ABSTRACT
During infection of host cells, a number of enveloped animal viruses are known to produce soluble forms of viral membrane glycoproteins lacking the transmembrane domain. The roles of such soluble glycoproteins in viral life cycles are incompletely understood, but in several cases they are believed to modulate host immune response and viral pathogenesis. Semliki Forest virus (SFV) is an enveloped alphavirus that infects cells through low-pH-dependent fusion and buds from the plasma membrane. Fusion is mediated by the E1 subunit of the SFV spike protein. Previous studies described the in vivo generation of E1s, a truncated soluble form of E1, under conditions in which budding is inhibited in mammalian host cells. We have here examined the properties of E1s generation and the biological activity of E1s. E1s cleavage required spike protein transport out of the endoplasmic reticulum and was independent of virus infection. Cell surface E1 efficiently acted as a precursor for E1s. E1s generation was strongly pH dependent in BHK cells, with optimal cleavage at a pH of ≤7.0, conditions that inhibited the budding of SFV but not the budding of the rhabdovirus vesicular stomatitis virus. The pH dependence of E1s production and SFV budding was unaffected by the stability of the spike protein dimer but was a function of the host cell. Similar to the intact virus and in vitro-generated E1 ectodomain, treatment of E1s at low pH in the presence of target membranes triggered specific acid-dependent conformational changes. Thus, under a variety of conditions, SFV-infected cells can produce a soluble form of E1 that is biologically active.
Enveloped viruses infect cells via fusion with a cellular membrane and produce progeny virus by budding through a cellular membrane containing the viral spike glycoproteins. Viral membrane glycoproteins are essential in mediating receptor binding, membrane fusion, and virus infection, and thus their packaging into viral particles is critical for the generation of infectious virions. Interestingly, during infection a number of enveloped viruses produce not only the full-length forms of viral glycoproteins but also truncated soluble forms lacking the transmembrane domain (5, 11, 37, 53). These soluble glycoproteins are inactive in virus assembly and membrane fusion but have been implicated in the pathogenesis of viruses such as vesicular stomatitis virus (VSV), Ebola virus, and respiratory syncytial virus (22, 24, 69).
Semliki Forest virus (SFV) is an enveloped alphavirus whose entry and exit pathways have been the focus of considerable research (14, 26, 60). Each virus particle contains an icosahedral nucleocapsid (T=4) consisting of the positive-sense RNA genome in association with 240 copies of the capsid protein. The lipid envelope surrounding the nucleocapsid excludes host cell proteins and contains a spike glycoprotein layer that is also organized with T=4 icosahedral symmetry (7, 12, 43, 50). The envelope contains two transmembrane polypeptides, E1 and E2, each with a mass of about 50 kDa, plus a peripheral polypeptide, E3 (∼10 kDa). Two hundred forty copies of these spike protein subunits are assembled into 80 spikes, each consisting of a trimer, (E1/E2/E3)3. E2 and E3 are synthesized as a precursor, p62, which noncovalently dimerizes with E1 in the rough endoplasmic reticulum. The E1/p62 heterodimers are transported through the secretory pathway to the cell surface, and at a late stage of transport, p62 is cleaved by the cellular protease furin into E2 and E3. The cytoplasmic domain of each of the E2 subunits associates with a capsid subunit in an interaction believed to drive the budding of virus at the plasma membrane. During spike protein biogenesis, the formation of the E1/E2 heterodimer is critical in correct protein folding, in providing the signal for E1 transport, and in the budding reaction (17, 39, 60). Budding is independent of p62 cleavage (54).
SFV infects a wide variety of host cells, including mammalian, avian, and insect cells. Infection occurs via receptor-mediated endocytic uptake and acid-triggered fusion of the virus membrane with the endosome membrane (see references 14, 26, 27, and60 for reviews). E2 is the receptor-binding subunit, while the E1 subunit contains the putative fusion peptide and mediates fusion via its interaction with the target membrane. The dimeric interaction of E2 and E1 acts to regulate the fusion reaction, and the dissociation of this interaction is the first biochemically detectable step during low-pH-triggered fusion (23, 54). Following dimer dissociation, the E1 subunit associates with the target membrane and forms a stable trypsin-resistant E1 homotrimer that is believed to be critical for fusion (16, 66). One key piece of evidence for this is that the mutation G91D in the E1 fusion peptide blocks virus fusion and inhibits E1 homotrimer formation (30). In addition to its low-pH requirement, membrane fusion of SFV is also strongly dependent on the presence of cholesterol and sphingolipid in the target membrane (31, 48, 51, 63, 68). A truncated soluble form of E1, termed E1*, has been generated in vitro by proteolytic removal of the hydrophobic transmembrane domain (28, 33). E1* undergoes similar low-pH-dependent membrane association and homotrimer formation as the intact E1 subunit but is inactive in membrane fusion due to the absence of the transmembrane domain (16, 33).
Alphavirus budding at the plasma membrane requires the expression of both capsid and spike proteins (39, 61, 71), and its efficiency is determined by several factors, including correct lateral interactions of the spike protein (10, 70), the expression of a small viral membrane protein, 6K (36, 38, 42), and the presence of cholesterol in the host cell membrane (41, 44, 63). In the absence of efficient virus budding, both the E1 and E2 spike protein subunits are rapidly turned over (9, 41, 71). However, under some conditions the E1 subunit, instead of being completely degraded like the E2 subunit, is released into the culture medium of infected BHK cells as a truncated soluble fragment, termed E1s (9, 71). E1s is generated when the spike proteins are expressed in the absence of capsid protein (71) or when the E1 subunit carries the G91D or G91A mutations in the fusion peptide (9). These mutations weaken the E1/E2 dimer interaction and make the virus temperature sensitive for assembly. At 37°C, the nonpermissive temperature, virus production is impaired and nearly all of the E1 is released as E1s (9). While E1s production is most efficient in cells infected with the capsid-minus or fusion peptide mutants, it is clear that small amounts of E1s are also released from wild-type (wt) SFV-infected BHK cells (9, 71).
Given the potential importance of SFV E1 cleavage in virus budding, we set out to further investigate its properties. Although the previous experiments indicated that E1s production occurred for the fusion peptide mutants with less stable E1/E2 dimer interaction, we found that E1 cleavage did not necessarily correlate with the lack or presence of a strong dimer interaction. Cleavage was independent of viral infection and was mediated by a host cell protease late in the secretory pathway. Similar to in vitro-generated E1 ectodomains, treatment of E1s at a low pH in the presence of target membranes triggered reactivity with an acid-specific monoclonal antibody and formation of a stable E1 homotrimer.
