|Home | About | Journals | Submit | Contact Us | Français|
The protease domain within the RUBV (rubella virus) NS (non-structural) replicase proteins functions in the self-cleavage of the polyprotein precursor into the two mature proteins which form the replication complex. This domain has previously been shown to require both zinc and calcium ions for optimal activity. In the present study we carried out metal-binding and conformational experiments on a purified cysteine-rich minidomain of the RUBV NS protease containing the putative Zn2+-binding ligands. This minidomain bound to Zn2+ with a stoichiometry of ≈0.7 and an apparent dissociation constant of < 500 nM. Fluorescence quenching and 8-anilinonaphthalene-1-sulfonic acid fluorescence methods revealed that Zn2+ binding resulted in conformational changes characterized by shielding of hydrophobic regions from the solvent. Mutational analyses using the minidomain identified residues Cys1175, Cys1178, Cys1225 and Cys1227 were required for the binding of Zn2+. Corresponding mutational analyses using a RUBV replicon confirmed that these residues were necessary for both proteolytic activity of the NS protease and viability. The present study demonstrates that the CXXC(X)48CXC Zn2+-binding motif in the RUBV NS protease is critical for maintaining the structural integrity of the protease domain and essential for proteolysis and virus replication.
RUBV (rubella virus), an important human pathogen causing German measles, is a simple, plus-strand RNA virus that is the sole member of the genus Rubivirus in the Togaviridae family of animal viruses. The RUBV genome encodes five proteins which are expressed from two ORFs (open reading frames): the two proteins involved in viral RNA replication that are translated from one [termed the ‘NS (non-structural) protein’ or ‘NS-ORF’] and the three structural proteins that form the virus particle from the second. Proteolysis is a common feature in expression of the proteins from both ORFs, as the primary translation product of the NS-ORF is cleaved into its two component proteins by a protease embedded in one of these proteins, whereas the structural-protein ORF translation product is processed co-translationally by the cellular enzyme signalase. The two replicase proteins, P150 and P90, are situated at the N- and C-terminal ends of the NS-ORF precursor respectively, and the embedded NS protease is at the C-terminus of P150 immediately upstream from the cleavage site. Other than cleaving the NS-ORF precursor, the NS protease has no known cellular targets .
Previous studies in our laboratories have shown that the RUBV NS protease is a papain-like protease with a catalytic dyad comprising Cys1152 and His1273 (P150 has 1301 residues) . We recently showed that the NS protease contains a linear EF-hand Ca2+-binding motif that plays a structural role in stabilizing the protease at physiological temperatures . To our knowledge, this is the only example of Ca2+ playing such a role in a viral protease. We also showed that Zn2+ is required for optimal proteolytic activity [4,5]. However, the exact Zn2+-binding ligands and the role Zn2+ plays in protease activity have yet to be identified. On the basis of our model structure of the RUBV NS protease, a cluster of cysteine residues (Cys1167, Cys1175, Cys1178, Cys1225 and Cys1227) are arranged in close proximity in space and could potentially form a Zn2+-binding pocket. Structural Zn2+-binding sites are generally co-ordinated by four cysteine residues and histidine residues as ligands, although there are no histidine residues nearby in this cysteine cluster. In the present study we expressed and purified a minimal metal-binding domain from the NS protease, RUBCa (comprising amino acids 1143–1252 of P150), which contains the EF-hand Ca2+-binding loop as well as all of the putative cysteine residues involved in Zn2+ binding. We used the well-folded RUBCa minidomain to study Zn2+ binding and the resulting changes in minidomain structure with various spectroscopic methods and, furthermore, took a site-directed-mutagenesis approach to identify the ligands involved in Zn2+ binding. Mutants were also introduced into a RUBV infectious cDNA replicon to correlate Zn2+ binding with NS protease activity and viability.
RUBCa, the minidomain containing the Zn2+- and Ca2+-binding motifs of the RUBV NS protease (amino acids 1143–1252 of P150) was expressed using the pGEX-2T expression vector . A series of cysteine-to-serine substitution mutations were created in this vector. These mutations were also introduced into the RUBV replicon RUBrep-HA/GFP (where HA is the haemagglutinin epitope and GFP is green fluorescent protein) . Transfection of Vero cells with replicon transcripts, determination of viability and assay of in vivo NS protease activity were performed as previously described .
