We report here that the partial knockdown of IRSp53 expression reduces HIV-1 particle release in host Jurkat T cells and in the model cell line HEK293T (Figure 1a,b), similar to our previously reported results in primary T lymphocytes (Thomas et al., 2015). Cells were treated with siRNA targeting IRSp53 or IRTKS (validated by extinction of the transfected ectopic IRSp53-GFP or IRTKS-GFP proteins – Figure 1—figure supplement 1b and c, respectively). In Jurkat T cells, we expressed the viral Gag proteins in the context of HIV-1(ΔEnv) in order to only monitor the late steps of the viral life cycle. Partial IRSp53 gene expression knockdown (resulting in a maximum of 50% protein depletion) reduced particle release threefold as compared to the control siRNA (Figure 1a, bottom), and a sixfold reduction was determined when taking into account the percentage of protein depletion. This reduction in HIV-1 particle release is highly significant (n = 3 independent experiments, p value = 0.00265, Student’s t-test, and ANOVA statistical test for multiple comparisons indicate p value = 0.0089) since the gene editing of IRSp53 cannot be complete, nor edited by CRISPR/Cas9 knockout, without being toxic for the cells. To compare the role of different I-BAR domain proteins from the same family, we also measured the effect of siRNA targeting IRSp53 and IRTKS (Figure 1b) on HIV-1 Gag VLP production in HEK293T cells (Figure 1—figure supplement 1a, see graph Figure 1b, and immunoblots Figure 1—figure supplement 1d,e). IRTKS shares similar protein domain organization and high sequence homology with IRSp53 (40% amino acid sequence identity and 59% sequence similarity, Figure 1—figure supplement2), and displays some functional redundancy with IRSp53 (Chou et al., 2017). IRTKS can also induce plasma membrane curvature (Saarikangas et al., 2009). Partial knockdown of IRSp53 (~50% protein depletion) resulted in a two- to threefold decrease in HIV-1 Gag particle production (n = 3 independent experiments, p value = 0.000487, Student’s t-test) (Figure 1b, bottom, Figure 1—figure supplement 1d). In contrast, knockdown of IRTKS (Figure 1b, Figure 1—figure supplement 1e) did not have any significant effect on HIV Gag particle release (n = 3 independent experiments, p value = 0.0924, Student’s t-test), thus precluding the possibility of redundant functions between IRSp53 and IRTKS in the context of HIV-1 Gag particle formation.

Electron microscopy imaging of siRNA IRSp53 treated HEK293T cells expressing HIV-1 Gag revealed arrested particle budding at the cell plasma membrane (Figure 1c, lower panel), as compared to the siRNA-control cells (Figure 1c, upper panel). While the control cells exhibited the normal phenotype of Gag-VLP budding from the cell plasma membrane, the IRSp53 knockdown cells displayed a series of viral buds arrested in assembly decorating the cell plasma membrane (Figure 1c, Figure 1—figure supplement 3). These results revealed an arrest in Gag assembly at the membrane and thus the involvement of IRSp53 in the assembly process. Since IRSp53 is an I-BAR protein involved in cell membrane curvature, we measured the curvature exhibited by HIV-1 buds in IRSp53 knockdown cells. While control cells displayed a range of HIV-1 Gag particles at different stages of assembly and budding, the cells knocked down for IRSp53 instead displayed arrested buds at an early assembly stage (Figure 1d). By measuring the dimensions of these arrested buds, we found that buds from cells knocked down for IRSp53 displayed a narrower range of curvature height (48 ± 22 nm), as compared to the control (85 ± 53 nm) (n = 145 buds from 14 different cells, p value = 1.053 × 10⁻²⁸, Kolmogorov-Smirnov test), while the bud widths presented no difference between siIRSp53 (135 ± 64 nm) and the control (140 ± 87 nm) (n = 145 buds from 14 different cells, p value = 0.0609, Kolmogorov-Smirnov test) (Figure 1e). The control cells thus exhibited a range of heights and widths consistent with the range of buds seen at the membrane of these cells. The result indicates that in the absence of IRSp53, the viral buds were unable to progress beyond a certain curvature.

