The Size and Conservation of a Coiled-coil Structure in the Ectodomain of Human BST-2/Tetherin Is Dispensable for Inhibition of HIV-1 Virion Release*

Background: BST-2 inhibits virus release by tethering virions to the cell surface. Results: The length of the BST-2 ectodomain can be increased or reduced without loss of function. Conclusion: The positioning of ectodomain deletions rather than their size determines the impact on BST-2 function. Significance: Understanding the importance of structural elements in BST-2 is critical for developing antiviral strategies. BST-2/CD317/tetherin is a host factor that inhibits HIV-1 release and is counteracted by HIV-1 Vpu. Structural studies indicate that the BST-2 ectodomain assumes a coiled-coil conformation. Here we studied the role of the BST-2 ectodomain for tethering function. First, we addressed the importance of the length and structure of the ectodomain by adding or substituting heterologous coiled-coil or non-coiled-coil sequences. We found that extending or replacing the BST-2 ectodomain using non-coiled-coil sequences resulted in loss of BST-2 function. Doubling the size of the BST-2 ectodomain by insertion of a heterologous coiled-coil motif or substituting the BST-2 coiled-coil domain with a heterologous coiled-coil motif maintained tethering function. Reductions in the size of the BST-2 coiled-coil domain were tolerated as well. In fact, deletion of the C-terminal half of the BST-2 ectodomain, including a series of seven consecutive heptad motifs did not abolish tethering function. However, slight changes in the positioning of deletions affecting the relative placing of charged or hydrophobic residues on the helix severely impacted the functional properties of BST-2. Overall, we conclude that the size of the BST-2 ectodomain is highly flexible and can be reduced or extended as long as the positioning of residues important for the stability of the dimer interface is maintained.

BST-2 was originally identified as a membrane protein in terminally differentiated human B cells of patients with multiple myeloma (20,21). BST-2 is a 30 -36-kDa type II transmembrane glycoprotein consisting of 180 amino acids (22). The protein is predicted to have an N-terminal transmembrane (TM) 2 region and a C-terminal glycosylphosphatidylinositol anchor (23). However, we recently found evidence to suggest that the C-terminal membrane anchor could be a second transmembrane domain rather than a glycosylphosphatidylinositol anchor (24). The two TM domains are separated by ϳ120 residues that constitute the ectodomain of the protein, which is predicted to form a coiled-coil structure (25)(26)(27)(28). The N-terminal half of the BST-2 ectodomain contains three cysteine residues, which are involved in the formation of covalent cysteinelinked dimers (20,21,29,30). Interestingly, two of the three cysteines can be mutated without loss-of-function; in fact, any one of the three cysteines can independently contribute to the formation of cysteine-linked dimers (29,30). In addition, BST-2 is modified by N-linked glycosylation at two sites in the BST-2 ectodomain (20,23,29,30); however, the functional importance of BST-2 glycosylation for inhibition of virus release is under debate (29,30). Finally, BST-2 protein associates with lipid rafts at the cell surface and on internal membranes (23,(31)(32)(33)(34) and the C-terminal TM region has been implicated with raft targeting of BST-2 (23).
A current model suggests that BST-2 tethers mature virions to the cell surface by means of its N-terminal TM region and C-terminal domain. Structural studies suggest that the BST-2 ectodomain assumes a rod-like structure of about 16 -17 nm (25)(26)(27)(28), which constitutes the maximum distance between membranes that can be bridged by BST-2. Indeed, immune electron microscopy identified BST-2 on virions tethered to the cell surface and on particles tethered to each other (30,(35)(36)(37)(38). Curiously, the distance between these structures frequently exceeded the 17-nm limit imposed by the BST-2 ectodomain, a phenomenon that has remained unexplained. Interestingly, Perez-Caballero et al. (30) demonstrated that an artificial teth-erin consisting of the N-terminal TM region of transferrin receptor, a coiled-coil ectodomain of the cytoplasmic dimeric protein dystrophia myotonica protein kinase, and a glycosylphosphatidylinositol anchor signal derived from urokinase plasminogen activator receptor is capable of inhibiting the release of HIV-1 virions tethered to the cell surface. This suggests that the secondary structure rather than the primary amino acid sequence of BST-2 is important for tethering function. This conclusion is confirmed by a recent study demonstrating that replacement of four consecutive residues by alanines is well tolerated across much of the BST-2 ectodomain (39).
