Virus Glycoprotein. This mutant also formed heterotrimers with a hybrid G. fusion have been identified in several viral fusion proteins

Transcription

1 JOURNAL OF VIROLOGY, OCt. 1990, p X/90/ $02.00/0 Copyright C) 1990, American Society for Microbiology Vol. 64, No. 10 A Fusion-Defective Mutant of the Vesicular Stomatitis Virus Glycoprotein MICHAEL A. WHITT,'* PANAYIOTIS ZAGOURAS,1 BRUCE CRISE,' AND JOHN K. ROSE"2 Departments of Pathology' and Cell Biology,2 Yale University School of Medicine, New Haven, Connecticut Received 10 May 1990/Accepted 30 June 1990 We have recently described an assay in which a temperature-sensitive mutant of vesicular stomatitis virus (VSV; mutant ts045), encoding a glycoprotein that is not transported to the cell surface, can be rescued by expression of wild-type VSV glycoproteins from cdna (M. Whitt, L. Chong, and J. Rose, J. Virol. 63: , 1989). Here we examined the ability of mutant G proteins to rescue ts045. We found that one mutant protein (QN-1) having an additional N-linked oligosaccharide at amino acid 117 in the extracellular domain was incorporated into VSV virions but that the virions containing this glycoprotein were not infectious. Further analysis showed that virus particles containing the mutant protein would bind to cells and were endocytosed with kinetics identical to those of virions rescued with wild-type G protein. We also found that QN-1 lacked the normal membrane fusion activity characteristic of wild-type G protein. The absence of fusion activity appears to explain lack of particle infectivity. The proximity of the new glycosylation site to a sequence of 19 uncharged amino acids (residues 118 to 136) that is conserved in the glycoproteins of the two VSV serotypes suggests that this region may be involved in membrane fusion. The mutant glycoprotein also interferes strongly with rescue of virus by wild-type G protein. The strong interference may result from formation of heterotrimers that lack fusion activity. Vesicular stomatitis virus (VSV) is the prototype rhabdovirus and, like other members of this group, contains a single membrane-spanning spike glycoprotein (G) in its membrane envelope. Infection by many enveloped viruses such as VSV involves binding of the virus to a receptor on the cell surface, internalization by endocytosis, and fusion of the viral envelope with the endosomal membrane (13, 18). For VSV, the membrane fusion event is catalyzed by G protein (7, 22) and occurs in the endosome, where G protein undergoes a conformational change induced by low ph. Viral and endosomal membrane fusion releases the viral nucleocapsid into the cytoplasm and is thus required for infection. Although the amino acid sequences involved in receptor binding and membrane fusion have been determined for several enveloped viruses (see reference 28 for a review), these sequences have not been identified for VSV or other rhabdoviruses. Analysis of mutations in G protein affecting binding, fusion, or release of nucleocapsids should be useful in understanding the early events of infection. We have recently described a system for identifying mutations affecting VSV G protein function. G protein expressed from cloned DNA is used to rescue a temperaturesensitive mutant of VSV (tso45) which makes a defective G protein that is not transported to the plasma membrane at the nonpermissive temperature (31). This system provides a sensitive assay for examining the ability of mutant glycoproteins to be incorporated into particles or to function in the subsequent steps of virus binding and entry. Using this rescue assay, we were able to demonstrate a requirement for at least part of the normal cytoplasmic tail of VSV G protein in its assembly into virus particles (31). By using this system to analyze a variety of VSV G proteins carrying mutations in the extracellular domain, it should be possible to localize domains of the glycoprotein involved in later steps of virus infectivity. * Corresponding author We previously described the generation of point mutations that introduced new sites for asparagine-linked (N-linked) glycosylation in the VSV G protein. Although one of these proteins was misfolded and retained in the endoplasmic reticulum, two others appeared to fold correctly and were transported to the plasma membrane (16, 17). Here we show that one of the mutants (QN-1) that is transported normally was incorporated into virus particles but that it did not rescue infectivity, apparently because it lacked fusion activity. This mutant also formed heterotrimers with a hybrid G protein that has the wild-type G protein ectodomain and interfered with rescue by wild-type VSV G protein. Although hydrophobic domains involved in membrane fusion have been identified in several viral fusion proteins (11, 28, 32), such a domain has not been identified in any rhabdovirus glycoprotein. Because the extra oligosaccharide at amino acid 117 in QN-1 immediately precedes a region consisting largely of uncharged amino acids (residues 118 to 136), we suggest that