The integral membrane protein tetherin has been associated with an eclectic mix of cellular processes, including restricting the release of a range of enveloped viruses from infected cells. The unusual topology of tetherin (it possesses both a conventional transmembrane domain and a glycosylphosphatidylinositol anchor), its localisation to membrane microdomains (lipid rafts) and the fact that its cytosolic domain can be linked (indirectly) to the actin cytoskeleton, led us to speculate that tetherin might form a ‘tethered picket fence’ and thereby play a role in the organisation of lipid rafts. We now show that knocking down expression of tetherin leads to changes in the distribution of lipid raft-localised proteins and changes in the organisation of lipids in the plasma membrane. These changes can be reversed by re-expression of wild-type tetherin, but not by any of a range of tetherin-based constructs, indicating that no individual feature of the tetherin sequence is dispensable in the context of its lipid raft organising function.
The classic Singer and Nicolson ‘fluid mosaic’ model of the plasma membrane has served us well (Singer and Nicolson, 1972). However, it is now clear that there is significant complexity and heterogeneity in the distribution of lipids and proteins in most biological membrane systems and that the plasma membrane should, ‘be viewed as a mosaic of microdomains’ (Maxfield, 2002). Much discussion in this area over the past decade or more has focussed on the term ‘lipid rafts’ which has been used as short-hand to describe sphingolipid- and cholesterol-rich microdomains in cell membranes. In fact the lipid raft hypothesis “proposes that the lipid bilayer is not a structurally passive solvent, but that the preferential association between sphingolipids, sterols, and specific proteins bestows cell membranes with lateral segregation potential” (Lingwood and Simons, 2010). The physical existence of lipid rafts was demonstrated some time ago (Pralle et al., 2000) and there is extensive evidence that lipid rafts are small, and that some types might be highly dynamic (reviewed by Lingwood and Simons, 2010; Simons and Toomre, 2000; Viola and Gupta, 2007). They have been implicated as being important in a broad range of fundamental and essential cellular functions. These include membrane trafficking, by helping to segregate proteins for delivery to specific locations; cell signalling, by providing platforms for the transient assembly of signalling complexes; regulated exocytosis, by forming sites in the plasma membrane for this to occur; pathogen entry/egress (including influenza and HIV) and toxin entry, by providing sites in the plasma membrane where these events take place (Brown, 2006; Dykstra et al., 2003; Fantini et al., 2002; Holowka et al., 2005; Lafont and van der Goot, 2005; Lencer and Saslowsky, 2005; Mañes et al., 2003; Pelkmans, 2005; Pike, 2003; Rosenberger et al., 2000; Salaün et al., 2005; Salaün et al., 2004; Simons and Toomre, 2000; van der Goot and Harder, 2001; Viola and Gupta, 2007). In addition, lipid rafts have been implicated as playing a role in several human diseases, e.g. Alzheimer's, Parkinson's, Prion, Cardiovascular and Autoimmune diseases (reviewed by Cordy et al., 2006; Fantini et al., 2002; Jury et al., 2007; Michel and Bakovic, 2007; Simons and Ehehalt, 2002).
Viola and Gupta highlight the fact that there have been several reports of integral membrane proteins that tether membrane rafts to the underlying actin cytoskeleton (Viola and Gupta, 2007). Indeed, the work of Kusumi and colleagues has developed the ‘picket fence’ model, originally proposed by Sheetz (Sheetz, 1983), in which specific integral membrane proteins are linked directly (or indirectly) to the actin cytoskeleton, thus acting as rows of pickets in the plane of the lipid bilayer, limiting the free diffusion of membrane lipids and proteins and thereby serving to establish membrane microdomains (Kusumi et al., 2005). This linking of integral membrane protein ‘pickets’ to the actin ‘fence’ not only generates membrane microdomains, it also ensures proximity of the actin cytoskeleton to the region of the plasma membrane containing the membrane protein ‘pickets’, thereby intimately coupling the architecture of the cytoskeleton with that of the plasma membrane. Consistent with this model is the fact that a growing number of integral membrane proteins have been shown to be linked to the actin cytoskeleton, often (directly or indirectly) via a member of the Ezrin–Radixin–Moesin (ERM) family of proteins (reviewed by Bretscher et al., 2000; Fanning and Anderson, 1999).
It is not hard to envisage how a ‘picket fence’ around the edge of a membrane raft could restrict free diffusion in the plane of the lipid bilayer. However, a ‘picket’ linked to both the underlying actin cytoskeleton ‘fence’ and the membrane raft would, potentially, serve a greater organizing role than a simple ‘picket fence’ in which the transmembrane (TM) domain of the integral membrane protein ‘picket’ lies outside the membrane raft. Such a ‘tethered picket fence’ could be formed by an integral membrane protein that has one or more TM domain lying within the raft and one or more TM domain lying outside the raft. Alternatively, a ‘tethered picket fence’ might be formed by integral membrane proteins that possess a conventional TM domain and a glycosyl phosphatidylinositol (GPI) anchor. GPI anchors have been shown to be preferentially localized to rafts (Brown and London, 1998a; Brown and London, 1998b; Simons and Ikonen, 1997), so a protein with a conventional TM domain and a C-terminal GPI anchor might exist with the GPI anchor within a membrane raft and the TM domain outside that raft. If the cytosolic domain of the protein were also linked to the actin cytoskeleton, this would generate a component of a ‘tethered picket fence’. CD317/tetherin is a protein with just such a topology (Kupzig et al., 2003) and one which, by oligomerisation (Schubert et al., 2010; Yang et al., 2010), has the potential to link adjacent lipid rafts and thereby generate/stabilize larger signaling platforms than would be formed by isolated lipid rafts, thereby contributing to the mesoscale domain organisation of the plasma membrane (Kusumi et al., 2011).