The goal of the current study was to gain additional insights into the role of the BST-2 ectodomain for tethering function. Our approach was two-pronged. First, we assessed the importance of the length and structure of the ectodomain by adding or substituting heterologous coiled-coil or non-coiled-coil sequences in the BST-2 ectodomain. Second, we introduced in-frame deletions at various positions and of various sizes to identify the minimum size of the BST-2 ectodomain. Results from our first approach revealed that extending or substituting the BST-2 ectodomain using non-coiled-coil sequences derived from the CD4 ectodomain inhibited BST-2 function. In contrast, substituting the BST-2 coiled-coil domain with a heterologous coiled-coil motif derived from vimentin maintained the tethering function, consistent with previous results involving artificial tetherin. Interestingly, addition of the vimentin coiledcoil to the existing BST-2 structure, which almost doubled the size of the BST-2 ectodomain, remained functional. These results suggest that there is significant tolerance with respect to increasing the length of the ectodomain. Results from our second approach revealed that much of the BST-2 ectodomain can be deleted without loss of function. Indeed, we found that approximately half of the ectodomain, including a series of seven consecutive heptad motifs, could be deleted without losing tethering function. Interestingly, the positioning of the deletions within the ectodomain was critical for BST-2 function. Utilizing molecular modeling, we found that the orientation of hydrophobic and hydrophilic residues along the dimer interface in the C-terminal part of BST-2 was critical for function. Overall, we conclude that the size of the BST-2 ectodomain can be reduced or enlarged with heterologous coiled-coil sequences, revealing a significant flexibility in the overall size of the protein, whereas changes to the heptad motifs or the register of the coil had variable effects depending on the positioning of the deletion.
Antisera-Rabbit polyclonal BST-2 antiserum, directed against the extracellular portion of BST-2, has been described (29,43) and is available through the NIH AIDS Research and Reference Reagent Program (aidsreagent.org; catalog number 11721). HA-specific mouse monoclonal antibody was from Roche Diagnostics.
Tissue Viral Infectivity-Virus stocks were prepared by transfection of 293T cells with the indicated plasmid DNAs. Virus-containing supernatants were harvested 24 h after transfection. Cellular debris was removed by centrifugation (3 min, 1,500 ϫ g) and clarified supernatants were filtered (0.45 M) to remove residual cellular debris. 100 l of viral stock was used to infect 5 ϫ 10 4 TZM-bl cells in a 24-well plate in a total volume of 1.1 ml. Infection was allowed for 48 h at 37°C. Media was removed and cells were lysed in 300 l of Promega 1ϫ reporter lysis buffer (Promega Corp., Madison, WI) and frozen at Ϫ80°C for a minimum of 30 min. To determine the luciferase activity in the lysates, 5 l of each lysate were combined with 20 l of luciferase substrate (Steady-Glo; Promega Corp., Madison, WI) and light emission was measured using a Modulus ii microplate reader (Turner Biosystems Inc., Sunnyvale, CA).

Replacing the C-terminal Coiled-coil Domain with a Noncoiled-coil Domain Results in Functional Inactivation of
BST-2-A current working model suggests that the BST-2 ectodomain functions as a bridge between the terminal membrane anchors to tether virions to the plasma membranes or to each other (for review, see Ref. 52). A known structural requirement for virion tethering is the formation of cysteine-linked BST-2 dimers (29,30). The cysteines involved in BST-2 dimerization are located in the N-terminal half of the ectodomain (see Fig.  1A). Structural analyses of BST-2 indicate that the BST-2 ectodomain forms a continuous ␣-helix that forms a parallel dimeric coiled-coil in its C-terminal region (25)(26)(27)(28). However, the precise role of the coiled-coil and, in particular, possible constraints regarding the size of the coiled-coil remain unclear.
