HeLa and BLV-FLK (a kind gift of Dr Arsène Burny and Dr Luc Willems; Universitaire des Sciences Agronomiques de Gembloux, Belgium) cells were maintained in Dulbecco's modified Eagle medium supplemented with 10% fetal bovine serum (FBS).

The Plasmid HTE-1 28 and pRSV-Rev 29 have been described. The plasmid pCMV-BLVenv-RRE was constructed by replacing a fragment of the HIV-1 envelope open reading frame in pCMVgp160ΔSA 30 with a genomic fragment spanning the entire BLV envelope. In brief, pCMVgp160ΔSA was digested with EcoR I, which cuts the recipient vector after the CMV early promoter but prior to the initiating ATG of the HIV-1 env sequences. The vector was subsequently digested with BglII, which removes the HIV-1 SU region but retains the HIV RRE. A fragment encompassing the entire BLV envelope open reading frame between a 5' Xho I site and a 3' BamH I site (nucleotides 4347–6997 of NC_001414) was ligated into the vector backbone using an EcoR I-Xho I linker. The resulting plasmid encodes BLV env including the natural BLV env stop codon placed upstream of the HIV RRE; the transcription unit is terminated by the SV40 poly A site and is expressed from the CMV early promoter.

Peptides (Table 1) were synthesized using standard solid-phase Fmoc chemistry and unless stated otherwise have acetylated N-termini and amidated C-termini. The peptides were purified by reverse-phase high-pressure liquid chromatography and verified for purity by MALDI-TOF mass spectrometry. All peptides were dissolved in dimethyl sulfoxide (DMSO), the concentration of peptide stock solutions was confirmed where possible by absorbance at 280 nm in 6 M guanidine hydrochloride and peptides were used at the final concentrations indicated. For the peptide P^BLV-ΔN, peptide concentration was estimated by Bradford assay at 5 two-fold serial dilutions from a stock solution using the P^BLV-ΔC peptide in concentrations verified by absorbance at 280 nm in 6 M guanidine hydrochloride to plot a standard curve. The HTLV-1-derived peptides are based on the sequence of HTLV-1 strain CR and conform closely to the consensus sequence for HTLV-1 and HTLV-2 strains, the BLV peptides conform to the consensus sequence for most BLV isolates.

Peptides to be biotinylated were reduced using immobilized Tris [2-carboxyethyl] phosphine (TCEP) reducing agent (Pierce), and subsequent biotinylation was carried out with EZ-Link^® Iodoacetyl-PEO₂-Biotin (Pierce), in both cases according to the manufacturer's protocols. The biotinylation reaction was quenched with cysteine. The biotinylated peptide was incubated for 30 mins at room temperature with either streptavidin-agarose (Gibco-BRL) or amylose-agarose (New England Biolabs) in a spin-column. Unbound peptide was recovered by centrifugation, the flow-through was re-applied to the column, and the incubation and centrifugation was repeated. The flow-through from the second centrifugation was used in syncytium interference assays; the peptide concentration of the amylose-agarose flow through was established by UV spectrometry as described above, and added to tissue culture medium to produce the final assay concentrations as indicated. In the case of the flow-through from the streptavidin-agarose column, volumes equivalent to those used with the amylose-agarose flow-through were added to the wells.

A two-fold serial dilution of peptide in DMSO was performed, and added in duplicate to 96-well microplates. Filtered PBS was added to give a total volume of 200 μl and a final DMSO concentration of 1.5 % in all wells. The plates were incubated at room temperature for 1 hr and the relative solubility of peptides was established by measuring forward scattered light using a NEPHELOstar laser-based microplate nephelometer (BMG LABTECH). Wells containing PBS and 1.5 % DMSO only were used as blanks. Data analysis was carried out using ActivityBase, and peptides giving readings up to and including 3-fold higher than the average reading for the DMSO control were considered to be in solution at the concentrations specified.

