CD36-specific antibodies block release of HIV-1 from infected primary macrophages and its transmission to T cells

To approach the potential involvement of host factors in the macrophage-specific location of HIV-1 assembly, we first analyzed, in HIV-1–infected macrophages at 7 d p.i., the intracellular distribution of three scavenger receptors and of the classical LDLR with respect to Gag localization by immunofluorescence (Fig. 1 A). We observed that staining specific for SR-BI, CD68, and LDLR was distinct from the one of Gag. In contrast, CD36 staining was not only present in membrane ruffles of the plasma membrane but also in internal compartments that stained positive for Gag (Fig. 1, A and B for quantifications) and for Env (not depicted), indicating that CD36 is associated with VCCs. Furthermore, analysis of infected macrophages at the ultrastructural level by immuno–electron microscopy (EM) revealed the presence of CD36 on Gag⁺ virions present in the lumen of VCCs (Fig. 1 C). We conclude that CD36 is associated with the VCCs and virions in HIV-1–infected macrophages.

The potential involvement of CD36 during the HIV-1 cycle in macrophages was evaluated by CD36 silencing. HIV-1–infected macrophages were transfected with siRNAs specific for CD36, luciferase as a negative control, and Tsg101 as a positive control (Fig. 1 D, scheme). CD36 silencing using three different siRNAs specific for CD36 led to a strong reduction (up to 65%, P ≤ 0.001 for the three siRNA on three independent experiments) of the amount of p24 Gag secreted in the supernatant (Fig. 1 E). The extent of the inhibition was similar to the one obtained upon silencing of Tsg101, an ESCRT protein crucial for viral budding (Garrus et al., 2001). Cell viability was similar in the different cell populations (Fig. 1 F). Efficiency of CD36 silencing ranged from 50 to 70% as determined by intracellular FACS staining (Fig. 1 G). Analysis of the kinetics of the inhibitory effect of CD36 silencing on p24 Gag release revealed that it was clearly visible at days 2–3 and 3–4 (65 and 48% reduction, respectively, for siRNA CD36#6; not depicted). In contrast, intracellular Gag levels measured at day 4 were not altered in any condition (not depicted). We conclude that CD36 expression is required for efficient release of HIV-1 particles by infected macrophages.

The presence of CD36 in the VCCs raised the question of the existence of similar compartments before HIV-1 infection. In uninfected macrophages, we observed that CD36⁺ compartments were CD9⁺ (Fig. 2 A) and Lamp1⁻ (not depicted), similar to VCCs in infected macrophages (Raposo et al., 2002; Pelchen-Matthews et al., 2003; Deneka et al., 2007; Jouve et al., 2007). To perform additional characterization of the CD36 compartments, we built a lentivector encoding CD36-GFP, a fusion protein known to be functional (Zhang and Crandall, 2007). CD36-GFP co-distributed with endogenous CD36 (not depicted) and in HIV-1–infected macrophages, with Gag (Fig. 2 B).

These results further suggested that CD36⁺ compartments in uninfected and infected macrophages were related. One of the major specific features of the VCC being its accessibility to external medium, we tested whether CD36⁺ VCC shared this property. Live macrophages co-infected with the lentivector CD36-GFP and HIV-1 Gag-iCherry ΔEnv were imaged before and immediately (in less than 10 s) after exposure to a fluorescent fluid phase marker, a 10-kD Dextran–Alexa Fluor 647 (Dex-647), as recently described (Gaudin et al., 2013). We observed that Dex-647 was able to access CD36⁺VCCs (see the large Gag⁺CD36⁺Dex-647⁺ compartment at the bottom of the cell in Fig. 2 C). Moreover, correlative EM performed on the very same cell established that this compartment contained both mature and immature viral particles (Fig. 2 D) and was therefore a bona fide VCC. Similar experiments performed on uninfected macrophages expressing CD36-GFP revealed that internal CD36⁺ compartments were also CD81⁺ (Fig. 2 E) and accessible to Dex-647 (Fig. 2 F). We conclude that CD36⁺ internal compartments in uninfected and infected macrophages share markers and possess plasma membrane connections allowing passive diffusion of small compounds.

These results prompted us to investigate whether HIV-1 could subvert preexisting CD36⁺ compartments and convert them into VCCs. Primary macrophages were infected first with the CD36-GFP lentivector together with simian immunodeficiency virus–like particles (SIV-VLPs) containing Vpx to increase the rate of infection (Goujon et al., 2008) and 4 d later with HIV-1 Gag-iCherry ΔEnv. Cells were imaged at day 0.5 after HIV addition by time-lapse epifluorescent microscopy at a rate of one image every 15 min for 2 d. At early time points, CD36-GFP was observed in cells that were devoid of Gag, and was present at the plasma membrane and in intracellular compartments (Fig. 3). With time, some macrophages exhibited a diffuse Gag-iCherry signal in the cytosol which then became concentrated in internal compartments, most of them being CD36-GFP⁺ (Fig. 3). We also unambiguously witnessed Gag recruitment and accumulation in preexisting CD36-GFP⁺ structures (Fig. 3 and Video 1). We conclude that Gag accumulates in preexisting CD36⁺ compartments in infected macrophages.

