Structural and dynamics characterization of norovirus protease

Abstract

Norovirus protease is an essential enzyme for proteolytic maturation of norovirus nonstructural proteins and has been implicated as a potential target for antiviral drug development. Although X-ray structural studies of the protease give us wealth of structural information including interactions of the protease with its substrate and dimeric overall structure, the role of protein dynamics in the substrate recognition and the biological relevance of the protease dimer remain unclear. Here we determined the solution NMR structure of the 3C-like protease from Norwalk virus (NV 3CLpro), a prototype strain of norovirus, and analyzed its backbone dynamics and hydrodynamic behavior in solution. ¹⁵N spin relaxation and analytical ultracentrifugation analyses demonstrate that NV 3CLpro is predominantly a monomer in solution. Solution structure of NV 3CLpro shows significant structural variation in C-terminal domain compared with crystal structures and among lower energy structure ensembles. Also, ¹⁵N spin relaxation and Carr–Purcell–Meiboom–Gill (CPMG)-based relaxation dispersion analyses reveal the dynamic properties of residues in the C-terminal domain over a wide range of timescales. In particular, the long loop spanning residues T123–G133 show fast motion (ps-ns), and the residues in the bII–cII region forming the large hydrophobic pocket (S2 site) undergo conformational exchanges on slower timescales (μs–ms), suggesting their important role in substrate recognition.

Introduction

Norovirus protease plays a pivotal role in norovirus infection, a major cause of acute gastroenteritis outbreaks,^1–2 through the proteolytic cleavage of viral polyprotein (∼200 kDa) into six nonstructural proteins in their mature or intermediate forms (Supporting Information Fig. S1).^3–7 Norovirus protease, therefore, has been highlighted as a potential target for antiviral drug development. The protease was also reported to mediate the cleavage of poly(A)-binding protein, leading to translation inhibition in the viral-infected cells.⁸

Norovirus protease is a member of the viral cysteine protease family that possesses a chymotrypsin-like fold. The protease has a cysteine residue that acts as a nucleophile instead of a serine residue in chymotrypsin.⁹ The structures of Norwalk virus 3C-like protease (NV 3CLpro),^{10, 11} Chiba virus 3C-like protease (CV 3CLpro),¹² and Southampton virus 3C-like protease (SV 3CLpro) with a Michael acceptor peptidyl inhibitor¹³ has been determined as a dimeric form using X-ray crystallography. These viruses belong to the genotype I strain; proteases from this strain show nearly identical sequence homology (Supporting Information Fig. S1). The overall fold of the protease is built by N-terminal antiparallel β-sheet domain and C-terminal β-barrel domain. Catalytic residues C139, H30, and E54 along with an oxyanion hole are situated within a deep cleft between N- and C-terminal domains. The C-terminal domain is predominantly involved in the interaction with the substrate, in which the binding occurs in the manner of antiparallel β-sheet via hydrogen bond network between the substrate P5-P2 and protease K162-A158 residues.^{12, 13} The enzyme's S1 pocket encompasses T134 and H157, which directly interact with the substrate P1 Gln (or Glu) residue. These two residues are conserved in all norovirus proteases (Supporting Information Fig. S1) and have been shown to be critical to proteolytic activity.^{14, 15} The S2 site forms a large hydrophobic pocket consisting of I109, R112, and V114 and accommodates bulky hydrophobic P2 (Leu or Phe) residues found in the cleavage site in viral polyprotein (Supporting Information Fig. S1).

Crystallographic studies also have shed light on the possible role of protein conformational dynamics in the substrate recognition.^10–13 Catalytically important residues, C139 and T134, are positioned near the base of the highly disordered loop (L121−G140). This loop possibly contributes to the substrate interaction at the S′ site¹³ and has no visible electron densities in the NV 3CLpro crystal structure.¹⁰ Protease–substrate interactions were shown to be sensitive to the mutation of this region and substrate P2 residue. Mutation of P2 residue to more hydrophobic residue caused tighter binding, and the mutation of I109 to alanine resulted in a rapid release of the substrate.¹⁶ Despite the highly conserved amino acid sequence, the loop containing S2 site (bII−cII β-sheet region) of the norovirus protease crystal structures in the presence and absence of a substrate-based inhibitor showed a significant structural difference,^10–13 suggesting that the movement of the loop containing S2 site is possibly involved in the substrate entry and the product release. These dynamic properties, however, remain poorly characterized to date.