To assess the functional importance of the coiled-coil sequence in BST-2, we constructed a series of BST-2 variants where the coiled-coil domain was replaced by two different segments of the vimentin coiled-coil or by a non-coiled-coil element derived from the ectodomain of human CD4. For consistency with immune recognition, we utilized a BST-2 construct containing a triple HA tag downstream of residue 148 near the C terminus of the ectodomain as described previously (24). We refer to this construct as BST-2 wt for the remainder of the study. We used coiled-coil prediction software to predict the coiled-coil region in BST-2. Five of the six predictors suggested the coiled-coil in the ectodomain starts downstream of residue 90, whereas one predicted residue 101 as start of the coiled-coil region (see "Experimental Procedures"). Based on these predictions, we created replacement variants rCD4, rCC1, and rCC2 by replacing residues 106 to 148 as described under "Experimental Procedures" and schematically depicted in Fig. 1A. Expression and dimerization of the resulting variants was confirmed by immunoblotting (Fig. 1B). The functional properties of the BST-2 variants were tested by determining their effects on HIV-1 particle release. For that purpose, 293T cells, lacking endogenous BST-2 expression, were transfected with the vpu-defective pNL4 -3/Udel plasmid DNA either alone or in combination with BST-2 wt, rCC1, rCC2, or rCD4 at virus:BST-2 ratios of 50:1, 20:1, and 10:1. Viral supernatants were collected 24 h later and used for the infection of TZM-bl indicator cells. Relative virus titers were determined by measuring the virus-induced expression of luciferase in the TZM-bl cells 48 h later. The signal produced by the NL4 -3/Udel virus in the absence of BST-2 was defined as 100% (Fig. 1C, 0.0 g of BST-2). As expected, BST-2 wt inhibited the release of NL4 -3/ Udel virions in a dose-dependent manner (Fig. 1C, solid circles). Similarly, BST-2 rCC1, encoding a 89-residue coiled-coil segment of vimentin, inhibited the release of vpu-defective HIV-1 (Fig. 1C, left panel, open circles). BST-2 rCC2 contains a 178residue coiled-coil segment of vimentin, which more than doubled the size of the ectodomain. This variant retained partial activity (Fig. 1C, middle panel, open circles). Finally, replacing the C-terminal coiled-coil region of BST-2 by a 114-amino acid non-coiled-coil segment of CD4 resulted in loss of function (Fig. 1C, right panel, open circles). These results therefore suggest that whereas a coiled-coil domain at the C-terminal region of the BST-2 ectodomain is important for tethering function, there may be limitations to the overall size of the ectodomain.
The Length of the BST-2 Ectodomain Is Flexible-To further address the importance of the size of the coiled-coil region, we tested whether the length of the BST-2 ectodomain could be extended by adding coiled-coil regions of vimentin to the existing BST-2 coiled-coil domain. As a control, a non-coiled-coil portion of the ectodomain of CD4 was added as well ( Fig. 2A). Regions of CD4 or vimentin were inserted just upstream of the HA tag to create iCD4, iCC1, and iCC2, respectively, as detailed under "Experimental Procedures." Thus, this series of BST-2 variants contains the full BST-2 ectodomain in addition to heterologous sequences inserted upstream of the HA tag. The resulting proteins encode ectodomains of 239 to 328 residues compared with 150 residues (i.e. residues  of the HAtagged wild type protein ( Fig. 2A). As before, protein expression and dimerization properties of the resulting variants were analyzed by immunoblotting (Fig. 2B).
We tested the ability of the iCD4, iCC1, and iCC2 variants to inhibit virus particle release. 293T cells were transfected with pNL4 -3/Udel plasmid DNA either alone or in combination with BST-2 wt, iCD4, iCC1, or iCC2 at virus:BST-2 ratios of 50:1, 20:1, and 10:1, and virus release was determined as in In contrast, BST-2 iCC1 containing an 89-residue coiled-coil element of vimentin was capable of inhibiting virus release with almost wild type efficiency when compared with BST-2 wt (Fig. 2B, left panel,  open circles). The lack of function of iCD4 could be explained by the inability of this sequence to form a coiled-coil because the vimentin coiled-coil element of similar size (iCC1) was functional. It is unclear why insertion of the 178-residue coiled-coil element of vimentin in iCC2 led to functional inactivation of BST-2. Immunoblot analysis under reducing conditions revealed reduced expression of iCC2 (Fig. 2B, left panel). Nevertheless, iCC2 appeared to efficiently dimerize as indicated by the complete absence of the 52-kDa monomer in the non-reducing gel (Fig. 2B, right panel). Therefore, an inability to form cysteine-linked dimers is unlikely to explain the functional defect. Instead, reduced protein stability could at least partially explain the functional defect. It is also possible, however, that the inability of iCC2 to restrict virus release reflects an upper limit to the size of the BST-2 ectodomain even if the coiled-coil structure is maintained. Finally, because the same vimentin fragment could not rescue BST-2 function in the context of the replacement variant (rCC2) in Fig. 1, the functional defect could indicate a general incompatibility of this vimentin segment with the BST-2 structure. Nevertheless, together with the data from Fig. 1, we conclude that changes to both the length and the primary sequence of the BST-2 ectodomain are tolerated provided a coiled-coil structure is maintained.