Since both Gag and IRSp53 target the cell plasma membrane upon interaction with PI(4,5)P₂ (Favard et al., 2019; Mattila et al., 2007; Prévost et al., 2015; Saarikangas et al., 2009; Sengupta et al., 2019; Takemura et al., 2017; Yandrapalli et al., 2016), we then tested if Gag and IRSp53 could associate directly or indirectly using immuno-precipitation (IP) assays (Figure 2a,b). Our results showed that IP of endogenous IRSp53 resulted in co-precipitation of Gag (Figure 2a, lane 1), as compared to the controls (lanes 2–4). Unfortunately, we could not assess the amount of IRSp53 pulled down by the antibody between conditions because the IgG signal masked the endogenous IRSp53 signal. To overcome this issue, we performed an IP/co-IP experiment of ectopic IRSp53-GFP/HIV-1 Gag proteins with an anti-GFP antibody, and confirmed the pull-down of IRSp53-GFP (Figure 2b, lanes 2 and 3) and the co-precipitation of Gag (lane 3) while nothing was detected in the controls (lanes 1 and 4). Input and flowthrough were in accordance with the results, showing less IRSp53-GFP in the flowthrough (lane 3). We then concluded that HIV-1 Gag and endogenous IRSp53, or ectopic IRSp53-GFP, were components of the same intracellular complex, interacting directly or indirectly through other factors or membrane domains. IRSp53 is a cellular protein that switches from the cytosol to the cell plasma membrane for inducing membrane ruffles upon activation by Rac1 and its effectors (Miki and Takenawa, 2002; Suetsugu et al., 2006). We have previously shown that Gag cellular expression triggers Rac1 activation (Thomas et al., 2015), on which IRSp53 membrane localization and function are dependent. Thus, here, we compared the relative membrane binding of IRSp53 upon cellular expression of HIV-1 Gag using membrane flotation assays (Figure 2c,d). In the absence of Gag (‘HEK293T control' cells), one could observe the presence of IRSp53 both in the cytosol (Figure 2c, fractions 6–8, labeled with the ribosomal S6 biomarker) and in the cell membranes fractions (Figure 2c, fractions 1–3, labeled with the membrane Lamp2 biomarker). The lysosomal membrane protein Lamp2 and the cytosolic ribosomal protein S6 were used as controls to validate the correct separation of the membrane and cytosolic fractions. Thus, at equilibrium, 19 ± 8% of IRSp53 was bound to the cell membranes (Graph, Figure 2d). The same experiment was repeated with cells expressing HIV-1 Gag, where 66 ± 9% of Gag was bound to the cell membranes (Graph, Figure 2d). Notably, we observed a twofold increase with 44 ± 5% of IRSp53 bound to the cell membranes upon HIV-1 Gag expression (Graph Figure 2d) (n = 5 independent experiments, p value = 0.000129, Student’s t-test), while none with IRTKS. This effect was comparable with the one of Tsg101, a protein of the ESCRT-I complex known to interact mainly with the p6 domain of Gag (Garrus et al., 2001; Pornillos et al., 2003; von Schwedler et al., 2003) and partially with the NC domain (El Meshri et al., 2018). The cellular endosomal sorting complex required for transport (ESCRT) machinery has been involved in the mechanism of vesicular budding of intracellular multi-vesicular bodies and also hijacked by the HIV-1 Gag protein for viral particle budding. Here, we observe a twofold increase in cell membrane binding of Tsg101 upon Gag expression, passing from 36 ± 10% without Gag to 79 ± 8% in the presence of Gag (Figure 2c) (n = 5 independent experiments, p value = 0.00517, Student’s t-test).

Furthermore, we examined if Gag/IRSp53 complexation was dependent on the p6 domain of Gag to reveal if this could be independent of ESCRT recruitment by Gag. We, thus, used a C-terminal mutant of Gag, GagΔp6, which is deficient in ESCRT-Tsg101 recruitment (von Schwedler et al., 2003), but is still capable of binding the plasma membrane and assembling particles that poorly bud. GagΔp6 viral particles are tethered and remain attached to the plasma membrane (see Floderer et al., 2018 for the characterization of Gag(i)mEos2Δp6). Our experiments revealed that Gag, Gag(i)mEos2, and GagΔp6(i)mEos2 were all pulled down with IRSp53 (Figure 2—figure supplement 1), showing that the addition of the internal mEos2 protein - a necessary tag for super resolution microscopy (SRM) imaging of Gag (see the following section) - did not affect the complexation of Gag with IRSp53. This was to show that the Gag used in the SRM imaging studies behaved similarly to wild-type Gag. Moreover, we showed that the p6 domain was not required for Gag/IRSp53 molecular interplay suggesting that it occurs before ESCRT recruitment.

Taken together, these results suggest that there is a complexation between HIV-1 Gag and IRSp53, reinforcing the idea of a strong molecular interplay between these two proteins directly or indirectly but in the same membrane domain. We observed that cellular Gag expression, possibly by triggering Rac1 activation (Thomas et al., 2015), favors cell membrane binding of IRSp53.

Our finding that IRSp53 and HIV-1 Gag are present in the same molecular complex at the cell membrane motivated us to assess whether IRSp53 was present specifically at the Gag assembly sites. Because HIV-1 assembly is ~100 nm in diameter (Floderer et al., 2018; Manley et al., 2008), we used PALM (Photo-Activated Localization Microscopy) coupled to dSTORM (direct Stochastic Optical Reconstruction Microscopy) with TIRF illumination, to investigate with high precision the localization of I-BAR proteins in Gag(i)mEos2 assembly sites at the plasma membrane (Figure 3, Figure 3—figure supplement 1, Figure 3—figure supplement 2).

We first checked our ability to correctly identify ongoing assembly sites by quantifying the sizes of Gag clusters. For this, we used purified HIV-1 Gag(i)mEos2 VLPs (Figure 3—figure supplement 3b). When using the same size estimation method described in Floderer et al., 2018, we measured individual VLP sizes around 130–140 nm (Figure 3—figure supplement 3b). VLP size was correctly estimated (114 ± 37 nm, median ± 1st IQR) from the size distribution of 800 different clusters identified in our VLP images using a density-based spatial scan (DBSCAN) clustering method (Figure 3—figure supplement 3b). Finally, DBSCAN showed us that these HIV-1 Gag(i)mEos2 cluster size distributions had similar median values for each cell type in each condition (HEK 293T (96 ± 44 nm, Figure 3—figure supplement 3c) or Jurkat T cells (105 ± 66 nm, Figure 3—figure supplement 3e) when IRSp53 was immunolabelled, HEK 293T (116 ± 68 nm, Figure 3—figure supplement 3d) or Jurkat T cells (96 ± 50 nm, Figure 3—figure supplement 3f) when IRTKS was immunolabelled), allowing us to directly compare IRSp53 and IRTKS organization close to these assembly sites.

