Several studies concerning the in vivo interaction between EIAV p9 and ALIX were previously designed using the Y2H assay 1521. For comparison, we examined for the first time the HIV-1 p6-ALIX interaction using a similar approach. Close characterization of the ALIX-binding site in HIV-1 p6 was accomplished by systematically introducing alanine mutations at every amino acid residue contained within the p6 minimal region (aa: 31–46) that had been previously implicated in ALIX recognition 4. These Gal4 DBD-p6 bait constructs were individually co-transformed into the yeast strain AH 109 with a prey plasmid encoding the ALIX protein (868 amino acid-long) fused to the Gal4 AD. Relative quantification of the protein/protein interaction strength was monitored by measuring the β-galactosidase activity in yeast cells cotransformed with bait and prey expressing plasmids.

As shown in Figure 1A, the alanine scan clearly revealed that both amino acid residues in the YPx_n L consensus sequence as well as the leucine triplet repeat sequence (Lxx)₄ are crucial for HIV-1 p6 to interact with ALIX. This motif overlaps completely with the helix-2 in p6 as identified by NMR analysis 20, thus indicating that the ability of complex formation in vivo closely depends on the complete integrity of the secondary structure of helix-2. In details, the alanine substitution has variable effect from complete abolition of binding for Y36A and L38A, severe reduction in binding for E34A, L35A, P37A, L41A, R42A and a moderate but significant reduction for L44A, while the substitution of all other residues were well tolerated. Collectively, the binding data of our p6 mutants, are in full agreement with experimental data obtained in vitro with p6-derived peptides 18, except for the poor binding activity of L35A mutant, that has not been previously found in an in vitro binding assay measured by SPR 161718. However, the same authors reported that the corresponding L22A mutation in p9, completely abrogates p9 binding to ALIX. Thus, both L35 in p6 and L22 in p9 late domains are critical residues in the binding to ALIX.

In subsequent experiments, we tested in the Y2H system the potential interaction between the p6 domain and different ALIX mutant constructs. Side-by-side comparison was carried out in the presence of the EIAV p9 domain. Unexpectedly, a truncation of the proline-rich region in ALIX (ALIX_ΔPRR) from amino acids 716 to 868 impaired ALIX binding to p6 in vivo, while p9 still bound to the ALIX mutant (Fig. 1B). Because in vitro ALIX deleted from PRR has a lower affinity for p6 than for p9 (dissociation constants measured by SPR are 60 μM and 1.2 μM, respectively) 16, our data point out to a major role of PRR as positive regulator for the ALIX-p6 interaction.

The region 541 to 582 contains essentially helix-7 residues in the arm 1 of the V domain (nomenclature is from 17) and could play a key role in ALIX oligomerization. Indeed, bioinformatic analyses using the MultiCoil prediction program 22 suggested that this region in ALIX has a high probability for forming a trimeric coiled-coil (with a maximum trimeric residue probability value of 0.691 for S575). Y2H analysis of ALIX_Δ541–582 (Fig. 1B) provides evidence that helix 7 (and probably oligomerization of ALIX protein) is dispensable for interaction with p6 and/or p9 in vivo.

From structural studies, the ALIX viral late domain-binding site has been mapped to a large hydrophobic pocket on the long arm of the V domain. Mutational experiments targeting amino acid residues, which form the surrounding walls, revealed in particular that substitutions V509A in α5 and F676D in α11 caused a dramatic effect on the ability of the protein to bind a p6-derived peptide in vitro 17.

The effect of these two mutations on interaction with p6 and p9 was evaluated in vivo using the Y2H assay. As expected, both the V509A and F676D mutations prevented the yeast cell growth on selective medium when tested against the bait-p6 protein (Fig. 1B). Quite different results were obtained with p9, since the V509A mutation was well tolerated. Taken together, these results are consistent with a model in which the intact conformation of the binding site is required for the efficient interaction between ALIX and the helix-2 amino acid residues in the HIV-1 p6 late domain. In this regard, it has been postulated that p6 may bind coaxially to the V domain hydrophobic pocket and form a four-helix bundle together with ALIX α- 4, α- 5 and α- 11 17. The molecular mechanisms by which p9 binds to ALIX are likely involving a less stringent process in terms of structural requirement and integrity of the late domain binding site. Indeed, the short YPDL tetrapeptide motif detected in p9 constitutes a specific binding epitope for AIP1 family members throughout the eukaryotic evolution 23. Moreover, this motif appears very stringent since the close YLDL motif within the Sendai virus M protein, binds to the Bro1 domain of ALIX between amino acid residues 1–211, i.e. outside of the p9 binding domain 24.

Isothermal titration calorimetry experiments performed on the HIV-1 p6-derived peptide (DKELYPLTSLRSLFGN) and the EIAV p9-derived peptide (QTQNLYPDLSEIKKE) have reported that both peptides interacted in vitro with ALIX with quite similar K_d values 18, while full-length p6 displayed a much lower ALIX-binding affinity when compared to p9 16. A possible explanation for such divergent behaviour is that p6 could exhibit a constrained conformation for ALIX binding. Analysis of the high resolution structure of p6 20 (Fig. 2A) suggests that the hinge region (aa: 19–32) in the vicinity of the ALIX-binding site (helix α- 2) may play such a structural function.