this domain may be directly involved in the membrane fusion activity of VSV G protein. MATERIALS AND METHODS Viruses, cell culture, plasmids, and rescue assay. Stocks of tso45 and the recombinant vaccinia virus vtf7-3 were prepared as described previously (31). All cdnas encoding G protein glycosylation mutants (see Fig. 1) were excised with BamHI from pjc119-based plasmids (16) and inserted into the unique BamHI site of pet-3 (25). Rescue of tso45 was performed as described previously (31) except that all plasmid transfections were mediated by a modification of the lipofection procedure (6), using dimethyldioctadecyl ammonium bromide (Sigma Chemical Co.) as the cationic lipid (J. Rose, L. Buonocore, and M. Whitt, submitted for publication). Virus binding assays. The procedure used to determine the amount of virus binding was a modification of that used by Matlin et al. (19). Virus particles radioactively labeled with [35S]methionine were recovered from the medium by cen-

2 4908 WHITT ET AL. trifugation after rescue of tso45 with the indicated G protein. A sample of the medium was titered by plaque assay on mouse L cells before centrifugation to determine the amount of infectious virus present. Virions were sedimented through a 20% sucrose cushion at 115,000 x g for 1 h and suspended in 100,ul of 10 mm Tris hydrochloride (ph 7.4), and the amount of radioactivity in the virus fraction was determined by scintillation counting. Confluent monolayers of mouse L cells were washed twice with ice-cold sodium bicarbonatefree Dulbecco modified Eagle medium (DMEM) containing 0.1% bovine serum albumin (BSA) and buffered to ph 6.4 to 6.6 with 2 mm 2-(N-morpholino)ethanesulfonic acid (MES) and 2 mm N-2-hydroxyethylpiperazine-N'-2-ethanesulfonic acid (HEPES) (binding medium) and were then incubated on ice for an additional 10 min in binding medium. The monolayers were then maintained at 4 C during the binding assay. Approximately 3 x 104 cpm of [35S]methionine-labeled virus was adsorbed to the cells for 1.5 h in 200,ul of binding medium, and the plates were rocked every 15 min. All assays were performed in duplicate. After binding, the inoculum was removed and the monolayers were washed four times with ice-cold binding medium. The cells were then solibilized in phosphate-buffer saline (PBS) containing 1% Nonidet P-40 (Sigma) and scraped from the dish with a rubber policeman. Duplicate samples were suspended in Ecoscint scintillation fluid (National Diagnostics), and the amount of labeled virus bound was determined. To determine the amount of virus internalized, cells were removed from the plate with PBS containing 50 mm EDTA and then pelleted at 4,000 rpm for 1 min in a microcentrifuge. The cell pellet was resuspended in ice-cold binding medium (ph 7.4) without BSA but containing proteinase K (0.5 mg/ml; Sigma) and incubated on ice for 45 min to remove bound virus remaining on the cell surface. The protease-treated cells were then washed three times with ice-cold binding medium (ph 7.4) containing 0.1% BSA. Cell-associated radioactivity was determined after solubilizing as described above. Metabolic labeling, immunoprecipitation, and flow cytometry. Baby hamster kidney (BHK-21) cells were infected with vtf7-3 and transfected with the appropriate plasmid(s). Cells transiently expressing G proteins were radioactively labeled with [35S]methionine in serum-free, methionine-free DMEM at 3 to 4 h posttransfection, using Trans[35S]abel (ICN Biomedicals, Inc.). Immunoprecipitation of single G proteins was performed as described previously (23). For coimmunoprecipitation of G protein heterotrimers, cells were solubilized in MNT buffer (ph 5.8) containing 1% Triton X-100 (4), nuclei and cytoskeleton were removed by centrifugation at 14,000 rpm for 2 min in a microcentifuge (Eppendorf), and the cell extracts were precleared by incubation with an irrelevant antiserum for 6 h on ice, followed by incubation with Pansorbin (Calbiochem, Inc.) for an additional 6 h on ice. The supernatants were then divided into three portions and incubated with the appropriate antisera for 6 h on ice. Immune complexes were recovered with Pansorbin and then washed with MNT buffer (ph 5.8) containing 1% Triton X-100. Labeled proteins were resolved on 10% polyacrylamide gels containing sodium dodecyl sulfate (SDS) (14) and visualized by fluorography (1). Quantitation of immunoprecipitated proteins was determined by scanning densitometry as described previously (15). Flow cytometry was performed as described previously (31) except that after fixation, the cells were washed once with PBS containing 10 mm glycine (PBS-glycine) and then incubated in PBS-glycine for 30 min to block any residual fixative. The fixed cells were then incubated in PBS-glycine containing wt G HS KT QN-1 H--wS (49) (25) (117) 179 qj I?