Interactions between the Transmembrane Segments of the Alphavirus E1 and E2 Proteins Play a Role in Virus Budding and Fusion

Abstract

The alphavirus envelope is built by heterodimers of the membrane proteins E1 and E2. The complex is formed as a p62E1 precursor in the endoplasmic reticulum. During transit to the plasma membrane (PM), it is cleaved into mature E1-E2 heterodimers, which are oligomerized into trimeric complexes, so-called spikes that bind both to each other and, at the PM, also to nucleocapsid (NC) structures under the membrane. These interactions drive the budding of new virus particles from the cell surface. The virus enters new cells by a low-pH-induced membrane fusion event where both inter- and intraheterodimer interactions are reorganized to establish a fusion-active membrane protein complex. There are no intact heterodimers left after fusion activation; instead, an E1 homotrimer remains in the cellular (or viral) membrane. We analyzed whether these transitions depend on interactions in the transmembrane (TM) region of the heterodimer. We observed a pattern of conserved glycines in the TM region of E1 and made two mutants where either the glycines only (SFV/E1^4L) or the whole segment around the glycines (SFV/E1^11L) was replaced by leucines. We found that both mutations decreased the stability of the heterodimer and increased the formation of the E1 homotrimer at a suboptimal fusion pH, while the fusion activity was decreased. This suggested that TM interactions play a role in virus assembly and entry and that anomalous or uncoordinated protein reorganizations take place in the mutants. In addition, the SFV/E1^11L mutant was completely deficient in budding, which may reflect an inability to form multivalent NC interactions at the PM.

Alphaviruses use a heterodimeric membrane protein complex, E1-E2, for their membrane assembly in an infected cell and for entry into uninfected cells by membrane fusion (39, 55). In the infected cell, the heterodimers assemble into higher oligomeric complexes that, at the plasma membrane (PM), interact with the viral nucleocapsid (NC) and drive virus budding and release (9, 28, 43). In the virus, the heterodimers are found to be organized into a fenestrated T=4 icosahedral glycoprotein shell around the particle, with trimeric spike-like protrusions at the 3-fold and quasi-3-fold (q3-fold) axes and pentameric and hexameric interactions around shell openings at the 5- and 2-fold axes. From the latter positions, C-terminal parts of the heterodimeric E1 and E2 subunits dive into and penetrate the viral lipid bilayer to meet and interact with the capsid (C)-protein monomers of the NC, which are arranged in pentameric and hexameric capsomers at the corresponding positions (5, 46). The membrane fusion process is directed by the E1 subunit and controlled by the E1-E2 intersubunit interaction (31, 47). The crystal structure of the Semliki Forest virus (SFV) E1 ectodomain has recently been solved, and it revealed an elongated protein with most of the polypeptide folded into β-sheets (21, 50). One end formed an immunoglobulin (Ig)-like fold, and the other harbored the putative (internal) fusion peptide (22). Fitting of this structure into the cryo-electron microscopy (cryo-EM)-derived density map of SFV indicated that the Ig-like domains constitute the shell regions around the 5- and 2-fold axes. From here, the E1 subunits raised obliquely to the membrane plane toward the q3- and 3-fold axes, forming the sides of the spikes and the lower parts of their tips (21). While the data suggested that E2 protects the E1 fusion peptide in the spike tip and that E1 mediates most of the interheterodimeric interactions in the shell region, they also pointed to the complexity of rearrangements that must occur during fusion activation. This must include both shell dissolution and E1 fusion peptide exposure (11). Indeed, earlier biochemical studies about SFV fusion with liposomes or cell membranes showed that the NP-40-resistant heterodimer dissociated and that E1 formed homotrimers instead (2, 49). It was recently shown through cryo-EM studies of acid-treated SFV that the activation process involves the release of heterodimer interactions around the 5- and 2-fold axes and a concomitant reciprocal displacement of the E1 and E2 subunits, which could explain how the fusion peptide is exposed and the E1 trimer is formed (14). The most significant finding was that the transmembrane (TM) parts of the heterodimers rearranged, from their original positions as parts of pentamers and hexamers directly above the C-protein capsomers around the 5- and 2-fold axes, into trimers at the q3- and 3-fold axes. This suggested that interactions of the E1 and E2 TM segments might regulate the assembly and fusion reactions of the alphavirus membrane. Indeed, cryo-EM studies of SFV and, recently, in Sindbis virus have shown that the two TM segments penetrate the lipid bilayer together, at an angle to each other (29, 53). Furthermore, genetic studies in which the adaptation of significantly growth-retarded E1-E2 alphavirus chimeras was followed suggested that mutations in regions flanking the TM segment on the external side of the membrane and in the TM segment itself of either E1 or E2 were important in improving the growth of the chimeras (18, 38, 52). In the present work, we provide additional evidence for TM interactions in the alphavirus. We observed a pattern of conserved glycine residues in the external part of the TM segment of E1 and tested its role in budding and entry by mutagenesis of SFV. Two mutants were made: SFV/E1^11L, where a segment around the conserved glycines was replaced by leucine residues, and SFV/E1^4L, where only the glycines were replaced. Phenotype analyses showed that both mutations concomitantly decreased the heterodimer stability and increased their ability to trimerize at a pH that is suboptimal for wild-type fusion. Surprisingly, fusion activity was not increased but rather decreased for both mutants, suggesting that incorrect or uncoordinated structural alterations occurred during fusion-activating conditions. Furthermore, SFV/E1^11L was completely deficient in budding. The reason for this remains unclear, though it may result from poor positioning of the NC-interacting E2 tails and thus from an inability to form multivalent heterodimer-NC interactions.