MATERIALS AND METHODS
Cells, viruses, and plasmids.BHK-21 cells were cultured at 37°C in Dulbecco's modified Eagle's (DME) medium containing 100 U of penicillin/ml and 100 μg of streptomycin/ml (P/S), 5% fetal bovine serum (FBS), and 10% tryptose phosphate broth (51). C6/36 cells, a clonal cell line derived fromAedes albopictus (20), were grown and maintained at 28°C in DME medium with P/S and 10% heat-inactivated FBS (44). COS-7 cells were cultured in DME medium with P/S, 10 mM HEPES (pH 7.4), and 10% FBS (35).
The wt SFV stock was a well-characterized plaque-purified SFV isolate (17) or was prepared from the pSP6-SFV-4 infectious clone by electroporation of the in vitro-transcribed RNA as previously described (36, 63). The sequences of these wt virus preparations were identical in the spike protein region (17). G91D was the previously characterized SFV E1 fusion peptide mutant with a change of glycine to aspartic acid at position 91 (9, 30, 35). Cells were infected via G91D RNA transcribed from the infectious clone containing the mutation. Mutant L (mL) was an SFV mutant in which p62 cleavage was blocked by the mutation of arginine to leucine at the E2 minus 1 position (54). An mL virus stock was prepared by in vitro transcription from the mL infectious clone, a gift from Peter Liljeström (54). A 75-cm2flask of BHK cells was then transfected using a mixture of 18 μg of mL RNA and 70 μg of Lipofectin reagent (Life Technologies, Gaithersburg, Md.) in Opti-MEM medium (Gibco/BRL, Gaithersburg, Md.) for 2 h at 37°C. The cells were then washed and further cultured at 37°C for 12 h before the collection of the stock. Since mL virus was not infectious, the stock was activated by trypsin cleavage and titered on BHK cells using a modified plaque assay with a trypsin-containing overlay (32, 54). VSV was the Indiana strain and was propagated and titered on BHK cells as previously described (67). All virus stocks were stored at −80°C in medium containing 10 mM HEPES (pH 7.4).
For analysis of transient expression, the structural proteins of wt and G91D virus were expressed in COS cells using the pCB3 vector containing the cytomegalovirus promoter, a polylinker region, the human growth hormone terminator, and the SV40 origin (3) (a gift from Michael G. Roth, Department of Biochemistry, University of Texas Southwestern Medical Center at Dallas). To simplify the construction, aSpeI/SalI fragment encoding the wt structural proteins was first subcloned from pSP6-SFV-4 into pGEM5ZF− (Promega, Madison, Wis.). The resultant plasmid, termed XYZ1, was then digested to generate a ∼900-bp XhoI/EcoRI fragment and a ∼3-kb HindIII/XhoI fragment. These were ligated in a three-way reaction with theHindIII/EcoRI fragment of the pCB3 construct to give pCB3-wt, encoding the complete wt structural proteins, capsid, p62, 6K, and E1. An EcoRV/BclI fragment from pL2-G91D (35) was exchanged with this fragment in the pCB3-wt construct to give pCB3-G91D containing the E1 G91D mutation.
Transfection of DNA or RNA into cells.COS-7 cells were plated 1 day prior to transfection in 35-mm plates and allowed to reach 40 to 50% confluency. Cells were washed twice with Opti-MEM and incubated with a mixture of 1.5 μg of pCB3-wt or pCB3-G91D DNA and 250 μg of DEAE-dextran (molecular weight, ∼5 × 105; Pharmacia, Piscataway, N.J.) in Opti-MEM for 1.5 h at 37°C. Cells were then cultured in 2 ml of fresh COS-7 culture medium containing 100 μM chloroquine for 4 to 5 h and further incubated in medium without chloroquine for an additional 40 to 44 h before metabolic labeling.
wt or G91D infectious RNA was prepared from the infectious clones and introduced into BHK cells via electroporation as previously described (9), and the cells were cultured for 6 h before pulse-chase analysis. G91D RNA was transfected into C6/36 and BHK cells in parallel by incubation for 4 h with 2.5 μg of RNA mixed with 9 μg of Lipofectamine (BHK) or 10 μg of Cellfectin reagent (C6/36) (Gibco/BRL), followed by culturing for an additional 12 h before pulse-chase analysis (40).
Virus infection, metabolic labeling, and cell surface budding assay.Budding of SFV and VSV from BHK and C6/36 cells was assayed using minor modifications of the assay previously described (41). Cells were infected with wt SFV, mL, or VSV at 10 PFU/cell for 1 h at 37°C and washed, and the incubation was continued for another 4 h. C6/36 cells were infected with wt SFV at 100 PFU/cell for 2 h at 28°C, washed, and further incubated for 4 h.
Following the above-described transfection or infection protocols, cells were starved in Met- and Cys-free medium for 15 min and pulse-labeled for 5 to 30 min in 0.6 ml of this medium containing 50 to 200 μCi of [35S]methionine-cysteine ([35S]Met-Cys, Pro-mix Cell labeling mix; Amersham Pharmacia Biotech, Arlington Heights, Ill.) and chased for various times in medium containing a 10-fold excess of Met and Cys (modified Eagle's medium [MEM] for BHK cells and Opti-MEM containing 0.2% bovine serum albumin for C6/36 cells).
To follow budding of SFV or VSV from the cell surface, virus-infected, pulse-labeled BHK or C6/36 cells were chased for 45 or 30 min, respectively, and derivatized with EZ-Link Sulfo-NHS-LC-Biotin (Pierce Chemical Co., Rockford, Ill.) as previously described (41). Biotin-tagged cells were then incubated in a water bath at 37°C for BHK cells or 28°C for C6/36 cells to permit budding. The incubation was performed in 1 ml of post-biotin incubation medium (MEM containing P/S and 10 mM HEPES without bicarbonate but supplemented with the equivalent concentration of NaCl) which had been preadjusted to the indicated pH.
To assay the effect of transport inhibitors on the E1 cleavage, BHK cells were pulse-labeled for 5 min, washed, and incubated in chase medium containing Brefeldin A (BFA) at a concentration of 5 μg/ml or carbonyl cyanide m-chlorophenyl hydrazone (CCCP) at a concentration of 50 μg/ml for 10 min on ice followed by 2 h at 37°C (29).
Following the above-described pulse-chase analysis or cell surface budding assay, medium samples were collected and protease inhibitors were added to final concentrations of 1 mM phenylmethylsulfonyl fluoride and 1 μg of pepstatin and leupeptin per ml. Cells were lysed on ice in 0.5 ml of a lysis buffer containing 1% Triton X-100 (TX-100) and protease inhibitors (29). Media and cell lysates were centrifuged at 10,000 rpm for 10 min to remove cell debris and nuclei (29), and the samples were then aliquoted for further analysis and stored as previously described (41). TX-100 was added to the medium at a final concentration of 1% as indicated.
