Subcellular Localization and Functional Characterization of GII.4 Norovirus-Encoded NTPase

ABSTRACT The genotype II.4 (GII.4) variants of human noroviruses (HuNVs) are recognized as the major agent of global gastroenteritis outbreaks. Due to the lack of an efficient cell culture system for HuNV propagation, the exact roles of HuNV-encoded nonstructural proteins (including Nterm, NTPase, P22, VPg, Pro, and RdRp) in viral replication or pathogenesis have not yet been fully understood. Here, we report the molecular characterization of the GII.4 HuNV-encoded NTPase (designated GII-NTPase). Results from our studies showed that GII-NTPase forms vesicular or nonvesicular textures in the cell cytoplasm, and the nonvesicular fraction of GII-NTPase significantly localizes to the endoplasmic reticulum (ER) or mitochondria. Deletion analysis revealed that the N-terminal 179-amino-acid (aa) region of GII-NTPase is required for vesicle formation and for ER colocalization, whereas the C-terminal region is involved in mitochondrial colocalization. In particular, two mitochondrion-targeting domains were identified in the C-terminal region of GII-NTPase which perfectly colocalized with mitochondria when the N-terminal region of GII-NTPase was deleted. However, the corresponding C-terminal portions of NTPase derived from the GI HuNV did not show mitochondrial colocalization. We also found that GII-NTPase physically interacts with itself as well as with Nterm and P22, but not VPg, Pro, and RdRp, in cells. The Nterm- and P22-interacting region was mapped to the N-terminal 179-aa region of GII-NTPase, whereas the self-assembly of GII-NTPase could be achieved via a head-to-head, tail-to-tail, or head-to-tail configuration. More importantly, we demonstrate that GII-NTPase possesses a proapoptotic activity, which can be further enhanced by coexpression with Nterm or P22. IMPORTANCE Despite the importance of human norovirus GII.4 variants in global gastroenteritis outbreaks, the basic biological functions of the viral nonstructural proteins in cells remain rarely investigated. In this report, we focus our studies on characteristics of the GII.4 norovirus-encoded NTPase (GII-NTPase). We unexpectedly find that GII-NTPase can perfectly colocalize with mitochondria after its N-terminal region is deleted. However, such a phenomenon is not observed for NTPase encoded by a GI strain. We further reveal that the N-terminal 179-aa region of GII-NTPase is sufficient to mediate (i) vesicle formation, (ii) ER colocalization, (iii) the interaction with two other nonstructural proteins, including Nterm and P22, (iv) the formation of homodimers or homo-oligomers, and (v) the induction of cell apoptosis. Taken together, our findings emphasize that the virus-encoded NTPase must have multiple activities during viral replication or pathogenesis; however, these activities may vary somewhat among different genogroups.

N oroviruses are members of the family Caliciviridae and contain a positive-sense single-stranded RNA genome (1). Noroviruses can be classified into seven genogroups, designated genotype I (GI), GII, GIII, GIV, GV, GVI, and GVII, based on their genome sequence similarity (1, 2). Among them, viruses from GI, GII, and, occasionally, GIV can infect humans and cause acute gastroenteritis (3). Viral strains in each genogroup can be further divided into distinct genotypes according to their nucleotide and amino acid sequence diversity. In particular, variants of the GII.4 genotype have been recognized as the most widespread strains that result in pandemic outbreaks since the mid-1990s (4)(5)(6).
Despite the fact that human noroviruses (HuNVs) are now the leading cause of acute nonbacterial gastroenteritis (1, 7), multiple aspects regarding the mechanisms of viral genome replication or viral pathogenesis are still poorly understood. The poor understanding of HuNV replication and pathogenicity is due to the lack of an efficient and reliable cell culture system or an animal system for multicycle replication of HuNV. Although progress has been made in recent years regarding the identification of human permissive target cells for HuNV infection in vitro (8,9) and the use of a plasmid-based replicon for HuNV propagation (10,11), these newly established model systems have not yet been combined for investigations of the molecular mechanisms associated with virus-host interactions. Currently, most of our knowledge about HuNV biology is derived mainly from studies with animal caliciviruses such as murine norovirus (MNV; a strain of GV noroviruses).
