Identification of a Novel Structural Protein of Arteriviruses

Abstract

Arteriviruses are positive-stranded RNA viruses with an efficiently organized, polycistronic genome. A short region between the replicase gene and open reading frame (ORF) 2 of the equine arteritis virus (EAV) genome was previously assumed to be untranslated. However, here we report that this segment of the EAV genome contains the 5′ part of a novel gene (ORF 2a) which is conserved in all arteriviruses. The 3′ part of EAV ORF 2a overlaps with the 5′ part of the former ORF 2 (now renamed ORF 2b), which encodes the G_S glycoprotein. Both ORF 2a and ORF 2b appear to be expressed from mRNA 2, which thereby constitutes the first proven example of a bicistronic mRNA in arteriviruses. The 67-amino-acid protein encoded by EAV ORF 2a, which we have provisionally named the envelope (E) protein, is very hydrophobic and has a basic C terminus. An E protein-specific antiserum was raised and used to demonstrate the expression of the novel gene in EAV-infected cells. The EAV E protein proved to be very stable, did not form disulfide-linked oligomers, and was not N-glycosylated. Immunofluorescence and immunoelectron microscopy studies showed that the E protein associates with intracellular membranes both in EAV-infected cells and upon independent expression. An analysis of purified EAV particles revealed that the E protein is a structural protein. By using reverse genetics, we demonstrated that both the EAV E and G_S proteins are essential for the production of infectious progeny virus.

Arteriviruses are enveloped, positive-stranded RNA viruses with a diameter of 40 to 60 nm (for a review, see reference 54). In addition to the family prototype, equine arteritis virus (EAV) (20), the recently established Arteriviridae family contains the lactate dehydrogenase-elevating virus (LDV) of mice, porcine reproductive and respiratory syndrome virus (PRRSV), and simian hemorrhagic fever virus (SHFV). The gene encoding the arterivirus replicase is related to that of the Coronaviridae, and despite a considerable difference in genome size, the replication and genome expression strategies of these two virus families are strikingly similar (Fig. 1) (14, 17, 54, 55). Consequently, the two virus groups have recently been united in the newly created order Nidovirales (9).

FIG. 1.

Overview of the replication cycle of arteriviruses based on the family prototype, EAV. The viral genome and the nested set of subgenomic mRNAs are indicated, with the small black boxes representing the common 5′ leader sequence. The previously identified genes (14, 16) are depicted, with the white boxes representing the translationally active ORF(s) for each mRNA.

The putatively icosahedral arterivirus core (diameter, 25 to 35 nm) consists of a 12.7- to 15.7-kb nonsegmented RNA genome and a single nucleocapsid (N) protein of 110 to 128 amino acids. The arterivirus envelope is relatively smooth and accommodates two main protein components: (i) the nonglycosylated, triple-spanning membrane protein M (16 to 20 kDa); and (ii) the major envelope glycoprotein (designated G_L in the case of EAV, VP-3P for LDV, GP₅ for PRRSV, and p54 for SHFV), which differs considerably in size depending on the virus (16, 18, 23, 30, 42, 47). In addition, arterivirus particles contain at least one minor membrane glycoprotein, denoted G_S for EAV (19), VP-3M for LDV (23), GP₂ for PRRSV (46), and p42 for SHFV (30). Furthermore, PRRSV particles have been reported to possess two additional, minor envelope glycoproteins (GP₃ and GP₄) (64).

With the exception of a small region of the SHFV genome (29, 53), a fairly consistent genome organization was described for all arteriviruses. The 5′ three-quarters of the polycistronic arterivirus genome (Fig. 1) contain two large open reading frames (ORFs), 1a and 1b, which encode the viral replicase. The replicase ORFs are translated directly from the genomic RNA, with ORF 1b expression involving a ribosomal frameshift (14). The resulting nonstructural precursor polyproteins are cleaved by three viral proteases (for reviews, see references 17 and 54). Downstream of the replicase gene, three of the four arterivirus genomes (those of EAV, LDV, and PRRSV) were reported to contain six smaller genes (ORFs 2 to 7), which mostly or exclusively encode structural proteins. The corresponding region of the SHFV genome contains three additional ORFs (covering about 1.6 kb) between ORF 1b and the equivalent of EAV/LDV/PRRSV ORF 5 (named ORF 7 in SHFV [53]). On the basis of limited sequence similarities, it was postulated that these ORFs (named ORFs 2a, 2b, and 3) arose from the duplication of the SHFV homologs of EAV/LDV/PRRSV ORFs 2 to 4 by an RNA recombination event (see Fig. 3B) (29). Except for this putative gene duplication in SHFV, the genetic compositions and gene orders of all arteriviruses are identical. The three most 3′-proximal genes (ORFs 5 to 7; ORFs 7 to 9 in SHFV) encode the three main structural components of the virus particle: the major membrane glycoprotein (G_L/VP-3P/GP₅/p54), the M protein, and the N protein, respectively (Fig. 1). The minor envelope glycoprotein (G_S/VP-3M/GP₂) of 25 to 30 kDa is encoded by EAV/LDV/PRRSV ORF 2.

FIG. 3.

Schematic overview of the revised genomic organization of the four arteriviruses. (A) Schema of the partial genomic organization of EAV, LDV, PRRSV, and SHFV in which the positions of the novel gene (in black) and previously identified ORFs are indicated. The TRSs used for the synthesis of mRNAs 2 and 3 (or mRNAs 4 and 5 in SHFV) are also depicted (arrows). (B) Comparison of the genomic organizations of the complete 3′ portions of the EAV and SHFV genomes. The hatched ORFs 2a, 2b, and 3 in SHFV have been proposed to be derived from a three-gene duplication (29). The SHFV ORFs 4a to 9 are assumed to be the homologs of EAV ORFs 2a to 7 (see the text).

The six to nine genes in the 3′ quarter of the arterivirus genome are expressed from a nested set of six to eight subgenomic mRNAs (Fig. 1). The generation of such a set of subgenomic transcripts is one of the hallmarks of the nidovirus replication cycle. The subgenomic mRNAs of arteriviruses and coronaviruses are 3′ coterminal, but they also carry a common 5′ leader sequence which is derived from the 5′ end of the viral genome. The leader is fused to the subgenomic mRNA bodies by a discontinuous transcription mechanism (reviewed in references 17, 37, 54, and 62). With the exception of the smallest mRNA, the nidovirus subgenomic mRNAs are structurally polycistronic (Fig. 1). However, they are assumed to be functionally monocistronic; i.e., only the most 5′-proximal gene of each mRNA is translated. Several exceptions to this rule have been reported for coronaviruses (reviewed in reference 40). More recently, mRNA 2 of the arterivirus SHFV was proposed to function as a bicistronic mRNA (29), although the translation of two ORFs from this mRNA remains to be confirmed experimentally.