Retrieval and analysis of viral proteins.Retrieval of biotin-tagged proteins or viruses was performed essentially as previously described using 25 μl of BioMag Streptavidin Ultra-Load Particles (magnetic streptavidin [mag-SA]) (PerSeptive Biosystems, Framingham, Mass.) (41) for one quarter of the medium from a 35-mm plate. Biotin-tagged intact viral particles were retrieved in the absence of detergent, while the biotinylated full-length and soluble glycoproteins were retrieved in the presence of 1% TX-100. The mag-SA particles were collected and washed using a magnetic tube holder (Magnetic Particle Concentrator MPC-E; Dynal, Oslo, Norway), and bound material was released by heating of the samples to 95°C in 1× sodium dodecyl sulfate (SDS) sample buffer for 5 min. VSV samples were reduced and alkylated as previously described (44). Both SFV and VSV samples were then analyzed by electrophoresis on 10 or 12% acrylamide gels, followed by fluorography (29). To recover the radiolabeled capsid protein in the biotinylated SFV virus particles, 3 μg of nonradiolabeled SFV was added to each sample with the SDS sample buffer. Gels were quantitated by phosphorimaging (ImageQuant v. 1.2; Molecular Dynamics, Inc., Sunnyvale, Calif.).
Release of SFV E1s and VSV Gs in the medium was measured, and the cleavage indices were calculated as follows. VSV G and Gs were well separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) (see Fig. 3, VSV gel), and thus the cleavage index for VSV G was defined as the amount of Gs divided by the sum of intact G plus Gs. It was difficult to directly quantitate the intact and cleaved forms of SFV E1 due to their similar electrophoretic mobilities. E1 and E2 are synthesized from a polyprotein precursor and have similar amounts of Met and Cys residues in their coding sequences (8 Met and 17 Cys in E1 and 9 Met and 16 Cys in E2). Previous data indicate that E2 is released from cells only in a heterodimeric complex with full-length E1 in intact virus particles (41). The amount of E2 recovered in the newly budded virus was therefore taken as equivalent to the amount of full-length E1 subunit, and E1s was calculated as the difference between the total released E1 and the full-length E1. The E1 cleavage index was defined as the amount of E1s divided by the total amount of E1.
Phase separation and immunoprecipitation.To better visualize E1s, medium samples from wt or G91D RNA-transfected cells were subjected to Triton X-114 (TX-114) phase separation (2, 28), to give a detergent phase enriched in intact transmembrane spike proteins and an aqueous phase enriched in soluble E1s. The aqueous and detergent phases, as well as control medium and cell lysate samples without phase separation, were immunoprecipitated with a polyclonal antibody against the SFV spike proteins followed by SDS-PAGE analysis and fluorography (29).
Assay of biological activity of E1s.Radiolabeled E1s was generated from 100-mm plates of BHK cells by infection, pulse-labeling, and chase but without biotinylation, followed by incubation for 1 h in 4.5 ml of the post-biotin incubation medium at pH 6.75, all as described above. The medium was collected and the cell debris was removed by low-speed centrifugation, and any budded intact virus was removed by ultracentrifugation for 30 min at 4°C at 54,000 rpm in a TLS55 rotor. The supernatant was concentrated, and the buffer was changed to morpholinoethanesulfonic acid-saline (20 mM morpholinoethanesulfonic acid [pH 7.0], 130 mM NaCl) using a Centricon-30 membrane (Amicon, Beverly, Mass.). E1*, the in vitro-generated E1 ectodomain, was tested in parallel as a positive control and was prepared by protease cleavage and purified by concanavalin A chromatography as previously described (16, 33). Liposomes were made by freeze-thaw and extrusion as previously described and contained phosphatidylcholine, phosphatidylethanolamine, sphingomyelin, and cholesterol (1:1:1:3) (“complete” liposomes) or this lipid mix without cholesterol (“minus-sterol” liposomes) (6, 16). E1s or E1* was mixed with liposomes at a final concentration of 1 mM lipid, treated at pH 7.0 or 5.5 for 10 min at 37°C, and adjusted to neutral pH (33). Samples were then assayed for reactivity with the acid-conformation-specific monoclonal antibody (MAb) E1a-1 (1, 29). Formation of the E1 homotrimer was assayed by its resistance to digestion at 37°C for 15 min with 200 μg of trypsin/ml as previously described (16). Samples were reduced and alkylated and were analyzed by gel electrophoresis and fluorography.
RESULTS
Role of virus infection in E1 cleavage.E1s production was first observed in cells infected with an SFV mutant with a deletion of the capsid gene (71). E1s is also produced in BHK cells infected with the SFV fusion peptide mutants G91D and G91A, with the most efficient E1s production occurring at 37°C, the nonpermissive temperature for assembly of these mutants (9). Low levels of E1 cleavage are also observed in wt SFV-infected BHK cells (9, 71).
We wished to determine the dependence of E1s production on virus infection as an approach to determining if cleavage was mediated by a cellular or viral protease. COS cells were therefore transfected with expression vectors encoding the structural proteins from either the wt SFV or the G91D mutant, and E1s production was evaluated by pulse-chase and SDS-PAGE analyses of the transiently expressing cells (Fig.1). Following a 15-min pulse with [35S]Met-Cys, both wt and G91D-transfected cells showed efficient expression of the viral p62 and E1 proteins (lanes 2 and 4), and no spike proteins were released in the medium (lanes 1 and 3). After a 2-h chase, p62 in both sets of cells was efficiently processed into mature E2 (lanes 6 and 8), indicating its correct folding and transport. No E2 protein was released in the chase medium of either wt or G91D-transfected cells (lanes 5 and 7), indicating that neither virus nor virus-like particles were produced. However, a protein migrating slightly faster than intact E1 was detected in the chase medium of G91D-expressing cells (lane 7). As is characteristic of E1s, this protein was immunoprecipitated by a monoclonal antibody against E1 and was water soluble as determined by its partitioning into the aqueous phase of a TX-114 phase separation (data not shown). In contrast, the intact membrane-bound spike proteins were found predominantly in the TX-114 detergent phase. Although E1s was not observed in the 2-h chase medium from wt cells (lane 5), longer chase times of wt cells revealed small amounts of E1s (data not shown), suggesting that cells expressing wt SFV structural proteins were also able to generate the soluble form of E1. Thus, E1 cleavage appeared to be independent of virus infection and occurred for the spike protein of both wt virus and a mutant virus with a less stable dimer.