CD36 is a receptor located at the plasma membrane that is internalized in macrophages after ligand recognition (Zeng et al., 2003; Collins et al., 2009). We first established that CD36-specific antibodies when incubated for 2 h at 37°C with uninfected macrophages are able to reach internal CD81⁺ compartments (unpublished data). We therefore asked whether CD36 antibody is also internalized in infected macrophages and if it can reach VCCs. To address these questions, HIV-1–infected macrophages were exposed to a mAb specific for CD36 at 37°C for 2 h. After fixation and appropriate staining, cells were analyzed by confocal microscopy. The CD36-specific mAb was able to access Gag⁺CD9⁺ compartments, whereas an isotype control used in the same conditions could not (Fig. 4 A). Incubation at 4 instead of 37°C prevented the CD36 mAb to access VCCs (Fig. 4 A), and stained regions of the plasma membrane. Moreover, simultaneous exposure of HIV-1–infected macrophages to antibodies specific for SR-B1 and CD36 at 37°C for 2 h led to the specific accumulation of the CD36 mAb into Gag⁺ compartments (Fig. 4 B). SR-B1 antibodies were also internalized but in intracellular compartments clearly distinct from the Gag⁺ ones. Such transport did not occur at 4°C for both types of antibodies. These data show that anti-CD36 mAb can be actively and specifically transported into VCCs in HIV-1–infected macrophages.

The accessibility of the VCCs to exogenously added CD36-specific mAb prompted us to investigate the effect of such antibodies on HIV-1 release by infected macrophages (Fig. 5 A, experimental design). Exposure of HIV-1–infected macrophages to CD36-specific mAb for 24 h led to a strong reduction (80% mean, P ≤ 0.001 in three independent experiments) of the amounts of p24 Gag released in the supernatant as compared with macrophages exposed to an isotype control antibody (Fig. 5 B). Conversely, the amounts of p24 Gag detected in the corresponding cell lysates increased (40% mean, P ≤ 0.001 in three independent experiments) in anti-CD36–treated cells as compared with isotype-treated cells (Fig. 5 C). As a result, the sum of the amount of p24 Gag found in the supernatant and corresponding cell lysates in each of the conditions remained roughly constant (Fig. 5 D). These antibody exposures did not have any effect on the cell viability (Fig. 5 E). Two CD36-specific mAbs, FA6-152 and CLB-IVC7, induced a potent inhibition of p24 Gag release (Fig. 5 F), indicating that the effect observed was not a peculiarity of a single mAb. In contrast, SMΦ, an IgM mAb also specific for CD36, did not affect p24 Gag release from macrophages (Fig. 5 G). Of note, we verified that SMΦ antibody binds CD36 by flow cytometry (unpublished data).

Titration revealed that as little as 32 ng/ml CLB-IVC7 (a CD36-specific mAb) was sufficient to induce a 30% reduction of the p24 Gag released in the culture supernatant of infected macrophages (Fig. 5 H). This inhibition was accompanied by an increase in the amount of p24 Gag found in the corresponding cell lysates (Fig. 5 I). Despite the presence of CD36 on the virions (Chertova et al., 2006; Fig. 1 B), addition of the anti-CD36 mAbs at similar concentrations used in the assays above (2 µg/ml) did not affect the capacity of the virions to infect reporter TZM-bl cells (not depicted). The anti-CD36 treatment did not affect the infectivity of the few virions still released, as measured using normalized amounts of p24 Gag (Fig. 5 J). This indicates that the measurement of the p24 Gag level reflects the amount of infectious virions released.

The inhibitory effect of the anti-CD36 mAb on p24 Gag release was already detectable after 2 h of antibody exposure (Fig. 5 K), suggesting a rapid mechanism probably independent of protein synthesis. Furthermore, we observed that a 1-d anti-CD36 mAb exposure followed by a 1-d culture in mAb-free medium still strongly inhibited the p24 Gag release by macrophages during the subsequent 24 h (Fig. 5 L). This suggested that the effect induced was long lasting and confirmed that it did not result from the neutralization of the released virions but rather from an effect at the cellular level, possibly within the VCCs. We conclude that exposure of infected macrophages to soluble anti-CD36 antibodies are sufficient to strongly inhibit HIV-1 release.

VCCs as well as viral particles are enriched in many host proteins such as the CD81 tetraspanin and the ICAM-1 adhesion molecules, in addition to CD36 itself (Chertova et al., 2006). To evaluate whether the inhibition of virion release was restricted to CD36-specific antibodies, similar experiments were performed using other antibodies. Antibodies specific for CD81 and ICAM-1 accessed VCCs when added exogenously to infected macrophages at 37°C for 2 h (Fig. 6, A and B) but not at 4°C (not depicted). However, exposure to antibodies specific for CD81, ICAM-1, or CD9 did not modify the amounts of p24 Gag released by macrophages (Fig. 6, C–E), indicating that antibody access to the VCC was not sufficient to induce the retention of the virions. Similarly, exposure to antibodies specific for other lipid receptors, including scavenger receptors (CD68, SR-B1, or LDLR), did not modify the release of p24 Gag as compared with their respective isotype controls (Fig. 6 F).