CD317 (aka BST2 or HM1.24 antigen (Goto et al., 1994; Vidal-Laliena et al., 2005) and recently designated tetherin (Neil et al., 2008) is an integral membrane protein with both a conventional TM domain and a GPI anchor (Elortza et al., 2006; Kupzig et al., 2003). Tetherin has been shown to play a role in regulating the growth and development of B cells and to be highly expressed in myeloma cells (Goto et al., 1994; Ishikawa et al., 1995; Ohtomo et al., 1999), its expression is upregulated by interferon-α (Blasius et al., 2006); elevated expression of tetherin leads to activation of NFkB and MAPK pathways (Matsuda et al., 2003), it has been shown to be elevated in expression in cells with elevated metastatic potential (especially metastasis to bone) (Cai et al., 2009; Walter-Yohrling et al., 2003) and in tamoxifen-resistant cells derived from a primary tumour (Becker et al., 2005); it has been implicated in cell adhesion (Yoo et al., 2011) as well as cell migration (Chai et al., 2010) and it causes retention of fully formed HIV viral particles at the surface of HIV infected cells (Neil et al., 2008; Van Damme et al., 2008). It is this last property that led to the protein being designated ‘tetherin’. Tetherin has subsequently been shown to restrict the release of a range of enveloped viruses from infected cells (Jouvenet et al., 2009; Mansouri et al., 2009; Sakuma et al., 2009a). Thus tetherin has been associated with an eclectic mix of cellular processes, prompting us to speculate that tetherin might play a fundamental organizational role within cells; a role that would impact upon diverse cellular functions.
Tetherin resides – at least at the cell surface – in lipid rafts with the TM domain apparently lying outside the raft (or at the interface of the raft and non-raft domains) and with the raft localisation being dependent upon the GPI anchor (Kupzig et al., 2003; Rollason et al., 2007). This places the N-terminal cytosolic domain of tetherin in a suitable position to interact with the actin cytoskeleton, which it does (indirectly) (Rollason et al., 2009). Consistent with this interaction is the fact that knocking down expression of tetherin in polarised epithelial cells leads to a complete loss of the sub-apical cortical actin network and a commensurate loss of apical microvilli (Rollason et al., 2009).
Thus, tetherin possesses the structural features that would be required to form a ‘tethered picket fence’. In addition, the extracellular domain of tetherin has been shown to form a disulphide bonded parallel coiled coil, thereby generating a dimer with two adjacent transmembrane domains and two adjacent GPI anchors separated by ∼17 nm (Hinz et al., 2010; Schubert et al., 2010; Swiecki et al., 2011; Yang et al., 2010). Furthermore, the parallel coiled coil dimer of the tetherin extracellular domain can exist as a head-to-head (anti-parallel) dimer in which the N-terminal 40 residues of two parallel coiled coils form a four helix bundle (Schubert et al., 2010; Yang et al., 2010). This suggested to us that, not only might tetherin form a ‘tethered picket fence’, but that it might also link adjacent lipid rafts and thereby play a role in organising membrane microdomains. We therefore set out to test the hypothesis that tetherin plays a role in the organisation of membrane microdomains and now present data that are entirely consistent with tetherin playing such a role, a role that might explain its association with a diverse range of cellular processes.
Tetherin knockdown leads to a redistribution of raft-localised proteins
We reasoned that if tetherin does play a role in the organisation of membrane microdomains, then, in cells in which expression of tetherin has been knocked down, we might detect a change in the distribution of proteins that are normally localised to lipid rafts. This is what we observed. The GDP loaded form of H-Ras is associated with the inner leaflet of the plasma membrane and is normally preferentially localised to lipid rafts; GFP tagged with the C-terminal 31 amino acids of H-Ras (GFP-H-Ras) is also preferentially localised to lipid rafts and is present in the more buoyant fractions following Na2CO3 extraction of cells and sucrose density centrifugation (Prior et al., 2001). We expressed GFP-H-Ras in control HeLa cells and in HeLa cells in which expression of tetherin had been knocked down using siRNA (Fig. 1A). The distribution of GFP-H-Ras on sucrose density gradients following Na2C03 extraction was markedly different for extracts from control cells compared to those from tetherin knockdown cells (Fig. 1B), with a clear skewing of GFP-H-Ras to the more dense ‘non-raft’ fractions in the tetherin knockdown cells.