Evaluation of E1 cleavage in the absence of virus infection. COS-7 cells were transfected with pCB3-wt or pCB3-G91D DNA encoding SFV structural proteins, as described in Materials and Methods. Forty-eight hours posttransfection, cells were pulse-labeled with 200 μCi of [35S]Met-Cys/ml for 15 min and chased for 0 or 2 h at 37°C. The medium samples (M) were collected, and the cells (C) were lysed. The radiolabeled spike proteins released in the medium and remaining in the cells were quantitated by immunoprecipitation with a polyclonal antibody against the spike proteins, followed by SDS–12% PAGE and fluorography. The data shown are a representative example of two experiments.
Cellular location of E1s production.We next wished to address the cellular site of E1s cleavage. Virus-infected cells were treated with the transport inhibitors CCCP or BFA. CCCP uncouples oxidative phosphorylation and inhibits endoplasmic reticulum (ER)-to-Golgi transport via energy depletion (47, 62), while BFA blocks anterograde transport to but not retrograde transport from the Golgi, resulting in an ER transport block and the recycling of the Golgi compartment to the ER (4).
For these experiments, BHK cells were transfected with infectious RNA derived from either the wt or G91D infectious SFV clones. The infected cells were pulse-labeled and chased for 2 h at 37°C in the presence of CCCP or BFA, and the spike proteins present in the medium and cell lysates were analyzed (Fig. 2). TX-114 phase separation was performed on the medium samples to distinguish between cleaved soluble E1 and intact membrane-bound E1. All samples were immunoprecipitated with a polyclonal antibody against the SFV spike protein and analyzed by SDS-PAGE. After a 2-h chase in the absence of inhibitors, p62 in lysates of either wt or G91D-infected cells was efficiently processed to E2 (Fig. 2A, lanes 1 and 4) and E3 (not detected in this electrophoresis system). In contrast, p62 cleavage was inhibited when the chase was performed in the presence of CCCP or BFA (Fig. 2A, lanes 2, 3, 5, and 6). In keeping with the known localization of p62 processing to a post-trans-Golgi network compartment (8), these results demonstrated efficient spike protein transport under control conditions and the inhibition of transport by CCCP or BFA. The chase medium from untreated wt cells contained intact E1 and E2 spike proteins, most of which partitioned into the detergent phase, indicative of virus budding (Fig. 2B and C, lanes 1, and data not shown). As predicted, when transport of wt spike protein was blocked by CCCP or BFA, virus budding was also blocked (Fig. 2B and C, lanes 2 and 3). In agreement with the previously reported assembly defect of G91D at 37°C (9), only water-soluble E1s was released in the chase medium of untreated G91D-infected cells (Fig. 2B and C, lanes 4). E1s production was completely inhibited when G91D-infected cells were chased in the presence of CCCP or BFA (Fig. 2B and C, lanes 5 and 6). Taken together, these results strongly suggest that the cleavage occurs at a late stage of the exocytic pathway, after transport to the Golgi compartment.
Effect of transport inhibitors on E1 cleavage. wt or G91D infectious RNAs were transcribed in vitro and transfected into BHK cells. The cells were incubated at 37°C for 6 h, pulse-labeled with 100 μCi of [35S]Met-Cys/ml for 5 min at 37°C, and then chased for 2 h at 37°C. To follow the effect of transport inhibitors on E1 cleavage, parallel cultures of cells were incubated with 50 μg of CCCP/ml or 5 μg of BFA/ml on ice for 10 min after the labeling and then chased at 37°C in the presence of these transport inhibitors. The media were collected, and the cells were lysed. TX-114 was added to the medium, and one aliquot of the medium sample was phase separated to yield aqueous and detergent phases which contained soluble and membrane proteins, respectively. Viral spike proteins in the medium (B), the aqueous phase of the medium (C), and the remaining cell lysates (A) were quantitated by immunoprecipitation with a polyclonal antibody against the spike proteins followed by SDS-PAGE analysis. The data shown are a representative example of two experiments. Note that the small amount of wt viral proteins present in the aqueous phase (panel C, lane 1) was not observed in a separate experiment.
The cell lysates of BFA-treated wt or G91D-infected cells contained a protein band migrating slightly faster than intact E1 protein (Fig. 2A, lanes 3 and 6). To rule out the possibility that this protein was an intracellular cleaved form of E1, TX-114 phase separation of the cell lysates was performed. The faster-migrating form of E1 partitioned into the detergent phase (data not shown), suggesting that it was most likely a differentially glycosylated form of E1 produced in the presence of BFA.
pH dependence of SFV and VSV glycoprotein cleavage and budding.The above-described transport inhibitor studies indicated that E1 cleavage occurred at a post-ER site in the secretory pathway. Both the E1 and E2 spike protein subunits are efficiently transported to the plasma membranes of G91D- and G91A-infected cells (9). Thus, data from these mutants do not further differentiate among possible post-ER sites of E1 cleavage, such as the Golgi or trans-Golgi network or after arrival at the plasma membrane. Previous results from Zhao and Garoff suggested that E1 cleavage might be occurring following delivery to the plasma membrane. Their studies showed that E1s was not recovered in cell lysates and that the cleavage kinetics correlated with delivery of the E2 subunit to the plasma membrane (71). However, in previous studies we observed E1s production from cells infected with G91P, an E1 fusion peptide mutant (58). The E1 subunit containing the G91P mutation localizes primarily to the ER and Golgi by immunofluorescence, and thus its cleavage suggested that E1s production might be occurring in the absence of cell surface transport. To address this issue, we used a sensitive cell surface biotinylation assay to examine E1 transport for this virus mutant and demonstrated that G91P E1 was delivered, albeit less efficiently, to the plasma membrane (S. G. Shome and M. Kielian, unpublished results).