Syncytium interference assays were performed by standard methods 2031. Briefly, HeLa cells for use as effector cells were transfected with the envelope expression vector pHTE-1 or with equal amounts of pCMV-BLVenv-RRE and pRSV-Rev using the Genejuice™ transfection reagent (Novagen) in accordance with the manufacturer's instructions. 24 h later, 3 × 10⁵ effector cells were added to 7 × 10⁵ untransfected HeLa target cells in six-well dishes (Nunc). Where appropriate, the co-culture was incubated in the presence of peptides at the concentrations specified. To assess the ability of the peptides to inhibit fusion induced by virally expressed BLV envelope, 2 × 10⁵ BLV-infected FLK cells were used as effectors and added to 8 × 10⁵ uninfected HeLa cells. After incubation at 37°C for 16 h, cells were washed twice with PBS and fixed in PBS + 3% paraformaldehyde. Assays were performed in triplicate and the number of syncytia (defined as multinucleated cells with 4 or more nuclei) from 10 low-power fields (LPF) per replicate was scored by light microscopy; some assays were stained using Giemsa. A syncytium formation value of 100% is defined as the number of syncytia formed in the absence of peptide but in the presence of 1.5% DMSO. The peptide concentration required to give 50% inhibition (IC₅₀) of syncytium formation was calculated using GraphPad Prism 4.

Although there are considerable differences in the amino acid sequence of class-1 fusion proteins from diverse viral groups there is exceptional conservation of secondary and tertiary structure. To compare the class-1 fusion proteins from the related retroviruses BLV and HTLV-1, the predicted coiled-coil regions of the BLV TM were identified using the program LearnCoil-VMF 32 and the BLV and HTLV-1 amino acid sequences were aligned using Clustal-W 33 (Figure 1A). The alignment revealed that for the TM 33% of the residues are identical and a further 10% are conservative substitutions. The homology is particularly evident in the predicted coiled-coil region incorporating the heptad repeat and in the LHR of the TM ectodomain (Figure 1A), the LHR lies distal to a CX₆CC motif common to oncoretroviral fusion proteins. The crystal structure of the HTLV-1 six-helix-bundle has been solved and the structure spans these regions of homology 16.

Using the crystal structure of the HTLV-1 TM as a template, we mapped on the coiled coil and LHR the location of amino acid residues that are conserved between the ectodomain of HTLV-1 and BLV TM (Figure 1B). Using this approach, we observed that for the core coiled-coil the majority of conserved residues map along the grooves formed by the interface of each pair of interacting N helices. Importantly, these grooves act as docking sites for the LHR as TM folds from the pre-hairpin intermediate to the trimer-of-hairpins. Moreover, many of the conserved amino acids of the LHR are located on the face of the LHR that interacts with the grooves on the coiled coil. By examining the location of substituted residues on the HTLV-1 TM it becomes clear that where there are amino acid substitutions on the BLV LHR there are complimentary or accommodating amino acid changes within the hydrophobic grooves of the core coiled coil (Figure 1C). For example, leucines 413 and 419 in the HTLV-1 LHR are conserved in BLV, and these leucines interact with eight coiled coil residues of which seven are identical in BLV and one is a conservative substitution (Figure 1C). In contrast, HTLV-1 LHR residues H409 and R416 interact with the side chains of six residues of the coiled coil, but H409 and R416 are not conserved in BLV and of the six interacting coiled coil residues four have diverged and only one residue is semi-conserved (Figure 1C). Overall, the analysis indicates that the majority of the conserved residues occupy positions that form the interacting surfaces of the trimer-of-hairpins. In agreement with these observations, those residues that do not involve the interacting surfaces of the TM are invariably solvent exposed on the trimer-of-hairpins and are subject to the highest degree of variation between the two viruses.

A synthetic peptide, P^cr-400, which mimics the LHR of the HTLV-1 TM is a potent inhibitor of envelope-catalysed membrane fusion 20. This peptide interacts directly and specifically with a recombinant coiled coil derived from HTLV-1 TM and substitution of critical amino acid residues within the peptide disrupts coiled coil binding and impairs the biological activity of the peptide 202122. These findings are consistent with the view that the peptide blocks membrane fusion by binding to the coiled coil of fusion-active envelope. As illustrated above, there are remarkable similarities in the interacting surfaces of the coiled coil and LHR between HTLV-1 and BLV (Figure 1). Considering the noted differences, it was not clear if the HTLV-1-derived synthetic peptide could inhibit membrane fusion mediated by BLV envelope. The HTLV-1 peptide inhibits viral entry by the divergent HTLV-2 but does not inhibit membrane fusion catalysed by a number of heterologous viral envelopes including HIV-1, feline immunodeficiency virus and vesicular stomatitis virus G protein 2023 (our unpublished results). Moreover, the HTLV-1 inhibitory peptide is unusual among C helix-based fusion inhibitors in that it includes both α-helical and extended non-helical peptide segments. It was therefore uncertain if peptides based on the LHR of BLV would, like the HTLV-mimetic peptide, display anti-fusogenic activity. We therefore compared the fusogenic activity of HTLV-1 and BLV envelope and examined the sensitivity of BLV envelope to inhibition by peptide inhibitors.

