Insight into the evolution of nidovirus endoribonuclease based on the finding that nsp15 from porcine Deltacoronavirus functions as a dimer

Nidovirus endoribonucleases (NendoUs) include nonstructural protein 15 (nsp15) from coronaviruses and nsp11 from arteriviruses, both of which have been reported to participate in the viral replication process and in the evasion of the host immune system. Results from a previous study of coronaviruses SARS-CoV, HCoV-229E, and MHV nsp15 indicate that it mainly forms a functional hexamer, whereas nsp11 from the arterivirus PRRSV is a dimer. Here, we found that porcine Deltacoronavirus (PDCoV) nsp15 primarily exists as dimers and monomers in vitro. Biological experiments reveal that a PDCoV nsp15 mutant lacking the first 27 amino acids of the N-terminal domain (Asn-1–Asn-27) forms more monomers and displays decreased enzymatic activity, indicating that this region is important for its dimerization. Moreover, multiple sequence alignments and three-dimensional structural analysis indicated that the C-terminal region (His-251–Val-261) of PDCoV nsp15 is 10 amino acids shorter and forms a shorter loop than that formed by the equivalent sequence (Gln-259–Phe-279) of SARS-CoV nsp15. This result may explain why PDCoV nsp15 failed to form hexamers. We speculate that NendoUs may have originated from XendoU endoribonucleases (XendoUs) forming monomers in eukaryotic cells, that NendoU from arterivirus gained the ability to form dimers, and that the coronavirus variants then evolved the capacity to assemble into hexamers. We further propose that PDCoV nsp15 may be an intermediate in this evolutionary process. Our findings provide a theoretical basis for improving our understanding of NendoU evolution and offer useful clues for designing drugs and vaccines against nidoviruses.