Dimer formation of the protease also was reported in crystallographic studies of NV 3CLpro^{10, 11} and SV 3CLpro.¹³ However, whether these proteases behave as dimers in solution remains unclear because the dimer interfaces observed in the crystallographic studies are quite different from each other despite their nearly identical amino acid sequences (Supporting Information Fig. S1).

In this study, we focus on the structural dynamics of NV 3CLpro to better understand its substrate recognition and catalytic function. Backbone ¹⁵N spin relaxation and CPMG relaxation dispersion analyses reveal the highly dynamic properties of C-terminal domain. The long disordered loop between cII and dII shows fast motion on pico- to nanosecond timescales, and the residues located in the bII−cII region undergo conformational exchange on slower timescales. Furthermore, we demonstrate from ¹⁵N spin relaxation and analytical ultracentrifugation analyses that NV 3CLpro is predominantly a monomer under the solution conditions analyzed here.

Results and Discussion

Solution structure determination of NV 3CLpro

We previously established complete backbone and side-chain chemical shift assignments and the chemical shift-based secondary structure prediction of NV 3CLpro under the same solution conditions used in this study.¹⁷ To solve the solution NMR structure, we analyzed ¹³C- and ¹⁵N-edited NOESY spectra that resulted in a total of 2135 nonredundant, inter-residue NOE restraints. A total of 198 backbone dihedral angle restraints were obtained by backbone chemical shift analysis and 82 hydrogen bond restraints were assigned by analyzing the ¹H-¹⁵N HSQC spectrum of ¹⁵N-labeled NV 3CLpro measured in 99% D₂O (Supporting Information Fig. S2). On the basis of these distance and dihedral angle constraints, the structural ensemble of NV 3CLpro was calculated using a simulated annealing procedure. Final structure calculation generated 100 structures, and we selected the 20 lowest energy structures for further analysis. The statistics of the final structural ensemble are given in Table I; we observed neither NOE distance violation greater than 5Å nor torsion angle violation greater than 5°, and 97.8% of the residues were found in the energetically most-favored and additionally allowed regions in a Ramachandran plot indicating good stereo-chemical quality of the calculated structures.

Table I.

Statistics for the final 20 solution structures of NV 3CLpro

Number of constraints
Total NOE distance constraints	2650
Intra-residue	515
Inter-residue	2135
Sequential (|i—j| = 1)	981
Medium range (1 < |i—j| < 5)	329
Long range (|i—j| ≥ 5)	825
Backbone dihedral angle constraints	198
ϕ	98
Ψ	100
Hydrogen bond restraints	82
Number of violations
Distance constraints (>0.5Å)	0
Dihedral angle constraints (>5°)	0
Ramachandran plot for the 20 lower energy structures
Residues in most favored region (%)	70.1
Residues in additionally allowed region (%)	27.7
Residues in generously allowed region (%)	1.9
Residues in disallowed region (%)	0.3
RMSD to mean structure
Backbone	1.5
Heavy atom	2.05
RMSD for structured region (1–173) without His tags and C-terminal flexible regions
Backbone	0.84
Heavy atom	1.14

The overall fold of NV 3CLpro solution structures agrees well with that of the crystal structures,^10–13 and consists of an incomplete β-barrel N-domain and an antiparallel β-barrel C-domain with active site residues (H30, E54, C139) at the cleft between two domains [Fig. 1(A)]. The backbone RMSD between the crystal structure (2FYQ)¹⁰ and our solution structure (2LNC) is 1.99Å [Supporting Information Fig. S3(A,B)], which is comparable to the RMSD value observed in systematic comparison of crystal and NMR structures ranging from 1.5Å to 2.5Å.¹⁸ A notable difference is an additional short β-strand structure (V72-E74) observed in the linker region connecting the N- and C-domains of the NV 3CLpro solution structure [Fig. 1(A)], which is consistent with our published chemical shift-based secondary structure prediction.¹⁷ The presence of the cross peak for backbone amide of E74 in the ¹H-¹⁵N HSQC spectrum of NV 3CLpro in 99% D₂O indicates the backbone amide hydrogen atom of E74 is protected by hydrogen bond formation (Supporting Information Fig. S2). Moreover, the proximity of the amide group of E74 to the carboxyl oxygen of V153 within β-strand eII corroborates the hydrogen bond formation in the very short β-strand. Because E74 is positioned within a 20 residue-long loop (eI–aII loop) connecting N-terminal and C-terminal domains, the observed hydrogen bond formation possibly contributes to the structural integrity of NV 3CLpro by stabilizing the eI−aII loop.