The RNA genome of HuNV is about 7.6 kb in length and consists of three conserved open reading frames (ORF), including ORF1, ORF2, and ORF3 (12). Like all RNA genomes of caliciviruses, the 5= end of HuNV genomic RNA is covalently linked to a virus-encoded protein known as VPg (viral protein, genome linked), and the 3= end contains a poly(A) tail (1, 12). ORF1 in the viral genome encodes a large nonstructural polyprotein that can be further proteolytically cleaved into six mature proteins, including Nterm (N-terminal nonstructural protein, or NS1-2), NTPase (or NS3), P22 (or NS4), VPg (or NS5), Pro (proteinase, or NS6), and RdRp (RNA-dependent RNA polymerase, or NS7). ORF2 and ORF3 encode the major capsid protein (VP1) and the minor capsid protein (VP2), respectively. According to an integrated model of noroviral genome replication, the incoming viral RNA directly acts as an mRNA for viral protein synthesis during infection (13,14). Following the synthesis of the ORF1 polyprotein precursor, Pro is responsible for the cleavage of the polyprotein into six mature nonstructural proteins. These viral nonstructural proteins may subsequently participate in the assembly of a replication complex where RdRp functions to catalyze the synthesis of the negative-sense RNA intermediate, genomic, and subgenomic RNA in cells. Although the principle functions of VPg, Pro, and RdRp in viral replication have been well documented, the exact roles of the other nonstructural proteins, including Nterm, NTPase, and P22, in cells still need further investigation.
For MNV or Norwalk virus (the prototype of GI noroviruses), Nterm and P22 previously were implicated in the remodeling of cell membranes, which may facilitate the assembly of the replication complex in the host cell (14)(15)(16)(17)(18). The NTPase protein of noroviruses is known to share sequence similarity with picornavirus 2C protein and the helicase of superfamily 3 (SF3) (19). However, biochemical characterization of the NTPase protein encoded by Southampton virus (SHV; a strain of GI noroviruses) revealed that the viral NTPase has both NTP-binding and NTP hydrolysis activities but lacks a helicase activity (19). More recently, Cotton et al. showed that when expressed individually, NTPase derived from MNV or Norwalk virus could induce the formation of vesicular structures in cells (20). Although they did not find colocalization of these vesicular structures with any particular organelle-specific markers, their studies suggested that the viral NTPase was dynamically associated with lipid membranes and microtubules. Despite the fact that the potential functions of these norovirus-encoded nonstructural proteins have been increasingly investigated, most of the published studies focused only on MNV or GI noroviruses (14). To date, very little information is available about the subcellular localization and function of the GII nonstructural proteins in cells.
In this report, we showed that the GII.4 nonstructural proteins expressed individually in cells displayed somewhat different intracellular localization patterns from those of the previously reported counterparts of GI and GV noroviruses (11,(20)(21)(22)(23). We then performed a detailed characterization of the GII.4 NTPase, referred to here as GII-NTPase. Confocal immunofluorescence analysis revealed that GII-NTPase could form vesicular or nonvesicular structures in the cytoplasm, and the nonvesicular fraction of GII-NTPase showed a significant colocalization with the endoplasmic reticulum (ER) or mitochondria. The critical regions of GII-NTPase required for vesicle formation, ER colocalization, and mitochondrial colocalization were mapped in this study. Additionally, GII-NTPase could form homodimers/ oligomers and physically interact with Nterm and P22 in the cell. Functional analysis revealed that GII-NTPase had a proapoptotic activity, which could be further enhanced by coexpression with Nterm or P22. Therefore, our findings strongly suggest that GII-NTPase is a multifunctional protein in cells.

RESULTS
Expression and subcellular localization of the GII.4 nonstructural proteins. To determine the subcellular localization of the GII.4 nonstructural proteins, a full-length cDNA clone of a GII.4 HuNV strain, HuNV/GII.4/YJB1/2009/Chiayi, isolated from Chang-Gung Memorial Hospital at Chiayi (Taiwan), was used as the starting material for the DNA cloning. The cDNA fragments corresponding to the coding regions of Nterm, NTPase, P22, VPg, Pro, and RdRp were inserted into an N-terminal FLAG-tagging expression vector. The resultant expression plasmids were designated pCMV-F-Nterm, pCMV-F-NTPase, pCMV-F-P22, pCMV-F-VPg, pCMV-F-Pro, and pCMV-F-RdRp. When these expression plasmids were individually transfected into 293T cells, all of these nonstructural proteins were expressed normally in 293T cells and displayed predicted molecular masses on SDS-PAGE ( Fig. 1A and B). The detected molecular masses of F-Nterm, F-NTPase, F-P22, F-VPg, F-Pro, and F-RdRp were 42,45,22,18,20, and 60 kDa, respectively. We subsequently analyzed their intracellular localization in human melanoma A7 cells using confocal microscopy (Fig. 1C). In confocal immunofluorescence analysis, we found that F-Nterm, F-NTPase, and F-P22 were exclusively localized in the cytoplasm, whereas F-VPg, F-Pro, and F-RdRp were distributed in both the cytoplasm and the nucleus (Fig. 1C). It is worth noting that F-Nterm consistently produced thread-like structures in the cell cytoplasm, whereas F-NTPase could form circular vesicular structures (Fig. 1C). In this study, we focused our attention on the analysis of the GII.4 norovirus-encoded NTPase, designated GII-NTPase, which consists of 366 amino acids (aa).