Preliminary experiments with a variety of BLV envelope expression constructs produced only low levels of BLV envelope expression and little fusogenic activity in syncytium formation assays (data not shown); this may, in part, be due to the nuclear retention of the envelope transcripts as observed for HIV-1 and HTLV-1. Therefore, we developed an envelope expression vector whereby BLV env was inserted downstream of the strong cytomegalovirus (CMV) early promoter, and immediately upstream of the human immunodeficiency virus Rev-response element (RRE). The RRE forms a region of extensive secondary structure in the mRNA that is recognized by Rev and the resulting ribonucleoprotein complex is subsequently exported out of the nucleus. The BLV envelope expression construct was examined for envelope-induced membrane fusion in syncytium formation assays. Briefly, HeLa cells were either transfected with pCMV-BLVenv-RRE or pRSV-Rev individually, or cotransfected with equal amounts of both vectors. These cells were then used as effector cells to induce syncytia when co-cultured with non-transfected cells. Neither vector induced syncytium formation when transfected alone, but cotransfection of effector cells with pCMV-BLVenv-RRE and pRSV-Rev resulted in the widespread formation of large syncytia (Figure 2). Furthermore, BLV envelope expressed in this system produced levels of syncytia that were comparable to that of HTLV-1 envelope expressed from pHTE-1 and consequently this approach was used to express BLV envelope for these studies.

To compare the inhibitory properties and specificity of LHR-based synthetic peptides from HTLV-1 and BLV a peptide based on the LHR of BLV was generated. The synthetic peptide designated P^BLV-391 includes residues Cys391 to Gln419 of BLV Env and spans a region that is equivalent to the HTLV-1 LHR-derived peptide P^cr-400 (Table 1). To aid comparison with TM, we refer to the residues of each peptide using the co-ordinates for the full-length envelope precursor (thus for the BLV-derived peptide residue 1 is referred to as Cys391). The BLV and HTLV-1 peptides share 45 % identity (Figure 1A, B), but it should be noted that only a fragment of the HTLV-1 LHR that is mimicked by P^cr-400 is resolved in the available HTLV-1 TM crystal structure (Table 1, Figure 1) 20.

Both HTLV-1 and BLV envelope induced widespread syncytium formation in cultures incubated in the absence of peptide inhibitors or in the presence of inactive control peptides (Figure 3A, B). However, in keeping with previous studies 202122, HTLV envelope-mediated syncytium formation was robustly blocked in a dose-dependent manner by P^cr-400 with an IC₅₀ of 0.28 ± 0.01 μM (Figure 3A). However, despite the marked conservation of amino acid sequence between the LHRs and coiled coils of HTLV-1 and BLV, P^cr-400 failed to inhibit membrane fusion induced by BLV envelope even at concentrations up to 15 μM (Figure 3B) and above (data not shown). Also, like the inactive control peptides, the BLV LHR-mimetic peptide at concentrations up to 20 μM (Figure 3A) and above (data not shown) failed to inhibit membrane fusion induced by HTLV-1 envelope. By contrast, the peptide P^BLV-391 specifically antagonized BLV envelope-mediated membrane fusion (Figure 3B) with a calculated IC₅₀ of 3.49 ± 0.03 μM; control peptides including C34 and P^cr-400 L/A did not interfere with BLV Env-induced membrane fusion (Figure 3B). In addition, P^BLV-391 robustly antagonized membrane fusion induced by virally expressed envelope as shown by the inhibition of syncytium formation between chronically BLV infected FLK cells and target cells (Figure 3C); whereas, the HTLV-1 peptide inhibitor did not block BLV-induced membrane fusion. Thus, it appears that the inhibitory properties of the LHR-mimetic peptides are highly specific to the virus from which they were derived.