~~~~~~~~I TM 336 Domain FIG. 1. Schematic representation of wild-type (wt) G protein and G-protein glycosylation mutants. The three glycosylation mutants HS, KT, and QN-1 were generated previously (16). The locations of N-linked oligosaccharides are indicated (i'), and the amino acid residues to which the additional glycans are attached are given in parenthenses. Amino acid numbering corresponds to the protein before signal sequence cleavage (24). The amino acid residues that are glycosylated in the wild-type G protein and all of the mutants are indicated above the wild-type sequence. The HS mutant contains a consensus site for N-linked sugar addition which is not glycosylated. M, transmembrane (TM) domain. 0.2% BSA for an additional 30 min. All subsequent antibody incubations and washes were performed in PBS-glycine- BSA. RESULTS The wild-type VSV G protein contains two N-linked oligosaccharides at amino acids 179 and 336. Three mutant VSV G proteins were made in our laboratory previously with single amino acid changes introduced to generate additional sites for oligosaccharide attachment at amino acids 25, 49, and 117 (KT, HS, and QN-1, respectively) (16). The positions of the mutations are shown in Fig. 1. Two of the additional sites were glycosylated (KT and QN-1); however, the site designated HS was not glycosylated although a consensus site for glycosylation was introduced (16). To examine the effects of these mutations on the abilities of the altered G proteins to assemble into and function in virus particles, we used a system in which transient expression of wild-type G protein from cloned cdna results in the rescue of the temperature-sensitive mutant of VSV (tso45) at the nonpermissive temperature (31). Cells were infected with tso45, and the indicated G protein was expressed by using the transient system of Fuerst et al. (8) at 40.5 C. The rescue of tso45 by each of the G-protein mutants was then determined by titration at 34 C of virus produced. We found that neither KT nor QN-1 rescued tso45 and that the titer obtained after rescue with HS was significantly lower than that obtained with wild-type G protein (Table 1). To determine whether the mutant proteins were expressed on the cell surface at the high temperature required for the rescue assay (40.5 C), we examined the expression of each of the mutant G proteins at both 37 and 40.5 C by flow cytometry. At 37 C, the relative surface expression of the three mutant G proteins was equivalent to that of wild-type G protein. At 40.5 C, QN-1 showed the same level of surface expression as did wild-type G protein; however, cells ex- f J. VIROL.

3 VOL. 64, 1990 GLYCOSYLATION MUTANT DEFECTIVE IN FUSION ACTIVITY 4909 TABLE 1. Surface expression and rescue with G protein glycosylation mutants Relative surface DNA expressiona Range of titers (PFU/ml)b 37 C 400C Wild-type GI x X 105 HS x x 103 KT x 101 QN x 101 a Determined by flow cytometry for cells expressing G proteins at the indicated temperatures in the absence of ts045. Values represent the amount of surface G protein in each culture relative to the amount wild-type G protein. b Values from at least three separate rescue experiments with each G protein. pressing KT showed only 5% and HS showed approximately 50% of the surface expression levels of wild-type G protein at the higher temperature (Table 1). Pulse-chase experiments showed that KT was degraded at 40.50C with a half-life of approximately 2 h, whereas HS had a half-life of approximately 3.5 to 4 h. In contrast, both QN-1 and wild-type G protein were relatively stable at 40.5 C, with half-lives of greater than 5 h (data not shown). The reduction in rescue efficiency observed with HS could therefore be explained by a shorter half-life resulting in a lower level of surface expression than for wild-type G protein. Similarly, the rapid degradation of KT and low levels of surface expression could explain the inability of KT to rescue tso45. These results also suggested that the single point mutation in HS and the mutation or the extra oligosaccharide on KT interfered with protein folding at 40.5 C. In contrast, the additional carbohydrate on QN-1 had no discernible effect on protein stability or transport at the elevated temperature. QN-1 is efficiently incorporated into tso45 virions. Although expression of QN-1 did not result in the production of infectious particles in the rescue assay, the protein was present on the cell surface at a level comparable to that of wild-type G protein. This result suggested that the extra oligosaccharide was either blocking incorporation of QN-1 into particles or affecting some early step in infection. To determine whether QN-1 was incorporated into virions, we examined the protein composition of [35S]methionine-labeled particles recovered from the medium of tso45-infected cells expressing QN-1. Radioactively labeled virus particles were collected by centrifugation and purified by sedimentation in a sucrose gradient. Fractions were immunoprecipitated and then analyzed by SDS-polyacrylamide gel electrophoresis. A peak of QN-1 was found in fractions that contained intact virions (31), indicating that QN-1 was incorporated into VSV particles (Fig. 2). Quantitation of QN-1 present in