Inhibition of HIV-1 release upon anti-CD36 mAb exposure remained unchanged in the presence of the FcR blocking reagent (Fig. 6 G), indicating that Fc receptors were not involved in the antibody-mediated effect. Exposure of HIV-1–infected macrophages to the Fab fragment prepared from the CD36-specific CLB-ICV7 mAb (Fig. 6 H, comparative titration of the mAb and its Fab) at three concentrations (Fig. 6 H, arrows) did not affect p24 Gag release, in contrast to the inhibition observed with the bivalent mAb (Fig. 6 I). These results establish that the inhibitory effect of the anti-CD36 mAbs on p24 Gag release by infected macrophages is highly specific and requires bivalency.

To evaluate the impact of the anti-CD36 mAb exposure on the HIV-1 cycle, we first looked at the maturation of the Gag precursor. The precursor p55 Gag is synthesized as a cytosolic protein. Once within the newly formed virions, p55 Gag is cleaved into four main polypeptides, including the Capsid p24 Gag. Exposure to the anti-CD36 mAb induced an intracellular accumulation of p24 Gag in infected macrophages as measured by immunoblot (Fig. 7 A). Quantification of the amounts of p55 and p24 Gag in cells from seven donors was performed from immunoblot analyses. Results expressed as the amounts of p24, p55, or total Gag (Fig. 7, B–D, respectively) normalized for the tubulin levels confirmed that anti-CD36 mAb exposure led to the accumulation of cell-associated p24 Gag that represents, in all likelihood, virions.

These results were confirmed by quantification of 3D reconstruction images of antibody-treated infected macrophages (Fig. 7 E, examples). Different parameters were extracted, including the sum of the Gag fluorescence intensity present in VCCs per cell (Fig. 7 F) and the total volume of the VCCs per cell (Fig. 7 G), which were both increased upon antibody exposure (statistically significant P ≤ 0.001).

EM analysis of anti-CD36– and isotype-treated infected macrophages revealed that both types of cells possessed VCCs containing mature and immature virions (Fig. 7 H). We also observed budding profiles at the limiting membrane of the VCCs under both conditions, suggesting that the antibody treatment did not affect the assembly process per se. Substantial amounts of anti-CD36 mAb were detected within VCCs of cells exposed to anti-CD36 mAb but not to the isotype control (Fig. 7 I). Interestingly, the staining appeared rather associated to the membranes of the virions, suggesting that the mAb may create a network through its binding to the CD36 present in the viral membranes. We conclude that the anti-CD36 mAb exposure induces the retention of virions within the VCCs of infected macrophages, increasing the volume of the VCCs and thus inhibiting the particle release to the extracellular medium.

Among the many ligands that CD36 can internalize in macrophages, MOxLDL has been widely studied (Endemann et al., 1993; Nozaki et al., 1995; Kunjathoor et al., 2002; Collins et al., 2009; Jaqaman et al., 2011). The effect of the anti-CD36 mAb could be related to the pathway induced by MOxLDL-CD36 interaction. To test this hypothesis, we first exposed HIV-1–infected macrophages to freshly thawed MOxLDL or to the anti-CD36 mAb or its isotype control for 10 or 30 min and measured the phosphorylation of JNK. As expected, MOxLDL induced via CD36 the expected JNK phosphorylation (Rahaman et al., 2006; Rios et al., 2012) after 10 min of exposure (Fig. 8 A), indicating that the MOxLDL used was biologically active. In contrast, anti-CD36 (or its isotype control) exposure had no effect on the levels of JNK phosphorylation, even after 30 min (Fig. 8 A), suggesting that conventional signaling through CD36 does not take place upon exposure to the CD36 mAb.

To evaluate the impact of MOxLDL exposure on the virus release, HIV-1–infected macrophages were exposed to biologically active MOxLDL together or not with the anti-CD36 mAb or its isotype control for 24 h. In all cases, the amount of p24 Gag measured in the supernatant was not modified by the presence of MOxLDL (Fig. 8 B) nor was the cell viability (Fig. 8 C).

We also tested whether the anti-CD36 mAb-mediated inhibition of virion release was dependent on the presence of LDLs in the culture medium. HIV-1–infected macrophages were starved for 4 h and then cultured for 24 h in medium complemented with lipoprotein-deficient serum, or without any fetal calf serum, and in the presence of the anti-CD36 mAb or its isotype control. In all cases the anti-CD36 mAb exposure led to a strong reduction in the amounts of p24 Gag secreted in the supernatant (Fig. 8 D). These data indicate that exposure to the anti-CD36 mAb does not induce a signaling pathway similar to the one induced by MOxLDL. Moreover, our data rule out the possibility that the anti-CD36 mAb effect on particle release is linked to MOxLDL interaction with CD36.

Besides MOxLDL, other endogenous ligands of CD36 include trombospondin 1 and type-I collagen (Asch et al., 1987; Tandon et al., 1989). We tested in a similar manner on infected macrophages whether overnight exposure to these two ligands exerts any effect on virion release. Although exposure to the anti-CD36 mAb induced a strong inhibition, none of the two ligands reduced the release of p24 Gag (Fig. 8 E). We checked that both ligands were biologically active on uninfected macrophages by analyzing the phosphorylation status of the JNK and Src kinases (Roberts et al., 2010; not depicted). Collectively, these data suggest that the anti-CD36 mAb effect on particle release is not related to the pathways induced upon CD36 binding to some of its endogenous multivalent ligands such as MOxLDL, trombospondin 1, or type-I collagen.