In HEK293T and Jurkat T cells, reconstructed dual-color PALM/STORM images exhibited Gag(i)mEos2 assembly sites close to or overlapping with IRSp53 (Figures 3a,e, Figure 3—figure supplement 1a, Figure 3—figure supplement 2a) whereas Gag clusters did not seem to overlap with IRTKS (Figures 3b,f, Figure 3—figure supplement 1b, Figure 3—figure supplement 2b). These results are consistent with the siRNA data presented in Figure 1b. In order to quantify these observations, we performed coordinate-based colocalization (Malkusch et al., 2012) (CBC) analysis of HIV-1 Gag and IRSp53 (or IRTKS) in close proximity of HIV-1 assembly sites. We first isolated the assembly sites by segmentation and retrieved their center positions. We then kept all HIV-1 Gag located within a distance of 80 nm from this center (70–80% of all cluster sizes measured by DBSCAN are found within this distance) and all the IRSp53 (or IRTKS) found in a distance of 150 nm from this center (~twofold the assembly site size, see Figure 3—figure supplement 4 for details on the process workflow). We chose this IRSp53 (or IRTKS) cutoff distance to minimize the contribution of cross-colocalization between different HIV-1 Gag clusters. In contrast to classical colocalization analysis, CBC takes into account the spatial distribution of biomolecules to avoid excessive colocalization due to local densities of single molecules and provides a colocalization value for each single-molecule localization. This CBC value ranges from −1 to +1, where −1 corresponds to anti-correlation, 0 indicating non-correlation, and +1 corresponds to perfect correlation between the two molecules. Since CBC values are calculated for each localization, we plotted the CBC values as frequency distributions for all localizations of IRSp53/Gag and IRTKS/Gag. As shown in Figure 3c (for n = 4 HEK293T cells) or in Figure 3g (for n = 5 Jurkat T cells), the CBC distribution for IRSp53/Gag has a higher proportion of values exhibiting high colocalization (27% in HEK293T cells or 26% in Jurkat T cells of CBC > 0.5) in comparison to IRTKS/Gag values (14% for both cell types, n = 5 for HEK293T and Jurkat T cells of CBC > 0.5) which instead show a very high proportion (close to 75%) of anti-correlation (CBC < 0) (Figure 3d,h). This comparison directly shows that IRSp53 displays stronger single-molecule colocalization with Gag in assembling clusters than IRTKS does.

Although CBC values give a quantitative value of the colocalization, it does not provide direct information on the average positions of IRTKS or IRSp53 molecules with respect to Gag molecules within the assembling clusters. Moreover, as for all the colocalization methods used to analyze molecule proximities, the CBC method is a strongly parameterize method, the results of which can depend on the parameter values (total distance of search for colocalization, number of searching circles…). Thus, to gain more insight into these colocalization quantification, we performed simulations to generate different patterns of PALM/STORM localizations (Figure 4), and analyzed them with the same set of parameters (total distance and distance steps) that the one we used for the experimental data. Since IRSp53 and IRTKS are membrane curvature sensors and since we observed that a lack of IRSp53 stops the assembling bud at half completion (Figure 1e), we hypothesis that they will be located around the ongoing assembly site. We therefore simulated an annulus of IRSp53 or IRTKS STORM localizations with different waists located at increasing distances from the center of Gag clusters (Figure 4a and Figure 4—figure supplement 1). Numerically generated images were then analyzed identically to our experimental images (see 'Materials and methods' for details) and CBC distributions were generated (Figure 4b). Experimental and simulated cumulative distribution functions of the CBC values were compared by performing a root mean square error (RMSE) quantification (Figure 4—figure supplement 1, g,h,i,j). The lowest RMSE value was considered as the best similarity between the two distributions. This comparison indicated that IRSp53 localization, on average, displays a restricted pattern around and in the assembly sites. This corresponds to a circular ring surrounding the assembly site at 80 nm from the center of the Gag budding sites with a width of 80 nm (Figure 4c,e). On the other hand, IRTKS was present as a large diffuse pattern located at 140 nm from the Gag assembly site center with a width of 200 nm, explaining why fewer IRTKS molecules were detected in the assembly sites (Figure 4d,f). Interestingly, the maps of the RMSE values show that, on average, the surrounding belt center (independently of its waist) is located nearer to the HIV-1 Gag assembly center in the case of IRSp53 as compared to that of IRTKS (Figure 4—figure supplement 1g,h,i,j). Our results thus show that IRSp53 indeed specifically localizes at HIV-1 Gag assembly sites at the cell plasma membrane, whereas IRTKS poorly does.

The involvement of IRSp53 around Gag assembly sites seems to be conserved regardless of the cell type, reinforcing the idea of a specific role for IRSp53 in HIV-1 Gag particle assembly.