fractions 6 and 7, relative to the amount of N or M in the same fractions, showed that the amount of QN-1 incorporated was similar to that found for particles rescued with wild-type G protein (31), indicating that the lack of infectivity observed after rescue of tso45 with QN-1 was not due to the inability of QN-1 to assemble into virus particles. Virus particles containing QN-1 bind to cells and are endocytosed. To determine whether the failure of particles M P MM - containing QN-1 to infect cells was due to a defect in the binding of these particles to cells, we compared the binding properties of particles containing either QN-1 or wild-type G protein. Rescued particles labeled with [35S]methionine were concentrated from the medium by centrifugation and ado 0 co I - _ MW _ - a 0 _ -0- QN F raci,;or Numt(x,i FIG. 2. Analysis of virus particles rescued with QN-1 by sucrose gradient centrifugation. Rescue and labeling conditions have been described elsewhere (31). Gradient fractions were immunoprecipitated with anti-vsv antiserum and electrophoresed on a 10% SDS-polyacrylamide gel. Fractions from the bottom (fraction 1) to the top (fraction 12) are shown. Wild-type VSV marker proteins (lane M) and a sample of the material loaded onto the gradient, after immunoprecipitation with anti-vsv antiserum (lane P), are also shown. Note the slower mobility of QN-1 (fraction 6) compared with the wild-type G protein marker. sorbed to confluent monolayers of L cells at 4 C. After adsorption, the cells were washed extensively, the monolayers were solubilized, and cell-associated radioactivity was determined. Particles rescued with wild-type G protein or QN-1 bound equally well to the monolayers; however, only 10 to 17% of the input radioactivity remained cell associated after washing. A similar low level of virus binding has been shown previously for wild-type VSV virions (19). At least 40% of the binding observed with particles containing QN-1 appeared to be specific because this fraction could be competed with increasing concentrations of wildtype VSV virions (Fig. 3A). Furthermore, particles containing QN-1 were internalized with kinetics similar to those of rescued virions containing wild-type G protein (Fig. 3B). This amount of specific binding correlated well with our previous results showing that only 30% of the virions have detectable amounts of G protein after rescue of tso45 with wild-type G protein (31). QN-1 is defective in cell-cell fusion activity. We next determined whether QN-1 had the membrane fusion activity characteristic of wild-type VSV G protein. When cells expressing wild-type G protein are incubated briefly at a ph of 6.0 or lower, the acid-induced fusogenic activity of VSV G protein causes cell-cell fusion (7, 22). Cells expressing either wild-type G protein or QN-1 were incubated for 1 min at ph 5.5 and then monitored for cell-cell fusion after incubation at 37 C in normal growth medium. Parallel cultures were also

4 4910 WHITT ET AL. J. VIROL. 1:4 C r.0. C s. '-U U CA Ul Us IL 10 bilized as de cell-associate A wild-type G protein after exposure to phs from 4.0 to 6.5, in increments of 0.5 ph unit. Cells expressing wild-type G protein formed syncytia after incubation at ph 6.0 or lower, but under no circumstances did cells expressing QN-1 form syncytia (data not shown). These results indicate that QN-1 is not able to induce membrane fusion and suggest that the lack of infectivity of particles containing QN-1 is due to a fusion defect. QN-1 interferes with rescue by wild-type G protein. The inability of particles containing QN-1 to infect cells and the El failure of cells expressing QN-1 to form syncytia after exposure to low ph suggested that the additional oligosaccharide on G protein was preventing fusion of the viral with the endosomal membrane. To determine. Cenvelope whether QN-1 might interfere with the function of wild-type UG protein (perhaps through the formation of hetero-oligojgvsv mers), we performed a rescue by expressing wild-type G protein and QN-1 simultaneously. Cells infected with tso45 and vtf7-3 were cotransfected with plasmids encoding QN-1 and wild-type G protein of the Indiana serotype (GI), _ B and the medium was assayed for infectious virus after incubation at 40.5 C. To control for any inhibitory effect resulting from the coexpression of two proteins in the rescue assay, we also examined the titers obtained after coexpressing G and an unrelated protein, the alpha subunit of human chorionic gonadotropin (hcg-ot) (12). Coexpression of QN-1 and wild-type G protein, at a ratio of 1:1.5, respectively, resulted in a 32-fold reduction in the titer of rescued virus compared with the titers found after rescue by coexpression of wild-type G and hcg-ot (Table 2). Increasing the amount of QN-1 expression 2-fold over that of wild-type G protein decreased infectivity greater than 180-fold. QN-1 and wild-type G protein can form heterotrimers when 01>1'1_1~ coexpressed. The strong inhibition of rescued virus infectiv ity resulting from the