Nidoviruses contain 5Ј-and 3Ј-untranslated regions (UTRs) at the terminal regions of their genomes, and they utilize a discontinuous negative-stranded RNA synthesis model to generate subgenomic mRNA (8,12). They have similar genomic organizations in which the two largest overlapping open reading frames (ORF1a and ORF1b) are located within the 5Ј-proximal two-thirds of the genome, and a ribosomal frame-shifting event is involved in the translation process (13,14). The viral RNA can be translated into 12 or 13 nonstructural proteins (nsps) 3 in arteriviruses and 16 nsps in most coronaviruses to exert different biological functions, such as the formation of double-membrane vesicles (DMVs) and replication-transcription complex (RTCs) formation (3,15,16). Viral RNA synthesis has been observed in DMVs of SARS-CoV and MHV (15,17). Among the nsps found in Nidovirales, NendoU is unique and can both excise single-and dsRNAs and specifically recognize uridylates to produce 2Ј-3Ј-cyclic phosphodiester products (5,18). The DMVs are associated with the production of dsRNA intermediates and can sequester them away from host dsRNA sensors to prevent the activation of the host innate immune response (15). A recent study reported that the coro-navirus nsp15 cooperates with the viral replication complex to limit the exposure of viral dsRNA to host dsRNA sensors, thereby enabling the evasion of the host cell immune system (15,19). In addition, the inhibition of IFN-␤ induction has been verified in PRRSV nsp11, but further investigations need to be conducted because a cytotoxic effect of WT nsp11 resulting in IFN-␤ inhibition was not excluded (20,21). Regardless of whether nsp11 contributes to the decline in innate immune functions, its importance in viral replication has been established (22).
Among the Coronaviridae, SARS-CoV nsp15 was first shown to have NendoU activity in vitro (23); similar findings were then obtained for MHV, HCoV-229E, turkey coronavirus (24 -27), and PRRSV (5). However, not all nidoviruses possess NendoU activity; some examples are the insect nidovirus Nam Dinh virus and roniviruses, all of which lack the EndoU domain (1,28). The active sites of three catalytic residues (two histidines and one lysine) in NendoU are highly conserved in its counterpart, XendoU (19,24), an endoribonuclease found in eukaryotes that processes the intron-encoded box C/D U16 small nucleolar RNA from its host pre-mRNA and produces 2Ј-3Ј-cyclic phosphate termini (29). An intriguing finding regarding NendoU and XendoU is the difference in their requirement for the manganese ion. XendoU is a monomer and requires Mn 2ϩ (29,30). Except for turkey coronavirus nsp15, Mn 2ϩ is generally essential for the catalytic activity of coronavirus nsp15 (25,31). However, Mn 2ϩ is dispensable and, to some extent, even inhibits the activity of PRRSV nsp11, implying the existence of different enzymatic mechanisms for NendoU in coronaviruses and arteriviruses (5).
The results of previous studies have indicated that nsp15 is important for nidovirus replication. More recent studies have shown that the replacement of the conserved Asp-6408 in HCoV-229E nsp15 with alanine (Ala) completely abolishes its EndoU activity and viral RNA synthesis, thus rendering the virus unrecoverable. A similar result was obtained in an MHV nsp15 mutant in which Ala was substituted for Asp-324 (24,26). The results reported by Kang et al. (26) indicated that Asp-324 is an important amino acid for the ionic-bond network because the mutant protein is insoluble when MHV nsp15 was expressed in Escherichia coli. Taken together, these results suggest that the failure of nsp15 to fold correctly may play an indirect role in viral replication by affecting the proteolytic processing of the neighboring components by preventing itself from folding correctly, ultimately leading to a nonviable phenotype (32). Furthermore, these findings regarding coronavirus nsp15 demonstrate its critical function in the evasion of the dsRNA sensors in macrophages and suggest that the disruption of the stability of nsp15 in coronaviruses may be helpful for designing new live-attenuated vaccines (19).
In a previous study, nsp15 was found to be active as a hexamer. This finding was confirmed for the nsp15s in several coronaviruses, including those of MHV, SARS-CoV, HCoV-229E, and turkey coronavirus nsp15 (25,27,33,34). Furthermore, the basic biofunctional unit of SARS-CoV nsp15 was confirmed to be a hexamer (33). Multiple sequence alignments and the site-directed mutagenesis experiments using SARS-CoV nsp15 showed that the N-terminal domain was essential for the oligomerization of nsp15 and that destruction of this motif strongly affected the catalytic activity of the enzyme (33). A previous study of the oligomerization's influence on SARS-CoV nsp15 activity showed that a mutant lacking the 28 N-terminal residues and 11 C-terminal residues was assumed a monomeric rather than a hexameric form, and the stabilization of the catalytic loop (residues 234 -249) by the supporting loop (residues 276 -295) was required to maintain its enzymatic activity. In addition, the supporting loop on the C-terminal domain was supported by the NTD of the adjacent subunit (13). Therefore, the disruption of the hexameric state of nsp15 can indirectly impact the two reactive histidines, namely His-234 and His-249, thus abolishing its enzymatic activity.
In this study, we performed oligomeric and functional analyses of PDCoV nsp15 to elucidate the mechanism underlying the function of NendoU during nidovirus replication and to identify potential drug targets for controlling PDCoV. Our data indicate that PDCoV nsp15 exists as a dimer and monomer rather than as a hexamer in solution and suggest that the dimeric form of PDCoV nsp15 may represent an intermediate in the evolutionary process of nidoviruses. Moreover, our findings further verified the conservatively catalytic sites among NendoU and provided a new hypothesis for the NendoU evolutionary process based on their different oligomeric forms.

Predicted three-dimensional structure of PDCoV nsp15
The predicted three-dimensional structure of PDCoV nsp15 is a homohexamer; two monomers interact with each other and form a hexamer with the NTDs facing toward the inside and the C-terminal domains (CTD) facing outward. In addition, the hexamer resembles a triangular scaffold to stand still (Fig. 1A); the basic oligomeric unit is a dimer, and the N-terminal domain is important for the oligomerization (Figs. 1B and 3B). In addition, the monomer consists of three parts: the intact N-terminal domain (Asn-1 to Arg-167); the C-terminal domain (Pro-193 to Gln-327); and a middle linker domain (Tyr-168 to Thr-192) (Fig. 1C).