To assess IRSp53 incorporation into HIV-1 Gag particles, we purified Gag-VLPs from cells transfected with Gag-mCherry and several GFP-tagged I-BAR domain proteins (Figure 5a). The IRSp53-I-BAR-GFP construct only contains the membrane curving I-BAR domain of IRSp53. PH-PLCδ-GFP, a PI(4,5)P₂ binding protein, was used as a control, because it binds PI(4,5)P₂ but does not generate membrane curvature. Fluorescent VLPs were purified from these transfected cells, then visualized for two colors (green: GFP and red: mCherry), and then a Mander’s coefficient was calculated as an indicator of incorporation of the ectopic (green) GFP-tagged proteins within the (red) Gag-mCherry VLPs and vice-versa (see Materials and Methods) (Figure 5a,b). We found a high correlation (Mander’s coefficient = 0.95–1) between IRSp53-GFP and Gag-mCherry (Figure 5b, left graph, red column), indicating that almost all Gag-mCherry VLPs contained IRSp53-GFP, while the reverse (green column) was not true, indicating that all the IRSp53-GFP-labelled vesicles produced by the cells were not all positive for Gag. When using the IRSp53-I-BAR domain alone, we also obtained a high Mander’s coefficient, that is, ~0.8 (Figure 5b). In contrast, for IRTKS-GFP, the Mander’s coefficient was 0.4–0.5, indicating no significant correlation between IRTKS-GFP and Gag-mCherry. This indicates that even if a domain (PH-PLCδ) or another I-BAR protein (IRTKS) can recognize PI(4,5)P₂ at the cell membrane, it is not sufficient for incorporation into Gag-VLPs. Finally, we found that the reverse correlation of GFP-tagged proteins with Gag-mCherry was under the random 0.5 coefficient (Figure 5b, right graph, green column), indicating that, on average, GFP-tagged vesicles produced by the cells did not contain Gag. Taken together, these results show the preferential incorporation of IRSp53-GFP into released HIV-1 Gag-mCherry VLPs.

To study the incorporation of endogenous IRSp53 in HIV-1 particles, cells were transfected with plasmids expressing either wild-type infectious HIV-1 or codon-optimized immature HIV-1 Gag protein (without genomic RNA). The virus particles were purified through a 20%-sucrose cushion or further through a continuous iodixanol density gradient (as in Grigorov et al., 2009 and Grigorov et al., 2006). IRSp53 was found to be associated with the viral particles in both conditions, that is, in infectious HIV-1 and in Gag VLPs, indicating that Gag alone is sufficient to recruit IRSp53 in the viral particles (Figure 5c,d, ‘IRSp53’). Tsg101 also showed an association with viral particles in both conditions (Figure 5c,d, ‘Tsg101’), as reported previously (Garrus et al., 2001; Hammarstedt and Garoff, 2004; Pornillos et al., 2003). In contrast, IRTKS was not associated neither with Gag-VLP nor HIV-1 particles (Figure 5c,d, ‘IRTKS’). Upon further purification (Figure 5e), IRSp53, and the ESCRT proteins, Tsg101 and ALIX, were found to be associated within the same fractions containing the HIV-1 Gag viral particles, together with other well-known viral particle cofactors such as CD81, CD63 tetraspanins (Grigorov et al., 2009; Grigorov et al., 2006). Thus, endogenous IRSp53 is most probably incorporated in HIV-1 particles in a Gag-dependent manner.

The results above demonstrate that not only is IRSp53 incorporated in Gag-VLPs, but it is present at the budding sites and its deletion strongly reduces HIV-1 particle release in a Gag-dependent manner by arresting its bud assembly at the cell plasma membrane (Figure 1). In order to advance molecular mechanistic understanding of the role of IRSp53 locally at Gag assembly sites, we assessed IRSp53 I-BAR/Gag interplay on model membranes giant unilamellar vesicles (GUVs). Previous in vitro studies showed that when placing IRSp53 I-BAR domain outside PI(4,5)P₂-containing GUVs, the I-BAR domain can deform GUV membranes, generating tubes toward the vesicle interior (Jarin et al., 2019; Prévost et al., 2015; Saarikangas et al., 2009; Jarin et al., 2021). Consistent with the previous observations, we observed I-BAR driven tubulations on GUVs at low I-BAR concentrations (0.005–0.06 μM, see Figure 6, Figure 6—figure supplement 1c as an example); surprisingly, when increasing I-BAR concentrations to 0.1 μM and up to 1 μM, we observed a decrease in the number of GUVs having tubes. Future work is required to investigate this seemingly puzzling observation.