Except for the 5′ and 3′ untranslated regions of the genome, which are likely to contain viral replication signals, the genomes of the four currently known arteriviruses were reported to be almost completely devoid of noncoding sequences (14, 28, 45, 49, 53). As a rule, the structural protein genes in the 3′ end of the genome overlap with both adjacent genes over distances that vary from a few to several hundred nucleotides (Fig. 1; see also Fig. 3B). Due to this extremely efficient organization of arterivirus genetic information, the subgenomic mRNA transcription-regulating sequences (TRSs), which precede every gene and are used to join the leader sequence to the mRNA body (17, 54), are always located within an upstream gene.

The region between ORF 1b and ORF 2, which separates the replicase from the structural protein genes (Fig. 1), was one of the few apparently noncoding gaps in the genomes of EAV, LDV, and PRRSV. However, we have now obtained convincing evidence that in the case of EAV this region contains the 5′ part of a 201-nucleotide (nt) gene (ORF 2a) which encodes a largely hydrophobic, 67-residue polypeptide with a basic C-terminal domain. The 3′ part of this novel ORF overlaps with the previously identified G_S gene (ORF 2, now renamed ORF 2b). Both ORF 2a and ORF 2b appear to be expressed from subgenomic mRNA 2. Homologous genes were identified at corresponding positions in the genomes of the three other arteriviruses. With the aid of a monospecific antiserum, the EAV ORF 2a protein was identified in purified virions. By using reverse genetics, we also showed that it is indispensable for the generation of infectious virus particles. Thus, we have identified a novel and essential structural component of arterivirus particles.

MATERIALS AND METHODS

Cells and viruses.

Baby hamster kidney (BHK-21) cells and chicken embryo fibroblasts were maintained in Glasgow minimum essential medium (GMEM; Life Technologies) containing 5 to 10% fetal calf serum (FCS) and antibiotics. Rabbit kidney (RK-13) and OST-7.1 cells (22) were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Life Technologies) supplemented with 5% FCS and antibiotics. Infection experiments with the Bucyrus strain of EAV (20) were performed in BHK-21 or RK-13 cells according to the protocol of de Vries et al. (16). BHK-21 cells were also used for transfection experiments with in vitro-generated transcripts from the EAV infectious cDNA clone pEAV030 (and derivatives thereof), using a previously described electroporation protocol (63). Modified vaccinia virus Ankara expressing the bacteriophage T7 DNA-dependent RNA polymerase (MVA-T7) (58) was propagated and titrated in chicken embryo fibroblasts. DNA transfection experiments for transient expression studies were carried out in OST-7.1 or BHK-21 cells.

Antisera.

An antiserum specific for the C terminus of the EAV ORF 2a protein was generated by injecting rabbits with the bovine serum albumin-coupled peptide NH₂-Gly-Arg-Ser-Leu-Val-Ala-Arg-Cys-Ser-Arg-Gly-Ala-Arg-Tyr-Arg-Pro-Val-COOH, using a previously described immunization scheme (56). The production and characterization of rabbit antisera recognizing other EAV nonstructural and structural proteins have been described before (16, 56). The localization of the EAV ORF 5-encoded glycoprotein G_L was studied with monoclonal antibody (MAb) 93B (26). The protein disulfide isomerase (PDI)-specific MAb 1D3 (65) and MAb F20/65-1-4 (32) were used as markers for the endoplasmic reticulum (ER) and Golgi complex, respectively.

Expression vectors.

Recombinant DNA plasmids were generated by standard techniques and protocols (1, 50). ORF 2a was placed downstream of a T7 RNA polymerase promoter by ligation of a 0.3-kb ApaI-PstI fragment (nt 9,740 to 10,051 of the EAV genome) from EAV cDNA clone 535 (15) into pBluescript SK(−) (Stratagene) to produce pAVI02a. The EAV ORF 6 expression vector pAVI16 has been described elsewhere (16).

Generation and analysis of knockout viruses and revertants.

The translation initiation codons of EAV ORF 2a and ORF 2b were mutated in the EAV infectious cDNA clone pEAV030 (63) (EMBL database accession no. Y07862). The ORF 2a start codon (Fig. 2) was changed to ATA (G-9754 to A). In addition, a T residue 4 nt downstream (T-9758) was replaced by C to engineer a novel StuI restriction site (AGGCCT) in pEAV030 which could be used as marker for the ORF 2a knockout mutation in construct pEAV030-2aKO. In construct pEAV030-2bKO, the ORF 2b start codon was changed to ACG (T-9826 to C) (Fig. 2). A second mutation, also translationally silent with respect to ORF 2a, was made 3 nt downstream (A-9829 to C) to destroy an Eco47III restriction site. The ORF 2a and ORF 2b knockout mutations were engineered in shuttle vectors by PCR mutagenesis (38), after which fully sequenced restriction fragments carrying the mutations were placed back into pEAV030. In vitro transcription reactions and RNA transfections were carried out as described before (63).

FIG. 2.

Sequence analysis of the ORF 2a-containing genomic region of EAV and the corresponding parts of the genomes of the three other arteriviruses. cDNA sequences as well as amino acid sequences of significant ORFs are given; potential translation initiation codons (ATG) are underlined. The arrows indicate the start sites of the newly identified ORFs in the sequences. The positions of the TRSs of the relevant mRNAs are also indicated.

Revertants of the pEAV030-2aKO mutant were obtained by picking individual plaques from infectious-center assays of transfected cells. Plaques were resuspended in DMEM and used to infect fresh BHK-21 cells, after which intracellular RNA was isolated as described before (13). The ORF 2a region in the 5′ end of mRNA 2 was reverse transcribed by using Moloney murine leukemia virus reverse transcriptase and a primer complementary to nt 10,301 to 10,318 of the EAV genome. The same primer and a primer corresponding to nt 63 to 89 (in the mRNA 2 leader sequence) were used for PCR amplification of the cDNA. The sequence of the resulting 752-bp DNA fragment was determined directly by cycle sequencing with an ABi PRISM Big Dye Terminator Cycle Sequencing Kit and an ABiPrism 310 automatic sequencer (Perkin-Elmer).

DNA transfection experiments.