Having shown a change in the distribution of a recombinant reporter protein in response to tetherin knockdown, we considered it important to test the effects of tetherin knockdown on the raft versus non-raft distribution of an endogenous protein. Flotillin 2 is an endogenous protein associated with the inner leaflet of the plasma membrane and one that is normally localised to lipid rafts (Stuermer et al., 2001). Following ice-cold Triton X-100 (TX100) extraction and sucrose density centrifugation, the majority of flotillin 2 in control cells was, as expected, detected in the more buoyant fractions from the gradient; these ‘detergent resistant membrane’ (DRM) fractions correspond to the position on the gradient where proteins that are associated with lipid rafts are found (Brown, 2006) (Fig. 1C, lamin KD). Indeed tetherin itself is enriched in these fractions (Fig. 1A, right) as we have shown previously (Kupzig et al., 2003). However, a significant redistribution of flotillin 2 across the gradient was detected in cells in which expression of tetherin had been knocked down (Fig. 1C, CD317 KD). A less dramatic change in the distribution of flotillin 2 was observed in lysates from cells in which expression of Annexin II (another putative membrane organising protein) had been knocked down (supplementary material Fig. S1). However, a similar change in the distribution of flotillin 2 in response to tetherin knockdown was observed following Na2CO3 extraction of cells and sucrose density centrifugation (supplementary material Fig. S1). The distribution of flotillin 2 across the gradient was essentially restored to that observed in control cells in tetherin knockdown cells expressing an siRNA resistant form of tetherin (Fig. 1C, WT rescue) (see also section on tetherin-based constructs). A similar redistribution from raft (in control cells) to non-raft (in tetherin knockdown cells) fractions was observed for GPI-YFP (supplementary material Fig. S1). GFP (or YFP) linked to the outer leaflet of the plasma membrane by a GPI anchor (GPI-GFP/GPI-YFP) has previously been used as a marker of lipid rafts (Crespo et al., 2002; Nichols et al., 2001). However, not all proteins that are normally enriched in DRMs have their distribution on sucrose gradients altered by knockdown of tetherin expression. The distribution of caveolin 1 on sucrose gradients was unaffected by tetherin knockdown (supplementary material Fig. S2).
Further evidence that tetherin knockdown leads to a redistribution of raft-localised proteins
We chose to undertake a series of fluorescence recovery after photobleaching (FRAP) experiments in order to complement the data from sucrose density gradients. GPI-YFP was expressed in control HeLa cells and in HeLa cells in which tetherin expression had been knocked down. Cells expressing GPI-YFP were then subjected to FRAP analysis (as described in Materials and Methods): GPI-YFP in tetherin knockdown cells showed a higher diffusional mobility than that in control cells (Fig. 2A upper panel; supplementary material Fig. S3A). No difference was observed in the diffusional mobility of CD99-GFP [a non-raft localised type 1 integral membrane protein (Darling et al., 1986)] between control cells and tetherin knockdown cells (Fig. 2A, lower panel). It is of note that depletion of cholesterol from cells has been reported to increase, to decrease or to have no effect (depending on the experimental system used) on the diffusion coefficient of ‘raft’ proteins in live cells (reviewed by Day and Kenworthy, 2009). In our hands, depletion of cholesterol led to an increase in the diffusional mobility of GPI-YFP, but not of CD99-GFP (supplementary material Fig. S3B).
Tetherin knockdown leads to changes in the proportion of ordered versus fluid regions of the plasma membrane
Given that the preceding data are consistent with tetherin playing a role in the organisation of lipid rafts, we reasoned that there would be differences between control cells and tetherin knockdown cells in terms of ordered (equating to lipid raft regions) versus fluid (equating to non-raft regions) regions of the plasma membrane. We therefore performed microscopy utilising the fluorescent probe Laurdan. As membranes transition from liquid-disordered to the liquid-ordered phase, characteristic of lipid raft domains in model membranes, the peak emission of Laurdan undergoes a 50-nm blue shift (Parasassi et al., 1997), a property which has been exploited to investigate the microdomain order of the plasma membrane in model lipid bilayers and in cells (Gaus et al., 2003). Generalized polarisation (GP) values (Parasassi et al., 1990), derived from Laurdan images of cells (pseudocoloured in Fig. 2B, bottom panels), give a measure of membrane order, with higher GP values denoting greater order. Methyl β cyclodextrin treatment (to deplete cellular cholesterol levels) leads to a decrease in the Laurdan GP signal (Gaus et al., 2006). It should be noted that this approach does not reveal individual domains but quantifies membrane order globally. In HeLa cells in which tetherin expression had been knocked down, GP values were significantly decreased in comparison to those from WT control cells or cells in which expression of the transferrin receptor had been knocked down (Fig. 2C–F). This is true whether one considers the whole cell (Fig. 2C,D) or only the plasma membrane (Fig. 2E,F). Furthermore, expression of the siRNA resistant HuCD317-HA-SR construct in tetherin knockdown HeLa cells restored GP values to those observed in control cells (Fig. 2C–F). GP values could be further decreased in tetherin knockdown HeLa cells by depletion of cholesterol levels using methyl β cyclodextrin (supplementary material Fig. S4).