The nonvesicular fraction of GII-NTPase significantly localizes to ER or mitochondria. Before we further characterized the subcellular localization and biological functions of GII-NTPase, we initially investigated whether adding a FLAG tag to the N terminus of GII-NTPase interfered with the native structure of the protein. Therefore, an anti-NTPase antibody was generated using bacterially expressed GII-NTPase (aa 49 to 366) as an immunogen and was used in immunofluorescence staining experiments. In the experiments, A7 cells that were transfected with pCMV-F-NTPase were dually labeled with the anti-NTPase and anti-FLAG antibodies ( Fig. 2A). As shown in Fig. 2A, full colocalization of the anti-FLAG and anti-NTPase signals was observed in transfected A7 cells by confocal microscopy, suggesting that the addition of a FLAG tag to GII-NTPase does not significantly affect the protein structure.
To further analyze the subcellular localization of GII-NTPase, A7 cells expressing F-NTPase were dual stained with an anti-FLAG antibody for F-NTPase and antibodies for organelle-specific markers, including protein disulfide isomerase (PDI; an ER marker), Golgi matrix protein of 130 kDa (GM130; a cis-Golgi marker), trans-Golgi network protein of 46 kDa (TGN46; a trans-Golgi marker), and early endosome antigen 1 (EEA1; an endosome marker). Additionally, MitoTracker (Mito), a mito-chondrial labeling dye, was used to stain mitochondria in the cell. Confocal microscopic analysis revealed that F-NTPase substantially colocalized with the signals of PDI (Fig. 2B, row a) or MitoTracker (Fig. 2B, row b) but not with those of GM130, TGN46, and EEA1 (Fig. 2B, rows c to e). It should be noted, however, that the vesicle structure formed by F-NTPase did not show any significant colocalization with PDI or MitoTracker (Fig. 2B, a-5 and b-5).
Mapping the critical regions of GII-NTPase involved in vesicle formation, ER colocalization, and mitochondrial colocalization. We next attempted to map the critical regions of GII-NTPase essential for vesicle formation and for the colocalization of ER or mitochondria. Since little is known about the structure and function of GII-NTPase, a series of the N-terminal or C-terminal deletion constructs of F-NTPase were generated ( Fig. 3A and B). At the start, the expression plasmids encoding F-NTPase deletion constructs were transfected into A7 cells, and the transfected cells were dual stained with anti-FLAG and anti-PDI antibodies (Fig. 3C). Confocal microscopic analysis showed that the C-terminal-deleted mutants, including F-NTPase(1-271) and F-NTPase(1-179), still retained the ability to form vesicles and colocalize with PDI (Fig.  3C, rows a to c). However, the F-NTPase deletions lacking the N-terminal region, such as F-NTPase(96 -366) and F-NTPase(213-366), completely lost their ability to form vesicles and to colocalize with PDI ( Fig. 3C, rows d and e). These results suggested that the N-terminal 179-aa region of GII-NTPase is important for vesicle formation and ER colocalization in the cell.