coexpression of QN-1 and wild-type G Time (min) protein suggested that QN-1 might be preventing wild-type G protein from inducing fusion of the viral envelope with the FIG. 3. In hibition of binding and endocytosis of particles con- endosomal membrane, thereby preventing infection. One taining QN-1. (A) Radioactively labeled virus particles containing mechanism by which QN-1 could interfere strongly with the QN-1 (2 x 1104 cpm) and various amounts of wild-type VSV were.. added to L cells in binding medium and incubated on ice for 3 h. The activity of wild-type G protein would be if QN-1 and monolayers viere washed to remove unbound virus and then solu- wild-type G protein formed mixed oligomers that were then Dscribed in Materials and Methods. The amount of fusion defective. Recently we have shown that wild-type and -d radioactivity was determined by scintillation count- some mutant forms of G protein can form heterotrimers ing, and the p'ercent inhibition of virus binding as a function of added efficiently when coexpressed (P. Zagouras, A. Ruusala, and wild-type VSV is shown. The total amount of radioactivity bound in J. Rose, unpublished data). To determine whether QN-1 the absence of wild-type VSV was 9.3% (1.8 x 103 cpm). (B) could form heterotrimers with other G proteins, we used a Radioactivelzy labeled particles containing either wild-type G protein coimmunoprecipitation assay. This assay relies on the use of -1 (El) were bound to cells as described above. After two different antibodies that can distinguish between the two ( * ) or QNwashing withkice-cold binding medium, the cells were incubated at teinsinvlvedin hat bee o 37 C for the times indicated and then treated with proteinase K to proteins involved heterotrimer formation. Because all of remove virus remaining on the cell surface. The amount of virus our antibodies recognize both QN-1 and wild-type G protein internalized a5s a function of time of incubation at 37 C is shown. The equally well, we examined whether QN-1 could form hetof virus bound was 9.6% for particles containing QN-1 erotrimers with a hybrid G protein (G23) which contains the total amount and 10.2% for particles containing wild-type G protein. extracellular and transmembrane domains of wild-type VSV G protein and the cytoplasmic domain of the infectious bronchitis virus p23 membrane protein (21). Immunoprecip- flow cytometry to determine the relative level itation of cell lysates with an antibody specific for the examined biy of surface elxpression for each protein. After exposure to ph cytoplasmic tail of infectious bronchitis virus p23 resulted in 5.5, cells e: xpressing wild-type G protein formed extensive the coprecipitation of G23 and QN-1. Likewise, antiserum syncytia, biut the cells expressing QN-1 showed no syncy- specific for the C-terminal portion of the VSV G protein tium formatlion (Fig. 4). The lack of fusion by QN-1 was not cytoplasmic domain precipitated both proteins (Fig. 5A). explained b' y reduced surface expression because the level of Sucrose gradient analysis of lysates from cells coexpressing expression observed by flow cytometry was slightly higher QN-1 and G23 followed by coimmunoprecipitation of the than in the Nwild-type G control. To determine whether QN-1 trimer peak with the cytoplasmic tail antibodies confirmed had a different ph requirement for fusion than wild-type G that QN-1 and G23 were in heterotrimers (data not shown). protein, we examined the fusion activity of either QN-1 or Because the determinants for trimer formation are in the _, - - '

5 j.-.-.i VOL. 64, 1990 GLYCOSYLATION MUTANT DEFECTIVE IN FUSION ACTIVITY m Sṙ~~~~~~~~~~~~~~~~' A 40 FIG. 4. Acid-induced syncytium formation in cells expressing QN-1 and wild-type G protein. HeLa cells were infected with vtf7-3 and transfected with plasmids containing the cdnas for either wild-type G, protein or QN-1. The cultures were rinsed once with PBS and then incubated for 1 min in fusion medium (PBS containing 10 mm MES and 10 mm HEPES, ph 5.5) at 7 h posttransfection. The fusion medium was removed, and the cultures were incubated in DMEM containing 5% fetal calf serum for 1 h to visualize cell-cell fusion. To ensure that the cultures were expressing equivalent amounts of QN-1 and wild-type G protein, surface expression was determined by flow cytometry from parallel cultures expressing QN-1 and wild-type G protein. (A) Mock-transfected; (B) wild-type G protein transfected; (C) QN-1 transfected. extracellular domain (3), it is therefore very likely that N-linked glycan at position 117 in the extracellular domain. wild-type G protein and QN-1 also form heterotrimers. Particles formed with the mutant glycoprotein are noninfectious, apparently because the QN-1 protein lacks all mem- If the strong inhibition of rescue caused by the coexpression of QN-1 with wild-type G protein were due to heterotrimer formation, then QN-1 should not interfere as strongly the failure of QN-1 to effect membrane fusion because cells brane fusion activity. We postulate a mechanism