Stimulation of NF-κB Activity by the HIV Restriction Factor BST2

Abstract

BST2 (HM1.24; CD317; tetherin) is an interferon-inducible transmembrane protein that restricts the release of several enveloped viruses, including HIV, from infected cells. Before its activity as an antiviral factor was described, BST2 was identified as an inducer of NF-κB activity. Here we show that human BST2 induces NF-κB in a dose-dependent manner. This activity is separable from the restriction of virus release: a YxY sequence in the cytoplasmic domain of BST2 is required for the induction of NF-κB but is dispensable for restriction, whereas the glycosylphosphatidylinositol (GPI) addition site in the protein's ectodomain is required for restriction but is largely dispensable for the induction of NF-κB. Mutations predicted to disrupt the coiled-coil structure of the BST2 ectodomain impaired both signaling and restriction, but disruption of the tetramerization interface differentially affected signaling. The induction of NF-κB by BST2 was impaired by inhibition of transforming growth factor β (TGF-β)-activated kinase 1 (TAK1) or by calcium chelation, suggesting potential linkage to the mitogen-activated protein kinase and endoplasmic reticulum (ER) stress response pathways. Consistent with a role for TAK1, BST2 coimmunoprecipitated with TAK1 and the TAK1-associated pseudophosphatase TAB1; these interactions required the YxY sequence in BST2. Moreover, signaling by BST2 was blocked by expression of an IκB-mutant that inhibits the canonical pathway of NF-κB activation. The expression of HIV-1 Vpu inhibited the induction of NF-κB by BST2; this inhibition required Vpu's ability to bind the cellular β-TrCP-E3-ubiquitin ligase complex. The expression of HIV-1 lacking vpu augmented the induction of NF-κB activity by BST2, suggesting that BST2 can act as a virus sensor. This augmentation was also inhibited by Vpu in a β-TrCP-dependent manner. The role of BST2 in the host-pathogen relationship is apparently multifaceted: signaling during the innate immune response, sensing of viral gene expression, and direct restriction of virus release. HIV-1 Vpu counteracts each of these functions.

INTRODUCTION

BST2 (bone marrow stromal cell antigen 2) (also known as tetherin) is an interferon (IFN)-inducible transmembrane and glycosylphosphatidylinositol (GPI)-anchored protein that restricts the release of several enveloped viruses from infected cells (1, 2). Viruses susceptible to BST2 include all retroviruses so far tested as well as members of the Rhabdoviridae, Paramyxoviridae, Filoviridae, and Herpesviridae families (reviewed in reference 3). Most of these viruses encode BST2 antagonists, which degrade the protein or remove it from the cell surface; the prototype BST2 antagonist is the HIV-1 accessory protein Vpu (1, 2).

Although the release of cell-free virions and cell-free infectivity can be dramatically inhibited by BST2, restriction of the cell-to-cell spread of virus is less effective (4, 5). This observation calls into question whether restriction of enveloped viruses is the sole function of this protein. Consistent with the possibility of additional functions, BST2 reportedly serves as the ligand for ILT7, a receptor on plasmacytoid dendritic cells that negatively regulates the expression of type I interferon (6). Moreover, BST2 reportedly stimulates the activity of the NF-κB family of transcription factors (7), although the determinants of this activity in BST2 and its consequences have until very recently been unknown.

Here we confirm that BST2 induces NF-κB activity. We show that this activity is genetically separable from the restriction of virion release, yet it requires conserved features of the protein. These features include a YxY motif in the cytoplasmic domain (CD) of BST2 that directs the interaction with a TAK1- and TAB1-containing signaling complex. The BST2 antagonist encoded by HIV-1, Vpu, inhibits the activation of NF-κB by BST2 in a manner dependent on its ability to bind the cellular β-TrCP-containing, cullin-1-based E3 ubiquitin ligase complex. In the absence of Vpu, however, the expression of HIV-1 augments the activation of NF-κB. This suggests that BST2, like the restriction factor Trim5α (8), serves not only as an effector protein of the innate immune response but also as a viral sensor predicted to trigger an inflammatory response.

MATERIALS AND METHODS

Plasmids, antibodies, cell lines, and reagents.