Figure 1.

Solution structure of NV 3CLpro (2LNC). (A) Ribbon representation of the lowest energy structure with catalytic residues (H30, E54, C139) shown as sticks. Secondary structure elements are labeled. Very short strand (V72-E74) is shown in green color. (B) Backbone overlays of the 20 lowest energy structures ensemble. α-helix is colored in red and β-strand in blue. A large structural variation was observed in bII–cII β-sheet region as well as N- and C-terminal segments.

Figure 1(B) shows the superposition of backbone atoms of the 20 lowest energy conformers for NV 3CLpro over residues 1−173. The backbone and heavy atom RMSD of these conformers excluding N- and C-terminal flexible regions are 0.84Å and 1.14Å, respectively. Although the N-terminal domain shows well-defined structures, a substantially larger displacement is observed in the C-terminal domain, especially in bII and cII strands (residues, 106−115) and the large loop region connecting cII and dII β-strands (residues, 122−137) (Supporting Information Fig. S3). No long-range NOE except inter-strands NOE between bII and cII strands is observed for these regions, suggesting the existence of substantial motion in these regions. This result agrees well with the published structures of norovirus proteases,^10–13 where the significant structural variability is found for these regions.

NV 3CLpro dynamics probed by ¹⁵N spin relaxation

Protein backbone and side-chain motions can be probed by NMR relaxation measurements, because these motions affect the rate at which an excited nucleus experiences the fluctuating magnetic filed to exchange energy and relaxes to its equilibrium state. We measured ¹⁵N-longitudinal relaxation rate (R₁), ¹⁵N-transverse relaxation rate (R₂), and heteronuclear nuclear Overhauser effect (heteronuclear NOE) to characterize fast time-scale dynamics of NV 3CLpro backbone. A total of 148 resonances are used in the spin relaxation analysis. Residues that are missing or show significant broadening include T4, S14, F18, I26, F39, G53, E54, G116, and S163. The pairs of L5/V24, W6/A52, F25/A158, and L73/C169 fully or partially overlapped. The N126 side-chain amide cross peak interfered with the precise intensity measurement of G76.

¹⁵N relaxation data summarized in Figure 2 reveal the highly flexible nature of NV 3CLpro. Significantly lower values of heteronuclear NOE [Fig. 2(A)] were observed for C-terminal residues and the long loop region between cII and dII spanning residues T123 to G133, indicating that these structural components exhibit extensive fast motion on sub-nanosecond timescales. The existence of the fast motion was also corroborated by a substantial decrease in R₂ (∼8.24 and 14.5 s⁻¹ for 500 and 800 MHz, respectively), an increase in R₁ value (∼1.92 and 2.76 s⁻¹ for 500 and 800 MHz, respectively), and lower R₂/R₁ ratios (Fig. 2). These results are also consistent with the lack of long-range NOE and the structural variation of the long loop region observed in the solution structure [Fig. 1(B)]. Structural analysis of SVpro with a peptide mimetic inhibitor suggests that the long loop of T123-D131 may form a putative binding site of a substrate P' side.¹³ Fast motion of the long loop region observed here likely contributes to accommodate the P' side of substrate in the protease molecule and release it after proteolytic cleavage. In addition, K162, G164, and N165 located on the loop between eII and fII strands showed lower heteronuclear NOE and higher R₁ values, suggesting the fast backbone dynamics of this region.

Figure 2.

¹⁵N-spin relaxation parameters of NV 3CLpro measured at 500 MHz and 800 MHz: (A) Steady-state heteronuclear NOE, (B) R₂, (C) R₁, and (D) R₂/R₁ values. R₂/R₁ values were calculated for spin relaxation data obtained at 500 MHz.

Residues V114 and H115 in the bII−cII region that form a hydrophobic pocket for the recognition of P2 residue¹² showed strikingly higher R₂ values and higher R₂/R₁ ratios (Fig. 2) The heteronuclear NOE data, a sensitive probe for fast motion, showed no noticeable changes in this region compared with the rest of the protein. The relaxation behavior of the bII−cII region is most probably because of conformational exchange on the micro- to millisecond timescale.^{19, 20} Because exchange contributions scale quadratically with the external field strength in principal, the substantial increase in R₂ values of the 800 MHz dataset observed for V114 and H115 further corroborates the conformational exchange on the μs–ms timescale in this region. This conformational dynamics of the bII−cII region may contribute to the formation of a large hydrophobic pocket that accommodate a bulky P2 side chain and thus cover a peptide portion of substrate as suggested by crystallographic studies.^{12, 13}

