Membrane Organization of Bluetongue Virus Nonstructural Glycoprotein NS3

Abstract

The smallest RNA segment (S10) of bluetongue virus (an orbivirus, family Reoviridae) encodes two closely related nonstructural proteins, the 229-amino-acid (aa) NS3 and the 216-aa NS3A. The proteins are found in glycosylated and nonglycosylated forms in infected cells (X. Wu, H. Iwata, S.-Y. Chen, R. W. Compans and P. Roy J. Virol. 66:7104–7112, 1992). The NS3/NS3A proteins have two hydrophobic domains (aa 118 to 141 and 162 to 182) and two potential asparagine-linked glycosylation sites (aa 63 and 150), one of which is located between the hydrophobic domains. To determine whether these features were used in the mature protein forms, we generated a series of mutants of the S10 gene and expressed them by using the vaccinia virus T7 polymerase transient-expression system. Our data indicate that both hydrophobic domains of NS3 span the cell membrane and that only the site at aa 150 is responsible for N-linked glycosylation of the NS3 proteins.

Bluetongue virus (BTV), a member of the Orbivirus genus, family Reoviridae, synthesizes four nonstructural proteins in infected cells (NS1, NS2, NS3, and NS3A) in addition to the seven structural proteins (VP1 to VP7) of the mature virus particle. Unlike the NS1 and NS2 proteins, the two smallest nonstructural proteins, NS3 (25.5 kDa) and NS3A (24 kDa) accumulate to very low levels in BTV-infected mammalian cells, preventing detailed analyses of their functions in these cells. Interestingly, more NS3 protein is made in BTV-infected insect cells (12), suggesting that there may be different functions for the proteins in different hosts or cell types. Both NS3 (229 amino acids [aa]) and NS3A (216 aa) proteins are encoded by BTV S10 RNA and are read from alternative in-frame methionine initiation codons in the S10 mRNA (for a review, see reference 22). We have reported previously the synthesis of NS3 proteins by using the baculovirus expression system (9). NS3A expression is more variable than that of NS3 and generally occurs at a much lower level (see below). When the first ATG codon is deleted from the S10 gene, NS3 synthesis is abolished and NS3A accumulates at higher levels (25). Intracellular and cell surface immunofluorescence studies have shown that newly synthesized NS3 proteins are transported to the Golgi apparatus and then to the cell membrane (25). We have also shown that the NS3 and NS3A proteins made in mammalian cells can be glycosylated and modified into heterogeneous polylactosaminoglycan-containing proteins (25). Immunoelectron microscopic studies of NS3 proteins expressed by recombinant baculoviruses have confirmed that the proteins are associated with both intracellular smooth-surfaced vesicles and the cell plasma membrane (15). When BTV virus-like particles are expressed by using baculovirus vectors in the presence of the NS3 proteins, the particles are secreted by budding through the cellular membrane but not in the absence of NS3 (15). These findings suggest a possible role for the NS3 proteins in the final stages of BTV morphogenesis and release of virions. Exactly how NS3 is involved or how the NS3 proteins are arranged in association with the cell membrane is not known. The BTV NS3 proteins may have some functional similarity to the rotavirus NS28 glycoprotein, which has been shown to mediate the binding of rotavirus particles to the rough endoplasmic reticulum (ER) and the acquisition of a transient envelope (1, 2).

BTV NS3 proteins have two conserved and putative transmembrane hydrophobic domains (designated HI and HII) located in the carboxy-terminal half of the molecule (25). In addition, there are two potential glycosylation sites, one at NS3 aa 63 and the other at aa 150. The latter glycosylation site is located between the two putative hydrophobic domains. In this study, we have determined the membrane organization of the NS3 proteins and have identified the single site of protein glycosylation. To accomplish this, we prepared 10 S10 deletion mutants, spanning various parts of the two hydrophobic domains and flanking sequences, and other site-specific mutants that modified the two potential glycosylation sites. Each modified S10 gene was used to synthesize mutant NS3 proteins by using the vaccinia virus T7 polymerase transient-expression system. The data obtained indicate that both hydrophobic domains are involved in membrane spanning but that only the aa 150 locus is used for N-linked glycosylation. Based on this, a model for the membrane organization of NS3 has been derived.

MATERIALS AND METHODS

Viruses and cells.

Vero cells were maintained in Eagle’s minimum essential medium (EMEM; GIBCO-BRL) containing 5% (vol/vol) newborn calf serum. HeLa T4⁺ cells were propagated in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum. A recombinant vaccinia virus expressing the T7 RNA polymerase (VVTF-7) was obtained from B. Moss (National Institutes of Health, Bethesda, Md.) and propagated in Vero cells. The infectivity titers of the virus were determined either by 50% tissue culture infective dose measurements or by plaque assays with confluent monolayers of Vero cells.

DNA manipulations and construction of recombinant transfer vectors.

