FRET Analysis of the Promiscuous yet Specific Interactions of the HIV-1 Vpu Transmembrane Domain

Abstract

The Vpu protein of HIV-1 functions to downregulate cell surface localization of host proteins involved in the innate immune response to viral infection. For several target proteins, including the NTB-A and PVR receptors and the host restriction factor tetherin, this antagonism is carried out via direct interactions between the transmembrane domains (TMDs) of Vpu and the target. The Vpu TMD also modulates homooligomerization of this protein, and the tetherin TMD forms homodimers. The mechanism through which a single transmembrane helix is able to recognize and interact with a wide range of select targets that do not share known interaction motifs is poorly understood. Here we use Förster resonance energy transfer to characterize the energetics of homo- and heterooligomer interactions between the Vpu TMD and several target proteins. Our data show that target TMDs compete for interaction with Vpu, and that formation of each heterooligomer has a similar dissociation constant (K_d) and free energy of association to the Vpu homooligomer. This leads to a model in which Vpu monomers, Vpu homooligomers, and Vpu-target heterooligomers coexist, and suggests that the conserved binding surface of Vpu TMD has been selected for weak binding to multiple targets.

Introduction

The human immune system has developed specialized defense mechanisms to combat viral infection. Viruses have adapted to this immune response by expressing gene products capable of counteracting host defenses. For example, Vpu, a small integral membrane protein encoded by HIV-1, is expressed in infected cells, and plays critical roles in viral release and virulence in vivo (1, 2, 3). Vpu is an 81-residue protein consisting of a single N-terminal transmembrane (TM) domain followed by a C-terminal helix-loop-helix domain, and has been associated with multiple functions in the host cell, including the downregulation of several cell-surface proteins as well as cation-specific ion channel activity (1, 4, 5). The helix-loop-helix domain functions to drive proper trafficking of Vpu within the host cell, to recruit various cellular proteins (βTRCP-2, AP-1), and is responsible for recognition of several cellular proteins targeted for downregulation by Vpu (including the CD4 receptor, CD74, and CD1d) (6, 7, 8, 9, 10, 11, 12). Interestingly, the TM domain (TMD) plays key roles in both functions of Vpu; the homooligomerization of the Vpu TMD has been shown to be sufficient for channel activity, whereas a subset of the cell surface proteins targeted by Vpu are recognized via TMD-TMD interactions (13, 14, 15). Host cell proteins targeted by the Vpu TMD include the coactivating natural killer (NK) cell receptor NTB-A, the activating NK cell receptor PVR, and the host restriction factor tetherin (13, 14, 15). The reduction in cell surface levels of these proteins provides HIV-1 with several biological advantages. A reduction in tetherin at the cell surface results in a dramatic increase in release of new virions from infected cells and thus an increased viral load within the host (5). Removal of PVR from the cell surface protects the infected host cell from being targeted by DNAM-1 NK cells (16). Finally, cells lacking NTB-A remain capable of activating NK cells but are unable to trigger sufficient NK cell degranulation (14). The fitness advantages of these Vpu activities are reflected in the pathogenicity of the different HIV-1 groups found in nature. Only the pandemic Group M viruses, and not the nonpandemic Group N, O, or P viruses, encode a fully functional Vpu that is capable of efficiently antagonizing both tetherin and the CD4 receptor (17).

The Vpu-mediated antagonism of NTB-A, PVR, and tetherin, has been shown to depend on the TMD of Vpu. Mutagenesis experiments performed in conjunction with flow cytometry and viral release assays have identified a conserved small-xxx-small amino acid motif (AxxxAxxxA) in the Vpu TMD as a putative site for all TMD interactions of Vpu (13, 18, 19). Based on proposed models of the Vpu homooligomer structure (20, 21), this alanine-rich face of the Vpu TM helix also mediates Vpu self-association, leading to a model in which Vpu oligomers must dissociate before interaction with human targets such as NTB-A, PVR, and tetherin. To our knowledge, there is no evidence for homooligomerization of the PVR or NTB-A TMDs; however, both full-length tetherin and the isolated tetherin TMD have been shown to exist as a parallel homodimer (22, 23). The full-length tetherin homodimer is stabilized by three conserved disulfide bonds in its extracellular coiled-coiled domain (24, 25), whereas the tetherin TMD homodimer forms in the absence of any obvious interaction motifs (22).

Although a structural model has been proposed for the heterooligomeric interaction between the TMDs of Vpu and tetherin (18), no such details have been reported for other Vpu TMD targets, including NTB-A and PVR. In fact, a direct TM interaction between the isolated TMDs of Vpu and NTB-A/PVR has not been previously demonstrated. In addition, the underlying energetics defining TMD interactions have not been explored for any of these systems, raising the question of how a single TMD is capable of mediating multiple homo- and heterooligomeric interactions although retaining target specificity. To address this, we have designed a Förster resonance energy transfer (FRET)-based assay to measure the binding of the Vpu transmembrane domain with the transmembrane domains of tetherin, PVR, and NTB-A. The association of TMD peptides in liposomal environments has been previously measured using FRET in systems of varying complexity, with the best characterized being TMD homodimers such as ErbB1, fibroblast growth factor receptor-3, or glycophorin A (26, 27, 28). More recent studies have investigated heterodimerization of transmembrane helices in the FtsB/FtsL complex of the Escherichia coli divisome (29), and several larger complexes such as the influenza A M2 channel tetramer and pentamers of phospholamban, have also been studied by FRET (30, 31). Likewise, heterodimeric interactions of ErbB family TMDs in detergent micelles have been studied by FRET, showing a hierarchy of interactions between family members, despite the presence of putative interaction motifs in each peptide (32). Although ErbB and FGFR display promiscuous binding within their respective protein families, there is significant sequence homology between the various interaction partners in each case (32, 33). Thus, neither of these systems display the level of promiscuity attributed to the Vpu TMD in terms of TMD recognition of nonhomologous binding partners.