Our group recently demonstrated that truncations at the N- or C-terminal end of P^cr-400 abolished fusion-inhibitory function 29. To test whether or not the N- and C-terminal leash regions are required for the activity of P^BLV-391, we synthesized two peptides, P^BLV-ΔN and P^BLV-ΔC, which lack nine amino acid residues at the N-terminus or C-terminus respectively (Table 1). The peptides retain an eleven-residue overlap, and have solubility profiles comparable to the parental peptide P^BLV-391 (Table 1). Unlike the parental peptide, the peptide derivatives P^BLV-ΔN and P^BLV-ΔC lacked detectable inhibitory activity in syncytium interference assays (Figure 4A). These data illustrate that amino acid residues contained within the regions Cys391 to Asp399, and Ser411 to Gln419, are critical to the activity of the mimetic peptide, and that both the amino-terminal and C-terminal regions are necessary but not sufficient for antagonism of membrane fusion. Importantly, the data also demonstrate that the central 11-residue region of the BLV peptide, equivalent to Ser400-Leu410 and homologous to the short C-terminal α-helix of the HTLV-1 trimer-of-hairpins is not sufficient for inhibition of syncytium formation.

Moreover, the BLV peptide was remarkably intolerant of even relatively small deletions. For example, a peptide, P^BLV-ΔCCF, in which only 3 amino acids were deleted from the N-terminus exhibited dramatically reduced ability to inhibit membrane fusion (Figure 4B). The P^BLV-ΔCCF peptide blocked syncytium formation by only 30% at 20 μM (Figure 4B), compared to > 95% for the parental peptide, and even at a concentration of 30 μM peptide P^BLV-ΔCCF achieved only 40% inhibition (data not shown). These results can be explained only in part by the decrease in peptide solubility at concentrations above 11 μM that is associated with the loss of the three N-terminal amino acid residues (Table 1). At peptide concentrations below 11 μM, P^BLV-ΔCCF is soluble under the conditions used in the syncytium interference assays and yet fails to inhibit membrane fusion (Figure 4B). It should be noted that disulphide formation between the peptide and envelope is not required for inhibitory activity, as reduction of P^BLV-391 and subsequent modification of the cysteine residues with the sulfhydryl reactive agent Iodoacetyl-PEO₂-Biotin failed to disrupt the inhibitory properties of the peptide (Figure 4C). Moreover, the activity of the biotinylated peptide was indistinguishable from that of the unmodified P^BLV-391, indicating that potential dimerization of the peptide through inter-molecular disulphide bonding does not influence peptide potency (Figure 4C). The first 3 amino acids of the BLV peptide, which includes the two cysteine residues and an adjacent phenylalanine, are conserved between HTLV-1 and BLV. Given the data obtained for the BLV peptide it is surprising to note that substitution of the cysteines with alanine did not affect the activity of the HTLV-1 inhibitor P^cr-400 22. Thus it seems that, at least for the BLV peptide, the first 3 amino acids aid peptide solubility and contribute in an important but, as yet, ill-defined way to the binding or orientation of the peptide within the target-binding site on TM.

Leucine residues in P^cr-400 play a key functional role in peptide activity 20. The crystal structure of the HTLV-1 TM 16 reveals that within the LHR several leucine and isoleucine residues reach down into deep pockets within the groove of the coiled coil. It appears that the LHR-derived peptide P^cr-400 makes similar contacts with the coiled coil and that these contacts are necessary for stable binding of the peptide to the coiled coil and thus are critical to the inhibitory activity of the peptide 22. Intriguingly, some but not all of these leucine and isoleucine residues are conserved between the LHRs of HTLV-1 and BLV. We therefore sought to determine the importance of these conserved residues to the inhibitory properties of the BLV LHR-mimetic peptide. Two peptides were synthesized, P^BLV-L/A in which all leucines were substituted with alanine, and P^BLV-L404/410A in which the Leu404 and Leu410 of BLV envelope were replaced by alanine (Table 1) these particular leucines are equivalent to the well-conserved Leu413 and Leu419 of HTLV-1 isolates. Syncytium interference assays revealed that compared to the parental peptide (P^BLV-391) the alanine-substituted peptides were severely compromised in their ability to inhibit membrane fusion (Figure 4D); in particular, P^BLV-L/A did not exhibit any discernible inhibition up to 20 μM (Figure 4D) or above (data not shown). Hence, the leucine residues are important to peptide function. Moreover, although P^BLV-L404/410A was just as soluble as the parental peptide (Table 1), P^BLV-L404/410A also failed to display any fusion-blocking activity up to 20 μM (Figure 4D); indicating that the leucines equivalent to BLV envelope residues 404 and 410 are particularly important to the inhibitory properties of the LHR-mimetic peptide.