Plasmids expressing BST2, Vpu from a codon-optimized sequence (pVphu), the complete HIV-1 genome (pNL4-3), a pNL4-3-derived mutant containing an out-of-frame deletion downstream of the Vpu initiator (Udel; here referred to as Δvpu), a pNL4-3-derived mutant encoding S52/56N substitutions in Vpu, IκB and dominant-negative IκB, and HLA-A2 and the AP1-luc reporter plasmid, have been previously described (9–15). Plasmids expressing BST2 were derived from pcDNA3.1 as described previously (16), except for the pcDNA3.1 derivative expressing BST2-C3A (10). Although the native sequences of bst2 upstream of the initiator codon were not preserved, they were, like the native sequence, suboptimal for translational initiation at methionine residue 1. The NF-κB-luc reporter plasmid and the β-galactosidase expression plasmid were purchased from Promega. The ISRE-luc reporter plasmid, the IRF-3 5D expression plasmid, and plasmids expressing FLAG-tagged MyD88, TRAF3, TRAF6, or TAB2 were provided by Sumit Chanda. Plasmids expressing FLAG-tagged TRAF2 and TAK1 were provided by Michael Karin. A plasmid expressing FLAG-tagged TAB1 was provided by Jun Ninomiya-Tsuji (17). Mutations were introduced into the coding sequences of pVphu and BST2 using the QuikChange kit from Stratagene and were verified by nucleotide sequencing. To introduce the A10,14,18F mutation into the provirus, the EcoRI-NheI fragment of pNL4-3 was mutagenized using overlap PCR, cloned using the TOPO TA method (Invitrogen), verified by sequencing, and used to replace the EcoRI-NheI fragment of pNL4-3 Udel. HEK293T cells and HeLaP4.R5 cells (transduced to express CD4 and CCR5) were obtained from Ned Landau. HT1080 cells were obtained from the ATCC. Rabbit antisera to BST2 and Vpu were obtained from the NIH AIDS Reference and Reagent Repository and contributed by Klaus Strebel. Antibody to HIV-1 p24 capsid was purchased from the ATCC. Anti-FLAG (monoclonal antibody M2) was purchased from Sigma. Immunoblot detection antibodies (horseradish peroxidase-conjugated anti-mouse and anti-rabbit IgG) were purchased from Thermo Scientific. Reagents for enhanced chemiluminescence were purchased from Pierce. BAPTA-AM was purchased from Invitrogen. (5Z)-7-oxozeaenol and phorbol 12-myristate 13-acetate (PMA) were purchased from Sigma. Tumor necrosis factor alpha (TNF-α) was purchased from R&D Systems.

Transfections.

Cells (HEK293T, HeLa clone P4.R5, and HT1080) were transfected using Lipofectamine2000 according to the manufacturer's instructions, including the suggested amounts of total DNA and lipid according to well size. The amount of DNA in each transfection was adjusted to the suggested maximum using an empty plasmid vector, usually pcDNA3.1.

Luciferase assays.

Cells (HEK293T) were transfected in duplicate using 24-well plates (200 ng of NF-κB-luc plasmid per well or as indicated in the figure legends) and including an expression plasmid for β-galactosidase (220 ng per well or as indicated in the figure legends) to normalize the data for variations in transfection efficiency and cell number. HeLa cells were transfected in quadruplicate and without a β-galactosidase expression plasmid. A portion of the cells from each well was lysed in Promega Bright Glo reagent for measurement of luciferase activity, and the remaining portion was lysed in Promega buffer E387A including protease inhibitor cocktail (Roche) before the addition of Galacton-Plus substrate for measurement of β-galactosidase activity (Applied Biosystems). Luciferase and β-galactosidase activities were measured as luminescence using a SpectraMax plate reader (Molecular Devices). The amounts of β-galactosidase expression plasmid were within the linear dose-response range of the assay, indicating that normalization of luciferase activity to that of β-galactosidase was appropriate.

Virion release assays.

Virion release assays were done as previously reported using HEK293T cells transfected with plasmids expressing BST2, HIV-1 pNL4-3, Vpu (pVphu), or the indicated mutated versions of these plasmids (16). The day after transfection, supernatants were collected and virus particles were pelleted through a 20% sucrose cushion before assay of capsid antigen by enzyme-linked immunosorbent assay (ELISA) (Perkin-Elmer).

Immunofluorescence microscopic and flow cytometric assays.

For immunofluorescence microscopy, HT1080 cells were transfected to express wild-type BST2 or various mutants. The next day, the cells were fixed, permeabilized, and stained using indirect immunofluorescence using the rabbit antiserum to BST2 and a monoclonal antibody to transferrin receptor as previously described (16). Images were acquired under identical conditions as a Z-series using a wide-field Olympus fluorescence microscope and SlideBook software. The images were deconvolved using the nearest-neighbor method, and projection images of the entire Z-series of representative fields were assembled into a composite figure using Adobe Photoshop software. For flow cytometry, HEK293T cells were transfected in wells of a 12-well plate with 400 ng of a plasmid expressing BST2 or various mutants along with 400 ng of a plasmid expressing green fluorescent protein (GFP). The next day, the cells were stained with a monoclonal antibody to BST2 (clone RS38) conjugated to allophycocyanin (APC) (BioLegend, San Diego, CA) and then fixed in 0.5% formaldehyde in phosphate-buffered saline and analyzed using an Accuri C6 flow cytometer (BD Biosciences). Cells were gated based on forward- and side-scatter characteristics, and the data were depicted as two-color dot plots of APC (BST2) versus GFP relative fluorescence intensity.

Immunoprecipitation.

HEK293T cells were transfected with 8 μg of a plasmid expressing the FLAG-tagged protein as indicated and 12 μg of a plasmid expressing BST2 or the indicated mutants in 10-cm² dishes. The next day, the cells were lysed in buffer containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, and 5% glycerol supplemented with protease inhibitor cocktail (Roche Diagnostics) for 30 min on ice and 5 min at room temperature. Lysates were cleared by centrifugation at 16,000 × g at 4°C for 10 min and then incubated with anti-FLAG M2-conjugated magnetic beads (Sigma) for 2 to 3 h at 4°C. Before immunoprecipitation, the beads were blocked with 2% bovine serum albumin in lysis buffer. After immunoprecipitation, the beads were washed three times (using lysis buffer as described above but containing 250 mM NaCl), and the precipitated proteins were eluted with Laemmli buffer and analyzed by immunoblotting.

Statistics.