involving with rescue by a glycoprotein with which it does not form expressing QN-1 did not form synctia after exposure to heterotrimers. To test this hypothesis, we performed a acidic ph. This mutation (a single amino acid change) does rescue assay by coexpressing QN-1 and the wild-type G not appear to have major effects on protein folding because protein of the New Jersey serotype (GNJ)- We have shown it does not interfere with G protein oligomerization (5), previously that GNJ is efficiently incorporated into tso45 particles and produces infectious virions (31); however, QN-1 (an Indiana serotype G protein) and GNJ did not form A heterotrimers, apparently because of extensive divergence ( IroILil ' (I) I- QN ei Ii (29 of the Indiana serotype and New Jersey serotype G protein ectodomain sequences (Fig. SB). The effect of QN-1 on \IIhbLk)i 2 I rescue by GNJ was, in fact, much smaller than that seen for the homologous GI protein (Table 2); however, there was still significant interference by QN-1. DISCUSSION I.) i) 1 > I.L (j S In this report, we have described the incorporation into virions of a mutated VSV G protein (QN-1) that has an extra B (.^};ia t>n- (I QN ; I TABLE 2. Infectivity of virus rescued by coexpression of _ N T_ wild-type and glycosylation mutant G proteins Expressed Titer Fold * ~4mm protein(s) (PFU/ml)a reductionb Wild-type G, 1.0 X G, + hcg-a (1:1)c 1.3 x GI + QN-1 (1.5:1) 3.1 x FIG. 5. Coimmunoprecipitation of QN-1-G protein heterotrimers. Cells transiently expressing the indicated G proteins were GI + QN-1 (1:2) 5.5 x Wild-type GNJ 1.9 X pulse-labeled with [35S]methionine for 30 min, chased with excess GNJ + QN-1 (1:3) 2.7 x unlabeled methionine for 1 h, and then solubilized as described in GNJ + QN-1 (1:8) 4.9 x Materials and Methods. (A) G proteins were immunoprecipitated from the detergent lysates with anti-gi cytoplasmic tail antiserum a Determined on mouse L cells; values are averages from three experi- (T; lanes 1, 4, and 7), anti-infectious bronchitis virus p23 cytoplasmic tail antiserum (23; lanes 2, 5, and 8), or anti-vsvi antiserum recognizing the ectodomain of GI protein (I; lanes 3, 6, and 9). (B) G %'K~~~~~~~~~~~~~~~~rAN' 'C % ments. A total of 20 jig of DNA was used for transfection in each rescue assay. b Reduction in titer resulting from coexpression of QN-1 and wild-type G protein compared with that obtained after rescue with wild-type G protein alone. c Numbers in parentheses are the relative amounts of each protein, determined by scanning densitometry of fluorograms (not shown), when coexpressed in the absence of tso45. proteins were immunoprecipitated with anti-gi cytoplasmic tail antiserum (T; lanes 1, 4, and 7), anti-vsvnj antiserum (N; lanes 2, 5, 8, and 9), or anti-vsvi antiserum (I; lanes 3, 6, and 9). G proteins in lane 9 were immunoprecipitated with an admixture of anti-vsv1 and anti-vsvnj antisera.

6 4912 WHITT ET AL. transport to the cell surface, incorporation into virus particles, or subsequent binding and endocytosis of virus particles. The events that bring about fusion of the viral envelope with the endosomal membrane are not understood in precise molecular detail for any virus fusogen. In the case of influenza virus hemagglutinin and several other viral glycoproteins, proteolytic cleavage is required before acid activation in order to expose a critical hydrophobic domain which is thought to insert into the opposing membrane and facilitate redistribution of the two bilayers (28). For VSV G protein, no posttranslational cleavage occurs and there are no obvious hydrophobic regions that could participate in the fusion reaction. It has been suggested that a domain at the N terminus of G protein may serve as the fusion peptide (26); however, a G protein having a mutation in the critical region retained fusogenic activity equivalent to that of wild-type G protein (33). How does the QN-1 mutation interfere with the fusion activity? The VSV G protein, like other fusogenic proteins that act at low ph (10, 13, 18, 27), is though to undergo a conformational change induced by low ph (2, 4). This change may expose a hydrophobic domain that inserts into a target membrane (11, 30). It is conceivable that the extra glycan or the mutation itself might prevent such a conformational change. The only assay available for this conformational change is the stabilization of the G protein trimer at phs below 6.0 (4). Because trimers formed by the QN-1 mutant are stable at low ph (5), the mutant is probably undergoing a conformational change, although there might well be differences in the conformation from wild-type G protein that prevent fusogenic activity. An alternative explanation is that the mutation has a more direct effect and falls in or near the fusion domain itself. The hydrophilic glycan would presumably be a strong inhibitor of any membrane binding or insertion. Inspection of the G protein amino acid sequence reveals that the mutation immediately precedes a sequence of 19 uncharged amino acids (residues 118 to 