New Antiviral Target Revealed by the Hexameric Structure of Mouse Hepatitis Virus Nonstructural Protein nsp15

Protein expression and purification.

The complete gene fragment encoding MHV-A59 nsp15 was cloned from the virus RNA genome by reverse transcription-PCR into the pGEX6p-1 vector (Amersham Biosciences), and the resulting plasmid was transformed into Escherichia coli BL21(DE3) cells. The recombinant glutathione S-transferase (GST) fusion protein, GST-nsp15, was purified by GST-glutathione affinity chromatography. The GST tag was removed by PreScission protease (Amersham Biosciences), leading to five additional residues (GPLGS) at the N terminus. It was noted that the expression of the wild-type (WT) nsp15 is toxic to E. coli, indicated by slower cell growth and low yield. An incidental F307L point mutation was identified because of its significantly improved expression under the same conditions as those of the WT. In addition, another three mutants, H262S, H277S, and K317S, were constructed, expressed, and purified using the same protocol; all resulted in higher yields than the WT nsp15.

Crystallization.

MHV-A59 nsp15 was crystallized by the hanging drop vapor diffusion method with a 3-μl drop volume. A 1.5-μl mixture of nsp15 at 10 mg/ml and the n-octyl β-d-thioglucopyranoside (OTG) detergent at a final concentration of 6 mM was added to the same volume of reservoir solution (0.6 to 1.0 M lithium sulfate, 0.5 M ammonium sulfate, and 0.1 M sodium citrate, pH 6.0), producing drum-shaped crystals with the space group P6₃22. Crystals grown under the same conditions but without the OTG detergent had a thin semicircular shape and diffracted poorly in the c* direction.

Data collection.

To improve their diffraction quality, the drum-like crystals of WT nsp15 were dehydrated in a buffer containing 0.2 M lithium sulfate, 3 M ammonium sulfate, 0.1 M sodium citrate (pH 6.0), 4% glycerol for 10 to 20 min and then transferred to a cryoprotectant solution that contained the same components as the dehydration buffer but with a lower concentration of ammonium sulfate (1 M) (to avoid forming salt crystals) and a higher concentration of glycerol (25%). A 2.7-Å-resolution data set of WT nsp15 was collected at 100 K using an in-house Rigaku MM-007 generator and a Rigaku R-AXIS IV++ detector. The crystal form belongs to the hexagonal space group P6₃22 (a = b = 86.4 Å, c = 221.1 Å, α = β = 90°, γ = 120°) with one copy of nsp15 per asymmetric unit and a V_M of 2.8 ų Da⁻¹ (5, 15), corresponding to a solvent content of 57%.

Crystals of the selenomethionyl (Se-Met) derivative F307L mutant were obtained in order to solve the phase problem. The Se-Met crystals shared the same space group and unit cell with the WT crystals and were dehydrated and cryoprotected following the same protocol. The F307L mutant of nsp15 was used, as it is easier to express than the wild type under the same conditions. A 2.15-Å-resolution data set of the Se-Met derivative was collected for the F307L mutant at 100 K using an SBC2 (3,000 by 3,000) CCD detector on beamline BL19-ID at the Advanced Photon Source (Argonne National Laboratory). Processing of diffraction images and scaling of the integrated intensities were performed using the HKL2000 software package (18).

Structure determination.

The structure of nsp15 was solved by the single-wavelength anomalous dispersion (SAD) method from an Se-Met derivative of the F307L mutant. Four of the five potential selenium atoms in one nsp15 monomer were located, and initial phases were calculated by the program SOLVE (27). Density modification (solvent flipping) and phase extension to 2.15 Å were performed using RESOLVE (27). The model of nsp15 was automatically traced using the program ARP/wARP (19) to approximately 80% completeness and then further manually built and refined using the programs O (11) and CNS (4). The F307L mutant crystal structure was refined at 2.15-Å resolution to a final R_work of 20.8% and R_free of 24.8%. The WT nsp15 structure was refined at 2.7-Å resolution to a final R_work of 19.5% and R_free of 25.7%. Refinement statistics are detailed in Table 1.

Gel filtration.

The oligomerization of MHV-A59 nsp15 was analyzed by eluting the nsp15 protein from a Superdex 200 HR10/30 gel filtration column (Pharmacia) with a buffer containing 25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM dithiothreitol (DTT) at a flow rate of 0.5 ml min⁻¹ and temperature of 16°C. The WT and mutants (H262S, H277S, K317S, and F307L) were analyzed in parallel under the same buffer conditions. All experiments were repeated more than three times, and representative data are shown.