All graphed values with the exception of those in Fig. 9 are the average of duplicates; error bars represent plus and minus one average deviation. The values in Fig. 9 are the average of quadruplicates; error bars represent plus and minus one standard deviation. Luciferase/β-galactosidase ratios were calculated for each well of the duplicates or quadruplicates before calculating the averages and the errors using Excel (Microsoft). The results shown are representative of at least two experiments.

Fig 9.

Effect of Vpu on the induction of NF-κB activity under conditions of endogenous BST2 expression. HeLa cells, which express endogenous BST2, were transfected with the indicated amounts of proviral plasmids, either wild-type HIV-1 or the indicated vpu mutants, along with 150 ng of the NF-κB-luc indicator plasmid. Mock indicates transfection of cells with the NF-κB-luc indicator plasmid but no proviral DNA. The next day, the cells were lysed and luciferase activity was measured by luminometry (left panel). The cell lysates were also analyzed by immunoblotting to detect HIV-1 Gag (p55 precursor and p24) and Vpu (right panel).

RESULTS

BST2 induces NF-κB activity.

We used reporter assays in which the expression of firefly luciferase is driven by a promoter including recognition sites for the transcription factor NF-κB or AP1 or the response element (ISRE) for interferon-stimulated genes (ISGs). The expression of BST2 by transient transfection in HEK293T cells, which lack endogenous BST2, induced NF-κB activity in a dose-dependent manner (Fig. 1A). AP1 activity was minimally if at all induced by BST2, although it was induced by phorbol myristic acetate (PMA) (Fig. 1B). The ISRE indicator was also not induced by BST2, although it was induced by a constitutively active mutant of interferon regulatory factor 3 (IRF-3 5D) (Fig. 1B) (18). The extent of NF-κB activity induced by BST2 was less than that induced by soluble TNF-α but comparable to that induced by the overexpression of transforming growth factor β (TGF-β)-activated kinase 1 (TAK1) (also known as mitogen-activated protein [MAP] kinase kinase kinase 7). In contrast to the expression of BST2, the overexpression of TAK1 induced AP-1 activity (Fig. 1B). These data indicate that the expression of BST2 specifically induces NF-κB activity and suggest that this induction is in some manner mechanistically distinct from the MAP kinase (MAPK) pathway. The data also suggest that this induction does not cross talk with pathways leading to type I IFNs or ISGs.

Fig 1.

BST2 induces NF-κB activity. (A) BST2-mediated induction of NF-κB activity compared to the induction of AP-1 activity and expression driven by an interferon-stimulated response element (ISRE). HEK293T cells were transfected in wells of a 24-well plate with plasmids encoding firefly luciferase under the control of promoters responsive to the various transcription factors (200 ng) together with increasing amounts of a plasmid expressing BST2; a plasmid expressing β-galactosidase (20 ng) was included as a normalization control. The next day, the cells were lysed and luciferase and β-galactosidase activities were measured by luminometry. Data are the ratio of relative light units (RLU) measured for luciferase activity to RLU measured for β-galactosidase activity. (B) Controls for the various indicators shown as fold inductions compared to the indicator alone. A plasmid expressing TGF-β-activated kinase 1 (TAK1) was used at a 30-ng or 680-ng amount, as indicated. TNF-α (TNF) was used at 1 ng/ml, and phorbol myristic acetate (PMA) was used at 100 ng/ml. A plasmid expressing IRF-3 5D was used (400 ng) as a positive control for the ISRE indicator. Data are the average of duplicates.

The YxY sequence in the BST2 CD is required for signaling but dispensable for restriction.

We examined mutations in the cytoplasmic domain (CD) of BST2 for their effect on the induction of NF-κB (Fig. 2A). Mutations of putative ubiquitination sites, including the N-terminal STS sequence as well as cysteines and lysines throughout the CD (16), modestly increased the protein's ability to induce NF-κB. Conversely, mutation of the YxY sequence, previously associated with the binding of BST2 to clathrin adaptors (19, 20), nearly abolished induction. We compared these effects on signaling to those on the activity of BST2 as a restriction factor by coexpressing the BST2 mutants with an HIV-1 provirus lacking vpu (Fig. 2B). Mutants with mutations of putative ubiquitination sites restricted as well as or slightly less well than wild-type BST2, whereas the YxY sequence mutant restricted slightly better than the wild type, probably due to its relatively increased level of expression (Fig. 2B). These data suggest a lack of correlation between the activities of restriction and induction of NF-κB. In particular, the data indicate that a conserved feature of the protein, the YxY sequence in the cytoplasmic domain, is dispensable for restriction yet required for the induction of NF-κB.

Fig 2.