When the transfected cells were dual stained with anti-FLAG antibody and Mito-Tracker, we found that full-length F-NTPase and F-NTPase(1-271) showed a similar extent of mitochondrial colocalization ( Fig. 4A and C, rows a and b). In contrast, further C-terminal deletion to aa 179, F-NTPase(1-179), completely abolished its localization to mitochondria (Fig. 4C, row c). Surprisingly, we found that two N-terminal deletion constructs, including F-NTPase(96 -366) and F-NTPase(213-366), in the cell colocalized perfectly with MitoTracker signals (Fig. 4C, rows d and e). These results were unexpected and prompted us to generate two more deletion constructs, F-NTPase(255-366) and F-NTPase(213-314) ( Fig. 4A and B). Confocal microscopy analysis showed that F-NTPase(255-366) still perfectly colocalized with mitochondria ( Fig. 4C, row g). However, F-NTPase(213-314) lost its ability to localize to mitochondria (Fig. 4C, row f). Based on the results described above, there are at least three important implications in the Besides the extreme C-terminal mitochondrion-targeting domain in GII-NTPase, we noticed that the mutant F-NTPase(1-271) partially localized to mitochondria to an extent comparable to that of full-length F-NTPase (Fig. 4C, compare rows a and b). The question therefore was raised as to whether F-NTPase(1-271) contained another mitochondrion-targeting domain. Since the N-terminal region might have an inhibitory activity toward the mitochondrial targeting of F-NTPase, we deleted the N-terminal 95-aa region from F-NTPase(1-271). Consistent with our hypothesis, high degrees of colocalization between F-NTPase(96 -271) and mitochondria were detected in confocal microscopic analysis (Fig. 4C, row h). Taken together, our findings proposed that GII-NTPase has two separate mitochondrion-targeting domains. Difference in the mitochondrial localization between GI-NTPase and GII-NTPase. As described above, GII-NTPase could localize to mitochondria in cells. We next investigated whether the NTPase protein derived from a GI norovirus strain, GI-NTPase, could also localize to mitochondria. The full-length GI-NTPase protein contains 363 amino acids and shares 52% identity (or 67% similarity) with GII-NTPase. When the GI-NTPase protein was expressed in A7 cells, vesicle-like structures induced by GI-NTPase could be observed in the cell; however, very limited colocalization between GI-NTPase and the signals of MitoTracker was detected (Fig. 5A, row a). Since GII-NTPase deletion constructs that lacks the N-terminal region, such as GII-NTPase(213-366) and GII-NTPase(255-366), gained the mitochondrial targeting function, the corresponding GI-NTPase deletions, including GI-NTPase(210 -363) and GI-NTPase(252-363), also were generated ( Fig. 5A and B). Unlike the GII-NTPase deletion mutants, GI-NTPase(210 -363) and GI-NTPase(252-363) displayed diffused distribution in both the nucleus and the cytoplasm and did not show any significant colocalization with mitochondria ( Fig. 5A, rows b and c). During the course of the experiments, we additionally found that GI-NTPase(252-363) was expressed at a much lower level than the full-length GI-NTPase and GI-NTPase(210 -363) (Fig. 5B). Currently, we do not know why GI-NTPase(252-363) was expressed at very low levels in transfected cells. It is possible that the specific deletion of GI-NTPase from aa 210 to 252 causes a decrease in mRNA stability, mRNA translation efficiency, or protein stability.
As shown above, the N-terminal 74-aa region of GII-NTPase(255-366) was demonstrated to be critical for supporting mitochondrial targeting. We generated another two domain-swapped mutants, GII (56)ϩGI and GII (32)ϩGI, which encompass the N-terminal 56 and 32 residues of GII-NTPase(255-366), respectively. Like the GII (74)ϩGI construct, both GII (56)ϩGI and GII (32)ϩGI also exhibited a wide spectrum of localization to mitochondria in transfected cells ( Fig. 5D and G). When the mitochondriontargeting ability of the NTPase chimeric mutants was evaluated, we found that the colocalization between these chimeric constructs and mitochondria was gradually reduced along with the shortening of the N-terminal region of GII-NTPase(255-366) in domain-swapped mutants ( Fig. 5D and F). The percentages of the type I and type II patterns (corresponding to high degrees of colocalization with the mitochondria) for GII-NTPase(255-366), GII (74)ϩGI, GII (56)ϩGI, and GII (32)ϩGI in transfected cells were found to be 100%, 57%, 38%, and 10%, respectively (Fig. 5F). Since no specific mitochondrial targeting signal could be identified in GII-NTPase(255-366) using the domain-swapping approach, our results suggested that the targeting of these NTPase chimeric constructs to the mitochondria is dependent upon their protein structure but not a contiguous linear sequence motif.
Interaction of GII-NTPase with itself and other nonstructural proteins. Due to the importance of the replication complex assembly during noroviral infection, we determined whether GII-NTPase physically interacts with itself or other nonstructural proteins. In the experiments, 293T cells first were cotransfected with the expression plasmid encoding a C-terminally myc-tagged NTPase(NTPase-myc) and the expression plasmid encoding F-Nterm, F-NTPase, F-P22, F-VPg, F-Pro, or F-RdRp (Fig. 6). Cell lysates of the transfected samples were immunoprecipitated using anti-myc antibody, and then the resultant immunoprecipitates were analyzed by Western blotting using anti-FLAG and anti-myc antibodies. We found that F-Nterm, F-NTPase, and F-P22, but not F-VPg, F-Pro, and F-RdRp, could be coimmunoprecipitated with NTPase-myc in the experiments (Fig. 6A, right). In confocal microscopic analysis, we also demonstrated that NTPase-myc specifically colocalized with F-Nterm, F-P22, and F-NTPase but not F-VPg, F-Pro, and F-RdRp in transfected A7 cells (Fig. 6B).

Mapping the interaction region of GII-NTPase with Nterm and P22.
