within ETA using specific mutations raises the question of whether such mutations can be combined rationally to achieve desired outcomes. This potential for fine-tuning is exemplified by combining ETA D164A, which enhances catalysis, with ETA W14A, which reduces catalysis. Therefore, we produced the double mutant of ETA W14A,D164A and observed that many CSPs from the individual mutations were additive. Specifically, CSPs induced by ETA D164A and W14A show cumulative effects, such as 72 shifting further in the same direction in the context of the double mutant (Fig. 5A). Amides that exhibit CSPs in opposite directions in the single mutations now display intermediate shifts in the double mutations, exemplified by R185 (Fig. 5B). These additive effects reinforce the notion of a dominant two-state sampling at the local level, directly influencing the equilibrium between inactive and active conformations. Ultimately, these additive effects culminate in predicted catalytic outcomes, with kcat values in between that of the individual mutations (Table 1).

Fig. 5. Combining mutations with a predictive outcome.

(A) H72 CSPs are shown for ETA WT (black), ETA W14A (maroon), and the double-mutation ETA W14A,D164A (blue) in the context of the inactive (D164E) and hyperactive (D164A) mutations from 15 N-HSQC spectra. (B) R185 is shown as in (A). (C) Catalytic activity is also additive for the double mutant, ETA W14A,D164A compared to ETA WT, W14A, and D164A.

CSPs observed within the double mutation ETA W14A,D164A provide further evidence for the segmental nature of allostery within ETA. While many CSPs show additive effects, some remain similar to the ETA WT or simply do not shift any further than that of D164A (extended data fig. S5B). This would be predicted considering that the individual W14A mutation “short circuits” many of the individual local samplings between inactive and active conformations. Furthermore, some of the local sampling would be expected to be close to that of the ETA WT considering that catalytic activity is similar. Thus, allosteric communication within ETA involves a complex interplay of partially coupled samplings of inactive and active forms.

DISCUSSION

Proteins, like all biological machines, rely on precise movements to perform their functions. For serine proteases, enzymes central to countless biological processes, these movements involve a delicate balance between inactive and active states. Our study provides the first direct evidence at atomic resolution that serine proteases inherently sample both conformations, confirming a long-standing hypothesis about their dynamic nature (9–11). We demonstrate how these movements can be identified and controlled, paving the way for a new era of enzyme engineering.

Using ETA as a model system, we identified a single residue, D164, that acts as a molecular “switch” to regulate dynamic sampling. Mutations at this site globally shifted the balance between inactive and active states, a shift that directly correlated with catalytic activity (19, 20). By harnessing NMR and MD simulations, we mapped these dynamic movements with atomic resolution, revealing how they shape the enzyme’s function (13–15). This dynamic sampling mechanism is likely shared among serine proteases and aligns with broader theories of conformational selection, where enzymes exist in equilibrium between multiple conformational states. For example, in thrombin, distinct structural rearrangements, such as shifts in W215, have been observed to sterically occlude substrate binding in the inactive state, representing a static barrier to activity (24). In contrast, the evidence in ETA suggests that activity is regulated by dynamic interconversion between inactive and active states, without the need for gross structural changes. This helps explain why x-ray structures of ETA mutants (e.g., D164A and D164E) appear nearly identical, as their ground states are largely identical. In contrast, NMR spectral positions and relaxation dispersion reveals substantial differences in the underlying conformational dynamics in solution. These findings reinforce the central role of dynamics in regulating serine protease activity and underscore the complementary strengths of NMR and x-ray techniques in studying enzymes that rely on conformational equilibria. Such findings also reinforce the well-known extension of the structure-function relationships to illustrate how structure-dynamics-function relationships must be considered. The ability to both observe and manipulate the conformational landscape represents a critical breakthrough in understanding enzyme dynamics.

Our findings have profound implications. First, they demonstrate that conformational sampling is not just a theoretical concept but a tangible feature that can be harnessed to predict and control enzyme function (12, 29). Second, this work underscores how localized changes, such as mutations at a single residue, can propagate through an enzyme to globally regulate its dynamics and function (25, 28). Last, the ability to fine-tune these dynamics suggests exciting applications, from designing custom enzymes for industrial processes to creating targeted therapeutics (17).

Our findings also have important implications for serine proteases in general, as multiple subfamilies are reliant on their N termini for proper activity, as we have revealed here for the ratcheting motion of the N terminus of ETA. For example, the S. aureus serine-like protease subfamily and human granzyme subfamilies must be precisely activated to facilitate their correct N-terminal residue (31, 32). These specific N-terminal residues fold into their respective serine proteases to allosterically activate them through unknown mechanisms, but illustrate how serine proteases have become reliant on their N termini.

The broader relevance of this work lies in its demonstration of how dynamic sampling, long recognized as an essential feature of enzymes, can be systematically identified, mapped, and controlled within a serine protease. While the existence of inactive and active conformations has been proposed in the past, our study provides direct evidence and a method to manipulate these states with precision. By linking enzyme dynamics to function, we establish a framework for understanding how conformational sampling underpins catalytic efficiency and specificity. This ability to rationally control enzyme movements offers an exciting avenue for advancing both fundamental enzymology and practical applications in biotechnology and medicine, illustrating the untapped potential within nature’s molecular machinery.