Enzyme activity assay.

The activity of MHV-A59 nsp15 was examined using fluorescent resonance energy transformation (FRET) (24, 25) following a previously published protocol (17). Two substrates were designed based on the core MHV transcription regulatory sequence (5′-UCUAAAC-3′) (9, 33); each contains a 6-carboxyfluorescein (6-FAM) and a tetramethylrhodamine (TAMRA) group linked respectively to the 5′ end and 3′ end of a four-nucleotide substrate. One substrate consists of the nucleotides dC-rU-dA-dA, and the other consists of dC-dT-dA-dA as a negative control. The excitation and emission wavelengths of 6-FAM are 498 nm and 518 nm, respectively; those of TAMRA are 558 nm and 576 nm, respectively. While the physical closeness of the 6-FAM and TAMRA groups in the intact substrate diminishes fluorescent emission at 518 nm, cleavage of the substrate dC-rU-dA-dA at the ribonucleotide uridine separates the two fluorophore groups and results in an increase in the 518-nm emission. Activity assays for WT nsp15 and H262S, H277S, K317S, and F307L mutants were performed under the same conditions in a buffer containing 50 mM HEPES (pH 7.5) and 50 mM KCl, which was diluted by 0.1% diethyl pyrocarbonate-treated water. The activity was assayed at 30°C, with the excitation wavelength at 485 nm and emission wavelength at 518 nm, using a fluorescence time scan with a Fluoroskan Ascent instrument (ThermoLabsystems, Helsinki, Finland). All experiments were repeated more than three times, and representative data are shown.

Protein structure accession numbers.

Coordinates and structure factors for the MHV nsp15 crystal structures have been deposited in the RCSB Protein Data Bank under accession numbers 2GTH (for the 2.7-Å WT crystal structure) and 2GTI (for the 2.15-Å F307L mutant structure).

Overall structure.

The cDNA coding for MHV nsp15 was cloned and amplified using PCR. The coded protein consists of amino acid residues 6506 to 6879 of pp1ab, which are renumbered 1 to 374 hereafter for convenience. The recombinant WT protein and a selenomethionyl derivative of the incidental mutant F307L were crystallized in different morphologies but with the same cell parameters and with similar structures. Unless specifically mentioned, the two crystal structures are discussed without distinction. In both crystal structures, residues in the region Tyr195-Leu216 and after Phe369 (modeled as alanine) could not be traced due to lack of interpretable electron density, although a protein sample of dissolved crystals was confirmed to be an intact peptide by sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis. Cα atom root mean square deviation (RMSD) between the two crystal structures is 0.44 Å.

Crystallographic symmetry-related nsp15 molecules form tight trimers, which further assemble into hexamers with whole symmetry D3 (Fig. 1A to C). According to the large contact surface area (discussed below), this hexamer is believed to be equivalent to the functionally required nsp15 hexamer in solution. The electrostatic potential on the hexamer surface is shown in Fig. 1D. The electrostatic surface is predominantly negative at the ends of the hexamer, with areas of positive and negative charge scattered around the sides. The dimensions of the MHV-A59 nsp15 hexamer are 96 by 96 by 110 Å, which are in excellent agreement with single-particle analysis of SARS-CoV nsp15 (9). Image reconstruction from transmission EM analysis of SARS-CoV nsp15 suggested that six subunits form a dimer of trimers with dimensions of ∼90 by 90 by 110 Å. Consistent with the SARS-CoV nsp15 EM structure, our MHV-A59 nsp15 structure shows cavities at both ends as well as a central core. Nevertheless, there is no evidence of the channels observed from a side-on view of SARS-CoV nsp15.

Monomer structure of nsp15.

