Journal of General Virology

Non-structural proteins 2 and 3 interact to modify host cell membranes during the formation of the arterivirus replication complex

Eric J. Snijder,¹ Hans van Tol,¹ Norbert Roos² and Ketil W. Pedersen²

¹ Department of Virology, Center of Infectious Diseases, Leiden University Medical Center, LUMC P4-26, PO Box 9600, 2300 RC Leiden, The Netherlands
² Department of Biology, Division of Electron Microscopy, University of Oslo, Norway

The replicase polyproteins of equine arteritis virus (EAV; family Arteriviridae, order Nidovirales) are processed by three viral proteases to yield 12 non-structural proteins (nsps). The nsp2 and nsp3 cleavage products have previously been found to interact, a property that allows nsp2 to act as a co-factor in the processing of the downstream part of the polyprotein by the nsp4 protease. Remarkably, upon infection of Vero cells, but not of BHK-21 or RK-13 cells, EAV nsp2 is now shown to be subject to an additional, internal, cleavage. In Vero cells, approximately 50 % of nsp2 (61 kDa) was cleaved into an 18 kDa N-terminal part and a 44 kDa C-terminal part, most likely by a host cell protease that is absent in BHK-21 and RK-13 cells. Although the functional consequences of this additional processing step are unknown, the experiments in Vero cells revealed that the C-terminal part of nsp2 interacts with nsp3. Most EAV nsps localize to virus-induced double-membrane structures in the perinuclear region of the infected cell, where virus RNA synthesis takes place. It is now shown that, in an expression system, the co-expression of nsp2 and nsp3 is both necessary and sufficient to induce the formation of double-membrane structures that strikingly resemble those found in infected cells. Thus, the nsp2 and nsp3 cleavage products play a crucial role in two processes that are common to positive-strand RNA viruses that replicate in mammalian cells: controlled proteolysis of replicase precursors and membrane association of the virus replication complex.

The replicative proteins (or 'replicases') of many positive-stranded RNA (+RNA) viruses associate with host cell membrane compartments and modify these to establish a complex that is specialized for viral RNA synthesis (see Carette et al., 2000; Chen & Ahlquist, 2000; Egger et al., 2000; Mackenzie et al., 1999; Pedersen et al., 1999; Schaad et al., 1997; Schlegel et al., 1996; van der Meer et al., 1999; and references therein). Among the intracellular changes that have been documented are membrane proliferation and the modification of different cellular organelles, like those of the endo- and exocytotic pathways, the nucleus, peroxisomes and mitochondria. Although such changes appear to be a general feature of +RNA virus replication in eukaryotic cells, little is known about the exact role of membranes in viral RNA synthesis. The membranes may play a structural and/or functional role by offering a suitable microenvironment for viral RNA synthesis or they may facilitate the recruitment of membrane-associated host cell proteins for the purpose of virus transcription. In many +RNA viruses, including all major groups of animal +RNA viruses, the membrane association of virus non-structural proteins (nsps) is linked to their generation from large polyprotein precursors by proteolytic processing. Consequently, membrane association or translocation, polyprotein cleavage and the initiation of viral RNA synthesis probably occur in a highly co-ordinated fashion during the initial stages of the replication of these viruses.

Equine arteritis virus (EAV) (Doll et al., 1957) is the prototype of the family Arteriviridae (Snijder & Meulenberg, 1998), which has been placed in the order Nidovirales, together with the family Coronaviridae (Cavanagh, 1997). The ancestral relationship between the two virus groups is most evident from the common features of their genome organization and expression. Arteri- and coronaviruses both (i) encode a similar array of functional domains in their replicase genes, (ii) use ribosomal frameshifting to express key replicative functions, (iii) employ extensive proteolytic processing of replicase precursor polyproteins (Ziebuhr et al., 2000) and (iv) generate a nested set of subgenomic mRNAs to regulate the expression of their structural genes (Lai & Cavanagh, 1997; Snijder & Meulenberg, 1998).

EAV replicase gene expression results in the generation of two multidomain precursor proteins, the 1727 amino acid ORF1a protein and the 3175 amino acid ORF1ab protein (den Boon et al., 1991). These polyproteins are cleaved by three internal, ORF1a-encoded proteases (Snijder et al., 1992, 1995, 1996). Our current understanding of EAV replicase processing is summarized in Fig. 1. The ORF1a protein can be cleaved at seven sites (Snijder et al., 1994, 1996; Wassenaar et al., 1997), yielding a number of processing intermediates and eight end-products, nsp1 to nsp8. The N-terminal cleavage products nsp1 and nsp2 are liberated rapidly by internal cysteine autoprotease activities (Snijder et al., 1994). The remaining nsp3–8 intermediate (96 kDa) is processed by the nsp4 serine protease (SP), which also cleaves the ORF1b-encoded polyprotein three times (van Dinten et al., 1996, 1999). The latter part of the replicase contains a set of highly conserved functions that have been implicated in viral RNA synthesis (den Boon et al., 1991). Among its processing products (nsp9 to nsp12) are the putative virus RNA-dependent RNA polymerase (RdRp, nsp9) and the recently characterized EAV helicase (Hel; nsp10) (Seybert et al., 2000), which also contains a putative N-terminal metal-binding domain (van Dinten et al., 2000).