Here, we provide evidence that dissociation of Vpu TMD homooligomers occurs before interaction with the TMD of each targeted host cell protein, such that the active form of Vpu is a monomer. Despite a lack of sequence homology, the target TMDs compete in our FRET studies for Vpu binding, suggesting that they share a common binding surface on Vpu. From lipid titration experiments we have calculated the dissociation constants (K_d) as well as the free energies of association (ΔG) for each TMD interaction within our system, with similar values for each oligomerization reaction. This supports a model in which Vpu does not have a clearly preferred target, allowing promiscuous interaction with multiple targets. These results provide, to our knowledge, new insight into the molecular mechanism of Vpu-mediated immune evasion.

Materials and Methods

Peptide synthesis, purification, and labeling

The transmembrane segments for Vpu, PVR, NTB-A, and tetherin were identified using the program TM-finder (http://tmfinder.research.sickkids.ca) (34). Based on this analysis, 0.1-mmol-scale synthesis of all TMD peptides (Table 1) was accomplished via automated solid phase FMOC chemistry on a PS3 peptide synthesizer (Protein Technologies, Tucson, AZ) using PalPEG resin (Applied Biosystems, Foster City, CA). Lithium chloride (50 mg/vial) was added during coupling to minimize hydrophobic interactions, preventing aggregation and truncation. Cleavage was carried out using a solution of 90% trifluoroacetic acid (TFA), 5% thioanisole, 3% 1,2-ethanedithiol, and 2% anisole for 2 h. The cleavage reaction was precipitated with ice-cold diethyl ether, isolated by centrifugation, resuspended in water, then frozen and lyophilized.

Table 1.

Vpu and Target Transmembrane Domain Peptides

Peptide Name	Amino Acid Sequence
TethTM	∗HHHHH19RCKLLLGIGILVLLIIVILGVPLIIFTIKAN49KKKK∗
VpuTM	1 MQPIQIAIVALVVAIIIAIVVWSIVIIEYRK31
VpuRD	MIPIIIAVILAVAVQAIVIVIVSWIIQEYRK
NTB-A TM	∗KK222TDTKMILFMVSGISIVFGFIILLLLVLRKRR252
PVR TM	∗KKK339MSRNAIIFLVLGILVFLILLGIGIYFYW368KKK∗

Predicted TM domain are underlined; N- and C-terminal polar tags are indicated with asterisks, where used.

TMD peptide purification was achieved by reverse-phase high performance liquid chromatography (HPLC) on either an 11 × 300 mm C4 or C8 column (Vydac), using either an acetonitrile/water or isopropanol/water gradient containing 0.1% TFA. HPLC fractions were lyophilized and peptide identity was confirmed by electrospray ionization mass spectrometry at the SPARC Biocentre (The Hospital for Sick Children, Toronto, ON, Canada).

N-terminal labeling of peptides with dansyl or dabsyl chloride (Invitrogen, Carlsbad, CA) was carried out on resin by adding 10 mg of dye and 50 μL of disopropyl ethylamine to 100 mg of resin in 2 mL of dimethylformamide. The reaction was capped with nitrogen and left rocking overnight, after which the resin was washed with methanol and cleaved normally. Purification of labeled peptide was carried out as described for unlabeled peptide. Labeling was confirmed by both HPLC and mass spectrometry and quantified using methods modified from Rath et al. (35). Briefly, the concentrations of dabsyl-labeled peptides in trifluoroethanol were calculated from their absorbance at 470 nm using a molar extinction coefficient of 2.96 × 10⁴ cm⁻¹ M⁻¹. Circular dichroism (CD) spectra were recorded and the mean residue ellipticity (MRE) at 222 nm was calculated for each peptide. This MRE was then used to backcalculate the concentrations of dansyl, dabsyl, or unlabeled peptides. This method was in good agreement with standard Trp absorbance methods where applicable (V and PVRTM). Fortunately, N-terminal labeling of peptides resulted in a significant increase in peptide retention during HPLC purification. This allowed for the isolation of completely labeled peptide, simplifying FRET analysis.

TMD peptide incorporation into liposomes

TMD peptides in a 1:2 trifluoroethanol/methanol solution were added to 1-palmitoyl-2-oleoylphosphatidylcholine (POPC) in chloroform. Final peptide concentrations of 1 mol % relative to phospholipid (peptide/lipid ratio of 0.01) were used for acceptor titration and competition experiments (see Figs. 2 and 3). The peptide-to-lipid ratios for lipid dilution experiments ranged from 0.67 to 0.02 mol %. The mixture was incubated to ensure complete dissolution and then dried under N_2(g). To ensure complete removal of solvent, the sample was resuspended in water, frozen, and lyophilized. The final peptide-containing liposomes were reconstituted in 50 mM sodium phosphate, pH 7.5, followed by five cycles of freeze-thawing. To reduce light scattering in spectroscopic measurements, small unilamellar vesicles were produced by 10 cycles of mild probe sonication, 1 s each, with sample cooling. Samples were left to equilibrate overnight in the dark at room temperature before fluorescence measurement.

Figure 2.