The YxY sequence in the BST2 cytoplasmic domain is required for signaling but dispensable for restriction. (A) HEK293T cells were transfected in wells of a 24-well plate with a plasmid encoding firefly luciferase under the control of a minimal promoter containing binding sites for NF-κB (200 ng) along with increasing amounts of a plasmid expressing BST2 (indicated); a plasmid expressing β-galactosidase (220 ng) was included as a normalization control. The next day, the cells were lysed and luciferase and β-galactosidase activities were measured by luminometry. Data are the ratio of relative light units (RLU) measured for luciferase activity to RLU measured for β-galactosidase activity and are the average of duplicates. (B) Restriction of HIV-1 virion release by BST2. HEK293T cells were transfected in wells of a 12-well plate with a plasmid encoding an HIV-1 provirus lacking vpu (1.6 μg) along with a plasmid encoding either no BST2, wild-type BST2, or the indicated mutants (30 ng). The next day, the supernatants were removed and the virions pelleted though a 20% sucrose cushion and then resuspended in the original culture volume before measurement of p24 Gag (capsid) concentration by ELISA. Data are the average of duplicates. (C) Cells from the experiment shown in panel B were analyzed by SDS-PAGE and immunoblotting for BST2 and p55 Gag precursor. In all panels, Y6/8A is a BST2 mutant in which the tyrosines at positions 6 and 8 are replaced with alanines, STS is a mutant in which the serine and threonine residues at positions 3 to 5 are replaced with alanines, and KKCC is a mutant in which the lysine residues at positions 18 and 21 are replaced with arginines and the cysteine residues at positions 9 and 20 are replaced with alanines. “KCST-less” is a mutant encoding the combined substitutions of STS and KKCC.

Mutations in the BST2 ectodomain tetramerization interface and the GPI anchor site also separate signaling from restriction.

We looked next at mutations within the BST2 ectodomain, with initial attention to those that disrupt the ability of the protein to multimerize and/or restrict virion release. The BST2 ectodomain has been observed to adopt two forms by X-ray crystallography: an oxidized, parallel, dimeric coiled coil and, under reducing conditions, a tetramer in which the N-terminal regions of the dimer are antiparallel (21–23). The extended coiled coil of the ectodomain has been proposed to facilitate partitioning of the N-terminal transmembrane helix and the C-terminal GPI anchor into different membranes, viral or cellular, during budding, thus enabling the oxidized dimer to cross-link budded virions to the cell (21). The reduced tetramer can potentially cross-link membranes by more complex topologies including all eight of its membrane anchors (23). Sets of mutations in the N-terminal third of the ectodomain, as well as in heptad repeats in the C-terminal two-thirds, reportedly abrogate restriction (21). In contrast, mutation of L70 to aspartate, which is sufficient to disrupt the tetramer interface, reportedly affects restriction only modestly (22).

We tested these ectodomain mutants for their ability to induce NF-κB activity. All the mutants were impaired in signaling activity, but “set2,” which encodes multiple substitutions at the C-terminal end of the coiled coil, and L70D were nearly completely defective (Fig. 3A). Consistent with previous reports, the “set” mutants were markedly impaired for restriction of virion release, whereas the L70D mutant was only modestly impaired (Fig. 3B). Thus, residue L70, which is well conserved among diverse mammalian species, is required for the induction of NF-κB but largely dispensable for the restriction of virion release. Since L70 is the key hydrophobic residue of the tetramer interface (22), tetramerization of BST2 might be specifically required for its signaling activity.

Fig 3.

Mutations in the BST2 ectodomain coiled coil separate signaling from restriction. (A) HEK293T cells were transfected in wells of a 24-well plate, and the induction of NF-κB was measured as described in the legend of Fig. 2. BST2 “Set” mutants contain multiple substitutions in the BST2 ectodomain and are described further in the text. L70D is a mutant in which the leucine residue at position 70 is replaced with an aspartic acid residue. Data are the ratio of relative light units (RLU) measured for luciferase activity to RLU measured for β-galactosidase activity and are the average of duplicates. (B) Restriction of HIV-1 virion release by BST2. Virion release from transfected HEK293T cells was measured as described the legend of Fig. 2. (C) Cells from the experiment shown in panel B were analyzed by SDS-PAGE and immunoblotting for BST2 and p55 Gag precursor.

We next turned our attention to the roles of the ectodomain cysteine residues that mediate dimerization, the asparagine residues that are substrates for glycosylation, and the C-terminal GPI anchor. BST2 mutated at cysteine residues 53, 63, and 91 (BST2-C3A) was modestly impaired in its ability to induce NF-κB activity, and its ability to restrict virion release was similarly impaired (Fig. 4A and B). In comparison to BST2-L70D, BST2-C3A displayed similar restrictive activity, but it retained relatively greater signaling activity. BST2 mutated at asparagine residues 65 and 92 (BST2-N2Q) was markedly impaired in its ability to induce NF-κB activity; it retained substantial ability to restrict virion release, although it was less restrictive than the L70D and C3A mutants (Fig. 4A and B). To evaluate the role of the GPI anchor, we created an in-frame deletion mutant of BST2, Δ156-162, in which all potential serine residues that could serve as sites of GPI modification were removed, while the hydrophobic C-terminal sequence that could serve as a membrane anchor was preserved (24). This mutant induced NF-κB activity almost as well as the wild-type protein (Fig. 4A), but it was completely defective for restriction of virion release (Fig. 4B). Like the L70D mutation, the deletion of residues 156 to 162 separated the signaling and restriction phenotypes, but in a converse manner: the GPI anchor is required specifically for the restriction of virion release, whereas the tetramerization interface is required specifically for signaling.

Fig 4.

The GPI anchor site of BST2 is required for restriction but not signaling. (A) HEK293T cells were transfected in wells of a 24-well plate, and the induction of NF-κB was measured as described in the legend of Fig. 2. Data are the ratio of relative light units (RLU) measured for luciferase activity to RLU measured for β-galactosidase activity and are the average of duplicates. (B) Restriction of HIV-1 virion release by BST2. Virion release from transfected HEK293T cells was measured as described in the legend of Fig. 2. (C) Cells from the experiment shown in panel B were analyzed by SDS-PAGE and immunoblotting for BST2 and p24 Gag (capsid). In all panels, L70D is as described in the legend of Fig. 2, C3A is a mutant in which the cysteine residues at positions 53, 63, and 91 are replaced with alanines, N2Q is a mutant in which the asparagine residues at positions 65 and 92 are replaced with glutamines, and Δ156-162 is a mutant encoding an in-frame deletion at the GPI anchor modification site.