136). Previously, Ohnishi suggested that residues 116 to 148 (residue numbering was changed to correspond to VSV G protein containing the signal sequence) may be involved in the fusion activity of VSV G protein because this sequence is highly conserved between the two VSV serotypes, although it contains several polar residues and is not strongly hydrophobic (20). Similarly, it has been proposed that a sequence of uncharged amino acids in the ectodomain of Semliki Forest virus El glycoprotein may be involved in membrane fusion (9). The sequence around the QN-1 mutation clearly deserves further investigation, since a fusion domain has not yet been identified in any rhabdovirus glycoprotein. The other interesting aspect of the fusion-defective mutant was that it interfered with infection when it was incorporated into VSV particles with the wild-type glycoprotein. When the mutant protein was expressed at levels only 2-fold greater than those of wild-type G protein, the virus titers were reduced 180-fold. Such strong interference might occur through the formation of fusion-inactive heterotrimers of QN-1 and wild-type G protein. Although we did not have antibodies that could distinguish between QN-1 and wildtype G protein, which prevented us from detecting QN-1- wild-type G protein heterotrimer formation directly, we did find that QN-1 is capable of forming heterotrimers with another mutant G protein containing a wild-type extracellular domain. The strong interference of infectivity observed after QN-1 and wild-type G protein coexpression suggests J. VIROL. that the presence of mutant subunits might well interfere with wild-type G protein fusion activity. Consistent with such a model, we found that when QN-1 was incorporated into virions with the VSV NJ glycoprotein (with which it does not form heterotrimers), there was much less inhibition of infectivity by QN-1. The expression of mutant subunits of oligomeric viral proteins to confer dominant negative effects on virus assembly and growth has been proposed as a strategy for controlling viral diseases such as acquired immunodeficiency syndrome (29). A dominant negative mutation that allowed assembly of the etiological agent and rendered it noninfectious could be more effective than a mutation blocking virus replication, since the host immune system might then respond more effectively to the particles produced. The results presented here suggest that a fusion-defective viral glycoprotein subunit might be very useful in preventing production of infectious virus. ACKNOWLEDGMENTS We thank Carolyn Machamer for helpful discussions during the course of this work and Debbie Brown for comments on the manuscript. We also thank Bernard Moss for the vaccinia virus expression system, Rob Edwards for performing the flow cytometric analysis, and Ann Curly-Whitehouse for photographic assistance. This work was supported by Public Health Service grants Al and CA from the National Institutes of Health. M.A.W. was supported by National Research Service award GM from the National Institute of General Medical Sciences. LITERATURE CITED 1. Bonner, W. M., and R. A. Laskey A film detection method for tritium-labelled proteins and nucleic acids in polyacrylamide gels. Eur. J. Biochem. 46: Crimmins, D. L., W. B. Mehard, and S. Schlesinger Physical properties of a soluble form of the glycoprotein of vesicular stomatitis virus at neutral and acidic ph. Biochemistry 22: Crise, B., A. Ruusala, P. Zagouras, A. Shaw, and J. K. Rose Oligomerization of glycolipid-anchored and soluble forms of the vesicular stomatitis virus glycoprotein. J. Virol. 63: Doms, R. W., D. S. Keller, A. Helenius, and W. E. Balch Role for adenosine triphosphate in regulating the assembly and transport of vesicular stomatitis virus G protein trimers. J. Cell Biol. 105: Doms, R. W., A. Ruusala, C. Machamer, J. Helenius, A. Helenius, and J. K. Rose Differential effects of mutations in three domains on folding, quaternary structure, and intracellular transport of vesicular stomatitis virus G protein. J. Cell Biol. 107: Felgner, P. L., T. R. Gadek, M. Holm, R. Roman, H. W. Chan, M. Wenz, J. P. Northrop, G. M. Ringold, and G. A. Danielsa Lipofection: a highly efficient, lipid-mediated DNA-transfection procedure. Proc. Natl. Acad. Sci. USA 84: Florkiewicz, R. Z., A. Smith, J. E. Bergmann, and J. K. Rose Isolation of stable mouse cell lines that express cell surface and secreted forms of the vesicular stomatitis virus glycoprotein. J. Cell Biol. 97: Fuerst, T. R., E. G. Niles, R. W. Studier, and B. Moss Eukaryotic transient-expression system based on recombinant vaccinia virus that synthesizes bacteriophage T7 RNA polymerase. Proc. Natl. Acad. Sci. USA 83: Garoff, H., A.-M., Frischauf, K. Simons, H. Lehrach, and H. Delius Nucleotide sequence of cdna coding for Semliki Forest virus membrane glycoproteins. Nature (London) 288: Gething, M.-J., R. W. Doms, D. York, and J. White Studies on the mechanism of membrane fusion: site-specific mutagenesis of the hemagglutinin of influenza virus. J. Cell Biol.