Homooligomerization of Vpu and tetherin TM domain peptides. Shown here is the donor quenching of dansyl-labeled VpuTM (A), or TethTM (B) in the presence of increasing mole fractions of dabsyl-labeled acceptor peptide. The peptide-to-lipid ratio was maintained at 0.01 in each sample by addition of unlabeled peptide; the total concentration of POPC was 1 mM. All experiments were carried out in 50 mM sodium phosphate buffer (pH 7.5) and liposomes were incubated overnight before measurements. Data is fit to the equation FRET Eff = K[1−(1−X)ⁿ⁻¹], where K is a constant and X is the acceptor mole ratio. The oligomer size, n, was equal to 2 for TethTM. An n of 3 was used for VpuTM, as this value resulted in the best fit of lipid titration experiments (Fig. S4), suggesting that this represents the average or most prevalent oligomeric species. Competition experiments with unlabeled peptide are shown for VpuTM (C) and TethTM (D). Each competition experiment consists of three samples: donor peptide only, 1:1 donor/acceptor, and 1:1:2 donor/acceptor unlabeled. Confidence intervals are reported as ns (p > 0.05), (^∗p ≤ 0.05), (^∗∗p ≤ 0.01), or (^∗∗∗∗p ≤ 0.0001).

Figure 3.

Heterooligomerization of VpuTM with its targets. Donor quenching of dansyl-labeled VpuTM is reported in the presence of increasing mole fractions of dabsyl-labeled target peptides (acceptor). The total peptide concentration was maintained at 10 μM by addition of unlabeled target peptide. Experiments were carried out in 50 mM sodium phosphate buffer (pH 7.5) and liposomes were incubated overnight before measurements. As in Fig. 2, experimental data was fit to FRET Eff = K[1−(1−X)ⁿ⁻¹], where n is equal to 2 for NTB-A and PVR, and 3 TethTM. To see this figure in color, go online.

CD spectroscopy

All CD measurements were performed using a model No. J-810 spectropolarimeter (JASCO, Oklahoma City, OK) and a 1-mm-path-length quartz cuvette (Hellma Analytics, Plainview, NY). Spectra were measured from 190 to 250 nm at a scan rate of 50 nm/min, and were recorded as an average of three measurements. Liposome samples contained 55 μM peptide in POPC at 4 mol % peptide relative to lipid, in 50 mM sodium phosphate buffer (pH 7.5). Spectra of samples containing peptides were subtracted from spectra of blank samples containing appropriate amounts of lipid and buffer. All CD measurements were made at room temperature.

Fluorescence spectroscopy

Dansyl fluorescence emission measurements were performed on a model No. C60 Spectrofluorimeter (Photon Technology International, Birmingham, NJ) scanning from 450 to 600 nm, using an excitation wavelength of 341 nm. Each emission spectrum was recorded as an average of three scans, acquired with integration set to 0.5 s, at a step size of 2 nm, and with excitation and emission slit widths of 4 nm. The area under the resultant curve (450–600 nm) was calculated for each measurement.

Analysis of FRET data

For unlabeled peptide competition experiments, the data are reported as normalized total fluorescence, with the integrated spectra of donor alone used to obtain a baseline fluorescence (F₀) to which all samples containing donor and acceptor (F) were normalized:

$$FRET=\left(\frac{F}{F_0}\right).$$ (1)

Acceptor and lipid titration experiments are reported in FRET efficiency (E), which is calculated by

$$E=1-\left(\frac{F}{F_0}\right).$$ (2)

The efficiency observed directly from experiments (E_expt) is a combination of FRET observed from specific oligomerization (E_oligo) as well as FRET due to the random colocalization of peptides (E_prox). In a lipid bilayer, donor- and acceptor-labeled peptides are limited to diffusion within the plane of the membrane, increasing the likelihood of FRET due to random proximity of donor and acceptor molecules. The estimated FRET due to proximity was subtracted from the experimental FRET efficiency for each peptide-to-lipid ratio tested in the lipid dilution experiments (see Fig. 6, A and B), resulting in only the FRET efficiency due to specific oligomerization:

$$E_{\text{oligo}}=E_{\text{expt}}-E_{\text{prox}}.$$ (3)

Values for E_prox, obtained using the method of Wolber and Hudson (36), are described in detail below.

Figure 6.

Thermodynamic analysis of Vpu TMD homo- and heterooligomeric assemblies. (A) Shown here is dilution of VpuTM (solid line) or TethTM (dashed line) in POPC liposomes, showing loss of FRET efficiency due to dissociation of the homooligomers. The TethTM data were fit to a monomer-dimer equilibrium using Eqs. 11, 12, and 13. VpuTM was modeled as a trimer using Eqs. 14 and 15. (B) Shown here is dilution of the VpuTM heterooligomers in POPC liposomes. Fit lines were generated from the K_d values calculated from Eqs. 16, 17, and 18 B. To see this figure in color, go online.

The FRET efficiency of an oligomer can also be presented as

$$E_{\text{oligo}}=f_{\text{oligo}}{p}_{\text{oligo}}{E}_r,$$ (4)

where f_oligo is the fraction of molecules in an oligomeric state, p_oligo is the probability of FRET occurring in an oligomer, and E_r is the FRET efficiency within the oligomer. The fraction of molecules in an oligomeric state can be computed as

$$f_{\text{oligo}}=\frac{n\left[\text{Oligo}\right]}{\left[\text{Total}\right]},$$ (5)

where [Oligo] is the concentration of the oligomeric species, [Total] is the total target peptide concentration, and n is the number of monomers making up the oligomer. The probability of FRET occurring in an oligomer is dependent on whether or not oligomers are formed exclusively by donor- or acceptor-labeled peptides, or by a mixture of the two. The probability of FRET occurring in a homooligomer can be calculated by the binomial distribution:

$$p_{\text{oligo}}=\sum _{k=1}^{n-1}\left(\begin{array}{c}n\\ k\end{array}\right)f_D^k{f}_A^{n-k}.$$ (6)