To get insight into the mechanism of action of the CD36 antibody, we needed to set up a more flexible cellular model than primary macrophages; thus, we derived HeLa cells expressing CD36. Stable expression of CD36 in HeLa cells did not modify the amounts of p24 Gag released upon HIV-1 infection (unpublished data). However, as observed with macrophages, exposure of HIV-1–infected HeLa-CD36 to CD36-specific mAb for 24 h led to a strong reduction (73% mean, P ≤ 0.001 on three independent experiments) of the amounts of p24 Gag released in the supernatant as compared with exposure to an isotype control antibody (Fig. 9 A). Conversely, and like in macrophages, the amounts of p24 Gag detected in the corresponding cell lysates increased (54% mean, P ≤ 0.001 on three independent experiments) in anti-CD36–treated cells as compared with isotype-treated cells (Fig. 9 B). In contrast, viral release from HeLa cells lacking CD36 remained unaffected by both antibodies (Fig. 9, A and B). Immunoblot analysis of the Gag contents of the cell lysates confirmed the results obtained by Elisa (unpublished data). We conclude that the inhibitory effect of the anti-CD36 mAb on particle release can be reproduced in infected HeLa-CD36 cells.

To approach the mechanism of action of the anti-CD36 mAb, we analyzed HIV-1– Gag–iGFP-infected HeLa-CD36 cells exposed to the mAb for 48 h by confocal microscopy and observed the presence of large clusters of Gag⁺CD9⁺ material at the plasma membrane of cells exposed to the anti-CD36 mAb that were absent in isotype-treated cells (Fig. 9 C). Time-lapse microscopy of HeLa-CD36 cells infected with HIV-1 Gag-iGFP after addition of the anti-CD36 mAb allowed us to witness the formation of these large clusters at the plasma membrane (Video 2). Ultrastructural analysis of HIV-1–infected HeLa-CD36 cells confirmed the presence at the plasma membrane of these large clusters of virions upon anti-CD36 exposure, but not in the isotype-treated control cells (Fig. 9 D). Collectively these results suggest that the anti-CD36 exposure induces virions tethering at the site of viral assembly, i.e., in the VCCs of macrophages and at the plasma membrane of HeLa-CD36 cells.

These results suggested that the anti-CD36 mAb effect did not rely on signal transduction via CD36. To further test the hypothesis, we took advantage of a previously published comparison of the cytoplasmic tail of CD36 with those of CD63, CD82, CD81, and CD9 (Primo et al., 2005). All these proteins are localized in the VCCs and present homologies (Fig. 9 E), as they possess a bulky/aromatic residue (position 4), a Cys (position 5), and a basic residue (position 9). This suggested that potential motifs of targeting to the VCCs may exist in the cytoplasmic tail of CD36 as well as other VCC proteins.

To identify key residues involved in the anti-CD36–mediated effect, we built a collection of CD36 mutants taking into account previous studies that have identified mutations altering several functions of CD36 (see Fig. 9 F and references in the legend). Each of the seven CD36 mutants, as well as the CD36 wild type, was cloned in a lentiviral vector that was used to establish stably expressing HeLa cells. Flow cytometry analysis revealed that all CD36 mutants were expressed at the cell surface, with the exception of the CD36^463* mutant (unpublished data). Importantly, all the CD36 mutants expressed at the cell surface were sensitive to the inhibitory effect of the anti-CD36 mAb, like wild-type CD36 (Fig. 9 G). This includes point mutants of the conserved cysteine residue (CD36^C464S) and a mutant carrying in its coding sequence a frame shift leading to substitution and premature stop codon (CD36^shift). These data suggest that the effect of the anti-CD36 mAb does not require the integrity of the CD36 cytoplasmic tail, and thus, in all likelihood, signal transduction through CD36 does not occur.

From a therapeutic perspective, we evaluated if our results could be extended to primary strains of HIV-1. Anti-CD36 mAb exposure of macrophages infected with the 132W primary isolate (Roesch et al., 2012) led to a strong reduction of the p24 Gag secreted in the supernatant (Fig. 10 A) and a concomitant increase of the cell-associated p24 Gag (Fig. 10 B) without affecting cell viability (Fig. 10 C). We conclude that anti-CD36 mAb exposure of macrophages infected with a primary HIV-1 isolate leads to the inhibition of viral secretion.

Cell-to-cell transmission is thought to be much more efficient than cell-free virus infection (Chen et al., 2007; Sourisseau et al., 2007). To test the effect of the anti-CD36 mAb on macrophage–to–T cell transmission, primary activated CD4⁺ T cells were added to HIV-1–infected macrophages together with the anti-CD36 mAb or its isotype control. After 24 h, AZT was added to stop new infections. The culture was left for an additional 24 h to allow Gag to accumulate into infected T cells, the proportion of which was measured by intracellular staining. As a positive control, we used anti–ICAM-1 mAb, which is known to inhibit cell-to-cell transfer of HIV-1 (Jolly et al., 2007; Wang et al., 2009). Importantly, the anti-CD36 mAb exposure like the anti–ICAM-1 mAb but not the isotype control led to a strong reduction of the fraction of T cells that became infected (Fig. 10 D). Therefore, exposure to anti-CD36 mAb can inhibit macrophage–to–T cell transmission of HIV-1.