The Structure of the Ectodomain of BST-2 at Its N Terminus Is Flexible-Recent studies revealed the crystallographic structure of the human and mouse BST-2 ectodomain (25)(26)(27)(28). We compiled the full-length structures of human BST-2 deposited into the RCSB PDB protein database using YASARA (Fig. 3A). Partial sequences and the murine structure were not included in this alignment. We found that whereas the structures of the C-terminal halves of the BST-2 ectodomain dimer are a near perfect match, the structures identified for the N-terminal por-FIGURE 1. BST-2 ectodomain requires a coiled-coil sequence at the C terminus. A, schematic structure of BST-2 variants rCC1, rCC2, and rCD4 compared with HA-tagged BST-2. Transmembrane domains (TM1, TM2) and the position of the HA tag are indicated as black and shaded boxes, respectively. Numbers indicate amino acid positions and the total number of residues for each of the proteins is indicated on the right. B, expression and dimerization of BST-2 variants. 293T cells were transfected with 1 g each of the indicated constructs together with 4 g of empty vector DNA. Mock-transfected 293T cells were included as negative control. Cells were harvested 24 h later, washed once with PBS, and resuspended in 400 l of PBS. Cell suspensions were divided into two equal aliquots (200 l each) and either mixed with an equal volume of reducing sample buffer (4% sodium dodecyl sulfate, 125 mM Tris-HCl, pH 6.8, 10% 2-mercaptoethanol, 10% glycerol, 0.002% bromphenol blue) or non-reducing sample buffer lacking 2-mercaptoethanol. Samples were heated for 10 to 15 min at 95°C with occasional vortexing of the samples to shear cellular DNA. Residual insoluble material was removed by centrifugation (2 min, 13,500 ϫ g in an Eppendorf Minifuge). Cell lysates were subjected to SDS-PAGE (12.5% for reduced samples; 10% for non-reduced samples); proteins were transferred to PVDF membranes and reacted with a BST-2-specific polyclonal rabbit antibody. Membranes were then incubated with horseradish peroxidase-conjugated secondary antibodies (GE Healthcare) and proteins were visualized by enhanced chemiluminescence (ECL, GE Healthcare). C, 293T cells were transfected with 5 g of NL4 -3/Udel together with increasing amounts (0, 0.1 g (50:1), 0.25 g (20:1), or 0.5 g (10:1)) of DNA encoding BST-2 wt (solid circles), rCC1, rCC2, or rCD4 chimera (open circles) as indicated. Virus-containing supernatants were collected 24 h after transfection and used for the infection of HeLa TZM-bl indicator cells. Virusinduced luciferase activity was recorded 48 h later and was used as measure of virus release. Values are expressed as mean Ϯ S.E. of three to six experiments. The infectivity of NL4 -3/Udel virus produced in the absence of BST-2 was defined as 100%. DECEMBER 28, 2012 • VOLUME 287 • NUMBER 53 tion appear to be more flexible with a hinge region around residue 80 (Fig. 3A). Cys 53 is displaced 12 Å between PDB codes 3NWH and 3MQC, whereas Cys 91 is only displaced 1.4 Å demonstrating the hinge region allows at least 12°of motion. The N-terminal region of the BST-2 ectodomain includes three cysteine residues (positions 53, 63, and 91), each of which can independently contribute to the formation of covalent cysteine-linked dimers. It also includes two N-linked carbohydrates at residues 65 and 92, which we previously reported to be dispensable for tetherin function (29).