The signaling-impaired BST2 mutants L70D and C3A are expressed aberrantly at the cell surface, but the Y6/8A mutant is not.

We sought to correlate BST2 activity with subcellular localization and cell surface expression. We used immunofluorescence microscopy to colocalize BST2 with the transferrin receptor, a marker of recycling endosomes, including those in the perinuclear region (Fig. 5A). Whereas wild-type BST2 and the N2Q, Y6/8A, and Δ156-162 mutants were each expressed predominantly within endosomal structures as previously described (2, 25), the L70D and C3A mutants exhibited intense expression along the cell perimeter, presumably at the plasma membrane. We confirmed this using flow cytometry to detect quantitatively the expression of BST2 at the cell surface (Fig. 5B). The BST2 N2Q and Y6/8A mutants were expressed at the cell surface indistinguishably from the wild-type protein, but the L70D and C3A mutants were expressed at aberrantly high levels. In contrast, the Δ156-162 mutant was expressed at lower surface levels than the wild type. These data suggested that the L70D and C3A mutants might signal inefficiently due to aberrant concentration at the cell surface, presumably due to inefficient internalization. In contrast, the Y6/8A mutant, although defective for signaling, had no such defects in subcellular localization.

Fig 5.

Signaling-impaired mutants BST2-L70D and -C3A are expressed at supranormal levels at the cell surface, but the signaling-defective BST2-Y6/8A is not. (A) Immunofluorescence images of HT1080 cells transfected to express wild-type BST2 (WT) or the indicated mutants. The cells were fixed, permeabilized, and stained for BST2 (red) or transferrin receptor (green). For each field, a projection image of a deconvolved Z-series is shown. (B) Flow cytometric analysis of BST2 expression at the cell surface. HEK293T cells were transfected to express BST2 or the indicated mutants along with GFP and then stained for cell surface BST2 and analyzed by two-color flow cytometry. Wild-type BST2 alone is shown in the upper left panel. These data for wild-type BST2 are also shown in each of the other panels in purple, with the data for each of the indicated BST2 mutants overlaid in green. Regions of similarity in the two-color plots between the mutants and the wild type appear black.

BST2 signals through TAK1 and the canonical NF-κB activation pathway.

The inability of BST2 to induce AP1 activity (Fig. 1) suggests that the induction of NF-κB might be to some extent distinct from the MAPK pathway. Nonetheless, a specific inhibitor of the kinase TAK1, which serves as a signaling intermediate downstream of TRAF6 and which activates IκB kinase (IKK) (26), inhibited the induction of NF-κB activity by BST2 as well as by TNF-α (Fig. 6A).

Fig 6.

Pathways of BST2-mediated signaling. (A) Induction of NF-κB activity by BST2 is decreased by an inhibitor of TAK1 kinase and by intracellular chelation of calcium. Left panel, HEK293T cells were transfected in wells of a 24-well plate with a plasmid encoding firefly luciferase under the control of a minimal promoter containing binding sites for NF-κB (80 ng) along with increasing amounts of a plasmid expressing BST2; a plasmid expressing β-galactosidase (40 ng) was included as a normalization control. Immediately after the transfection, 1 μM (5Z)-7-oxozeaenol (5OZ), a selective inhibitor of TAK1), 40 μM BAPTA-AM, a cell-permeative calcium chelator, or nothing was added. The next day, the cells were lysed, and luciferase and β-galactosidase activities were measured by luminometry. Data are the ratio of relative light units (RLU) measured for luciferase activity to RLU measured for β-galactosidase activity and are the average of duplicates. Right panel, fold induction of NF-κB activity in HEK293 cells transfected as described above but without a plasmid expressing BST2 and treated with 1 ng/ml TNF either alone or with 5OZ or BAPTA-AM. (B) Dominant-negative IκB blocks the induction of NF-κB activity by BST2. HEK293T cells were transfected and analyzed as for panel A, but the transfection mixtures included 200 ng of NF-κB-luc indicator, 220 ng of β-galactosidase expression plasmid, and, in addition to a dose range of plasmid expressing BST2, 20 ng of a plasmid expressing either wild-type IκB or dominant-negative IκB (DN-IκB). (C) BST2 coimmunoprecipitates with MyD88, TAB1, TAB2, and TAK1. HEK293T cells were transfected to express FLAG-tagged MyD88, TRAF2, TRAF3, TRAF6, TAB1, TAB2, or TAK1 and untagged BST2. The next day, the cells were lysed and immunoprecipitated with anti-FLAG conjugated beads. Immunoprecipitates (IP-FLAG) were analyzed for the FLAG-tagged proteins and BST2 as indicated. (D and E) Interaction with TAK1 and TAB1 requires the tyrosine residues at positions 6 and 8 but not the leucine residue at position 70 in BST2. HEK293T cells were transfected to express either FLAG-tagged TAK1 (D) or FLAG-tagged TAB1 (E) and untagged BST2 or the indicated mutants. The next day, the cells were lysed and immunoprecipitated with anti-FLAG conjugated beads. Lysates and immunoprecipitates were analyzed for the FLAG-tagged proteins and BST2 as indicated.