Thus, data from (i) Na2CO3 or TX100 extraction followed by sucrose density gradient centrifugation, (ii) FRAP analysis, and (iii) Laurdan microscopy are all consistent with tetherin playing a role in organising lipid rafts in the plasma membrane. This led us to question which features of tetherin are required for this function.
Structural studies have shown that the majority of the extracellular domain of tetherin exists as a parallel coiled coil, with the dimer so formed being stabilised by three intermolecular disulphide bonds (Hinz et al., 2010; Schubert et al., 2010; Swiecki et al., 2011; Yang et al., 2010). This fairly rigid [structural studies suggest some flexibility (Schubert et al., 2010; Yang et al., 2010)] bar [about 17 nm long (Hinz et al., 2010; Schubert et al., 2010; Swiecki et al., 2011; Yang et al., 2010)] would therefore separate the N- and C-termini of tetherin, allowing one, the C-terminal GPI anchor, to reside within a lipid raft, and the other, the N-terminal transmembrane domain, to reside outside the raft (or at the boundary between raft and non-raft lipids) as we have previously speculated (Kupzig et al., 2003). We therefore initially chose to address the relative importance of the N- and C-terminal regions of tetherin in organising lipid rafts. Five tetherin-based constructs were designed (Fig. 3A): the first lacks the entirety of the protein N-terminal to the transmembrane domain (ΔN-term); this construct was expected to remain raft-associated, but the absence of the cytosolic region was predicted to abrogate the indirect interaction with the actin cytoskeleton. In a similar vein, a mutant missing the carboxyl terminal GPI anchor (ΔGPI) was constructed; this construct was anticipated to interact with the cytoskeleton but to no longer partition into rafts: indeed, Neil et al. (Neil et al., 2008) found that a human tetherin ΔGPI mutant retained the ability to reach the plasma membrane, but was no longer able to inhibit HIV virion release. Two additional ΔGPI mutants were produced where the GPI anchor was substituted with a TM domain from another cell surface integral membrane protein; these chimaeric proteins were predicted to reach the plasma membrane and interact with the actin cytoskeleton as normal, but to exhibit divergent lipid raft localisation: insertion of the CD8α TM domain was predicted to lead to localisation to non-raft regions of the plasma membrane, given that CD8α localises to non-raft microdomains (Pang et al., 2007); in contrast, insertion of the CD44 TM domain was expected to result in raft-association, as for wild-type tetherin, because CD44 localises to lipid rafts and its TM domain is a critical determinant of this localisation (Neame et al., 1995; Perschl et al., 1995). To maximise cell surface delivery of these ΔGPI-CD8 and ΔGPI-CD44 constructs, a fifteen amino acid cytosolic sequence, previously demonstrated to facilitate efficient exit from the ER and subsequent transport and PM insertion of a series of CD8 mutants (Jackson et al., 1990), was attached to the C-terminus of the CD8/CD44 TM domains (see Materials and Methods for details). Finally, a fifth construct was created in which the three cysteine residues in the extracellular domain were mutated to alanines; this C3A mutant was expected to be no longer able to form disulphide bond stabilised dimers, but to remain raft-associated and to interact with the actin cytoskeleton. Thus, this group of tetherin constructs was designed to address the importance of several features of tetherin in its role in raft organisation; namely the role of the N-terminal cytosolic region, localisation to lipid rafts (and, with the ΔGPI-CD8/CD44 constructs, the specific requirement for a GPI anchor, as opposed to any other form of raft-localisation domain), and disulphide bond mediated stabilisation of the coiled coil dimer. All constructs were made in a background template containing (a) synonymous mutations making them resistant to the siRNA used to knock down expression of tetherin and (b) an HA epitope tag at the C-terminus of the extracellular domain, as utilised by others (Neil et al., 2008) (construct HuCD317-HA-SR in Fig. 3A).
HuCD317-HA-SR constructs (Fig. 3A) were expressed in either HEK293T cells [which do not express endogenous tetherin (Neil et al., 2008)] or in HeLa cells in which expression of endogenous tetherin had been knocked down using siRNA (this was done to eliminate any mislocalisation caused by heterodimerisation with residual endogenous tetherin) and their localisation determined by immunofluorescence analysis using an anti-HA antibody as primary antibody. The proteins encoded by the constructs exhibited similar sub-cellular localisation to that of the wild-type protein (Fig. 3B), with most cells examined containing a distinct intracellular pool of protein along with clear cell surface labelling. Experiments were then carried out in order to compare levels of expression of the different tetherin-based constructs at the cell surface. Live cells were incubated with the anti-HA antibody for 20 minutes at 4°C (NB the HA epitope is in a region of tetherin that is exposed at the extracellular side of the plasma membrane in surface-localised tetherin), fixed and processed for immunofluorescence microscopy (Fig. 3C). All tetherin constructs could be detected at the cell surface. This was confirmed by FACS analysis of HEK293T cells that had been transfected with the different constructs and complemented by immunoblot analysis (supplementary material Fig. S5).
