1. Introduction
As recently as 1980, the measles virus (MeV) killed 2,600,000 people per year, however, the effective use of a live attenuated vaccine has led to a significant drop in fatalities. Unfortunately, MeV remains endemic in many developing countries, causing over 100,000 deaths per year (World Health Organization statistics, [
1]). The MeV is a small RNA virus, classified in the genus
Morbillivirus, which encodes six transcription units and at least eight proteins. Two of these proteins, the fusion (F) and haemagglutinin (H) proteins, are glycoproteins, embedded as functional oligomers in the surface of the viral envelope [
2]. H directs attachment to one of two known receptors, SLAMF1 (signalling lymphocyte activation molecule F1) or nectin-4, while F initiates the membrane fusion events required for genome invasion [
2,
3]. SLAMF1 and nectin-4 are found on separate cells in vivo, immune and epithelial, respectively, contributing to a MeV life cycle that involves infection of both the lymphatic system and various epithelia [
3]. One of the characteristic features of a MeV infection is the formation of syncytia, or multinucleated cells, both in vitro and in vivo, i.e., in the lymph node, thymus, and respiratory tract [
4,
5]. Syncytia formation, a process which occurs when uninfected cells become fused by neighboring, F- and H-expressing infected cells is possible because MeV glycoproteins are functional at neutral pH [
6]. Morbilliviruses can also spread by canonical particle formation (viral budding) and both processes contribute to viral dissemination and pathogenesis in the host, however, the mechanisms underpinning the equilibrium between budding and cell–cell fusion are poorly understood.
Although MeV can potently inhibit innate immune signaling, particularly through the action of its accessory protein V [
7], there is evidence that a robust type I interferon (IFN) response is mounted in infected cells, both in vivo and in vitro, i.e., in patient peripheral blood mononuclear cells (PBMCs) [
8], mouse models [
9], and primary dendritic cell cultures [
10]. Amongst the upregulated genes observed in IFN-simulated cells are a range of well-characterized interferon stimulated genes (ISGs) that facilitate an antiviral state, including the restriction factor bone marrow stromal antigen 2
(BST2) and its resulting protein, known as BST2, Tetherin or CD317.
BST2 is a well-characterized IFN-inducible protein that is capable of restricting the release of a broad range of enveloped viruses [
11]. Its principal mechanism of action is to tether nascent virus to the cell surface, preventing effective release [
11,
12]. The antiviral properties of BST2 were first characterized by Neil et al., who, at the same time, identified a virally encoded antagonist, HIV-1 Vpu [
12]. The BST2 protein is capable of restricting both cell-free and cell-cell routes of HIV-1 transmission [
13]. Its antagonist, Vpu, in turn, is capable of redirecting this protein away from sites of HIV-1 budding through the hijack of membrane trafficking [
14]. Since its discovery, BST2-mediated restriction and viral antagonists have been identified in a range of viruses including Ebola virus [
15], hepatitis C virus [
16], dengue virus [
17], and herpes simplex virus [
18]. More recently, BST2 has been characterized as both an important immune sensor, through induction of proinflammatory gene expression via activation of NF-κB [
19], and as a modulator, through immunoglobulin-like transcript 7 (ILT7) interactions [
20]. These signaling pathways are activated in part by phosphorylation of sequences in the cytoplasmic tail of BST2 that resemble hemi-immunoreceptor tyrosine-based activation motifs (hemITAMs) [
19].
Since morbilliviruses, including MeV, frequently remain cell-associated and spread extensively via cell–cell fusion, rather than through cell-free virus, we investigated whether BST2 overexpression was capable of restricting this aspect of the viral life cycle specifically. Using virulent viruses and quantitative cell–cell fusion assays (based on viral glycoproteins from field isolates) we have demonstrated that BST2 from separate mammalian species can restrict morbillivirus cell–cell fusion when overexpressed in cells. This process is dependent on an overall reduction in viral glycoprotein levels, a decrease that is, in part, dependent on the BST2 GPI anchor.
2. Materials and Methods
2.1. Cells
The HEK293T, HEK293T stably expressing human SLAMF1 (293-hSLAM), and Vero stably expressing human or canine SLAMF1 (Vero-h/cSLAM cells) were maintained in DMEM media containing 10% fetal bovine serum (FBS) at 37 °C with 5% CO
2. 293-hSLAMcells were generated, as described previously, using a lentivirus transduction system and maintained with 1 μg/mL Puromycin [
21]. The Vero-h/cSLAM cells were maintained with 0.4 mg/mL geneticin.