To further map the critical region of NTPase responsible for the interaction with Nterm or P22, different deletion constructs of NTPase-myc (Fig. 7A) were cotransfected with the plasmid expressing F-Nterm or F-P22 into 293T cells. Immunoprecipitation was carried out using anti-FLAG antibody, and then the precipitated lysates were analyzed by Western blotting using anti-FLAG and anti-myc antibodies (Fig. 7B and D). Results from the Western blot analysis revealed that the full-length NTPase-myc, NTPase(1-271)myc, and NTPase(1-179)-myc, but not NTPase(96 -366)-myc and NTPase(213-366)-myc, could be coimmunoprecipitated with F-Nterm or F-P22 ( Fig. 7B and D, bottom). During the course of the experiments, we noticed that the levels of NTPase(96 -366)-myc and NTPase(213-366)-myc apparently were lower than those of other NTPase-myc deletion mutants in cell lysates ( Fig. 7B and D, top). The reduced levels of these two mutant proteins in the immunoprecipitation experiments were not due to their lower protein expression in cells. In fact, when the same transfected cells were lysed with Laemmli sample buffer, we found that all mutant proteins were expressed equally (data not shown), indicating that NTPase(96 -366)-myc and NTPase(213-366)-myc have reduced solubility in the immunoprecipitation assay buffer. In confocal microscopic analysis, we also demonstrated that NTPase(1-179)-myc, but not NTPase(96 -366), was able to interact with F-Nterm and F-P22 in transfected A7 cells ( Fig. 7C and E).
Since Nterm or P22 was capable of interacting with GII-NTPase as described above, we then investigated whether Nterm or P22 modulated the proapoptotic activity of GII-NTPase. Compared to the control cells, cells expressing Nterm or P22 alone did not significantly induce the cleavage of PARP and caspase 3 (Fig. 10B,  lanes 2 and 3). However, coexpression of Nterm or P22 with GII-NTPase in cells substantially augmented the levels of cleaved PARP and caspase 3 (Fig. 10B, lanes  5 and 6).

DISCUSSION
Although the genotype GII.4 strains of HuNV are the major agents of global gastroenteritis outbreaks in recent years, studies dealing with the biological functions of the GII.4 nonstructural proteins are seldom performed. Previous studies from GI or GV noroviruses have shown that all viral nonstructural proteins (including Nterm, NTPase, P22, VPg, Pro, and RdRp) were intimately associated with the viral replication complex in the cell cytoplasm during viral infections (21)(22)(23)(24)(25). However, here we showed that when expressed individually in cells, three GII.4 nonstructural viral proteins, including VPg, Pro, and RdRp, were distributed in both the cytoplasm and the nucleus (Fig. 1C). In fact, a GII.3 HuNV-based cell culture model has been previously established by Katayama et al. (10) using a full-length viral replicon DNA vector, and we noticed that RdRp, and probably also Pro, expressed from the GII.3-based replicon vector in COS7 cells was localized in both the cytoplasm and the nucleus (10). Although the main functions of the noroviral nonstructural proteins are for viral genome replication in the cytoplasm, the nuclear distribution of specific nonstructural viral proteins implies that they possess diverse biological actions in host cells. Actually, the nuclear localization of nonstructural viral proteins is not uncommon and has been described in a number of other related RNA virus families (26)(27)(28)(29). In the case of picornaviruses, both 3C protease (equivalent to the noroviral Pro) and 3D polymerase (equivalent to the noroviral RdRp) are known to enter the host nucleus (26)(27)(28)(29). Numerous studies have shown that the picornaviral 3C protease could cleave cellular transcriptional regulators or polyadenylation processing factors in the nucleus to modulate host gene expression (26)(27)(28), whereas the picornaviral 3D polymerase could target the pre-mRNA processing factor 8 (Prp8) to prevent mRNA maturation (29). Thus, for future research, it would be interesting to understand whether the nuclear localization of the GII norovirusencoded Pro or RdRp modulates specific host metabolic pathways. Although the Nterm, NTPase, and P22 proteins derived from GII.4 noroviruses were found exclusively in the cytoplasm, we would like to know whether these GII.4 viral proteins have the same intracellular localizations or functions as those of the previously reported counterparts derived from the GI and GV noroviruses. Here, we particularly  (20) showed that the NTPase protein from MNV (a strain of GV noroviruses) or Norwalk virus (a strain of GI noroviruses) could induce the formation of vesicle-like structures in the cell cytoplasm. In agreement with their published data, we also found that GII-NTPase was capable of inducing the formation of vesicles in the cytoplasm. Currently, we do not know whether the circular vesicular structure is made up of only GII-NTPase or the combination of GII-NTPase with other cellular proteins. In addition to vesicular structures, a nonvesicular fraction of GII-NTPase was also detected diffusely within the cytoplasm. When specific organelle markers were costained with GII-NTPase, vesiclelike structures formed by GII-NTPase did not colocalize with any tested organellespecific