Although still incomplete, current knowledge points to macrophages as key players in the course of HIV-1 infection and as a potential reservoir from which infection can rebound upon arrest of highly active antiretroviral therapy. It is thus crucial to decipher the life cycle of HIV-1 within macrophages, especially the late phases that remain poorly understood. In macrophages, HIV-1 assembly occurs in internal compartments, often referred to as VCCs, that probably originate from the plasma membrane (Deneka et al., 2007; Welsch et al., 2007). Our study reveals that HIV-1 hijacks preexisting CD36⁺ compartments for viral assembly and storage, and that CD36 expression is required for completion of the viral cycle. We also show that exposure to antibodies specific for CD36 induces the intracellular retention of the virions in infected macrophages and thus inhibits their release and transmission to T cells.

Our results showing a relationship between the VCC and preexisting compartments present in uninfected macrophages shed new light on the mysterious nature of the VCC (Tan and Sattentau, 2013). We show, both in uninfected and in HIV-infected macrophages, the presence of CD36⁺CD9⁺CD81⁺Lamp1⁻ compartments that can be connected to the surface. Furthermore, we show by time-lapse microscopy, for the first time to our knowledge, that internal CD36⁺ compartments preexist in macrophages and represent the site where newly synthesized cytosolic Gag accumulates, converting these compartments into VCCs with time. These data are consistent with previous studies (Deneka et al., 2007; Welsch et al., 2011; Chu et al., 2012a; Pelchen-Matthews et al., 2012) and indicate together that HIV-1 takes advantage of the existence of these macrophage-specific compartments. This feature of the macrophage may explain, at least in part, why VCCs are exclusive to macrophages.

How the viral precursor p55 Gag is recruited to preexisting CD36⁺ compartments remains unclear but may be linked to the capacity of both proteins to preferentially associate with membrane rafts (Nguyen and Hildreth, 2000; Ono and Freed, 2001; Zeng et al., 2003; Wolf et al., 2007). Although we did not find any evidence for a direct CD36–Gag interaction, it may occur indirectly, for instance through CD9. Indeed, we observed in uninfected macrophages an extensive co-distribution of CD36 and CD9. Moreover, CD36 has been shown to interact with CD9 (Miao et al., 2001; Huang et al., 2011; Kazerounian et al., 2011; Heit et al., 2013) and p55 Gag to sequester CD9 in plasma membrane rafts (Krementsov et al., 2010).

In macrophages, CD36 motion is subjected to linear confinement due to cytoskeleton constraints thereby promoting receptor encounters and clustering (Jaqaman et al., 2011). Interestingly, ligand internalization via CD36 occurs through the formation of multimolecular complexes together with β1 or β2 integrins, and CD9 and/or CD81 (Heit et al., 2013), which are components of the VCC (Deneka et al., 2007; Pelchen-Matthews et al., 2012) and of the membrane of HIV-1 particles produced by macrophages (Chertova et al., 2006). The tetraspanins CD9 and CD81 are present in membrane microdomains often referred to as TEM (tetraspanin-enriched microdomains) that are distinct from lipid raft microdomains (Nydegger et al., 2006). Importantly, through its oligomerization, p55 Gag promotes the coalescence of TEM and lipid rafts in HeLa cells. Therefore, from all these studies, we can hypothesize that Gag oligomerization at the cytosolic face of CD36⁺ membranes can be instrumental in the recruitment and concentration of TEMs together with lipid rafts, bringing together several tetraspanins through a network of membrane interactions. Such recruitment may favor the formation of platforms conducive to viral budding. Future work will aim at elucidating the precise role of CD36 in the HIV-1 cycle in macrophages.

Another major finding of the present study is that exposure of infected macrophages to CD36-specific antibodies inhibits viral release by inducing their intracellular retention. The viral content within the anti-CD36 antibody-treated cells was greater than in cells treated with an isotype control antibody and the VCCs were larger. Importantly, this corresponded to a reduced ability of the macrophages to transfer virus to primary T cells. In addition, this inhibitory effect appears to be virus-specific, as the viral production of influenza-infected macrophages remained unaffected by their exposure to the anti-CD36 mAb (unpublished data).

The anti-CD36 antibody effect on HIV-1 release was exclusive, as antibodies specific for other scavenger receptors or other VCC markers such as CD81, ICAM-1, and CD9 did not affect HIV-1 release. Antibodies specific for CD36, CD81, and ICAM-1 were internalized into most of the Gag⁺ compartments after 2 h of incubation at 37°C. In an experimental system similar to ours (3 h at 37°C), anti-CD63 antibodies accessed VCCs in HIV-1–infected macrophages (Deneka et al., 2007). Others (Chu et al., 2012b) and us have found that antibodies specific to CD9, CD63, CD81, and ICAM-1 do not access VCCs at 4°C or in fixed nonpermeabilized cells. It is therefore likely that, in HIV-1–infected macrophages, specific antibodies can be internalized into VCCs through an active mechanism that does not take place at 4°C but requires time to become detectable by immunofluorescence (mAbs were only weakly detected in VCCs 1 h after incubation in our experiments and in another study [Koppensteiner et al., 2012]). However, although other antibodies could be taken up into the VCC, only exposure to anti-CD36 antibodies led to the intracellular retention of virions.