The recombinant plasmid pGEM10.10 (a plasmid containing a BamHI cassette of the S10 gene of BTV serotype 10 [25] in a pGEM backbone) was used to prepare NS3 derivatives with internal deletions (D1, D2, etc.) in the HI or HII site or, by the introduction of a stop codon, removal of 74 aa of the carboxy terminus (CT), giving the derivatives pGEMHID1, pGEMHID2, pGEMHID3, pGEMHID4, pGEMHIID1, pGEMHIID2, pGEMHIID3, pGEMHIID4, pGEMHIID5, and pGEMCT (see Fig. 1). PCRs were used with the parental plasmid and synthetic oligonucleotides to produce the fragments lacking the desired amino acid coding sequences. To make the HIDI deletion, a T7 primer (Table 1, oligonucleotide M) was used as the forward primer and oligonucleotide G was used as the reverse primer to copy the amino-terminal portion of the gene and oligonucleotide H was used as the forward primer and an SP6 primer (oligonucleotide C) was used as the reverse primer to copy the downstream carboxy-terminal portion, with primers G and H providing a unique BglII restriction site (AGATCT) at the 5′ and 3′ ends of the two respective PCR products to cut and ligate the two gene fragments before cloning in pGEM and placing the modified gene under the control of a T7 promoter. By this means, aa 118 to 126 was removed from the S10 open reading frame and replaced by the dipeptide RS (representing the introduced BglII restriction site at the deletion locus). By using a similar strategy, the T7 and oligonucleotide G primers and oligonucleotide I and the SP6 primers were used to replace NS3 aa 118 to 136 with the dipeptide RS to make the HID2 mutant. For the HID3 and HID4 mutants, a different strategy was used; this method took advantage of a unique SpeI site in the NS3 gene (ACTAGT, nucleotides 391 to 396, aa 124 to 126). For the HID3 mutant, oligonucleotides A and C were used to provide a PCR product with an SpeI site preceding the downstream sequence starting at aa 137. For the HID4 mutant, oligonucleotides B and C were used, with oligonucleotide B providing an SpeI site preceding the sequence starting at aa 147. Each PCR product was digested with SpeI and PstI and ligated to pGEM10.10 previously cut with SpeI and PstI (the PstI site is located in the downstream sequence flanking the S10 gene). By this means, aa 127 to 136 was deleted in the HID3 mutant and aa 127 to 146 was deleted in the HID4 mutant and no additional amino acids were added. For the HIID1 mutant, we took advantage of a unique restriction AvaII restriction site in the NS3 gene (GGACC, aa 151 to 152). In this case, oligonucleotide D, having a 5′ AvaII site, and primer C were used to obtain a PCR product representing the NS3 sequence downstream from aa 163 and, after digestion with AvaII and XbaI (a unique site preceding the flanking SP6 promoter), ligated to AvaII- and XbaI-cut pGEM10.10. The recombinant plasmid thus lacked NS3 aa 153 to 162, and no additional amino acids were introduced. Similarly for the HIID2 or HIID3 mutant, oligonucleotide E (or F) with a 5′ AvaII site and oligonucleotide C were used to make PCR products that were digested with AvaII and XbaI and ligated to AvaII- and XbaI-digested pGEM10.10. In HIID2, aa 153 to 172 was deleted from NS3, and in HIID3, aa 153 to 182 was deleted, with no additional amino acids provided. For the HIID4 mutant, oligonucleotides J and M and oligonucleotides K and C were used to make the required upstream and downstream PCR products, respectively, while for the HIID5 mutant, oligonucleotides J and M were used to derive the upstream sequence and oligonucleotides L and C were used to derive the downstream sequence. The presence of BglII sites at the 5′ ends of J, K, and L allowed the respective PCR products to be cut and ligated and, after digestion with BamHI at the two ends of the modified NS3 genes, inserted into pGEM previously cut with BamHI. As before, the deleted amino acids (HIID4, aa 163 to 172; HIID5, aa 163 to 182) were replaced by the dipeptide RS. For pGEMCT, oligonucleotide N was used to introduce a TAG stop codon in lieu of NS3 aa 156 and, in conjunction with oligonucleotide C, to derive a PCR product that could be cut at the 5′ AvaII site and downstream XbaI site and ligated to AvaII- and XbaI-cut pGEM10.10 as described above for the HIID1-3 mutants. The sequences of the final constructs were confirmed by dideoxynucleotide sequencing.

FIG. 1.

Schematic of the NS3 protein and the various mutant constructs. Shown are the locations of the second initiation codon used to synthesize NS3A, the two hydrophobic domains (HI and HII), the two potential glycosylation sites (GSI and GSII), the various deletion mutants that were prepared in the HI (HID1, HID2, etc.), or HII (HIID1, HIID2, etc.) regions of NS3, and the deletion mutant from aa 155 to the carboxy terminus (CT). See the text and Table 2 for details.

TABLE 1.