Lipid raft localisation (or otherwise) of tetherin-based constructs
The design of the tetherin-based constructs used in this study was such that any GPI anchored proteins, and the protein possessing a CD44 transmembrane domain, would be predicted to be present in lipid rafts, whereas the remaining proteins would not. The presence (or otherwise) in lipid rafts of the proteins encoded by the different constructs was assayed by sucrose density centrifugation. The partitioning of the proteins encoded by the different tetherin constructs was predominantly as had been predicted (Fig. 3D; supplementary material Fig. S6). Endogenous tetherin is primarily localised to low density, lipid raft (DRM), fractions (Kupzig et al., 2003) and, as expected, a significant pool of HuCD317-HA-SR was also found in DRMs (Fig. 3D; supplementary material Fig. S6); the sizable fraction of the protein excluded from DRMs is likely due to the fact that the protein is being overexpressed, and suggests that segregation into rafts is saturable, with any protein that exceeds this ‘capacity’ being excluded. Both the ΔN-terminus and C3A mutants remained raft associated which, given that they retain the GPI anchor (Fig. 3A), is as expected, although it indicates that neither disulphide bond stabilisation of the coiled coil dimer, nor the presence of the N-terminal cytosolic domain are essential for localisation to lipid rafts. Disappointingly, the ΔGPI-CD44 chimaera failed to partition efficiently (or reproducibly) into raft domains (Fig. 3D; supplementary material Fig. S6), indicating that the CD44 TM domain cannot functionally replace the GPI anchor, at least with regards to retaining localisation of tetherin to rafts. As expected, neither the ΔGPI mutant nor the ΔGPI-CD8 chimaera partitioned into DRMs (Fig. 3D; supplementary material Fig. S6).
Rescue of tetherin knockdown cells by expression of tetherin constructs
We had shown that expression of wild-type tetherin in cells in which tetherin expression had been knocked down ‘rescued’ the wild-type phenotype in terms of raft organisation (e.g. Fig. 1C; Fig. 2B–F). The constructs illustrated in Fig. 3A were designed to shed light on the features of the tetherin sequence that are most important for this function. The different tetherin constructs were therefore expressed in HeLa cells in which expression of tetherin had been knocked down using siRNA, and analysis of any ‘rescue’ of the wild-type phenotype in terms of raft organisation – as assayed by FRAP and sucrose density gradients – was again performed. None of the constructs (other than wild-type tetherin, Fig. 3A) effectively rescued the wild-type phenotype (Fig. 4). This suggested that no single feature of tetherin is dispensable in terms of its role in organising lipid rafts, and led us to seek a (potentially more sensitive) biochemical screen for tetherin function to complement those used to date.
NF-κB activation following tetherin expression
We noted that a screen for activators of the NF-κB and MAPK pathways had identified tetherin as one of the most potent activators of these pathways (Matsuda et al., 2003). We reasoned that such activation may occur as a result of tetherin impinging directly on the NF-κB and MAPK pathways via e.g. some form of downstream signalling from the tetherin cytosolic domain. Alternatively, and consistent with a model in which tetherin plays a role in organising lipid rafts, expression of tetherin (either at elevated levels, or in cells that do not normally express tetherin) would be predicted to stabilise lipid rafts (as is the case when HuCD317-HA-SR is expressed in tetherin knockdown cells; Fig. 2B–F). Lipid rafts have been shown to serve as signalling platforms in a variety of cell types (most notably B-cells) (Dykstra et al., 2003; Dykstra et al., 2001; Harder and Engelhardt, 2004; Simons and Toomre, 2000) and have been implicated in NF-κB signalling (e.g. Oakley et al., 2009; Waterfield et al., 2010), so any stabilisation of lipid rafts would stabilise these signalling platforms and thereby extend the longevity of any downstream signalling, thereby activating NF-κB dependent pathways. We initially confirmed that expression of tetherin leads to an activation of the NF-κB pathway (Fig. 5). We then went on to demonstrate that this effect is dose-dependent, with the magnitude of NF-κB pathway activation increasing in direct proportion to the amount of tetherin that is expressed at the cell surface (as determined by flow cytometry) (Fig. 5A). We then used this NF-κB activation assay to screen the tetherin constructs previously described. None of the constructs were as effective as wild-type tetherin in their ability to activate the NF-κB pathway; however, it is of note that the construct which is most effective in activating the NF-κB pathway (to a level about 50% of that of WT tetherin) is that in which the GPI anchor of tetherin has been replaced by the transmembrane domain of CD44 (Fig. 5B: N.B. these data are normalised for the level of cell surface expression of each tetherin-based construct). Thus, although we could not reproducibly detect the ΔGPI-CD44 chimaera in DRMs, its expression does lead to detectable activation of the NF-κB pathway, suggesting that this assay might provide a more sensitive screen for tetherin function. Expression of each of the remaining tetherin constructs does give rise to some activation of the NF-κB pathway, but this is markedly less than that observed with WT tetherin and is entirely consistent with the failure of these constructs to ‘rescue’ the organisation of lipid rafts in cells in which tetherin expression has been knocked down. However, we note that there is no clear correlation between the degree to which individual constructs restore the presence of flotillin 2 in the buoyant fractions of sucrose density gradients and the efficiency with which they activate the NF-κB pathway, so we cannot exclude the possibility that tetherin impacts upon the NFkB signalling pathway in a way other than by organising membrane microdomains. This led us to consider the effect of reduced tetherin expression on another signalling process that has been shown to be facilitated by lipid raft formation. Thus, based on the published work of the Schwartz lab (del Pozo et al., 2004), we assayed recruitment of Rac1 to membranes in control HeLa cells and in HeLa cells in which expression of tetherin has been knocked down. We find that in mock transfected cells (i.e. processed as for transfection with siRNA, but no siRNA added), 49% (±13%) of Rac1 is in the cytosol and 51% (±13%) is membrane associated whereas in cells transfected with siRNA to target tetherin knockdown, 77% (±4.7%) of Rac1 is in the cytosol and only 23% (±4.7%) is membrane associated (n = 3). These results are consistent with tetherin having the capacity to play a general role in the organisation of lipid rafts and thereby impact multiple, raft-dependent signalling pathways.