To directly test if the cell surface pool of E1 was indeed a direct precursor for E1s, we used a cell surface protein biotinylation protocol that efficiently labels both E1 and E2 (41). This biotinylation method is the basis of an assay for the budding of cell surface spike proteins into progeny SFV particles (41). In the process of optimizing conditions for the budding assay, we found that budding was decreased when the biotin-tagged cells were incubated at pH 7 compared to the level observed at pH 8 (Y. E. Lu and M. Kielian, unpublished results). The decreased budding was accompanied by the production of E1s. To examine these issues further, we therefore tested whether cell surface E1 could be cleaved to form E1s and also the pH dependence of E1s production (Fig.3A). SFV-infected BHK cells were metabolically labeled with [35S]Met-Cys and chased to allow the delivery of the newly synthesized viral spike proteins to the cell surface. The cells were derivatized with biotin on ice and then incubated in medium of the indicated pH at 37°C to permit virus budding, all as previously described (41). The total biotin-tagged radiolabeled spike proteins present in the cells at the start of virus budding were quantitatively retrieved with mag-SA (Fig. 3A, 0-h CS lane). Abundant cell surface E1 and E2 subunits were specifically recovered, and no E1s was retrieved from the cell surface. The biotinylated spike proteins released during a 3-h, 37°C incubation were specifically recovered by mag-SA retrieval of the budding medium in the presence of detergent (Fig. 3A). As previously observed (41), biotinylated E1 and E2 were released at pH 7.5 and 8.0, with little biotinylated E1s observed. When the budding incubation was carried out at pH 7.0 or below, recovery of E2 protein in the medium was decreased and an increasing amount of E1s was detected (note the shift of the center of the E1 band from the pH 7.5 lane [E1] compared to the pH 7.0 lane [E1s]). At pH values of 6.5 or 6.75, the predominant biotinylated spike protein released was E1s, with little intact E1 or E2. These results thus indicate that E1s was produced from cell surface, biotin-tagged E1 under slightly acidic or neutral pH conditions, while production was inhibited under slightly basic conditions. Similar pH dependence was observed for E1s production from nonbiotinylated cells (data not shown).
pH dependence of SFV and VSV glycoprotein cleavage. BHK cells were infected with either SFV or VSV for a total of 5 h, radiolabeled for 15 min with [35S]Met-Cys at 100 μCi/ml for SFV and 200 μCi/ml for VSV, and chased for 45 min, and the cell surface proteins were derivatized with biotin (41). The biotinylated cells were then incubated for 3 h either on ice at pH 8.0 (referred to as time zero) or at 37°C at the indicated pH. The media were collected, and TX-100 was added to a final concentration of 1%. Biotinylated proteins in one-quarter of the medium were quantitatively retrieved with mag-SA as described in Materials and Methods. The cells that were incubated on ice were lysed in a TX-100-containing buffer, and one-quarter of the lysate was retrieved to quantitate the level of biotinylated cell surface proteins at time zero (labeled CS). Retrieved samples were analyzed by SDS-PAGE (A and B) and quantitated by phosphorimaging (C). To compare the pH dependence of glycoprotein cleavage, release of SFV E1s and VSV Gs was expressed as a cleavage index, calculated as described in Materials and Methods.
The strong pH dependence of E1s production was surprising, and we explored whether a similar pH dependence existed for production of the soluble form of another virus spike protein known to be produced by proteolytic cleavage during virus infection. Extensive studies in the literature reported that cells infected with VSV produce a soluble form of the spike glycoprotein, Gs, lacking the transmembrane domain and the cytoplasmic tail (5, 15, 21, 37). While some Gs is clearly produced intracellularly, at least part of the total Gs appears to result from cleavage of a cell surface pool (5, 37). We biotin tagged the cell surface G protein in VSV-infected BHK cells and compared the pH dependence of Gs generation with that for SFV E1s (Fig.3B). Intact G protein was labeled at the cell surface before the post-biotin incubation (Fig. 3B, CS lane). When the pH of the incubation medium was 7.0 or lower, the faster-migrating Gs form was released into the medium. Similar to SFV E1 cleavage, very little Gs was observed when the pH of the incubation medium was in the slightly basic range of pH 7.5 or 8.0. Quantitation and direct comparison of SFV E1 cleavage and VSV G cleavage showed a similar pH dependence, although cell surface G protein was cleaved much less efficiently than cell surface E1 protein (Fig. 3C).
Alphavirus budding is dependent on the expression and interaction of both the spike and capsid proteins (13, 39, 61, 71). In previous studies, we observed an inverse relationship between spike protein degradation and virus budding in cholesterol-containing or cholesterol-depleted insect cells, although in these experiments E1 was fully degraded rather than released as E1s (41). In order to test the correlation between E1s production and SFV budding, we compared the pH dependence of these two processes. As a measure of the overall effects of pH incubation on the cells, we measured VSV budding in parallel. While the budding of rhabodoviruses is also dependent on both the viral glycoprotein and the nucleocapsid, the requirement for the viral spike protein is less stringent than that of alphavirus budding (45), and the production of Gs is less pronounced at an acidic pH (Fig. 3). We therefore speculated that VSV budding would be less susceptible to the potential effects of incubation at mildly acidic pH.
Our previously developed assay was used to measure the budding of virus particles from the surfaces of SFV- or VSV-infected BHK cells (41). Newly synthesized radiolabeled viral glycoproteins were tagged with biotin at the cell surface, the cells were incubated in media of the indicated pH, and the biotinylated virus particles that were released in the media were retrieved with mag-SA in the absence of detergent and analyzed by SDS-PAGE (Fig.4A and B). The recovery of the viral internal proteins (SFV capsid protein and VSV N, P, and M proteins) confirmed that under these conditions the assay retrieved complete intact viral particles (see also reference 41 for more detailed controls). Budding was quantitated for SFV by following E2 retrieval, since this spike protein subunit is not released in the absence of budding and since the capsid protein pool is much larger than that of the spike proteins (41). VSV budding was quantitated by M protein retrieval (44). SFV budding was optimal at a slightly basic pH, with maximum efficiency observed between pH 7.5 and 8.0 (Fig. 4A and C). As the pH value of the incubation medium became more acidic, the budding of SFV showed a marked decrease while the production of E1s increased. Almost no budding was detected at pH values of 6.5 and 6.75. Similar results were obtained from nonbiotinylated cells (data not shown). In contrast, budding from VSV-infected cells was essentially independent of incubation at pH values from 6.5 to 8 (Fig. 4B and C), although Gs was produced when the cells were incubated at a mildly acidic pH (Fig. 4B). Thus, although BHK cells could efficiently bud virus particles under mildly acidic conditions, as exemplified by the VSV results, the budding of SFV was severely depressed under these conditions. It is not yet clear if the cleavage of E1 to E1s at a mildly acidic pH was the cause or an effect of inefficient SFV budding.
pH dependence of SFV and VSV budding. BHK cells were infected with either SFV or VSV, labeled, chased, derivatized with biotin, and incubated to allow the incorporation of biotinylated viral spike protein into budded virus particles, all as described in the legend to Fig. 3. The post-biotin incubation media were collected, and the biotinylated virus particles and cleaved glycoproteins were quantitatively retrieved in the absence of detergent. The mag-SA-bound material was analyzed by SDS-PAGE (A and B). To compare the SFV and VSV budding efficiencies, the amount of E2 (SFV) or M protein (VSV) was quantitated by phosphorimaging, and budding at each pH was expressed as a percentage of the E2 or M release at pH 8.0 (C). The data are representative of four (SFV) and one (VSV) experiments.