Oligonucleotide primers used for mutagenesis

Primer	Sequence
A	CCA CTA GTG TGC ACC CTT TCA AGT GAT
B	CCA CTA GTG AAA ATA GGG ACC AAA AC
C	ATT TAG GTG ACA CTA TA
D	CCG GGG ACC CTT AAT CCA ATG CTT GGC GTT GTC AAC TTG GGA GCA
E	CCG GGG ACC GGA GCA ACT TTT TTG ATG ATG GTT TGC GCA AAG ACT
F	CCG GGG ACC AAG AGT GAA AGA GCC TTG AAT CAA CAG ATA GAT ATG
G	CCA GAT CTT CGC TTC TTC TTC AAT TCA
H	CCA GAT CTG CTG CTG TGG TTG CG
I	CCA GAT CTT GCA CCC TTT CAA GTG AT
J	CCA GAT CTG CTT TTA AAC CAT GAA GG
K	CCA GAT CTG GAG CAA CTT TTT TGA TGA
L	CCA GAT CTA AGA GTG AAA GAG CCT TG
M	AAT ACG ACT CAC TAT AG
Na	TGG GAC CAA AAC AGA ATA GCC TTC ATG GTT TAA A

^aThe TAG used in oligonucleotide N to introduce a stop codon for the pGEMCT construct and in lieu of NS3 aa 156 is underlined. 

Oligonucleotide-directed site-specific mutagenesis.

Site-directed mutagenesis was carried out with the MUTA-GENE M13 in vitro mutagenesis kit (Bio-Rad). Two oligonucleotide primers, 5′ GATGTCAAGCACAACTG and 5′ GGCCTTCAAAATAAGTGGGACCA, were synthesized to provide, respectively either an N-to-S mutation at aa 63 (underlined) or an N-to-S mutation at aa 150 (underlined) to disrupt the potential glycosylation sites (GSI and GSII [see Fig. 1]). To make the mutants, the NS3 gene was initially recloned into the BamHI site of M13mp18 and the product was transformed into Escherichia coli MV1190. Single-stranded uracil-containing viral DNA was isolated by growing the M13 phage in E. coli CJ236 (Dut⁺ Ung⁺). Each synthetic oligonucleotide was annealed to the single-stranded template, and DNA was synthesized in the presence of T7 DNA polymerase and T4 DNA ligase (23). The ligated products were transformed into MV1190 cells, and recombinants were checked by dideoxynucleotide sequencing of the derived single-stranded phage DNA. Each potential glycosylation site was mutated individually, leaving the other site intact. The mutant NS3 genes were finally cloned into the BamHI site of a pGEM3Zf(+) vector.

T7 polymerase transient expression.

HeLa T4⁺ and Vero cells were grown to 80 to 90% confluency and infected at a multiplicity of infection of 5 to 10 for 1 h at 37°C with VVTF-7 (10). The infected cells were rinsed three times with prewarmed phosphate-buffered saline (PBS), and 0.5 to 1 ml of Opti-MEM (a reduced-serum medium [GIBCO-BRL]) was added to each infected-cell monolayer. Recombinant plasmid DNA (5 to 20 μg) was mixed gently with an equal volume of Lipofectin (GIBCO-BRL) in a polystyrene tube (8). The DNA-Lipofectin mix was incubated at room temperature for 15 to 20 min before being added to the infected cells. The cells were then incubated at 37°C for the indicated periods before being labeled or subjected to immunofluorescence analysis.

SDS-PAGE analyses.

Protein samples were electrophoresed in 10 or 15% polyacrylamide slab gels (0.75 mm thick) (ATTO Corp.) with an acrylamide/bisacrylamide ratio of 37.5:1 (Bio-Rad). After electrophoresis, the gels were fixed and treated with Enlightning (Du Pont) for 30 min at room temperature with gentle agitation. They were then dried at 60°C in a Savant gel dryer for 1 h, and the distribution of proteins was determined by autoradiography with Kodak X-OMAT XAR-5 film.

Radiolabeling and immunoprecipitation of viral proteins.

At 18 h posttransfection, the cells were washed three times with prewarmed PBS and incubated in 0.5 to 1 ml of MEM (or on occasion DMEM) lacking cysteine and methionine. The intracellular pools of these amino acids were depleted for 1 h at 37°C before labeling. The cells were labeled with 50 to 100 μCi of [³⁵S]methionine per ml in protein-labeling mixture (Du Pont, NEN, Boston, Mass.) for 2 h at 37°C. To label with [³H]mannose, medium was removed 18 h posttransfection. The cells were then incubated in MEM supplemented with 50 μCi of [³H]mannose. The cells were labeled for 1 h and then chased for 30 min in MEM lacking added mannose. In either case, the radiolabeled cells were washed three times with ice-cold PBS and lysed by the addition of chilled cell lysis buffer (50 mM Tris-HCl [pH 7.5], 150 mM NaCl, 1% Triton X-100, 0.1% sodium dodecyl sulfate [SDS], 20 mM EDTA). The lysate was centrifuged at 13,000 × g for 10 min at 4°C to pellet the nuclei and cell debris. The supernatant was recovered and transferred to a fresh tube. A sample of this extract was analyzed by SDS-polyacrylamide gel electrophoresis (PAGE) after the addition of Laemmli gel sample buffer (17).