Oligomerization of PDCoV nsp15 differs from that of other CoVs nsp15s
The size-exclusion chromatography results of PDCoV nsp15 show that it is not a hexamer but mainly a monomer and a dimer (Fig. 2, A and B). Thus, PDCoV (Deltacoronavirus), SARS-CoV (Betacoronavirus), and PEDV (Alphacoronavirus), as three representative viruses of the Coronaviridae, were selected among all coronaviruses for nsp15 oligomerization analysis among the whole coronaviruses to compare their different oligomeric stations in vitro. In our work, low production of WT protein was also observed due to its bacterial toxicity, which has been previously reported (20,33). Previous studies have shown that the key catalytic site histidine does not participate in stabilizing functional oligomerization in coronaviruses and arteriviruses (20,33). According to the sequence alignment results, the catalytic sites of His-219, His-234, and Lys-269 in PDCoV are highly conserved compared with those in SARS-CoV and PEDV nsp15s. Furthermore, several substrate-binding sites present in PDCoV nsp15 (Thr-273, Asp-276, Asp-305,
The size-exclusion chromatography results revealed that the oligomeric forms of PDCoV nsp15 were a dimer and a monomer (ϳ78.5 and 33.3 kDa, respectively). In contrast, both PEDV nsp15 and SARS-CoV nsp15 were hexamers (ϳ228.6 kDa) ( Fig.  2A). The peak elution volumes of PDCoV nsp15 were ϳ15.43 and 13.92 ml. PEDV and SARS-CoV nsp15s showed similar peak elution volumes of ϳ12.07 ml, which were reflected by the bio-standard protein elution results (Fig. 2B). To further verify the specific distributions of these oligomers, we performed an analytical ultracentrifugation experiment. The hexamer proportions of PEDV and SARS-CoV nsp15s were ϳ81.2 and 91.4%, with predicted molecular masses of ϳ187.0 and 218.0 kDa, respectively (Fig. 2D). In the case of PDCoV nsp15, dimers and monomers occupied ϳ36.1 and 47.5%, with a predicted mass of ϳ81.1 and 37.5 kDa, respectively (Fig. 2E). The calculated masses and the sedimentation coefficient (s 20, w ) data are shown in Table 1. The proteins used in the size-exclusion chromatography experiment were then visualized by SDS-PAGE (Fig. 2C).
In conclusion, our result verified that the oligomerization of PDCoV nsp15 was different from that of other CoV nsp15s.

NTD of PDCoV nsp15 is important for its dimerization
The three-dimensional structure of SARS-CV nsp15 (PDB code 2rhb) consists of six homomonomers and forms a compact homohexamer via the interaction of one NTD (1S-27I) with the other CTD (Fig. 3A). The predicted structure of PDCoV nsp15 shows that it is also a homohexamer and that part of the NTD is involved in its higher oligomerization (Figs. 1A and 3B).
To verify that part of the NTD region (1N-27N) participates in the dimerization of PDCoV nsp15, we analyzed a PDCoV nsp15 mutant with its NTD (1N-27N) truncated, called

Evolution of nidovirus endoribonuclease
protein elution results (Fig. 2B). The AUC results showed that both dimers and monomers were present in PDCoV nsp15 (N-terminal truncated), and they represented ϳ19.8% and 61.0% with predicted masses of ϳ72.9 and 37.1 kDa, respectively. In addition, few tetramers (ϳ136.0 kDa) were also observed ( Fig. 4D). Compared with those in PDCoV nsp15 (H234A), the percentage of monomers in PDCoV nsp15 (N-terminal truncated) increased by ϳ13.5%, and the percentage of dimers decreased by ϳ16.3% (Fig. 4D). Additionally, PDCoV nsp15 (N-terminal replaced) had a high proportion of monomers (ϳ72.9%) with a predicted mass of ϳ36.2 kDa and few other oligomers (Fig. 4E). The calculated masses and the sedimentation coefficient (s 20, w ) data of the AUC are shown in Table 1. The proteins eluted in the size-exclusion chromatography experiment were then visualized by SDS-PAGE (Fig. 4C).
The fact that PDCoV nsp15 NTD is important for its dimerization has been validated in PDCoV nsp15 (N-terminal truncated). The reason to explain why PDCoV nsp15 is a dimer

Evolution of nidovirus endoribonuclease
and monomer but not a hexamer remains unclear because the oligomeric form of PDCoV nsp15 (N-terminal replaced) remains almost the same as that of PDCoV nsp15 (H234A) in vitro, and it suggests that there should be other regions impeding the hexamerization in PDCoV nsp15. Thus, the following experiments were performed.
In conclusion, this completely oligomeric change found in SARS-CoV nsp15 Asp-259 -Phe-279 indicated that the region (251H-261V) of PDCoV nsp15 could be a reason to explain why PDCoV nsp15 fails to form a hexamer. In addition, both of the oligomeric forms in SARS-CoV nsp15 (N-terminal replaced) and PDCoV nsp15 (N-terminal replaced) remain the same as those of SARS-CoV nsp15 (H234A) and PDCoV nsp15 (H234A), which is additional evidence to prove that NTD is important for their dimerization but not the hexamerization.

PDCoV nsp15 is active as a dimer
In arterivirus, the biofunctional unit of the PRRSV NendoU is a dimer. Thus, the unit differs from that of MHV and SARS-CoV NendoUs; both of them function as hexamers (27,33). The predicted structure of PDCoV nsp15 contains three conserved catalytic sites (His-219, His-234, and Lys-269) and four binding sites (Thr-273, Thr-276, Asp-305, and Tyr-323) that form a pocket for RNA substrate binding (Fig. 6A), which was also reflected by the sequence alignment results (Fig. 3C). Furthermore, the interaction between the two monomers in PDCoV nsp15 was depicted; the NTD on monomer A interacts with the supporting loop on CTD of monomer B. The interactions between the Asp-225, Ser-227, Pro-228, Val-229, Ile-230, and Gly-232 residues on the active loop and the Asn-265, Ser-267, Ser-268, Lys-269, and Cys-272 residues on the supporting loop are depicted. Furthermore, these extensive interactions between the supporting loop and the adjacent monomer NTD were also analyzed and determined in LIGPLOT (35). Interactions between the residues Asn-258, Cys-259, Val-260, Ser-268, Asn-270 and Val271 from the supporting loop and residues Lys-9, Tyr-10, Tyr-11, Lys-12, Asn-13, and Cys-14 of the adjacent monomer NTD are observed (Fig. 6B). To further verify the enzymatic activity of these sites, FRET experiments were performed, and the results confirmed that these positions were the functional sites on PDCoV nsp15. The assay results showed the decreased activity of all PDCoV nsp15 mutants (H219A, H234A K269A, D276A, and the N-terminal truncation). PDCoV nsp15 mutants (T273A, D305A, and Y323A) could not be obtained due to insolubility. Excision by the PDCoV nsp15 mutants (H219A, H234A, and the N-terminal truncated) was dramatically reduced compared with that by WT PDCoV nsp15, and H219A retained approximately one-third of the activity of WT PDCoV nsp15. In addition, the D276A mutant showed a smaller decrease in catalytic activity compared with other mutants. Our experimental results can be explained based on the positions of the predicted catalytic sites within the predicted three-dimensional structure of PDCoV nsp15 (Fig.  6A). The H234A mutant shows the greatest activity decrease among all of these mutants because it lies in the center of the enzymatic groove (Fig. 6A). Therefore, compared with the other two catalytic sites, His-234 may play an important role in the interaction with the RNA substrate. The loss of activity observed in the PDCoV nsp15 (N-terminal truncated) mutant Evolution of nidovirus endoribonuclease may be due to the dimeric disruption of PDCoV nsp15, which may affect its normal biological function (Fig. 6B). Among the SARS-CoV nsp15 mutants, the nsp15 Asp-259 -Phe-279 mutant displayed the highest enzymatic activity, and its activity decreased by three-fourths compared with that of WT PDCoV nsp15. The second most active mutant was the N-terminal truncated mutant, followed by the N-terminal replaced mutant, and then the H234A mutant (Fig. 6C). The proteins from the enzymatic assay were then visualized by SDS-PAGE (Fig. 6D). Taken together, the enzymatic assay results further verified that the NTD and His-251 to Val-261 regions were important for the dimerization of PDCoV nsp15 and that disruption of its The calculated molecular masses are indicated by a red vertical line for the N-terminal truncated mutant with three peak volumes corresponding to predicted masses of ϳ112.4, 69.4, and 27.9 kDa; and the predicted masses of N-terminal replaced mutant were ϳ87.6 and 35.6 kDa; they were determined by fitting to the calibration curve as described for B. B, calculated molecular masses of the nsp15 protein peaks with the values obtained for known calibration standards (Bio-Rad and GE Healthcare). The calculated molecular mass of nsp15 peaks was determined by fitting to the calibration curve (K av ϭ volumes of elution (V es /24)). C, SDS-PAGE analysis of PDCoV nsp15 (N-terminally truncated) and PDCoV nsp15 (N-terminally replaced). The elution volume is labeled as described for A. Molecular mass markers are shown. D and E, sedimentation velocity analysis of PDCoV nsp15 (N-terminally truncated) (red), PDCoV nsp15 (H234A) (black), and PDCoV nsp15 (N-terminally replaced) (green) with their major peaks is depicted in D and E. The sedimentation coefficient (s 20, w ) and the calculated masses are shown in Table 1.