MATERIALS AND METHODS

Cell culture, virus, and antibodies.

BHK-21 cells (catalog no. CCL-10; American Type Culture Collection, Rockville, Md.) were grown as described previously (42). The monoclonal antibodies (MAb) UM 1.13 (37) and UM 8.139 (1) were used as ascites preparations. Polyclonal rabbit anti-E2 antibody serum was prepared in the lab (J. Wahlberg, unpublished data).

Plasmid constructs.

The plasmid pSP6-SFV4 (24) was used as the source for wild-type SFV (SFVwt). The mutant pSFV/E1^4L, where the conserved glycines (G415, G416, G418, and G423) in the TM region of E1 were changed into leucines, was constructed by a three-fragment ligation containing the SpeI-BssHII fragment of pSP6-SFV4, the BssHII-NdeI fragment of pSFV-Helper1 (24), and an NdeI-SpeI fragment of the fusion PCR product. The fusion PCR used pSFV-Helper1 as the template, the amplification primers 5′ GAGCCCCCGAAAGACCACAT 3′ (the 5′-end primer) and 5′ GAGCGAGGAAGCGGAAGAGC 3′ (the 3′-end primer), and the fusion primers 5′ CTTCTGTTGGCCTTCGCAATCCTCGCTATCCTGGTGCTG 3′ (the 5′-end primer) and 5′ ATTGCGAAGGCCAACAGAAGAAGCGAGATTTTCTGCACC 3′ (the 3′-end primer). To construct the pSFV/E1^11L plasmid, where all 11 nonleucine amino acids from Ile413 to Ala424 in the E1 protein were changed into leucine, the mutations were first introduced in pSFV-Helper1 by site-directed mutagenesis as described previously (7) with the primer 5′ CCGCACTATCATGGGTGCAGAAACTCCTGCTTCTTCTGTTGCTCCTCTTACTCCTCCTTATCCTGGTGCTGGTTGTGGTC 3′. pSFV/E1^11L was then assembled by using the SpeI-BssHII fragment of pSP6-SFV4, the BssHII-NdeI fragment of pSFV-Helper1 as above, and the NdeI-SpeI fragment of the mutated pSFV-Helper1. The complete subgenomic regions of both plasmids, which encode the virus structural proteins, were sequenced.

Infection, transfection, and metabolic labeling.

Cells were infected (16, 35) or transfected by electroporation (41) of in vitro-transcribed RNA (25). Cells were labeled with [³⁵S]methionine essentially as described previously (42). In short, cells were depleted of methionine (starved) for 30 min by incubation in methionine-free medium, labeled for the indicated time with [³⁵S]methionine (50 μCi per 3.5-cm-diameter dish; Amersham Biosciences AB, Uppsala, Sweden), and incubated in an excess (1 mM) of unlabeled methionine (chase) for the indicated time. The chase media were clarified by low-speed centrifugation (5 min at a maximum relative centrifugal force [RCF_max] of 2,600 × g in a model 16 F24-11 rotor [Eppendorf AG, Hamburg, Germany]), and the virus was collected by centrifugation (1.5 h at an RCF_max of 37,000 × g in a model JA 18.1 instrument [Beckman Coulter AB, Bromma, Sweden]). The cells were solubilized in 10 mg of sodium dodecyl sulfate (SDS)/ml or 10 mg of Triton X-100 (TX-100)/ml in 50 mM Tris-0.15 M NaCl-2 mM EDTA-0.2 μM phenylmethylsulfonyl fluoride-0.1 μg of N-ethylmaleimide/ml-0.1 μg of pepstatin A/ml, pH 7.6 (lysis buffer). Aliquots of the cell extracts and media pellets were analyzed by SDS-polyacrylamide gel electrophoresis (PAGE) under nonreducing conditions. Labeling with [³H]palmitic acid was done by an equivalent method (34) where starvation was omitted. The labeling was for 30 min in minimal essential medium (MEM) supplemented with 50 μCi of [9,10-³H(N)]palmitic acid (NET 043; PerkinElmer, Boston, Mass.)/ml followed by 2 h in isotope-free medium (chase).

Purification of virus.

[³⁵S]Methionine-labeled SFV collected in chase medium (30-min labeling starting at 6.5 h posttransfection; 3-h chase) was sedimented through a cushion of 5% (wt/vol) iodixanol (Optiprep; Axis-Shields Pol, Oslo, Norway) in 50 mM Tris-100 mM NaCl-0.5 mM EDTA, pH 7.4 (TNE), by centrifugation for 1 h at an RCF_max of 1.6 × 10⁵ × g in a Beckman SW40 rotor at 5°C. The pelleted virus was soaked in TNE overnight, resuspended, and sedimented in a linear 5 to 30% iodixanol gradient in TNE for 1.5 h at an RCF_max of 2.2 × 10⁵ × g in a Beckman SW41 rotor at 5°C and fractionated from the top. Peak fractions were identified by liquid scintillation counting and pooled.

Immunoprecipitation.

Aliquots of iodixanol-purified [³⁵S]SFV or SDS extracts from SFV-expressing cells were mixed with 10 volumes of 10-mg/ml TX-100 in lysis buffer and immunoprecipitated with the MAb UM 1.13 or UM 8.139 or a polyclonal antibody (PAb) against E2 as described previously (48), omitting the precleaning step. To precipitate intact virus (in the absence of detergent) and avoid unspecific binding, 1 μl of antibody was mixed with 50 μl of protein A-Sepharose slurry (20% [vol/vol] in 10 mM Tris, pH 7.5) and rotated for 3 h at 5°C. Then, 70 μl of [³⁵S]SFV diluted in lysis buffer without detergent and 50 μl of fetal calf serum were added and the samples were rotated for 16 h at 5°C, washed four times in 0.5 ml of 10 mM Tris, pH 7.5, and prepared for SDS-PAGE.