Subconfluent monolayers of OST-7.1 or BHK-21 cells were grown in 10-cm² dishes or on glass coverslips, washed once with DMEM (OST-7.1 cells) or GMEM (BHK-21 cells), and infected with MVA-T7 at a multiplicity of infection of 5 to 10 PFU per cell. After a 1-h incubation at 37°C, the inoculum was removed and the cells were washed twice with medium lacking serum and antibiotics. The cells were given 200 μl of a transfection mixture consisting of 5 μg of plasmid DNA, 10 μl of Lipofectin (Life Technologies), and 190 μl of DMEM (OST-7.1 cells) or GMEM (BHK-21 cells) and incubated for 10 min at room temperature. Next, 800 μl of prewarmed medium was added and the cells were incubated at 37°C, after which they were either fixed for immunofluorescence or metabolically labeled at various time points.

Preparation of radiolabeled cell lysates.

Metabolic labeling experiments were carried out with EAV- and mock-infected BHK-21 cells, RNA-transfected BHK-21 cells, and MVA-T7-infected and plasmid-transfected BHK-21 or OST-7.1 cells, using established procedures (19). The DNA-transfected cells were kept at 37°C, and the labeling was started at 6 h after infection with MVA-T7, following a starvation period of 30 min. EAV- and mock-infected cells and RNA-transfected cells were incubated at 39.5°C throughout the experiment. BHK-21 cells transfected with pEAV030, pEAV030-2aKO, or pEAV030-2bKO RNA were labeled from 10 to 14 h posttransfection. Labeling of EAV- and mock-infected BHK-21 cells took place at 7 h post infection (p.i.), again following a 30-min starvation period. In each case the labeling medium (DMEM without l-cystine, l-glutamine, and l-methionine [BioWhittaker], supplemented with 5% dialyzed FCS, 10 mM HEPES [pH 7.4], and 2 mM l-alanyl-l-glutamine) contained 150 μCi of protein labeling mix ([³⁵S]Met-[³⁵S]Cys; Amersham Pharmacia Biotech or New England Nuclear) per ml. After being labeled, the cells were placed on ice, and total lysates of cells and media were prepared. This was done in the absence or presence of the sulfhydryl blocking agent iodoacetamide, as previously described (19), with the modification that the sodium dodecyl sulfate (SDS) concentration in the postnuclear supernatants was increased to 0.5%. Alternatively, cells were rinsed twice with complete medium containing 2 mM l-methionine and 4 mM l-cysteine hydrochloride monohydrate and incubated for different chase periods prior to lysis. When indicated, the translation inhibitor cycloheximide (500 μM) was present in the chase medium.

Preparation of radiolabeled virions.

Subconfluent BHK-21 monolayers (80-cm² flasks) were infected with EAV at a multiplicity of infection of 20 PFU per cell. At 6 h p.i., the medium was replaced by Met- and/or Cys-free medium containing either 100 μCi of [³⁵S]Met (ICN), 100 μCi of [³⁵S]Cys (ICN), or 100 μCi of [³⁵S]Met-[³⁵S]Cys per ml as indicated in the text. Three hours later, an equal amount of label was added to the medium, followed at 12 h p.i. by one-sixth volume of complete medium. The culture supernatants were harvested at 20 h p.i., and cell debris was removed by low-speed centrifugation. The [³⁵S]Met-[³⁵S]Cys-labeled virus was pelleted through a cushion of 20% (wt/wt) sucrose in 20 mM Tris-HCl (pH 7.6)–100 mM NaCl–1 mM EDTA (TNE) by overnight centrifugation in an SW 28 rotor (Beckman) at 15,000 rpm and 4°C and used for immunoprecipitation analyses. The [³⁵S]Met- or [³⁵S]Cys-labeled virus preparations were subjected to polyethylene glycol precipitation (16), resuspended in ice-cold TNE, and further purified by ultracentrifugation through a 20 to 50% (wt/wt) sucrose gradient for 16 h at 30,000 rpm and 4°C, using an SW 41 Ti rotor (Beckman). Gradient fractions (500 μl) were collected, and a part of each aliquot was analyzed by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) (36).

Protein analysis.

Immunoprecipitations, endoglycosidase H digestions, and glycopeptidase F treatments were performed by established procedures (19). Immunoprecipitated proteins were analyzed in SDS–20% polyacrylamide gels under nonreducing or reducing conditions. Prior to fluorography, gels were fixed in 10% acetic acid–50% methanol–0.005% Coomassie brilliant blue R250 for 30 min, soaked in 1 M sodium salicylate (30 min) (10), and dried.

Immunofluorescence assays.

The preparation of cells for indirect immunofluorescence assays was done essentially as described previously (61). The ORF 2a protein-specific rabbit antiserum was used at a dilution of 1:200 to 1:400. Tissue culture supernatants of the anti-EAV G_L (MAb 93B), anti-PDI (MAb 1D3), and anti-Golgi protein (MAb F20/65-1-4) hybridoma cell lines were used at dilutions of 1:40, 1:10, and 1:20, respectively. As secondary antibodies, a Cy3-conjugated donkey anti-rabbit immunoglobulin G antibody and a fluorescein isothiocyanate-conjugated goat anti-mouse immunoglobulin G antibody (both from Jackson ImmunoResearch Laboratories) were used at dilutions of 1:400 to 1:800 and 1:100, respectively. The samples were examined with an Olympus fluorescence microscope.

Electron microscopy.

For cryoimmunoelectron microscopy, EAV-infected BHK-21 or RK-13 cells were fixed in 200 mM HEPES (pH 7.4) containing 4% paraformaldehyde and 0.1% glutaraldehyde. Cells were scraped from the dish, pelleted, and incubated in 2.3 M sucrose for 1 h at room temperature. Subsequently, cell pellets were mounted on silver pins, flash frozen, and stored in liquid nitrogen. The specimens were sectioned with a Reichert Ultracut S ultramicrotome equipped with a Reichert FCS cryo attachment using a Drukker International diamond knife. Immunocytochemical labeling of thawed cryosections was performed essentially as described by Griffiths et al. (31), and specimens were examined in a Philips CM100 transmission electron microscope.

Arterivirus sequences and computer-assisted sequence analyses.

The sequence of the structural-protein-coding region of the genome of SHFV strain LVR 42-0/M6941 (53) and the full-length genomic sequences of the EAV Bucyrus strain (14), the Lelystad isolate of PRRSV (45), and the LDV P strain (49) (accession no. U63121, X53459, M96262, and U15146, respectively) were extracted from the EMBL/GenBank database. Deduced protein sequences were analyzed by using DNASIS software (version 2.1; Hitachi Software Engineering) or with the aid of software available from the Dutch National Expertise Center for Computer-Aided Chemistry and Bioinformatics (CAOS/CAMM), Nijmegen, The Netherlands), including the GCG (version 8.1; Genetics Computer Group, Madison, Wis.) and EGCG (extended GCG; version 8.1.0; Peter Rice, The Sanger Centre, Cambridge, United Kingdom) packages.