The radiolabeled proteins were immunoprecipitated by incubating cell lysates with a polyclonal NS3 antiserum (9) for 4 to 8 h at 20°C or overnight at 4°C. To each preparation of antigen-antibody complexes, protein A-Sepharose CL-4B (100 mg/ml; Promega) was added, and the mixture was incubated overnight at 4°C. The immunoprecipitated proteins were pelleted by centrifugation at 13,000 × g for 10 min at 4°C. The precipitate was washed three times with ice-cold cell lysis buffer and resuspended in Laemmli gel sample buffer. The proteins were analyzed on by SDS-PAGE (15% polyacrylamide) as described above.

Endoglycosidase F treatment.

³⁵S-labeled and immunoprecipitated proteins were treated with endoglycosidase F after dissociation of the protein A-Sepharose-bound protein. To accomplish this, the complexes were treated with 1% SDS in 10 mM sodium acetate buffer (pH 5.2) and heated to 95°C for 2 min. After centrifugation, the supernatant was diluted threefold with 0.5% Triton X-100 and incubated overnight with 0.4 U of endoglycosidase F (Boehringer Mannheim) at 37°C (24). The samples were then treated with Laemmli gel sample buffer and analyzed by SDS-PAGE (15% polyacrylamide).

Immunofluorescence.

HeLa T4⁺ cells were seeded onto 22-mm coverslips. They were infected and transfected as described above. After a further 16 h of incubation at 37°C, the cells were washed three times with cold PBS and fixed for 20 s with methanol-acetone (50:50, vol/vol). The methanol-acetone was removed, and 2 ml of PBS was added to prevent the cells from drying. The cells were then treated with mouse anti-NS3 antibodies for 45 min at 37°C. They were washed three times with PBS, and fluorescein isothiocyanate (FITC)-conjugated mouse immunoglobulin G was added. Incubation with the secondary antibody was carried out for 45 min at 37°C. The cells were washed three more times with PBS, and tetramethylrhodamine-5-isothiocyanate (TRITC)-conjugated wheat germ agglutinin was added. The incubation was carried out for 45 min at 37°C (3). Finally, the cells were washed with PBS, mounted in glycerol-PBS (90:10, vol/vol), and examined under confocal microscope.

RESULTS

Mutagenesis of the two hydrophobic domains of NS3 and expression of the mutant protein forms by using the T7 transient-expression system.

Earlier studies demonstrated that S10 gene products of BTV-10 are found in the endoplasmic reticulum, the Golgi complex, and the cell surface plasma membrane (25). The data suggested that the S10 gene products function as integral membrane proteins. Depending on their arrangement, integral membrane proteins span the membrane one or more times with the amino and carboxy termini of the polypeptide chain on the same or a different side of the lipid bilayer depending on the configuration. Membrane-spanning sequences generally consist of some 20 (or more) hydrophobic and uncharged amino acids. Analysis of the 229-aa S10 gene product indicates that the NS3 proteins contain two regions of hydrophobicity. Domain I (HI) is located at aa 118-141, and domain II (HII) is located at aa 162 to 182. Preceding HI, the sequence of amino acids is KLKSDLSELKKKR (aa 105 to 117); following HI, the sequence is DMSVAFKINGT (aa 142 to 152, with the putative glycosylation signal underlined). The sequence that precedes HII is KINGTKTEVPSWFK (aa 148 to 161, with the putative GSII signal similarly underlined), and the sequence that follows HII is KSERALNQQ (aa 183 to 191) (see Table 2).

TABLE 2.

NS3 mutants used in this study

NS3 mutant	Position of mutated amino acids	Amino acid sequence deletedb	Glycosylation	Transport to Golgi
NS3			+	+
pGEMHID1a	Δ118–126	AIIHTTLLV	−	+
pGEMHID2a	Δ118–136	AIIHTTLLVA AVVALLTSV	−	+
pGEMHID3	Δ127–136	AAVVALLTSV	−	+
pGEMHID4	Δ127–147	AAVVALLTSV CTLSSDMSAFV	−	+
pGEMHIID1	Δ153–162	KTEVPSWFKS	±	+
pGEMHIID2	Δ153–172	KTEVPSWFKS LNPMLGVVNL	±	+
pGEMHIID3	Δ153–182	KTEVPSWFKS LNPMLGVVNLGATFLMMVCA	+	+
pGEMHIID4a	Δ163–172	LNPMLGVVNL	+	+
pGEMHIID5a	Δ163–182	LNPMLGVVNL GATFLMMVCA	+	+
pGEMCT	Δ156–229	VPSWFKSLN PMLGVVNLGA TFLMMVCAKS ERALNQQIDM IKKEVMKKQS YNDAVRMSFT EFSSIPLDGF EM PLT	+	+
GSIc	63		+	+
GSIIc	150		−	+

^aIn these constructs, the deleted amino acids were replaced by the dipeptide RS. 