For our Gag membrane binding assay, we first used a high concentration of IRSp53 I-BAR domain (0.5 µM), while keeping the PI(4,5)P₂ concentration constant. This was done in order to prevent the generation of tubes by IRSp53 I-BAR domain and to focus on analyzing the membrane-binding efficiency of Gag in the presence of IRSp53 I-BAR domain on flat, non-deformed GUVs. We found that Gag binding to GUV membranes is increased ~7-fold when IRSp53-I-BAR domain was introduced first on GUVs before adding Gag (median value 6.7), compared to the condition of Gag only (median value 0.9) (p <0.0001, Student’s t-test) (Figure 6a, ‘Gag only’ vs. ‘I-BAR + Gag’). However, in the condition where Gag was introduced before adding the I-BAR domain, Gag intensity on GUV membranes increased only ~3-fold as compared to the Gag only condition (p < 0.0001, Student’s t-test) (Figure 6a, ‘Gag only’ vs. ‘Gag + I-BAR’). Notably, by comparing I-BAR + Gag and Gag + I-BAR conditions, we observed a twofold higher Gag intensity on GUV membranes in the first condition (p < 0.0001, Student’s t-test) (Figure 6a, ‘I-BAR + Gag’ vs. ‘Gag + I-BAR’). Taken together, these results show that IRSp53 I-BAR domain facilitates Gag membrane binding on GUV in favor of increasing Gag concentration locally. Furthermore, given that in our cell experiments, we observed a relative increase of IRSp53 bound to the cell membranes upon HIV-1 Gag expression (Figure 2d), we also assessed if HIV-1 Gag could facilitate the membrane binding of IRSp53 I-BAR domain. Consistent with our cell experiment results, we observed a similar increase of the membrane binding (~1.5-fold) of IRSp53 I-BAR domain in the presence of HIV-1 Gag on GUVs (Figure 6—figure supplement 1a,b) suggesting a strong interplay between Gag and IRSp53 I-BAR domain.

Given these results, and that IRSp53 is a membrane curving protein involved in the early stages of cell protrusion generation (Disanza et al., 2013; Sathe et al., 2018), we asked whether the local membrane deformation induced by IRSp53 could be a preferred location for HIV-1 Gag assembly. We incubated GUVs with IRSp53 I-BAR domain at a low concentration (0.05 µM), which allows for the generation of inward membrane tubes, followed by the addition of Gag (Figure 6b, see Materials and Methods). This experiment revealed that Gag was sorted preferentially to the tips of the tubes generated by the IRSp53 I-BAR domain (Figure 6b,c,d, Video 1 and Figure 6—figure supplement 1c,d). We note that the majority of the tubes in GUVs were moving too rapidly, preventing us from acquiring images with high spatial resolution (see Videos 2 and 3). However, we observed that the Gag signals appeared dotty inside GUVs (Figure 6—figure supplement 2 and Videos 2 and 3), which is very different from the IRSp53 I-BAR domain signal that is clearly all along the tubes (Figure 6—figure supplement 1c). Moreover, for tubes that were not moving too fast, we found that for all the tubes (17 tubes, protein concentrations: 0.05 μM unlabeled I-BAR domain and 0.3 μM AX488 Gag), the Gag signal was exclusively located at the tips of the tubes (Figure 6d and Video 1). Finally, we observed that addition of HIV-1 Gag resulted in the formation of shorter I-BAR tubules as compared to GUVs incubated with the IRSp53-I-BAR domain alone (Figure 6—figure supplement 1c,d), indicating an interference in I-BAR tubule elongation when Gag sorted to the tubule tips, suggesting that Gag usurps the IRSp53 tubulation function.

Taken together, these results demonstrate that HIV-1 Gag binding to membranes is enhanced locally by the presence of IRSp53 I-BAR domain, and that Gag preferentially binds to highly curved membranes generated by the I-BAR domain of IRSp53.

A rabbit polyclonal anti-IRSp53 antibody (Merck Millipore), a rabbit polyclonal anti-IRTKS (Bethyl), a mouse monoclonal anti-CA (NIH AIDS Reagent Program), a rabbit polyclonal anti-GFP (Invitrogen), a mouse monoclonal anti-human CD63 (Santa Cruz Biotechnology), a mouse monoclonal anti-human CD81 (Santa Cruz Biotechnology), a rabbit monoclonal anti-human TSG101 (Abcam), and a rabbit polyclonal anti-human ALIX (Covalab) and secondary anti-rabbit Atto647N antibody (Sigma) were used in this study.

The plasmid expressing HIV-1 codon-optimized Gag (pGag(myc), named pGag), the plasmid expressing Pol and Env-deleted HIV-1 (named pNL4.3ΔPolΔEnv was a gift of E.Freed, HIVDRP, NIH, USA) encoding Gag alone with its packageable viral RNA (Chen et al., 2014b) and the plasmid expressing full wild-type HIV-1 (named pNL4.3) were described previously (Favard et al., 2019). Plasmids IRSp53-GFP, IRTKS-GFP, PinkBAR-GFP, and IRSp53-I-BAR-GFP were obtained from the University of Helsinki (Finland) (Saarikangas et al., 2009). Plasmids expressing PH-PLCδ-GFP was a gift of B.Beaumelle (IRIM, France), Gag(i)mCherry (named Gag-mCherry), Gag tagged with internal photo-activable mEos2 (named Gag(i)mEos2), p6-deleted Gag tagged with mEos2 (named pGagΔp6-mEos2) were described in Floderer et al., 2018.

Stealth siRNA (Invitrogen) targeting IRSp53 (BAIAP2) and IRTKS (BAIAP2L2), and Smartpools (Dharmacon) targeting IRSp53 (BAIAP2) or random sequence for siRNA controls were used in this study.