Immunofluorescence.

Cells grown on coverslips were fixed in ice-cold methanol and prepared for indirect immunofluorescence as described previously (31) except that 5% fetal calf serum was used to block unspecific binding and that coverslips were mounted in Fluorsafe reagent (Calbiochem-Novbiochem Corp., La Jolla, Calif.). For surface labeling, the cells were first incubated with primary antibody at 0°C for 20 min, washed three times for 5 min each at 0°C, and then fixed with methanol. Tetramethyl rhodamine isothiocyanate (TRITC)-conjugated goat anti-mouse IgG (catalog no. 1030-03; Southern Biotechnology, Birmingham, Ala.) or TRITC-conjugated goat anti-rabbit IgG (catalogue no. 31670; Pierce Biotechnology, Rockford, Ill.) was used as secondary antibody.

Sucrose density gradients.

Transfected and [³⁵S]methionine-labeled BHK-21 cells (10-min pulse starting at 6.5 h posttransfection; 1.5-h chase) were treated for 10 min in either MEM (pH 7.4) or MEM supplemented with 20 mM bis(2-hydroxyethyl)imino-tris(hydroxymethyl)methane (Bis-Tris; final pH 6.4) and then dissolved in 10 mg of TX-100/ml in lysis buffer. Cell nuclei were removed by low-speed centrifugation, and the TX-100 extracts were separated on linear 5 to 20% (wt/wt) sucrose gradients in 30 mM Tris-100 mM NaCl-1.25 mM EDTA-1 mg of TX-100/ml, pH 7.4, in a Beckman SW50.1 rotor at an RCF_max of 2.2 × 10⁵ × g for 14 h at 5°C. The gradients were fractionated from the top and analyzed by SDS-PAGE and phosphorimaging.

Titration of virus and sequencing.

Virus titers were determined by plaque assay on BHK-21 cells as described previously (35). To sequence revertant viruses, total RNA was purified from infected BHK-21 cells by using TRIzol reagent (catalog no. 15596-026; Gibco/Invitrogen, Carlsbad, Calif.) and used as the template for cDNA synthesis and fragment amplification with the ProSTAR ultra high-fidelity reverse transcriptase-PCR system (catalog no. 600166; Stratagene, La Jolla, Calif.).

Transmission electron microscopy.

Cells were fixed in 2% glutaraldehyde in 0.1 M sodium cacodylate buffer containing 0.1 M sucrose and 3 mM CaCl₂, pH 7.4; postfixed in 2% osmium tetraoxide in 0.07 M sodium cacodylate buffer containing 1.5 mM CaCl₂, pH 7.4; and embedded in LX-112 resin (Ladd Research, Williston, Vt.), and sections were contrasted with uranyl acetate followed by lead citrate as described previously (42) and examined with a Leo 906 (Zeiss, Oberkochen, Germany) transmission electron microscope at 80 kV.

pH-induced trimerization.

Transfected and [³⁵S]methionine-labeled BHK-21 cells (10-min pulse starting at 6.5 h posttransfection; 1.5-h chase) grown on 24-well plates were washed once and then incubated either in 20 mM Bis-Tris-0.15 M NaCl, pH 7.0, 6.8, 6.6, 6.4, 6.2, or 6.0, or in 20 mM sodium succinate in MEM, final pH 5.8 or 5.6, for 10 min at room temperature and solubilized in 10 mg of TX-100/ml in lysis buffer. The extracts were cleared by low-speed centrifugation, and aliquots were prepared for SDS-PAGE. Parallel samples were heated to either 37 or 95°C prior to electrophoresis. The amount of E1 present as monomer and as SDS-resistant, temperature-labile trimer was measured, and the relative amount of E1 in trimer form was calculated. The relative trimerization of wild-type E1 at pH 5.6 was defined to be 1.0 and used to normalize the figures. The maximal amount of E1 trimer in SFVwt-transfected cells varied between 20 and 35% of the total amount of E1 protein in the different experiments.

Cell-cell fusion assay.

BHK-21 cells were transfected and plated on coverslips in 24-well plates. At 2 h posttransfection, the growth media were replaced with a suspension of untransfected BHK-21 cells (indicator cells) in complete BHK medium. At 8 h posttransfection, trimerization was induced as described above. After the pH treatment, the solutions were replaced by complete BHK medium and the cells were grown at 37°C in 5% CO₂ for 90 min, fixed in ice-cold methanol, and stained by indirect immunofluorescence with PAb rabbit anti-E2 as the primary antibody. The number of stained cells (n_cell) and nuclei (n_nuclei) in stained cells was counted, and the fusion index (F_i) was calculated (F_i = 1 − n_cell/n_nuclei) (51).

SDS-PAGE.

Protein samples were prepared as described previously (40), except that 63 mM sodium phosphate, pH 7.0, was used as the buffer (3) and 0.8 mM methionine was included, and samples were separated on SDS-10% PAGE gels (20). Samples of pelleted or purified virus were supplemented with 1 μl of BHK cell extract (≈10⁷ BHK cells solubilized in 1 ml of 10-mg/ml TX-100/ml in lysis buffer; nuclei were removed by low-speed centrifugation) per 10 μl of sample buffer. After electrophoresis, the gels were soaked in 1 M sodium salicylate, dried, and exposed to Fuji X-ray films. Quantitation was done on a Bas 2000 phosphorimager using Image Gauge version 3.3 software (Fujifilm Sverige AB, Stockholm, Sweden). ¹⁴C-Methylated molecular weight standards (CFA 626) were from Amersham Biosciences AB.