Therefore a p6 mutant deleted for amino-acids S25 to Q28 (see location in Fig. 2A) referred as p6_ΔSQKQ was produced and was tested for interaction with ALIX by the Y2H assay. p9-ALIX, p6-ALIX, and p6_ΔSQKQ-ALIX gave rise to detectable growth on selective media when incubated for 3 days at 30°C, indicating that protein-protein interaction has occurred (Fig. 2B). The binding affinity quantified by in situ α-galactosidase staining using X-α-Gal as a substrate revealed a quite stronger interaction between p6_ΔSQKQ-ALIX as compared to p6-ALIX and/or p9-ALIX. When tested for binding to the truncated ALIX_ΔPRR, the p6_ΔSQKQ mutant supported significant growth on selective media. α-galactosidase staining was however reduced as compared to that observed in yeast co-expressing p9. Under similar conditions, the p6-ALIX_ΔPRR cotransfectants were found unable to grow as expected from data described in Figure 1. The absence of a significant growth on selective media of yeast co-expressing p9-Gal4AD, p6-Gal4AD, p6_ΔSQKQ-Gal4AD and ALIX-Gal4DBD ruled out the possibility that the different bait and prey proteins tested could directly activate the Gal4 responsive promoter and thus validated the specificity of the above described interactions.

To confirm these data, GST-pull down assays were then carried out with extracts from cells expressing either ALIX-HA or ALIX_ΔPRR-HA. This truncation was used because the removal of the proline-rich region has been described to improve the efficiency of in vitro interaction 4. After their expression in E. coli, the following fusion proteins, GST, GST-p6, GST-p9 and GST-p6_ΔSQKQ were bound to glutathione-Sepharose beads, and allowed to interact with either ALIX-HA or ALIX_ΔPRR-HA. After extensive washings, the complexes were eluted, subjected to electrophoresis under denaturing conditions, transferred to a PVDF membrane and reacted with an anti-HA monoclonal antibody. As shown in Figure 2C, the three GST constructs bound to both ALIX-HA and ALIX_ΔPRR-HA proteins in the following strength order: GST-p9>GST-p6_ΔSQKQ>GST-p6 while the control GST displayed no detectable binding activity.

The overexpression of wild type ALIX has been shown to partially rescue budding defects of HIV-1 particles with a p6 domain containing mutations in the PTAP motif (called PTAP/LIRL), i.e. unable to recruit the ESCRT I component Tgs101 1721. We therefore tested the ability of ALIX to alleviate the release defect of HIV-1 PTAP/LIRL mutated viruses containing or not the SQKQ deletion. We used a previously described complementation assay 621: HIV-1 proviral plasmid (NLδp6) that lacks the p6 domain was cotransfected into 293T with a plasmid expressing a truncated HIV-1 Gag protein (Gagδp6) fused to either the PTAP/LIRL p6 domain or the PTAP/LIRL p6_ΔSQKQ domain of HIV-1, together with an expression vector for ALIX or empty vector. As shown in Figure 2D, the coexpression of ALIX led to an increase in viral particle production and infectivity by both HIV-1 PTAP/LIRL p6 virus and HIV-1 PTAP/LIRL p6_ΔSQKQ. Similar effect of ALIX on PTAP/LIRL p6_ΔSQKQ was observed, although with reduced efficiencies. In summary, the deletion amino acid residues located in the p6 hinge region (ΔSQKQ) enhanced binding to ALIX, and partially allowed the rescue of HIV-1 PTAP/LIRL upon ALIX overexpression. This limited enhancing effect of this deletion on HIV-1 PTAP/LIRL p6 upon ALIX overexpression is indicative of a negative modulation played by the hinge region of HIV-1 p6. This negative modulation would be part of the highly complex process that optimises the HIV-1 budding.

To isolate a minimal region in ALIX that was still able to bind to p6_ΔSQKQ we used a previously described Y2H assay called Y2H-TPCR 25 (Fig. 3). Briefly, a library of random ~300 bp long PCR fragments derived from the ALIX cDNA was subcloned downstream to the Gal4 AD, and screened for potential interaction against the Gal4 DBD/p6_ΔSQKQ bait. After selection on selective SD/-Trp/-Leu/-His/-Ade medium, one ALIX fragment encompassing residues 391–510 was found to bind to p6_ΔSQKQ. This fragment also interacted with p9, but not with p6 or with the double mutant p6_ΔSQKQ Y36A unable to bind to ALIX as reported above, thus demonstrating that the interaction was specific (Fig. 3inset). It is worth noticing that ALIX_391–510 fragment partially rebuilds the arm 2 of the V-shape domain and encompasses the great majority of the hydrophobic surface residues which presumably contact the late domains 17. Remarkably, the minimal p6 and p9 binding site ALIX_391–510 that we identified are present in the truncated ALIX_409–71515, ALIX_364–716, and ALIX_1–503 21 fragments known to bind the YPDL motif.