Here n is the size of the oligomer, k is the number of donors in the oligomer, and f_D and f_A are the mole fractions of donor and acceptor, respectively. This sum excludes the terms in which the oligomer will be made up of all acceptors, when k = 0, and when the oligomer is made up of all donors, when k = n. When probing the formation of Vpu-target heterooligomers p_oligo = 1 as any oligomers that form must have both a donor and an acceptor present. If the distance between the acceptor and donor dyes in an oligomer is smaller than the Förster radius of the FRET pair (i.e., in a parallel oligomer), then E_r = ∼1. Vpu and its target peptides have known orientations in cellular compartments and are all predicted to form parallel oligomers with the exception of Vpu-tetherin, which has been shown to be antiparallel. As the R₀ of the dansyl/dabsyl FRET pair approaches the bilayer thickness, an E_r of 0.5 was used for tetherin.

Combining Eqs. 3 and 4 relates oligomer concentration to FRET efficiency, which gives

$$\left[\text{Oligo}\right]=\frac{E_{\text{Oligo}}\left[\text{Total}\right]}{n{p}_{\text{oligo}}{E}_r}.$$ (7)

The oligomer fraction can then be used to calculate a dissociation constant given by

$$K_{d_{\text{Oligo}}}=\frac{{\left[\text{Monomer}\right]}^n}{\left[{\text{Oligomer}}_n\right]},$$ (8)

in which the monomer concentration [M] = [Total] − n[Oligo]. Finally, the free energy of oligomerization is defined as

$$\Delta G=RT\text{ln}\left(K_d\right).$$ (9)

Tetherin homodimerization

The equilibrium of tetherin homodimer formation in the absence of Vpu can be shown as

$$2Teth\stackrel{K_{d_{\text{Teth}}}}{\mathrm{\leftrightarrows }}Tet{h}_2.$$

The dissociation constant of this equilibrium can be computed as

$$K{d}_{\text{Teth}}=\frac{{\left[\text{Teth}\right]}^2}{\left[Tet{h}_2\right]},$$ (10)

where [Teth] and [Teth₂] are the tetherin monomer and dimer concentrations, respectively. The total tetherin concentration can be written as [Teth_Total] = [Teth] + 2[Teth₂], which can be rearranged and substituted into Eq. 9 to give

$$K_{d_{\text{Teth}}}=\frac{{\left(Tet{h}_{\text{Tot}}-2Tet{h}_2\right)}^2}{Tet{h}_2}.$$ (11)

The solution for this quadratic is

$$\left[Tet{h}_2\right]=\frac{4Tet{h}_{\text{Tot}}+K_{d_{\text{Teth}}}-\sqrt{{\left(4Tet{h}_{\text{Tot}}+K_{d_{\text{Teth}}}\right)}^2-16Tet{h}_{\text{Tot}}^2}}{8},$$ (12)

and can be related to FRET efficiency by

$$\left[Tet{h}_2\right]=E_{\text{Teth}}\left[Tet{h}_{\text{Tot}}\right].$$ (13)

Equations 12 and 13 were used to fit the FRET data in Fig. 5 A to calculate Kd_Teth, which was subsequently used in Eq. 9 to calculate the free energy of association, ΔG_Teth, for the tetherin homodimer.

Figure 5.

Target TM domain peptides compete for interaction with VpuTM. Unlabeled competition experiments were carried out using donor-labeled VpuTM in the presence of acceptor-labeled TethTM and equal molar amounts of each unlabeled target peptide. Competition experiments were completed under identical conditions to the VpuTM/TethTM homooligomer experiments (Fig. 4). Confidence intervals are reported as ns (p > 0.05) or ^∗(p ≤ 0.05).

Vpu homooligomerization

Like tetherin homooligomerization, Vpu homooligomer formation can be represented as

$$\begin{array}{c}nVpu\stackrel{K_{d_{\text{Vpu}}}}{\mathrm{\leftrightarrows }}Vp{u}_n,\\ K_{d_{\text{Vpu}}}=\frac{{\left(Vp{u}_{\text{Tot}}-nVp{u}_n\right)}^n}{Vp{u}_n},\end{array}$$ (14)

$$\left[Vp{u}_n\right]=\frac{E_{\text{Vpu}}\left[Vp{u}_{\text{tot}}\right]}{n\left(p_{\text{oligo}}\right)}.$$ (15)

Equations 14 and 15 were used to separately model Vpu homooligomerization for all possible oligomeric states proposed. This ranged from dimer (n = 2) to pentamer (n = 5). Unlike the quadratic function that results in tetherin homodimerization, Eq. 14 becomes quintic when n = 5, which cannot be solved algebraically. Therefore, we calculated the concentration of all the Vpu oligomers, [Vpu_n], for each individual sample using Eq. 15. The dissociation constant, Kd_Vpu, was then computed for each sample according to Eq. 14, and was repeated for n = 2 to n = 5 (Fig. S4). All K_d values for each value of n were averaged to give a value of Kd_Vpu for each oligomeric state, which was then used in Eq. 9 to give ΔG_Vpu.