Structural Requirements for Tethering Function of BST-2
In an attempt to characterize the functional importance of the N-terminal region of the BST-2 ectodomain, we introduced deletions within this region such that at least one cysteine residue remained (Fig. 3B). Mutants ⌬51-80 and ⌬51-88 retain cysteine 91, whereas mutant ⌬55-87 retains cysteines 53 and 91. Finally, mutant ⌬81-94 contains cysteines 53 and 63. All mutants were expressed and formed cysteine-linked dimers (Fig. 3C). The functional properties of these BST-2 deletion mutants were tested in a virus release assay as described for the previous figures. Interestingly, deletion of residues 51-80 did not affect tethering activity of BST-2 (Fig. 3D). However, dele-tion of eight additional residues (⌬51-88) that extended into the central hinge region resulted in functional inactivation of BST-2 (Fig. 3D).
Crystallography data indicate that the BST-2 ectodomain forms a continuous ␣-helix of 3.6 residues per turn. Hydrophobic residues are spaced along this helix about every fourth position so that they form an interface along one surface of the helix with another molecule of BST-2. The intervening positions consist of aliphatic or hydrophilic residues, which interact with the membrane or the extracellular fluid. One possible explanation for the lack of function of BST-2 ⌬51-88 could be a change in the residue spacing on the ␣-helix relative to wild type BST-2 that might induce misplacement of hydrophobic interactions throughout the ␣-helices. To test this hypothesis, we created an additional mutant, BST-2 ⌬55-87. This construct was designed to maintain the spacing of hydrophobic and hydrophilic residues in the ␣-helix, in-frame within the region we previously deleted. Analysis of this BST-2 variant in our virus release assay revealed that this mutant was unable to efficiently inhibit virus release (Fig. 3D) suggesting that our hypothesis was incorrect. Of note, a deletion affecting the hinge region (⌬81-94) was also poorly functional (Fig. 3D). Because the deletion in BST-2 ⌬51-88 also affects the hinge region, it is possible that the hinge region in BST-2 is critical for tetherin function. Taken together our data suggest a critical function for the central hinge region in BST-2, whereas upstream sequences encoding two of the three cysteine residues and one of the two glycosylation sites are expendable for the tethering function of BST-2. Of note, the functional importance of the central hinge region in the BST-2 ectodomain was not recognized in a recent study employing alanine scanning mutagenesis (39).
Maintaining the Register of the ␣-Helix in the C-terminal Region of the BST-2 Ectodomain May Be Important for Tethering Function-To test the minimum sequence requirements in the C-terminal region of BST-2 we created a series of in-frame deletions as depicted in Fig. 4A. Expression and dimerization of the resulting constructs was assessed by immunoblot analysis (Fig. 4B). Initially, mutants ⌬96 -116, ⌬96 -127, and ⌬131-148 were tested for their ability to inhibit virion release of NL4 -3/ Udel. Although alanine scanning mutagenesis of the region encompassed by residues 96 to 127 was previously reported to be without phenotype (39), neither mutant ⌬96 -116 nor ⌬96 -127 was able to inhibit virus release (Fig. 4C). In contrast, dele-tion of residues 131 to 148 in mutant ⌬131-148, previously found to be sensitive to alanine scanning mutagenesis (39), was tolerated and did not affect BST-2 function (Fig. 4C). We reasoned that structural elements such as the register of the helix encoded by residues 96 to 127 might be important for BST-2 function. To test this hypothesis we designed an additional mutant, BST-2 ⌬96 -118, that extended the deletion present in the inactive ⌬96 -116 variant by 2 amino acids to bring the register of the ␣-helix back in sync with the wild type helix. Deletion of two additional residues did indeed restore the tethering activity of BST-2 (Fig. 4C, ⌬96 -118). These results suggest that deletions in the C-terminal coiled-coil domain of BST-2 are tolerated as long as the proper register of the BST-2 encoded ␣-helix is maintained.