MATERIALS AND METHODS

Protein expression and purification

The sequence for mature ETA of residues 38 to 280 (UniProt ID P09331) was cloned into pET21 with an N-terminal 6xHis-tag and the plasmid was transformed into BL21 (DE3) cells at 37°C. Unlabeled protein was growth in Luria broth (LB). For labeled proteins, three labeling schemes were used that included ¹⁵N-labeled ammonium chloride for only ¹⁵N-labeled proteins used for standard ¹⁵N-HSQC spectra, ¹⁵N-labeled ammonium chloride grown in D₂O for ¹⁵N,²H-labeled growths for R2-CPMG relaxation dispersions, and ¹⁵N-labeled ammonium chloride with ¹³C-labeled glucose and D₂O for ¹⁵N,¹³C,²H-labeled growths for assignment spectra. A typical growth comprised 4 liters at 37°C with ampicillin selection and induced at 0.6 to 0.8 optical densities (ODs; 600 nm) with isopropyl β-d-1-thiogalactopyranoside (IPTG) for 3 to 4 hours. Labeled proteins were grown via media switching, which begins in LB similarly to unlabeled growths but cells spun at 0.4 OD (600 nm) and the media was switched to 1/4th the volume. Cells were allowed to adapt for approximately 1 hour until induction with IPTG and at least 5 to 6 hours.

For purification, both the unlabeled proteins and ¹⁵N-labeled proteins were retrieved from soluble preparations and the modification for refolding of deuterated proteins is described afterwards. All purified on an AKTA Pure FPLC (Cytiva Life Sciences, St. Louis, MO). For soluble preparations, cells were lysed via sonication and applied to a column with Ni affinity (Sigma-Aldrich) in “Ni-A buffer” [50 mM Na₂HPO₄ (pH 7), 500 mM NaCl, and 10 mM imidazole] and eluted with “Ni-B buffer” [50 mM l Na₂HPO₄ (pH 7), 500 mM NaCl, and 400 mM imidazole]. Protein elutions were concentrated and further purified via size exclusion chromatography using a Superdex-75 column (Cytiva, 120 ml total bed volume) in “NMR buffer” [50 mM Na₂HPO₄ (pH 6) and 50 mM NaCl]. For deuterated proteins (¹⁵N,²H- and ¹⁵N,²H,¹³C-labeled proteins), the only difference was that both Ni-A and Ni-B were supplemented with 5 M guanidine-HCl and the proteins and the elutions from Ni-affinity columns were refolded through 1 M arginine as previously described before size exclusion chromatography (6, 33).

Monitoring catalytic activity

The assay for monitoring ETA activity was previously reported (21). In brief, a SpectraMax Plus 384, Molecular Devices (San Jose, CA), was used with a 1-mm cell and 300 μl total volumes to observe cleavage of the Glu-Ophenyl substrate (Bachem, Bubendorf, Switzerland). Activity was monitored at 270 nm using 20 μM ETA 340 nm using an extinction coefficent of 1500 M⁻¹ cm⁻¹ and the initial velocities were calculated in triplicate using the first 30 to 60 s of data.

NMR spectroscopy

All samples were prepared in NMR buffer at 500 μM protein with 5% D₂O. All ¹⁵N-HSQC spectra were collected identically with 90 increments in nitrogen and 2048 points in the direct dimension. Full backbone assignment spectra were collected for ETA WT and D164E using ²H,¹⁵N,¹³C-labeled proteins. Assignment spectra were collected on a Bruker 600 spectrometer equipped with a cryo-probe at 25°C that included an HNCA, HN(co)CA, HNCACB, and a HN(co)CACB. Three-dimensional spectra used nonuniform sampling reconstructed to 72 and 96 increments in the nitrogen and carbon dimensions, respectively (34). R2-CPMG experiments were collected for ETA WT, D164E, and D164A using ²H,¹⁵N-labeled proteins with transverse relaxation-optimized spectroscopy selection as previously described (29, 35). Both the ETA WT and D164A mutant were also collected on a Bruker 600-MHz spectrometer for subsequent fits. R2-CPMG dispersions were fit to extract the associated parameters using CPMG-Fit (provided by D. M. Korzhnev, University of Connecticut).