Evolution of nidovirus endoribonuclease
dimerization could decrease its enzymatic activity, which suggests that the biofunctional unit of PDCoV nsp15 may be a dimer in vitro. In addition, the predicted catalytic and binding sites on PDCoV nsp15 could impact its enzymatic activity to different degrees.

Presumption for the evolution of Nidovirales NendoU
As the counterpart of NendoU, XendoU has been reported to be a monomer (36). According to the three-dimensional structural alignment of XendoU and SARS-CoV nsp15 (Fig. 7B), a similar structural feature on the CTD was found and reflected by the highly conserved catalytic sites of the active loop, which have also been found in other NendoUs, such as PRRSV nsp11 and PDCoV nsp15 (Fig. 7A). We hypothesize that their active loops are stabilized by the supporting loops in PRRSV nsp11, PDCoV, and SARS-CoV nsp15s. However, XendoU, could be active as a monomer via the undergirding support of residues (Lys-113 to Lys-125 and Asn-133 to Phe-144), and the additional upper residues of Gln-145 to Cys-158 could also provide stability to hold up the active loop (His-162 to His-178) (13);  Table 1. F, size-exclusion experiment result of SARS-CoV nsp15 (N-terminally replaced) (green) with the elution volume of ϳ11.84 ml and the calculated molecular mass of ϳ259.4 kDa were determined by fitting to the calibration curve as described in C.

Evolution of nidovirus endoribonuclease
these interactions and the predicted supporting loop on Xen-doU are depicted (Fig. 7A). In PRRSV nsp11, the active loop (His-129 to His-144) and the supporting loop (Val-162 to Thr-179) stand out from the center compared with these features in XendoU; thus, the supporting loop alone was not sufficient for its stabilization, and dimerization was required for the supporting loop to make PRRSV nsp11 functional. In addition, the same condition was found in PDCoV nsp15 with the interaction between the active loop (His-219 to His-234) and the supporting loop (Gln-257 to Thr-273) (Fig. 7A). In addition, the active loop (His-234 to His-249) and the supporting loop (Lys-276 to Ile-295) were also found in SARS-CoV nsp15. However, PDCoV nsp15 was an exception for being composed of dimers and monomers among other nsp15s ( Fig. 2A).
Thus, we hypothesize that PDCoV nsp15 is an intermediate during the NendoU evolutionary process of the development of high-order oligomerization to better exert its function via structural analysis (Fig. 7C).