Human embryonic kidney cells (HEK 293T-ATCC-CRL-1575TM) were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM, GIBCO) and human Jurkat T lymphocytes (ATCC-CRL-2899TM) were maintained in RPMI (GIBCO). All cell lines were tested negative for mycoplasma thanks to a MycoAlert Mycoplasma Detection Kit (Lonza Bioscience) done every month. Media was supplemented with 10% fetal bovine serum (FBS, Dominique Dutscher) and complemented with sodium pyruvate and antibiotics (penicillin-streptomycin). Cells were grown at 37°C in a 5% CO₂ atmosphere. Transfection was performed by using the calcium phosphate precipitate method on HEK293T cells (as described in Grigorov et al., 2006) and the AMAXA (Lonza) method according to the manufacturer's instructions for the Jurkat T cells (as in Thomas et al., 2015). Based on different plasmid conditions, the cells were transfected (2 × 10⁶ cells/transfection) with a total of 8 µg of plasmids. The amount of transfected plasmid was normalized by adding pcDNA3.1 empty plasmid DNA. The cell medium was replaced by fresh medium 6 hr post-transfection and the experiments were performed 24–48 hr post-transfection. SiRNA transfections in cells were performed with either RNAiMax (Invitrogen) or calcium phosphate buffer in HEK293T cells or by electroporation for Jurkat T cells. One day prior to transfection, 2 × 10⁵ cells/well were seeded in 2 mL of growth medium without antibiotics in a six-well plate. Transfection was performed using the manufacturer’s protocol. After 24 hr of siRNA transfection, the cells were again transfected using the phosphate calcium buffer method. These cells were incubated at 37°C in a 5% CO₂ atmosphere for 24/48 hr.

Based on different plasmid conditions, HEK293T cells (2 × 10⁶ cells) were transfected with pGag alone or pGag/pIRSp53-GFP plasmids (8 µg total) and the amount of transfected plasmid was normalized by adding pcDNA3.1 ‘mock’ plasmid. The cell medium was replaced 6 hr post-transfection. After 24 hr post-transfection, the cells were washed with cold 1× phosphate buffer solution (PBS) prior to collection with 800 µL of chilled lysis buffer (50 mM TRIS-HCl [pH = 7.4]; 150 mM NaCl; 1 mM EDTA; 1 mM CaCl₂; 1 mM MgCl₂; 1% Triton, 0.5% sodium deoxycholate; protease inhibitor cocktail [Roche] one tablet/10 mL lysis buffer). The cells were incubated on ice for 30 min and then centrifuged at 13,000 rpm/15 min/4°C. The supernatant was collected in a new tube and the pellet was discarded. For each condition, 1000 µg of protein (the collected supernatant) was incubated with 1 µg of anti-IRSp53 or anti-GFP antibody on a tube rotator overnight at 4°C. About 25 µL of beads (Dynabeads Protein A, Life Technologies) was added to each tube of protein-antibody complex and incubated for 2 hr on the tube rotator at 4°C. The samples were then washed five times with the lysis buffer, followed by addition of 20 µL 2× Laemmli’s buffer to the beads. The samples were denatured at 95°C for 10 min and then processed for Western blot.

About 50 µg of each protein (intracellular in cell lysates) samples or 20 µL of purified VLP samples were mixed with SDS loading dye, deposited, and resolved on a 10% SDS-PAGE gel. The gels were then transferred on to polyvinylidene PVDF membranes (Amersham). Immunoblotting was performed by incubating the membranes overnight with primary antibody at 4°C, and 2 hr with horseradish-peroxidase (HRP)-conjugated secondary antibody at room temperature. The Western blot signals were detected using ECL Prime/ECL Select substrate (Amersham) and images were taken using ChemiDoc (Bio-Rad).

After 24 or 48 hr post-transfection, cell culture supernatants containing Gag-VLPs were collected, filtered through a 0.45 µm filter, and clarified at 800×g for 5 min at 4°C. The supernatant was then purified by loading it on a cushion of 25% sucrose in TNE buffer (25 mM Tris-HCl, 4 mM EDTA, and 150 mM NaCl) and ultracentrifuged at 100,000×g for 1 hr 30 min at 4°C in an SW41Ti rotor (Beckman Coulter). Dry pellets were resuspended in TNE buffer at 4°C overnight. Gag-VLP release was estimated by performing anti-CAp24 immunoblot and by quantifying Gag signal in the blots using Fiji software as described in Thomas et al., 2015. The calculation for Gag-VLP release is: % of Gag in VLP = Gag_released/(Gag_released + Gag_{intracellular normalized to GAPDH}).

For each condition, 4 × 10⁶ cells were transfected and viral supernatants were harvested 48 hr post-transfection, as described above. The cells were washed with ice-cold PBS and resuspended in Tris-HCl containing 4 mM EDTA and 1× complete protease inhibitor cocktail (Roche). Every step was then performed at 4°C. Cell suspensions were lysed using a Dounce homogenizer, then centrifuged at 600×g for 3 min to obtain Post-Nuclear Supernatants (PNS). A cushion of 820 μL of 75% (wt/vol) sucrose in TNE buffer was loaded at the bottom of an ultracentrifuge tube and mixed with 180 μL of PNS adjusted to 150 mM NaCl. About 2 mL and 300 μL of 50% (wt/ml) sucrose cushion followed by 0.9 mL of 10% (wt/ml) sucrose cushion were then layered to obtain the gradient that was then centrifuged in a Beckmann SW60Ti rotor at 35,000 rpm, 4°C, overnight. Eight fractions of 500 μL were collected from the top to the bottom and analyzed by Western blotting.

siRNA treated HEK293T cells were fixed in 4% paraformaldehyde and 1% glutaraldehyde in 0.1 M phosphate buffer (pH 7.2) for 48 hr, washed with PBS, post-fixed in 1% osmium tetroxide for 1 hr, and dehydrated in a graded series of ethanol solutions. Cell pellets were embedded in EPON resin (Sigma) that was allowed to polymerize at 60°C for 48 hr. Ultrathin sections were cut, stained with 5% uranyl acetate and 5% lead citrate, and deposited on collodion-coated EM grids for examination using a JEOL 1230 transmission electron microscope.