Heptad Motifs in the BST-2 Ectodomain Are Dispensable for BST-2 Function-An important structural feature of a coiledcoil is the presence of heptad motifs. In classical coiled-coils, the 7-amino acid heptad motif places hydrophobic residues in the a and d positions and charged residues in the e and g positions to form stabilizing salt bridges resulting in "knobs into holes" packing of the helices (53). Heptad repeats in the BST-2 ectodomain were reported for the C-terminal half of the BST-2 ectodomain (25,26) and are indicated as green and yellow boxes in Fig. 5A. The deletion mutants shown in Fig. 4 affected many of the heptad repeats, which could have contributed to their functional properties. In an effort to determine whether heptad repeats in the BST-2 ectodomain play a critical role, we created two deletion mutants, BST-2 ⌬97-147 and ⌬95-146. Both variants lack all seven heptad motifs but they differ slightly in the positioning and the size of the deletions (Fig. 5A). Both constructs were expressed with similar efficiency as wild type BST-2 and they formed cysteine-linked dimers even though dimerization appeared to be less efficient for both constructs (Fig. 5B). Analysis of the functional properties of these mutants revealed that despite their close similarity, BST-2 ⌬97-147 was able to inhibit virus release, albeit somewhat less efficiently than the wild type protein, whereas BST-2 ⌬95-146 was completely non-functional (Fig. 5C).
To better understand the basis for the functional differences between BST-2 ⌬96 -116 and ⌬96 -118 as well as ⌬97-147 and ⌬95-146, we modeled these mutants on PDB structure 2XG7 (26) using FOLDX within YASARA (Fig. 6B). The amino acid alignment is shown in Fig. 6A. Interestingly, the structures derived for the non-functional BST-2 ⌬96 -116 and ⌬95-146 show more hydrophobic residues on the outside of the helix (indicated in orange color in Fig. 6B) compared with the active BST-2 ⌬96 -118 and ⌬97-147. In addition, charged residues (indicated in red and blue color in Fig. 6B) are located on the inside of the coil in BST-2 ⌬96 -116 and ⌬95-146, inconsistent with the typical location outside the interface for mediating interactions with the extracellular fluid. These data suggest that changes to charge distributions that are unrelated to alterations of the coil register can affect the stability of the coiled-coil and determine the functional properties of BST-2. Overall, our data suggest that the formation of a cysteine-linked coiled-coil dimer of BST-2 is critical for tethering activity. Importantly, however, the presence of C-terminal heptad motifs is not critical for proper assembly of BST-2 dimers and inhibition of virus release.

DISCUSSION
BST-2 is thought to provide a physical tether between otherwise fully detached virions and the cell surface. One of the structural prerequisites for virus tethering is the formation of cysteine-linked dimers (29,30). Indeed, structural models of the BST-2 ectodomain suggest that BST-2 forms a parallel dimer that forms a 16 -17-nm long rod-like coiled-coil (25)(26)(27)(28), which must be anchored at both ends in a membrane. Exactly how a cell-virus tether is formed or how many BST-2 dimers are required to effectively tether a virion to the plasma membrane is currently unclear. It is also not clear exactly how Vpu prevents BST-2 from tethering nascent virions to the cell surface. Our previous studies on antibody-mediated neutralization of BST-2 indicate that inhibition of virus release by BST-2 involves a very early step of virus assembly prior to Gag proteins reaching the cell surface (54). However, much of the Vpu-BST-2 interaction remains to be investigated.
Previous work by Hammonds and co-workers (39) used alanine scanning mutagenesis to identify regions of functional importance in the BST2 ectodomain. In that study, alanine mutagenesis was well tolerated across much of the BST-2 ectodomain and only changes near the N and C termini of the ectodomain significantly affected BST-2 function (39). These results may not be surprising if one takes into consideration that alanine is the least destabilizing residue to add to the interaction surface of an oligomeric coiled-coil (55)(56)(57) and that replacing four consecutive residues at a time by alanine would not be expected to greatly disrupt the structure and interaction of the BST2 coiled-coil (58,59). Because of these limitations, previous alanine scanning mutagenesis did not recognize the importance of the central hinge region nor the relative unimportance of the size of the BST-2 ectodomain.