The relaxation parameters, R₁, R₂ and hetero-nuclear NOE values, were then subjected to a Lipari−Szabo model-free analysis.^{21, 22} The model-free analysis was conducted with the program Modelfree.²³ The amide bond length was fixed at 1.02Å, and the ¹⁵N chemical-shift anisotropy value of –160 ppm was used in the calculations. On the basis of 3D structural coordinates of NV 3CLpro and the selected set of ¹⁵N relaxation data (as described in the Experimental section), the diffusion tensor of NV 3CLpro was best described by an isotropic tumbling model. Of the 148 residues subjected to the analysis, 24 residues could not be fitted with any model (Supporting Information Fig. S4). Figure 3 shows the order parameters (S²), the internal correlation times (τ_e), and the chemical exchange contribution (R_ex) for NV 3CLpro. Except for terminal residues, the long loop between cII and dII spanning residues 123−133 shows the largest local motional freedom. The average S² value for this region is 0.46, which is significantly lower than the overall average S² value of the protein, 0.86 [Fig. 3(A)]. Most residues in the loop can be best fitted to an extended model-free formalism, model 5, in which the fast internal dynamics are characterized by the squared generalized order parameter, , and the relatively slow internal dynamics is characterized by the squared generalized order parameter, , and the effective correlation time τ_e (>500 ps) [Fig. 3(B) and Supporting Information Fig. S4].²⁴ In general, uncertainty of τ_e values decreases when S² has lower value^{21, 25}; therefore, τ_e values ranging from 1.3 to 2.0 ns for this flexible loop prove the existence of extensive backbone motion on the pico- to nanosecond timescale [Fig. 3(B)]. The R_ex terms characterizing motions in the micro- to millisecond time regime are plotted in Figure 3C. The R_ex value is directly proportional to the difference in chemical shift values between the exchanging conformations. Significantly larger R_ex values were observed for V114 and H115 positioned within the bII−cII loop, suggesting the presence of μs–ms time-scale motions.

Figure 3.

Model-free parameters. (A) S², (B) τ_e, and (C) R_ex of NV 3CLpro backbone amides, derived from a model-free analysis of the relaxation data at 500 MHz presented in Figure 2.

NV 3CLpro dynamics probed by ¹⁵N relaxation dispersion

For direct characterization of the μs–ms conformational dynamics, we performed the CPMG relaxation dispersion experiment on ¹⁵N-labeled NV 3CLpro at 25°C. Among the backbone amide resonances with good signal-to-noise ratios, a total of eight resonances (C83, L97, R100, V114, H115, G124, A159, and T166) show dispersion with Δ≥ 2 s⁻¹, where Δis the difference in measured values at low (ν_cp = 40 Hz) and high (ν_cp = 800 or 920 Hz) CPMG frequencies. Representative relaxation dispersion curves are plotted as a function of CPMG field strength, ν_cp in Figures 4 and S5. Consistent with ¹⁵N-spin relaxation measurements (R₁, R₂, and hetero-nuclear NOE) and model-free analyses, V114 and H115 at the bII−cII loop region show considerable relaxation dispersion (Fig. 4), proving the μs–ms conformational dynamics in this region that may be involved in the formation of a large hydrophobic pocket for substrate recognition.

Figure 4.

¹⁵N relaxation dispersion profiles for NV 3CLpro. Representative ¹⁵N relaxation dispersion curves for residues located at bII–cII loop region, V114 and H115. The experiments were conducted at 25°C at a static magnetic field strength of 11.7 T. Error bars were estimated by peak intensity analysis of duplicate experiments for specific ν_cp values.