RESULTS

Construction of mutants and their initial characterization.

As a way to locate putative interaction regions in the TM segments of the E1 and E2 membrane proteins, we screened these segments for conserved features. Interestingly, E1 had a pair of glycine residues in the external part of its TM segment (positions 415 and 416 in SFV) that were conserved in most alphaviruses (Table 1). The exceptions were Sleeping Disease virus, Barmah Forest virus, and Ross River virus, which had only one of the two glycines conserved. Further, most alphaviruses appeared to have at least one additional glycine positioned 2 to 8 residues further inwards, i.e., toward the C terminus, in the E1 TM segment (positions 418 and 423 in SFV). In contrast, E2 showed no obvious conservation apart from the general hydrophobicity expected in a membrane anchor. To investigate whether the conserved glycines in the TM segment of E1 mediated any important interactions, we constructed two mutants: first, a general one where all 11 nonleucine amino acids from Ile413 to Ala424 in E1, including the conserved glycine residues, were changed into leucine (SFV/E1^11L [Fig. 1 ]), and then, a specific one where only the conserved glycines (G415L, G416L, G418L, and G423L) were changed into leucines (SFV/E1^4L [Fig. 1]). We first analyzed whether the TM mutations affected virus assembly. To this end, we transcribed the plasmids containing the mutated genomes into replication-competent RNA, transfected BHK-21 cells, and followed the viral proteins by pulse-chase analyses. Figure 1 shows that both wild-type and mutant genomes directed the synthesis of C, E1, the E2 precursor p62, and a 107-kDa protein (lane 1) which corresponds to an uncleaved and unglycosylated p62-6K-E1 polyprotein that has failed to become translocated in the endoplasmic reticulum (ER) (12). With time, the 107-kDa band disappeared, probably by degradation, and the p62 protein was cleaved into E2 (lanes 2 and 3). The wild-type-like migration of the membrane proteins of both mutants suggested that their processing was correct and thus that they were inserted normally into the ER membrane despite the E1 TM mutations and then further glycosylated and oligomerized into transport-competent heterodimers as in the wild type. Analysis of the chase media (lanes 4 through 6) showed that the SFV/E1^4L virus was released at a slightly lower rate than the wild type. The virus production was found to be 62% ± 14% (n = 5) of that of SFVwt production. A similar production deficiency, 60% ± 13% (n = 4), was found also in mosquito cells (C6/36), which is another natural host of SFV. In contrast, no virus was seen in the media of SFV/E1^11L-transfected cells (panel C), only an E1 fragment that had earlier been characterized and is thought to reflect the proteolytic release of E1 ectodomains from the cell surface (54). Reducing the growth temperature to 28°C did not change the release phenotype of SFV/E1^11L, and we conclude that this mutant is severely deficient in budding. To test the overall replicative function of the SFV/E1^4L virus, we measured its specific infectivity. To this end, [³⁵S]methionine-labeled virus was purified on a linear iodixanol density gradient and then analyzed for infectivity and labeled viral protein by plaque titration and SDS-PAGE, respectively. The specific infectivity of the SFV/E1^4L mutant was calculated and found to be 26% ± 3% (n = 4) of that of the wild type. This suggests that there is a major defect in the entry functions of the SFV/E1^4L mutant, which is in contrast to SFV/E1^11L, where the major defect was clearly in budding.

TABLE 1.

Comparison of the E1 TM segment and flanking amino acids (408 through 438 in SFV) in different alphaviruses (30)

Virusa	Sequence of domain regionc		GenBank accession no.
	ExternalTM	Cytoplasm
	408b    413        420                             430	437
SFV	...S W V Q KI S G G L G A F A I G A I L V L V V V T C I G L	R R	CAA01423
SDV	...R W A G RI V G N P S G P V S S S L A V T Y C V V	K K C...	NC_003433
SPDV	...G P A M RW A G G I V G T L V V L F L I L A V I Y C V V	K K C...	SPA012631
BFV	...Q W L A HT T S G P L T I L V V A I I V V V V V S I V V C A	R H	U73745
ONNV	...S W V Q KI T G G V G L V V A I A A L I L I I V L C V S F S	RH	AF079456
CHIKV	...S W V Q KI T G G V G L V V A V A A L I L I V V L C V C		L37661
RRV	...T W V Q RM A S G L G G L A L I A V V V L V L V T C I T M	R R	NC_001544
AURAV	...N W I T AL M G G I S S I A A I A A I V L V I A L V F T A	Q H R	NC_003900
SINV	...S W L F AL F G G A S S L L I I G L M I F A C S M M L T S T	R R	NC_001547
HJV	...N W L F AM L G G A S S L I V V G L L V L A C S S M I I N T	R R	U52586
WEE	...N W L L AL F G G A S S L I V V G L I V L V C S S M L I N T	R R	J03854
EEE	...S W L K VL V G G T S A F I V L G L I A T A V V A L V L F F	H R H	L20951
VEE	...T W L T SL L G G S A V I I I I G L V L A T I V A M Y V L T	N Q K H N	L01442

^aAbbreviations: SDV, sleeping disease virus; SPDV, salmon pancreas disease virus; BFV, Barmah Forest virus; ONNV, O'Nyong Nyong virus; CHIKV, Chikungunya virus; RRV, Ross River virus; AURAV, Aura virus; SINV, Sindbis virus; HJV, Highlands J virus; WEE, western equine encephalitis virus; EEE, eastern equine encephalitis virus; VEE, Venezuelan equine encephalitis virus.