RESULTS

Identification of a previously unnoticed but conserved arterivirus gene.

Analysis of the genetic information on the four arteriviruses (14, 28, 45, 49, 53) has resulted in the identification of a set of putative genes and a set of conserved mRNA TRSs in the 3′-terminal portions of their genomes. The coding assignments for many of these relatively large arterivirus genes (Fig. 1 and 3B) have now been firmly established, and the corresponding proteins have been characterized (see reference 54 for a review). Several small but potentially functional ORFs of 50 to 100 codons were also identified. However, since most of these small ORFs completely overlapped with a larger gene and were not the most 5′-proximal ORF in one of the viral subgenomic mRNAs, their expression was considered unlikely.

Following the recent construction of infectious cDNA clones for EAV (27, 63), we set out to establish the functions of the various viral gene products. To this end, the expression of individual genes was abolished by mutating the translation initiation codon (unpublished data). During a detailed analysis of the region containing the 5′ end of ORF 2 (Fig. 2 and 3A), it was noticed that (i) the AUG codon of ORF 2 appears to be in a particularly poor context (UUGAUGC) for translation initiation by conventional ribosomal scanning (ACCAUGG being the optimal context [34]) and (ii) another AUG codon in a reasonably favorable context (GUGAUGG) is located upstream of ORF 2 in subgenomic mRNA 2. This upstream AUG codon, which overlaps with the opal termination codon of ORF 1b (Fig. 2), is followed by a potential ORF of 66 additional codons, encoding a polypeptide with a calculated molecular mass of 7.4 kDa. We have termed this newly detected EAV reading frame ORF 2a, whereas the former ORF 2 will henceforth be referred to as ORF 2b.

To assess the significance of our finding, the sequences of the other three arteriviruses were examined. In the LDV genome (Fig. 2), a similar situation was encountered: an AUG codon (context, GUAAUGG) overlaps with the ORF 1b termination codon and is followed by an ORF of 69 additional codons. This small LDV ORF was already noticed by Chen et al. upon examination of the 3′ part of the LDV genome but was not studied or described in any detail (11). It specifies a polypeptide with properties that are remarkably similar to those of the EAV 7.4-kDa polypeptide (see below). Again, the initiation codon of the downstream ORF 2b (the former ORF 2) is in a suboptimal context (AAGAUGC) and is in fact the fourth AUG codon downstream of the mRNA 2 TRS. Also, in the PRRSV sequence, the homolog of EAV/LDV ORF 2a was readily identified, although in this case the relative position of the two ORFs is reversed (Fig. 2 and 3A). The AUG codon of the ORF encoding GP₂ (the PRRSV equivalent of EAV G_S) precedes the initiation codon of the EAV/LDV ORF 2a homolog by 5 nt. Consequently, the nomenclature for these two ORFs should also be reversed (Fig. 3A).

Finally, the SHFV sequence was an interesting case, in view of the postulated duplication of the original ORFs 2 to 4 by RNA recombination (Fig. 3B) (29). The SHFV homolog of the novel ORF was identified downstream of the putative three-gene duplication (ORFs 2a, 2b, and 3) (Fig. 3B). It starts upstream of and partially overlaps with ORF 4b (formerly ORF 4) and fills the small noncoding gap between ORFs 3 and 4b. Consequently, the novel SHFV gene was designated ORF 4a. Remarkably, in all arteriviruses the expression of the novel ORF seems to require the use of a bicistronic mRNA, since only one known TRS is present immediately upstream of ORFs 2a and 2b in EAV/LDV/PRRSV and ORFs 4a and 4b in SHFV (Fig. 2).

In conclusion, homologous short ORFs were identified at corresponding positions in the genomes of all four arteriviruses. For reasons explained below, and to avoid possible confusion due to the different numbering of the novel ORF (2a, 2b, or 4a) in the genomes of different arteriviruses, we will from now on refer to the product encoded by the newly discovered gene as the envelope (E) protein.

Computer analysis of the EAV E protein and its homologs.

Like all other arterivirus structural proteins, the E protein is clearly most conserved between LDV and PRRSV (43% identical residues) (Fig. 4A). The EAV E protein occupies an intermediate position, whereas its SHFV equivalent is most divergent and is also 10 to 13 residues longer than its three homologs. Still, two noticeable properties have been conserved in all four proteins: (i) a central hydrophobic domain of 30 to 40 residues (Fig. 4B) and (ii) a cluster of basic amino acid residues in the hydrophilic C-terminal domain of the protein (Fig. 4A). Although an N-terminal signal sequence is lacking, the presence of the central hydrophobic region strongly suggests that the E protein is an integral membrane polypeptide. In principle, the hydrophobic core of the protein is long enough to span the membrane twice, but computer analyses aimed at predicting the number of transmembrane domains and the topology of the E protein in the membrane produced inconsistent results.

FIG. 4.

Analysis of arterivirus E proteins. (A) Alignment of the sequences of the E proteins of EAV, LDV, PRRSV, and SHFV. Absolutely conserved residues are indicated by asterisks; basic amino acid residues are underlined, to show their clustering in the C-terminal domain of the protein; dashes correspond to gaps introduced for optimal alignment. The conserved potential N-terminal myristoylation site (G-[not E, D, R, K, H, P, F, Y, or W]-X-X-[S, T, A, G, N, or C]-[not P]) and the internal phosphorylation site for casein kinase II ([S or T]-X-X-[D or E]) are indicated by arrows. (B) Hydrophobicity analysis of the arterivirus E proteins, using the method of Kyte and Doolittle and a 9-residue moving window (35). aa, amino acids.

To identify possible sites for co- or posttranslational modifications, the deduced E protein sequences were searched for the presence of processing signals by using the PROSITE database (2). Interestingly, the N terminus of each of the E proteins contains a potential N-myristoylation site (Gly-2) followed by a phosphorylation site for casein kinase II (Ser-22) (Fig. 4). The EAV and SHFV E proteins further possess potential tyrosine-based endocytosis signals (YRPV and YHTLP, respectively) at their extreme C termini, but this feature is not conserved in the LDV and PRRSV E proteins (5).

Identification and characterization of the EAV E protein.