7 VOL. 64, 1990 GLYCOSYLATION MUTANT DEFECTIVE IN FUSION ACTIVITY : Gething, M.-J., J. White, and M. Waterfield Purification of the fusion protein of Sendai virus: analysis of the NH2- terminal sequence generated during precursor activation. Proc. Natl. Acad. Sci. USA 75: Guan, J.-L., and J. K. Rose Conversion of a secretory protein into a transmembrane protein results in its transport to the Golgi complex but not to the cell surface. Cell 37: Kielian, M., and A. Helenius Entry of alphaviruses, p In S. Schlesinger and M. J. Schlesinger (ed.), The togaviridae and flaviviridae. Plenum Publishing Corp., New York. 14. Laemmli, U. K Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature (London) 227: Machamer, C. E., R. Z. Florkiewicz, and J. K. Rose A single N-linked oligosaccharide at either of the two normal sites is sufficient for transport of vesicular stomatitis virus G protein to the cell surface. Mol. Cell. Biol. 5: Machamer, C. E., and J. K. Rose Influence of new glycosylation sites on expression of the vesicular stomatitis virus G protein at the plasma membrane. J. Biol. Chem. 263: Machamer, C. E., and J. K. Rose Vesicular stomatitis virus G proteins with altered glycosylation sites display temperature-sensitive intracellular transport and are subject to aberrant intermolecular disulfide bonding. J. Biol. Chem. 263: Marsh, M., and A. Helenius Virus entry into animal cells. Adv. Virus Res. 36: Matlin, K., H. Reggio, A. Helenius, and K. Simons The pathway of vesicular stomatitis entry leading to infection. J. Mol. Biol. 156: Ohnishi, S.-I Fusion of viral envelopes with cellular membranes. Curr. Top. Membr. Transp. 32: Puddington, L., C. E. Machamer, and J. K. Rose Cytoplasmic domains of cellular and viral integral membrane proteins substitute for the cytoplasmic domain of the vesicular stomatitis virus glycoprotein in transport to the plasma membrane. J. Cell Biol. 102: Riedel, H., C. Kondor-Koch, and H. Garoff Cell surface expression of fusogenic vesicular stomatitis virus G protein from cloned cdna. EMBO J. 3: Rose, J. K., and J. E. Bergmann Altered cytoplasmic domains affect intracellular transport of the vesicular stomatitis virus glycoprotein. Cell 34: Rose, J. K., and C. J. Gallione Nucleotide sequences of the mrna's encoding the vesicular stomatitis virus G and M proteins determined from cdna clones containing the complete coding regions. J. Virol. 39: Rosenberg, A. H., B. N. Lade, D.-C. Chui, S.-W. Lin, J. J. Dunn, and F. W. Studier Vectors for selective expression of cloned DNAs by T7 RNA polymerase. Gene 56: Schlegel, R., and M. Wade Biologically active peptides of the vesicular stomatitis virus glycoprotein. J. Virol. 53: Skehel, J., P. Bayley, E. Brown, S. Martin, M. Waterfield, J. White, I. Wilson, and D. Wiley Changes in the conformation of influenza virus hemagglutinin at the ph optimum of virus-mediated membrane fusion. Proc. Natl. Acad. Sci. USA 79: Stegmann, T., R. W. Doms, and A. Helenius Proteinmediated membrane fusion. Annu. Rev. Biophys. Chem. 18: Trono, D., M. B. Feinberh, and D. Baltimore HIV-1 Gag mutants can dominantly interfere with the replication of the wild-type virus. Cell 59: White, J., M. Kielian, and A. Helenius Membrane fusion proteins of enveloped animal viruses. Q. Rev. Biophys. 16: Whitt, M. A., L. Chong, and J. K. Rose Glycoprotein cytoplasmic domain sequences required for rescue of a vesicular stomatitis virus glycoprotein mutant. J. Virol. 63: Wilson, I., J. Skehel, and D. Wiley Structure of the hemagglutinin membrane glycoprotein of influenza virus at 3 A resolution. Nature (London) 289: Woodgett, C., and J. K. Rose Amino-terminal mutation of the vesicular stomatitis virus glycoprotein does not affect its fusion activity. J. Virol. 59: