Anti-Tetherin Activities of HIV-1 Vpu and Ebola Virus Glycoprotein Do Not Involve Removal of Tetherin from Lipid Rafts

Abstract

BST-2/tetherin is an interferon-inducible host restriction factor that blocks the release of newly formed enveloped viruses. It is enriched in lipid raft membrane microdomains, which are also the sites of assembly of several enveloped viruses. Viral anti-tetherin factors, such as the HIV-1 Vpu protein, typically act by removing tetherin from the cell surface. In contrast, the Ebola virus glycoprotein (GP) is unusual in that it blocks tetherin restriction without apparently altering its cell surface localization. We explored the possibility that GP acts to exclude tetherin from the specific sites of virus assembly without overtly removing it from the cell surface and that lipid raft exclusion is the mechanism involved. However, we found that neither GP nor Vpu had any effect on tetherin's distribution within lipid raft domains. Furthermore, GP did not prevent the colocalization of tetherin and budding viral particles. Contrary to previous reports, we also found no evidence that GP is itself a raft protein. Together, our data indicate that the exclusion of tetherin from lipid rafts is not the mechanism used by either HIV-1 Vpu or Ebola virus GP to counteract tetherin restriction.

INTRODUCTION

BST-2/tetherin is an interferon-inducible antiviral factor that prevents the release of newly formed enveloped viruses from the cell surface (64, 89). Diverse families of enveloped viruses are susceptible to tetherin restriction, including the retroviruses, filoviruses, arenaviruses, and herpesviruses (43, 44, 56, 79). This suggests a nonspecific mechanism of restriction, most likely a physical tether between the cell and viral membranes (19, 25, 28, 29, 64, 74). In turn, several viruses have been found to express anti-tetherin factors that increase virus release, including the Vpu (64, 89), Env (27, 48), and Nef (42, 82, 94, 96) proteins from the primate lentiviruses, the K5 protein from Kaposi's sarcoma-associated herpesvirus (56), and the glycoprotein (GP) from Ebola virus (44). These anti-tetherin factors can also promote the release of heterologous viral particles, suggesting that they act directly on tetherin itself and not by influencing specific pathways of viral assembly and release.

Most viral anti-tetherin factors appear to counteract tetherin by removing it from the cell surface, using intracellular sequestration (18, 33, 27, 48) and/or targeting of the protein to a degradative pathway (4, 17, 25, 26, 60, 55, 56, 61). However, we have previously reported that the Ebola virus GP is able to prevent tetherin restriction without significantly removing tetherin from the cell surface (54). In addition, GP is the only anti-tetherin factor so far identified that also has activity against a wholly artificial tetherin-like molecule (art-tetherin) composed of functionally related protein domains with no sequence homology to the native protein (54, 74). Together, these findings suggest that Ebola virus GP uses a distinct mechanism to counteract tetherin restriction.

The Ebola virus GP exists in several different forms. The major gene product is a secreted, nonstructural soluble glycoprotein (sGP) (80, 93), and there are two additional secreted proteins, GP_1,2Δ (15) and ssGP (58). Transcriptional editing also generates a type I transmembrane glycoprotein that forms the trimeric spikes in the viral envelope (41, 80, 92, 93), and it is this full-length membrane-bound form of GP that counteracts tetherin restriction (44).

Several mechanisms can be proposed to account for the ability of the membrane-bound form of Ebola virus GP to block tetherin and art-tetherin restriction without removing the proteins from the cell surface. First, the expression of GP could alter some feature of the cell membrane, or its associated cytoskeletal elements, in a way that leads to loss of tetherin function. Alternatively, the presence of GP at sites of virus assembly could physically shield the viral particles from access by tetherin. Finally, GP could act to exclude tetherin from the specific sites where virus assembly occurs while still retaining the protein at the cell surface. In consideration of the last possibility, it is noteworthy that several enveloped viruses assemble at specific membrane microdomains termed lipid rafts (46) and that tetherin is also present in these domains (47, 57). Therefore, it is possible that the colocalization of tetherin and virions to lipid rafts or other membrane microdomains is necessary for tetherin restriction and that the expression of Ebola virus GP blocks this association without overt removal of tetherin from the cell surface.

Lipid rafts are small (10 to 200 nm), heterogeneous, dynamic membrane microdomains that are rich in sterols and sphingolipids (9). They are present both at the plasma membrane and at internal membranes, where they play a role in protein sorting (22). Resident raft proteins can be identified through the isolation of detergent-resistant membrane (DRM) fractions that remain intact after treatment with nonionic detergents such as Triton X-100 (TX-100) under cold conditions (8, 10). Raft proteins are often targeted to these domains by posttranslational modifications such as glycophosphatidylinositol (GPI) anchors or palmitoylation (8, 11). Tetherin's localization in DRMs is dependent on the presence of a C-terminal sequence that has been proposed as a GPI anchor (47, 57), although an alternate role as a second transmembrane domain has also been suggested (3). The C-terminal region of tetherin is also essential for its antiviral restriction activity (64, 74). Finally, both HIV-1 (35, 52, 66, 70) and Ebola virus (5, 73) have been reported to bud from lipid rafts, although this has been challenged for Ebola virus (16, 86).

In the present study, we examined the colocalization of HIV-1 virus-like particles (VLPs), tetherin, and either GP or Vpu in lipid rafts, in an attempt to understand the role that such domains could play in both tetherin restriction and its counteraction. Contrary to previous reports (5), we found no evidence that Ebola virus GP is itself a raft protein. Furthermore, although both tetherin and HIV-1 VLPs were detected in DRMs, the expression of neither GP nor Vpu removed tetherin from these microdomains. Finally, although Vpu was able to prevent the colocalization of tetherin and HIV-1 VLPs at the cell surface, the Ebola virus GP did not prevent this association. Consequently, the mechanism by which GP counteracts tetherin restriction is still unclear, but it does not involve the specific removal of tetherin from lipid rafts.

MATERIALS AND METHODS

Cell lines.

HeLa and 293T cells were obtained from the American Type Culture Collection. All cells were maintained in D10 medium (Dulbecco's modified Eagle's medium [DMEM] [Mediatech, Herndon, VA] supplemented with 10% fetal bovine serum [FBS] [Denville Scientific, Metuchen, NJ]).

Plasmids.

Plasmid pHIV-1-pack expresses HIV-1 Gag-Pol and Rev and generates HIV-1 VLPs (1). Plasmid Gag-yellow fluorescent protein (Gag-YFP) was provided by Paul Spearman (Emory University) (14). Plasmid pCAGGS-FLAG-VP40 produces Ebola virus VP40 VLPs (44) and was obtained from Paul Bates (University of Pennsylvania). Plasmid pcDNA-Vphu encodes a human codon-optimized form of HIV-1 Vpu (67), and plasmid Vphu-HcRed is a C-terminally tagged version (91). Expression plasmids for human tetherin/BST-2 (pCMV6-XL5-Bst2), an ectodomain hemagglutinin (HA) tag version (tetherin_HA), and an artificial tetherin-like protein (art-tetherin) (74) in the same plasmid backbone were previously described (54). Plasmid pEboGP expresses Ebola virus Zaire GP-8A (54), and site-directed mutagenesis was used to change residues Cys670 and Cys672 to alanines in plasmid pEboGP_C670/672A. EboGP was also recloned into plasmid pCMV5-XL5 to give pEboGP-X, and a myc tag (EQKLISEEDL) was added to the C terminus to give pEboGP_Myc.

Production and analysis of HIV-1 and Ebola virus VLPs.

HIV-1 and Ebola virus VLPs were generated in HeLa cells or 293T cells by transient transfection with plasmid pHIV-1-pack (8 μg) or FLAG-VP40 (500 ng), essentially as described previously (54) but using TurboFect transfection reagent (Thermo Scientific, Glen Burnie, MD). For HIV-1 VLP analysis, cell lysates were harvested and viral particles were collected from the supernatant at 24 h posttransfection and analyzed by Western blotting against p24, as described previously (54). For Ebola virus VLPs, cell lysates and supernatants were harvested at 48 h posttransfection, and supernatants were filtered prior to layering over a 20% sucrose cushion and ultracentrifuged at 30,000 rpm (Beckman SW41Ti rotor) for 2 h. Lysates and pelleted Ebola virus VLPs were analyzed by Western blotting using a mouse anti-FLAG antibody (Sigma-Aldrich, St. Louis, MO) at a 1:2,000 dilution, followed by horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG (1:10,000) (Sigma-Aldrich). Specific proteins were visualized using the enhanced chemiluminescence (ECL) detection system (Amersham International, Arlington Heights, IL). Exposed and developed films were scanned and quantified using the public domain NIH ImageJ software. The intensity of either p24-reacting or FLAG-reacting bands on Western blots was measured and the ratio of the signal in VLPs to that in lysates obtained. Fold enhancement of VLP release was calculated by normalizing all values to the tetherin- or art-tetherin-expressing controls. Statistical analysis was performed using one-way analysis of variance (ANOVA) followed by Dunnett's multiple-comparison test from GraphPad Prism (GraphPad Software, La Jolla, CA).

Isolation of DRMs.

HeLa or 293T cells were transiently transfected with the indicated plasmids using TurboFect reagent (Thermo Scientific). At 24 h, cells were washed with ice-cold phosphate-buffered saline (PBS), scraped from dishes, and incubated on ice for 30 min in 250 μl TNE buffer (10 mM Tris-HCl [pH 7.4], 150 mM NaCl, 5 mM EDTA, 10 mM phenylmethylsulfonyl fluoride [PMSF]). The cells were then lysed for 30 min on ice by addition of 250 μl of TNE plus 0.5% Triton X-100 (TX-100) (Sigma-Aldrich), 60 mM octyl glucoside (OG) (Sigma-Aldrich), or 1% Brij 97 (Sigma-Aldrich). For methyl-β-cyclodextrin (MBCD) (Sigma-Aldrich)-treated samples, cells were preincubated in serum-free DMEM for 2 h prior to the addition of 10 mM MBCD and then incubated for an additional 1 h before lysis in TNE plus 0.5% TX-100, as described above. All lysates were homogenized by passage through a 22-gauge needle and clarified by centrifugation at 5,000 rpm for 5 min at 4°C. Clarified lysates were mixed with an equal volume of 80% sucrose in TNE, placed in an ultracentrifuge tube, and overlaid with 2.5 ml of 30% sucrose in TNE, followed by 1 ml of 5% sucrose in TNE. Samples were centrifuged for 16 h at 35,000 rpm (Beckman SW55Ti rotor) at 4°C. Ten equal fractions of approximately 450 μl were collected from the top (fraction 1) of the gradient, and 3% of each fraction was analyzed by SDS-PAGE and Western blotting.

Western blotting.

Tetherin was detected in cell lysates using a 1:20,000 dilution of polyclonal rabbit anti-BST-2 serum (AIDS Research and Reference Reagent Program [ARRRP]; from Klaus Strebel), followed by HRP-conjugated goat anti-rabbit IgG (1:12,000) (Santa Cruz Biotechnology, Santa Cruz, CA). The HA-tagged proteins tetherin_HA and art-tetherin were detected using mouse anti-HA antibody 12CA5 (1:1,000) (Roche Applied Science, Indianapolis, IN), followed by HRP-conjugated goat anti-mouse IgG (1:10,000) (Sigma-Aldrich). CD59 was detected using a 1:100 dilution of mouse anti-CD59 antibody (Santa Cruz Biotechnology), and human transferrin receptor (TfR) was detected using a 1:500 dilution of mouse anti-transferrin receptor antibody (Invitrogen, Carlsbad, CA). Vpu was detected using 1:1,000 dilutions of rabbit anti-Vpu antiserum (ARRRP; from Frank Maldarelli and Klaus Strebel), and Ebola virus GP was detected using rabbit antisera raised against Ebola virus sGP, kindly provided by Paul Bates (University of Pennsylvania), followed by a 1:12,000 dilution of HRP-conjugated goat anti-rabbit IgG. Specific bands were visualized by ECL.

Confocal microscopy.

HeLa cells in 10-cm plates were transfected with specific expression plasmids and 24 h later were replated either onto coverslips coated with poly-l-lysine (Sigma-Aldrich) or in Lab-Tek chamber slides (Thermo Scientific). The cells were incubated for an additional 24 h at 37°C and then fixed and permeabilized using 100% cold methanol at −20°C for 20 min prior to antibody staining. Alternatively, cells transfected with Gag-YFP were fixed by incubation in 4% paraformaldehyde for 20 min at room temperature, followed by 3 washes in PBS and permeabilization by 10 min of incubation in 0.1% TX-100 at room temperature, followed by 3 additional PBS washes.

For cholesterol depletion studies, transfected cells were processed as described above except that prior to fixation in 100% methanol, the cells were incubated in either 10 mM MBCD in serum-free DMEM for 30 min at 37°C or 1% TX-100 for 15 min at 4°C. For cells that received both treatments, they were first incubated with MBCD, washed in PBS, and then incubated with 1% TX-100, followed by fixation.

Tetherin_HA was detected using a monoclonal mouse anti-HA antibody (Roche Applied Science) at a 1:500 dilution, EboGP_Myc was detected using a polyclonal goat anti-myc antibody (Abcam, Cambridge, United Kingdom) at a 1:1,000 dilution, and untagged EboGP was detected using mouse anti-Ebola virus GP monoclonal antibody 42 (a gift from Yoshihiro Kawaoka, University of Wisconsin) (88). CD55 was detected using specific mouse antiserum (Santa Cruz Biotechnology) at a 1:500 dilution. Vpu, transferrin receptor, and the trans-Golgi network (TGN) were detected as previously described (33). The conjugated secondary antibodies used were donkey anti-mouse Alexa Fluor 488, donkey anti-goat Alexa Fluor 488, donkey anti-rabbit Alexa Fluor 594, donkey anti-goat Alexa Fluor 594, donkey anti-sheep Alexa Fluor 594, and donkey anti-mouse Alexa Fluor 647 (Invitrogen). Processed cells were imaged and analyzed as previously described (54).

Colocalization analyses of confocal images were performed by calculating Pearson correlation coefficients and performing intensity correlation analysis (ICA), using NIH ImageJ software, essentially as described previously (33). Product difference from the mean (PDM) images were displayed as pseudocolored images with a scale bar, with areas of colocalization and positive PDM values represented in orange and negative PDM values represented in blue. To aid visualization, surface plot profiles were created from the PDM images, with higher peaks indicating stronger colocalization.

Quantitative reverse transcription-PCR (qRT-PCR).

Total cellular RNA was extracted from 293T or HeLa cells using TRIzol LS reagent (Invitrogen). A 0.5-μg portion was treated with DNase I (New England BioLabs Inc., Ipswich, MA) and reverse transcribed in a 30-μl reaction mixture containing 1× first-strand buffer (Promega, Madison WI), 200 units of Moloney murine leukemia virus (MMLV) RT (Promega), 100 pmol random hexamers (Genewiz, South Plainfield, NJ), 1.33 mM each deoxynucleoside triphosphate (dNTP) (Promega), and 20 units of RNAseOut (Invitrogen). The reaction mixture was incubated for 1 h at 37°C and inactivated for 15 min at 95°C. A volume of cDNA product corresponding to 16.7 ng of input RNA was amplified in a 20-μl reaction mixture containing 1× SYBR buffer (Applied Biosystems, Carlsbad, CA) and 12.5 pmol each of the specific forward and reverse primers: RICH2 1364-1387 F (5′-GTG GAT CCA GGC TTC CAA TGT CC-3′) and RICH2 1693-1666 R (5′-GCC AGTAAT GTT GAA CTC TAT CTC CCC-3′) or GAPDH F (5′-TGC MTC CTG CAC CAC CAA CT-3′) and GAPDH R (5′-YGC CTG CTT CAC CAC CTT C-3′). The reaction was monitored on a 7500 Fast system (Applied Biosystems) and the relative RICH2 levels analyzed using the ΔΔCT method normalized to the GAPDH (glyceraldehyde-3-phosphate dehydrogenase) levels.

RESULTS

Conserved cysteines in the Ebola virus GP cytoplasmic tail are not required for anti-tetherin activity.

Previously, Ebola virus GP was reported to be present in DRMs isolated from either GP-transfected 293T cells or HepG2 cells infected with Ebola virus Zaire (5). Furthermore, two conserved cysteines in the cytoplasmic tail of GP were found to be necessary for this distribution (5). The two cysteine residues are highly conserved among the Filoviridae and are sites for modification with palmitic acid (23, 38), a known marker for lipid raft association. Therefore, as an indirect measure of the importance of lipid raft localization for GP's anti-tetherin activity, we examined the activity of GP containing alanine substitutions at these residues (EboGP_C670/672A).

The alanine-substituted GP retained the same cellular distribution as the wild-type protein (Fig. 1A), being present both at the cell surface and also in a perinuclear location that partially overlapped with the resident endoplasmic reticulum (ER) protein calnexin (data not shown). We compared the abilities of the GPs to counteract tetherin and promote the release of HIV-1 Gag-Pol-based VLPs using HeLa cells, which naturally express tetherin. The HIV-1 Vpu protein was included as a control anti-tetherin factor. We confirmed the anti-tetherin activity of the wild-type GP and found that the C670/672A mutant was also able to enhance VLP release, although less efficiently than the wild-type protein (mean, 81%; n = 2 experiments) (Fig. 1B). We repeated these analyses using 293T cells transiently transfected with either native tetherin or an artificial tetherin-like protein (art-tetherin) (74) (Fig. 1C), where we confirmed our previous finding that GP can counteract both proteins, while Vpu has activity only against native tetherin (54). In this system, we found that the C670/672A mutant had about half the activity of the wild-type GP against native tetherin but was equally as effective against art-tetherin. Of note, art-tetherin is a considerably weaker inhibitor of HIV-1 VLP release than tetherin (54).

Fig 1.

anti-tetherin activity of Ebola virus GP. (A) Confocal microscopy analysis of HeLa cells expressing either Ebola virus GP (EboGP) or the mutant EboGP_C670/672A, detected with anti-GP antibody. (B) HIV-1 Gag-Pol VLPs were generated in HeLa cells by transfection with plasmid pHIV-1-pack, together with either a control cytomegalovirus (CMV) expression plasmid (Ctrl.) or an expression plasmid for Vpu, EboGP, or EboGP_C670/672A. VLP release was analyzed by Western blotting of lysates and pelleted supernatants using the indicated antibodies. HIV-1 Gag-Pol VLPs (C) or Ebola virus VP40 VLPs (D) were produced in 293T cells in the presence of either tetherin or artificial tetherin (art-tetherin), together with either a control CMV expression plasmid (Ctrl.) or an expression plasmid for Vpu, EboGP, or EboGP_C670/672A. VLP release was analyzed by Western blotting of lysates and pelleted supernatants using anti-p24 or anti-FLAG antibodies. For each type of VLP, the fold enhancement of VLP release in the presence of Vpu or GP was calculated relative to control (tetherin only) levels. Graphs show means plus standard deviations for n = 3 independent experiments, with a P value of <0.05 (*) or <0.01 (**) indicated.

We next examined the effects of GP and Vpu expression on the release of minimal Ebola virus VLPs composed of the VP40 protein (32, 40). Such an analysis is complicated by the fact that GP itself has been reported to stimulate the production of VP40 VLPs when expressed in 293T cells (5, 51, 75), suggesting a tetherin-independent mechanism. In agreement, we also observed a small (1.5-fold) increase in VLP release by the wild-type GP, although there was no statistically significant enhancement caused by the C670/672A mutant (Fig. 1D). As expected, when tetherin was coexpressed, VP40 VLP release was significantly inhibited. However, this was successfully counteracted by GP, resulting in a 6-fold increase in VLP release over the level achieved in the presence of tetherin alone (Fig. 1D). The C670/672A mutant was also able to counteract tetherin, although, similar to the situation we observed with the HIV-1 VLPs, its activity was only about half as strong as that of the wild-type GP. However, in the case of Ebola virus VLPs, we cannot rule out that the superior activity of the wild-type GP was due to an additional, non-tetherin-mediated effect contributing a 1.5-fold increase in VLP release.

We also noted that Vpu was less effective than the wild-type GP in the VP40 VLP system, causing only a 3-fold increase in VLP release. While this difference could also be attributed to the additional effect of wild-type GP on VP40 VLP release, it is also possible that some specificity exists between the two different VLP systems, since Vpu was more effective at stimulating HIV-1 VLPs than Ebola virus GP (Fig. 1C). Finally, we observed that both the wild-type and mutant GPs were equally effective at counteracting art-tetherin, effectively restoring VP40 VLP release levels back to the no-tetherin levels (Fig. 1D). As with the HIV-1 VLPs, we found that art-tetherin was a much weaker inhibitor of VP40 VLP release than native tetherin.

Tetherin association with DRMs and sensitivity to lipid raft-defining disruption methods.

Tetherin's characterization as a lipid raft protein is based on its resistance to solubilization in cold nonionic detergents such as Triton X-100 (TX-100), which results in flotation on sucrose density gradients (47). This method typically localizes DRM proteins to fractions 2 to 4 out of a total of 10 fractions harvested, while the solubilized proteins sediment in fractions 7 to 10. Using this technique, we analyzed tetherin from both HeLa cells, which naturally express tetherin, and transfected 293T cells, since these are two common systems used to study tetherin biology. A known lipid raft protein, CD59 (85), was included as a control.

In HeLa cells, we found that 44% of the total tetherin was associated with the DRM fractions 2 to 4 and 53% with the soluble fractions 7 to 10 (Fig. 2A). CD59 gave a similar distribution, with 40% in DRMs and 58% in soluble fractions. Since TX-100 insolubility can also be caused by association of proteins with cytoskeletal elements (6, 36, 62, 81) and the cytoplasmic domain of tetherin indirectly associates with the actin cytoskeleton (77), we next investigated whether tetherin's retention in the DRM fraction truly reflected a lipid raft distribution or could instead be caused by such an association. To do this, we treated lysates with octyl glucoside (OG), a nonionic detergent that efficiently solubilizes DRMs (8, 59) but does not disrupt associations with the cytoskeleton (63, 81). Following this treatment, we found a large increase in the amounts of both tetherin (76%) and CD59 (73%) that were solubilized (Fig. 2A). This behavior is consistent with a lipid raft distribution for both proteins and further demonstrates that although tetherin can coimmunoprecipitate with elements of the cytoskeleton (77), it is not itself a component of the cytoskeleton in HeLa cells.

Fig 2.

Tetherin localization to DRMs in HeLa cells and transfected 293T cells. Mock-transfected HeLa cells (A) or 293T cells transfected with a tetherin expression plasmid (B) were lysed in cold TNE buffer containing either 0.5% Triton X-100 (TX-100) or 60 mM octyl glucoside (OG). A third set of cells were pretreated by incubation with methyl-β-cyclodextrin (MBCD) before lysis in TNE plus 0.5% TX-100. Cell extracts were ultracentrifuged through sucrose density gradients, and 10 fractions were collected and analyzed for the presence of tetherin or CD59, a known lipid raft protein, by Western blotting using specific antisera. Under these conditions, DRM proteins are found in fractions 2 to 4. Graphs show the percentage of total tetherin (left) or CD59 (right) in each fraction for the indicated treatments.

We further confirmed tetherin's association with lipid rafts by incubating the cells with methyl-β-cyclodextrin (MBCD) prior to TX-100 treatment and DRM analysis. MBCD depletes cholesterol from membranes and thereby disrupts lipid rafts (37). This treatment resulted in the movement of both tetherin (86%) and CD59 (89%) into soluble fractions (Fig. 2A). Since MBCD removes cholesterol only from the plasma membrane (45, 65, 69), the remaining tetherin (10%) and CD59 (8%) in the DRM fractions likely represent protein associated with intracellular rafts.

We repeated these analyses using 293T cells transfected with tetherin (Fig. 2B), where we observed marked differences from the HeLa cells. First, a much greater percentage (73%) of tetherin was present in the DRM fractions. This cell type-dependent shift to the DRM fractions was also observed for the control raft protein, CD59. In addition, we noted that tetherin in 293T cells was present both as a slower-running smear (30 to 40 kDa) that is typical of the mature, glycosylated, post-ER form of tetherin and as a distinct, single protein species at 24 kDa that corresponds to an immature form of the protein (2). This immature form of tetherin was the only species that we detected in the soluble fractions. A similar pattern for rat tetherin has been observed in H4IIE cells, where the majority of tetherin in the DRM fractions was the mature, glycosylated form, while the immature form sedimented in the soluble fractions (47). We occasionally also observed the immature form of tetherin in HeLa cell lysates (see examples in Fig. 3 and 4), although it was a much less prominent species than in the transfected 293T cells.

Fig 3.

Effect of Vpu on tetherin DRM distribution. DRMs were isolated from HeLa cells transfected with an empty expression plasmid or Vpu. Ten sucrose gradient fractions were collected and probed for tetherin, Vpu, the control raft protein CD59, and the nonraft protein TfR. Graphs show the percentage of total tetherin in each fraction in the presence or absence of Vpu for n = 3 independent experiments. Error bars represent standard deviations.

Fig 4.

Effect of GP on tetherin DRM distribution. (A) TX-100 DRMs were isolated from HeLa cells transfected with an empty expression plasmid, Ebola virus GP (EboGP), or EboGP_C670/672A. Ten gradient fractions were collected and probed for tetherin, EboGP, and CD59. Graphs show the percentage of total tetherin in each fraction in the presence or absence of the indicated GP for n = 3 independent experiments. (B) Brij 97 DRMs were isolated from HeLa cells transfected with an empty expression plasmid or EboGP. Gradient fractions were collected and analyzed as described above. The graph shows the percentage of total tetherin in each fraction in the presence or absence of EboGP for a single experiment. (C) TX-100 DRMs were isolated from HeLa cells transfected with art-tetherin alone or together with EboGP. Ten gradient fractions were collected and probed for art-tetherin using an anti-HA antibody and for EboGP and CD59. The graph shows percentage of total art-tetherin in each fraction in the presence or absence of GP for n = 3 independent experiments. Error bars represent standard deviations.

When 293T cell lysates were treated with OG, only 15% of CD59 remained in the DRM fractions 2 to 4, while a considerably greater fraction (50%) of tetherin was retained (Fig. 2B). This suggests that the exogenous tetherin expressed in transfected 293T cells is more tightly associated with the membrane cytoskeleton than either a control raft protein, CD59, in the same cells or the endogenous tetherin present in HeLa cells. Of note, tetherin has been reported to associate with the cytoskeletal protein RICH2, (77), and qRT-PCR analysis found that the relative levels of RICH2 mRNA were 2.3-fold higher in 293T cells than in HeLa cells (data not shown). Such an association could account for a considerable part of tetherin's DRM residency in 293T cells. Similarly, we found that although CD59 was effectively solubilized by MBCD incubation prior to TX-100 extraction, with a shift from 6% to 57% of the total protein in the soluble fractions, the same treatment solubilized only 44% of tetherin (Fig. 2B). Notably, the majority of the tetherin that was solubilized by MBCD was the immature form of the protein, suggesting that most of the mature, and therefore cell surface, tetherin was relatively insensitive to cholesterol depletion in these cells. Possibly this could reflect different lipid compositions of 293T and HeLa cells, such that TX-100 resistance in 293T cells involves both cholesterol and sphingomyelin and removing cholesterol alone is not enough (83). Alternatively, if the tetherin expressed in 293T cells is more tightly associated with the membrane cytoskeleton, this could retain the protein in the DRM fractions of the gradient, even when cholesterol was removed.

In summary, these data reveal cell type differences between HeLa and 293T cells in how proteins behave in the assays used to assign lipid raft association. Such variations have previously been reported for other cell types (72, 83). Since the endogenous tetherin in HeLa cells was almost completely solubilized by both the OG and MBCD treatments that specifically disrupt lipid rafts, we chose to restrict subsequent studies to examine tetherin's lipid raft association to HeLa cells.

Vpu does not move tetherin out of DRMs.

Recently, it was shown by two independent reports that Vpu resides within DRMs expressed in transfected 293T cells (21, 78). Ruiz et al. found that certain mutations within the transmembrane domain of Vpu reduced its anti-tetherin activity and prevented DRM distribution, suggesting that lipid raft association is functionally important for this activity (78). In contrast, Fritz et al. analyzed mutants with mutations in the Vpu hinge region which abrogated DRM association but did not affect antagonism of tetherin restriction (21). Our analysis of Vpu in transfected HeLa cells identified only a minor population of Vpu in DRM fractions 2 to 4, and we also observed similar low levels of Vpu in fractions 5 and 6, where we did not detect either tetherin or CD59. Overall, most of the Vpu protein was found in the soluble fractions (Fig. 3).

We asked whether Vpu could interfere with the localization of tetherin in DRMs by coexpressing Vpu in HeLa cells. Although we noticed a decrease in the absolute levels of tetherin, which is a known consequence of Vpu coexpression (4, 17, 55, 60, 61), we observed no impact on tetherin's distribution within the sucrose gradient, where it remained mostly associated with the DRM fractions (Fig. 3). Since Vpu removes tetherin from the cell surface, and sequestration in the TGN is one suggested mechanism (18, 33), it is likely that this DRM-associated tetherin under these conditions is located in intracellular rafts, including those in the TGN.

Ebola virus GP does not sediment in DRMs and has no effect on tetherin's distribution.

A single previous study using transfected 293T cells identified GP as a lipid raft protein (5). However, our DRM analysis of GP expressed in HeLa cells found no evidence to support such a characterization (Fig. 4A), and similar results were also obtained with 293T cells (data not shown). Furthermore, we did not observe any difference in the sucrose gradient profiles of the wild-type or the C670/672A mutant GPs (Fig. 4A). Finally, when tetherin was analyzed in the presence of either the wild-type or cysteine mutant GPs, there was no reduction in its DRM association. These data suggest that similar to Vpu, Ebola virus GP does not act to exclude tetherin from lipid rafts.

The lack of DRM partitioning for Ebola virus GP that we observed following Triton X-100 solubilization does not necessarily rule out a lipid raft location, since this treatment can also disrupt the association of proteins that interact only weakly with lipid rafts (39, 83). We therefore repeated these analyses using a less stringent nonionic detergent, Brij 97 (formally known as Brij 96), that results in DRMs that retain significantly more plasma membrane proteins (12). However, in agreement with the Triton X-100 data, we found that the Brij 97 DRMs did not contain GP, and there was no difference in tetherin's profile in the presence or absence of GP under these isolation conditions (Fig. 4B). The slight upward shift in tetherin's distribution in the gradient under conditions of Brij 97 extraction compared to Triton X-100 extraction are typical of the characteristics of the different detergents (83) and were duplicated by the CD59 control protein.