^bThe sequences that contribute to the two hydrophobic domains are underlined. 

^cIn these mutants, there is an N-to-S change at the indicated positions. 

To determine which of the hydrophobic domains are membrane-spanning sequences, or whether both are, a series of mutant NS3 genes were prepared. As illustrated in Fig. 1 and described in Materials and Methods, deletions of 9 to 30 aa were introduced on either the HI or the HII hydrophobic domain and some of the flanking sequences and from aa 156 to the carboxy terminus to generate the panel of NS3 mutants listed in Table 2. For domain locus I, four deletion mutants lacking either aa 118 to 126 (HID1), aa 118 to 136 (HID2), aa 127 to 136 (HID3), or aa 127 to 147 (HID4) were made. For domain locus II, five mutants lacking either aa 153 to 162 (HIID1), aa 153 to 172 (HIID2), aa 153 to 182 (HIID3), aa 163 to 172 (HIID4), or aa 163 to 182 (HIID5) were made. The carboxy-terminal mutant was constructed with a stop codon introduced after aa 155 so that the product of the open reading frame lacks 74 residues of the carboxy end (CT, aa 156 to 229), including the whole of the HII domain. For none of the mutants was the putative glycosylation signal at aa 150 to 152 (NGT, GSII) removed or modified, nor was the GSI glycosylation signal at aa 63 to 65 (NTT) changed. The mutant genes were cloned into pGEM plasmids and expressed in mammalian cells (HeLa T4⁺) with a VV T7 polymerase transient-expression system. This system involves the infection of cells with a recombinant VV (VVTF-7) that expresses the bacteriophage T7 RNA polymerase, followed by transfection with a plasmid containing a foreign gene under the control of a T7 promoter. Transcription of the foreign gene is mediated by the VV-encoded T7 RNA polymerase expressed in the cytoplasm of the virus-infected cells.

The expression of the mutant proteins was analyzed by immunoprecipitation of ³⁵S-labeled infected-cell proteins with monospecific polyclonal NS3 antibodies (9). Mock-transfected cells gave no identifiable labeled products after a similar immunoprecipitation (Fig. 2). All HID mutants with mutations in the region 118 to 147 (HID1, HID2, HID3, and HID4 respectively) and HIID mutants with mutations in the region 153 to 182 (HIID1, HIID2, HIID3, HIID4, and HIID5) were analyzed. Figure 2 shows the protein profiles obtained for all the HID and HIID mutants by comparison with the control unmodified NS3 proteins. As indicated, the latter included the major NS3 protein, the minor NS3A species, and a glycosylated form of NS3 designated GNS3 (Fig. 2). Whether a glycosylated form of NS3A that comigrated with NS3 was present is not known. Between experiments, NS3A was not always observed, possibly due to variations in the fidelity of translation from the first AUG codon in the S10 mRNA. The HID mutants HID1 and HID2 (Fig. 2) gave major protein bands slightly smaller than the native NS3, while very few, if any, changes in mobility were detected for HID3 and HID4. No bands equivalent to the glycosylated forms of the proteins were evident, indicating that glycosylation was inhibited when the HI domain was modified in the manner used. This was further characterized, as discussed below. Interestingly, and although different experiments and exposures were used in the analyses, the migration patterns of various HID mutant NS3 and NS3A protein bands, although indicating a reduction in the size of the species for some mutants, were not reduced as much as expected (see Fig. 2 for a comparison), implying that the truncations of the HID hydrophobic domain may have influenced the amount of SDS binding and hence the mobilities of the mutant proteins.

FIG. 2.

Expression of native NS3, NS3 HID, and NS3 HIID mutant proteins. HeLa T4⁺ cells were infected with VVTF-7 at a multiplicity of 10, transfected in the presence of Lipofectin with either a plasmid containing the wild-type NS3 gene, i.e., pGEM10.10, or with the indicated HID mutant NS3 plasmids (Table 2), and labeled with [³⁵S]methionine as described in Materials and Methods. Infected cells were lysed and immunoprecipitated with an NS3-monospecific polyclonal antiserum, and the proteins were analyzed by SDS-PAGE (15% polyacrylamide) and autoradiography. For the unmodified gene, the positions of the glycosylated GNS3 and nonglycosylated NS3 and NS3A bands are indicated. The mutant genes gave bands corresponding to the expected NS3 and NS3A deletion products.

Deletion mutants flanking or spanning all or parts of the second hydrophobic domain (HIID) also expressed mutant NS3 proteins. The changes in the NS3 sizes were generally greater for the HIID mutants than for the HID mutants. For example, the 20-aa HIID2 (Fig. 2) deletion mutant (aa 153 to 172) migrated significantly faster than the 21-aa HID4 deletion mutant (aa 127 to 147). For the HIID1 and HIID2 mutants, few, if any, of the mutant glycosylated NS3 species (Fig. 2) were evident. However, for the HIID3, HIID4, and HIID5 mutants (Fig. 2), major bands that migrated slower than the mutant NS3 bands were observed, suggesting that glycosylated forms of the protein were present in these mutants.