HEK293T cells expressing HIV-1 Gag/Gag(i)mEos2 cultured on poly-l-lysine (Sigma) coated 25 mm round #1.5 coverslips (VWR) were fixed using 4% PFA + 4% sucrose in PBS for 15 min at room temperature. Samples were subsequently quenched in 50 mM NH₄Cl for 5 min. Samples were then washed in PBS and then blocked for 15 min iat room temperature using 1% BSA in PBS and subsequently in 0.05% Saponin in 1% BSA in PBS. Samples were stained using a 1:100 dilution of the primary antibodies (rabbit polyclonal anti-human IRSp53, Sigma and rabbit polyclonal anti-human IRTKS antibody, Bethyl) for 60 min at room temperature. Samples were washed three times for 5 min using 1% BSA in PBS followed by 60 min staining using a 1:2000 dilution of the anti-rabbit Atto647N antibody (Sigma). Samples were washed three times for 5 min with PBS and stored in light protected container in +4°C until imaged. Samples were mounted on a StarFrost slide with a silicon joint with the STORM buffer (Abbelight). Cells were imaged within 60 min after application of the STORM buffer.

Single-molecule localization microscopy was performed on a Nikon inverted microscope equipped with 405, 488, 561, and 642 nm lasers, an EMCCD Evolve 512 Photometrics camera (512 × 512, 16 µm pixel size) with an oil immersion objective 100× NA1.49 Plan Apochromat. PALM imaging of Gag mEos2, activation was performed with laser irradiance set to 0.3 kW/cm² for 405 nm conversation and ~2.2 kW/cm² for 561 nm excitation. Illumination was performed over a 25 × 25 μm² area in the sample (1/e² spatial irradiance distance) in TIRF-mode. About 20–50,000 images were acquired for each cell with 50 ms integration time. The mean precision localization in PALM measurements was found to be 20 ± 5 nm (Figure 3—figure supplement 3a). 2D-STORM imaging of Atto647N was performed using a ~5 kW/cm² irradiance with the 642 nm excitation. About 25,000 images were acquired for each condition. Tetraspeck 100 nm multicolor beads (Life Technologies) as fiducial markers to correct for drift and chromatic abberation.

SMLM acquisitions were analyzed using the ThunderSTORM plugin in Fiji (Ovesný et al., 2014). The mean precision localization in PALM measurements was found to be 20 ± 5 nm (mean ± sd) and 27 ± 9 nm for STORM (Figure 3—figure supplement 3). During post-processing, a density filter was applied first to eliminate the background noise by identifying and discarding ‘isolated’ localizations, with a threshold of a minimum of five neighbors in a 50 nm radius for a molecule to be considered ‘not isolated.’ In the next step, molecules that converged to the same position were identified as duplicates and removed, keeping the localization with the smallest uncertainty as the valid coordinate. Furthermore, molecules reappearing with one ‘off-frame’ within 20 nm were merged into one single localization, with the first one appearing considered as the valid localization. Following these steps of post-processing, the sample drift was corrected by the drift correction module using fiducial markers described above. Each acquisition had at least two fiducial markers in the field of illumination. The resulting list of localizations was then used to reconstructing the respective PALM and STORM images and for the CBC analyses using the ThunderSTORM module. The module DBSCAN of the super-resolution quantification software SR Tesseler (Levet et al., 2015) was used to analyse the PALM localizations for quantification of Gag cluster sizes. In order to monitor the localization of I-BAR proteins in the vicinity of Gag assembling particles, the Gag particles were segmented by thresholding using Fiji, to generate a binary mask of the PALM images. The centers of each Gag assembling cluster were then determined and a custom MATLAB (Mathworks) code was used to extract localizations in a radius of 80 nm around each Gag cluster center and to extract I-BAR proteins localizations belonging to a disk of 150 nm radius around the center of each Gag clusters (see Results section). These subsets of coordinates were then used to calculate the experimental CBC, with the algorithm developed by Malkusch et al., 2012 and implemented in the ThunderSTORM plugin of Fiji (see Figure 3—figure supplement 4). The CBC values are calculated from single-molecule localization data of two species (Gag and IBAR proteins [IRSp53 or IRTKS]). A CBC value is assigned to each single localization of each species. We analyzed the distributions of these CBC values by plotting and comparing the distribution histograms of the CBCs obtained in the two conditions (IRSp53 vs. IRTKS).