^bThe numbering refers to the position in the SFV sequence.

^cBoldface indicates the conserved glycines.

FIG. 1.

C-terminal sequence of the E1 protein and pulse-chase analysis. (Top) Sequence of the E1 protein (amino acids 408 through 438) of SFVwt (A), SFV/E1^4L (B), and SFV/E1^11L (C), with the TM region underlined. (Bottom) BHK-21 cells transfected with in vitro-transcribed RNA expressing SFVwt (panel A), SFV/E1^4L (panel B), or SFV/E1^11L (panel C) were labeled with [³⁵S]methionine for 10 min and chased for either 5 min (lanes 1 and 4), 60 min (lanes 2 and 5), or 120 min (lanes 3 and 6). Detergent extract corresponding to 0.73% of the cells (lanes 1 through 3) and pelleted particles from 18% of the culture media (lanes 4 through 6) in each dish were analyzed by SDS-PAGE under nonreducing conditions. The migration of molecular weight standard proteins is indicated to the left.

Stability of the E1-E2 heterodimer.

In order to directly assess the effect of the TM mutations on heterodimer stability, we measured the sensitivity of the heterodimers towards acid-induced dissociation. Earlier studies have demonstrated that E1-E2 heterodimers become efficiently dissociated at the optimal pH for membrane fusion (pH 5.6) (48). To enhance any changes in heterodimer stability, we made the analyses at pH 6.4, a mildly acidic pH that should dissociate only a fraction of the wild-type heterodimers. Transfected cells that had been labeled with [³⁵S]methionine and chased for 1.5 h to allow maximal accumulation of labeled E1-E2 heterodimers at the PM were treated at either neutral-pH or pH 6.4 conditions for 10 min, solubilized, and sedimented in a sucrose gradient containing TX-100. We followed the heterodimer dissociation by monitoring the E2 subunit. The analyses showed that the wild-type heterodimer was marginally dissociated at a neutral pH and slightly more at pH 6.4 (Fig. 2). The amounts of E2 monomer in the wild-type-transfected cells were 5.6% ± 1.1% (n = 4) and 9.0% ± 2.0% (n = 4) of total E2 in the gradient at neutral-pH and pH 6.4 conditions, respectively. The heterodimer of SFV/E1^4L was also largely stable at a neutral pH but significantly more dissociated at pH 6.4: 1.4 ± 0.14 (n = 3) times that in wild-type-transfected cells at the same pH conditions. The heterodimer of SFV/E1^11L, on the other hand, was unstable already at a neutral pH and the most dissociated at pH 6.4 conditions. Here, the E2 monomers amounted to about two times that in wild-type-transfected cells at the same pH condition: 1.9 ± 0.28 (n = 4) times under neutral-pH, and 2.2 ± 0.27 (n = 4) times at pH 6.4, conditions. This clearly shows that both TM mutations decreased the stability of the E1-E2 heterodimer.

FIG. 2.

Oligomeric state of the E2 protein in TX-100 extract of cells. BHK-21 cells were transfected with SFVwt (solid line), SFV/E1^4L (broken line), or SFV/E1^11L (dotted line) RNA, labeled, and chased for 1.5 h as described in the legend to Fig. 1 and treated for 10 min at pH 7.4 (A) or pH 6.4 (B). TX-100 extracts of the cells were separated on a linear sucrose gradient containing 1 mg of TX-100/ml. The gradients were fractionated from the top and analyzed by SDS-PAGE and phosphorimaging. Shown is the relative amount of E2 in each fraction as the arithmetic mean value from three (SFV/E1^4L) or four (SFVwt and SFV/E1^11L) separate experiments. The size of the wild-type monomer fractions and the relative increase in the monomer fractions of the mutants ± 1 standard deviation are given in Results.

Entry functions of the SFV/E1^11L and SFV/E1^4L mutants.

The lower stability found in the heterodimers of the mutants could have an effect on the subsequent acid-induced trimerization of E1 and thereby also on the E1 fusion function. This was also tested using transfected cells. We first measured the ability of E1 to form homotrimers in response to treatment with low pH. To this end, transfected cells were labeled with [³⁵S]methionine and chased for 1.5 h as above and then treated at increasingly acidic conditions, solubilized, and analyzed by SDS-PAGE under nonreducing conditions. Parallel aliquots were treated either at 37°C to preserve any E1 trimers or at 95°C to dissociate all membrane protein oligomers into monomers prior to electrophoresis. The relative trimerization was calculated as a fraction of the wild-type E1 trimer at pH 5.6 and plotted as a function of pH (Fig. 3A). The results showed that the wild-type E1 trimerized upon treatment with a pH in the range of 6.4 to 5.6 and with increasing efficiency as the pH was decreased. Surprisingly, both the E1^11L and E1^4L variants showed increased sensitivity to trimerize at an intermediate low pH. Already at pH 6.2 was the trimerization of the mutants maximal, whereas the wild type needed a pH of 5.8 or 5.6 to reach the same level of trimerization. The final amount of the E1 trimer was similar in the mutants and the wild type. This suggested that the fusion function of the mutants would be easier to activate by acid conditions. Nevertheless, when the capacity to induce cell-cell fusion in response to increasingly acidic conditions was tested, we found that both mutants were significantly less efficient than the wild type (Fig. 3B). Still, the proteins of both mutants and of the wild type induced cell-cell fusion in response to pH treatment in the expected range of pH 6.4 to 5.6 (22). This shows that the increased sensitivity to trimerize by low pH that was seen in the TM mutants did not correlate with an increased fusion capacity.

