Specific Interaction of HIV-1 and HIV-2 Surface Envelope Glycoproteins with Monolayers of Galactosylceramide and Ganglioside GM3*

, Cellular glycosphingolipids mediate the fusion between some viruses and the plasma membrane of target cells. In the present study, we have analyzed the interaction of human immunodeficiency virus (HIV)-1 and HIV-2 surface envelope glycoproteins from distinct viral isolates with monolayers of various glycosphingolipids at the air-water interface. The penetration of the viral glycoproteins into glycosphingolipid monolayers was detected as an increase in the surface pressure. We found that HIV-1 recombinant gp120 (IIIB isolate) could penetrate into a monomolecular film of a -hydroxylated galactosylceramide (GalCer-HFA), while ceramides, GluCer, and nonhydroxylated GalCer were totally inac-tive. The glycoproteins isolated from HIV-1 isolates LAI and NDK and from HIV-2(ROD) could also interact with a GalCer-HFA monolayer, whereas gp120 from HIV-1(SEN) and HIV-1(89.6) did not react. These data corre-lated with the ability of the corresponding viruses to

The main receptor for type 1 and type 2 human immunodeficiency viruses (HIV-1 and HIV-2, respectively) 1 is CD4, a 55-kDa glycoprotein expressed by a subset of T-lymphocytes and macrophages (1). This receptor is recognized by the surface envelope glycoproteins of both viruses, i.e. HIV-1 gp120 (2) and HIV-2 gp105 (3). Most of our knowledge of the interaction between human immunodeficiency viruses and the plasma membrane of target cells concerns HIV-1. Indeed, the binding of HIV-1 to the CD4 receptor induces conformational changes in gp120 (4), resulting in the exposure of its third variable domain (V3 loop). The fusion process is then initiated by secondary interactions between the V3 loop and a cellular HIV-1 coreceptor that can be either the ␣-chemokine receptor CXCR4 (5,6) or the ␤-chemokine receptors CCR2b, CCR3, or CCR5 (7)(8)(9), according to the cell type infected and the tropism of the virus (10).
Alternative pathways of HIV infection have been described for several CD4-negative cell types including fibroblasts (11), neural (12), or intestinal epithelial cell lines (13). Galactosylceramide (GalCer), a glycosphingolipid mainly expressed in intestinal and neural tissues, has been shown to mediate HIV-1 gp120 binding to some CD4-negative cells (14,15). Correspondingly, HIV-1 entry into CD4 Ϫ /GalCer ϩ cells could be blocked by anti-GalCer antibodies (16 -19) or synthetic soluble analogs of the glycosphingolipid (20). The domain of HIV-1 gp120 devoted to GalCer recognition has been mapped to the V3 loop, based on the following data. (i) GalCer recognition could be prevented by anti-gp120 antibodies specific for the V3 loop (21). (ii) Multibranched V3 synthetic peptides including the consensus hexameric motif of HIV-1 GPGRAF inhibited gp120 binding to GalCer and HIV-1 infection of intestinal cells (22). Curiously, these peptides were also characterized as inhibitors of HIV-1 and HIV-2 infection in CD4 ϩ lymphocytes and macrophages (23). Since they were designed to mimic the V3 loop and potentially act as molecular decoys for the V3 loop binding sites on the target of CD4 ϩ cells, our first interpretation was that the multibranched V3 peptides could bind to the HIV-1 coreceptors CXCR4 or CCR5. However, this turned out not to be the case, and the cellular binding sites for these peptides on CD4 ϩ lymphocytes were rather identified as glycosphingolipids, i.e. gangliosides GM3 and GD3 (24). These data raised the interesting possibility that GM3 and/or GD3 could represent alternative binding sites for HIV-1 gp120 on the surface of CD4 ϩ cells. Glycosphingolipids can serve as attachment sites for a number of pathogens including bacteria, parasites, or viruses (25). In addition, these lipids may play an active role in viral fusion, e.g. by increasing the lateral membrane tension and/or the curvature of the membrane (26). In this respect, GalCer has been shown to promote membrane fusion of Semliki forest virus (27) while neutral glycolipids terminating in galactose were found to mediate myxovirus-induced membrane fusion (28). The aim of the present study was to analyze the interaction of HIV-1 and HIV-2 surface envelope glycoproteins from distinct isolates with monolayers of various glycosphingolipids at the air-water interface. In these experiments, films of pure glycolipids (i.e. without carrier lipids such as phosphatidylcholine) were prepared to mimic the glycosphingolipid microdomains of the outer leaflet of the plasma membrane (29).