2.2. Viruses and Viral RNA Quantification
The MeV-GFP, an EGFP-expressing recombinant MeV (strain IC323), was generated as reported previously [
22]. Virus stocks were grown and titred in Vero hSLAM cells. PPRV, Turkey 2000, a wild-type strain was grown in Vero cSLAMcells. Virus titres were calculated by TCID50 using the Reed-Müench method following a single freeze–thaw cycle at −80 °C. Viral genome was detected using a SYBR-based strand-specific RT-qPCR protocol targeting the N gene, using a standard curve for quantification.
2.3. Plasmids
The BST2 ORFs from human and ovine genes were amplified by RT-PCR from HEK293T and sheep epithelial cell lines, respectively, using Superscript II Reverse Transcriptase (ThermoFisher, Waltham, MA, USA) and Kod HiFi DNA polymerase (MerckMillipore, Burlington, MA, USA). Primer sequences were designed so as to incorporate an N-terminal FLAG tag peptide sequence into the resultant BST2 protein. MeV and PPRV N, F and H ORFs were amplified by RT-PCR from cells infected with the virulent MeV-IC323, MeV-Dublin, and PPRV-Turkey 2000 virus strains, respectively. All cell–cell fusion assays were performed with MeV-Dublin strain constructs. No significant difference was observed in the BST2-mediated restriction of IC323 or Dublin based cell–cell fusion. The Dublin MeV H expression construct was amplified to include an N-terminal HA-tag sequence. The cDNAs were directionally cloned into the multiple cloning sites of the eukaryotic expression vector pcDNA3.1 (ThermoFisher) under the control of a CMV promoter. The MeV F and H constructs expressing these viral glycoproteins with truncated cytoplasmic tails, were generated as described previously [
23,
24], by mutagenic PCR and cloned into pcDNA3.1, as were the BST2 mutants ∆-GPI (lacking 19 C-terminal amino acids) and ∆-TM (lacking 46 N-terminal amino acids). Human SLAM was generated by RT-PCR from Vero-hSLAM cells and cloned into a lentivirus expression system, as described previously [
21]. All primer sequences and restriction endonuclease cloning strategies are available upon request.
2.4. Infections
The 293-hSLAM cells were plated at a density of 1 × 105 cells per well in 24 well dishes. The following day, cells were transfected with 500 ng of plasmid DNA (pcDNA3.1, pcDNA3.1-BST2, pcDNA3.1-∆-GPI or pcDNA3.1-∆-TM) using Transit X-2 transfection reagent (Mirus, Madison, WI, USA) and Optimem (ThermoFisher), as per the manufacturer’s instructions. Twenty-four hours later the media was removed and cells were infected with MeV-GFP at various MOIs (as determined by TCID50) in a 500 μL inoculum volume. After 1 h of incubation at 37 °C the inoculum was removed and fresh media was added to the cells. At various times post infection, the supernatant from infected cells was removed and frozen (to quantify released virus). Fresh media was then added to the remaining cells and these were frozen to quantify the cell-associated virus. All experiments were carried out with biological triplicates. For phase-contrast microscopy, MeV-GFP infected cells were visualized by phase-contrast microscopy using an inverted UV microscope (Nikon Eclipse TE2000-5 microscope coupled with a Nikon HB-10101AF super high-pressure mercury lamp) equipped with a Hamamatsu C472-95 digital camera (Sony, Minato-ku, Japan).
2.5. Pseudotyped Viruses
HEK293T cells were plated at a density of 7.5 × 105 cells per well in 6 well dishes. The following day they were transfected with 3.5 µg each of pcDNA3.1 constructs expressing MeV F and H with 30 and 24 amino acid cytoplasmic tail truncations, as well as 1.5 µg of p8.91 (encoding for HIV-1 gag-pol) and 1 µg of CSFLW (the luciferase reporter-expressing lentivirus-backbone). Supernatants containing pseudotyped virus (MeV-PP) were harvested at 72 h post transfection, clarified by centrifugation, and frozen to −80 °C. The target 293-hSLAM cells were plated at a density of 1 × 105 cells per well in 24 well dishes one day prior to transduction/infection for 72 h. Firefly luciferase activity in these cells was assayed using the Luciferase Assay System (Promega, Madison, WI, USA) according to the manufacturer’s instructions and a Promega GloMax multimode plate reader.