The data we have presented are consistent with tetherin playing a role in organising lipid rafts. The loss of expression of tetherin (by the use of siRNA) from cells in which it is normally expressed leads to (a) redistribution of lipid raft markers (from both the inner and outer leaflet of the lipid bilayer) in sucrose density gradients, (b) an increase in the diffusional mobility of a lipid raft marker in the outer leaflet of the plasma membrane at the cell surface, and c) a reduction in the order of lipids in cell membranes. Lingwood and Simons comment that “Lipid rafts are fluctuating nanoscale assemblies of sphingolipid, cholesterol, and proteins that can be stabilised to coalesce, forming platforms that function in membrane signaling and trafficking” (Lingwood and Simons, 2010). Tetherin would appear to be playing a role in stabilising lipid rafts, or a subset of rafts that are tethered to the actin cytoskeleton, and the fact that its expression leads to activation of the NF-κB pathway is consistent with that stabilisation enabling the formation of a signalling platform. It is of note that structural studies indicate that the parallel coiled coil dimer of the tetherin extracellular domain can exist as a head-to-head (anti-parallel) dimer in which the N-terminal 40 residues of two parallel coiled coils form a four helix bundle (Schubert et al., 2010; Yang et al., 2010). Formation of this four helix bundle appears to be required for tetherin to function effectively in restricting HIV release, as a point mutation (L70D) which abrogates its formation (while not affecting formation of the parallel coiled coil dimer) is significantly reduced in its capacity to restrict HIV release from infected cells (Schubert et al., 2010). This implies that tetherin can exist as a head-to-head anti-parallel dimer of parallel dimers in cells. Such an organisation would allow tetherin to link adjacent lipid rafts, thereby clustering signalling platforms and at the same time generating a network of barriers on the outer leaflet of the plasma membrane (Fig. 6). Such barriers are likely to be in close proximity to the cell surface (if not in actual contact with it) given that structural studies indicate that there are only a few amino acids between the ends of the extracellular coiled coil of tetherin and its membrane anchors [three amino acids between the transmembrane domain and the N-terminus of the coiled coil and two between the C-terminus of the coiled coil and the GPI anchor (Yang et al., 2010)]. The fact that the cytosolic domain of tetherin can also interact (indirectly) with the actin cytoskeleton (Rollason et al., 2009) only serves to re-enforce the membrane organising role that it is capable of playing, since this allows tetherin to link a stabilised raft platform with the underlying actin cytoskeleton. Thus, the known structure of tetherin provides an explanation for its capacity to serve as an organiser of lipid rafts and it joins a growing number of proteins [e.g. tetraspanins, annexins, galectins, caveolins, cavins, flotillins, MYADM (Aranda et al., 2011; Claas et al., 2001; Frick et al., 2007; Grewal et al., 2010; Hill et al., 2008; Lajoie et al., 2009; Stuermer et al., 2001)] that perform roles in the organisation of membrane microdomains.
It is of note that expression of the HIV accessory protein Vpu leads to the degradation of tetherin; this counteracts the restriction on HIV release imposed by tetherin (Neil et al., 2008; Van Damme et al., 2008). However, since loss of tetherin expression also leads to a change in the organisation of lipid rafts, the effects on cell physiology of Vpu-mediated downregulation of tetherin expression are potentially more diverse than has been appreciated to date. Furthermore, the increased expression of tetherin in response to type 1 interferon (Blasius et al., 2006; Kawai et al., 2008) has implications for the organisation of lipid rafts in tetherin-expressing cells.