FIG. 3.

Acid-induced E1 trimerization (A) and cell-cell fusion (B). (A) BHK-21 cells transfected with SFVwt (solid line), SFV/E1^4L (broken line), or SFV/E1^11L (dotted line) RNA were labeled with [³⁵S]methionine at 6.5 h posttransfection, chased for 1.5 h, treated at the indicated pH, and solubilized. Parallel samples were treated at either 37 or 95°C and analyzed by SDS-PAGE under nonreducing conditions. The amount of temperature-labile E1 trimer was measured, and the relative amount was calculated. Shown is the arithmetic mean value (n = 4) ± 1 standard deviation (error bars). (B) Aliquots of the transfected cells in panel A were grown on coverslips with indicator cells, treated at the indicated pH, shifted back to a neutral pH, grown for 90 min, and fixed. The extent of fusion (F_i) was calculated and plotted as a function of pH. Shown is the arithmetic mean (n = 4) ± 1 standard deviation (error bars).

Transport and maturation of the SFV/E1^11L membrane proteins.

To investigate the basis of the assembly defect of SFV/E1^11L, we measured the cell surface presentation and maturation of its membrane proteins by surface-specific immunostaining and biotinylation. For immunostaining, a PAb against the E2 subunit was bound either to living cells on ice or to fixed and permeabilized cells. Figure 4 shows that the E2 of SFV/E1^11L (panel A), like that of SFVwt (panel B), was present on the surface. The internal staining pattern in both SFV/E1^11L- and SFVwt-transfected cells (panels C and D, respectively) with a strong perinuclear signal and dark nuclei showed that the membranes of the surface-stained cells were intact. Nontransfected cells that were surface stained in parallel gave no signal, showing that the temperature had been low enough to stop the unspecific endocytosis of the primary antibody (data not shown). In the biotinylation assay, we used transfected cells that had been pulse-labeled and -chased for 5 or 30 min. At the shorter chase time, labeled wild-type heterodimers should not have yet reached the PM, while at the longer chase time, their PM appearance should have just become evident. Both E2 and E1 of SFV/E1^11L were found to be specifically biotinylated after 30 min of chase (Fig. 5). The efficiency was calculated to about 0.6 times that of the wild-type proteins, suggesting some deficiency in the routing of the mutant heterodimer. There was no biotinylation of E1 or E2 at the 5-min chase (lanes 5 and 6) nor of the cytoplasmic C protein at either time point (lanes 5 through 8), showing the specificity of the assay.

FIG. 4.

Immunofluorescence analysis of the surface expression of E2. BHK-21 cells were transfected with SFV/E1^11L (A and C) or SFVwt (B and D) RNA. At 6 h posttransfection, cells were either chilled to 0°C (A and B) or fixed and permeabilized (C and D) prior to binding of a PAb against E2 and then prepared for immunofluorescence by using a rhodamine-labeled secondary antibody.

FIG. 5.

Surface biotinylation of SFV-expressing cells. Proteins in BHK-21 cells transfected with SFVwt (odd lanes) or SFV/E1^11L (even lanes) RNA were labeled with [³⁵S]methionine for 10 min and chased for 5 min (lanes 1, 2, 5, and 6) or 30 min (lanes 3, 4, 7, and 8), chilled to 0°C, and biotinylated. Total TX-100 extracts of the cells (lanes 1 through 4) and the biotinylated proteins captured from the TX-100 extracts (lanes 5 through 8) were analyzed by SDS-PAGE under nonreducing conditions. Shown is a fluorogram of the resulting gel. Molecular weights are shown to the left.

Spike formation of SFV/E1^11L.

To test whether the TM mutations affected the capacity of the heterodimers to trimerize into spikes and/or higher-order structures like spike hexamers or even a spike lattice, we used the MAb UM 1.13. UM 1.13 is a neutralizing, E2-specific MAb that detects its cognate epitope in intact particles. In Fig. 6A, we show that UM 1.13 readily binds to intact virus (lane 1) but is completely unreactive for TX-100-solubilized heterodimers (lane 2). In parallel, a control antibody, UM 8.139, recognized its epitope both before and after solubilization of the virus (lanes 3 and 4). This suggests that the UM 1.13 epitope is maintained only in the glycoprotein shell and that the MAb can be used to assay for spike or spike lattice formation in the cell. When we compared the UM 1.13 staining of SFV/E1^11L- and SFVwt-expressing cells (Fig. 6B and C, respectively), we found similar staining patterns, consisting of dots and short rods, that were partially on the surface but also inside the cell, with the highest concentration in the perinuclear area. This indicates that correct spikes or possibly spike lattices are formed by the SFV/E1^11L mutant.

FIG. 6.

Presence of the UM 1.13 epitope in SFV-expressing cells. (A) Antibody specificity. [³⁵S]Methionine-labeled SFVwt was immunoprecipitated with the MAb UM 1.13 (lanes 1 and 2) or UM 8.139 (lanes 3 and 4) in the absence (lanes 1 and 3) or presence (lanes 2 and 4) of TX-100 and analyzed by SDS-PAGE under nonreducing conditions. (B and C) Immunofluorescence analysis. BHK-21 cells transfected with SFV/E1^11L (B) or SFVwt (C) RNA were fixed with ice-cold methanol and reacted with UM 1.13, which was visualized by a rhodamine-labeled secondary antibody.

Ultrastructural analysis of the SFV mutants.

Electron micrographs of cells transfected with SFV/E1^11L RNA (Fig. 7A and B) showed abundant NCs (arrowheads) in the cytoplasm. The PM was traced in sections of 30 different SFV/E1^11L-transfected cells that all contained cytoplasmic NCs and either cytopathic vacuole 1 (CPV-1) or CPV-2 (see below), but PM-bound NCs, budding profiles, or extracellular mature virus could not be observed. These features are typical of wild-type-transfected cells (panel D) and their absence therefore suggests that stable heterodimer-NC interactions are not formed at the PMs of cells expressing the SFV/E1^11L mutant. We also screened the cells for the appearance of CPV-2 structures. An alphavirus infection results in at least two types of CPVs in the cell: CPV-1 and CPV-2. The former appears as large vacuoles with spherical, 100-nm-diameter invaginations. These have been shown to consist of endosomal and/or lysosomal membranes converted into centers of viral replication and are established early in infection (10). CPV-2s appear mostly as tubular structures of membranes that are studded with dense, 30-nm-diameter spheres on the surface. They form at a later time and are most likely Golgi-derived membranes with NCs bound to their cytoplasmic face (13), possibly via the E2 tail. At 12 h posttransfection with SFV/E1^11L, NCs in CPV-2 structures were readily detected in the cells (panel B). This suggests that the heterodimer was able to expose a binding site for the NCs, although this was not productive in virus budding at the PM. Figure 7 also shows the wild-type-like budding of SFV/E1^4L (panel C) with complete virus particles (arrow) on the surface and NCs (arrowheads) both in the cytoplasm and close to the PM.

FIG. 7.

Ultrastructure of SFV-expressing cells. BHK-21 cells transfected with SFV/E1^11L (panels A and B), SFV/E1^4L (panel C), or SFVwt (panel D) RNA were fixed at 8 h (panels A, C, and D) or 12 h (panel B) posttransfection and prepared for electron microscopy. Complete virus (arrows), NC structures (arrowheads), and CPVs (CPV-1 and CPV-2) are indicated.

Cytoplasmic exposure of the E2 tail.

The NC-heterodimer contact in an alphavirus is mediated by the internal, C-terminal tail of the E2 protein. The morphological analysis suggests that this interaction was defective in the SFV/E1^11L mutant. One possibility is that the tail exposure on the cytoplasmic side had been incorrect during the biosynthesis of the mutant heterodimer. During the initial synthesis of the C-p62-6K-E1 polyprotein, the E2 tail functions as a signal peptide for the 6K peptide. The E2 tail is released by signal peptidase cleavage on the luminal side of the ER membrane and translocated back to the cytoplasmic side, where it is acylated, predominantly by palmitic acid (17, 23, 45). The E2 protein has three acylation sites in the E2 tail, all of which are probably used (15), and one potential site in the TM region (Cys385). In case the E2 tail fails to translocate, the three sites in the tail will probably not be acylated. Thus, in one possible scenario, the TM mutations would disturb the tail translocation back to the cytoplasm and hence significantly reduce or possibly completely abolish the E2 acylation. To test whether the TM mutations affected acylation of the E2 tail, parallel samples of transfected cells were labeled with either [³H]palmitic acid or [³⁵S]methionine and analyzed by SDS-PAGE and phosphorimaging. Figure 8 shows that E2 proteins from both mutants were ³H labeled approximately as efficiently as the wild type. The relative acylation of E2 was found to be three times that of E1 in the wild type and in both mutants. This is consistent with the acylation of all potential sites in both E2 and E1. The lack of ³H labeling (acylation) in the nontranslocated 107-kDa band (lanes 4 through 6) shows that the labeling was specific. This suggests that the E2 tail is positioned on the cytoplasmic side of the membrane.

FIG. 8.

Acylation of spike proteins. Parallel samples of BHK-21 cells transfected with SFVwt (lanes 1 and 4), SFV/E1^4L (lanes 2 and 5), or SFV/E1^11L (lanes 3 and 6) RNA were labeled with either [³⁵S]methionine (lanes 1 through 3) or [³H]palmitic acid (lanes 4 through 6) for 30 min and chased for 1 h. SFV spike proteins were recovered from SDS extracts of the cells by immunoprecipitation with a combination of PAb against E1 and E2 and analyzed by SDS-PAGE under nonreducing conditions. Molecular weights are shown to the left.

Revertants.

To find out whether the SFV/E1^11L reverted into budding-competent mutants, growth media from cells transfected with SFV/E1^11L were analyzed by plaque titration. We found that the media contained infectious virus that gave rise to plaques that were smaller (about 30% of the wild-type diameter) and much fewer in quantity (about 10⁵ to 10⁶ times fewer) than the wild type. The viruses from 30 small plaques were purified by two consecutive rounds of plaque formation. They were then used to infect fresh BHK-21 cells, and their phenotype was analyzed by pulse-chase analysis. All of the plaque-isolated viruses were found to have a partially restored budding capacity (about 5 to 10% of the wild-type level). Sequencing of the TM segments of both E1 and E2 in 12 of these revertants showed that all had a single change in the 12-leucine stretch of the mutant E1: either L424P (11 revertants) or L417P (one revertant, R9). Figure 9 summarizes the analysis of the phenotype of one L424P revertant (R20). Pulse-chase analyses showed that protein synthesis and trimming were wild type-like but that particle release was reduced (panel A). Electron microscopy (panels B and C) showed NCs (arrowheads) both in the cytoplasm and lined up under the PM. In some areas, wild-type-like budding was seen (panel B, arrow), but defective budding, resulting in multicore particles, was also common (panel C). This suggests that heterodimer-NC interaction was partially restored in the revertant.

FIG. 9.