Our analysis reveals that for the ectodomain of the TM the majority of the amino acid residues that are conserved between HTLV-1 and BLV map to the interacting surfaces of the trimer-of-hairpins. Moreover, a BLV homologue of the HTLV-1 LHR-derived peptide inhibitor also exhibits robust but highly specific inhibitory activity against BLV-induced membrane fusion. Significantly, conserved leucine residues are critical to the inhibitory activity of both peptides. Encouraged by these results and to gain greater insight into the mechanism of fusion and the likely contacts made by P^BLV-391 with the coiled coil, we constructed a homology model of the BLV trimer-of-hairpins that is based on the crystal structure of the HTLV-1 TM (Figure 1B) 16.

Having identified the predicted BLV coiled-coil (Figure 1A), the Clustal-W alignment of the TM ectodomain sequences of BLV and HTLV-1 (Figure 1A) permitted the substitution of the BLV residues onto the HTLV-1-derived scaffold, consisting of the complete trimer of N-helices and a single LHR. The geometry of the crude model was improved by simulated annealing and energy minimisation in explicit solvent with the GROMACS (Groningen Machine for Chemical Simulations) package using the GROMOS96 43a1 force field 34. It should be noted that, compared to the HTLV-1 trimer of hairpins, there are two additional residues in the predicted BLV chain-reversal region at positions 380 and 381 of BLV envelope. Since these residues are within a flexible loop there is insufficient information to model these residues with any degree of accuracy therefore these residues are omitted in the current model. Nonetheless, the restraint provided by the disulphide bond between Cys384 and Cys391 coupled with a high level of sequence conservation within the heptad repeat region and within the LHR suggests that the model is likely to be a reasonably accurate representation of the interaction between the LHR and the coiled coil. The model for the BLV coiled coil and LHR is presented in Figure 5A.

Consistent with the sequence alignment and the structure of the HTLV-1 TM ectodomain (Figure 1), the BLV TM model indicates that Leu394 and Ile396 likely project into a hydrophobic pocket at the membrane-distal end of the core coiled-coil (Figure 5B). It also implies that Ile401, Leu404 and Leu407, which all lie on the same side of the putative α-helix of the LHR, are oriented such that they project into the groove of the coiled coil. Notably, Leu410 is predicted to make a significant contact with a deep pocket situated towards the membrane-proximal end of the core coiled-coil. Therefore, the BLV coiled coil and LHR model is highly consistent with the experimental data and provides a molecular explanation for the loss of activity associated with substitutions in the BLV LHR-derived peptide.

The accumulated experimental data correlate well with the structural model, implying that predications based on the BLV trimer-of-hairpins model are likely to be informative. The homology model of the BLV TM ectodomain (Figure 6) suggests that Arg403, a residue within the predicted α-helix of the LHR and mimicked by P^BLV-391 peptide, may be electrostatically unfavourable for efficient binding of the C-terminal LHR into the groove of the core coiled-coil. We predicted that removing this unfavourable charge interaction would improve the binding of the peptide to the BLV coiled coil and thereby improve the inhibitory activity of the peptide. We therefore synthesized a peptide, P^BLV-R403A, which incorporated an alanine residue in place of the arginine equivalent to Arg403 of Env (Table 1). As anticipated, substitution of the arginine residue resulted in a modest but highly consistent and significant (p < 0.0001, Student's t-test) improvement in peptide potency when compared to P^BLV-391. The peptide P^BLV-R403A is more than twice as potent as P^BLV-391 in syncytium interference assays, with a calculated IC₅₀ of 1.56 ± 0.05 μM compared to 3.49 μM ± 0.03 μM for P^BLV-391 (Figure 6). The data show that a single amino-acid substitution in the predicted short α-helix of the LHR-mimetic peptide increases the ability of the peptide to block membrane fusion and provides further support for the utility of the model of the BLV TM core.