Materials and Methods
CD317 expression constructs and plasmids
The previously described (Rollason et al., 2009) hairpin siRNA oligonucleotides 5′-CCAGGTCTTAAGCGTGAGA-3′ (corresponding to base pairs 432–450 of the human CD317 sequence) and 5′-CTGGACTTCCAGAAGAACA-3′ (corresponding to base pairs 807–825 of the human lamin A/C sequence) were used to knockdown CD317 and lamin A/C, respectively. The transferrin receptor (TfR) was knocked down using the siRNA oligonucleotide 5′-GCAGTGCCTTCCATAATTA-3′, corresponding to base pairs 3371–3389 of the human transferrin receptor (TfR) sequence. An HA-tagged human CD317 (HA-tagged between amino acids 154 and 155) construct (a gift from Stuart Neil, Guy's Hospital, London, UK), which has been described previously (McNatt et al., 2009), was used to generate an siRNA resistant, HA-tagged CD317 construct via PCR-based site-directed mutagenesis (SDM; Stratagene), by making three synonymous mutations in the siRNA recognition site of HA-tagged human CD317; all deletion/substitution mutants were subsequently generated in this background. Constructs lacking the N-terminal cytosolic region (residues 1 to 20) or the GPI motif (residues 162 to 180), hereafter termed ΔN-term and ΔGPI, respectively, were constructed using PCR primers directed to the relevant sequences. A ΔGPI-CD8 construct was generated by annealing sequence encoding the entirety of the CD8α TM domain (residues 183 to 203 of transcript variant 1, with cytosolic residues 179 to 182 as linker) to the ΔGPI PCR product, by using complementary overlapping primers; the pIRESneo2-CD8α was courtesy of Matt Seaman (University of Cambridge, Cambridge, UK). A ΔGPI-CD44 mutant was constructed in the same manner, using residues 266 to 290 of transcript variant 4, with 270 to 290 being the TM region; pcDNA3.1-CD44 was a gift from Ashley Toye (University of Bristol, Bristol, UK). To ensure the two chimeric ΔGPI mutants would be efficiently trafficked to the cell surface, a KYKSRSSSSSSSSSS sequence was placed at the extreme carboxyl terminus, after the CD8/CD44 TM; this motif liberates normally ER-retained proteins, enabling their delivery to the cell surface (Jackson et al., 1990). A construct in which all three cysteine residues in the extracellular region, C53, C63 and C91, were converted to alanines, hereafter called C3A, was generated by SDM PCR; these three cysteines have previously been shown to be essential for dimerisation of CD317 (Andrew et al., 2009; Sakuma et al., 2009b).
The ‘artificial tetherin’ construct, consisting of part of the transferrin receptor (TfR) cytosolic domain, its TM region and part of the extracellular stalk, the coiled-coil of DMPK (dystrophia myotonica protein kinase), an HA epitope tag and the C-terminus (including GPI addition motif) of uPAR (urokinase plasminogen activator receptor) ligated together to form a protein that mimics tetherin's topology, has been described previously (Perez-Caballero et al., 2009), and was a gift from Paul Bieniasz (Rockefeller University, New York, NY, USA).
Immunofluorescence confocal microscopy was performed as described previously (Rollason et al., 2009). Briefly, HeLa cells grown on coverslips were co-transfected with CD317 siRNA and CD317-HA-siRNA resistant expression constructs (X-tremeGENE; Roche) and cultured for a further 48 hours before processing for immunofluorescence analysis. Cells were then immediately fixed and permeabilised in methanol to allow detection of whole-cell distribution of protein, incubated with the primary anti-HA antibody (Covance) for 1 hour, washed with PBS and then incubated with Alexa-Fluor 488-conjugated secondary donkey anti-mouse antibody for 1 hour. To assay protein delivery to the cell surface, cells were instead incubated with the primary anti-HA antibody for 20 minutes on ice prior to fixation in 3% formaldehyde, washed with PBS, and incubated with the same Alexa-Fluor 488-conjugated secondary antibody for 1 hour. Labelled cells were imaged using a confocal laser-scanning microscope (AOBS SP2; Leica) equipped with Ar (458, 476, 488, 496, 514 nm lines) and 405 nm diode lasers attached to an inverted epifluorescence microscope (DMRBE2; Leica). Images were collected using a 63×NA 1.4 oil immersion objective and processed with Leica and Photoshop (Adobe) software.
For Laurdan microscopy, HeLa cells grown on coverslips and either mock transfected, transfected with TfR siRNA, CD317 siRNA or CD317 siRNA and WT HuCD317 HA-SR, were, 48 hours post-transfection, incubated with 10 µM Laurdan (Invitrogen) in serum-free OPTI-MEM for 30 minutes at 37°C, with or without transferrin Alexa-Fluor-647 (Invitrogen). Cells were washed twice in PBS, fixed in 4% formaldehyde, and then incubated with wheat germ agglutinin (WGA)-TexasXRed (Invitrogen) for 10 minutes at RT. Cells were then incubated with primary mouse anti-CD317 antibody (mock, CD317 KD) or mouse anti-HA antibody (WT HuCD317-HA-SR rescue) for 1 hour, prior to incubation with Alexa-Fluor-647-conjugated secondary donkey anti-mouse antibody for 1 hour. Imaging of Laurdan-treated cells was performed with the same confocal microscope set-up employed for immunofluorescence, with excitation of Laurdan, via the 405 nm diode laser, detected in the ranges 410–460 nm and 470–530 nm, using a 100× NA 1.4 oil immersion objective. Collected images were then used to construct GP images as previously described (Gaus et al., 2006) using ImageJ software (National Institutes of Health, Bethesda, MD). GP values were analysed with Excel (Microsoft) and Prism (GraphPad) software, and statistical analysis performed using Student's t-test.