We noticed that the Δ156-162 mutant did not generate the higher-molecular-weight species attributed to fully glycosylated BST2 (Fig. 4C) (27); this suggested that this GPI anchor-defective BST2 mutant might not escape the endoplasmic reticulum (ER). Since the Δ156-162 mutant induced NF-κB activity nearly as efficiently as the wild-type protein, we considered that BST2 might signal from the ER by triggering the ER stress response, perhaps due to its unusual topology and oligomerization properties. ER stress leads to the induction of NF-κB activity through several pathways, one of which involves the release of calcium from the ER (reviewed in reference 28). Consistent with signaling through the ER stress response, the cell-permeative calcium chelator BAPTA-AM effectively inhibited the induction of NF-κB activity by BST2 (Fig. 6A), whereas it affected induction by TNF-α minimally if at all.

The transcriptional activity of NF-κB is induced by several mechanisms, many of which rely on the degradation of the inhibitory protein IκB, which sequesters preformed NF-κB in an inactive complex in the cytoplasm (29, 30). A dominant-negative mutant of IκB, in which the serine residues phosphorylated by IKK to enable binding to β-TrCP and an SCF E3 ubiquitin ligase complex are replaced with alanines (11), cannot be degraded in response to IKK, and this protein blocked the induction of NF-κB by BST2 (Fig. 6B). The overexpression of wild-type IκB was less effective. These data support a “canonical” pathway of NF-κB activation by BST2 dependent on the degradation of IκB.

Consistent with the inhibitory activity of IκB, we identified BST2 in immunoprecipitates of several proteins that stimulate NF-κB activation via the canonical pathway, including MyD88, TAB1, TAB2, TAK1, and to a lesser extent TRAF2 (Fig. 6C). The interactions of BST2 with TAK1 and TAB1, a regulatory pseudophosphatase that associates with TAK1, required the tyrosine residues at positions 6 and 8 in BST2 but not the leucine residue at position 70 (Fig. 6D and E). These data suggest that the interaction of BST2 with a TAK1/TAB1 signaling complex specifically requires the YxY sequence in the cytoplasmic domain of BST2.

HIV-1 Vpu inhibits BST2-mediated signaling.

Vpu counteracts the activity of BST2 as a viral restriction factor by decreasing the concentration of the protein at the cell surface, in some cases with concomitant degradation (2, 31). These effects require an interaction between the transmembrane domains of the two proteins (32), which is mediated by an AxxxAxxxA sequence on one face of the Vpu transmembrane helix (33, 34), and, in the case of degradation, the interaction of Vpu with β-TrCP (31, 35, 36), the substrate adaptor for the same SCF E3 ubiquitin ligase complex that ubiquitinates IκB (37). Vpu inhibited the induction of NF-κB activity by BST2 (Fig. 7A), and this activity required the β-TrCP binding motif in Vpu (serines 52 and 56 of the DSGxxS sequence). The inhibition of NF-κB induction was independent of the interactive face of the Vpu transmembrane helix (see the “VpuA3/F” mutant). This initially counterintuitive result is probably due to the ability of high levels of Vpu to saturate β-TrCP, thus blocking the degradation of IκB and inhibiting the activation of NF-κB by a general mechanism that does not require direct interaction with BST2 (38). The effects of Vpu on the expression of BST2 were as anticipated: the expression of BST2 was decreased by wild-type Vpu, increased by Vpu-2/6 (the mutant unable to interact with β-TrCP), and unaffected by Vpu mutants containing the triple A/F substitution, which disrupts the interaction with BST2 (Fig. 7B).

Fig 7.

HIV-1 Vpu inhibits BST2-mediated signaling. (A) HEK293T cells were transfected in wells of a 24-well plate with a plasmid encoding firefly luciferase under the control of a minimal promoter containing binding sites for NF-κB, along with increasing amounts of a plasmid expressing BST2 and fixed amounts (65 ng) of plasmids expressing either Vpu, the indicated Vpu mutants, HLA-A2, or nothing (BST2 only). Vpu3A/F is the A10,14,18F mutant of the interactive face of the Vpu transmembrane helix, Vpu2/6 is the S52,56N mutant of the β-TrCP binding site, and Vpu3A/F,2/6 is a combination mutant. A plasmid expressing β-galactosidase was included as a normalization control. Data are the ratio of relative light units (RLU) measured for luciferase activity to RLU measured for β-galactosidase activity and are the average of duplicates. (B) Cells from parallel transfections set up as for panel A and containing 10 ng of BST2 expression plasmid were analyzed by SDS-PAGE and immunoblotting for BST2.

BST2 signaling is augmented by viral gene expression.