markers, including PDI, MitoTracker, GM130, TGN46, and EEA1 (Fig. 2B). These results are also consistent with the main findings reported by Cotton et al. (20) using GI and GV NTPase proteins. However, we showed here that the nonvesicular fraction of GII-NTPase substantially colocalized with the signals of PDI or MitoTracker (Fig. 2B). Deletion analysis of GII-NTPase revealed that the N-terminal 179-aa region is required for vesicle formation and for colocalization with PDI (Fig. 3), whereas the C-terminal region is responsible for the colocalization with MitoTracker (Fig. 4). Two functional mitochondrion-targeting domains in the C-terminal region of GII-NTPase were identified in the study and are located in the regions from aa 96 to 271 and from aa 255 to 366 of GII-NTPase (Fig. 4). Interestingly, although GII-NTPase contains two intrinsic mitochondrion-targeting domains, the full-length GII-NTPase is not predominantly localized to mitochondria (Fig. 2B, row b). These findings indicate that the N-terminal portion of GII-NTPase has an inhibitory action in its mitochondrial targeting. Due to the fact that GII-NTPase could interact with a variety of viral or cellular proteins during viral genome replication, we believe that different protein-protein interactions critically affect the targeting of GII-NTPase to mitochondria, ER, or other unknown areas. Further investigations may be required to determine whether and how viral or cellular proteins dynamically regulate the intracellular localization of GII-NTPase.
Potential mechanisms and roles of GII-NTPase targeting to mitochondria. Sequence analysis has revealed that the mitochondrion-targeting domains of GII-NTPase do not share significant homology with any known cellular or other viral proteins that are destined for mitochondria. Although GI-NTPase protein used in the study shares 52% amino acid identity with GII-NTPase and forms vesicle-like structures in the cytoplasm, the GI-NTPase protein did not produce substantial colocalization with mitochondria (Fig. 5A). This is consistent with the previous report that NTPase derived from Norwalk virus did not localize to mitochondria (20). In domain-swapping experiments using both GII-NTPase(255-366) and GI-NTPase(252-363), we showed that along with the shortening of the N-terminal part of GII-NTPase(255-366) in domainswapped mutants, their mitochondrial targeting ability was also gradually reduced ( Fig.  5D and F). These findings indicate that the N-terminal part of GII-NTPase(255-366) appears to be more important for the regulation of its mitochondrial targeting than the C-terminal part. However, we failed to identify a specific linear sequence motif within GII-NTPase(255-366) sufficient for the mitochondrial targeting using the domainswapping approach. From these findings, it is possible that the targeting of GII-NTPase(255-366) to mitochondria is mediated through a specific protein structure rather than a short primary sequence motif. Additionally, it should be noted that most domain-swapped mutants presented in the study exhibited a wide spectrum of localization to mitochondria in transfected cells ( Fig. 5D and G). The nonuniform patterns of mitochondrial colocalization for these domain-swapped mutants implied that intrinsic differences in cultured cells (e.g., mitochondrial fusion, fission, mitophagy, or biogenesis) influence their mitochondrial targeting.
Although the functional significance regarding the mitochondrial targeting of GII-NTPase in cells still remains obscure, three potential roles can be proposed. First, for many positive-sense RNA viruses, viral genome replication usually occurs in or on intracellular membranes. The association between GII-NTPase and mitochondria via the mitochondrion-targeting domains may expand the usage of intracellular membranes to facilitate cytoplasmic viral genome replication. Second, GII-NTPase is known to have NTP-binding and NTP-hydrolyzing activities (19), which may be involved in the modulation of central energy metabolism by ATP hydrolysis in cells. The association between GII-NTPase and mitochondria via the mitochondriontargeting domains may increase the energy supply for viral replication. Third, it is well known that the maintenance of mitochondrial homeostasis is critical for cell survival under stress conditions. Thus, the association between GII-NTPase and mitochondria via the mitochondrion-targeting domains may be linked to prevention or promotion of cell apoptosis at specific stages during viral replication. To distinguish these possibilities, further investigations are needed. Moreover, it is important to note that the mitochondrion-targeting domains of GII-NTPase expressed in cells not only fully colocalized with mitochondria but also significantly affected mitochondrial morphology (Fig. 4C, rows d, e, g, and h). Unlike the normal cell that generally contains a heterogeneous pattern of mitochondrial morphology with the combination of small fragmented globules and tubular threads (30)(31)(32), the cells expressing GII-NTPase(96 -366), GII-NTPase(213-366), GII-NTPase(255-366), or GII-NTPase(96 -271) consistently exhibited a homogeneous pattern of mitochondrial morphology with either globules or tubular threads (Fig. 4, rows d, e, g, and h).