Vpu NTB-A/PVR heterooligomerization

Heterooligomerization of Vpu with its human targets NTB-A and PVR is convoluted by its own homooligomerization as both interactions will occur simultaneously in any sample. However, conducting FRET experiments with exclusively donor-labeled Vpu and acceptor-labeled targets allows for the direct observation of heterooligomerization. A bilayer containing Vpu as well as NTB-A or PVR (NP) will result in the formation of three species of Vpu: homopentamers (Vpu₅), Vpu NTB-A/PVR heterodimers (VpuNP), and Vpu monomers (Vpu), the equilibria of which are

$$\begin{array}{c}3Vpu\stackrel{K_{d_{Vpu}}}{\mathrm{\leftrightarrows }}Vp{u}_3,Vpu+NP\stackrel{K_{d_{VpuNP}}}{\mathrm{\leftrightarrows }}VpuNP,\\ K_{d_{\text{VpuNP}}}=\frac{\left[\text{Vpu}\right]\left[\text{NP}\right]}{\left[\text{VpuNP}\right]},\end{array}$$ (16)

where [NP] is the concentration of free NTB-A or PVR and [VpuNP] is the concentration of Vpu dimerized with either PVR or NTB-A. As tetherin also undergoes homooligomerization, we will treat it separately below. Similar to tetherin homodimerization, the concentration of heterodimer, VpuNP, formed relates to FRET efficiency and total Vpu concentration as

$$\left[\text{VpuNP}\right]=E_{\text{VpuNP}}\left[Vp{u}_{\text{Tot}}\right].$$ (17)

In this context, there are only two species of NTB-A or PVR that can be formed—monomeric NTB-A/PVR, or Vpu-bound NTB-A/PVR. Equation 17 can be used to calculate the concentration of [VpuNP], and from this, [NP] can be calculated by [NP] = [NP_Tot] – [VpuNP]. However, Vpu is also involved in homooligomerization, thus [Vpu] ≠ [Vpu_Tot} – [VpuNP]. Instead [Vpu_Tot] – [VpuNP] = [Vpu_unbound], where [Vpu_unbound] is a mixture of Vpu monomer and trimeric Vpu. The fit from Fig. 6 A was then used to backcalculate the concentration of trimer and monomer concentrations of Vpu from [Vpu_unbound]]. The Vpu monomer concentration was then used in Eq. 16 to calculate Kd_VpuNP and subsequently ΔG_VpuNP.

Vpu tetherin heterooligomerization

Vpu-tetherin heterooligomerization is treated in a similar fashion to Vpu NTB-A/PVR heterooligomerization with the addition of parameters describing tetherin homooligomerization. As discussed in the Results and Discussion below, we have proposed two possible models that describe Vpu-tetherin heterooligomerization. Model 1 is where only the tetherin homodimer binds to Vpu, forming a heterotrimer. Model 2 is where both the tetherin homodimer and tetherin monomer bind to Vpu, resulting in a mixture of Vpu-tetherin heterodimers and heterotrimers. The equilibria of both models are as shown in Table 2.

Table 2.

Vpu-Tetherin Heterooligomerization Models

Model 1	Model 2
$3Vpu\rightleftarrows Vp{u}_3,$	$3Vpu\rightleftarrows Vp{u}_3,$
$Vpu+Tet{h}_2\rightleftarrows VpuTet{h}_2,$	$2Vpu+Tet{h}_2+\text{Teth}\rightleftarrows VpuTet{h}_2+VpuTeth,$
$2Teth\rightleftarrows Tet{h}_2$	$2Teth\rightleftarrows Tet{h}_2$
$$\left[VpuTet{h}_2\right]=\frac{E_{VpuTet{h}_2}\left[Vp{u}_{Tot}\right]}{E_r}$$ (18A)	$$\left[VpuTet{h}_{1/2}\right]=\frac{E_{VpuTet{h}_{1/2}}\left[Vp{u}_{Tot}\right]}{E_r}$$ (18B)
$$K{d}_{VpuTet{h}_2}=\frac{\left[Vpu\right]\left[Tet{h}_2\right]}{\left[VpuTet{h}_2\right]}$$ (19A)	$$K{d}_{VpuTet{h}_{1/2}}=\frac{{\left[Vpu\right]}^2\left[Tet{h}_2\right]\left[Teth\right]}{\left[VpuTet{h}_2\right]\left[VpuTeth\right]}$$ (19B)

In Model 1, the concentration of the VpuTeth₂ heterotrimer, [VpuTeth₂], was calculated using Eq. 18 A, whereas [Vpu] was found using the same methods discussed above for Vpu-NTB-A/PVR heterooligomer formation. Tetherin homodimer and monomer concentrations were then calculated using [Teth_unbound] = [Teth_Tot] − 2[VpuTeth₂] and the tetherin fit from Fig. 5 A. All component concentrations were then used in Eq. 19 A to calculate Kd_VpuTeth2. In Model 2, [VpuTeth_1/2], calculated by Eq. 18 B, is a mixture of Vpu bound to either tetherin monomer [VpuTeth], or tetherin dimer [VpuTeth₂]. As Vpu did not have a notable effect on tetherin homodimerization (Fig. S5), we assumed that the tetherin dimer fraction remained constant whether or not Vpu was present. Therefore, the ratio of tetherin dimer to monomer in isolation would be equivalent to the ratio of Vpu bound to tetherin dimer to Vpu bound to tetherin monomer. Equations 18 B and 19 B were then used to calculate the parameters in Model 2.