The refined model of nsp15 consists of residues from the N terminus (including an additional serine residue from the GST tag) to Phe369. The nsp15 monomer contains nine α-helices (α1 to α9) and 21 β-strands (β1 to β21). The secondary structure elements and their locations in the primary structure are illustrated in Fig. 2. The overall fold of the nsp15 monomer can be viewed as a butterfly-shaped molecule (Fig. 2A) with approximate dimensions of 40 by 55 by 70 Å. It contains three domains: the N-terminal “wing” domain (or N-domain; residues 1 to 63), the middle “body” domain (or M-domain; residues 64 to 194), and the C-terminal “wing” domain (C-domain; residues 217 to 396). The N-terminal domain contains two α-helices (α1 and α2) and a three-strand antiparallel β-sheet (β1-β2-β3), and it is connected to the middle domain through the 60-to-69 loop. The middle domain contains one major mixed β-sheet (β4-β7-β8-β14-β9/β15), three small ones (β5-β6, β10-β11, and β12-β13), and two short α-helices (α3 and α4). Skipping the missing region from Tyr195-Leu216, which is presumably a flexible interdomain linker, the visible C-terminal domain contains two β-sheets (β16-β17-β18 and β19-β20-β21) and five α-helices (α5 to α9). The interface between the N-terminal and middle domains buries a solvent-accessible surface (SAS) of ∼1,300 Ų, and that between the middle and C-terminal domains buries ∼1,800 Ų. SAS calculations were performed using CCP4 (5). Such extensive interactions are likely to endow the three-domain organization of nsp15 with high stability. On the other hand, the multidomain organization provides a large surface area and structural motifs to interact with other macromolecules as well as to form a homooligomer. Three previously identified residues critical to the activity of nsp15, namely His262, His277, and Lys317, are close to each other in 3D structure and are located on a concave surface in the C-domain of one monomer, where they are distributed around the edge of a cavity (Fig. 2D). In the WT nsp15 structure, His277 N_ɛ2 (3.06 Å), Lys317 N_ζ (3.18 Å), and Glu363 O_ɛ2 (2.83 Å) all interact with a bound water molecule, while Glu363 O_ɛ2 forms a hydrogen bond with His262 N_ɛ2 (2.65 Å). Other strictly conserved residues distributed around the putative active site include Gly266, Gly274, Gly275, Ile280, Val319, Cys320, and Asp324. Among the nine cysteine residues in an nsp15 monomer, only Cys80 and Cys95 are potentially able to form an intramolecular disulfide bond. Nevertheless, their S_γ-S_γ distance is 4.0 Å, which is thus in the oxidized state. Interestingly, the surface loop where Cys95 resides is involved in nsp15 oligomerization; therefore, the reduced oxidation state of the potential disulfide bridge might influence the stability of the oligomer.

A whole-monomer and domain-by-domain search in DALI failed to identify any structural homologs, indicating the novelty of the MHV-A59 nsp15 structure and fold.

Monomer-monomer interactions and trimer assembly.

nsp15 monomers are organized into trimers inside the crystal (Fig. 3). There are two sets of crystallographically independent threefold axes in the nsp15 crystal form, and each one assembles the nsp15 monomers into a distinct trimer. The first type of trimer buries 2,600 Ų SAS per monomer (54% contributed by hydrophobic groups), and the second one buries 1,400 Ų SAS per monomer. As a result of the more extensive monomer-monomer interactions, the first trimer is believed to be biologically relevant and will be discussed in more detail. This trimer has the shape of a three-footed pot, with the N-terminal domain forming the bottom, the C-terminal domain shaping the rim as well as three ears, and the middle domain contributing to the wall of the cavity and forming the three feet. The central spherical cavity of the trimer is about 35 to 40 Å in diameter, and the opened end is slightly smaller in diameter (∼20 Å). The closed end has a hole with a diameter of less than 10 Å. The outer surface of the foot and ear from each monomer forms a continuous ridge. In such a trimer, the small N-terminal domain packs against a deep cleft between the middle and C-terminal domains (Fig. 3B). Although extensive hydrogen bond networks are observed in the monomer-monomer interface, no intermolecular β-sheet exists in the trimer. Intermonomer hydrogen bonds include those of Lys30-Asn45 (3.2 Å), Lys30-Thr47 (2.9 Å), Glu38-Arg73 (2.6 Å), Asp39-Lys90 (3.1 Å), Asn157-Thr309 (3.0 Å), Lys163-Ser313 (3.3 Å), Gly165-Ser313 (3.0 Å), and Asp166-Ser313 (2.8 Å). Residue Phe307 is located in the β17 strand and would be partially solvent exposed in a monomer. In both the WT and F307L crystals, however, the side chain at this position becomes completely buried in the trimer interface and faces Ile63 from a neighboring monomer.

Trimer-trimer interactions and hexamer architecture.