Fig. 1. Proteolytic processing scheme of the EAV replicase. Depicted are the EAV ORF1a and ORF1ab polyproteins, their cleavage sites and the current subunit nomenclature (Ziebuhr et al., 2000). The three EAV protease domains (PCP, CP and SP) and their cleavage sites (arrowheads) are shown in matching colours. Hydrophobic (H) domains in the ORF1a protein are depicted as black boxes. In the ORF1b-encoded polypeptide, four major domains conserved in nidoviruses have been depicted: RdRp, putative RNA-dependent RNA polymerase; M, putative metal-binding domain; Hel, RNA helicase; N, conserved C-terminal domain specific to nidoviruses. The lower part of the figure shows the two processing pathways for the ORF1a protein (Wassenaar et al., 1997). In the scheme of the major pathway, the presumed interaction between nsp2 and nsp3 (see text) is indicated in red. The approximate position of the internal nsp2 cleavage site (nsp2N/nsp2C junction), which is processed exclusively in Vero cells (see text), is indicated by a green arrowhead.

Immunofluorescence (van der Meer et al., 1998) and electron microscopy (EM) (Pedersen et al., 1999) studies have revealed that most EAV replicase subunits and viral RNA synthesis co-localize to the perinuclear region of infected cells. They are associated with intriguing double-membrane vesicles (DMVs), which are induced upon arterivirus infection and appear to be derived from the endoplasmic reticulum (ER) (Pedersen et al., 1999). Very similar membrane changes can be induced, in the absence of EAV infection, upon expression of the ORF1a-encoded replicase subunits nsp2 to nsp7 in an alphavirus-driven expression system (Pedersen et al., 1999). Hydrophobic domains located in nsp2, nsp3 and nsp5 were postulated to mediate the association of the EAV replicase with membranes. Taken together, these data suggest strongly that the formation of a membrane-bound scaffold for the arterivirus replication complex is an important function of the ORF1a protein.

An interesting observation during analysis of the processing of the EAV ORF1a protein was the fact that two alternative pathways can be followed for processing of the C-terminal part (Fig. 1). Either the nsp4/5 site ('major pathway') or the nsp5/6 and nsp6/7 sites ('minor pathway') are processed (Wassenaar et al., 1997). Cleavage of either of these sites is believed to render the alternative site(s) non-accessible. In an expression system, the presence of liberated nsp2 was found to determine whether the nsp4 SP could cleave the nsp4/5 site of the nsp3–8 precursor (Wassenaar et al., 1997). It is possible that a specific folding or post-translational organization of the protein is required for this proteolytic event. In the nsp2/nsp3–8 complex, nsp2 is likely to have a strong interaction with nsp3, since the two proteins were previously found to co-immunoprecipitate, even under quite stringent conditions (Snijder et al., 1994). Here, we show that nsp2 can be subject to an additional, internal cleavage, which is specific to Vero cells. Furthermore, both nsp2 and nsp3 were found to associate with membranes upon their individual expression by means of an alphavirus expression vector. However, only the expression of a self-cleaving nsp2–3 polyprotein induced the formation of the DMVs that are so typical of arterivirus infection.

Cells, virus and antisera. Baby hamster kidney (BHK-21), rabbit kidney (RK-13) and Vero (African green monkey kidney) cells were used for infection experiments with the EAV Bucyrus strain (Doll et al., 1957) following the protocol described by de Vries et al. (1992). Rabbit anti-EAV replicase antisera recognizing nsp1, nsp2, nsp4 and nsp7–8 have been described previously (Snijder et al., 1994). Because the anti-nsp3 antiserum (Pedersen et al., 1999) became available only recently, it could not be included in the immunoprecipitation studies described in this paper. The anti-nsp4 antiserum was a 1:1 mixture of the 4M and 4C rabbit antisera. Mouse monoclonal antibodies (MAbs) were used to visualize the cellular enzyme protein disulphide isomerase (PDI) (MAb 1D3; Vaux et al., 1990) and to detect hexahistidine-tagged expression products (MAb 13/45/31; Zentgraf et al., 1995).

Labelling and immunoprecipitation of EAV nsps. Cells were infected (m.o.i.=10) with an EAV stock grown and titrated in BHK-21 cells. To achieve the same m.o.i., Vero cells had to be infected with tenfold more virus than was used for either BHK-21 or RK-13 cells. All infected cell cultures were incubated at 39.5 °C. Radioactive labelling of intracellular proteins with [³⁵S]methionine/[³⁵S]cysteine (Expre³⁵S³⁵S protein labelling mix, DuPont NEN) was carried out between 5 and 8 h post-infection (p.i.) according to Snijder et al. (1994). Pulse–chase experiments were performed at 8 h p.i. by using a 15 min pulse and the labelling and chase protocol described previously (Snijder et al., 1994). Cells were lysed in the buffer described by de Vries et al. (1992) containing the protease inhibitors PMSF (400 µM), leupeptin (4 µM) and aprotinin (30 µM). Immunoprecipitations and SDS–PAGE were carried out essentially as described by de Vries et al. (1992). For all antisera, the immunoprecipitation buffer contained 0.5 % SDS.