Proximity corrections

A general approximation of FRET between donors and acceptors distributed randomly in a 2D plane, such as a lipid bilayer, has been previously described by Wolber and Hudson (36). Here we use their method to remove FRET arising out of the random colocalization of peptides from our experimental data. As our experimental conditions do not allow for the control of peptide orientation in the bilayer, any estimation of FRET due solely to proximity must include parameters that account for random FRET that has occurred on the same side of the membrane as well as cross-bilayer interactions. The calculations require the 2D concentration of acceptors (per unit area), the Förster radius R₀, and the distance of closest approach R_e (which was varied to account for cross-bilayer FRET). The 2D concentration of acceptors was calculated using the acceptor-to-lipid ratio and the area of a POPC headgroup, 62.7 Å² (37). The R₀ value of the dansyl/dabsyl FRET pair is 33 Å (30). The distance of closest approach for random FRET occurring within the same plane was zero (as we assumed the space taken up by the dye itself was negligible), where the distance of closest approach for random FRET occurring across the bilayer was the bilayer’s hydrophobic thickness 29.2 Å² (37). The values for random FRET from both sides of the bilayer were summed and subtracted from experimental data for each lipid-to-peptide ratio used.

Statistical analysis

Statistical analysis was carried out using the softwares GraphPad Prism 6.0 (GraphPad Prism, La Jolla, CA) or SigmaPlot 11.0 (Systat Software, Chicago, IL). In all cases error bars represent the standard error of the mean. Where appropriate, confidence intervals were calculated using unpaired tests and are reported in nanoseconds (ns) as: p > 0.05, ^∗p ≤ 0.05, ^∗∗p ≤ 0.01, ^∗∗∗p ≤ 0.001, ^∗∗∗∗p ≤ 0.0001.

Results and Discussion

TMD peptides retain nativelike secondary structure in POPC liposomes

The interaction between isolated TMDs of Vpu and tetherin has been documented previously using both in-cell assays and in vitro NMR, although the energetics of this binding event have not been characterized (18, 19). Similarly, little is known about the TMD interactions between Vpu and NTB-A/PVR. To investigate the energy landscape of Vpu-target interactions within the membrane, we have synthesized peptides containing the transmembrane domains of all four proteins, as well as a peptide composed of a scrambled Vpu TMD sequence (VpuRD) previously shown to be incapable of antagonizing target proteins (Table 1) (15). Where required, polar tags have been added to the termini of the TMD peptides. This strategy has been shown to improve solubility and facilitate the purification of TMD peptides without effecting their ability to form their native oligomeric states (38).

Before probing the peptides for potential TMD interactions it was necessary to first determine if they adopted the expected secondary structure upon reconstitution into liposomes. In this study, POPC liposomes were used because this is an abundant phospholipid in human membranes and has favorable physiochemical properties, including zwitterionic charge and a low phase transition temperature. Isolated synthetic peptides were incorporated into POPC liposomes by codissolution in organic solvent and CD spectra were recorded (Fig. 1). As expected, all peptides display the characteristic minima at 208 and 222 nm, indicating that all TMD peptides adopt a nativelike α-helical structure in model membranes.

Figure 1.

Secondary structure of TMD peptides reconstituted into POPC liposomes. CD spectra were recorded for each peptide at a peptide-to-lipid ratio of 0.04 and are reported as MRE. All spectra were recorded at 25°C in 50 mM sodium phosphate buffer (pH 7.5). The peptide concentration was 55 μM in each case. To see this figure in color, go online.

Homooligomerization of Vpu and tetherin

Our group has previously shown that the tetherin TMD forms homodimers in various membrane mimetics, and self-assembly of the Vpu TMD has been demonstrated by several groups (22, 39, 40). Although the size estimates of the Vpu homooligomer have varied (20, 39, 40), a pentameric structure is consistent with most experimental data. However, some studies have suggested the presence of multiple oligomeric states of Vpu in lipid bilayers and detergent micelles (21, 41). To confirm the formation of tetherin and Vpu TMD homooligomers in our system, we carried out FRET experiments, which involve the nonradiative transfer of energy from a fluorescent donor to an acceptor. The dansyl fluorophore can be selectively excited in the presence of dabsyl-labeled peptides (Fig. S1) and this FRET pair has been previously used in TM oligomerization studies (30, 42). The quenching of fluorescence emission from donor-labeled peptides was measured in the presence of increasing amounts of acceptor-labeled peptide, while maintaining a constant peptide-to-lipid ratio through addition of unlabeled peptide (Fig. 2, A and B). Whereas a linear response to this titration is indicative of dimer formation, departure from linearity corresponds to the formation of higher-order oligomers (28, 30, 43). The linear fit for the TethTM titration supports our previous observation of a tetherin TMD dimer, whereas the results of the Vpu titration is best fit to a nonlinear curve, consistent with formation of higher-order oligomers from trimer to hexamer in size (Fig. S2). Here the curves describing trimer, tetramer, and pentamer formation provide equally good fits to the acceptor titration data, preventing unambiguous determination of Vpu oligomer size. Therefore, each oligomer size was treated separately in analysis of lipid dilution FRET experiments (described below) to determine the best fit.

Competition experiments using unlabeled TMD peptides were carried out to ensure that the FRET data reflect specific oligomerization, rather than nonspecific peptide interactions caused by proximity in the lipid bilayer. The results for Vpu (Fig. 2 C) and tetherin (Fig. 2 D) both show a significant reduction of FRET after the addition of unlabeled peptide, indicating that the interactions observed are specific. Both NTB-A and PVR TMDs were also tested for self-association, but neither showed significant FRET in the presence of acceptor-labeled peptide or any change in FRET signal during unlabeled competition experiments, indicating that the transmembrane domains of both NTB-A (Fig. S3 A) and PVR (Fig. S3 B) do not form homooligomers.