Two nsp15 trimers of the type described above further interact in a base-to-base fashion to form a hexamer through crystallographic twofold symmetries (Fig. 1). The two sets of three feet are slightly displaced relative to each other, if viewed along the crystallographic threefold axis shared by the two trimers. The trimer-trimer interface buries 3,600 Ų SAS from each trimer, with 47% contributed by nonpolar groups. Around the waist of the hexamer, Cys109 from each nsp15 monomer forms a twofold symmetry-restricted disulfide bond with its counterpart across the trimer-trimer interface. Deep in the center of the hexamer, the helix α1 at each N terminus goes face to face following another crystallographic twofold symmetry with its counterpart to form a close contact. An extensive expansion of the nsp15 N terminus should prohibit the formation of the hexamer, which is consistent with the observation that all known nsp15 proteins have uniform N termini. As in the monomer-monomer interface within a trimer, no intermolecular β-sheet exits in the trimer-trimer interface. The visible C terminus (residue 369) is located near both the intermonomer and intertrimer interfaces, but it is buried in neither. Similarly, the missing region of Tyr195-Leu216 is projected to be on the surface of the hexamer. This flexible region is unlikely to interfere with either the trimer or hexamer formation. Furthermore, this region is located on the opposite side of the C-terminal domain from where the putative active-site residues His262, His277, and Lys317 cluster together.

A detailed analysis of the distribution of conserved residues in the nsp15 structure further supports the biological relevance of the crystallographically observed hexamer. Figure 2C shows high sequence homology among CoV nsp15 proteins. For simplicity, we focus on the strictly conserved residues in the sequence alignment. Most of the identical residues are concentrated in two regions in the 3D structure of nsp15: (i) the interface between the N-terminal and middle domains (conserved N-M region) and (ii) the C-terminal domain. The first region is heavily involved in both trimer and hexamer formation; therefore, these conserved residues play important roles in several levels of nsp15 structural organization. The tip of the N-terminal domain, however, is not the most conserved region. Nevertheless, during the trimer formation it forms a complementary interaction with an equally less conserved concave surface area formed by the middle and C-terminal domains. On the other hand, most of the conserved C-terminal domain residues do not directly participate in intermonomer contacts. The largest cluster of conserved residues in the C-terminal domain forms the top face of the pot and the entry rim. In nsp15, the top face of this cluster and its surrounding region are particularly rich in acidic residues, including Glu236, Asp240, Asp243, Asp245, Asp246, Asp247, Glu261, Asp328, and Asp329. Another 3D cluster of conserved residues in the C-domain consists of the three above-mentioned key residues in catalysis (i.e., His262, His277, and Lys317) and their surrounding region. Furthermore, the highly conserved but mobile C-terminal tail is close to the putative active site and may also play a functional role in catalysis.

Regarding the structural basis for the requirement of oligomerization for the enzymatic activity (see below), we have identified an extended patch on the hexamer surface of concentrated conserved residues. It starts from the putative active site on one C-terminal domain, extends across the conserved N-M region in a neighboring molecule in the trimer, and further spans across the trimer-trimer interface via a crystallography twofold symmetry. The hexamer formation may enhance the enzyme activity in one of at least two possible ways: (i) the extended conserved surface patch facilitates the substrate recognition, and (ii) the catalytic machinery is composed of parts from different domains that are brought together in the hexamer.

Besides the conserved residues, the sequence deletion-insertion pattern shown in Fig. 2C is also consistent with our structure-based observations. There are six deletion-insertion regions among the CoV nsp15 proteins compared in Fig. 2C. Four of them (around positions 174, 145, 210, and 340) align on the outer surface ridge, including the Tyr195-Leu216 region mobile inside the crystal; the fifth one, near position 229, is at the toe of the pot foot; and the last one is near position 298 and is located inside the pot cavity. All six deletion-insertion regions are solvent exposed and unlikely to interfere with the trimer and hexamer organization of nsp15 in any of the coronaviruses included in the comparison. In addition, these variable regions are in locations distinct from the putative active site. Furthermore, most of the deletion-insertion segments are in regions of non-α, non-β secondary structure, except the one located in the β14-β15 turn. In the MHV-A59 nsp15 crystal structure, this β-turn protrudes into solvent. Removing this β-turn would not compromise the stability of the monomer or higher order oligomers.

Mn²⁺ dependence of the endoribonuclease activity and catalytic residues.

To verify that our recombinant nsp15 has RNA substrate specificity, we performed an enzyme activity assay for two substrates that differ in only one nucleotide. For an enzyme concentration of 0.2 μM and substrate concentration of 1.2 μM, the reaction velocity of 5′-FAM-dC-rU-dA-dA-3′-TAMRA hydrolysis is 1.8 (±0.22) nM s⁻¹, while 5′-FAM-dC-dT-dA-dA-3′-TAMRA was not cleaved by the enzyme, indicating that nsp15 is uridine specific and has no hydrolysis activity against deoxythymidine. Since the only chemical differences between the two substrates were deoxythymidine C5 methylation and C2′ deoxidation, the discrepancy in the enzyme activities for the two substrates is likely to stem from the C5 methyl and C2′ hydroxyl groups. Furthermore, based on previous studies on NendoU in which the C2′ hydroxyl group played a critical role in substrate cleavage (10), we propose that the 2′ hydroxyl group is also involved in the nsp15-catalyzed hydrolysis reaction.