Discussion
The newly emerged PDCoV belongs to the Deltacoronavirus family, whose members can cause similar clinical symptoms to those of PEDV. The highest mortality rate is observed in suckling and weaned pigs with no effective vaccination strategies proposed (4,6). The large-scale outbreak of PDCoV first emerged in Ohio and Indiana in 2014, and then PDCoV immediately spread across the United States, Canada, China, and Thailand, causing serious economic losses to the pork industry worldwide (37,38). In addition, few studies on PDCoV nsp15 have been reported; thus, to clarify its biofunction, three-dimensional structure or oligomeric distribution could help us to better understand its exact function in coronaviruses.
NendoU is highly conserved for specifically recognizing the U-pyrimidine and producing 3Ј-5Ј-phosphate products (24). In addition, the conserved biofunction of NendoU relies on the conserved catalytic sites (two histidines and one lysine) (Fig.  6A). The stability of the catalytic center in NendoU is the primary requirement to maintain its natural enzymatic capacity, and any degrees of its structural disruption can directly or indirectly destroy its enzymatic activity. The basic biofunctional unit in SARS-CoV nsp15 is a hexamer (33), and it includes a dimer of trimers with the NTDs in the center and the CTDs facing outward, allowing them to possess six active sites to better exert their function in SARS-CoV and MHV nsp15 (27,31,39). In addition, for PRRSV nsp11, the dimer is an active unit, and the disruption of its dimerization can markedly impair its endoribonuclease activity (20). Here, we found that both mono-mers and dimers were present in PDCoV nsp15. In addition, decreased enzymatic activity was observed in the PDCoV nsp15 (N-terminal truncation) mutant (Fig. 6C), which indicates that it functions as a dimer. Furthermore, the SARS-CoV nsp15 Asp-259 -Phe-279 mutant showed a completely oligomeric change from hexamers to a mixture of monomers and dimers, with the monomer predominating in solution. However, SARS-CoV nsp15 (N-terminal replaced) still remains as a hexamer (Fig. 5F), which needs to be further verified in vivo via reverse genetics system. An alternative hypothesis that the residues required for hexamer formation were lost or mutated during the evolutionary process in PDCoV nsp15 is possible, because the region (His-251-Val-261) is 10 amino acids shorter than the equivalent region (Asp-259 -Phe-279) on SARS-CoV nsp15. In conclusion, the disruption of the oligomerization can impair the normal enzymatic activity of NendoU. In addition, further investigations of nsp15s need to be performed to verify whether their different oligomeric forms impact their biofunction or the interaction with the RTCs and other nsps in vivo by using the reverse genetics or replicon systems. In this study, we attempted to screen the PDCoV nsp15 crystal under various conditions. Unfortunately, we were unable to obtain the structure.
As the counterpart of NendoU, XendoU is also an endoribonuclease and was purified from Xenopus laevis oocyte nuclear extracts (29). A previous study revealed that the oligomeric state of XendoU was monomeric, and the three-dimensional structure of XendoU showed that it adopted a different oligomeric pattern to that of NendoU. The support of residues (Lys-113-Lys-125 and Asn-133-Phe-144) undergirding and the additional upper residues (Gln-145-Cys-158) could also provide stability to hold up the active loop (His-165-His-178) with the predicted supporting loop adjacent to its active loop to form into a compact structure, with a globular shape (13). In addition, the highly conserved catalytic sites in XendoU (His-162, His-178, and Lys-224), are also found in PRRSV nsp11, PDCoV, and SARS-CoV nsp15s on the three-dimensional structural level (Fig. 7A), serving as additional evidence of their close relationship at an evolutionary aspect. The conservative catalytic residues may provide a novel target for drug design against nidoviruses in the future.
Regarding the origin of NendoU, we hypothesize that nidoriviruses may acquire this gene from XendoU in eukaryotic cells, and this gene may evolve in the virus to better adapt to the outside environment, a possibility that has already been put forward in previous studies (22,40). The different oligomeric

Evolution of nidovirus endoribonuclease
characteristics of XendoU and NendoU further support this hypothesis. In eukaryotic cells, XendoU exists as a monomer, and in the arterivirus PRRSV nsp11, it exists mainly as a dimer.
In PDCoV nsp15, we found it was a mixture of monomers and dimers, and in several Alpha-, Beta-, and Gammacoronavirus nsp15s, the proteins have been identified as hexamers (25, 27,  33,34). Furthermore, according to the molecular clock analysis, the times to which these four coronaviruses could be traced were ϳ2400 BCE, 2800 BCE, 3300 BCE, 3000 BCE, corresponding to Alphacoronavirus, Gammacoronavirus, Betacoronavirus, and Deltacoronavirus, respectively (4). PDCoV is the second oldest among the four genera. This result reflects the possible evolutionary course among all coronavirus NendoUs. PDCoV nsp15 is the intermediate status in the transition process of forming a hexamer from the dimeric form in PRRSV nsp11, and all the NendoUs have the same ancestor, namely XendoU (Fig. 7C). In addition, a relevant study showed that approximately half of the proteins could interact with other copies of themselves and assemble into homomeric complexes, which was often assumed to be a functionally beneficial result of evolutionary selection (11).