The signaling function of BST2 might be united with its activity as a restriction factor if the protein functions as a “virus sensor” that triggers a host response to viral replication. In particular, BST2 aggregates at sites of viral assembly by an unclear mechanism (39–42), and this phenomenon could enable BST2 to “sense” the assembly of enveloped viruses. To test this, we asked whether HIV-1 gene expression would augment the induction of NF-κB by BST2 (Fig. 8). Expression of a vpu-negative HIV-1 genome augmented the induction of NF-κB activity by BST2 in a synergistic manner. Vpu, when provided in trans, suppressed this augmentation, but the suppression was substantially less when the β-TrCP binding motif in Vpu was mutated (Fig. 8A and B). As was seen for signaling by BST2 alone, the interactive face of the Vpu transmembrane helix was largely dispensable for suppression of virally augmented NF-κB activity. When these mutations were studied in the context of the complete viral genome, however, a role for the interactive alanine face of the Vpu was observed, and the viral genome expressing wild-type Vpu suppressed the activation of NF-κB below that induced by BST2 alone (Fig. 8C and D). These data support a model of BST2 as a virus sensor that leads to the activation of NF-κB. The data also suggest that HIV-1 Vpu antagonizes this effect primarily through its ability to usurp the SCF E3 ubiquitin ligase complex that leads to NF-κB activation.

Fig 8.

BST2 signaling is augmented by viral gene expression, and the augmentation is counteracted by Vpu. (A) Induction of NF-κB activity by BST2 alone or together with cotransfected HIV1 proviral DNA lacking vpu (Δvpu), either alone or together with a plasmid expressing either wild-type (wt) Vpu, Vpu2/6, or Vpu3A/F. HEK293T cells in wells of a 24-well plate were transfected with 80 ng NF-κB-luc and 40 ng β-galactosidase plasmids, 620 ng of Δvpu proviral plasmid where indicated, 60 ng Vpu expression plasmid or the indicated Vpu mutants, and the indicated amounts of BST2 expression plasmid. NF-κB activity was measured as described in the legend of Fig. 1. Data are the average of duplicates. (B) Immunoblot of cells from the experiment of panel A probed for BST2, p24 Gag, and Vpu. (C) Induction of NF-κB activity by BST2 alone or together with cotransfected wild-type HIV-1 proviral plasmid encoding vpu (WT HIV), Δvpu, or proviral plasmids encoding the vpu mutant 2/6 (S52,56N) or 3A/F (A10, 14, 18F). The experiment was done as for panel A except that 640 ng of proviral plasmids was used. NF-κB activity was measured as described above; data are the average of duplicates. (D) Immunoblot of cells from the experiment of panel C probed for BST2, p24 Gag, and Vpu.

The induction of NF-κB activity by HIV1 gene expression in cells that express BST2 endogenously is inhibited by Vpu.

All of the above-described experiments utilized transient transfection of HEK293T or HT1080 cells to control the expression of BST2. To further validate the effects of Vpu on the induction of NF-κB activity, we repeated the experiments using complete viral genomes in HeLa cells, which express BST2 constitutively (Fig. 9). Expression of the HIV genome by transfection induced NF-κB activity in a dose-dependent manner, and this induction was greater in the absence of vpu. Maximal induction occurred when the viral genome encoded the Vpu2/6 mutant, indicating that the inhibition of NF-κB activity by Vpu requires its β-TrCP binding motif. Although not as active as the vpu-negative mutant, the viral genome encoding the VpuA3/F mutant nonetheless induced NF-κB activity to an extent greater than the wild-type virus; this suggests that Vpu's ability to inhibit the activation of NF-κB in the setting of viral replication and endogenous expression of BST2 is partly due to its ability to interact with BST2 and not solely due to saturation of β-TrCP.

DISCUSSION

We have confirmed the ability of the antiviral restriction factor BST2 to induce activity of the NF-κB family of transcription factors (7). This activity of BST2 requires certain sequences that are dispensable for the protein's ability to restrict the release of enveloped viruses from infected cells, namely, a YxY sequence in the protein's cytoplasmic domain and a key hydrophobic residue, L70, that is required for tetramerization of the BST2 coiled-coil ectodomain (22). Conversely, at least one feature of the protein that is required for restriction is nearly dispensable for the induction of NF-κB activity: the GPI anchor sequence. BST2 thus appears to contain conserved features that are specifically required for either signaling or restriction, suggesting that both properties of the protein are biologically important.

What is the mechanism of induction of NF-κB by BST2? While this remains to be better defined, sequences in both the cytoplasmic domain and the ectodomain of the protein provide revealing clues. The YxY sequence in the cytoplasmic domain is critical to signaling. This sequence is well conserved among placental mammals, although it is missing in cats and at least one strain of mice due to sole expression of a BST2 isoform using an initiator codon corresponding to Met13 in the human protein (43, 44). The only previously described function of the YxY sequence is its ability to serve as an endocytosis and trans-Golgi retrieval signal through interaction with subunits of the clathrin adaptor protein complexes (19, 20). One possibility, then, is that BST2 signals from an endosomal compartment and that the YxY sequence is required to direct it there. The data here weigh against this, however, because the Y6/8A mutant of BST2 is localized normally within endosomes and is not expressed at supranormal levels at the cell surface. Alternatively, the YxY sequence could mediate an interaction with a signaling adaptor or kinase. The coimmunoprecipitation data here indicate that this is indeed the case: BST2 interacts with the kinase TAK1 and the TAK-associated regulatory cofactor TAB1 in a YxY-dependent manner. Apparently, the critical role of the YxY sequence is not to direct the internalization of BST2 from the plasma membrane but rather to direct the interaction of BST2 with the TAK1/TAB1 signaling complex.