Sindbis virus-based expression vectors. The previously described Sindbis virus expression vector pSinEAV(261–1677)His, a pSinRep5-derivative (Bredenbeek et al., 1993) expressing a C-terminally hexahistidine-tagged version of nsp2–7 (Pedersen et al., 1999; Wassenaar et al., 1997), was used to engineer similar vectors expressing nsp2, His-tagged nsp3 and His-tagged nsp2–3. These constructs were created by PCR deletion mutagenesis and were named pSRE-nsp2, pSRE-nsp3His and pSRE-nsp2+3His. pSRE-nsp2 was constructed by deleting the sequences encoding nsp3 to nsp7 and the His tag from pSinEAV(261–1677)His. To obtain pSRE-nsp2+3His, the nsp4 to nsp7 coding sequences were deleted from pSinEAV(261–1677)His and the His tag was fused in-frame to the nsp3 coding sequence. pSRE-nsp3His was obtained by deleting the nsp2 coding sequence from pSRE-nsp2+3His and fusing the engineered upstream ATG codon to the 5´ end of the nsp3 coding sequence. A control Sindbis virus expression vector for the green fluorescent protein (pSinRep/GFP) was kindly provided by C. M. Rice (Washington University, St Louis, MO, USA). BHK-21 cells were transfected by electroporation with infectious RNA transcribed from pSinRep5-derived vectors (van Dinten et al., 1997).

Immunofluorescence assays (IFAs) and electron microscopy. For indirect IFAs, transfected cells were seeded on cover slips, fixed with paraformaldehyde at 6, 9 or 12 h after transfection and processed as described by van der Meer et al. (1998). For EM, transfected cells were seeded in tissue culture dishes. Protocols for conventional Epon embedding and ultrathin sectioning and for cryoimmuno EM have been described by Pedersen et al. (1999). EM specimens were examined in a Philips CM100 transmission electron microscope.

Comparison of ORF1a polyprotein processing in different cell lines reveals a Vero cell-specific, internal cleavage of nsp2

RK-13, BHK-21 and Vero cells are the cell lines commonly used for the cultivation of EAV and the molecular-biological characterization of its life-cycle. Our previous biochemical studies on EAV replicase processing were all carried out in infected or transfected RK-13 cells (Snijder et al., 1994; van Dinten et al., 1996; Wassenaar et al., 1997). We have now compared the processing of the EAV ORF1a protein in the three cell lines mentioned above (Fig. 2). Cells were infected at high m.o.i. and ³⁵S-labelled between 5 and 8 h p.i., an interval roughly corresponding to the peak of viral RNA synthesis. Cell lysates were prepared and EAV replicase products were immunoprecipitated by using a panel of rabbit antisera described previously (Snijder et al., 1994).

SDS–PAGE analysis of the comparison (Fig. 2) revealed that the major pathway of ORF1a protein processing was essentially identical in RK-13 and BHK-21 cells (Fig. 2, lanes B and R). As outlined in Fig. 1 (major pathway), nsp1 liberates itself very efficiently; as before, nsp1-containing processing intermediates were not detected. Interpretation of the immunoprecipitations with the other sera is complicated by the strong interaction described previously between nsp2 and nsp3 (and nsp3-containing intermediates) (Snijder et al., 1994). Consequently, the nsp2 immunoprecipitation showed nsp2 itself, but also co-immunoprecipitation of nsp3, nsp3–4 and a small quantity of nsp3–8. Conversely, in the nsp4 and nsp7–8 immunoprecipitations, nsp2 was co-precipitated due to its respective interactions with nsp3–4 and nsp3–8 (Fig. 1). The nsp4 immunoprecipitation also showed mature nsp4 and the nsp3–4 intermediate, which has a considerable half-life due to the slow processing of the nsp3/4 junction (Snijder et al., 1994). Other polypeptides recognized by the nsp7–8 serum were nsp5–8, nsp5–7 and trace amounts of the somewhat smaller products produced via the alternative, minor pathway (Fig. 1; Wassenaar et al., 1997), which migrate in the 25–35 kDa region of the gel. Except for the nsp1 immunoprecipitation, all immunoprecipitations showed multiple, high molecular mass precursors that are explained by the extension of the ORF1a protein with the ORF1b protein upon ribosomal frameshifting (den Boon et al., 1991; van Dinten et al., 1996, 1999).

Fig. 2. Immunoprecipitation analysis of EAV ORF1a protein processing in infected BHK-21 (B), RK-13 (R) and Vero (V) cells. Viral proteins were ³⁵S-labelled from 5 to 8 h p.i. and EAV nsps were immunoprecipitated with the antisera indicated at the bottom of each panel. Size markers (arrowheads) and the positions of the major processing intermediates and end-products are indicated. The products of the newly described, internal cleavage of nsp2 (61 kDa) in Vero cells are indicated as nsp2N (18 kDa) and nsp2C (44 kDa) and are underlined.