It has been proposed that the endoribonuclease activity of coronavirus nsp15 is Mn²⁺ dependent, although the structural basis of this dependence remains unclear. Enzyme activity assays were performed with 0.25 μM WT nsp15 and 1 μM 5′-FAM-dC-rU-dA-dA-3′-TAMRA substrate in the presence of various concentrations of manganese. The enzyme activity increased linearly when the concentration of Mn²⁺ was increased from 1 to 5 mM. These results indicate that the MHV-A59 nsp15 is a manganese-enhanced uridine-specific endoribonuclease.

We further performed enzyme activity assays for WT nsp15 and the H262S, H277S, and K317S mutants under identical conditions. With the concentration of enzyme at 0.25 μM and that of the substrate 5′-FAM-dC-rU-dA-dA-3′-TAMRA at 1 μM, the activity of the mutants was at least 10-fold less than that of the WT one monitored using fluorescent absorption at 518 nm (Fig. 4A). Thus, the H262S, H277S, and K317S mutations severely diminish the enzyme activity, revealing that these three residues are located at a catalytic site. Previous studies of similar mutations in SARS-CoV nsp15 and NendoU suggest that a decrease in their ion binding or substrate binding capacity may be an additional or alternative reason for the loss of endoribonuclease activity (9, 10).

The close proximity of His262, His277, and Lys317 around a cavity suggests that they may coordinate a manganese ion. In order to verify if this region is the manganese binding site, we attempted to soak MHV-A59 nsp15 crystals with manganese. In a resulting 2.5-Å structure, while there appeared to be strong electron density in the cavity supporting a bound metal ion, we did not observe the expected hexacoordination environment for manganese (data not shown). Furthermore, inspection of the crystal structure does not suggest that Mn²⁺ binding is disrupted by the crystal packing.

MHV-A59 nsp15 is fully active only as a hexamer.

Elution of purified MHV-A59 WT nsp15 by gel filtration yields three distinct 280-nm absorption peaks (Fig. 4B). The first peak was eluted at 8.2 (±0.1) ml and is believed to be a random aggregation state with no biological function, since this species has no enzymatic activity and could not be crystallized. According to the elution pattern of the Pharmacia standard marker proteins (data not shown), the second peak was eluted at 11.4 (±0.1) ml, corresponding to a molecular weight consistent with a hexamer of nsp15. The third peak appeared at 14.9 (±0.2) ml, corresponding to the monomer of nsp15. The areas under the first and second peaks are larger than that under the third one (Fig. 4B). Interestingly, the equilibrium between the three peaks of MHV-A59 nsp15 was not affected by Mn²⁺ concentration or reducing agents such as DTT. In contrast to MHV-A59 nsp15, it was shown previously that, under similar conditions, SARS-CoV nsp15 exhibits two distinct UV absorption peaks at 11.41 ml and 15.1 ml, respectively, and 90% of the total protein was hexameric (9). When performed with the same 500-μg quantity of enzyme under identical conditions, the elution profiles of H262S, H277S, and K317S mutants were similar to that of the WT nsp15. Each mutant was eluted at three distinct 280-nm absorption peaks, with the second peak appearing at 11.4 (±0.1) ml and the third peak appearing at 14.6 (±0.7) ml (H277S is shown in Fig. 4A). However, the detailed ratio of the three peaks was different between the three mutants and WT nsp15. Thus, replacement of His262, His277, and Lys317 with serine did not influence hexamer formation in solution. The lower activity of these active-site mutants does not result from the different proportions of oligomeric species but from disruption of the catalytic site. In addition, based on the structure, these active catalytic site residues do not participate in interactions between monomers.