2.6. Fusion Assays
The HEK293T cells were plated out to a cell density of 7.5 × 10
5 cells per well in 6 well dishes. The following day, effector cells were transfected with 500 ng each of MeV or PPRV F and H expression constructs, 500 ng of the 1–7 fragment of rLuc-GFP [
25] and 1 μg of either the blank vector control (pcDNA3.1) or a BST2 expression vector (as indicated). Separately, target cells were transfected with 1 μg of lentivirus vectors expressing human or ovine SLAMF1, as well as 500 ng of the 8–11 fragment of rLuc-GFP. All transfections were performed using TransitX transfection reagent (Mirus), according to the manufacturer’s instructions. Following 48 h of incubation, effector and target cells were washed, counted, and co-cultured at a ratio of 1:1 in white-walled 96 well plates to a final density of 1 × 10
5 cells per well. Then, 16–24 h later the Renilla luciferase activity in fused cells was measured (in a Promega GloMax multi-mode plate reader) by removing the media and adding 2 μg/mL of cell-permeable coelenterazine 400A (Biotium, Fremont, CA, USA), in PBS. Normally, five or more co-culturing replicates were performed for each biological condition.
2.7. Incucyte Fluorometric Quantification
GFP fluorescence, including total GFP intensity and average green object size, were quantified using an Incucyte S3 real-time imager (Essen Bioscience, Ann Arbor, MI, USA) with cells being maintained under the same conditions listed previously (37 °C with 5% CO2). Phase images were captured regularly and masking applied to identify individual cells. Concurrently, GFP fluorescence was quantified using a built-in fluorescence detection filter.
2.8. Protein Labeling and Quantification
To examine protein co-expression, 293-hSLAM cells were plated at a density of 1 × 10
5 cells per well in 24 well dishes, transfected with relevant combinations of BST2 and viral protein expression constructs using Transit X2 transfection reagent (Mirus), and lysed at 16–24 h post transfection. All protein samples were generated in 1X radio-immunoprecipitation assay (RIPA) buffer containing protease inhibitors (ThermoFisher). Briefly, existing growth media was removed and cells were washed in phosphate buffered saline (PBS) before being pelleted by centrifugation. Pelleted cells were then resuspended in 1X RIPA and left on ice for 10 min before repeated centrifugation at high speed (16,000×
g) for a further 10 min at 4 °C. Protein lysate-containing supernatants were then stored at −20 °C until required. Samples for western blot were analyzed by SDS-PAGE, semi-dry, PVDF-based, transfer, and blotting in TBS-Tween containing 5% (
w/v) milk powder. All primary antibodies were incubated overnight at 4 °C. The following antibodies were used: anti-MeV nucleocapsid (N505) and anti-morbillivirus/MeV haemagglutinin (cytoplasmic tail) (rabbit polyclonal at 1:1000, gifted from R. Cattaneo, [
26]), anti-FLAG (1:1000, Cell Signaling (CS), 9A3), anti-GAPDH (1:1000, 14C10, CS), anti-HA (1:1000, CS, C29F4), anti-tubulin (1:1000, CS, 9F3) and standard HRP-linked secondary antibodies (CS). For flow cytometry analysis of transfected 293-hSLAMs, cells were immunolabeled in PBS with 1% BSA, 0.01% NaN
3, and protease inhibitors (ThermoFisher) together with the PE-conjugated anti-SLAMF1 antibody (BD, 559592, 1:100). Labeled, or isotype-control labeled cells, were then fixed in a solution containing 2% PFA, PBS, and 0.01 % NaN
3 and cells were analyzed using a CyAn Analyzer flow cytometer (Beckman Coulter, Brea, CA, USA). Following appropriate gating the mean fluorescence intensity of SLAMF1 positive cells was calculated from triplicate analyses. For immunofluorescence analysis by confocal imaging transfected Vero hSLAM cells (24 h post transfection) were fixed in 4% PFA PBS, permeabilized in 0.2% TX-100 PBS, and blocked and stained in 1% BSA PBS. The antibodies used for staining were anti-FLAG (CS, 1:100), anti-HA (CS, 1:100), anti-PPRV H (C77 mAb, 1:100) and standard fluorophore-conjugated secondary antibodies (Invitrogen, Carlsbad, CA, USA). Slides were mounted using Mowiol mounting medium (Merck Millipore) containing Hoescht 33342 DNA stain. To visualize the cells we used a Leiss LSM 510 Meta Confocal Microscope.