The RUBCa minidomain and its cysteine-to-serine mutants were expressed as a GST-fusion protein in Escherichia coli BL21(DE3) as previously described . The proteins were purified by following the protocols for GST-fusion protein purification  using glutathione–Sepharose 4B beads (GE Healthcare). The GST tag was subsequently cleaved with thrombin, followed by further purification by gel-filtration (Superdex 75) and cation-exchange chromatography (Hitrap SP; GE Healthcare). The molecular mass of RUBCa was confirmed by MALDI–TOF-MS (matrix-assisted laser-desorption ionization–time-of-flight MS) (see Supplementary Figure 1 at http://www.BiochemJ.org/bj/417/bj4170477add.htm) in the Advanced Biotechnology Core Facilities of Georgia State University. The concentration of RUBCa was measured by its A280 using an absorption coefficient of 19 630 M−1 · cm−1 calculated as described previously . Metal elements in the purified protein were screened by ICPMS (inductively coupled plasma MS) at the Chemical Analysis Laboratory at the University of Georgia.
The PAR colorimetric assay, with slight modification, was used to determine the amount of protein-bound Zn2+ as previously described . Protein samples were extensively dialysed against Chelex-100-(Bio-Rad)-pretreated water to remove background metal ions and then digested in a total volume of 100 μl with 100 μg/ml protease K (Sigma) in Chelex-100-treated HSD buffer (50 mM Hepes/KOH, 200 mM NaCl and 5 mM dithiothreitol), pH 7.0, at 56°C for 30 min. Subsequently, an identical volume of HSD containing 5 mM iodoacetamide and 200 μM PAR (Sigma) was added. Absorbance over the range 300–600 nm was measured using an RF-1501 UV spectrometer (Shimadzu). HSD containing 0–15 μM standard ZnCl2 solution was used to create a standard curve (see Supplementary Figure 2 at http://www.BiochemJ.org/bj/417/bj4170477add.htm). Carbonic anhydrase, which contains a single Zn2+-binding site, was used as positive control.
Fluorescence emission spectra were measured on a PTI lifetime fluorimeter at 25°C. For intrinsic tryptophan fluorescence, spectrum measurements were carried out at protein concentrations of 2–5 μM in 10mM Tris, 100mM KCl and 5mM DTT (dithiothreitol), pH 7.4, with slit widths of 4 and 8 nm for excitation and emission respectively. The emission spectra were collected from 300 to 400 nm with an excitation wavelength of 282 nm. ANS (8-anilinonaphthalene-1-sulfonic acid) fluorescence emission spectra were recorded from 400 to 600 nm, with an excitation wavelength of 390 nm. Protein samples (2 μM) were added to the solution containing 40 μM ANS, 10 mM Tris, 100 mM KCl and 5 mM DTT at pH 7.4. For metal titration, aliquots of 5–10 μl of zinc stock solution (20–80 μM) were gradually added into the sample solution. The apparent Zn2+ binding constant was obtained by fitting the acquired data to the following equation:
where f is the fractional change of fluorescence intensity, Kd is the dissociation constant for zinc, and [P]T and [M]T are the total concentrations of protein and Zn2+ respectively.
Tryptophan fluorescence quenching experiments were carried out at 25°C by adding 10–20 μl aliquots of 4 M acrylamide to the sample solution containing 2 μM protein. The excitation wavelength was set at 282 nm and the emission spectra were acquired from 300 to 400 nm. The integrated area of fluorescence intensity between 300 and 400 nm was used for data analysis by fitting to a modified Stern–Volmer equation:
where KSV is the dynamic or collisional quenching constant, V is the static quenching constant, [Q] is the concentration of added acrylamide, and F0 and F are the integrated fluorescence intensity in the absence and presence of acrylamide respectively.