To analyze the expression of the mutant proteins at different times postinfection, the NS3 products were analyzed at 16, 20, and 24 h posttransfection. No significant difference was observed among the three time points (data not shown).

To confirm which mutants made glycosylated products, mutant proteins labeled with [³⁵S]methionine were recovered by immunoprecipitation as described in Materials and Methods and treated with endoglycosidase F to remove the glycans from the proteins. Endoglycosidase F-treated and untreated NS3 and mutant proteins were subsequently resolved by SDS-PAGE. The results are shown in Fig. 3. For the unmodified gene NS3, treatment with endoglycosidase F significantly reduced the GNS3 band, thereby confirming that GNS3 is a glycoprotein (Fig. 3B), while the NS3 band became more prominent (in this experiment, NS3A was not identified). For the HID mutants, no changes in the mobility patterns of the treated and untreated proteins were apparent (Fig. 3A), agreeing with the above observations that the NS3 proteins were not glycosylated in these mutants. Likewise, little difference was observed for the HIID1 and HIID2 mutants (Fig. 3B), indicating that the presence of carbohydrate in these NS3 products was minimal. However, for the HIID3, HIID4, HIID5 (Fig. 3B), and CT (data not shown for CT) mutants, treatment with endoglycosidase F led to a reduction of the amounts of the glycosylated forms and a major increase in the amounts of the nonglycosylated forms.

FIG. 3.

Endoglycosidase F treatment of immunoprecipitated native NS3 and mutant NS3 proteins. NS3 and the mutant proteins were labeled with [³⁵S]methionine and immunoprecipitated. The immunoprecipitated proteins were treated (lanes +) or not treated (lanes −) with endoglycosidase F as described in Materials and Methods. (A) The HID mutants did not show any detectable differences between the profiles of treated or untreated proteins. (B) The HIID mutants HIID1 and HIID2 also did not show any detectable difference, but significant differences in the profiles were observed between the treated and untreated samples for both the native and HIID3 to HIID5 mutant proteins.

Further, in a separate experiment, in vivo labeling of the mutant proteins with [³H]mannose showed that in contrast to strong radiolabeling of the products of HIID3, HIID4, HIID5, and the CT mutants, HIID1 and HIID2 gave very low levels of incorporation and none of the HID mutants incorporated radioactivity, as anticipated (data not shown). Therefore, it was concluded that sugar was added to only a minority of the NS3 expression products.

Processing of NS3 mutant proteins with modified hydrophobic domains.

The intracellular location of the native and mutant NS3 proteins was analyzed by indirect immunofluorescence. We have shown previously (25) that the NS3 gene products use a cycling pathway in the ER and Golgi complex in mammalian cells. In view of this, we investigated the localization of the NS3 protein in infected and transfected cells by using a double-labeling procedure to allow the location of the proteins to be identified (see Materials and Methods). Double-labeling involved treating cells with the FITC-tagged monospecific NS3 antibody followed by labeling with TRITC coupled to wheat germ agglutinin (a lectin that binds to glycoconjugates consisting of N-acetylglucosamino and sialic acids and which is routinely used as a marker for the Golgi region).

In preliminary single-staining studies, i.e., with cells treated with only FITC conjugated to NS3 antiserum no difference in labeling was observed between the early and late stages of infection (data not shown). However, by using the double-labeling method, together with confocal microscopy, which allowed the Golgi complex to be identified concomitantly, it was found that if the cells were processed late postinfection and posttransfection, the Golgi complex was distorted, affecting its labeling. Therefore, to localize the NS3 proteins specifically in the Golgi complex, double staining was carried out only at early stages posttransfection.

Analyses of the intracellular distributions of all four HID mutants and all five HIID mutants by confocal microscopy indicated that in each case the mutant proteins, like the native NS3 protein (Fig. 4a), were transported into the Golgi complex (Fig. 4c and d), indicating that the mutant proteins were processed via the ER to the Golgi complex whether or not the HI or HII domains were partially or totally deleted. The typical staining patterns obtained for these mutants are shown for HDI3 and HDII2 in Fig. 4c and d, respectively, being as representative of HDI and HDII mutants. The CT mutant also behaved in a similar manner, as shown in Fig. 4e.

FIG. 4.

Cellular localization of the NS3 mutant proteins. Confocal microscopy was performed on permeabilized HeLa T4⁺ cells that were infected with VVTF-7 and then transfected with either a plasmid containing the wild-type NS3 gene (pGEM10.10) (a) or with the mutant NS3 plasmids (HID3 [c], hIID2 [d], and CT [e]). All cells were incubated with a mouse anti-NS3 antibody and then labeled with FITC-conjugated mouse IgG (green) and subsequently with Golgi-specific TRITC-conjugated wheat germ agglutinin (red). All four panels show both types of staining, indicating the localization of NS3 within the Golgi complex. Note the green staining around the perimeter of the cell in panel a, indicating the surface staining. The cells were also infected with VVTF-7 (b and f) and transfected with pGEM10.10 similarly to those in panel a but either treated with the FITC conjugates (b) or treated only with TRITC conjugates to show the Golgi staining (f).