NV 3CLpro is a monomer in solution

The oligomeric state of NV 3CLpro was evaluated by ¹⁵N relaxation and analytical ultracentrifugation experiments. As previously described, the diffusion tensor of NV 3CLpro was best fitted with an isotropic tumbling model. Analysis with an anisotropic tumbling model produced the axially symmetric model with D_||/D⊥ = 1.12 (D_|| = 1.28 × 10⁷ s⁻¹, D⊥ = 1.20 × 10⁷ s⁻¹) and the full asymmetry model with D_x = 1.19 × 10⁷ s⁻¹, D_y = 1.20 × 10⁷ s⁻¹, D_Z = 1.28 × 10⁷ s⁻¹, but no significant improvement was observed over the isotropic tumbling model according to an F-test. Isotropic tumbling model analysis yielded the overall correlation time, τ_m, of 13.02 ns, which is reasonably close to the theoretical value of NV 3CLpro monomer, 10.8 ns. We performed further sedimentation equilibrium and velocity analyses by analytical ultracentrifugation to test the monomeric state of NV 3CLpro directly (Fig. 5). First, sedimentation equilibrium of NV 3CLpro was analyzed at dilute conditions (0.015, 0.03, and 0.06 mM) and at three rotor speeds [Fig. 5(A)]. Global analysis of the data gave a reasonable fit for a monomer model with the molecular weight of 23 kDa, which was slightly higher than that of NV 3CLpro monomer, 20 kDa. Subsequent analysis of sedimentation velocity at three different concentrations of NV 3CLpro (30, 200, and 600 μM) indicated that NV 3CLpro is in a monomeric state without larger oligomeric species [Fig. 5(B)]. The selectivity curve analysis of size exclusion chromatography data also shows M_w is around 25 kDa (data not shown). We also confirmed from ¹H-¹⁵N HSQC measurements of NV 3CLpro with different protein concentrations (from 0.2 to 0.8 mM) that protein concentration-dependent chemical shift changes were negligible. On the basis of these biophysical data, we conclude that NV 3CLpro is predominantly a monomer in solution without any concentration dependence. The enzymatic activity of NV 3CLpro used here was confirmed in our previous study by using two synthesized fluorogenic substrates derived from P7-P7′ and P5-P2′ residues on the NS2/3 cleavage site in ORF1 of Norwalk virus (Supporting Information Fig. S1).²⁶

Figure 5.

Sedimentation equilibrium profiles for NV 3CLpro and sedimentation velocity profiles for different concentrations of NV 3CLpro. Panel (A) shows selected sedimentation equilibrium profiles. The global fit (red line) is shown for sedimentation of 0.015 mM NV 3CLpro at the three higher speeds (18,000, 22,000, and 26,000 rpm). Panel (B) shows g(s*) profiles of the indicated NV 3CLpro concentrations at 49,000 rpm.

Contrary to our results, previous X-ray studies have shown that the protease forms dimer in the crystal structures.^{10, 11, 13} However, significantly different dimer interfaces were observed among these X-ray structures of norovirus proteases despite highly conserved amino acid sequences among norovirus proteases (Supporting Information Fig. S1).^{10, 11, 13} It should be noted that the same preparation of NV 3CLpro was used for X-ray crystallography¹¹ (3UR6) and NMR studies (2LNC). It is unlikely that the biologically relevant dimer would have different interface contacts as observed in the crystal structures. Our ¹⁵N relaxation studies reveal fast timescale (ns-ps) dynamics of the long loop region (T123-G133) and slow time-scale (μs–ms) motion in the bII–cII region. From available crystal structures, these structural components are buried in the dimer interfaces.^{10, 13} Because the bII–cII region plays a critical role in the substrate recognition by forming an arch (large S2 hydrophobic pocket) that encompasses the P2 residue on the large viral polyprotein, the restriction of conformational dynamics because of the dimer interface formation would have had a negative impact on the substrate binding and release.

NV 3CLpro dynamical properties and functional implication

Here, we characterize NV 3CLpro solution structure and backbone dynamics using NMR spectroscopy to gain insight into NV 3CLpro dynamical properties that have not been addressed in previous crystallographic studies.^10–13 Our data reveal that the C-terminal domain of NV 3CLpro is highly dynamic. The backbone amide groups with lower order parameter, S², are mapped onto the NV 3CLpro solution structure [Fig. 6(A)]. ¹⁵N nuclear spin relaxation studies of NV 3CLpro reveal that the backbone of the long loop (T123−G133) is extensively flexible on the sub-nanosecond timescales, as indicated by lower values of heteronuclear NOE, lower R₂ values (Fig. 2), and lower values of S² [Figs. 4(A) and 6(A)]. These observations agree with the lack of electron density for this region in the crystal structure of NV 3CLpro.¹⁰ The bII–cII region, on the contrary, is characterized by motions on a slow timescale, as reflected by higher R₂ values and CPMG relaxation dispersion of V114 and H115 [Figs. 2, 5, and 6(B)]. A recent Ala scanning mutagenesis study of CV 3CLpro also showed that most of the residues critical to the activity and stability of the protease reside in the C-terminal domain.¹⁴ These observations suggest that the dynamical properties of the C-terminal region presented in this study are likely important for NV 3CLpro function.

Figure 6.