The homology modelling of the RUB NS protease domain was constructed using the comparative structure modelling programs SWISSMODEL and MODELLER . The leader protease of FMDV (foot-and-mouth-disease virus; PDB code: 1qmy), which shares over 15% sequence similarity and a highly similar secondary-structural arrangement (see Supplementary Figure 3 at http://www.BiochemJ.org/bj/417/bj4170477add.htm), serves as the best available template for structure modelling [11,12]. Energy minimization was subsequently performed using SYBYL. The secondary-structure prediction was carried out by using the program PredictProtein . The metal-binding site was further modified on the basis of the Zn2+ binding sites in the proteinase with four cysteine liganding residues such as found in SARS-CoV (severe-acute-respiratory-syndrome-associated coronavirus)  by fixing the backbone atoms and allowing free rotation of sulfur atoms on the side chain using the software DeepView, and energy minimization was subsequently carried out using SYBYL.
Our previous studies demonstrated the requirement of Zn2+ for proteolytic activity of the RUBV NS protease in vitro and further narrowed down the potential Zn2+-binding sequences to within residues 1143–1301 of the P150 replicase protein (Figure 1) [4,5]. On the basis of its primary- and secondary-structure similarity (see Supplementary Figure 3) to the well-characterized leader protease of FMDV (PDB code 1qmy), we built a model structure for the RUBV NS protease by homology modelling . This model proved accurate in our previous study on the linear EF-hand Ca2+-binding domain (residues 1197–1225) within RUBV NS protease . As shown in Figure 1(B), Cys1152 and His1273 form the active site of this papain-like protease. A cluster of cysteine residues (Cys1167, Cys1175, Cys1178, Cys1225 and Cys1227), which is conserved in all genotypes of RUBV  and located ≈20 Å (1 Å = 0.1 nm) away from the active site, was found to be arranged in close proximity in space and could potentially form a Zn2+-binding pocket (Figure 1B). Binding of Zn2+ was studied using the RUBCa minidomain spanning residues 1143–1252 of P150, including the putative Zn2+-binding pocket; use of a larger polypeptide proved impossible, because of solubility problems .
Using MBP (maltose-binding protein) fusions, we previously demonstrated the binding of Zn2+ ions to various sub-regions of the NS protease domain with a radioactive-65Zn2+ blotting assay . Although MBP itself did not bind Zn2+, we were not able to rule out the possibility that an artificial Zn2+ binding pocket might be formed by residues from both the MBP tag and the RUBV NS protease domain. In addition, the 65Zn2+ blotting does not provide quantitative estimation of Zn2+ binding constants. With the tagless RUBCa minidomain in hand, we were able to confirm that the protease domain itself bound Zn2+. The purified minidomain was first subjected to metal-element screening by using ICP-MS, which revealed that it bound 0.6 mol of zinc and 0.7 mol of calcium ion per mol of protein simultaneously. The amount of minidomain-bound Zn2+ was further confirmed with an independent colorimetric assay. In this latter assay, after purifying the RUBCa minidomain and removing background metal ions with Chelex 100 resin, the Zn2+ content was determined with the metal indicator PAR, which detects protein-bound Zn2+ released by protease K digestion. The released Zn2+ forms a Zn–(PAR)2 complex accompanied by a large increase in intensity at 500 nm (Δε=6.6×104 M−1 · cm−1 at pH 7.0) within 3 ms  (see Supplementary Figure 2). The amount of RUBCa-bound Zn2+ ions ([Zn2+]/[P]) was found to be 0.69±0.03. Thus the bacterially expressed cysteine-rich RUBCa minidomain from the RUB NS protease was capable of binding Zn2+ ions by itself.
We next asked whether the binding of Zn2+ ions would lead to induction of conformational changes using fluorescence spectroscopy. The RUBCa minidomain contains three aromatic tryptophan residues, one of which, Trp1231, is in close proximity (≈ 5 Å) to the proposed cysteine-rich zinc-binding pocket (Figure 1B). In comparison with fully exposed tryptophan (as seen in L-tryptophan solution) or highly buried tryptophan residues [as seen in domain 1 of the cell-adhesion molecule CD2 (cluster of differentiation 2)], the tryptophan residues in RUBCa were relatively well protected (Figure 2 and Table 1), indicating the structural integrity of the minidomain. Since aromatic residues are highly sensitive to local or global conformational changes, we took advantage of this feature to probe Zn2+-induced conformational changes by monitoring intrinsic tryptophan fluorescence. As shown in Figure 3(A), with the gradual addition of Zn2+ ions to the minidomain, the emission maximum blue-shifted from 339 nm to 335 nm with a concomitant increase of fluorescence intensity by at least 15%, indicating that Zn2+ ions caused conformational changes and led to the shielding of aromatic residues from the solvent.