Identification of the carbohydrate addition site(s) in NS3 proteins by site-specific mutagenesis.

To identify the carbohydrate addition site(s) of NS3, the N residues at aa 63 and 150 were individually mutated to S residues to provide the NS3 mutants GSI and GSII, respectively (see Materials and Methods). Recombinant vv-infected HeLa T4⁺ cells were transfected with each of the mutant plasmids, and the NS3 proteins were immunoprecipitated using the available monospecific NS3 antiserum. As shown in Fig. 5A, bands corresponding to GNS3 were identified for the [³⁵S]methionine-labeled native NS3 gene and the GSI but not the GSII mutant genes. When the proteins were treated with endoglycosidase F (Fig. 5B), the GNS3 band was reduced in the GSI mutant but the pattern of [³⁵S]methionine labeling for the GSII mutant was unchanged. From these results, it was concluded that only aa 150 was used for the addition of sugars to the protein.

FIG. 5.

Analyses of GSI and GSII mutant NS3 proteins. (A) Results of immunoprecipitation of [³⁵S]methionine-labeled native NS3, GSI mutant, and GSII mutant NS3 proteins. (B) Results of untreated (lanes −) and endoglycosidase F-treated (lanes +) preparations of immunoprecipitated [³⁵S]methionine-labeled native NS3, GSI, and GSII proteins.

Localization of the GS mutant proteins in cells and on the cell surface.

The S10 gene products have been previously reported to be associated with smooth-surfaced vesicles and the plasma membrane (15). To determine whether the GSI and GSII mutations affected the display of NS3 molecules on the cell surface and to investigate whether mutation at either of the potential glycosylation sites affected the location of NS3, the surfaces of cells were examined along with the cellular localization of the expressed protein. The transient expression of the GSI and GSII mutant proteins was analyzed as described above by labeling mammalian cells previously infected with VVTF-7 and transfected with the respective plasmid DNAs. For both the GSI and GSII mutants, the proteins were identified in the Golgi complex of the cells, suggesting that both mutant proteins were processed through the ER and the Golgi apparatus (Fig. 6).

FIG. 6.

Confocal microscopy of GSI and GSII mutant transfected cells showing the surface and intracellular localization of the NS3 proteins. The mutant proteins are localized in the Golgi complex in both GSI and GSII. However, GSII (b) failed to reach the surface, in contrast to GSI (a), which it was transported to the cell surface.

Interestingly, and as shown in Fig. 6, the results obtained indicated that the accumulation of the GSII mutant protein was abolished, since the surface-labeled cells were comparable to the nontransfected control. By contrast, the GSI mutant products were identified on the cell surface, as for the native NS3 protein (Fig. 4). A summary of the results concerning the patterns of glycosylation and association with the Golgi complex for NS3 and the various mutants that were analyzed is presented in Table 2.

DISCUSSION

Orbiviruses are nonenveloped viruses that in their mature forms lack integral membrane glycoproteins, although it has been reported that VP5, one of the outer proteins of the BTV virion, may be glycosylated in certain environments (26); however, this has not been confirmed in other studies (unpublished data). We have demonstrated previously (25) that transiently expressed BTV NS3 and NS3A proteins exist in both unmodified and glycan-modified forms in mammalian cells and that they are involved in the release of virus-like particles when coexpressed with BTV VP2, VP3, VP5, and VP7 in insect cells by using baculovirus vectors (15).

One of the questions that has been addressed in this study is the presence and site of NS3 glycosylation. Many integral membrane proteins become modified by N-linked glycosylation on regions of the proteins that have the necessary recognition sites (N.X.S/T) exposed on the luminal side of the ER membrane. In general, these are then processed further and transported inside the infected cells in a sequential manner, with the Golgi complex playing an important role in the intracellular sorting and trafficking of glycoproteins (5, 6, 19–21). The mature glycoproteins may contain low-mannose chains derived from the sugar residues added in the ER or complex chains resulting from the trimming and addition of other ligands in the Golgi complex (16).