The effect of E1/E2 dimer association on SFV budding and E1s production.The SFV E1/E2 spike protein heterodimer interaction is pH sensitive, being highly stable at slightly basic pH and destabilized at pH values of ≤∼7.0 (65). Given the similarity of this pH dependence to that of SFV budding and the importance of heterodimer formation in the budding reaction (17, 39, 60), we speculated that stabilization of the E1/E2 dimer at basic pH resulted in more efficient virus budding and decreased E1 cleavage. To test this theory, we utilized an SFV mutant, mL, which has a mutation at the tetrabasic cleavage site of p62 that blocks processing and therefore results in the production of virus particles containing p62/E1 dimers (54). Compared to the E2/E1 dimer, the p62/E1 dimer is less pH sensitive and is not dissociated until treatment at pH values of ≤5.0 (54, 65). BHK cells were infected with wt or mL virus, and virus budding was tested at pH values ranging from 6.5 to 8.0. The biotinylated virus particles in the medium and any biotin-tagged E1s were retrieved with mag-SA in the absence of detergent and analyzed by SDS-PAGE (shown in Fig.5A for mL). Capsid protein was recovered from both mL and wt SFV samples, suggesting the retrieval of intact virus particles (data not shown). Budding was quantitated for both viruses by following the retrieval of either p62 or E2 (Fig. 5B). The results demonstrated that, similar to wt budding, mL budding was inhibited at a mildly acidic pH and was more efficient at a basic pH. E1s was efficiently produced from either mL- or wt-infected cells at an acidic pH. Quantitation of budding demonstrated that budding of mL had a pH dependence comparable to or even slightly more basic than that of wt SFV (Fig. 5B). These results thus suggested that dimer stability was not primarily responsible for basic pH optimum of SFV budding.
pH dependence of E1 cleavage and budding for SFV mL. BHK cells were infected with the p62 cleavage-defective mL and radiolabeled, chased, and biotin-derivatized as for Fig. 3, followed by incubation in media of the indicated pH for 60 min at 37°C. The media were collected and retrieved with mag-SA in the absence of detergent, followed by SDS-PAGE analysis (A) and quantitation by phosphorimaging (B). The pH dependence of wt SFV budding was measured in parallel and similarly quantitated (B). Virus budding efficiency at each pH was calculated as a percentage of the E2 (wt) or p62 (mL) release at pH 8.0 (B). The data are representative of four (wt SFV) or two (mL) experiments.
pH dependence of SFV budding and E1 cleavage in insect cells.The results to this point suggested that E1s cleavage was not dependent on virus infection, that it required transport of the spike protein out of the rough ER, and that both cleavage and budding showed a strong pH dependence that was distinct from the pH sensitivity of the spike protein heterodimer interaction. These results suggested that the pH dependence of cleavage and budding might be conferred by the host cell, and we therefore examined their properties in an alternative host cell. We utilized C6/36 mosquito cells, a cell type that we have used extensively in previous studies of the cholesterol requirements for SFV and Sindbis virus membrane fusion and budding (40, 41, 44, 63). C6/36 cells were cultured under control cholesterol-containing conditions, infected with SFV, and used in the cell surface budding assay at the indicated pH values (Fig.6). The E2 in C6/36 cells migrated as a doublet in SDS-PAGE (Fig. 6A), presumably due to the known glycosylation differences between mammalian and insect cells (19). In contrast to the efficient cleavage of E1 in BHK cells at mildly acidic pH, no faster-migrating form of E1 was detected in the medium from C6/36 cells incubated at pH 6.5 to 8.0 (Fig. 6A). This was confirmed by direct comparison of the E1 cleavage indices for C6/36 cells and BHK cells (Fig. 6B). Quantitation of virus budding showed that budding from C6/36 cells was as efficient at pH 7.0 as at a more basic pH, such as pH 7.5 or 8.0 (Fig. 6A and data not shown). Budding declined somewhat at an acidic pH but still showed ∼30% efficiency at pH 6.5 and ∼65% efficiency at pH 6.75 compared to the efficiency of budding at pH 8.0. In contrast, parallel budding experiments with BHK cells showed budding efficiencies of ∼5% at pH 6.5, ∼10% at pH 6.75, and 25% at pH 7.0. Taken together, these results indicated that the pH dependencies of SFV budding and E1 cleavage were cell type specific.
pH dependence of SFV E1 cleavage and budding in insect cells. C6/36 cells were infected with SFV at a concentration of 100 PFU/cell for a total of 6 h, pulse-labeled for 15 min with 50 μCi of [35S]Met-Cys/ml, chased for 30 min, and derivatized with biotin. The cells were then incubated at 28°C for 1 h in medium of the indicated pH to allow virus budding. Virus particles and biotinylated proteins in the medium were retrieved with mag-SA in the absence of detergent and analyzed by SDS-PAGE (A). SFV budding and cleavage in BHK cells was assayed in parallel as for Fig.4. Although no detectable E1s was produced by the insect cells, cleavage indices for both cell types were calculated as described in Materials and Methods and compared in panel B.
Detection of E1s cleavage in insect cells.The absence of E1s production by wt SFV-infected C6/36 cells at an acidic pH suggested that either the cells did not contain the necessary enzymatic activity or they were insensitive to incubation at an acidic pH. To differentiate between these two possibilities, we utilized the G91D SFV mutant and examined budding and E1s production in G91D-infected C6/36 or BHK cells. G91D has a strong budding defect at 37°C, buds inefficiently even at the permissive temperature of 28°C, and produces significant amounts of E1s at either temperature (9). Infectious G91D RNA was introduced into C6/36 cells or BHK cells by transfection, the cells were pulse-labeled and chased for 4 h at 28°C, and the samples were immunoprecipitated to follow viral spike proteins in the cells and chase medium (Fig.7). Radiolabeled E1 and E2 spike proteins were expressed in both BHK and C6/36 cells (lanes 1) and released into the medium during the chase period (lanes 2), although the efficiency of release was lower in the C6/36 cells. The chase medium from both cell types contained what appeared to be a doublet migrating close to the position of the authentic E1 subunit. In order to better differentiate between E1s and intact E1, medium samples from both cell types were analyzed by TX-114 phase separation (lanes 3 and 4). E1 and E2 spike proteins in the medium from BHK cells were predominantly recovered in the detergent phase, indicating that they are intact membrane-bound forms (lane 4), while the aqueous phase contained the faster-migrating E1s (lane 3). The detergent phase of the chase medium from C6/36 cells contained equimolar amounts of the intact E1 and E2 subunits (lane 4), suggesting that similarly to BHK cells, C6/36 cells could produce G91D virus particles at 28°C. The aqueous phase of the C6/36 chase medium contained a band migrating at the position of E1s (lane 3, C6/36 cells). When C6/36 cells were chased at 37°C, the nonpermissive temperature for G91D virus assembly, increased production of E1s was observed (data not shown). Thus, these results indicated that C6/36 cells were capable of producing E1s when infected with the G91D mutant, indicating that the enzymatic activity required for cleavage was present. The lack of E1s production during incubation of wt-infected C6/36 cells at low pH may be due to the relative insensitivity of insect cells to acidic-pH medium (49).