The mechanism of the antibody-mediated inhibition of HIV-1 release does not seem to rely on signaling through CD36. Indeed, we show that the inhibitory effect of the mAb on HIV-1 release can be reproduced in HeLa cells transduced with CD36 wt, but also with CD36 deleted from (or substituted in) most of its cytoplasmic tail. This includes CD36 mutated in key residues of the cytoplasmic tail required for CD36 signaling (Fig. 9 F, references). We also show that endogenous ligands of CD36, MOxLDL, TSP1, and COL1 do not inhibit viral particle release by infected macrophages while they induce signaling.

Similarly, Fc receptor signaling does not seem to be involved in the CD36-specific mAb inhibitory effect on viral release. Indeed, HeLa-CD36 cells, which do not express FcR, are susceptible to the antibody-mediated inhibition. In addition, all our experiments have been performed in parallel using isotype controls, which did not affect viral release. Similarly, antibodies specific for ICAM-1, CD81, and CD9 that are able to bind viral particles, as well as engage FcRs, had no effect on viral release. (Of note, the anti–ICAM-1 mAb is the same isotype as the anti-CD36 mAb.) Finally, blockage of FcR binding and signaling using a commercial FcR antagonist had no effect on the anti-CD36 mAb-mediated inhibition of particle release by macrophages.

Different hypotheses can be evoked to explain the antibody-induced retention, including the clustering of the virions within the VCC lumen through mAb binding to CD36 on the viral membrane and on the limiting membrane of the VCC. Immuno-EM analysis of HIV-1–infected macrophages exposed to anti-CD36 mAb reveals that the inhibitory mAb is associated with virions present within the lumen of the VCCs. Similar experiments performed with HIV-1–infected HeLa-CD36 cells revealed clustering of the virions at their assembly site, i.e., at the plasma membrane. In all likelihood, exogenously added anti-CD36 mAb access the VCCs of infected macrophages, where they retain virions. Moreover, our data are compatible with the idea that the antibody effect could take place without affecting the virions themselves, at least in the short term, because free virions remain equally infectious in the presence of the anti-CD36 antibody.

Further supporting the clustering effect of the mAb, the anti-CD36 mAb FA6-152, but not its Fab fragment, mediates the inhibition of virion release, suggesting that bivalency of the mAb is required. In contrast to FA6-152, SMΦ mAb does not inhibit HIV-1 release. Both mAbs are specific for the same CD36 region (155–183; Daviet et al., 1995), but their epitopes have not been precisely defined and can be different. The fact that SMΦ is an IgM and FA6-152 an IgG1 supports the idea of a clustering induced by the bivalent ligand (FA6-152) and not by multivalent ligands like MOxLDL or SMΦ. It can be proposed that the density of CD36 on the virion and VCC membranes, and the valence and affinity of the mAbs are critical for the clustering effect. However, more complex mechanisms cannot be excluded. For instance, the tetherin/BST2 protein has been recently shown to be important for HIV-1 viral cycle in macrophages, possibly by mediating retention of the virions in the VCCs (Chu et al., 2012a). In our system, the anti-CD36 antibody exposure did not modify the level of expression of tetherin/BST2 even after 2 d of treatment (unpublished data). However, this does not rule out a possible contribution of tetherin/BST2 in the observed retention of virions within VCCs.

Our data show that the anti-CD36 antibody-induced retention of virions into VCCs results in a powerful inhibition of virus release and of macrophage to T cell transmission of HIV-1. Given that HIV-1 is a fragile virus, the increased intracellular retention induced by repeated exposures could lead, in time, to a strong reduction of the infectious capacity of the intracellular stocks of virions. In addition, we recently showed that the amounts of virus released by infected macrophages and their infectious capacity both decrease with time p.i. (Gaudin et al., 2013). Therefore, it may be possible to use anti-CD36 mAbs to complement existing antiviral treatments to target viral reservoirs within macrophages.

Goat polyclonal antibody anti-p24 Gag, mouse mAbs specific for human CD81 (clone TS81), CD68 (clone KP1), and SRA1 (clone 7H1G1), and rabbit antibodies specific for human LDLR (clone EP1553Y) were purchased from Abcam. We also used rabbit polyclonal antibody anti–SR-B1 (Novus Biologicals), mAbs specific for CD9 (MEM-61; Immunotools), ICAM-1 (clone HA58; BD), and CD36 (clones CLB-IVC7 and TR9 from Immunotools and Acris Antibodies, and clone FA6-152 from STEMCELL Technologies), unlabeled isotype controls (BD), and secondary antibodies conjugated with Alexa Fluor 488 or 647 (Life Technologies), Cy3, or Cy5 (Jackson ImmunoResearch Laboratories). FITC- and PE-labeled isotype controls, as well as anti-p24 Gag (KC57-FITC or KC-57 RD1), were purchased from Beckman Coulter, whereas FITC- and PE-labeled anti-CD36 (CLB-IVC7 and TR9) were purchased from Immunotools. Alexa Fluor 647–coupled anti-CD36 (clone SMΦ), its isotype control, and rabbit polyclonal anti-CD9 antibodies (H-110) were purchased from Santa Cruz Biotechnology, Inc. Human antibodies specific for EGFP and rabbit antibodies anti–α-tubulin were given by A. El Marjou (Institut Curie, Paris, France). The phospho-SAPK/JNK (Thr183/Tyr185; 81E11) rabbit mAb from Cell Signaling Technology was used. Antibodies specific for p17 Gag, as well as p55 Gag (National Institutes of Health [NIH] AIDS reagents 4811 from P. Spearman, Children’s Healthcare of Atlanta and Emory University School of Medicine, Atlanta, GA), and HIV immunoglobulins (used at 0.37 mg/ml in the infectivity assay on TZM-bl cells) were obtained through the NIH AIDS Research and Reference Reagent Program, Division of AIDS, National Institute of Allergy and Infectious Diseases, NIH. Fab fragments were produced using the Mouse IgG1 Fab and F(ab′)2 Preparation kit (Thermo Fisher Scientific). The human FcR Blocking Reagent (Miltenyi Biotec) was tested by FACS for its capacity to stain the plasma membrane of primary macrophages and was used according to the manufacturer’s guidelines. Each tube of MOxLDL (Kalen Biomedical) was used for a single experiment within 1 wk from the shipment date to avoid oxidation. TSP-1 was purchased from EMD Millipore, whereas collagen type I was purchased from Sigma-Aldrich. We also used dextran coupled to Alexa Fluor 647 10,000 MW (Molecular Probes), IL2 (eBioscience), PHA-L (Sigma-Aldrich), AZT (Sigma-Aldrich), and rhM-CSF (Immunotools) dissolved in ultrapure water (25 µg/ml stock).