Nonetheless, localization of BST2 on internal cellular membranes might well be critical for signaling. Residue L70, which is also well conserved among diverse mammalian species and required for tetramerization of the BST2 ectodomain in vitro, is defective for signaling and strikingly mislocalizes to the plasma membrane. The cysteine-linked dimerization mutant C3A is also impaired for signaling and mislocalizes to the plasma membrane, although each of these defects is not as severe as in the case of the L70D mutant. These phenotypes of the L70D and C3A mutants suggest that multimerization and internalization are prerequisites for signaling (model in Fig. 10). Potentially consistent with this, the Δ156-162 mutant of BST2, which lacks the potential for GPI modification, is poorly expressed at the cell surface yet signals nearly normally. Moreover, this mutant seems not to generate the higher-molecular-weight forms associated with fully mature glycosylation, and it appears to colocalize relatively poorly with transferrin receptor. These observations led us to speculate that the Δ156-162 mutant might not escape the ER efficiently, and if so, its signaling activity might reveal an ER-based mechanism. The inhibition of BST2 signaling by treatment of cells with the calcium chelator BAPTA-AM supports the possibility that NF-κB induction occurs via an ER stress response. A hypothesis consistent with all of these data is that BST2 signals from the ER via TAK1, but this requires further evaluation. Ultimately, the induction of NF-κB activity by BST2 depends on the canonical pathway involving the degradation of IκB, because it is blocked by expression of an IκB mutant that is able to bind NF-κB but unable to bind β-TrCP. That BST2 activates NF-κB via the canonical pathway is consistent with its interaction with the TAK1/TAB1 signaling complex.

Fig 10.

Models for intrinsic signaling by BST2 and for BST2 as a virus sensor. (A) Intrinsic signaling. BST2 is internalized into endosomal compartments and/or remains within the ER, where it recruits a signaling complex including TAK1, TAB1, and TAB2, and potentially MyD88 and TRAF2, leading to activation of NF-κB by a canonical pathway. The activation of NF-κB is inhibited by substitutions in the BST2 ectodomain that impair dimerization (C3A) or tetramerization (L70D) and by substitutions in the cytoplasmic domain that impair binding to the TAK1/TAB signaling complex (Y6/8A). HIV Vpu inhibits the activation of NF-κB in a manner dependent upon its ability to bind β-TrCP. (B) Virus sensing. HIV assembly and/or gene expression stimulates the induction of NF-κB by an as-yet-unclear mechanism and at unclear intracellular sites. This stimulation allows BST2 to “sense” the virus, resulting in increased NF-κB activity, which presumably induces the production of inflammatory cytokines by the infected cell. Vpu inhibits this, again dependent upon its ability to bind β-TrCP but also to a lesser extent upon its ability to bind BST2.

Induction of NF-κB activity by BST2 is a compelling phenomenon if this represents a mechanism for the cell to sense viral replication. Consistent with this notion, the expression of HIV-1 augmented the activation of NF-κB by BST2, as long as the wild-type Vpu protein was not expressed. HIV-1 Vpu can antagonize the induction of NF-κB by usurping β-TrCP-based SCF E3 ubiquitin ligase complexes, an observation consistent with activation of NF-κB by a canonical pathway. The direct interaction between Vpu and BST2 seems relatively less important to the inhibition of signaling. Nonetheless, the interactive face of Vpu's transmembrane helix contributes to the antagonism of signaling when Vpu is expressed in cis in the viral genome, and it is essential for antagonizing BST2-mediated restriction (32–34). These observations suggest that BST2's activity as a restriction factor is a biologically important function that is specifically counteracted by Vpu and is independent of its signaling activity.

How does viral gene expression augment the signaling activity of BST2, and what are the consequences for the host cell and organism? The notion that aggregation of BST2 during viral assembly would stimulate signaling is attractive, because BST2 aggregates at viral assembly sites (39–42) and mutational analysis suggests the potential importance of BST2 oligomerization in signaling. An alternative possibility is that viral augmentation of BST2-dependent activation of NF-κB requires the viral envelope glycoprotein (Env). Consistent with this hypothesis, the Env proteins of primate lentiviruses have recently been reported to induce NF-κB activity by a signaling mechanism involving TAK1 (45). Moreover, expression of high levels of Env, a transmembrane protein, could induce an ER stress response in cells primed by the expression of BST2.

While this paper was in revision, two independent reports highlighted the ability of BST2 to induce NF-κB activity (46, 47). Together with the present paper, these reports indicate a critical role for the YxY sequence of BST2 in signaling, support a canonical pathway of NF-κB induction, and suggest that BST2 signals at least in part via multimerization and via TAK1. Less certain is whether signaling activity is a general property of mammalian BST2 orthologues and whether viral proteins other than Vpu antagonize this activity. Also uncertain are the intracellular sites from which BST2 signals and how viral gene expression is detected and augments signaling. Although we have shown discordances between the intrinsic signaling and the restrictive activities of BST2 through the analysis of mutants, whether the ability of the protein to sense viral gene expression correlates with its ability to restrict virion release remains to be clarified.

The augmentation of BST2-mediated signaling by viral replication represents a novel virus-sensing mechanism for the infected cell. In one possible scenario, cells primed to express BST2 by a systemic or local interferon response would respond to infection by HIV-1 or another enveloped virus by activating a transcriptional program driven by increased NF-κB activity. This program would likely lead to the expression of inflammatory cytokines (47, 48), which would in turn provoke a local cellular response at the specific site of viral replication in the host organism. Such a response would presumably contribute to the antiviral host defense.