Backbone dynamics of NV 3CLpro revealed by ¹⁵N spin relaxation and relaxation dispersion experiments. (A) Residues with lower backbone amide S² values are mapped on the solution structure of NV 3CLpro. The mean value of S² for NV 3CLpro is 0.86 (<S²>=0.86). The standard deviation, σ, is 0.22. Ribbon color is scaled as follows: red, S² < 0.64 (<S²>−σ), pink, 0.64 < S² < 0.86. (B) Residues undergoing conformational exchange (Δ≥ 2 s⁻¹) are plotted as blue sphere. Catalytic residues, C139, H30, and E54, are also shown as sticks.

T134 and H157 are key residues for the recognition of the substrate P1 residues (Gln or Glu) in the cleavage site on the polyprotein; the side-chain of P1 Gln (or Glu) is sandwiched between T134 in the long loop and H157 in eII β-strand, and forms hydrogen bond with these two residues.¹³ Because the S1 site forms a relatively large pocket for the accommodation of the large P1 side chain,^{12, 13} the fast dynamics of the large loop (T123 to G133) adjacent to T134 with a higher degree of motional freedom (lower S² values are shown in Fig. 3) most likely contribute to an optimized conformational contact with the large P1 side chains.

It is now widely recognized that protein dynamics on timescales of microseconds or slower play a pivotal role in biological processes such as molecular recognition, enzymatic catalysis, and allosteric regulation.^27–29 Crystal structure of SV 3CLpro complexed with a peptide-mimetic inhibitor showed that the inhibitor interacts with bII–cII β-sheet region and the hydrophobic pocket formed by I109, Q110, R112, and V114 is responsible for the recognition of bulky hydrophobic P2 residues such as Leu and Phe in the norovirus polyprotein sequence (Supporting Information Fig. S1).¹³ These studies also suggested that, in conjunction with K162 between eII and fII β-strands, the side chain of Q110 serves as a clamp for P4 residues^{12, 13}; therefore, the slow conformational exchange in the bII–cII β-sheet region and the fast dynamics of the loop containing K162, as revealed by ¹⁵N spin relaxation and relaxation dispersion analyses, may play a significant role in the substrate recognition and subsequent release of the cleaved products. Comparison of the dynamics between the apo- and holo-enzyme using a substrate/inhibitor, however, would be required to provide a clear understanding of the dynamical properties of NV 3CLpro associated with its wider substrate specificity and catalytic function. Because the long disordered loop and the bII–cII β-strand region are structurally connected, it will be particularly intriguing to study how these dynamics are fine-tuned in response to the substrate binding.

Understanding the structural dynamics of NV 3CLpro will be useful in elucidating the mechanism of drug resistance and lead to successful drug discovery and development. Currently, we are studying the comparative analysis of the NMR structures and X-ray structure in complex with GC376, a potent inhibitor we recently reported.¹¹ Also, we plan to perform dynamics measurements of NV 3CLpro with the inhibitor and mutant enzymes identified during the selection of drug-resistant viruses.

Materials and Methods

Protein expression and purification

Uniformly ¹³C and/or ¹⁵N-labeled NV 3CLpro were expressed in Escherichia coli BL21 cells (Invitrogen) with a stretch of N-terminal six histidine tags as previously described.¹⁶ M9 growth media supplemented with 1 g/L ¹⁵NH₄Cl, and 2 g/L ¹³C-d-glucose (Cambridge Isotope Laboratories) was used to grow the cells. Protein purification was performed with Ni-NTA affinity chromatography (QIAGEN) followed by further purification using size exclusion chromatography on a Superdex 75 (GE healthcare).

NMR spectroscopy

NMR spectra were recorded at 25°C on Varian VNMRS 500 MHz and Bruker Avance 800 MHz NMR spectrometers equipped with 5 mm triple-resonance cryogenic probes and z-axis pulse field gradients. Complete backbone and side-chain chemical shift assignments and the chemical shift-based secondary structure prediction were achieved previously.¹⁷ NOE distance restraints were obtained from ¹⁵N-NOESY-HSQC (100 ms mixing time) and ¹³C-NOESY-HSQC (120 ms mixing time) spectra were collected on a 0.8 mM¹³C/¹⁵N-NV 3CLpro prepared in 50 mM sodium phosphate buffer, pH 6.5, with 100 mM NaCl and 3 mM NaN₃ in 90% H₂O/10% D₂O. The NOESY experiments were performed on a Bruker Avance 800 MHz spectrometer. ¹H–¹⁵N HSQC spectrum was also collected on ¹⁵N-NV 3CLpro in 99% D₂O to identify residues involved in hydrogen bond. All the spectra were processed with NMRpipe.³⁰ The programs CARA³¹ and Sparky³² were used to analyze the datasets.