To gain quantitative information on the extent of Zn2+-induced conformational changes, we then carried out an experiment to measure fluorescence quenching by acrylamide in the presence and absence of Zn2+ (Figure 2). The binding of Zn2+ to RUBCa provided protection against the quenching by acrylamide. In the presence of Zn2+, the dynamic quenching constant, KSV, decreased by ≈40% from 10.5±0.4 M−1 to 6.3±0.2 M−1 (Table 1).
Zn2+-induced conformational change in the RUBCa mini-domain was further confirmed by monitoring ANS fluorescence. The anionic amphiphilic ANS was used as a hydrophobic probe to examine the conformational properties of proteins. As shown in Figure 3(B), on the addition of RUBCa, the emission peak of ANS fluorescence blue-shifted from 510 to 500 nm and the intensity increased by ≈30%, suggesting that some of the hydrophobic regions of the purified RUBCa were exposed to the solvent and thus accessible to ANS. The addition of increasing amounts of Zn2+ caused a red shift (by 2 nm) and ≈10% decrease of the fluorescence intensity (Figure 3B). Taken together, these data indicate that the binding of Zn2+ induced significant conformational changes in RUBCa by reducing the exposure of hydrophobic regions to the surface.
By monitoring tryptophan intensity change, a dissociation constant of 202±74 nM for Zn2+ was obtained by assuming a 1:1 binding ratio (inset to Figure 3A). Consistent with this result, an apparent dissociation constant of 468±321 nM was obtained by monitoring the ANS fluorescence intensity change as a function of Zn2+ concentration (inset to Figure 3B).
The Zn2+ content of the RUBCa minidomain containing the WT (wild-type) sequence or one or more cysteine-to-serine substitution mutations (Figure 1A) were compared using the PAR colorimetric assay. As shown in Figure 4(A), except for the WT RUBCa ([Zn2+]/[P] = 0.69±0.03) and the mutant C1167S ([Zn2+]/[P] = 0.67±0.05), all of the other cysteine-to-serine mutants [C1175S, C1178S, C1225S, C1227S, C1167S/C1175S, C1175S/C1178S, treble mutant (C1167S/C1175S/C1178S) and C1225S/C1227S] showed a significantly lower metal to protein ratio ([Zn2+]/[P] = 0.08±0.29). With the quintuple mutant (when all five of the cysteine residues were replaced by serine), the minidomain failed to bind Zn2+. Thus Cys1175, Cys1178, Cys1225 and Cys1127, but not Cys1167, were involved in the co-ordination of Zn2+ in the binding pocket.
We further introduced the individual or combined cysteine-to-serine substitution mutations into RUBrep-HA/GFP, a replicon in which the RUBV structural protein ORF is replaced with a GFP reporter gene. Upon transfection of in vitro replicon transcripts into Vero culture cells, the transcripts were translated to produce the NS-ORF precursor which is proteolytically cleaved if it contains a functional NS protease. Cleavage is detected through the presence of an HA epitope tag in P150 by Western blotting using anti-HA antibodies [3,6]. As shown in Figure 4(B), NS protease cleavage of the precursor (termed P200) to P150 was efficient in WT RUBrep-HA transfected cells as well as in RUBrep-HA/GFP-C1167S transfected cells (the other cleavage product, P90, was not detected by anti-HA antibodies). Vero cells transfected with RUBrep-HA/GFP constructs with single cysteine-to-serine mutations at positions 1175, 1178, 1225, 1227, similar to the construct with a C1152S substitution at the catalytic site that served as the negative control, were unable to undertake cleavage (that is, only the P200 precursor was detected). In addition, cleavage in Vero cells transfected with RUBrep-HA/GFP constructs with combined cysteine-to-serine mutations (C1167S/C1178S, C1175S/C1178S, the treble mutant, C1225S/C1227S and the quintuple mutant) was abolished (Figure 4B). These findings correlated exactly with the effects of these mutations on Zn2+ binding in that mutants unable to bind Zn2+ lacked protease activity, whereas the Cys1167 mutant able to bind Zn2+ exhibited protease activity.