Phenotype of SFV/E1^11L revertants. (Top) Sequence of the E1 protein (amino acids 408 through 438) of revertants R9 and R20, with the TM region underlined. (Bottom) Pulse-chase analysis (A) and ultrastructure at 8 h postinfection (B) of BHK-21 cells infected with SFV/E1^11LR20 as detailed in the legends to Fig. 1 and 7, respectively, are shown. Defective budding, resulting in multicore particles, also occurred (C).

DISCUSSION

Apart from their role in assembly interactions, it is well established that heterodimeric interactions between the E1 and E2 subunits contribute a major regulatory mechanism for the viral fusion activity. In the p62-E1 precursor of the heterodimer, the fusion activity of E1 is suppressed by a very stable subunit interaction (48). This facilitates safe routing of the heterodimer from the ER through the mildly acidic compartments of the secretory pathway to the PM, where virus budding occurs (6). The cleavage of p62 into E2 potentiates the fusion activity by increasing the acid sensitivity of the complex towards dissociation (6, 31, 33). In the endosomes, the acid pH will finally activate the fusion function of E1. This occurs following the structural alterations initiated by heterodimer dissociation and ending at E1 trimerization (2, 49). The E1 trimers probably represent the inactive end products, and the fusion-active form is most likely an intermediate structure. The subunit interactions in the E1-E2 heterodimer have so far not been characterized in any greater detail. Cryo-EM and genetic studies suggest ectodomain intersubunit interactions that involve the fusion peptide of E1 (8), the “pre” part E3 of the p62 precursor (27), and the region around amino acid 4 of the SFV E2 protein (44) and around amino acids 129 through 248 in Sindbis virus E2 (18, 52). Biochemical studies of proteolytically released ectodomains of solubilized E1-E2 heterodimers and acid-induced E1 trimers unequivocally demonstrate that subunit interactions responsible for heterodimer interactions reside in the ectodomain (19, 50). In this study, we have used a genetic approach to test the TM interaction of membrane protein subunits of SFV in assembly and entry functions. Such interactions have been implicated by cryo-EM studies (14, 29, 53) and also by adaptation studies of Sindbis virus-Ross River virus chimeras (18, 38, 52). We scanned the sequence for conserved amino acids, found a pair of glycines in the TM segment of E1, and made two mutants, SFV/E1^4L and SFV/E1^11L, to investigate their impact. Both mutants showed a weaker interaction between E1 and E2 in the heterodimer that was coupled to an increased sensitivity to acid-induced reorganization into E1 homotrimers. However, the fusion capacity of the mutant heterodimers was decreased, suggesting that the formation of fusion-active intermediates was poorer in these mutants than in the wild type. One possibility is that both the TM and ectodomain interactions have to be released by low pH in a concerted fashion and that this does not occur in the mutants. If the TM interaction is dissociated too early, this may lead to the formation of incorrect intermediate structures that are likely to have poor fusion function. The second effect of our TM mutations was reduced budding. In this respect, the SFV/E1^11L was severely inhibited while the SFV/E1^4L showed only a slight reduction. The TM segments of E1 and E2 traverse the membrane close together (29, 53). Under the membrane, the heterodimer-NC interaction is established by a tyrosine motif in the E2 tail that binds to a pocket in the C protein (36). The icosahedral symmetry, which is equal in both the spike protein lattice and the NC, allows a multivalent binding that favors budding (5, 9). A schematic drawing of this interaction is shown in Fig. 10A. In the SFV/E1^4L mutant, the weak heterodimer interaction may change the relative position, or just allow a certain movement, of the E2 tail, which may make multivalent NC interactions more difficult to obtain and thus reduce budding (Fig. 10B). Note that the SFV/E1^4L mutant still posseses 4 amino acids with small side chains (serine and alanine) in the external half of its TM segment (Table 1), making a slight curvature of the helix likely. In contrast to the other mutant, the SFV/E1^11L was unable to form stable spike-NC interactions and did not bud at the PM (Fig. 7) although the presentation of its NC-interacting E2 tail appeared to be largely correct (Fig. 8) and a wild-type-like spike lattice appeared to form (Fig. 6). A possible reason for this inability would be that the polyleucine stretch in SFV/E1^11L forms a too-stable α-helix that cannot bend (26). This may displace the TM helix of E1^11L and thereby allow large variations in the position of the E2 tail (Fig. 10C), which could decrease the probability of multivalent NC binding dramatically. This stiff-helix model is supported by the revertants in which a single leucine in the polyleucine stretch was replaced by a proline (Fig. 9). Such a change is known to introduce a kink (4) or even a hinge (32) in TM α-helices. Both proline and glycine residues have been found to function as hinges (32), but proline probably prevails in the revertants because a single mutation is sufficient to change a leucine into a proline while two changes are needed to produce a glycine. Further, Sleeping Disease virus and Barmah Forest virus, which both lack the glycine pair in the TM segment of E1, carry a proline in a nearby position (Table 1). Ross River virus, the third alphavirus with only one conserved glycine, has serine, an amino acid with a small side chain, in the position of the first glycine and a double glycine 2 residues further toward its C terminus (Table 1). The proline in the SFV revertants studied here may restore the bend of the TM segment of E1 and might thereby restrict the movement of the E2 tail in a way that promotes multivalent NC binding and thus budding (Fig. 10D).

FIG. 10.

Schematic drawing of the TM segments of E1 and E2 in SFVwt (A), SFV/E1^4L (B), SFV/E1^11L (C), and SFV/E1^11LR20 (D) showing their proposed conformation and interaction with the C protein. The tyrosine motif in the E2 tail is indicated with a star.