Three-dimensional structure prediction
The predicted three-dimensional structure of PDCoV nsp15 was accomplished in the SWISS-MODEL website (https:// swissmodel.expasy.org/), and then the structural figures were generated with the PyMOL molecular visualization system (Schrödinger).

Gene cloning and plasmid constructs
The genes encoding nsp15s were cloned from the PDCoV strain CHN-HN-2014 (GenBank TM accession number KT336560) and the PEDV FJZZ strain (GenBank TM accession number KC140102.1). A codon-optimized SARS-CoV nsp15 gene (GenBank TM accession number NP_828872.1; gi:29837507) was synthesized by GenScript Corp. (GenScript, Nanjing, China). The PDCoV nsp15 mutants (H234A, N-terminal truncated, N-terminal replaced), the SARS-CoV nsp15 mutants (H234A, N-terminal truncated, and N-terminal replaced), and the PEDV nsp15 mutant (H241A) used in the analytical ultracentrifugation and size-exclusion chromatography experiments were separately cloned into the pET-42b(ϩ) vector with the NdeI and XhoI restriction sites, and a His 6 tag was added at the C terminus. The alanine-substituted PDCoV nsp15 mutants (H219A, H234A, K269A, T273A, D276A, D305A, and N-terminal truncated, each with a C-terminal GST tag) were cloned into the pGEX-6p-1 vector by homologous recombination with a C-terminal GST tag for the enzymatic assays. In addition, pET-30a(ϩ) and pET-28a(ϩ) containing the WT PDCoV nsp15 gene with a C-terminal His 6 tag were transformed into E. coli DE3 to obtain the WT PDCoV nsp15. For the oligomerization experiment, several important sequences of Phe-145-Ile-164 and His-251-Val-261 on PDCoV nsp15 were replaced by the equivalent sequences of Thr-152-Phe-179 and Asp-259 -Phe-279 on SARS-CoV nsp15 to create SARS-CoV nsp15 Thr-152-Phe-179 and SARS-CoV nsp15 Asp-259 -Phe-279 , respectively. These constructs, each with a His 6 tag at the C terminus, were cloned into pET-42b(ϩ) after the digestion of the plasmid with the NdeI and HindIII restriction enzymes. All of the point mutations and sequence substitution mutations were constructed by the overlapping PCR technique and were confirmed by the GenScript Co.

Protein expression and purification
The recombinant plasmids described above were transformed into the E. coli strain Trans BL21 (DE3) pLysS (Beijing TransGen Biotech, Co., Ltd.). The cells transformed with plasmids encoding proteins with GST tags were cultured at 37°C in LB medium containing 50 g/ml ampicillin. LB medium containing 50 g/ml kanamycin was used for the cells expressing proteins with His 6 tags. When the OD 600 value of the cell density reached 0.6 -0.8, the cells were induced with 0.8 mM isopropyl ␤-D-1-thiogalactopyranoside. To maximize the protein yield, the cells expressing proteins with GST tags were cultured at 37°C for an additional 4 h. Protein expression and purification were conducted according to our previously reported procedure (19).
The harvested proteins were further purified on a Superdex 200 gel-filtration column (GE Healthcare) and were washed with buffer B 2 (20 mM Tris-HCl and 200 mM NaCl, pH 7.4) when the volume of the filtered protein was less than 2.0 ml in the condensing column for centrifugation at 4°C. The proteins were flash-frozen in liquid nitrogen and stored at Ϫ80°C until used in biochemical analyses. The concentrations of PDCoV, SARS-CoV, and PEDV nsp15 mutant proteins were determined based on the absorbance values of the samples at 280 nm (A 280 ) using a DS-11 FXϩ spectrophotometer combined with a highly sensitive 1-l UV-visible absorbance spectrophotometer (DeNovix, Inc.).