After transfection of in vitro replicon transcripts into cells, the replicon is capable of replication, as evidenced by production of the GFP reporter gene, but not cell-to-cell spread [3,6]. As expected, Vero cells transfected with transcripts containing mutations that compromised the protease activity showed no GFP fluorescence, indicating an inability to replicate (Table 2). Although the mutant C1167S was able to undertake cleavage of the precursor P200, it exhibited no fluorescence signal at all when transfected into Vero cells (see Supplementary Figure 4 at http://www.BiochemJ.org/bj/417/bj4170477add.htm). It seems likely that Cys1167 is not directly required for the Zn2+-dependent cleavage process, but is essential for the efficient replication of RUBV.
The importance of Zn2+ in the life cycles of viruses has been well recognized. Viral Zn2+-binding proteins include proteins that interact with nucleic acids and are important in transcription and/or replication, such as UL52 (unique long region 52) of HSV-1 (herpes simplex virus type 1), a member of a helicase–primase complex , the nucleocapsid of retroviruses (HIV-1 and murine-leukaemia virus) [17,18], the V protein of Sendai virus involved in RNA editing , helicases of nidoviruses such as SARS-CoV and equine-arteritis virus [20, 21], the NS5A replicase component of hepatitis C virus (HCV) [22-24], and several viral proteases, such as the papain-like cysteine protease 2 of SARS-CoV , nsp1 (NS protein 1) papain-like cysteine protease of equine-arteritis virus (a nidovirus) , the leader protease of encephalomyocarditis virus  and the NS3 (NS protein 3) protease of HCV .
In our previous study , a model of the RUBV NS protease, based on the FMDV leader papain-like cysteine protease with which it shares reasonable (albeit distant) homology, revealed a putative cysteine-rich Zn2+-binding pocket with no histidine residues present in close proximity. This pocket was present in the previously characterized RUBCa minidomain, which we purified away from the GST expression tag, discovering that it bound Zn2+ with apparent nanomolar affinity.
Our previous study  reported that this minidomain also contains an EF-hand Ca2+-binding motif that is required for structural stability and optimal protease activity. The requirement for both Ca2+ and Zn2+ as cofactors in a protease is reminiscent of thermolysin, a thermostable metalloprotease produced by the Gram-positive bacteria Bacillus thermoproteolyticus. The only bound Zn2+ in thermolysin is directly involved in the enzymatic activity, whereas the binding of four Ca2+ ions is required for thermostability [29,30]. In the present study we showed that the minidomain is well folded in the absence of Zn2+, which is different from the siutation in zinc-finger proteins, which require Zn2+ binding for folding. Our conformational analyses clearly demonstrate that the bound Zn2+ plays a structural role by maintaining structural integrity and shielding some hydrophobic regions from bulk solvent. Site mutagenesis of the putative Zn2+-binding ligands identified Cys1175,Cys1178,Cys1225 and Cys1227, but not Cys1167, as co-ordinating ligands, and this corresponds exactly to protease activity in vivo, which requires the four earlier cited residues, but not C1167.
We previously reported that mutagenesis of Cys1175, Cys1178 and Cys1227 results in the loss of 65Zn2+-binding activity, whereas mutagenesis of Cys1167 does not [4,5], a finding similar to that in the present study, but mutagenesis of Cys1225 did not reduce 65Zn2+ binding, in contrast with the results of the present study. In our previous study , the mutagenesis and 65Zn2+ binding experiments were performed using MBP fusions in which Cys1167, Cys1175 and Cys1178 are present in one fusion while Cys1225 and Cys1227 are present on a different one, which would destroy the Zn2+-binding pocket identified in the present study. In our previous study , Cys1175, Cys1178 and Cys1227 mutants were found to lose in vitro protease activity (using a coupled transcription/translation system), a finding similar to that obtained in the present study. In the previous study, a Cys1167 mutant lost activity (this mutant retained 65Zn2+ binding), whereas a Cys1225 mutant retained in vitro protease activity , a result in contrast with the in vivo protease activity obtained in the present study.