We also performed a set of numerical simulations of super resolution images in Fiji. For this, we generated images by randomly choosing a molecule position within a radius of 60 nm from the cluster center for Gag [a compromise between the values found in the EM images (Figure 1e) and those obtained by DBSCAN analysis in the SMLM experimental images (Figure 3—figure supplement 3)]. For I-BAR proteins, we randomly assign positions within belts of different waists (from 0 to 100 nm) located at different distances (from 0 to 160 nm) from the Gag cluster center (see Figure 4 and Figure 4—figure supplement 1). Each localization was then randomly assigned a brightness such that the final pointing precision obtained at the end of the process for the Gag localization was equivalent to the experimental one. We then convolved each of these positions by a Gaussian 2D PSF, to generate a diffraction-limited image in which random noise was introduced using the ‘add noise’ function of Fiji. Super-resolution images of these simulated localizations were then reconstructed using the ThunderSTORM plugin of Fiji with the same parameters as the one used in the experimental data reconstruction. Finally, once the data set of localizations for both numerically simulated Gag clusters and associated surrounding I-BAR proteins were obtained, they were analyzed with the CBC function of ThunderSTORM with the same parameters as in the experimental analysis (analyzing all the neighbor positions within 10 successive radii of 20 nm each to estimate the CBC). Each localization was attributed CBC values and we compared the experimental cumulative distribution functions of CBC values to the simulated cumulative distribution functions using RMSE measurements, in order to establish the most probable configuration of I-BAR belts around the ongoing Gag assembly site.

After 24 hr seeding, 2 × 10⁶ HEK293T cells were transfected with 8 µg of I-BAR-GFP expressing plasmid with or without 8 µg of pGag/pGag(i)mCherry plasmids (2/3 and 1/3, respectively). After 24 hr transfection, cell media (9 mL) was filtered before performing VLPs purification by ultracentrifugation in an SW41Ti rotor (Beckman Coulter) at 29,000 rpm, for 1 hr 30 min, at 4°C, on a 20% sucrose cushion in TNE buffer. Dry pellets were resuspended in 110 µL of TNE and allowed to sediment on round 25 mm coverslips for 45 min in an AttoFluor Cell Chamber (Invitrogen). VLPs were imaged with a Nikon Ti Eclipse 2 TIRF microscope. Images were taken with an Evolve EMCCD camera – 512 photometrics, using a NA = 1.45, 100× objective and using 488 and 561 nm lasers.

Images were acquired with a Zeiss LSM780 (for fixed cells) or a Nikon Eclipse Ti-2 in TIRF mode (for fluorescent viral particles). Colocalization analysis based on Mander’s coefficients was performed using JaCOP (Just another Colocalization Plugin) (Bolte and Cordelières, 2006). Mander’s coefficients are defined as M1=∑iAi,coloc∑iAi and M2=∑iBi,coloc∑iBi, A and B being the two respective channels (mCherry and GFP). 0 < M < 1, with 1 full colocalization and 0.5 random colocalization. The M1 and M2 coefficients were calculated for several images and then represented as column graphs with red columns representing the degree of overlap of mCherry images with GFP images, and green columns representing the inverse.

Cell culture medium of HEK293T (2.5 × 10⁶ cells plated) transfected with 8 µg of pNL4.3ΔpolΔenv plasmid was collected 48 hr after transfection and filtered using a 0.45 µm filter. The medium was then ultracentrifuged on a 20% sucrose cushion in TNE using an SW41Ti rotor (Beckman) at 40,000 rpm for 1 hr 30 min. A solution with 0.25 M sucrose, 1 mM EDTA, 10 mM Tris-HCl pH 7.4 was used to diluted the 60% iodixanol stock solution (OptiPrep from Sigma) and to prepare a 40% and 20% iodixanol solution. About 1.5 mL of each dilution (60%, 40%, and 20% iodixanol) was successively layered in an SW55Ti tube (Beckman) and the pellet of VLPs obtained after ultracentrifugation on a 20% sucrose cushion in TNE was loaded from the top. Tubes were ultracentrifuged at 50,000 rpm in an SW55Ti rotor (Beckman) at 4°C for 3 hr. Then, 20 fractions of 200 µL were collected from the top of the tube to the bottom. About 20 µL of each fraction was loaded for Western blotting.

Brain total lipid extract (131101P) and brain L-α-phosphatidylinositol-4,5-bisphosphate (PIP₂, 840046P) were purchased from Avanti Polar Lipids/Interchim. BODIPY-TR-C5-ceramide, (BODIPY TR ceramide, D7540) and Alexa Fluor 488 C5-Maleimide (AX 488) were purchased from Invitrogen. β-casein from bovine milk (>98% pure, C6905) and other reagents were purchased from Sigma-Aldrich. Culture-Inserts 2 Well for self-insertion were purchased from Ibidi (Silicon open chambers, 80209).

Recombinant mouse IRSp53 I-BAR domain was purified and labeled with AX488 dyes, as previously described (Prévost et al., 2015; Saarikangas et al., 2009). Recombinant HIV-1 immature Gag protein was purified by J. Mak as described in Yandrapalli et al., 2016 and labeled with Alexa488 maleimide dyes (Invitrogen). Briefly, a 200 µM solution of the maleimide dye was incubated overnight at 4°C with a 20 µM solution of the Gag purified protein in a buffer of pH 8.0 with 1M NaCl and 50 mM Tris-HCl. Post incubation, the labeled mixture was subjected to dialysis with the Slide-A-Lyzer Mini Dialysis Device (Thermo Fisher Scientific), following the manufacturer’s instructions to remove the excess unbound dye from the solution.

Image analysis was performed by using Fiji (Schindelin et al., 2012).

All notched boxes show the median (central line), the 25th and 75th percentiles (the bottom and top edges of the box), the most extreme data points the algorithm considers to be not outliers (the whiskers), and the outliers (crosses).