Structure calculations

A total of 2135 distance constraints (981 sequential, 329 medium range, 825 long range) were derived from the NOESY spectra. All lower limits of distance constraint were set to 1.8Å and a distance of 0.6Å was added to the upper distance bound of NOE constraint involving methyl groups. Backbone torsion angle restraints were obtained by chemical shift analysis using TALOS+ program.³³ Structure calculations were performed using the simulated annealing protocol in the CNS program.^{34, 35} After initial structure generation, hydrogen bond restraints (d_NH–O = 2.0–2.8Å, d_N–O = 2.7–3.6Å) were assigned based on the proximities of donors and acceptors in the initially generated structures and resonance peaks observed in the ¹H–¹⁵N HSQC spectrum of ¹⁵N-NV 3CLpro collected approximately 12 h after the solvent exchange to 99% D₂O. Hydrogen bond restraints together with distance and torsion angle restraints were used for subsequent structure calculations. In the final structural calculation (4000 steps for the high-temperature annealing stage, 8000 steps for the first cooling stage using torsion angle dynamics, and 8000 steps for the second cooling stage using Cartesian dynamics), an ensemble of 100 structures was generated, and the 20 lowest energy structures were selected for further analysis. Assessment of the structural quality and calculation of coordinate RMSD were carried out using PROCHECK³⁶ and MOLMOL,³⁷ respectively. The statistics of NMR structure determination are summarized in Table I.

15N-relaxation experiments

¹⁵N spin relaxation measurements were carried out on ¹⁵N-NV 3CLpro using established pulse sequences.^38–40 The data were collected at 25°C on both Varian VNMRS 500 MHz and Bruker Avance 800 MHz spectrometers equipped with cryogenic probes. The ¹H and ¹⁵N carrier frequencies were set to the water resonance and 117 ppm, respectively. For the longitudinal relaxation rate, R₁, and transverse relaxation rate, R₂, measurements, NMR spectra at 500 MHz were acquired with a spectral width of 6510 Hz (13 ppm) over 1024 complex points in the ¹H dimension and 1925 Hz (38 ppm) over 256 complex points in the ¹⁵N dimension with 24 transients. NMR Spectra at 800 MHz were acquired with a spectral width of 11160 Hz over 1024 complex points in the ¹H dimension and 3081 Hz over 256 complex points in the ¹⁵N dimension with 28 transients. Delays of 10, 50, 100, 250, 500, 750, 1000, and 1500 ms for the 500 MHz dataset and of 10, 70, 150, 250, 400, 750, and 1250 ms for the 800 MHz were arrayed for R₁ measurements. R₂ values were measured using relaxation delays of 10, 30, 50, 70, 110, 150, 190, and 250 ms for the 500 MHz dataset and 20, 50, 70, 90, and 130 ms for the 800 MHz. The recycle delay between transients was set to 2.5 s for both R₁ and R₂ measurements. Heteronuclear steady-state NOE values were determined by the ratio of cross-peak intensities of spectra recorded in the presence and in the absence of 3 s of ¹H saturation. Twenty-four and 32 transients were collected for the 500 and 800 MHz datasets, respectively. All spectra were processed using NMRPipe,³⁰ and the program Sparky³² was used to measure peak intensities and to determine R₁ and R₂ values by a nonlinear least square fitting. Errors in R₁ and R₂ were obtained from the fits by Sparky. The uncertainty of NOE values was estimated by comparison of the replicate NOE spectra.

Relaxation data, R₁, R₂, and ¹⁵N–¹H NOE values obtained at 500 MHz were used for model-free analysis. Initial estimation of the overall rotational diffusion tensor was performed using the program TENSOR2.⁴¹ The following criteria were used to select residues for tensor optimization; residues within the defined secondary structures were selected. Residues having very fast internal motion (lower NOE values, which is less than 0.7, and lower R₂/R₁ ratio) and residues experiencing conformational exchange (higher R₂/R₁ ratio) on μs–ms were then removed from input for TENSOR2. The resulting residues included 9–11, 16, 17, 19, 21, 47, 49, 81–84, 86, 98, 102, 104-107, 109, 140, 141, 146, 147, 150, 151, 153, 155, 156, and 166–168. The isotropic rotational correlation time, τ_m, was also calculated by the Stokes–Einstein–Debye relation⁴²:

(1)

in which η is the viscosity of the solvent, r is the effective hydrodynamic radius of the protein, k is the Boltzmann constant, and T is the temperature. Protein hydrodynamic radius, r, was calculated as follows:

(2)

where ν is the specific volume of the protein (0.73 cm³/g), M_w is the molecular weight, N_A is Avogadro's number, and r_w is a hydration layer surrounding the protein, which is assumed to be 3.0Å.⁴² Model-free analysis^{21, 22} of relaxation data was carried out using the Modelfree 4.0 software²³ and TENSOR2,⁴¹ in which ¹H–¹⁵N bond distance and ¹⁵N CSA were set to 1.02Å and −160, respectively. In the Modelfree analysis, the following five combinations of model-free parameters were fitted to the relaxation data with the initial estimate of rotational correlation time²²: (1) S², (2) S², and τ_e, (3) S² and R_ex, (4) S², τ_e, and R_ex, and (5) , τ_e, and .

15N-relaxation dispersion experiment

Relaxation dispersion profiles for the backbone ¹⁵N nuclei of NV 3CLpro were recorded using relaxation-compensated constant-time CPMG pulse sequence^{43, 44} on a Varian VNMRS 500 MHz spectrometer at 25°C. Dispersion data were collected with a 50 ms constant time relaxation delay, T_cp, and CPMG field strength, ν_cp, between 40 and 920 Hz. Data collection was repeated for one ν_cp to calculate the standard deviation. The effective transverse relaxation rate, , comprising the transverse relaxation rate in the absence of exchange, R₂, and the component because of chemical exchange, R_ex, was calculated via (ν_cp) = (−1/T_cp) ln[I(ν_cp)/I₀], where I(ν_cp) is the peak intensity at a given CPMG field strength and I₀ is the peak intensity without the CPMG relaxation segment. Relaxation dispersion curves were obtained by plotting as a function of ν_cp.

Analytical ultracentrifugation experiments

Sedimentation velocity experiments were conducted using an Optima XL-I ultracentrifuge using the An-60 Ti rotor at 20°C in 20 mM sodium phosphate buffer (pH 6.5) including 0.1M NaCl. Sedimentation was monitored by absorbance using double sector cells with a final loading of 400 μL per cell. Sedimentation was at 49,000 rpm with scans made at 5-min intervals. To evaluate the change in oligomeric state of NV 3CLpro, we evaluated the changes in sedimentation velocity patterns at 30, 200, and 600 μM NV 3CLpro. Sedimentation was monitored at 280 nm for the lower level (30 μM), at 298 nm for the middle level (200 μM), and at 305 nm for the higher level (600 μM). Sedimentation data were analyzed using DCDT+ software version 1.16, and sedimentation coefficients were calculated by using g(s*) fitting function in DCDT+ software. Buffer density and viscosity were calculated using Sednterp version 1.08 (http://www.jphilo.mailway.com [Please confirm that the URL is current and active.]). The partial specific volumes of proteins were calculated from their amino acid composition using Sednterp (0.7345 mL/g at 20°C).

Sedimentation equilibrium studies were also conducted on NV 3CLpro at concentrations of 15, 30, and 60 μM in sodium phosphate buffer. Using a charcoal-filled Epon centerpiece with 6-sector cells, 60 μL samples were overlaid over 30 μL of Fluorinert® FC-43 silicon oil (3M Industrial Chemical Products Division), and 150 μL of matched solvent was introduced in the reference positions. Equilibrium at 4°C was attained at 10,000, 14,000, 18,000, 22,000, and 26,000 rpm. Following equilibration at the lowest speed during a 30-h period, at least an 8-h equilibration time was used for the transition to a higher speed. Sedimentation equilibrium data were globally fit to monomer-only or monomer-dimer models using Beckman software (version 4.0) that was provided with the Optima XL-I ultracentrifuge.

Databank accession numbers

Structural coordinates and experimental constraints have been deposited in the Protein Data Bank with accession code 2LNC.

Glossary

Abbreviations

AUC

Analytical ultracentrifugation

CPMG

Carr–Purcell–Meiboom–Gill

CV 3CLpro

Chiba virus 3C-like protease

HSQC

Heteronuclear single quantum coherence

NOE

Nuclear Overhauser effect

NV 3CLpro

Norwalk virus 3C-like protease

RMSD

Root mean square deviation

SV 3CLpro

Southampton virus 3C-like protease

Supplementary material

Additional Supporting Information may be found in the online version of this article.