Size-exclusion chromatography assay
The PDCoV nsp15 mutants (H234A, N-terminal truncated, and N-terminal replaced), the SARS-CoV nsp15 mutants (H234A, N-terminal truncated, N-terminal replaced, and Asp-259 -Phe-279), and the PEDV nsp15 mutant (H241A) were purified through a nickel-nitrilotriacetic acid high-affinity filter and then passed through a HiLoad Superdex 200 gel-filtration column (GE Healthcare) in elution buffer (20 mM Tris-HCl, 200 mM NaCl, pH 7.4) at a flow rate of 0.8 ml/min at 4°C. For the analysis of oligomerization, ϳ1 mg of each protein was subjected to chromatography on a Superdex 200 10/300 GL column in the same elution buffer described above at a flow rate of 0.5 ml/min at 4°C, and the eluted proteins were then assessed by SDS-PAGE analysis. Size-exclusion standards were separated on the column following our previously reported procedure (19). The obtained data were analyzed using Origin 8.0 software. The predicted weight-averaged molar masses were calculated using DNASTAR (version 7.1) software.

Analytical ultracentrifugation assay
To confirm the masses of the individual proteins, sedimentation velocity analysis was conducted in an XL-A model centrifuge (Proteome Lab) at 18°C and 45,000 rpm in 400-l double-sector cells. The sedimentation boundary was monitored every 3 min at a wavelength of 280 nm, resulting in 110 scans. The collected data were analyzed using Sedfit software with the model-based distribution of Lamm equation solutions c(s). The data obtained from the size-exclusion chromatography experiments and sedimentation velocity analysis were transformed into curves for analysis in Origin 8.0 software. DNASTAR (ver-Evolution of nidovirus endoribonuclease sion7.1) software was used for protein weight-averaged molar mass prediction.

Enzymatic activity assay
The endoribonuclease activity assay was performed as follows: the WT PDCoV nsp15 and mutant proteins (H219A, H234A, K269A, D276A, and N-terminal truncated) with GST tags at their C-terminal ends and the GST protein as a negative control, each at a concentration of 2 M, were mixed with 1 M RNA substrate (5Ј,6-carboxyfluorescein-dA-rU-dA-dA-6-carboxy-N,N,NЈ,NЈ-tetramethylrhodamine-3Ј, purchased from GenScript Corp.) (32) in the reaction buffer (50 mM KCl, 50 mM HEPES, pH 7.5), 5 mM MnCl 2 , and 1 mM DTT dissolved in water with 0.1% diethyl pyrocarbonate) and incubated at 25°C for 30 min. The same protocol was used to assess the endoribonuclease activity of the SARS-CoV nsp15 mutants (H234A, N-terminal truncated, N-terminal replaced, and Asp-259 -Phe-279) with the His 6 tag. The CFX96 Touch TM Real-Time PCR Detection System (Bio-Rad) was used to quantify the 5Ј,6-carboxyfluorescein value of each protein for 30 min at wavelengths ranging from 492 to 518 nm every 30 s. Each protein was assayed three times, and the obtained data were analyzed via GraphPad Prism software 5.0 (GraphPad Software Inc., CA). The values obtained in the triplicate measurements are shown. All of the proteins tested in this assay were also analyzed by SDS-PAGE.

Sequence alignment and phylogenetic tree analysis
The genes encoding SARS-CoV (JF29292), PEDV (KU558702), and PDCoV nsp15s (KP757891) were separately cloned from the full-length genomes of these three viruses, and the amino acid prediction results were obtained by analyzing the gene sequences using DNASTAR (version 7.1) software. The amino acid sequences of SARS-CoV, PEDV, and PDCoV nsp15s were prealigned on the website server of Clustal Omega (https:// www.ebi.ac.uk/Tools/msa/clustalo/) 4 (41), and the resulting file was then uploaded to the ESPript 3 website (http:// espript.ibcp.fr) 4 (42) to produce the final alignment result. MEGA 6.0 software was used to build the phylogenetic tree to visualize the evolutionary relationships between Nidovirales and Coronaviridae via the distance-based neighbor-joining method. Bootstrap values were calculated from 1000 replicates of the alignment. The different subgenotypes are indicated.
Author contributions-A. Z. and Y. S. data curation; A. Z. software; A. Z. and Y. S. formal analysis; A. Z. validation; A. Z. visualization; A. Z. writing-original draft; A. Z., Z. S., J. S., and Q. X. project administration; Y. S. and G. P. supervision; Y. S. and G. P. funding acquisition; Y. S. and G. P. writing-review and editing; G. W. methodology; L. F., S. X., Z. F. F., and G. P. resources.