In previous studies, we found an ER-Golgi cycling pathway of transiently expressed NS3 protein in HeLa T4⁺ and Vero cells, indicating that NS3 is an intregal membrane protein, and showed that part of the NS3 population can be N-linked glycosylated (25). Based on the S10 gene sequence, there are two potential sites for the attachment of such glycans, one at aa 63 and the other at aa 150. However, resolution of the expressed products by SDS-PAGE suggested that only one site was used, based on the approximately 2-kDa larger glycosylated species (GNS3) compared with the nonglycosylated form (NS3). This conclusion is tentative, since NS3A is also encoded by the S10 gene and NS3A is some 2 kDa smaller than NS3. Thus, if only NS3A were glycosylated, two glycans could be accommodated. In the present study, the use of radioactive mannose or methionine precursors and treatment of the products with endoglycosidase F clearly showed that for the native protein only one glycan is present (GNS3) in the glycosylated form of NS3. Variable amounts of NS3A were detected between experiments, making it difficult to detect the glycosylated form of NS3A. However, it could be investigated by using an S10 gene from which the first AUG codon is deleted.

Analysis of the GSI and GSII mutants has shown that only the aa 150 site is glycosylated. Thus, when the GSI site was modified, no change in the mobility of GNS3 was detected. Further, when the GSII site was modified, no carbohydrate label was incorporated. Also, treatment of the methionine-labeled GSI and GSII mutants with endoglycosidase F reduced the amounts of GNS3 in the GSI mutant but had no effect on the GSII mutant. Thus, the data support the view that NS3 is glycosylated only at aa 150 and not at aa 63.

The presence of carbohydrate on some of the NS3 products raises the question of its function. In the present investigation, indirect immunofluorescence and surface-labeling studies of the glycosylation mutants have suggested that deletion of the GSII site (aa 150) resulted in partial processing of the NS3 protein, since although the protein could be detected in association with the Golgi complex, none could be detected in association with the cell surface even though both the GSI mutant and native NS3 were readily detected. The absence of the GSII mutant protein from the cell surface may be a result of a failure of transport of the protein due to the lack of proper folding that requires glycosylation and the presence of mature forms of the protein (polylactosaminoglycans [7, 18]). For some proteins, correct folding of the newly synthesized chain has been found to be necessary for transport from the ER or Golgi apparatus to the cell surface (4, 11, 13, 14, 16). Alternatively, the absence of the GSII mutant from the cell surface may be due to the proteolysis caused by the absence of protective carbohydrate.

The presence of two potential membrane-spanning sequences (HI and HII), with the glycosylation site at aa 150 between them, raises the questions whether both hydrophobic domains are involved in membrane-spanning membranes and how they are arranged. Depending on the organization, the amino- and carboxy-terminal sequences of NS3 could be located on the same or different sides of the membrane. Deletion of 9 to 21 aa in HI of NS3 (and adjacent sequences) resulted in the complete loss of glycosylation of the protein products. Nevertheless, all the mutant products were processed through the Golgi complex. Since glycosylation at GSII was abolished, the conclusion is that the HI domain, when it is intact, is responsible for introducing the protein into the ER to allow subsequent glycosylation.

The deletions encompassing HII, on the other hand, showed some interesting features. For the CT deletion, in which all of the HII region was deleted together with the remaining carboxy-proximal and end sequence, the protein was recovered in the ER and Golgi complex and was glycosylated. This result agrees with the results of the HI deletion and confirms that the HI region is involved in spanning the membrane, allowing glycosylation at the GSII site. The deletion of aa 153 to 182 (HIID3), 163 to 172 (HIID4), or 163 to 182 (HIID5) gave the same phenotypes as the CT deletion. For HIID5, just the HII region was deleted (i.e., in this respect like the CT deletion), while for HIID4, only the leading half of the first 20 residues of HII was deleted with, in both cases, replacement of the hydrophobic amino acids with the dipeptide RS (derived from the manipulations). For the HIID3 mutant, all the HII region was deleted as well as 9 aa preceding the HII region. However, a different phenotype was obtained with the HIID1 and HIID2 mutants. In both cases, the proteins were processed and associated with the ER and Golgi complex; however, the results indicated that only a small portion of the products were glycosylated. For HIID1, only the 9 aa preceding the HII domain (aa 153-161) and the following serine were deleted, whereas for the HIID2 mutant, the 9 aa preceding HII as well as the initial 10 aa of HII were deleted (i.e., similar to the HIID3 mutant, except that for this mutant all the HII region was deleted). The most likely explanation for the low level of glycosylation is that removal of aa 153 to 161 leads to the glycosylation signal (aa 150 to 152) becoming the leading edge of the HII domain, compromising its ability to be glycosylated. Another possible interpretation is that glycosylated forms at the GSII site of these mutants are not stable and are degraded. The fact that the GSII mutant proteins are not detected on the surface of cells whereas the native or GSI mutant proteins are detected suggests an important role for the glycosylated forms of the protein, with the glycans most probably serving to protect the protein from degradation either in the route to incorporation onto the cell surface or thereafter.

Based on the results obtained and our previous studies NS3,like rotavirus NSP4, has two hydrophobic domains, and it seems likely that both the amino- and carboxy-terminal ends of the protein are cytoplasmic. However, further work is required to confirm this organization and to determine which, if any, of the HI, HII, or other mutants allow the release of BTV virus-like particles from cells when coexpressed with the BTV VP2, VP3, VP5, and VP7 proteins.