Finally, since GP also counteracted restriction by art-tetherin, we examined the DRM association of art-tetherin and asked whether GP had any impact on this. We found that art-tetherin also localized to DRMs in HeLa cells, in a pattern similar to that for both the native tetherin protein and CD59. However, as observed with native tetherin, coexpression of GP had no effect on this distribution (Fig. 4C).

Tetherin and Ebola virus GP colocalize at cell surface by confocal microscopy, but only tetherin is in cell surface rafts.

Our biochemical DRM analyses identified tetherin, but not GP, as being present in DRMs. Such DRM fractions contain a mixture of proteins at the cell surface, as well as those trafficking through the secretory and endocytic pathways (34, 68). It was previously suggested that a significant portion of tetherin in DRMs must come from intracellular rafts, since only about ∼30% of total tetherin was found at the surface, while much more than 30% was in DRMs (57). We therefore more specifically examined the colocalization and lipid raft association of the cell surface populations of tetherin and Ebola virus GP.

First, we observed tetherin and GP at the cell surface using confocal microcopy. To facilitate the studies, we used a myc-tagged version of GP, which we confirmed had full anti-tetherin activity (Fig. 5A), and an HA-tagged version of tetherin that we have previously shown is just as restrictive as the untagged protein (54). Coexpression of EboGP_Myc and tetherin_HA in HeLa cells resulted in a high degree of colocalization, giving a Pearson coefficient of 0.774 when the entire cell shown was scored (Fig. 5B), which increased to 0.902 when only the cell surface was analyzed. In contrast, there was minimal colocalization between the two proteins at intracellular sites, where tetherin is found in the TGN but Ebola virus GP is not. Finally, we noted that tetherin's distribution at both the surface and intracellular locations was unaffected by GP coexpression (Fig. 5B).

Fig 5.

Cell surface expression of tetherin and GP. (A) anti-tetherin activity of myc-tagged Ebola virus GP (EboGP_Myc), assessed by HIV-1 VLP release from HeLa cells, is equivalent to that of the untagged GP (EboGP-X). (B and C) Confocal microscopy analysis of cellular distribution and colocalization between Ebola virus GP (EboGP_Myc) and tetherin_HA (B) or art-tetherin (C) in HeLa cells. Control cells were transfected with an empty expression plasmid. Merged images reveal that both tetherin and art-tetherin have a high degree of overlap with EboGP_Myc. Insets show magnifications of areas in dotted white lines. The Pearson correlation coefficients (r values) are shown for overlapped red and green pixels in merged images. (D) Confocal microscopy analysis of the effects of the indicated treatments on cell surface proteins in HeLa cells transfected with either a control CMV expression plasmid, for analysis of endogenous CD55 and TfR, or an expression plasmid for tetherin_HA or EboGP_Myc. Both tetherin and CD55 exhibited characteristics of lipid raft proteins, being removed from the surface only by the combination of MBCD and TX-100. Ctrl., untreated cells.

We performed a similar analysis using art-tetherin, where, in contrast to the initial report that described this protein as having the same distribution as the native tetherin (74), we found some differences. Specifically, art-tetherin had no prominent TGN localization and an overall stronger cell surface signal than tetherin (Fig. 5C). When art-tetherin was coexpressed with EboGP_Myc, we found that the two proteins strongly colocalized both at the cell membrane and at intracellular sites, resulting in a Pearson coefficient for the whole cell of 0.890. As with native tetherin, coexpression of Ebola virus GP had no effect on art-tetherin's distribution (Fig. 5C).

Next, we analyzed whether GP or tetherin was present in cell surface lipid rafts by repeating the confocal microscopy analyses under conditions where lipid rafts are disrupted by cholesterol depletion (47). First, HeLa cells were incubated with 1% TX-100 at 4°C, a treatment that solubilizes nonraft proteins (24), and we confirmed that this significantly reduced the cell surface staining of a control nonraft protein, transferrin receptor (TfR). In contrast, the surface staining patterns for a control raft protein, CD55 (30), and tetherin_HA, were unaffected (Fig. 5D). The intracellular perinuclear pool of tetherin that we have previously identified as being in the TGN (33) was also solubilized by the TX-100 treatment, as was previously observed for rat tetherin (47). Several studies using polarized epithelial cells have reported that raft proteins destined for the plasma membrane first become detergent insoluble in the TGN (8, 53, 90). However, our observation of a TX-100-soluble perinuclear pool of tetherin would suggest that the majority of intracellular tetherin in HeLa cells is not raft associated and likely contributes to the significant amount of soluble tetherin we find upon TX-100 extraction (Fig. 2A). Finally, we observed that Ebola virus GP was also extracted from the surface by TX-100 incubation, indicating that it is not in lipid rafts at the cell surface and confirming our findings from the sucrose gradient studies.

We next treated cells with MBCD, which extracts cholesterol exclusively from the plasma membrane but leaves lipid rafts in the TGN intact (45, 65, 69). As expected, no proteins were depleted from the cell surface by this treatment alone, and the resolution limitations of microscopy mean that it is also unlikely that we would detect any effects on protein distribution caused by the disruption of surface raft microdomains in this way. However, the combination of MBCD and TX-100 extraction results in the solubilization of lipid raft proteins that are not removed by either treatment alone and has previously been demonstrated to effectively extract tetherin (47). In agreement, we found that this combination completely removed both CD55 and tetherin_HA from the surface, leaving only punctate intracellular protein aggregates, thereby confirming their localization in cell surface rafts (Fig. 5D).

Overall, these confocal microscopy studies agreed with the biochemical data, confirming that while both tetherin and GP are present at the cell surface, only tetherin is present in lipid rafts. Despite these differences in raft residency, we observed strong colocalization at the surface between the two proteins, which likely reflects the limits of resolution of standard confocal microscopy. Indeed, lipid raft microdomains are too small to be seen without first artificially clustering rafts by copatching with a raft marker such as GM1 (31). The alternative possibility remains that GP is also present in lipid rafts but is significantly more susceptible to TX-100 extraction than either tetherin or classic raft marker proteins such as CD55 and CD59.

Impact of Vpu and Ebola virus GP on tetherin association with lipid rafts, visualized by confocal microscopy.

We next performed confocal microscopy studies to analyze the distribution of tetherin in cell surface rafts when either Vpu or GP was coexpressed. Our previous studies have shown that Vpu antagonism includes sequestration of tetherin in the TGN (33), and we also observed that coexpression of tetherin_HA and Vpu resulted in an intense intracellular colocalization. Interestingly, this compartment remained intact after treatment with TX-100 (Fig. 6A), which contrasted with the almost complete solubilization of intracellular tetherin_HA by TX-100 that we observed when it was expressed alone (Fig. 5D). When cells were treated with the combination of MBCD followed by TX-100 solubilization, two distinct populations were observed: ∼60% of cells had very little staining for either Vpu or tetherin_HA, while ∼40% of cells retained staining for both proteins, colocalized around the nucleus. Together these data suggest that Vpu sequesters tetherin in detergent-resistant membranes located in a perinuclear compartment, most likely the TGN (18, 33).

Fig 6.

Effects of Vpu and GP on cell surface raft association of tetherin. Confocal microscopy analysis of the effects of the indicated treatments on tetherin localization in HeLa cells transfected with tetherin_HA and either Vpu (A) or EboGP_Myc (B) is shown. Two representative fields are shown for the MBCD-plus-TX-100 treatment of tetherin_HA plus Vpu, showing the different observed outcomes. Ctrl., untreated cells.

In contrast, when we analyzed the effects of coexpression of tetherin_HA and Ebola virus GP (Fig. 6B), we observed no impact on the distribution of tetherin compared to when it was expressed alone (Fig. 5D). Overall, these results confirm the findings of the DRM analyses and support the conclusion that GP does not affect tetherin's partitioning into cell surface DRMs.

Ebola virus GP does not prevent tetherin and HIV-1 Gag colocalization at the cell surface.

Previously, tetherin has been shown to colocalize at the cell surface with budding viral particles, including HIV-1 and MLV Gag VLPs (43, 64), Vpu-deleted HIV-1 virions (60), and both Marburg virus (43) and Ebola virus (44) VP40 VLPs. Since expression of GP did not alter either the overall cell surface distribution or DRM partitioning of tetherin, we asked instead whether GP was preventing the specific colocalization of tetherin with assembling viral particles. To do this, we coexpressed HIV-1 Gag (Gag-YFP) and tetherin_HA in HeLa cells and analyzed the distribution of the proteins by confocal microscopy. For merged images of interior planes of the cells, the surface was mostly teal in color, reflecting overlap between Gag (green channel) and tetherin (blue channel) (Fig. 7A). The high degree of colocalization was confirmed by determining the product of the difference from the mean (PDM) for each pixel (Fig. 7A) and by calculating the Pearson coefficient for the two channels (Fig. 7C). We also confirmed these findings when focusing on the top of cells, just above the cover glass (Fig. 7B and D).

Fig 7.

Tetherin colocalization with HIV-1 Gag at the cell surface. HeLa cells were transfected with HIV-1 Gag-YFP and tetherin_HA alone (Ctrl.) or together with either Vphu-HcRed or EboGP_Myc. (A) Confocal microscopy images were taken at internal slices through the cell. Merged panels show the overlap between all three channels; insets, indicated by the dotted white lines, are magnified merges between the green and blue channels only. The product of the differences from the mean (PDM) images are shown for the Gag-YFP and tetherin_HA channels as pseudocolored panels with a scale bar, where a more orange color indicates a large PDM value and high degree of colocalization, while blue indicates no overlap. (B) Merged confocal microscopy images are shown for internal slices (far left column) and the corresponding top slice showing the image plane of the plasma membrane (green and blue channels merged only). PDM images are shown for merged Gag-YFP and tetherin_HA channels, with surface plots of the PDM values to indicate the strength of the correlation between the Gag-YFP and tetherin_HA channels. (C) The Pearson correlation coefficients were calculated for the Gag-YFP and tetherin_HA channels for 6 cells per condition, using merged images of internal confocal slices, as shown in panel A. Prior to calculation, intracellular regions were masked to focus only on the cell surface. (D) The Pearson correlation coefficients were calculated for the Gag-YFP and tetherin_HA channels for 8 cells per condition, using merged images from the top of the cell, as shown in panel B. Error bars represent standard deviations.

As expected, the coexpression of Vpu moved tetherin away from the surface and therefore prevented its colocalization with Gag (43, 60, 64). This resulted in the mostly green color in the merged confocal panels, the blue scale color in the PDM images, and the low Pearson coefficient scores when either interior slices or the top of the cell was analyzed. Although a residual low level of tetherin remained at the plasma membrane, this population did not colocalize strongly with Gag, as shown by the small peaks in the PDM surface plot (Fig. 7B). In contrast, GP expression had no impact on either the distribution of cell surface tetherin or its colocalization with HIV-1 Gag by any of these measures.

Ebola virus GP does not disrupt HIV-1 Gag lipid raft localization.

It has previously been reported that a fraction of HIV-1 Gag proteins can be isolated from DRMs from various cell lines (35, 52, 66, 70) and that lipid rafts play a role in HIV-1 assembly (52, 70). We therefore examined whether the DRM association of Gag was in any way disrupted by GP. Such an effect could physically separate Gag and tetherin in membrane microdomains and thereby block tetherin's restriction of virus release.

We first analyzed the DRM association of HIV-1 VLPs in HeLa cells. We found that the profile matched previous reports (70, 71), with a fraction of the immature Pr55^Gag protein detected in the DRM fractions (Fig. 8A). When either Vpu or GP was coexpressed, there was no change in this profile. Since the amount of Pr55^Gag in the DRMs (fractions 2 to 4) in HeLa cells was quite low (average of 8.7% of the total), we also repeated these analyses in 293T cells cotransfected with tetherin and GP. Here, a greater portion of Pr55^Gag (32%) was now located in the DRM fractions (Fig. 8B). However, neither the coexpression of tetherin nor the combination of tetherin and GP had any effect on Pr55^Gag. Statistical analysis of the percentage of Pr55^Gag in DRMs or the soluble fractions for both cell types is shown in Fig. 8C.

Fig 8.