The F307L mutant is detrimental to endoribonuclease activity. Under the same conditions as those for WT nsp15, the activity of the F307L mutant was extremely low and showed no response to Mn²⁺. Meanwhile, unlike other low-activity mutants being analyzed, the elution profile of the F307L mutant through a Superdex 200 gel filtration column is in sharp contrast to that of the WT nsp15. Two peaks were observed at 8.3 (±0.1) ml and 14.0 (±0.1) ml, respectively, and the first peak was much weaker (15%) than the monomer peak eluted later (85%) (Fig. 4B). Thus, the F307L mutant exists in two forms, with the second peak being the dominant species in the buffer. This species corresponds to a molecular mass of about 72 kDa and may be a mixture of monomer, dimer, and trimer. These results suggest that the cause of activity loss in the F307L mutant is different from that of active-site mutations. Since Phe307 is buried within the trimer interface, replacement of Phe307 with leucine may disturb the hydrophobic interactions between two monomers, preventing the formation of stable trimers and hexamers. It appears that the crystallization process of F307L selected the trimer form for crystal packing, thus shifting the equilibrium between species.

To further clarify the relationship between hexamer formation and activity, we performed enzyme activity assays for the hexamer and monomer species of WT nsp15 separately after collecting the elution fractions. Since we did not observe the WT trimer species from gel filtration experiments, activity of the trimer could not be assayed. The dynamic parameters for the hexamer species are the following: K_m, 24 ± 3.5 μM; k_cat, 0.20 ± 0.03 s⁻¹; for the monomer species they are the following: K_m, 27 ± 3 μM; k_cat, 0.20 ± 0.02 s⁻¹. The k_cat values for the hexamer and monomer species are identical, indicating their catalytic activities are the same, but the K_m value for the hexamer is slightly lower than that for the monomer, suggesting its substrate binding ability may be slightly greater than that of the monomer species. Furthermore, the activity of WT nsp15 can be enhanced by the presence of Mn²⁺ (Fig. 4C). Therefore, we conclude that the full activity of nsp15 requires a stable hexamer formation, which is believed to favor substrate binding. This notion is strongly supported by the results from our crystallography analysis on the structure of nsp15 hexamer. Consistent with this rationale, previous studies concerning SARS nsp15 also point to the hexamer as the enzymatically active state (9). In contrast to the three active-site mutants, SARS CoV nsp15 E56A (56 in MHV) and D212A (240 in MHV) point mutants behave the same as the WT in both gel filtration and activity assays, while E3A (3 in MHV) shows a larger monomer population (52%) and lower level of activity. Inspection of equivalent positions in the crystal structure of MHV-A59 nsp15 provides a clear explanation to the varied functional behaviors of these mutants at conserved positions throughout the coronavirus family. Both Glu56 and Asp240 are located on the monomer surface and are not directly involved in oligomerization, while the side chain of Glu3 forms a 2.9-Å hydrogen bond with the main-chain amide nitrogen of Leu2 of the nsp15 monomer across the trimer-trimer interface. Thus, this observation once again confirms that hexamerization is important for nsp15 activity. Furthermore, the activity of the D212A mutant indicates that the corresponding ear region in the hexamer is less likely to contribute to the enzymatic activity.

Conclusions.

The scientific significance of this study is as follows. First, the crystal structure of MHV-A59 nsp15 demonstrates a new protein fold. Second, the MHV-A59 nsp15 structure provides a model for elucidation of the structure-function relationship in the XendoU family. It becomes the first member of this family to have its 3D structure unveiled. Other members of this family are likely to share the same novel fold and organization. Third, we firmly demonstrate that the specific catalysis of nsp15 at the uridine nucleotide depends both on stable hexamer formation and Mn²⁺ ion binding. Mutations disturbing the monomer-monomer interface are detrimental to the activity of nsp15. Fourth, the conserved amino acids His262, His277, and Lys317 constitute the active-site residues and are critical for the enzyme activity of nsp15. Endoribonuclease activity is essential for the virus life cycle and is unusual among positive-strand RNA viruses. The high sequence conservation of the nsp15 enzyme in coronaviruses, including SARS, provides a new potential avenue for the design of antiviral therapeutics. One possibility may be to use an approach similar to the design of broad-spectrum coronavirus M^pro inhibitors in which a conserved region of the enzyme was targeted (29). The current studies on MHV-A59 nsp15 provide rich structural information for the development of inhibitors and other antiviral drugs. The 3D structure of MHV-A59 nsp15 also adds to the increasing knowledge base of coronavirus replicase/transcriptase structures (2), which also includes the SARS-CoV nsp3 ADRP domain (22), SARS-CoV nsp5 (M^pro) (2), the hexadecameric SARS-CoV nsp7-nsp8 complex (32), SARS-CoV nsp9 (6, 26), and SARS-CoV nsp10 (Z. Rao, personal communication).