pVSV-G (BD), pEGFP N1 (Takara Bio Inc.), and psPax2 (Addgene number 12260) plasmids were used. pHIV-1 Gag-iGFP, which carries an R5-tropic envelope with the V3-loop V92th014.12 (Koppensteiner et al., 2012), and pHIV-1 Gag-iGFPΔEnv were provided by M. Schindler (Institute of Virology, Helmholtz Zentrum Munich, Munich, Germany). The pHIV-1 Gag-iCherryΔEnv was obtained by subcloning the region from NarI to BamHI from the pHIV-1 Gag-iGFP into pUNO-mcs vector (InvivoGen). Then, the GFP was replaced with the mCherry sequence amplified with the primers mCherry_F_MluI (5′-TGTACAAACG­CGTGTGAGCAA­GGGCGAGGAG-3′) and mCherry_R_XbaI (5′-TTGGCTTCTA­GACTTGTACAG­CTCGTCCATGC-3′) using the restriction sites XbaI and MluI. The fragment obtained, containing the mCherry in place of the GFP, was then recloned in the pHIV-1 Gag-iGFPΔEnv backbone using BssHII and SpeI. To build the lentiviral vector pEXP DH1 CD36, the CD36 cDNA from the pENT221 CD36 (DNASU reference HsCD00045106) was cloned with Gateway technology (Life Technologies) into a pDEST DH1, which is a modified version of pCDH1 (System Bioscience) in which the MCS has been replaced by a Gateway cassette. To construct CD36-GFP, pENT221 CD36 was used as a template to amplify the CD36 cDNA with the primers CD36 N1 forward (5′-ACCGCGGGCC­CGGGATATGGG­CTGTGACCGGA­ACTG-3′) and CD36 N1 reverse (5′-CATGGTGGCG­ACCGGTGGTTT­TATTGTTTTCG­ATCTGCATGCAC-3′). The PCR product was then cloned into the pEGFP N1 using the restriction sites XmaI and AgeI. The resulting cDNAs CD36-GFP were cloned with Gateway technology back into pDONR201and then to a pDEST DH1. CD36 mutants were produced with the QuikChange II Site-Directed Mutagenesis kit (Agilent Technologies) using the pEXP DH1 CD36 as template. Primers were designed with the QuikChange Primer Design Program (Agilent Technologies).

Purification, differentiation, and culture of monocytes and macrophages, as well as culture of the cell lines Ghost, HEK293T (CRL-11268; ATCC), and TZM-bl, were performed as previously described (Gaudin et al., 2012). HeLa-CD36 cells were produced by lentiviral transduction of the CD36 gene. Puromycin selection (at 4 µg/ml; InvivoGen) was performed to obtain a pure population of transduced cells. HeLa cells, as well as HeLa cells expressing the different CD36 mutants, were cultured in DMEM with GlutaMAX completed with 10% FCS (Gibco) and 1% penicillin-streptomycin (Gibco).

The following virus strains have been produced for the present study: NL(AD8) (NLAD8), NL4-3, HIV-1 Gag-iGFP, HIV-1 Gag-iGFP ΔEnv, and HIV-1 Gag-iCherry ΔEnv. Viruses were produced by cotransfection of the appropriate proviral cDNA, and the pVSV-G plasmid into 293T cells using GeneJuice (EMD Millipore). Supernatants were harvested 72 h after transfection, filtered, and stored at −80°C. Virus preparations were titrated by infecting the Ghost reporter cell line and infectivity was measured 24 h p.i. by flow cytometry as previously described (Mörner et al., 1999). For infection experiments, macrophages were infected at a multiplicity of infection (MOI) of 0.2 and kept in culture for indicated times. HeLa cells were infected at MOI 0.5. The 132W primary isolate (used at 10 ng p24 Gag for 10⁵ macrophages) was a gift of O. Schwartz (Pasteur Institute, Paris, France). Lentiviral vectors have been produced for pEXP DH1 CD36, pEXP DH1 CD36-GFP, and for all the CD36 mutants used in Fig. 9. Transducing lentiviruses were produced with the same protocol described for the HIV-1 viral strains with the addition of the packaging psPax2 plasmid. SIV-VLP were produced and used as previously described (Goujon et al., 2008; Manel et al., 2010).