The analysis of ORF1a protein processing in Vero cells (Fig. 2; lanes V) revealed a number of striking differences from the other two cell lines. The nsp2 antiserum precipitated a prominent additional band of approximately 18 kDa. At the same time, less nsp2 (61 kDa) was precipitated compared with the corresponding samples from the other cell lines. Furthermore, the nsp4 and nsp7–8 immunoprecipitations of the Vero cell sample revealed the appearance of a prominent additional band of about 44 kDa. The sizes of the two novel bands suggested strongly that they were derived from the cleavage of a proportion of the nsp2 at an internal site. The strong recognition of the 18 kDa protein by the anti-nsp2 antiserum, which was raised by using a 9 amino acid N-terminal peptide (Snijder et al., 1994), suggested that this product contained the nsp2 N-terminal domain. Only a trace amount of the 44 kDa product was precipitated by the anti-nsp2 antiserum (see also below). The anti-nsp4 and anti-nsp7–8 antisera precipitated much larger amounts of the 44 kDa product. Taken together, these data suggested cleavage of nsp2 into an N-terminal 18 kDa product (nsp2N) and a C-terminal 44 kDa (nsp2C) product. The co-precipitation of the latter product, for which an antiserum is not available, upon use of the anti-nsp4 and anti-nsp7–8 antisera (Figs 2 and 3) suggested that it is the C-terminal part of nsp2 that interacts with nsp3 and nsp3-containing processing intermediates.

Internal cleavage of nsp2 in Vero cells occurs shortly after translation

In order to study the cleavage of nsp2 in Vero cells in more detail, we performed a pulse–chase experiment (Fig. 3). At 8 h p.i., infected Vero cells were ³⁵S-labelled for 15 min. Cells were lysed either directly after the pulse or after chase periods of 1, 2 or 4 h. Immunoprecipitations were performed with the same set of antisera used in Fig. 2.

The incorporation of label into nsp2 was found to increase slightly during the first hour of the chase. This has been seen before (Snijder et al., 1994) and may be explained by completion of polyprotein synthesis after termination of the labelling and/or processing of nsp2 from short-lived precursors. The analysis made it clear that most of the nsp2N and nsp2C can be detected immediately after the pulse labelling. Furthermore, both cleavage products and their nsp2 precursor were found to be relatively stable proteins. A densitometric analysis of the nsp2 and nsp2N bands from the immunoprecipitation of the 1, 2 and 4 h chase samples with the anti-nsp2 serum (Fig. 3) indicated that, in the long run, approximately 50 % of the nsp2 molecules were cleaved. However, directly after the pulse labelling, there was clearly more nsp2N relative to nsp2 (Fig. 3; nsp2 panel, lane P), suggesting that the nsp2N/nsp2C cleavage occurred prior to cleavage of the nsp2/nsp3 site. For this estimation, the nsp2N C terminus was assumed arbitrarily to be residue 435 of the ORF1a polyprotein, which would give 18 kDa for the size of the nsp2N product, starting from the nsp2 N terminus at Gly-261 (Snijder et al., 1992). The actual position of the nsp2N/nsp2C border is unlikely to differ from this position by more than 50 amino acids, a difference that would not affect the outcome of this analysis significantly. The densitometric data were corrected for the methionine/cysteine content of the two products.

Fig. 3. Pulse–chase analysis of EAV ORF1a protein processing in infected Vero cells. Viral proteins were ³⁵S-labelled for 15 min at 8 h p.i. Cells were either lysed immediately after the pulse (lanes P) or the label was chased for 1, 2 or 4 h. EAV nsps were immunoprecipitated with the antisera indicated at the bottom of each panel.. Size markers (arrowheads) and the positions of processing intermediates and end-products are indicated (see also Fig. 1). The products of the newly described, internal cleavage of nsp2 (61 kDa) in Vero cells are indicated as nsp2N (18 kDa) and nsp2C (44 kDa) and are underlined.

Taken together, these data indicated that the internal cleavage of nsp2 is relatively rapid and that the nsp2 molecules that are not cleaved within the first hour after their synthesis remain uncleaved. The pulse–chase analysis presented in Fig. 3 also revealed that a minor, but increasing quantity of nsp2C was precipitated when the nsp2 antiserum was used. We interpret this to be co-precipitation of a proportion of the nsp2C molecules with a complex containing uncleaved nsp2 and an nsp3-containing intermediate (most likely nsp3–4). This could indicate that these complexes may contain multiple copies of nsp3, interacting with either nsp2 or nsp2C, or that a single nsp3 molecule can interact with multiple nsp2/nsp2C subunits.

Expression of nsp2 and nsp3 from alphavirus vectors

We have reported previously that the expression of EAV nsp2–7 from the 26S promoter of a Sindbis virus-based RNA vector (Bredenbeek et al., 1993) induced striking membrane re-arrangements in BHK-21 cells (Pedersen et al., 1999). Double membranes and DMVs, strongly resembling those found in infected cells, were observed and were labelled for various replicase subunits in cryoimmuno EM. We employed this expression system to delineate the sequences required for DMV formation in more detail. The nsp2–7 expression vector was used as the basis for making deletion variants expressing nsp2, nsp2–3, nsp2–4, nsp3, nsp3–4, nsp3–8 and nsp5–7. With the exception of nsp5–7, each of these expression products is able to process itself due to the action of the nsp2 and/or nsp4 proteases (data not shown).