VpuTM binds specifically to the transmembrane domains of tetherin, NTB-A, and PVR

To characterize the association between VpuTM and the target TMDs, FRET experiments were carried out in which the donor quenching of labeled VpuTM was measured in the presence of increasing amounts of acceptor-labeled target peptides (Fig. 3). FRET was observed between donor-labeled VpuTM and all three acceptor-labeled targets. Again, the specificity of these interactions was validated with unlabeled control experiments (Fig. 4, A–C). The specificity of these interactions was further demonstrated using a scrambled Vpu peptide (VpuRD), which was unable to bind specifically to any of the target peptides (Fig. 4, D–F). It is important to note that this study does not differentiate between parallel and antiparallel interactions, as the R₀ value of the dansyl/dabsyl FRET pair is close to the thickness of the lipid bilayer (30). Previous studies have confirmed that Vpu-Vpu and tetherin-tetherin self-interactions are parallel in nature, whereas the Vpu/tetherin heterooligomer is antiparallel. When accounting for the known topologies of Vpu, PVR, and NTB-A in cellular membranes, it is expected that these heteroassociations will be parallel (44, 45).

Figure 4.

Competition experiments with unlabeled target peptides confirm the interactions between Vpu and target TMD peptides. Relative fluorescence yields are shown for samples containing donor-labeled Vpu with acceptor-labeled and/or -unlabeled (A) TethTM, (B) NTB-A TM, and (C) PVR TM. Negative controls using VpuTM RD are shown for (D) TethTM, (E) NTB-A TM, and (F) PVR TM. Competition experiments were completed using the same donor/acceptor-unlabeled ratios described in Fig. 2. Confidence intervals are reported as ns (p > 0.05), (^∗p ≤ 0.05), (^∗∗p ≤ 0.01), or (^∗∗∗∗p ≤ 0.0001).

The acceptor titration experiments for Vpu binding to PVR and to NTB-A (Fig. 3) show a linear response, suggesting that a heterodimer is formed in each case. However, the titration curve for the Vpu-tetherin interaction is nonlinear, signifying the formation of a higher-order heterooligomeric species. McNatt et al. (46) observed the binding of Vpu monomers to tetherin dimers in cross-linking experiments performed using the full-length proteins in vivo, suggesting that the Vpu binding face on the tetherin TMD is distinct from the tetherin homodimer interface. Thus, whereas our data cannot rule out the possibility of simultaneous formation of both 1:1 and 1:2 VpuTM/TethTM complexes, the nonlinear response in the Vpu tetherin acceptor titration experiment demonstrates that a heterotrimer is likely to be present in our system.

Target peptides compete for binding with Vpu

Establishing that NTB-A, PVR, and tetherin each bind to the Vpu TMD raised the question of whether all three target peptides share the same binding face on Vpu. To address this, unlabeled competition experiments were carried out in which dansyl fluorescence was measured for samples containing donor-labeled Vpu and acceptor-labeled tetherin in the presence and absence of unlabeled NTB-A or PVR (Fig. 5). In each case fluorescence was restored in the presence of unlabeled peptide, indicating that NTB-A and PVR TMD peptides are able to compete with the tetherin TMD for association with Vpu.

Currently there are no proposed Vpu binding surfaces for the NTB-A or PVR TMDs, but the ability of these peptides to compete with tetherin for binding to Vpu in liposomes suggests that these proteins are likely to also compete for binding in cell membranes. This supports the hypothesis that all three peptides share the same binding face on VpuTM, or that there is significant overlap between binding sites. This is supported by previously reported in vivo mutagenesis of the alanine face of Vpu (A10/A14/A18), in which loss of the small-XXX-small motifs were shown to reduce Vpu-mediated antagonism of tetherin and PVR (13, 15). Remarkably, the TMDs of the three Vpu targets studied here do not share any identifiable sequence motifs that could account for their affinity for the same face of the Vpu TMD. This is particularly intriguing because the Vpu TMD is able to form both antiparallel interactions with the tetherin TMD and parallel heterodimers with PVR or NTB-A, using the same binding surface in each case. Because this face of the Vpu helix is also proposed to be critical for Vpu homooligomerization (20), our data support a model in which the Vpu monomer interacts with target proteins. This could potentially result in the disassembly of higher-order Vpu oligomers capable of conducting ions, a functionality that is been highly conserved across the pandemic strains of HIV-1, although it remains of unknown biological significance (47).

Energetics of Vpu-target interactions

To determine if different binding energetics for each interaction may play a role in target specificity, we calculated the relative affinities of VpuTM-target oligomers, using FRET data recorded at varied lipid-to-peptide ratios, an approach validated by previous studies (29, 48). As transmembrane peptides are solubilized in a lipid bilayer rather than bulk solvent, the lipid-to-peptide ratio is the key factor in determining the fraction of oligomeric species (49). Although the unlabeled FRET competition experiments described above demonstrate the presence of a specific interaction in each case, determination of accurate dissociation constants (K_d) and free energies of association (ΔG) requires subtraction of proximity effects from the lipid titration data. The method of Wolber and Hudson (36), described in detail in the Materials and Methods, was used to estimate the FRET efficiency caused by proximity effects and subtract it from our lipid dilution experiments. The corrected FRET efficiency can then be used to calculate the amount of oligomerization, and from this the K_d value for each TMD interaction. Fig. 6, A and B, shows the proximity corrected FRET data used in the calculation of K_d and ΔG values for TMD homo- and heterooligomers, respectively. Determination of these parameters for the VpuTM homooligomer was initially impeded because the acceptor titration data for VpuTM could not be fit to a single oligomeric state with any confidence (Fig. S2). Thus, the lipid titration data were fit to curves calculated assuming that VpuTM existed as a single species ranging from dimer to pentamer in size. The results in Fig. S4 show a best fit to a VpuTM trimer, but this likely represents the average FRET efficiency of an equilibrium mixture containing Vpu oligomers of various sizes. Such a mixture of species has previously been observed in chemical cross linking and analytical ultracentrifugation studies in which dimers and trimers were the dominant species, along with smaller numbers of tetramers, pentamers, and high-order oligomers (21). To simplify the following Vpu-target calculations we assumed a model in which Vpu is trimeric.