Measure of HIV-1 p24 Gag was performed using the Innotest HIV Ag mAb Screening kit (Ingen). Cell viability was evaluated using the CellTiter-Glo (Promega). For infectivity assays, virion-containing supernatants were normalized for equal amounts of p24 Gag (final concentration of 10 ng/ml) and infectious titers were determined using TZM-bl indicator cells as previously described (Gaudin et al., 2012).

Luciferase-specific, Tsg101-specific (Gaudin et al., 2012), and CD36-specific (Hs_CD36_1/6/7; QIAGEN) siRNA were used. siRNAs (100 nM final) were diluted in 250 µl OptiMEM (Gibco) mixed with 5 µl INTERFERin (Polyplus) and left at room temperature for 15 min. Complexes were added drop-wise onto the macrophage cultures. Cells were assayed 3 d later. Knockdown efficiency was checked by flow cytometry.

For immunofluorescence, cells were stained and imaged as previously described (Gaudin et al., 2012). For antibody feeding assay (Fig. 4; and Fig. 6, A and B), cells were precooled at 4°C for 20 min before replacing the medium with ice-cold medium containing the antibody of interest or its isotype control. Cells were kept at 4°C or moved to 37°C for 2 h depending on the condition before fixation and staining. Confocal images were collected as snapshot or 3D stacks with a focal step size of 0.5 µm. NIS Element, MetaMorph, and ImageJ software were used for image processing. No filter or background subtraction was used. Live imaging was performed using a Nikon Biostation IM-Q using a 20× objective. To test the connection of VCCs to the plasma membrane (Fig. 2), cells were imaged with a microscope (Ti Inverted; Nikon) fitted with a video-rate confocal system consisting of a spinning disk confocal head (Yokogawa). HIV-1–infected macrophage samples were prepared for EM as previously described (Gaudin et al., 2012).

For quantification of the Pearson’s coefficient, 3D stacks (z = 0.5 µm) of infected macrophages were acquired with a confocal system (A1R; Nikon) and analyzed with the NIS Element Ar software. For each cell, a background of 100 for each channel was subtracted and plans containing no signal were discarded. Then the Pearson’s coefficient of each plan of the stack was calculated, generating mean Pearson’s coefficient per cell.

To quantify the total volume of VCCs per cell, the total intensity of Gag into VCCs per cell, and the number of VCCs per cell, confocal plans with a z-step of 0.25 µm were acquired for >134 cells per condition. VCCs were 3D-reconstructed and analyzed in a blind manner using IMARIS 7.2.3 software. Gag fluorescence intensity corresponding to the VCCs was detected, smoothed at 0.15 µm, and a background subtraction was applied based on largest sphere diameter (0.8 µm) criteria. Automatic segmentation was performed based on a region growing estimated diameter of 0.8 µm. Then, objects were filtered to take into account only VCCs >0.050 µm³. Automated statistical analysis was run to extract the different parameters for each condition.

Analyses were performed as previously described (Gaudin et al., 2012), except for pJNK assay where lysis buffer was supplemented with 5 mM NaF and 1 mM β-glycerophosphate.

Infected macrophages were washed extensively and exposed to the various antibodies and isotype controls at 37°C for various periods of time. All antibodies were dialyzed in PBS 1× and filtered before use. Content in p24 Gag in the supernatant and in cells lysed in the CellTiter-Glo buffer for 2 h at 4°C was determined by Elisa.

Macrophages (10⁵ cells/well) were infected with VSV-G–pseudotyped NL4-3 at MOI 1.5, washed extensively 6 d later, and received 1.5 × 10⁵ CD4⁺ T lymphocytes/well. These CD4⁺ T lymphocytes were purified by negative selection (Miltenyi Biotec) from heterologous PBMCs that had been activated 48 h before in RPMI, 10% FCS, 2.5 µg/ml PHA-L, and 30 U/ml IL2. 25 µM AZT was added at 0 h (as a negative control) or 24 h after the beginning of the co-cultures. At 48 h, CD4⁺ T cells were collected, washed, fixed with 1% PFA, permeabilized, stained for Gag with KC57-FITC mAb, and analyzed by flow cytometry in a flow cytometer (Accuri C6; BD).

Video 1 corresponds to Fig. 3. Video 2 corresponds to Fig. 9.

The authors greatly acknowledge the Nikon Imaging Center@Institut Curie-Centre National de la Recherche Scientifique, as well as S. Amigorena, N. Manel, and A.-M. Lennon-Dumenil for discussions. We thank Aditi Varthaman for editing the English version of the manuscript.

Work was supported by grants from Agence Nationale de Recherche contre le SIDA (ANRS) and Ensemble contre le SIDA (Sidaction) to P. Benaroch and fellowships to R. Gaudin (ANRS), S. Berre (Curie Institute), M. Desdouits (ANRS), and B. Cunha de Alencar (Fondation pour la Recherche Médicale).

The authors declare no competing financial interests.