2.9. Phylogenetic Analysis
A comparison of BST2 amino acid sequences was performed using the Vector Nti package (ThermoFisher), particularly the AlignX embedded software. The sequences analyzed were as follows: XP_006747308 Leptonychotes weddellii (seal), XP_865603 Canis lupus familiaris (dog), NP_001230014 Felis catus (cat), NP_004326 Homo sapiens (human), XP_004277750 Orcinus orca (whale), NP_001171522 Ovis aries (sheep, BST2B), DAA28235 Bos taurus (cow), and NP_001171521 Ovis aries (sheep, BST2A).
2.10. Statistical Analysis and Data Handling
All experimental data sets contain a minimum of three biological replicates. Statistical analysis was performed using an unpaired, one-tailed t-test (*, p < 0.05; **, p < 0.005; ***, p < 0.0005; ****, p < 0.0001) within the GraphPad Prism file.
4. Discussion
Our observation that mammalian BST2 proteins target morbillivirus haemagglutinin proteins contrasts with both the established restriction mechanism of BST2 (in tethering nascent virus to the cell surface [
11]) and more recent observations that many vGPs, e.g., Ebola GP [
15,
29,
30], HIV-2 Env [
31], and HSV-1 gM [
32], have evolved as direct BST2-antagonists. Importantly, the mechanisms underpinning vGP-mediated inhibition of BST2 are likely to be evolutionarily distinct, since both HSV-1 gM and HIV-2 Env have a Vpu-like mechanism for sequestration of BST2 from the cell surface [
31,
32], while Ebola virus GP is thought to use an alternative approach reliant on the concerted action of its glycan cap and membrane-spanning domain [
29,
30]. This complexity is also evident when comparing HSV-1 and HSV-2 that are related viruses that have evolved separate mechanisms for vGP restriction of BST2 [
32,
33]. Interestingly, our results suggest that morbillivirus vGPs have not evolved any BST2-antagonistic phenotype and are actually, in direct contrast, sensitive to inhibition by this protein, particularly, evident in our fusion assays and infected cells. Our data indicate that this inhibition is due to colocalization of BST2 and morbillivirus GPs in intracellular compartments, a reduction in H expression at the cell surface, and, lastly, a BST2 dose-dependent reduction in overall H protein levels. Although Narkpuk et al. observed a similar BST2-mediated down-regulation of transient protein expression in cells [
34], the reduction in H protein we observed was specific to this vGP, since neither the viral N protein nor split rLuc-GFP reporter were affected by transient BST2 co-expression. In addition, the trafficking and surface expression of the stably expressed and extensively glycosylated SLAMF1 was also not affected by BST2 expression, which is further indicative of a specific interaction between BST2 and morbillivirus H. Finally, the inhibition of fusion did not appear to correlate to BST2 overexpression induced activation of NF-κB, since the phosphorylation-signaling deficient tyrosine mutant Y6,8A was still capable of inhibiting fusion.
Targeted mutational analysis of BST2 demonstrated that inhibition of MeV fusion and replication is reliant on this protein’s GPI anchor which is consistent with this domain being important in viral restriction [
11]. It is important to highlight, however, that although the ∆-TM BST2 mutant was efficient at inhibiting MeV cell–cell fusion it did not reduce the overall level of H protein, unlike the full-length BST2 protein. This intriguing observation, in combination with the altered cellular localization of MeV H when co-expressed with this mutant points to a bipartite effect of BST2 on H, both at the level of cellular trafficking and, potentially, protein degradation. Although the ultimate fate of morbillivirus H proteins remains unclear at this juncture, we hypothesize that BST2 overexpression targets these proteins for proteosomal degradation and this is an area of continued work in our laboratory.
What specifically makes morbillivirus H proteins a target for BST2 is still not known. Since trimers of F and tetramers of H must fold and preassemble as functional oligomers during intracellular trafficking, it is attractive to postulate that morbillivirus-specific aspects of this process are targeted by BST2. One clue to support this hypothesis is that MeV vGPs do not pseudotype well onto lentiviruses, such as the defective HIV-1 system commonly used in laboratories [
23,
24]. We hypothesize that this occurs because MeV buds from different plasma membrane micro-domains to HIV-1, a process governed by the cytoplasmic tails of MeV vGPs. Accordingly, the defect in MeV vGP pseudotyping is overcome through removal of their cytoplasmic tails [
23,
24,
35]. BST2 may, therefore, have evolved to inhibit virus budding at only specific domains of the cell surface. This hypothesis is strengthened by our observation that MeV genome levels were higher in the supernatant from BST2-transfected cells, although the infective particle number was the same, indicating a perturbation of the normal processes occurring during MeV budding. However, an alternative explanation is the leakage of viral genomes from infected BST2-expressing cells, as there is no direct evidence the detected genomes are from enveloped particles.