Multiple functional activities for the N-terminal region of GII-NTPase. The full-length GII-NTPase contains 366 amino acids, in which a transmembrane helical motif located from aa 5 to 24 and an AAAϩ ATPase domain located from aa 157 to 291 could be predicted via InterPro tools (http://www.ebi.ac.uk/interpro/) (33). Functional analysis of GII-NTPase in the study has revealed that the N-terminal 179-aa region may have multiple activities in cells (summarized in Fig. 11). Besides acting as an inhibitory region in mitochondrial targeting, the N-terminal 179-aa region is also necessary and sufficient for mediating vesicular formation (Fig. 3), ER colocalization (Fig. 3), the interaction with Nterm or P22 (Fig. 7), the assembly of homodimers or homo-oligomers ( Fig. 8 and 9), and the induction of cell apoptosis (Fig. 10). Currently, we do not know whether the localization of GII-NTPase to the ER is mediated through a short ERtargeting sequence within the protein or through an interaction with other ERassociated cellular proteins. This issue will be addressed in future studies. Furthermore, the relationship between the diverse activities in the N-terminal region of GII-NTPase still remains to be determined. Due to the striking morphological differences between vesicular and nonvesicular (ER localization) structures formed by the N-terminal 179-aa region of GII-NTPase, the compositions of both vesicular and nonvesicular structures might be different in the cell. In this case, it may be interesting to know whether the formation of vesicular or ER-associated nonvesicular structures is linked to lipid droplets (LDs), the lipid-rich cytoplasmic organelles critical for lipid metabolism and energy homeostasis (34). Since LDs often vary in size in a given cell and can be divided into various types with different protein compositions (35), characterizing the association of GII-NTPase with specific LDs is under way. Additionally, although GII-NTPase could form homodimers or homo-oligomers through either a head-to-head, tail-to-tail, or headto-tail configuration, the regularly shaped vesicles were only observed in a head-tohead configuration (Fig. 9A, row c). These findings indicate that vesicular formation is attributed to a higher-order self-interaction of the N-terminal region of GII-NTPase.
Previous studies from MNV have showed that MNV infection triggered apoptosis in RAW264.7 cells through the mitochondrial pathway, which was characterized by the activation of caspase 9 and caspase 3 as well as downregulation of survivin (36). Furthermore, Herod et al. (37) demonstrated that only expression of the MNV ORF1 polyprotein in HEK293 cells was sufficient to induce cell apoptosis. However, no single nonstructural protein of noroviruses is known to be associated with induction of apoptosis. We showed here that the expression of GII-NTPase alone is capable of promoting apoptosis (Fig. 10A), although the mechanism employed by GII-NTPase to induce apoptosis still remains unknown. Surprisingly, we found that the proapoptotic activity of GII-NTPase was independent of its localization to mitochondria (Fig. 10A). Although Nterm or P22 alone did not show significant activity on cell apoptosis, we did find that Nterm or P22 augmented the proapoptotic activity of GII-NTPase. These findings strongly suggest that cooperative interactions among different nonstructural proteins during viral infection critically modulate cellular metabolism, thereby affecting host cell death and survival.
In summary, we report the subcellular localization and various functions of GII-NTPase in the present study. Our findings emphasize that the NTPase protein encoded by noroviruses must have multiple activities during viral replication; however, these activities may vary somewhat among different genogroups.

MATERIALS AND METHODS
Cell cultures and transfections. 293T is a human embryonic kidney cell line that harbors the E1 region of adenovirus and the simian virus 40 T antigen (38). A7 is a melanoma cell line (39). 293T and A7 cells were cultured in high-glucose DMEM (Dulbecco's modified Eagle's medium) supplemented with 10% fetal bovine serum. Transfection experiments were performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. Briefly, 293T cells (8 ϫ 10 5 ) and A7 cells (1.5 ϫ 10 5 ) were grown in 6-well culture plates for 16 to 24 h and then transfected with 3 g of plasmid DNA in 7.5 l of Lipofectamine and 1 g of plasmid DNA in 2.5 l of Lipofectamine, respectively.