Production of E1s in G91D-infected BHK and insect cells. BHK or C6/36 cells were transfected with G91D infectious RNA using Lipofectamine or Cellfectin reagents, respectively. After 16 h of incubation at 28°C, cells were pulse-labeled with 200 μCi of [35S]Met-Cys/ml for 30 min at 37°C (BHK cells) or 28°C (C6/36 cells) and chased for 4 h at 28°C. The media were collected, and the cells were lysed. TX-114 was added to the medium samples, and an aliquot of each was phase separated as for Fig. 2 to yield an aqueous phase containing soluble proteins and a detergent phase containing the transmembrane forms of E1 and E2. The cell lysates, medium samples, and both phases of the media were immunoprecipitated as described in the legend to Fig. 1 and analyzed by electrophoresis on 12% acrylamide gels. To better visualize the cleaved products, the medium samples in the BHK cell panel (lanes 2, 3, and 4) were exposed twice as long as lane 1, and the medium samples in the C6/36 cell panel (lanes 3 and 4) were exposed twice as long as lanes 1 and 2. The data are representative of three experiments.
Biological activity of E1s.The above-described studies demonstrated that both insect cells and BHK cells were able to generate the soluble form of E1. We next wanted to test whether this in vivo-generated E1s fragment was competent to respond to acidic pH. Upon low-pH treatment, either full-length viral E1 or the in vitro-generated E1* ectodomain undergoes specific conformational changes resulting in exposure of acid-conformation-specific epitopes and formation of a highly stable, protease-resistant E1 homotrimer (26, 27). These acid-dependent conformational changes are promoted by the presence of cholesterol and sphingolipid-containing target membranes. Radiolabeled E1s was generated from wt SFV-infected BHK cells by pulse-labeling and chase at a slightly acidic pH (pH 6.75). Any intact virions were removed by centrifugation, and the E1s in the chase medium was concentrated, mixed with complete or cholesterol-deficient liposomes, and treated at pH 7.0 or 5.5 (see Materials and Methods). As a control, E1* was generated by in vitro protease cleavage (33), purified, mixed with chase medium, and similarly concentrated and assayed. Following pH treatment, E1s and E1* were immunoprecipitated either with a polyclonal antibody to detect the total SFV spike proteins present (Fig.8A) or with the acid-specific MAb E1a-1 to detect exposure of the acid-specific epitope, which was previously mapped to residue 157 of the E1 subunit (1) (Fig. 8B). Negligible amounts of neutral-pH-treated E1* or E1s reacted with MAb E1a-1, demonstrating that the in vivo-generated E1s was not released from cells in the “acid” conformation (Fig. 8B, lanes 1 and 2). Upon low pH treatment, both E1s and E1* were precipitated by MAb E1a-1 (Fig. 8B, lanes 3 and 4), and quantitation showed that the immunoreactivity represented 23% of the total E1s and 30% of the total E1*. Conversion of either E1* or E1s was inhibited when the acid treatment was performed in the presence of sterol-deficient target membranes (Fig. 8B, lanes 5 and 6), and quantitation showed that under these conditions MAb E1a-1 precipitated only 0.5% of the total E1s.
Biological activity of E1s. Radiolabeled E1s or E1* was mixed with either complete liposomes (+sterol) or liposomes lacking sterol (−sterol), treated at pH 7.0 or 5.5 for 10 min, and adjusted to neutral pH. Aliquots of the samples were then treated as follows. (A) The total spike proteins in the reaction were quantitated by immunoprecipitation with a polyclonal rabbit antibody against the spike protein. (B) The samples were immunoprecipitated with MAb E1a-1, a monoclonal antibody specific for the acid conformation of E1. (C) Samples were digested with 200 μg of trypsin/ml for 15 min at 37°C to evaluate the trypsin-resistant homotrimer. All samples were then analyzed by SDS-PAGE and fluorography, and equivalent exposures of all E1s lanes and of all E1* lanes are shown. The data shown are a representative example of three separate experiments.
Aliquots of these same reactions were tested for the presence of the E1 homotrimer by digestion with trypsin at 37°C (16) (Fig.8C). Both E1s and E1* were highly trypsin resistant following low-pH treatment in the presence of complete liposomes (Fig. 8B, lanes 3 and 4) but not after neutral pH treatment (lanes 1 and 2) or acid treatment in the presence of sterol-deficient liposomes (lanes 5 and 6), in keeping with the previous findings of the acid and cholesterol dependence of this conformational change for E1* (16, 33). Quantitation showed that 87% of the total E1s and 74% of the total E1* were resistant to trypsin digestion after acid treatment with complete liposomes. By comparison, only 3% of the total E1s and 1% of the total E1* were trypsin resistant after acid treatment with sterol-deficient liposomes. Thus, these data indicate that the in vivo-generated E1s form of the SFV fusion protein showed acid- and cholesterol-dependent conformational changes similar to those previously described for the E1* form.
DISCUSSION
Examples from several unrelated virus families demonstrate that soluble forms of viral glycoproteins can be generated via various pathways and can play important roles in virus pathogenesis. The filovirus Ebola virus produces two major forms of its glycoprotein, a membrane-bound form that is translated from edited mRNA and incorporated into virus particles and a soluble, secreted form that is translated from unedited mRNA (11, 55, 56, 64, 69). The paramyxovirus respiratory syncytial virus (RSV) synthesizes a soluble, secreted form of the glycoprotein G from a second initiation codon (53). The rhabdoviruses VSV and rabies virus produce the soluble Gs form of the rhabdoviral fusion protein via several pathways. A significant amount of Gs is generated intracellularly either by premature termination of translation or by proteolysis (references15, 18, and 46; our data not shown). This soluble G can oligomerize with intact G in the ER and is then rapidly transported to the cell surface and efficiently secreted (57). Several reports have demonstrated that Gs cleavage can also occur at the cell surface (5, 21, 37). In agreement with these results, our studies demonstrated that biotin-tagged G could be cleaved to form Gs (Fig. 3). Such Gs cleavage was increased by incubation at a mildly acidic pH, similar to previous reports of low-pH-dependent Gs production for rabies virus (25, 46). We found that VSV budding from BHK cells still occurred efficiently at mildly acidic pH, in keeping with studies showing that rhabdovirus budding can occur even in the absence of glycoprotein expression (45) and that spikeless rabies virus particles are produced at a low pH (25).