Fluorescence recovery after photobleaching
Live cell imaging was carried out at 25°C in CO2 independent media (Gibco) with 10% FCS on a Leica TPS SP5 system. Images were collected at 1-second intervals using a 63× oil immersion objective at 1× zoom, after bleaching 10–20 times at full laser power. The pinhole was set to 5 Airy units and the bleach area was a circle of 4–6 µm. The diffusion coefficient was calculated as D = 0.244×r2/t0 (Schmidt and Nichols, 2004) where t0 is the time taken for half the fluorescence to recover (half life), this value is calculated by the Leica FRAP wizard by fitting the data to a single exponential curve.
Detergent-resistant membrane isolation
Cells were grown in 10-cm plates to 50% confluency, transfected with siRNA and/or plasmid and incubated for a further 48 hours. On ice, cells were scraped into 2 ml of TNE+1% Triton X-100 and passed eight times through a 21 gauge needle. After 30 minutes incubation on ice the lysate was brought up to 40% sucrose by addition of 2 ml of 80% sucrose in TNE in a 12 ml centrifuge tube. 5 ml of 35% sucrose in TNE was layered on top followed by 1 ml each of 15% sucrose, 1% sucrose and TNE. The gradients were spun at 34,000 rpm in a Sorval TH.641 swing out rotor for 18 hours at 4°C. 1 ml fractions were taken and the protein precipitated by addition of 0.25 volume of 100% TCA (trichloroacetic acid). Fractions were resuspended in sample buffer (10% sodium dodecyl sulphate, 10% β mercaptoethanol). For analysis of Ras proteins the protocol of Prior et al. (Prior et al., 2001) was used. Cells were scraped into 0.5 M Na2CO3. After homogenization through a 23-gauge needle, lysates were mixed with an equal volume of 90% sucrose in MES-buffered saline (MBS; 25 mM MES, pH 6.5, 150 mM NaCl). We overlaid 2 ml of 45% sucrose/lysate sequentially with 2.4 ml of 35% sucrose, 2 ml of 30% sucrose, 2 ml of 25% sucrose and 2 ml of 5% sucrose, and centrifuged it in a Sorval TH.641 rotor for 16 hours at 48,000 rpm. Ten 0.8-ml fractions were collected from the top of the gradient and were re-centrifuged at 100,000 g for 30 minutes. Membrane pellets were resuspended and separated by SDS-PAGE prior to immunoblotting.
Luciferase reporter assay
Luciferase assays were performed in 96-well plates. In each well of a black 96-well plate (Greiner), 1×104 293-T cells were seeded and, 24 hours later, transfected with 50 ng of CD317 or control plasmid together with 50 ng of reporter plasmid and 12.5 ng of transfection control plasmid, using 0.4 µl Genejuice (Merck Chemicals); total DNA levels were equalised with sheared salmon sperm DNA (Sigma). Twenty-four hours post-transfection, cells were harvested and assayed using the Dual-Glo Luciferase System (Promega), according to the manufacturer's instructions. The reporter plasmid, pNF-κB-Luc, contains Firefly luciferase downstream of an NF-κB responsive promoter; the transfection control plasmid, pRL-SV40 (Promega), contains Renilla luciferase downstream of the constitutive SV40 promoter. Negative and positive controls were performed using pGL3, where Firefly luciferase is under the control of no promoter, and pFC-MEKK (Stratagene), respectively. Each treatment was carried out in octuplicate.
To take protein expression variations into account, flow cytometry was performed contemporaneously with the luciferase assay. In each well of a 12-well plate, 1.27×105 293-T cells were seeded and, 24 hours later, transfected with the same mixture of plasmids as were the 96-well plates, except that each well of a 12-well plate was treated with 12.7 times the amount of transfection mixture used for a well of a 96-well plate. 24 hours post-transfection, cells were washed in PBS and resuspended in PBSA (PBS, 1% BSA) containing primary anti-HA antibody, and incubated for 1 hour. Cells were then washed once in ice-cold PBS, and incubated with PE conjugated anti-mouse secondary antibodies for 1 hour at 4°C. Fluorescence signals were measured using a FACS CantoII-F60 machine (BD Biosciences, Oxford, UK). Data were analyzed using Flowjo 7.2.5 software (Flowjo, Ashland, OR, USA). Each treatment was performed in duplicate. Subsequent to data analysis, luciferase data were normalised to mean PE fluorescence signals.
We thank Paul Bieniasz, Stuart Neil, Matthew Seaman and Ashley Toye for plasmids, Katie Blakemore for artwork in Fig. 6 and the Chugai Pharmaceutical Company for the gift of the HM1.24 monoclonal antibody.
P.G.B. and R.R. conceived, designed and performed experiments and contributed to writing the paper. I.P. helped design experiments and interpret results. D.M.O. and K.G. trained P.G.B. in Laurdan microscopy, designed Laurdan experiments and interpreted data from those experiments. G.B. conceived and designed experiments and wrote the paper. All authors commented on drafts of the paper.
We thank the Wellcome Trust (studentship WT086783MA) and the Biotechnology and Biological Sciences Research Council [grant number BB/G021031/1 to G.B.] for funding and the Medical Research Council for an Infrastructure Award and Joint Research Equipment Initiative Grant to establish the School of Medical Sciences Cell Imaging Facility. Deposited in PMC for release after 6 months.