Isolation of GI and GII norovirus strains. Stool samples were collected from patients with acute gastroenteritis in Chang-Gung Memorial Hospital at Chiayi (Taiwan). The protocol was approved by the Institutional Review Board of Chang-Gung Memorial Hospital (IRB number 97-1761B). Viral RNAs were extracted from stool supernatants using a QIAamp viral RNA minikit according to the manufacturer's instructions (Qiagen, Germany). Subsequently, viral RNAs were converted into cDNA by reverse transcription, followed by DNA cloning and sequencing. Complete genome sequences of two norovirus strains corresponding to a GII.4 strain, HuNV/GII.4/YJB1/2009/Chiayi, and a GI strain, HuNV/GI/CYY1/ 2009/Chiayi, were used in the study. The GenBank accession numbers for the genome sequences of the GII.4 strain, HuNV/GII.4/YJB1/2009/Chiayi, and the GI strain, HuNV/GI/CYY1/2009/Chiayi, are MG049692 and MG049693, respectively. As noted, the amino acid sequence of GII-NTPase used here shows 99% amino acid identity (364 out of 366 residues) with most of the previously published sequences for different GII.4 variants. The GI-NTPase protein from the HuNV/GI/CYY1/2009/Chiayi strain is also highly homologous (82 to 99% amino acid identity) with those of the previously reported GI noroviruses.
Plasmid construction. The full-length cDNA clone of a GII.4 norovirus strain, HuNV/GII.4/YJB1/2009/ Chiayi, was constructed by assembling different overlapping cDNA clones that were generated through reverse transcription. The full-length viral cDNA clone then was used as a template in PCR for the amplification of a target DNA region. To construct the expression plasmids encoding Nterm, NTPase, P22, VPg, Pro, and RdRp, the corresponding DNA fragments amplified by PCR were cloned into pFLAG-CMV-2 (Sigma). The resultant expression plasmids were designated pCMV-F-Nterm, pCMV-F-NTPase, pCMV-F-P22, pCMV-F-VPg, pCMV-F-Pro, and pCMV-F-RdRp. The expression plasmid pCMV-NTPase-myc was generated by inserting the NTPase coding region into pCMV-3Tag-4 (Stratagene), a C-terminal myctagging vector. Similarly, different expression plasmids encoding NTPase deletions were generated by insertion of PCR-amplified cDNA fragments into either pFLAG-CMV-2 or pCMV-3Tag-4. To construct the plasmids expressing the full length or deletion constructs of GI-NTPase, viral DNA fragments were obtained through PCR amplification using a cDNA clone of the GI norovirus strain (HuNV/GI/CYY1/2009/ Chiayi) as a template. The domain-swapped mutants between GI-NTPase(252-363) and GII-NTPase (255-366) were constructed by using the overlapping PCR method (40).
Confocal immunofluorescence analysis. A7 melanoma cells (1.5 ϫ 10 5 ) were seeded on coverslips in 6-well tissue culture plates and transiently transfected with the expression plasmids. After 24 h of transfection, cells were fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) at room temperature for 10 min and then washed with PBS three times for 5 min each. For mitochondrial staining, live cells were incubated with 250 nM MitoTracker (M7512; Thermo Fisher Scientific) at 37°C for 15 to 20 min prior to cell fixation. For immunofluorescence staining, the fixed cells were permeabilized with 0.1% Triton X-100 in PBS for 8 min at room temperature. Following a rinse with PBS, the cells were incubated with blocking solution (CAS-Block; Invitrogen) at room temperature for 30 min and then were treated with specific primary antibodies for 1 h. The primary antibodies against FLAG (A8592; Sigma), PDI (MA3-019; Thermo Fisher Scientific), GM130 (610822; BD Transduction Laboratories), TGN46 (NBP1-49643; Novus Biologicals), and myc (sc-40; Santa Cruz) were obtained commercially. After three further washing steps, appropriate secondary antibodies were added to treat cell samples, which included Alexa Fluor 488-conjugated goat-anti-mouse IgG (A-11001; Thermo Fisher Scientific) or goat-anti-rabbit IgG antibody (A-11008; Thermo Fisher Scientific), Alexa Fluor 633-conjugated goat-anti-mouse IgG antibody (A-21052; Thermo Fisher Scientific), or Alexa Fluor 594-conjugated goat-anti-mouse IgG antibody (A-11005; Thermo Fisher Scientific). Staining with 4=-6-diamidino-2-phenylindole (DAPI) was performed at room temperature for 15 min. Cells were mounted in CitiFluor (Agar Scientific) and were examined with a Leica confocal laser scanning system (TCS-SP5II). Image processing was carried out using LAS AF Lite software (Leica).
Coimmunoprecipitation. Coimmunoprecipitation analysis was performed as described previously (42). Briefly, 293T cells were cotransfected with expression vectors encoding myc-tagged NTPase and