The available data on alphaviruses suggest that two types of proteolysis can act to break down the cell surface pool of transmembrane spike proteins. The E1 subunit can be cleaved to the truncated E1s form and released from the cell (9, 71). Alternatively, the spike protein E1 and E2 subunits can be degraded, a more generalized process of spike protein turnover that does not produce specific fragments resolved by SDS-PAGE (41). Both degradative processes can apparently act in concert, since E1s production occurs under conditions in which the E2 spike subunit is degraded (9, 71). The highest levels of either form of proteolysis are observed in conjunction with inefficient virus budding. Thus, E1s is produced in BHK cells infected with a virus lacking the capsid protein (71), in BHK cells infected with wt virus and incubated at a pH of ≤7.0, as described here, or in either C6/36 or BHK cells infected with virus containing the E1 G91D mutation, particularly at the nonpermissive assembly temperature of 37°C (Fig. 7) (9). In all three of these examples, virus budding was either completely blocked or greatly decreased. Generalized degradation of both the E1 and E2 subunits was observed in C6/36 cells at early times of wt infection, when budding is inefficient (41). Similar to E1s production, degradation occurs following spike protein delivery to the cell surface, suggesting a possible endocytic route of cleavage. This rapid spike protein turnover is inhibited at later times of infection when efficient budding is observed. In addition, even at late times of infection, cholesterol-depleted C6/36 cells show inefficient budding of wt virus and high levels of spike protein degradation, while a less cholesterol-dependent SFV mutant has both increased budding efficiency and decreased spike protein turnover (41). Taken together, these results establish a strong correlation between the efficient proteolysis of the cell surface SFV spike protein and inefficient virus budding. It is not clear if spike protein degradation acts to inhibit virus budding or is itself a secondary effect of inefficient budding. In this regard, it is interesting that E1s was produced from COS cells expressing the G91D mutant spike protein but not from cells expressing the wt spike protein, even though no budding could occur for either expressed protein (Fig. 1). This may suggest that E1s cleavage is controlled upstream of budding and that it could act to regulate the concentration of budding-competent spike proteins at the plasma membrane.
It is not clear what determines whether the cell surface E1 subunit will be cleaved to form E1s versus undergoing a more generalized degradation. Previous work on the G91D and G91A mutants correlated their efficient E1s production and lack of budding with a dramatic decrease in the stability of the E2/E1 dimer (9). However, mL virus producing only the more acid-stable p62 form of the spike protein dimer nevertheless still produced E1s, and it did not support efficient budding at acidic pH. Conversely, wt-infected C6/36 cells incubated at an acidic pH did not produce E1s, even though the pH dependence of the spike dimer would presumably be the same in this cell type. C6/36 cells were capable of cleaving E1 from the G91D mutant, demonstrating that the requisite enzymatic activity was present in this cell line. Taken together, these results suggest that a mechanism other than or in addition to dimer stability is involved in E1s cleavage and the pH dependence of SFV budding.
What might be the site(s) and proteases responsible for the production of SFV E1s? The cleavage of the expressed G91D spike protein in COS cells demonstrated that E1s generation was independent of virus infection, ruling out a possible role of virus infection-dependent factors in E1 cleavage. A major degradation pathway for cellular membrane glycoproteins is that mediated by the proteasome complex, in which newly synthesized misfolded membrane proteins are translocated out of the ER into the cytosol, ubiquitinated, and degraded by proteasomes (see reference 52 for a review). Unlike proteasome degradation, however, production of E1s was blocked when transport of the spike protein out of the ER was inhibited (Fig. 2), clearly indicating that E1 cleavage was not mediated by the ubiquitin/proteasome pathway. Unlike soluble forms of some other viral glycoproteins, SFV E1s was not detected in cell lysates but rather was found exclusively in the medium (references 9 and71; data presented here). We have shown that E1s cleavage occurred after delivery of E1 to the plasma membrane, as defined by biotin derivatization. Cell surface spike proteins might be cleaved by a plasma membrane-associated protease or could be internalized by endocytosis, cleaved by endocytic proteases, and recycled back to the plasma membrane to be released into the medium. The E2 protein has a sequence within its cytoplasmic tail, LTPYALT, containing a tyrosine residue that could potentially act as a signal for endocytosis. The LTPY sequence is highly conserved and interacts with the nucleocapsid protein during budding, suggesting a possible mechanism that might decrease endocytic uptake late in infection (34, 59, 72). We found that the wt spike protein was efficiently endocytosed during early times of infection in insect cells when budding was inefficient (data not shown). However, it was not clear if such uptake could cause either E1s production or the generalized degradation of the E1 and E2 spike subunits. In addition, the G91D spike protein is an efficient substrate for E1s production even though morphological studies demonstrate that this spike protein does interact with the nucleocapsid at the plasma membrane (9). Attempts to use protease inhibitors to reduce spike protein cleavage and rescue budding were unsuccessful (unpublished results). Clearly, further studies will be required to establish the enzyme(s) responsible for E1s cleavage and spike protein turnover, the cellular site(s) of cleavage, and the importance of endocytic uptake in these processes.
It is intriguing that wt SFV-infected cells can produce a significant amount of E1s even at a neutral pH. The role, if any, of E1s in the pathogenesis of alphaviral disease is unknown. Our experiments demonstrated that, similar to the E1* ectodomain, E1s was able to undergo specific conformational changes that were dependent on both low pH and cholesterol-containing target membranes. The lack of association of E1s with the receptor-binding E2 subunit would presumably preclude its endocytic uptake and exposure to endosomal low pH in vivo. Nonetheless, the in vivo-generated E1s form of the alphavirus spike protein may prove useful in providing an alternative experimental method to study E1 mutants that are difficult to produce in virus particles.
ACKNOWLEDGMENTS
We thank Anna Ahn for very helpful technical assistance, particularly with the experiments shown in Fig. 8. We also thank Xinyong Zhang for preparation of the pCB3-wt construct, Michael G. Roth for providing the pCB3 vector, and Peter Liljeström for providing the SFV wt and mL infectious clones and helpful advice in their use. We also thank the members of our lab for helpful discussions and suggestions and for critical reading of the manuscript.
This work was supported by a grant to M.K. from the Public Health Service (R01 GM57454), by the Jack K. and Helen B. Lazar Fellowship in Cell Biology, and by Cancer Center Core Support Grant NIH/NCI P30-CA13330.
FOOTNOTES
- Received 20 April 2001.
- Accepted 31 May 2001.
- Copyright © 2001 American Society for Microbiology