To investigate whether EAV ORF 2a is expressed and to study the characteristics of its product, an antiserum was raised by using a synthetic peptide which represented the 17 most C-terminal residues of the predicted EAV E protein sequence. Immunoprecipitations using this antipeptide serum and the control preimmunization serum were carried out on lysates of EAV- or mock-infected BHK-21 cells which had been labeled with ³⁵S[Met]-³⁵S[Cys] from 7 to 8 h p.i. The antipeptide serum specifically precipitated a protein of the expected size (approximately 8 kDa; predicted size, 7.4 kDa) from the lysate of EAV-infected cells (Fig. 5A, left panel). Both the pre- and postimmunization sera also brought down the 12-kDa EAV N protein. As described before, this is due to the fact that the N protein binds to the protein A molecules on the surface of the Staphylococcus aureus cells used to precipitate the immune complexes (60).

FIG. 5.

Identification and stability of the EAV E protein in infected cells and upon independent expression in the MVA-T7 system. (A) The left panel shows an immunoprecipitation analysis of lysates of EAV (V)- or mock (M)-infected BHK-21 cells that were ³⁵S labeled for 60 min at 7 h p.i. Immunoprecipitations were performed with the E protein-specific antiserum (E) or the corresponding preimmunization serum (pE). The positions of the EAV nucleocapsid protein (N), the full-length E protein (E), and a truncated version of the E protein (E_i), which was only observed after a long exposure of the gel, are indicated. The right panel shows the results of a pulse-chase experiment. Cells were pulse labeled for 15 min and subsequently chased for the time periods (in minutes) indicated above the lanes. Protein synthesis during the chase was inhibited by the addition of cycloheximide. The numbers on the right represent the sizes (in kilodaltons) of marker proteins (M_r) run in the same SDS–20% polyacrylamide gel. (B) Corresponding analysis of the independently expressed EAV E protein, using lysates of MVA-T7-infected and pAVI02a (2a)- or pAVI16 (6)-transfected OST-7.1 cells that were labeled at 6 h p.i., as described above for the EAV-infected cells. Ab, antibody.

To confirm that the 8-kDa protein indeed represented the EAV E protein, ORF 2a was expressed from plasmid pAVI02a by using the MVA-T7 expression system (58). The EAV ORF 6 expression vector pAVI16 was used as a negative control. After metabolic labeling of the transfected cells from 6 to 7 h p.i., cell lysates were prepared and used for immunoprecipitation studies. As shown in the left panel of Fig. 5B, the antipeptide serum mainly recognized an 8-kDa protein in the lysate of pAVI02a-transfected cells. Furthermore, a minor 5.5-kDa polypeptide was immunoprecipitated (indicated as E_i in Fig. 5B). Neither protein species was precipitated when preimmunization serum or lysates from pAVI16-transfected cells were used. The largest expression product comigrated perfectly with the 8-kDa protein precipitated from EAV-infected cell lysates (data not shown) and was therefore concluded to be the full-length EAV E protein.

The precise nature of the minor pAVI02a expression product (5.5 kDa) is obscure, but its recognition by the antipeptide serum suggests that it is an N-terminally truncated version of the E protein. This product might result from translation initiation at the second AUG codon in ORF 2a (Fig. 2). However, since various amounts of the truncated E protein were found in different experiments, it is also possible that the 5.5-kDa protein was a degradation product which was generated during or after cell lysis. Remarkably, the E protein always migrated as a fuzzy band during SDS-PAGE. This was not merely the result of its low molecular mass, since marker proteins within the same size range showed up as discrete bands (Fig. 5). When analyzed in gels containing 10% glycerol to reduce radial diffusion, the E protein exhibited the same heterogeneous electrophoretic mobility (data not shown).

Pulse-chase experiments using EAV-infected cells and the MVA-T7 expression system showed that the EAV E protein is very stable. In EAV-infected cells, most of the E protein labeled during a 15-min pulse was still present after a 4-h chase (Fig. 5A, right panel). Similar results were obtained when the E protein was expressed independently (Fig. 5B, right panel). The pulse-chase analyses further revealed that the E protein did not undergo detectable molecular mass changes during the chase and therefore seems not to be extensively processed after its synthesis. Consistent with the absence of potential N-glycosylation sites, treatment of the E protein with endoglycosidase H or glycopeptidase F did not influence its apparent molecular mass (data not shown). Further, by preparing cell lysates in the presence of an alkylating agent, we investigated whether the single Cys residue in the E protein formed intermolecular disulfide bonds (19). Using this method, covalently linked multimers of the EAV E protein were not detected in EAV-infected BHK-21 cells or upon independent expression (data not shown).

Subcellular localization of the EAV E protein.

The subcellular localization of the E protein in EAV-infected BHK-21 cells and upon expression in the MVA-T7 system was studied by immunofluorescence microscopy (Fig. 6A). From the extensive reticular and perinuclear staining, it was clear that the E protein associated with intracellular membranes. Double labeling of EAV-infected cells (Fig. 6A, left panel) for the E and G_L proteins revealed that part of the E molecules localized to the Golgi complex (61). Another fraction of the E protein resided in the ER, as evidenced by colocalization with the ER-resident folding factor PDI. The staining pattern strongly resembled that of the EAV M protein, which has been found to localize to the ER and the Golgi complex (unpublished data). Inspection of cells fixed earlier or later in infection did not reveal any clear changes in the subcellular localization of the E protein (data not shown).

FIG. 6.

Subcellular localization of the EAV E protein. (A) Immunofluorescence analysis of the localization of the EAV E protein in infected BHK-21 cells (8 h p.i.) (left panel) and upon independent expression in the same cells, using the MVA-T7 system and expression vector pAVI02a (7 h p.i.) (right panel). Cells were double labeled for the EAV E protein and either an ER marker (PDI) or a marker of the Golgi complex (the EAV G_L glycoprotein in infected cells, or the protein recognized by MAb F20/65-1-4 in transfected cells). In both systems, part of the E protein was seen in the Golgi complex, but most of it colocalized with PDI. Staining of mock-infected cells and MVA-T7-infected, pAVI16-transfected cells with the E protein-specific antiserum did not yield a detectable signal (data not shown). Bar, 25 μm. (B) Immunogold labeling of cryosections of EAV-infected RK-13 cells (8 h p.i.) with the E protein-specific antiserum and protein A coupled to 10-nm-diameter gold particles. A specific but not very abundant labeling of the membranes of the Golgi complex and ER (not shown) was observed. Bar, 100 nm.

To examine whether the intracellular distribution of the E protein is an intrinsic property or it depends on the presence of other EAV components, we also studied its localization upon independent expression in the MVA-T7 system. BHK-21 cells were transfected with pAVI02a and stained for the E protein and either PDI or a poorly defined Golgi protein. As is evident in the right panel of Fig. 6A, the intracellular distribution of the E protein largely overlapped with that of PDI. Still, a limited amount of the protein was found in the Golgi complex. Comparable results were obtained with pAVI02a-transfected OST-7.1 cells (data not shown). Incubation of transfected cells for another 2 h in the presence of the translation inhibitor cycloheximide did not significantly alter the subcellular distribution of the E protein (data not shown).

To investigate the subcellular localization of the E protein at the ultrastructural level, EAV-infected cells were processed for cryoimmunoelectron microscopy. A specific but not very abundant labeling of the E protein was observed. The protein was detected in the ER (data not shown) and the Golgi complex (Fig. 6B), and weak labeling of the plasma membrane was also observed in some cells (data not shown). Taken together, our data clearly show that the intracellular form of the E protein is membrane associated.

The EAV E protein is a virion component.

To determine whether the E protein is present in virions, ³⁵S[Met]-³⁵S[Cys]-labeled EAV particles were prepared by metabolic labeling of infected BHK-21 cells and purified by sedimentation through a 20% (wt/wt) sucrose cushion. As a negative control, the culture supernatant from mock-infected cells was used. Direct analysis of the virus pellet by SDS-PAGE showed that it contained an 8-kDa protein which comigrated with the E protein from lysates of EAV-infected cells (data not shown). Immunoprecipitations with the E protein-specific antiserum confirmed the identity of this 8-kDa polypeptide (Fig. 7A). To verify that the EAV E protein was a virion protein and not a nonstructural contaminant of virus preparations, ³⁵S[Met]- and ³⁵S[Cys]-labeled virus particles were successively subjected to polyethylene glycol precipitation and sucrose density gradient centrifugation. This analysis revealed that the E protein band was concentrated in the gradient fractions which also contained the bulk of the other known structural polypeptides of EAV (Fig. 7B). Because these fractions also contained the highest level of specific infectivity, we concluded that the EAV E protein is indeed a structural protein.

FIG. 7.

Identification of the E protein in EAV particles. (A) Analysis of pellets obtained after ultracentrifugation through a 20% (wt/wt) sucrose cushion of supernatants from mock- or EAV-infected BHK-21 cells that were labeled with ³⁵S[Met]-³⁵S[Cys]. Pellets were analyzed directly (−) or resuspended and subjected to immunoprecipitation analysis with E protein-specific antiserum (E) or the preimmunization serum (pE). The positions of the EAV E protein (8 kDa), N protein (apparent molecular mass, 14 kDa), M protein (16 kDa), and the small (G_S) and large (G_L) glycoproteins (25 kDa and 30 to 42 kDa, respectively) are shown at the left. The positions and sizes (in kilodaltons) of marker proteins (M_r) analyzed in the same gel are indicated at the right. Ab, antibody. (B) Sucrose density gradient centrifugation of [³⁵S]Met- or [³⁵S]Cys-labeled EAV preparations. The numbers of the gradient fractions and the position of each sample relative to the top and bottom of the centrifuge tube are indicated, as are the positions of the EAV structural proteins E, N, M, G_S, and G_L. Note that the N protein does not contain any Cys residues.

The EAV E and G_S proteins are both required for virus infectivity.

The recent construction of EAV infectious cDNA clone pEAV030 (63) allowed us to test the importance of EAV ORF 2a expression by means of reverse genetics. To abolish the expression of ORF 2a, its translation initiation codon was mutated from AUG to AUA. This mutation does not affect the translation of ORF 2b or that of replicase ORF 1b, whose termination codon overlaps with the ORF 2a start codon (Fig. 2). As a control, a similar strategy was employed to inactivate the expression of ORF 2b.

Full-length transcripts of the ORF 2a and ORF 2b knockout constructs (pEAV030-2aKO and pEAV030-2bKO, respectively) were generated in vitro and transfected into BHK-21 cells. RNA derived from the wild-type plasmid pEAV030 served as a positive control. The transfected cells were ³⁵S[Met]-³⁵S[Cys] labeled from 10 to 14 h posttransfection, and viral protein synthesis was analyzed by immunoprecipitation with a set of EAV-specific antisera (Fig. 8). This biochemical analysis confirmed the complete inactivation of ORF 2a (E protein) expression in EAV030-2aKO and the abrogation of ORF 2b (G_S protein) expression in EAV030-2bKO. The expression of the replicase proteins (e.g., nsp2 [Fig. 8]) and other structural proteins (e.g., M and N [Fig. 8]) was not affected.

FIG. 8.

Viral protein synthesis in BHK-21 cells transfected with EAV030-2aKO (2a⁻), EAV030-2bKO (2b⁻), and wild-type (wt) EAV030 RNA. ³⁵S labeling was carried out from 10 to 14 h posttransfection, and immunoprecipitations were performed with antisera recognizing the replicase cleavage product nsp2 (56), the EAV E protein (see the text), the EAV G_S protein (16), and a mixture of antibodies recognizing the EAV M and N proteins (16). The positions of the different structural proteins and nsp2 are displayed at the left. Numbers on the right indicate sizes in kilodaltons. Ab, antibody.

At 12 and 24 h after electroporation, cells were fixed and processed for immunofluorescence staining (data not shown). As before (63), antisera recognizing EAV nonstructural and structural proteins were used to confirm that genome RNA replication and the transcription of subgenomic mRNAs were taking place. Analysis of samples fixed at 12 h after electroporation showed that a transfection efficiency of 40 to 50% had been achieved. The mutant viral genomes apparently replicated with the same efficiency as the wild-type EAV030 transcript. The production of subgenomic mRNAs was not substantially affected, as evidenced by the fact that comparable staining intensities were obtained for the major structural proteins G_L, M, and N in cells transfected with any one of the three transcripts. Moreover, localization of the three major structural proteins was not influenced by the inactivation of E or G_S protein expression.

Immunofluorescence analysis of the cells fixed after 24 h (i.e., after approximately two cycles of virus replication) strongly suggested that both EAV030-2aKO and EAV030-2bKO were defective in the production of infectious progeny. Whereas efficient spread of virus to initially uninfected (i.e., untransfected) cells was observed with the wild-type transcript, the number of EAV-positive cells in the EAV030-2aKO and EAV030-2bKO transfections remained constant. Nevertheless, a small number of the EAV030-2aKO-transfected cells now appeared to express the E protein, suggesting that revertants of the ORF 2a AUG knockout mutation had been generated. Analysis of cells fixed at later time points (48 and 72 h posttransfection) confirmed this hypothesis: spread of the progeny of EAV030-2aKO (but not of EAV030-2bKO) was observed, and a rapidly increasing number of cells stained positive for the E protein.

The reversion of EAV030-2aKO to a wild-type phenotype was further analyzed in infectious-center assays. Immediately after the electroporation, transfected cells in 10-fold serial dilutions were mixed with sufficient untransfected BHK-21 cells to give a confluent monolayer. The cells were allowed to attach to the surface of the culture dish for 4 h, after which an agarose overlay was applied. After 3 days, 31 plaques were obtained from a dilution containing about 60 cells transfected with the wild-type transcript. This value is consistent with the transfection efficiency of about 50% observed in the immunofluorescence assay. The pEAV030-2bKO transcript did not yield any plaques in the lowest dilution, in which about 60,000 transfected cells (again estimated to be 40 to 50% positive) were plated. Similar results were obtained with AUG knockouts of other structural genes (unpublished data), indicating that in general the reversion frequency of this kind of mutation was on the order of 10⁻⁴ (or less).

In contrast, the EAV030-2aKO transfection yielded 32 and 3 plaques in dilutions containing 60,000 and 6,000 cells, respectively. Although this was 1,000-fold less than for the wild-type construct, the (reproducible) difference with similar knockout mutations for other genes was remarkable. Since the average size and morphology of the EAV030-2aKO plaques were similar to those of the wild-type virus, we consider it unlikely that inactivation of ORF 2a expression results in reduced or delayed virus yields or in the release of poorly infectious virus. In view of our observations during the immunofluorescence analysis (see above), it was more likely that revertants of the ORF 2a knockout mutation arose with a relatively high frequency (see also Discussion).

Subsequently, three putative EAV030-2aKO revertants were plaque purified and used to infect fresh BHK-21 cells. Intracellular RNA was isolated and used to amplify the region containing ORF 2a by reverse transcription (RT)-PCR. Direct sequence analysis of the RT-PCR product confirmed that the ORF 2a start codon had been restored in these revertants: the AUA mutant codon had reverted back to AUG. Interestingly, an additional, translationally silent mutation 4 nt downstream (T-9758 to C) which had been introduced to engineer a StuI marker restriction site had not been repaired. The latter observation ruled out contamination with the wild-type virus and confirmed that the G-9754-to-A substitution, which had knocked out ORF 2a expression in the EAV030-2aKO mutant, indeed reverted with an unusually high frequency.

DISCUSSION

A newly discovered structural protein of arteriviruses.

In this paper, we have described the identification of a previously unnoticed gene (ORF 2a) in the genome of EAV. As expected from its conservation in the other three arteriviruses (Fig. 2 to 4), this novel ORF indeed represents a functional gene, and its product was detected in EAV-infected cells (Fig. 5 and 6) and in virus particles (Fig. 7). Our experiments showed that the protein is very stable and localizes to the ER and Golgi complex both in EAV-infected cells and when expressed independently. Finally, we employed reverse genetics to demonstrate that expression of EAV ORF 2a (and also of ORF 2b) is required to produce infectious progeny virus. In view of its predicted hydrophobic nature, its membrane association, and its presence in virus particles, we have chosen the provisional name envelope (E) protein for the EAV ORF 2a product.

So far, four structural proteins had been identified for EAV, LDV, and SHFV (16, 23, 30), while PRRSV was reported to possess six virion proteins (42, 46, 47, 64). The structural roles of the N protein, M protein, and major glycoprotein (G_L/VP-3P/GP₅/p54) are undisputed. The last protein has been demonstrated to serve as a target for neutralizing antibodies in EAV, LDV, and PRRSV (see reference 54 for references). The products of EAV/LDV ORF 2b and PRRSV ORF 2a (formerly ORF 2) are only minor structural components (16, 19, 23, 46). In the case of PRRSV, the ORF 4 protein was also shown to be present in virions and was identified as a second target for neutralizing antibodies (48, 64, 68). Recently, the ORF 4 protein was also identified as a structural protein in EAV (unpublished data). The structural role of the ORF 3 protein is still being debated, in view of the apparently conflicting data on its presence in virus particles (21, 33, 41, 64, 68). Mardassi et al. (41) recently reported that the PRRSV ORF 3 protein is a secretory protein that is not associated with virions, and similar claims have been made for the LDV ORF 3 protein (24).

In any case, the newly discovered E protein further extends the collection of arterivirus structural proteins, which already was quite unusual among positive-stranded RNA viruses in terms of the number of proteins and their (predicted) structural properties. Arterivirions can now be assumed to contain at least six, and possibly even seven, structural proteins. With the exception of the N protein, all of these polypeptides are integrated into or associated with the viral envelope.

The arterivirus E protein is expressed from a bicistronic mRNA.

Our data probably constitute the first experimental proof of a functionally bicistronic subgenomic mRNA in arteriviruses, since we have now detected the translation products of EAV ORFs 2a and 2b in infected cells. We cannot formally exclude the possibility that ORF 2a is expressed directly from the viral genome by reinitiation after termination of ORF 1b translation. However, a strong argument against this possibility is the observation that ORF 2a is the most 5′-proximal ORF in subgenomic mRNA 2. If ORF 2a was translated directly from mRNA 1, one would expect leader-to-body joining to occur downstream of the ORF 2a start codon, thereby allowing the efficient translation of ORF 2b from mRNA 2. It is also unlikely that two separate mRNAs are generated to express ORFs 2a and 2b. A previous analysis of the EAV mRNA leader-to-body junctions (13, 15) indicated that there is only one functional signal for subgenomic mRNA synthesis in the region directly upstream of ORF 2b. This TRS is located in the 3′ end of ORF 1b, 35 nt upstream of the ORF 2a initiation codon, and not between the ORF 2a and 2b translation start sites (Fig. 2). Thus, it is probably the expression of ORF 2b that requires a less orthodox translational mechanism, whereas ORF 2a is most likely expressed from mRNA 2 by conventional ribosomal scanning (34) and translation of the most 5′-proximal ORF.

A detailed analysis of the situation in the other three arteriviruses strongly suggests that bicistronic mRNAs are also used to express ORFs 2a and 2b in LDV and PRRSV and ORFs 4a and 4b in SHFV. Again, in each of these viruses, only a single functional TRS has been identified upstream of these pairs of genes (11, 29, 44). Remarkably, the relative position of the genes encoding the E protein and the G_S homolog (GP₂) in PRRSV is reversed compared to that in EAV, LDV, and SHFV (Fig. 3). It is noteworthy, however, that Thr-16 in GP₂ of most European PRRSV strains can be aligned with Met-1 of the LDV ORF 2b protein and that American PRRSV isolates even possess a Met residue at this position (data not shown). Thus, compared to its LDV equivalent, PRRSV GP₂ appears to contain an N-terminal extension of 15 amino acids, a change which is linked to the reversed gene arrangement. The consequences of the altered gene order for PRRSV ORF 2a and 2b translation are unknown, since their initiation codons are both in a favorable context and are separated by only 2 nt.

The mechanism of translation of the second ORF of mRNA 2 in EAV/LDV/PRRSV and mRNA 4 in SHFV is unclear. In all cases, the initiation codon of the upstream ORF (2a or 4a) is in a good context. This argues against a leaky scanning mechanism, even though no additional AUG triplets are present in the region between the initiation codons of the two ORFs in EAV, PRRSV, and SHFV (Fig. 2). Furthermore, the removal of the ORF 2a translation initiation codon in construct pEAV030-2aKO did not have a dramatic effect on the expression level of the G_S protein (Fig. 8). The presence of an internal ribosomal entry site which directs translation of the more-downstream ORF is an alternative explanation. Such translation-regulating RNA sequences have been identified in a number of other viral and cellular mRNAs (for a review, see reference 51). However, since the major part of internal ribosomal entry site elements is usually located upstream of the translation initiation codon, the situation in both SHFV (with its presumed gene duplication upstream of ORFs 4a and 4b) and PRRSV (in which the order of ORFs 2a and 2b is reversed) seems to argue against the presence of a conserved RNA structure in this region of the genome.

High reversion frequency of the ORF2a knockout mutation.

As described above, a significantly larger number of revertants arose in cells transfected with EAV030-2aKO than in EAV030-2bKO-transfected cells. A possible explanation for this difference is the fact that the ORF 2a initiation codon was mutated from AUG to AUA while the ORF 2b AUG codon was changed to ACG. During replication, the EAV genome (and antigenome) may be susceptible to the activity of a double-stranded RNA adenosine deaminase(s) which is present in the host cell (3). This enzyme can convert adenosine into inosine, which preferentially forms a base pair with cytidine (instead of uridine), thereby causing a U-to-C transition in the complementary strand and, upon a subsequent round of replication, the replacement of the original A by a G in the plus strand. Thus, the AUA mutant codon at the 5′ end of ORF 2a might be preferentially converted into AUG, which would explain the high reversion frequency of this mutation compared to that of the inactivated ORF 2b start codon (ACG). Similar A-to-G hypermutation phenomena have previously been reported for a number of negative-stranded RNA viruses and retroviruses (see reference 3 for a review) but, to our knowledge, not for any positive-stranded RNA virus.

Properties and function of the arterivirus E protein.

The apparent molecular mass of the EAV E protein (as estimated by SDS-PAGE) closely matched the molecular mass calculated from the primary amino acid sequence, and no alterations in size were observed during pulse-chase experiments. Still, we cannot exclude the possibility that the E protein undergoes co- or posttranslational modifications which do not have a noticeable effect on its electrophoretic mobility. Indeed, careful inspection of a number of gels revealed that the E protein always migrates as a doublet of between 6.5 and 10.5 kDa. This might indicate that (at least) two forms of the protein exist, which might explain its fuzzy appearance in SDS-PAGE gels. As described in the Results, the N terminus of the arterivirus E proteins invariably contains a potential N-myristoylation site followed by a phosphorylation site for casein kinase II. The possible significance of these sequence elements is underlined by the fact that they are also conserved in a group of PRRSV isolates in which the N terminus of the E protein is extended by 3 amino acids (see, e.g., reference 43). A similar combination of posttranslational modifications has been found in the myristoylated, alanine-rich C kinase substrate (MARCKS), which cycles between membrane and cytosolic compartments in a regulated fashion by means of a myristoyl-electrostatic switch (reviewed in reference 4). Modification of the E protein by casein kinase II would require its N terminus to be located on the cytoplasmic side of the membrane. Myristoylation might occur both in the cytoplasm and in the ER (7) and could have many implications for E protein function, in terms of both protein-lipid and protein-protein interactions. Myristoyl moieties, for example, have been reported to play a role in virus particle formation, receptor recognition, virion stability, and virus disassembly (for a review, see reference 12). Likewise, phosphorylation of the E protein could modulate its function as a result of a change in three-dimensional structure or intracellular trafficking.

The hydropathy profile of the E protein suggests that it is an integral membrane protein with an uncleaved signal-anchor sequence in the central part of the molecule. If the protein spans the membrane only once, the large size and hydrophobicity of the internal hydrophobic domain, the small acidic N terminus that may be easily translocated, and the abundance of basic residues at the C-terminal side of the signal-anchor sequence (67) would be most compatible with a type III orientation (N_exoC_cyt [57]). Alternatively, the E protein may span the lipid bilayer twice, with both termini located at the cytoplasmic face of the membrane. In the latter case, the conserved Asp-25 may interact in the plane of the membrane with the first downstream Arg (residue 51 in EAV) to stabilize the polytopic transmembrane structure. Finally, we cannot exclude the possibility that the E protein can adopt multiple transmembrane configurations, as has been described for some other viral envelope proteins (see, e.g., references 39 and 59).

The arterivirus E protein structurally resembles a number of small hydrophobic proteins found in other RNA viruses (for a review, see reference 8). All of these proteins seem to disturb lipid bilayers and have been implicated in diverse processes, such as (i) the formation of aqueous pores or selective ion channels, (ii) the vesiculation of intracellular membranes, (iii) the induction of membrane curvature, and (iv) the pinching-off reaction of budding virions. The expression of these small membrane-active proteins may further result in the development of cytopathic changes and inhibition of the secretory pathway and may be required to enhance particle release or to promote virus uncoating.

Interestingly, coronaviruses also code for a small hydrophobic membrane protein (reviewed in reference 52). This so-called E protein was detected both in virus-infected cells and in virions and appears to play a crucial role during virus morphogenesis (25). In recent studies, coexpression of the M and E proteins of mouse hepatitis virus was shown to be both necessary and sufficient to generate coronavirus-like particles (6, 66). In view of the evolutionary relationship between the two virus groups and the structural similarities between their triple-spanning M proteins, it is tempting to speculate that the E proteins of arteriviruses and coronaviruses are functional homologs. On the other hand, the coronavirus E protein seems to be present in virus particles at a much lower rate than the EAV E protein, and the other structural proteins of arteri- and coronaviruses are clearly different (for reviews, see references 9, 17, and 54). Furthermore, the simultaneous expression of EAV ORFs 2a, 5, and/or 6 in vaccinia virus-based expression systems did not result in the production of subviral particles (data not shown). Thus, either the E protein of arteriviruses is functionally distinct from that of coronaviruses or the production of arterivirus-like particles requires the expression of additional viral genes.