HIV-1 Gag DRM association in the presence of Vpu or GP. (A) DRMs were isolated from HeLa cells transfected with a HIV-1 Gag-Pol expression plasmid (pHIV-1-pack) and either a control CMV expression plasmid (Gag) or together with an expression plasmid for Vpu or Ebola virus GP (EboGP). Ten gradient fractions were collected for each condition and probed for HIV-1 Pr55^Gag (using an anti-p24 antibody), Vpu, GP, or the control raft protein CD59. (B) DRMs were isolated from 293T cells transfected with pHIV-1-pack and either a control CMV expression plasmid, tetherin alone, or tetherin plus EboGP and analyzed as described above. (C) The graph shows the percentage of total Pr55^Gag in either the DRM fractions (fractions 2 to 4) or the soluble fractions (fractions 7 to 10) of the gradients, with and without Vpu or GP expression, for HeLa cells (n = 3 independent experiments) and 293T cells (n = 2 independent experiments). Error bars represent standard deviations. No significant differences were found between the distributions of Pr55^Gag in the absence or presence of Vpu or GP.

The totality of our data leads us to conclude that Ebola virus GP has no effect on HIV-1 Gag colocalization with tetherin and no effect on the association of either tetherin or Gag with lipid rafts. Therefore, it seems that GP is able to neutralize the effect of tetherin without grossly altering the localization of any of the components of this interaction.

DISCUSSION

Tetherin inhibits the release of a wide variety of enveloped viruses and has been proposed to do this through the formation of a physical tether between viral and host cell membranes (19, 64, 74). This model of tetherin action is supported by the observation that a wholly artificial tetherin (art-tetherin), composed of similar protein functional domains, is also able to restrict enveloped virus release (74). Various viruses express tetherin antagonists that typically act by removing tetherin from the surface, either by promoting its degradation or by sequestering the protein at intracellular sites (18, 25, 33, 48, 55, 56, 60). The Ebola virus GP is unique in that it counteracts tetherin restriction without either degrading the protein or removing it from the cell surface, and it is also able to counteract art-tetherin (54). The mechanism it uses to do this is unknown.

Although both confocal microscopy and fluorescence-activated cell sorter (FACS) analyses have shown that tetherin remains at the cell surface when GP is expressed (54), we considered the possibility that the microlocalization of tetherin could be altered by the presence of GP, so that it was either excluded from the specific sites of virus assembly or affected in its ability to form a physical tether. The previously reported localization of both tetherin (47, 57) and GP (5) to lipid rafts suggested a possible platform for such an interaction, especially since lipid rafts also serve as sites for the assembly of various viruses (7, 13, 86).

Using both biochemical and confocal microscopy methods, we were able to confirm that tetherin is a lipid raft protein and that a fraction of HIV-1 Pr55^Gag is also found in lipid rafts. However, our analysis of GP found no evidence to support a lipid raft location for this protein in either HeLa or 293T cells. The literature about Ebola virus and rafts is sparse and contradictory, with one study suggesting rafts as the site of assembly of Ebola virus based on the colocalization of its matrix protein, VP40, with clusters of surface raft marker proteins (73) and another study finding that VP40 could not localize within lipid rafts unless Tsg101 was coexpressed (50). We also examined the lipid raft association of VP40 to determine if its expression was necessary for GP to localize to lipid rafts, and we found that VP40 also did not reside within TX-100 DRMs (data not shown). For GP itself, one study has localized filovirus GPs to DRMs in both transiently transfected 293T cells and infected HepG2 cells (5), but another study found that GP did not associate at the cell surface with the HIV-1 Env, a known raft protein (49). Finally, although both tetherin and GP are present at the cell surface, a precise colocalization has not been observed (44, 75). While our data overwhelmingly suggest that GP is not a raft protein, it remains possible that these different observations reflect a weak association of GP with rafts that is not preserved during either TX-100 or Brij DRM analyses.

Since tetherin and HIV-1 Gag have been reported to colocalize at the surface (60, 64), we also asked more broadly whether GP expression affected the extent of this association, regardless of whether DRMs were involved. We were able to demonstrate a strong colocalization between surface-associated Gag and tetherin, and we confirmed that Vpu significantly reduced this association by removing tetherin from the cell surface. In contrast, GP had no impact on the degree of colocalization between Gag and tetherin.

How does GP block tetherin's action without removing it from sites of virus assembly? One possibility is that GP interacts directly with tetherin in a manner that blocks its ability to form a virus-cell tether. In support of such a mechanism, it has previously been reported that GP can coimmunoprecipitate with tetherin (44), although the form of GP that was detected was an immature, and presumably intracellular, species. Since the overall cellular distribution of tetherin was unaffected by GP coexpression and since substantial colocalization between the two proteins occurred only at the plasma membrane, our observations do not support direct interactions occurring in intracellular compartments, and we have also been unable to demonstrate a specific interaction between tetherin and the mature form of GP (data not shown). In addition, the lack of specificity in GP's action implied by its activity against art-tetherin does not easily fit with a model of specific protein-protein interactions, unless a common feature shared by both native tetherin and art-tetherin is being recognized by GP.

Alternative nonspecific mechanisms of action to account for GP's activity against both tetherin and art-tetherin restriction include interfering with a cellular cofactor that is necessary for tether formation or altering some other feature of the assembly site. Such effects could even be something that GP does as a contribution to the Ebola virus assembly process. In support of such a scenario, GP has recently been reported to impede the recognition of several cell surface proteins, including β1 integrin (CD29) and major histocompatibility complex (MHC) class I (84, 87, 95), through a mechanism of glycan-mediated steric hindrance (20, 76). However, the mucin-like domain of GP that is necessary for this effect (20) is not required for GP to counteract tetherin (44).

In summary, we have demonstrated that Ebola virus GP can prevent tetherin restriction without reducing its steady-state levels, altering its cellular distribution, or removing it from sites of virus assembly. In particular, GP's action does not exclude tetherin from lipid rafts. A lack of involvement of lipid rafts was also demonstrated for Vpu's anti-tetherin activity. It will be interesting to determine whether the ability of GP to counteract both tetherin and art-tetherin restriction arises from an action that is part of the normal process of Ebola virus assembly, where tetherin could be considered a bystander target, or whether it indeed reflects a specific tetherin antagonism.