Our current study focused on the characterization of functionally critical elements in the BST-2 ectodomain using insertion, substitution, and deletion mutagenesis. The fact that in a previous study the BST-2 ectodomain could be replaced by a ϳ120 residue heterologous sequence consisting of 32 residues of the transferrin receptor extracellular stalk plus 75 residues of the cytoplasmic dimeric protein dystrophia myotonica protein kinase coiled-coil, already demonstrated that the primary sequence of the BST-2 ectodomain was not strictly required for tethering function (30). Our own observation showing that the coiled-coil region in the C-terminal region of the BST-2 ectodomain could be substituted with a coiled-coil region of vimentin but not with a region from the CD4 ectodomain ( Fig.  1) is consistent with this conclusion and, in addition, highlights the importance of a C-terminal coiled-coil structure. Interestingly, we found that in addition to flexibility in the primary sequence of the coiled-coil, there is also significant flexibility in its size. Indeed, we were able to almost double the size of the BST-2 ectodomain by adding a segment of the vimentin coiledcoil without loss of function (Fig. 2). Surprisingly, whereas those data point to a critical role of a coiled-coil structure of the BST-2 ectodomain, we were able to significantly reduce the size of the BST-2 ectodomain without loss of tethering function. In fact, we were able to delete more than 50 residues of the C-terminal region of the BST-2 ectodomain (Fig. 5, ⌬97-147). This result is particularly striking because the deletion in BST-2 ⌬97-147 removes most of the coiled-coil domain including all seven heptad motifs thought to be critical for the formation of a coiled-coil (see Fig. 5A). We were also able to delete up to 30 residues from the N-terminal portion of the BST-2 ectodomain, including two of the three cysteines critical for dimerization (Fig. 3, ⌬51-80), again without loss of function. In all cases Green highlights indicate motifs that perfectly match the HXXHEXE consensus motif where H is hydrophobic or uncharged and E is a charged residue; yellow highlights depict motifs in which one residue in the heptad sequence does not match the canonical motif. Black and shaded areas represent the transmembrane domains and the HA tag, respectively. B, expression and dimerization of BST-2 variants was tested as described in the legend to Fig. 1B. C, 293T cells were transfected with 5 g each of NL4 -3/Udel together with varying amounts of BST-2 wt (closed circles) or mutant DNA (open circles) as indicated. Virus-containing supernatants were collected 24 h later and used to infect HeLa TZM-bl indicator cells. Analysis of the data were performed as described in the legend to Fig. 1C. Values are expressed as mean Ϯ S.E. of nine experiments. The signal produced by NL4 -3/Udel virus in the absence of BST-2 was defined as 100%. DECEMBER 28, 2012 • VOLUME 287 • NUMBER 53

Structural Requirements for Tethering Function of BST-2
where functional inactivation was observed, differences in protein expression or the ability to form cysteine-linked dimers did not account for the observed effects on virus release with the possible exception of variant iCC2 (Fig. 2B).
Structural data indicate that the BST-2 ectodomain forms a continuous ␣-helix in the crystal lattice (26 -28). It was therefore conceivable that loss of function in at least some of our deletion mutants was caused by structural disturbances due to changes in the register of the helix. In an attempt to fit our data into a functional model, we compared the functional properties of mutations that differed by only two or three amino acids. We have two sets of mutants (⌬96 -116/⌬96 -118 and ⌬97-147/ ⌬95-146) that fit this description (Fig. 6). Two of these mutants (⌬96 -118 and ⌬97-147) are functional, whereas the other two (⌬96 -116 and ⌬95-146) are not (Figs. 4 and 5). The amino acid alignment of these mutants against BST-2 wt (Fig. 6A) indicates that the functional properties of these BST-2 variants are not determined by the size of the deletions because the deletion in BST-2 ⌬97-147 (functional) is much larger (51 amino acids) than in the non-functional BST-2 ⌬96 -116 (21 residues). Interestingly, however, we found that in both cases, the non-functional variant includes a charged residue in the dimer interface that was not present in the functional variant. Molecular modeling of the mutants based on published x-ray crystallographic data of the BST-2 ectodomain suggests differential positioning of these charged residues on the helices that could explain the functional properties of the variants. In both cases, the charged residue, Glu 98 in ⌬96 -116 (Fig. 6B, red) and Arg 95 in ⌬95-146 (Fig. 6B, blue), points into the dimer interface, whereas adjacent hydrophobic residues (Fig. 6B, orange) point toward the water phase of the helix. In contrast, in the functional variants the hydrophobic residues point toward the dimer interface and could thus stabilize the structure of the BST-2 dimer. Taken together our data demonstrate significant flexibility in the size of the BST-2 ectodomain as long as the positioning of residues critical for the formation of the dimer interface is conserved.