The self-association of VpuTM adds another dimension to determining the K_d and ΔG values for Vpu-target interactions because three possible Vpu-containing species can be formed: monomeric Vpu, homooligomeric Vpu, and target-associated Vpu. Merzlyakov et al. (50) have used an elegant method for FRET data analysis to calculate the energetics of heterodimerization of two peptides that also form homodimers. Here, we have modified this approach to account for the higher-order homooligomerization of VpuTM. Additionally, although it was clear from the acceptor titration experiments that Vpu formed heterodimers with PVR and NTB-A, there was evidence of Vpu-tetherin heterotrimers, as described above. The addition of excess VpuTM to samples containing both donor- and acceptor-labeled TethTM did not have an effect on tetherin TMD dimer formation (Fig. S5), and thus two possible models can explain Vpu-tetherin oligomer formation: 1) the Vpu monomer binds only to tetherin dimers, or 2) Vpu binds to both monomeric and dimeric tetherin.

We have calculated the K_d and ΔG values for all VpuTM-target interactions, including both possible VpuTM-TethTM models. These are summarized in Fig. 7, and show that the promiscuous interactions observed for Vpu are relatively weak relative to other TM domain interactions. In each case, the free energy of association for two helices forming a homo- or heterodimer is from −1.7 to −2.6 kcal/mol. For VpuTM homooligomers or Vpu-TethTM heterooligomers, this value is shown in parentheses in Fig. 7. Formation of VpuTM or VpuTM-TethTM trimers is more energetically favorable due to the additive effects of multiple TMD-TMD interactions. Although the association of VpuTM with PVR-TM is slightly weaker than binding to TethTM or NTB-A TM, all of these are relatively weak interactions when compared with known strong TM helix dimers such as glycophorin A, which has a ΔG of dimerization from −5 to −5.7 kcal/mol in phosphatidylcholine membranes (51). The ΔG values reported here for Vpu-target interactions (and for individual Vpu monomer contributions to the ΔG of homotrimers) are closer in magnitude to those reported for receptor tyrosine kinases, such as ErbB1 and fibroblast growth factor receptor-3, which have ΔG values for dimerization of ∼−2 kcal/mol (26, 27, 52).

Figure 7.

Thermodynamic and kinetic analysis of the VpuTM domain interactions with target TM domain peptides. Free energies (kcal mol⁻¹) and dissociation constants are shown for homooligomeric and heterooligomeric interactions of Vpu and target transmembrane domain peptides. Parameters were calculated from the data shown in Fig. 6, A and B. Values in parentheses are for the contribution of individual monomers to the Vpu homotrimer, or the Vpu-tetherin heterotrimer.

The observation of uniformly weak Vpu TMD interactions may have several possible explanations. First, there may be a requirement for release of Vpu before degradation of the target proteins, maintaining Vpu levels in the infected cell. Second, it may not be possible to design a TMD that is capable of strong interactions with all possible targets, especially given the capacity for Vpu to engage in both parallel and antiparallel helical interactions. Third, strong interactions would reduce the free monomer present, possibly preventing interaction with multiple partners. It is not known if an increase in TM affinity will correlate with an increased level of Vpu-mediated antagonism in vivo. Thus, weak interactions that result in antagonism of multiple targets are likely to provide a greater fitness advantage to HIV-1 than strong binding to a single target. This may explain why the alanine-rich motif in the TMD of Group M Vpu has been strongly conserved despite the high HIV-1 mutation rate. The impact of heterodimer formation on the putative ion channel activity of Vpu remains unclear, but our observations may suggest a careful balance between the ability to target host cell proteins through their TMDs while maintaining the capacity for homooligomerization.

The subtle variations in affinity for each binding event suggest the possibility that different intermolecular interactions define the binding of Vpu to different targets, even if the same face of VpuTM is required in each case. A broad range of mechanisms through which transmembrane helix associations can be stabilized have been identified: the small-xxx-small motif, cation-π interactions, π-π stacking, ILV packing, and Cα-H-backbone hydrogen bonds can all contribute to helix-helix dimer stabilization (53). In the case of Vpu, the highly conserved small-xxx-small motif is very likely to play a key role in target TM association. This model of Vpu tetherin TMD association involves the ³⁰VxxxIxxLxxxL⁴¹ face of tetherin packing into the small-xxx-small motif of Vpu (18). However, without atomic level structures for each complex, it is not known what contacts this motif makes with each target helix. Both PVR TM and NTB-A TM contain multiple faces that display large hydrophobic residues, which could pack against the alanine-rich face of VpuTM. Studies of glycophorin A have demonstrated that relatively conservative mutations in β-branched amino acids (i.e., Val to Ile) can significantly enhance dimerization levels (54). Thus, small changes in the binding face displayed by target TMDs could explain any slight variations in affinity toward VpuTM. To our knowledge, this is the first study showing that a single TMD can bind with similar affinity to highly varied targets that do not share a common membrane topology or amino acid sequence motifs. Further work will be necessary to identify the precise mechanism of helix-helix association between Vpu and its different targets, and to determine whether this variance plays a biological role in preferential target selection.

Author Contributions

G.B.C. carried out experiments and analyzed the data. S.E.R. assisted in data analysis and interpretation. S.S. and G.B.C. designed experiments and wrote the manuscript.

Editor: Kalina Hristova.