Previously, we found that nsp2–7 expression results in an IFA staining that is indistinguishable from the staining observed in infected cells (Pedersen et al., 1999). Thus, similar IFAs with our anti-replicase antisera were used for a first analysis of the novel set of expression vectors. This screening revealed that only co-expression of nsp2 and nsp3 in the form of nsp2–3 (construct SRE-nsp2+3His) or nsp2–4 (data not shown) produced the dense, perinuclear staining observed in EAV-infected cells (Fig. 4 C, D). The presence of a His tag at the C terminus of nsp3 allowed us to demonstrate the exact co-localization of nsp2 and nsp3His in a double-labelling experiment (Fig. 4 D). Upon individual expression of nsp2 (construct SRE-nsp2) or nsp3 (construct SRE-nsp3His), both proteins seemed to associate with membranes (Fig. 4 A, B), but did not produce the perinuclear staining observed upon co-expression. Remarkably, the expression of nsp2 strongly reduced the amount of labelling for the ER-resident protein PDI (Fig. 4 A).

Fig. 4. Double immunofluorescence labelling showing the intracellular distribution of the nsp2, nsp3His and nsp2+3His expression products expressed from Sindbis virus-based RNA vectors. Cells were fixed at 8 h post-transfection. (A) Nsp2-expressing cells double-stained with antisera recognizing nsp2 and the ER-resident protein PDI. Note that nsp2 expression severely reduces the labelling for PDI. (B) Nsp3His-expressing cells stained with the anti-nsp3 antiserum. (C) EAV-infected cells showing the typical perinuclear staining observed in IFAs with most anti-replicase antisera, including those recognizing nsp2 and nsp3 (van der Meer et al., 1998). (D) Cells expressing the nsp2+3His protein were double-stained with a rabbit antiserum recognizing nsp2 and a mouse monoclonal antibody recognizing the His tag of nsp3His. Note the complete co-localization of the two cleavage products of the nsp2+3His polyprotein and the similarity of the labelling to that in (C).

Expression products with either nsp3 or nsp5 at their N terminus (nsp3–4, nsp3–8 and nsp5–7) also appeared to associate with membranes (data not shown; see also Discussion). This may be due to the presence of hydrophobic domains from nsp3 or nsp5 at the extreme N termini of these proteins, which may function as signal sequences. However, the significance of this phenomenon is unclear, since these sequences are normally located internally in the full-length ORF1a polyprotein.

EAV nsp2 and nsp3 interact in the formation of DMVs

Our analysis of the morphological changes in cells transfected with SRE-nsp2, SRE-nsp3His and SRE-nsp2+3His was extended by using EM. Serial sections did not reveal the presence of double-membrane structures in Epon-embedded cells expressing nsp2 only (Fig. 5 C), nsp3 only or GFP, which was used as negative control. In contrast, closely apposed ER membranes and DMVs were abundant in cells expressing the nsp2+3His protein (Fig. 5 A). Fig. 5(B) shows that, as in EAV-infected cells (Pedersen et al., 1999), the outer membrane of DMVs can be continuous with the ER membrane, creating a neck-like connecting structure. These results confirmed that, as suggested by the IFA results shown in Fig. 4, the expression of EAV nsp2 and nsp3 is necessary and sufficient to induce DMV formation. The morphology of the DMVs induced upon nsp2+3His expression was somewhat more variable (both in size and shape) compared with the structures seen in EAV-infected cells.

In order to extend this structural analysis, we analysed the same set of samples by using cryoimmuno EM. In cells expressing the nsp2+3His polyprotein, both nsps localized exclusively to mature DMVs and to the closely apposed ER membranes (Fig. 6 A, B), from which the DMVs appear to be forming (Figs 5 B and 6 C). In cells expressing only nsp2 or only nsp3, the formation of closely apposed membranes or DMVs was not observed. The labelling for these individually expressed proteins was dispersed. Although intracellular membranes were labelled slightly above the background, the results were difficult to interpret (data not shown). The negative control expressing GFP did not show any labelling.

Both the production of nsps by polyprotein proteolysis and the membrane association of the replication complex are common properties of +RNA viruses that replicate in mammalian cells. In the case of EAV, it has become increasingly clear that the interaction between the nsp2 and nsp3 replicase subunits plays a crucial role in both processes. We have shown here that co-expression of nsp2 and nsp3 is both necessary and sufficient to induce the formation of double-membrane structures that are strikingly similar to those that have been shown to carry the EAV replication complex in infected cells (Pedersen et al., 1999; van der Meer et al., 1998). Previously, it was also concluded that nsp2 plays an essential role as co-factor in the major processing pathway of the ORF1a protein, most likely through its interaction with the nsp3 part of the nsp3–8 precursor (Fig. 1; Wassenaar et al., 1997). Since the nsp4 protease, which performs all cleavages in nsp3–8, remains fully active in the absence of nsp2, it was postulated that the nsp2 co-factor does not affect the proteolytic activity of nsp4 itself. Instead, the complex between nsp2 and nsp3–8 may have a specific conformation that allows processing of the nsp4/5 site, the first step in the major processing pathway (Fig. 1).

Fig. 5. Formation of DMVs from paired ER membranes upon nsp2–3 expression. Epon sections of transfected BHK-21 cells fixed at 8 or 12 h post-transfection are shown. Bars, 100 nm. ER, Endoplasmic reticulum; PM, plasma membrane; M, mitochondrion. (A) Overview of an nsp2–3His-expressing cell at 12 h post-transfection with construct SRE-nsp2+3His. ER membranes are tightly apposed, forming DMVs of irregular shape and size (arrows). The neighbouring untransfected cell shows no signs of closely apposed ER membranes or DMVs, as judged by serial sections. (B) A higher-magnification image of an nsp2–3-expressing cell (8 h post-transfection) demonstrating the apparent formation of a DMV from the ER. The arrow indicates a continuous (closed) inner membrane while the outer membrane is continuous with the ER, forming a neck-like connection (arrowhead). (C) Typical image of a cell expressing nsp2 (SRE-nsp2 transfection after 8 h). Closely apposed ER membranes or DMVs could not be detected in serial sections of this sample.

In view of these crucial functions of nsp2, it is quite remarkable that about 50 % of this protein was now found to be internally cleaved upon EAV replication in Vero cells. The internal cleavage site was estimated to be close to residue 435 of the ORF1a protein. However, this estimation was based solely on the migration of nsp2N (18 kDa) and nsp2C (44 kDa) upon SDS–PAGE, which cannot be considered a very reliable method. In principle, one of the three EAV proteases may be involved in the nsp2 cleavage, although potential cleavage sites for these enzymes appear to be lacking from the amino acid sequence of the region that should contain the nsp2N/nsp2C junction (residues 385–485 of the ORF1a protein). Instead, this part of nsp2 stands out for its high content of basic residues (13 %) and prolines (19 %). Together with the fact that the internal nsp2 cleavage is completely lacking in both BHK-21 and RK-13 cells, these observations suggests strongly the involvement of a host cell-specific protease. Such an enzyme might be located in the cytoplasm, but it might also reside in the lumen of intracellular compartments like the ER and act on a luminal domain of a partially translocated nsp2 protein. The latter mechanism is employed during the processing of polyproteins generated by members of the family Flaviviridae (see Ryan et al., 1998; and references therein).

EAV nsp2 contains two somewhat hydrophobic regions around residues 450 and 490 (ORF1a polyprotein numbering). If one of these spans the membrane, a luminal cleavage downstream of this domain would generate two products with sizes that are relatively close to those estimated from SDS–PAGE gels for nsp2N and nsp2C. The main hydrophobic domain of nsp2 is found between residues 520 and 640 and is large enough to span the lipid bilayer several times. Although it is clear that, at least at some point, the nsp2 N- and C-terminal domains (containing the nsp2 protease and the nsp2/3 cleavage site) must be on the same side of the membrane, the exact topology of this unusual non-structural protein remains to be elucidated.

The functional implications of the internal nsp2 cleavage for EAV replication in Vero cells are unclear. Although the EAV replication cycle is somewhat delayed in Vero cells compared with BHK-21 or RK-13 cells (15 h versus 12 h), it is premature to attribute this difference to the internal processing of nsp2. Previous IFA studies have not revealed any major differences between the nsp2 labelling patterns in the three different cell lines (van der Meer et al., 1998), suggesting that the nsp2N cleavage product remains associated with membranes. This observation could be explained by assuming that nsp2N indeed contains one of the hydrophobic domains of nsp2 and/or becomes part of a stable, membrane-associated complex before its partial internal cleavage. Thus, at the moment of nsp2N/nsp2C cleavage, both parts of the protein may be anchored within the replication complex by interactions with other partners or interactions between the N- and C-terminal domains of nsp2 itself. For nsp2C, an obvious partner is nsp3, on the basis of the observed co-immunoprecipitation with nsp3-containing intermediates (Figs 2 and 3). Furthermore, both nsp2N and nsp2C contain clusters of conserved Cys residues (Snijder et al., 1994, 1995) with unknown functions.

Fig. 6. Cryoimmuno EM analysis of DMVs formed upon nsp2–3His expression. Cells were fixed at 8 or 12 h post-transfection. Bars, 100 nm. (A) Labelling by using the anti-nsp2 antiserum of a DMV still connected to closely apposed ER membranes (arrows). (B) Tightly apposed ER membranes, a DMV connected to the ER and a 'mature' DMV, all labelled with the anti-nsp2 antiserum (arrows). (C) Labelling with the anti-nsp3 antiserum, demonstrating that nsp3 localizes to the DMVs in nsp2–3-expressing cells (big arrow). The small arrow indicates a region of the DMV where the inner and outer membranes are continuous, creating a connection between the DMV interior and the cytoplasm.

Despite the fact that detailed information on the biochemical properties of nsp2 and the mechanism of its membrane association/translocation is lacking, our IFA studies show that the protein by itself can associate with membranes, probably those of the ER. Remarkably, the expression of nsp2 seems to reduce the amount of PDI in the cell dramatically, an observation that has occasionally also been made in EAV-infected cells, late in infection (van der Meer et al., 1998). Although EM studies did not reveal any morphological changes of the ER in these cells, this phenomenon may still signify biochemical changes resulting from the interaction of nsp2 with the ER. The membrane association of individually expressed nsp3 (Fig. 4 B) is not very surprising, since computer analysis predicts the hydrophobic nsp3 N terminus to be a quite reasonable signal sequence. Some of the expression products with nsp3 at their N terminus (e.g. nsp3–4 and nsp3–8) were even partially transported to the Golgi complex (data not shown). Since staining of the Golgi complex with anti-replicase antisera has never been observed in EAV-infected cells (van der Meer et al., 1998), one consequence of the nsp2–nsp3 interaction appears to be the retention of nsp3-containing proteins in the ER membrane.

The IFA (Fig. 4) and EM (Figs 5 and 6) studies presented in this paper have clearly shown that co-expression of nsp2 and nsp3 leads to the formation of paired membranes and DMVs (Fig. 5 B) that are labelled for the two proteins (Fig. 6). Apparently, the third major hydrophobic domain in the ORF1a protein, that in nsp5, is dispensable for the formation of these structures. Using the EAV infectious cDNA clone (van Dinten et al., 1997) and Sindbis virus expression vectors, we should now be able to dissect the interaction between nsp2 and nsp3 in more detail, e.g. by using site-directed and deletion mutagenesis. These future studies can be expected to shed light on the co-ordination of replicase polyprotein processing and membrane association, the mechanism of DMV formation (Pedersen et al., 1999; Schlegel et al., 1996) and the role of DMVs in viral RNA synthesis.

We gratefully acknowledge Leonie van Dinten, Jessika Dobbe, Sophie Greve, Yvonne van der Meer and Fred Wassenaar (Department of Virology, Leiden University Medical Center) and Espen Stang, Andreas Brech and Tove Bakar (Department of Biology, Oslo University) for technical assistance and comments. We thank Sasha Gorbalenya for helpful discussions and for reviewing the manuscript. We are indebted to Dr S. Fuller (EMBL, Heidelberg, Germany) for the anti-PDI monoclonal antibody, Dr H. Zentgraf (DKFZ, Heidelberg, Germany) for the anti-His tag antibody and Dr C. M. Rice (Washington University, St Louis, USA) for the SinRep/GFP vector.

Bredenbeek, P. J., Frolov, I., Rice, C. M. & Schlesinger, S. (1993). Sindbis virus expression vectors: packaging of RNA replicons by using defective helper RNAs. Journal of Virology 67, 6439–6446.

Carette, J. E., Stuiver, M., Van Lent, J., Wellink, J. & Van Kammen, A. (2000). Cowpea mosaic virus infection induces a massive proliferation of endoplasmic reticulum but not Golgi membranes and is dependent on de novo membrane synthesis. Journal of Virology 74, 6556–6563.

Cavanagh, D. (1997). Nidovirales: a new order comprising Coronaviridae and Arteriviridae. Archives of Virology 142, 629–633.

Chen, J. & Ahlquist, P. (2000). Brome mosaic virus polymerase-like protein 2a is directed to the endoplasmic reticulum by helicase-like viral protein 1a. Journal of Virology 74, 4310–4318.

den Boon, J. A., Snijder, E. J., Chirnside, E. D., de Vries, A. A. F., Horzinek, M. C. & Spaan, W. J. M. (1991). Equine arteritis virus is not a togavirus but belongs to the coronaviruslike superfamily. Journal of Virology 65, 2910–2920.

de Vries, A. A. F., Chirnside, E. D., Horzinek, M. C. & Rottier, P. J. M. (1992). Structural proteins of equine arteritis virus. Journal of Virology 66, 6294–6303.

Doll, E. R., Bryans, J. T., McCollum, W. H. & Crowe, M. E. W. (1957). Isolation of a filterable agent causing arteritis of horses and abortion by mares. Its differentiation from the equine abortion (influenza) virus. Cornell Veterinarian 47, 3–41.

Egger, D., Teterina, N., Ehrenfeld, E. & Bienz, K. (2000). Formation of the poliovirus replication complex requires coupled viral translation, vesicle production, and viral RNA synthesis. Journal of Virology 74, 6570–6580.

Lai, M. M. C. & Cavanagh, D. (1997). The molecular biology of coronaviruses. Advances in Virus Research 48, 1–100.

Mackenzie, J. M., Jones, M. K. & Westaway, E. G. (1999). Markers for trans-Golgi membranes and the intermediate compartment localize to induced membranes with distinct replication functions in flavivirus-infected cells. Journal of Virology 73, 9555–9567.

Pedersen, K. W., van der Meer, Y., Roos, N. & Snijder, E. J. (1999). Open reading frame 1a-encoded subunits of the arterivirus replicase induce endoplasmic reticulum-derived double-membrane vesicles which carry the viral replication complex. Journal of Virology 73, 2016–2026.

Ryan, M. D., Monaghan, S. & Flint, M. (1998). Virus-encoded proteinases of the Flaviviridae. Journal of General Virology 79, 947–959.

Schaad, M. C., Jensen, P. E. & Carrington, J. C. (1997). Formation of plant RNA virus replication complexes on membranes: role of an endoplasmic reticulum-targeted viral protein. EMBO Journal 16, 4049–4059.

Schlegel, A., Giddings, T. H., Jr, Ladinsky, M. S. & Kirkegaard, K. (1996). Cellular origin and ultrastructure of membranes induced during poliovirus infection. Journal of Virology 70, 6576–6588.

Seybert, A., van Dinten, L. C., Snijder, E. J. & Ziebuhr, J. (2000). Biochemical characterization of the equine arteritis virus helicase suggests a close functional relationship between arterivirus and coronavirus helicases. Journal of Virology 74, 9586–9593.

Snijder, E. J. & Meulenberg, J. J. M. (1998). The molecular biology of arteriviruses. Journal of General Virology 79, 961–979.

Snijder, E. J., Wassenaar, A. L. M. & Spaan, W. J. M. (1992). The 5´ end of the equine arteritis virus replicase gene encodes a papainlike cysteine protease. Journal of Virology 66, 7040–7048.

Snijder, E. J., Wassenaar, A. L. M. & Spaan, W. J. M. (1994). Proteolytic processing of the replicase ORF1a protein of equine arteritis virus. Journal of Virology 68, 5755–5764.

Snijder, E. J., Wassenaar, A. L. M., Spaan, W. J. M. & Gorbalenya, A. E. (1995). The arterivirus Nsp2 protease. An unusual cysteine protease with primary structure similarities to both papain-like and chymotrypsin-like proteases. Journal of Biological Chemistry 270, 16671–16676.

Snijder, E. J., Wassenaar, A. L. M., van Dinten, L. C., Spaan, W. J. M. & Gorbalenya, A. E. (1996). The arterivirus nsp4 protease is the prototype of a novel group of chymotrypsin-like enzymes, the 3C-like serine proteases. Journal of Biological Chemistry 271, 4864–4871.

van der Meer, Y., van Tol, H., Krijnse Locker, J. & Snijder, E. J. (1998). ORF1a-encoded replicase subunits are involved in the membrane association of the arterivirus replication complex. Journal of Virology 72, 6689–6698.

van der Meer, Y., Snijder, E. J., Dobbe, J. C., Schleich, S., Denison, M. R., Spaan, W. J. M. & Krijnse Locker, J. (1999). Localization of mouse hepatitis virus nonstructural proteins and RNA synthesis indicates a role for late endosomes in viral replication. Journal of Virology 73, 7641–7657.

van Dinten, L. C., Wassenaar, A. L. M., Gorbalenya, A. E., Spaan, W. J. M. & Snijder, E. J. (1996). Processing of the equine arteritis virus replicase ORF1b protein: identification of cleavage products containing the putative viral polymerase and helicase domains. Journal of Virology 70, 6625–6633.

van Dinten, L. C., den Boon, J. A., Wassenaar, A. L. M., Spaan, W. J. M. & Snijder, E. J. (1997). An infectious arterivirus cDNA clone: identification of a replicase point mutation that abolishes discontinuous mRNA transcription. Proceedings of the National Academy of Sciences, USA 94, 991–996.

van Dinten, L. C., Rensen, S., Gorbalenya, A. E. & Snijder, E. J. (1999). Proteolytic processing of the open reading frame 1b-encoded part of arterivirus replicase is mediated by nsp4 serine protease and is essential for virus replication. Journal of Virology 73, 2027–2037.

van Dinten, L. C., van Tol, H., Gorbalenya, A. E. & Snijder, E. J. (2000). The predicted metal-binding region of the arterivirus helicase protein is involved in subgenomic mRNA synthesis, genome replication, and virion biogenesis. Journal of Virology 74, 5213–5223.

Vaux, D., Tooze, J. & Fuller, S. (1990). Identification by anti-idiotype antibodies of an intracellular membrane protein that recognizes a mammalian endoplasmic reticulum retention signal. Nature 345, 495–502.

Wassenaar, A. L. M., Spaan, W. J. M., Gorbalenya, A. E. & Snijder, E. J. (1997). Alternative proteolytic processing of the arterivirus replicase ORF1a polyprotein: evidence that NSP2 acts as a cofactor for the NSP4 serine protease. Journal of Virology 71, 9313–9322.

Zentgraf, H., Frey, M., Schwinn, S., Tessmer, C., Willemann, B., Samstag, Y. & Velhagen, I. (1995). Detection of histidine-tagged fusion proteins by using a high-specific mouse monoclonal anti-histidine tag antibody. Nucleic Acids Research 23, 3347–3348.

Ziebuhr, J., Snijder, E. J. & Gorbalenya, A. E. (2000). Virus-encoded proteinases and proteolytic processing in the Nidovirales. Journal of General Virology 81, 853–879.

JGV Direct table of contents

© 2001 SGM

This article is now available in the May 2001 print issue of JGV (vol. 82, 985–994). The complete issue of the journal may be seen in electronic form on JGV Online.