This hypothesis, in turn, relates to the absence of significant restriction of MeV release that we observed in our experiments, which was a surprising result given the broad specificity and mechanism of action of this restriction factor [
11]. Although it has previously been demonstrated that BST2 can inhibit MeV replication in vitro, these studies quantified only the released virus in the supernatant and did not correlate this to either cell-associated virus yields or the expression of viral proteins and BST2 [
9]. Regardless, the observation that released virus was significantly affected by BST2 expression is interesting and markedly contrasts with our own findings. This discrepancy may relate, in part, to the virus strain and receptor used. While we used the virulent IC323 MeV strain and HEK293T cells overexpressing the natural SLAMF1 receptor, Holmgren et al., in 2015 used the attenuated Edmonston vaccine strain which has an extended receptor tropism (binding CD46, in addition to wild-type receptors SLAMF1 and nectin-4) and demonstrably defective innate immune antagonists, especially the accessory protein V that blocks interferon signaling [
9,
36,
37]. Interestingly, the V protein from human parainfluenza virus type 2 (hPIV-2) has recently been shown to interact directly with BST2 to antagonize hPIV-2 restriction [
38]. This interaction, specific to C-terminal Tryptophan (Trp) residues in V and the GPI anchor of BST2, leads to re-localization of BST2 from the cell surface without apparent degradation [
38]. Since the MeV V protein has a conserved a Trp-containing C-terminal domain [
39], it may also be capable of an equivalent restriction of BST2. These putative interactions, as well as a comparison of IC323 and Edmonston V proteins, may explain the strain-specific effect of BST2 on MeV egress and are the focus of ongoing work in our laboratory. Strain-specific interactions with BST2, particularly between lab-adapted and virulent-strain viral proteins have been reported elsewhere, i.e., the HA of multiple influenza virus A strains were shown to variably antagonize BST2 [
40], indicating these observations might not be limited to accessory proteins such as V.
Although the broad specificity of mammalian BST2s was evident from our studies, the mechanism of restriction varied, dependent on both the BST2 sequence and target protein. While we observed colocalization between MeV H and
H. sapiens BST2 this was less evident for PPRV H and the shorter BST2A isoform of
O. aries. However, a more specific colocalization between PPRV H and BST2 was seen with the longer B isoform, in keeping with studies from Murphy et al., which demonstrated BST2B-specific sequestration of Jaagsiekte sheep retrovirus (JSRV) Env protein into the Golgi apparatus [
28]. The duplication of the
O. aries BST2 gene has been reported previously [
28,
41], in particular, the absence of N-linked glycosylation sites and GPI-anchor sequences in the longer B isoform, characterizations that are supported by our own western blot analysis of this protein. While this points to varying mechanisms of restriction, it should be highlighted that in all cases, including examples of cross-species restriction (e.g., MeV H and
O. aries BST2), a gross reduction in H protein was seen. This was most evident in the BST2-transfected PPRV infected cells where there was clear perturbation of H trafficking in areas of infected cell syncytia expressing larger levels of the overexpressed BST2 protein. Given the importance of BST2 in determining virus host susceptibility and disease pathogenesis [
42] the cross specificity of mammalian BST2 proteins against morbillivirus H proteins is of interest and an area for continued investigation.
Our studies focused on the overexpression of BST2 in vitro highlighting specific dysregulation of the morbillivirus H protein. Further work is required to examine the effect of endogenous BST2 on morbillivirus glycoprotein activity, at baseline or IFN-induced levels, and this is the focus of ongoing work in our laboratory. In addition, it remains to be determined what effect BST2-specific inhibition of morbillivirus H proteins has on viral infection in vivo. Intriguingly, although Holmgren et al. demonstrated upregulation of BST2 (via a type I IFN response) in primary murine neurons and the brains of intracranially infected mice, its removal, in related KO mice studies, had no effect on pathogenesis [
9]. This may be due to a MeV-vaccine-specific phenotype or, alternatively, a reflection of the built-in redundancy of the innate immune response following type I induction. Given the complex bitropic life cycle of morbilliviruses in SLAMF1-positive immune cells and nectin-4 positive epithelia, the role of restriction factors, including BST2, in the innate response to infection is an area of increasing interest. Our research also highlights the advantage of using quantitative assays modeling aspects of the viral life cycle, e.g., the MeV vGPs cell–cell fusion assay to characterize such restriction factors.