X-ray crystallography

Crystallization screens and hanging-drop plates were purchased from Molecular Dimensions (Holland, OH, USA). Chemicals used for protein crystallization were purchased from ChemImpex Inc. (Wood Dale, IL, USA). Crystal mounts, loops, and a UniPuck system were purchased from MiTeGen (Ithaca, NY, USA). ETA D164E was crystallized under previously determined crystal conditions [50 mM sodium phosphate (pH 7.0) with 22.5% polyethylene glycol (PEG) 4000] in a 2:2-μl protein:mother liquor ratio at room temperature with a protein concentration of 15 mg/ml (21). ETA D164A was subjected to high-throughput crystallization screening carried out with commercially available screens in small-volume sitting-drop trays using a Crystal Gryphon LCP robot (Art Robbins Instruments; Sunnyvale, CA, USA). ETA D164A was crystallized under the MCSG1 G10 condition (0.1 M magnesium formate + 17.5% PEG 3350) in a 2:2-μl protein:mother liquor ratio at room temperature with a protein concentration of 17.5 mg/ml (36). All crystals were grown using the hanging-drop vapor diffusion method in 24-well hanging-drop plates and cryoprotected with the same mother liquor supplemented with 20% (v/v) glycerol. Diffraction data for the ETA D164E and D164A crystals were collected at the Canadian Light Source BM beamline on a Detectris Pilatus3 S 6M at a wavelength of 1.18069 Å. All data were indexed, integrated, and scaled with DIALS (version 3.8.0) (37) and imported into the CCP4i suite (version 8.0.009) (38) with AIMLESS (version 0.7.9) (39). Phases for all structures were solved using molecular replacement (MOLREP version 11.9.02) (40) with WT ETA [Protein Data Bank (PDB): 1AGJ] (21). Refinement was done using phenix.refine (version 1.20.1_4487) (41) in conjunction with manual model building in COOT (version 0.8.9.2) (42). B factors were refined anisotropically with the exception of water molecules, whose B factors were refined isotropically. Model geometry was analyzed and optimized based on suggestions by MolProbity (version 4.5.2) (43). Data collection and model statistics for both structures are summarized in table S2.

MD simulations

MD simulations were used to generate time-evolved trajectories for the WT ETA protein and five of its mutants: D164A, D164E, D164N, D164R, and D164S. All simulations were performed using Amber 22 program (44). The initial coordinates for the WT protein were obtained from the PDB (1AGJ) (21). Mutations at the D164 position were introduced using PyMOL (45). Protonation states of the residues were determined at pH 7 using the PROPKA method (46). Each system was prepared using the tleap module of the AmberTools suite. The protein systems were solvated in octahedral boxes of water molecules with a 10 Å buffer distance between the protein and the edges of the solvent box. Amber ff19SB force-field parameters were used for protein atoms, while the TIP3P model was used for the water molecules. The systems were neutralized with Na⁺ or Cl⁻ counterions, and additional ions were added to achieve a physiological concentration of 150 mM. Energy minimization was performed on the prepared systems with restraints on protein atoms, gradually relaxing these restraints from 500 to 0 kcal/mol·Å² over six rounds. Each round consisted of 2000 steps of steepest descent followed by 3000 steps of conjugate gradient. Temperature control was maintained using a Langevin thermostat, with an initial temperature of 100 K gradually raised to a target temperature of 300 K. The collision frequency was set to 1.0 ps⁻¹. The systems were equilibrated over 2.5 ns under an NVT ensemble in five rounds, with restraints on protein atoms gradually reduced from 500 to 5 kcal/mol·Å². Subsequently, the systems were equilibrated under an NPT ensemble with isotropic pressure scaling to a reference pressure of 1.0 atm for 10 ns without any restraints. A Monte Carlo barostat with a coupling constant of 1.0 ps was used to maintain the pressure at 1.0 atm. The SHAKE algorithm was used to constrain bond lengths involving hydrogen atoms, allowing for a larger time step of 2 fs (47). Long-range electrostatic interactions were calculated using the particle mesh Ewald method, with short-range interactions cut off at 9 Å (48). Production runs for the WT, D164A, and D164E mutants were conducted for a minimum of 30 μs each, while the D164N, D164R, and D164S mutants were run for 25 μs each totaling 165 μs of simulation time across all systems. Another simulation was conducted for the WT protein using the same protocol, with the N-terminal helix (20 residues) removed and the free end capped with an acetyl group, for a duration of 10 μs.

Principal component analysis

PCA was used to investigate the dominant modes of motion and explore the conformational space of the ETA protein and its mutants. The analysis was conducted using the CPPTRAJ module of AmberTools (49). MD trajectories were preprocessed by aligning them to the first frame to eliminate overall translational and rotational motions. A covariance matrix was then calculated based on the x, y, and z coordinates of backbone atoms (N, Cα, C, and O) from the WT trajectory, capturing the correlated internal motions of the enzyme. Diagonalizing this matrix produced the eigenvectors and corresponding eigenvalues, representing the principal components (PCs) of motion. To compare the mutants’ conformational dynamics to the WT, eigenvectors derived from the WT were projected onto both WT and mutant trajectories, with all trajectories aligned to a common reference frame. This approach allowed for a direct comparison of the mutants’ conformational dynamics with those of the WT. The PC with the largest eigenvalue (PC1) corresponded to the greatest variance in conformational sampling within the trajectories. To gain structural insights, the conformational motions associated with the PC1 eigenvector were visualized using Visual Molecular Dynamics (50), providing a deeper understanding of the structural changes driving the observed dynamics.

Supplementary Materials

The PDF file includes:

Figs. S1 to S5

Tables S1 and S2

Legend for movie S1

Other Supplementary Material for this manuscript includes the following:

Movie S1