The discrepancy with these previously obtained results could arise for several reasons. First, in our previous studies , the expressed NS protease regions were never successfully separated from the MBP tag, and thus we could not unambiguously rule out the contribution of Zn2+ binding by the fusion tag. Additionally, identification of Zn2+-binding ligands was done using site-directed mutagenesis and a 65Zn2+ blotting assay in which the MBP–NS protease region fusions were not in solution (although subjected to renaturing conditions following blotting). It should also be noted that binding using the 65Zn2+ binding assay was assessed as ‘all or none’ by visual inspection, whereas the quantitative PAR assay used in the present study revealed that the decrease in Zn2+ binding exhibited by the single mutants was in the range of 50–65%, potentially leading to a false negative result (that is, identifying a mutant as retaining Zn2+ binding when it did not). Thirdly, the in vitro coupled transcription–translation assay was carried out under aerobic, non-reducing conditions that may not truly reflect the reducing environment of cell-based assays. Given the high numbers of cysteine residues in the NS protease (six in the less-than-100-amino acids region between residues 1152 and 1227), a reducing environment is more favourable for Zn2+ binding and protein folding by preventing formation of mismatched disulfide bonds. It is to be noted that, in the in vitro assay, cleavage of the P200 precursor was only partially complete, whereas, in the in vivo assay, cleavage was nearly complete. Given this, we consider that cell-based assay carried out in the preset study is more reliable than the in vitro assay.
In summary, site-directed mutagenesis analyses carried out on the bacterially expressed minidomain within the RUBV NS protease have defined a CXXC(X)48CXC Zn2+-binding motif that is required for protease activity. On the basis of conformational studies on the minidomain, Zn2+ plays a structural role in protease activity. Given the necessity of post-translational processing of the P200 precursor by this protease in the formation of replication complexes, it is not surprising to find that mutation of any of these residues impairs replication of a RUBV replicon. Interestingly, a nearby cysteine residue, Cys1167, is necessary neither for Zn2+ binding nor protease activity, but is still required for replication. Tetrahedral Zn2+ binding sites have been found that play structural roles in a number of other proteases of positive-sense RNA viruses [31-33], and the requirement of Zn2+ ion as a structural cofactor in these proteases is not a rare scenario. The spacing of the ligand residues of the RUBV NS protease Zn2+-binding site CXXC(X)48CXC is reminiscent of the CXXC(X)31CXC spacing of Zn2+-binding ligands in the PLpro (papain-like protease) of SARS-CoV [33,34]. Mutational analyses have revealed that the four cysteines in SARS-CoV involved in the binding of Zn2+ are also crucial for the proteolytic activity of PLpro. The crystal structure of SARS-CoV PLpro further confirms the mutational analyses and reveals that the Zn2+ ion is co-ordinated with four cysteine residues residing on two β-hairpins by adopting a tetrahedral geometry . The four cysteine residues are located at or close to the termini of anti-parallel β-strands. Similar to the Zn2+-binding site in SARS-CoV PLpro, our results suggest that the Zn2+-binding site in the protease domain of RUBV are also formed by four cysteine residues. However, the four cysteine residues in RUBV protease are flanked by helices instead of β-strands. By fixing the backbone atoms and allowing free rotation of sulfur atoms on side chain in our model structure of the RUBV protease domain, the minimal distance between two sulfur atoms of proposed cysteine ligands falls into the range of 3.5–6.0 Å, which is in good agreement with the average value of 4.0 Å in SARS-CoV PLpro. Further structural determination of the RUBV protease domain will unambiguously reveal the detailed co-ordination properties of Zn2+-binding pocket of RUBV protease.
Y. Z. is a fellow of the Molecular Basis of Disease Area of Focus at the Georgia State University.
This work is supported in part by a grant from the National Institute of Allergy and Infectious Diseases [number R01 AI21389] to T. K. F. and J. J. Y., and in part by grants from the National Institutes of Health [numbers R01 GM 62999, GM081749] to J. J. Y.