MATERIALS AND METHODS
Chemicals-GalCer with or without a hydroxy-substituted fatty acyl chain (GalCer-HFA and GalCer-NFA, respectively), ceramide with or without a hydroxy-substituted fatty acyl chain (Cer-HFA and Cer-NFA, respectively), GluCer, and gangliosides GM3 and GD3 were purchased from Sigma (Les Ulis, France Cells and Viruses-HIV-1 and HIV-2 viruses were produced in peripheral blood mononuclear cells (PBMC). The isolates used in this study were the laboratory strains HIV-2(ROD) (30), HIV-1(LAI) (31), and HIV-1(NDK) (32) and two primary isolates, HIV-1(89.6) (33) and HIV-1(SEN), obtained in our laboratory from an infected patient. Viral production was followed by measurement of HIV-1 p24 antigen using an ELISA capture assay (DuPont, Les Ulis, France). For HIV-2(ROD), the peak of viral production was evidenced by measurement of the reverse transcriptase activity in culture supernatants (18). A stably transfected Chinese hamster ovary (CHO) cell line expressing HIV-1 gp120 (IIIB isolate, BH10 clone) was kindly provided by Celltech through the Medical Research Council. The gp120 secreted by this cell line was referred to as recombinant gp120 throughout the manuscript.
HIV-1 and HIV-2 Infection of HT-29 Cells-HT-29 cells were exposed to the indicated virus at a multiplicity of infection of 1 median tissue culture infectious dose per cell for 16 h at 37°C. The cells were then extensively washed, trypsinated to remove excess inoculum, and seeded in new flasks. For HIV-1 isolates, viral production was monitored by dosing the p24 antigen in cell-free culture supernatants as described above. The presence of the proviral genome in HT-29 cellular DNA was determined by semi-quantitative polymerase chain reaction using the Amplicor detection kit (Roche, Neuilly, France). For HIV-2(ROD), viral production was monitored by measurement of the reverse transcriptase activity in culture supernatants.
HIV-1 and HIV-2 Surface Envelope Glycoproteins-The envelope glycoproteins were purified by lectin affinity chromatography (34) with slight modifications. Briefly, 500 ml of either viral supernatant or CHO cell supernatant containing recombinant gp120 were treated with 0.5% Empigen BB, Calbiochem Corp., La Jolla, CA) for 2 h at room temperature. The viral glycoproteins were then purified on a column of snowdrop lectin Galanthus nivalis (GNA) coupled to agarose (Sigma) as described (34). Each preparation was found highly purified and contained few contaminating proteins as assessed by SDS-polyacrylamide gel electrophoresis. The envelope glycoproteins were further characterized by ELISA using rabbit polyclonal antibodies raised against recombinant gp120 or, in the case of HIV-2 gp105, the serum from an infected patient. Binding of HIV-1 gp120 to CD4 was analyzed by ELISA according to Moore et al. (35).
Surface Pressure Measurements-Monolayer experiments were car-ried out in a laboratory equipped with a filtered air supply. The surface pressure was measured with a fully automated computerized Langmuir film balance from KSV (Oriel, Courtaboeuf, France). After being dissolved in a mixture of hexane:chloroform:ethanol, 11:5:4 (v:v:v) (36), lipids were spread inside a teflon tank. In all experiments, the subphases were 10 ml of pure water obtained by filtration through a milli-Q water purification system (Millipore, Saint-Quentin, France). For the determination of isotherms, the surface pressure was measured as a function of area surface. When the barrier of the film balance was run across a subphase of pure water, no change in surface tension was observed. As a further precaution against the introduction of surface active contaminant with the pure water, the surface layer of water was removed with a pasteur pipette connected to a water-driven aspirator after the barrier had been run across the subphase. This compression cleaning cycle was repeated twice before films of lipids were spread. Films of lipids were compressed at a rate of 10 mm/min. All isotherms were run at least twice in the direction of increasing pressure. Each run was performed with a fresh film and subphase. To measure the interaction of the V3 peptide or HIV-1/HIV-2 surface envelope glycoproteins with lipid monolayers, the lipids were spread inside the teflon tank, and various concentrations of ligand were added into the subphase. The increase in surface pressure was then measured as a function of ligand concentration. Equilibrium was reached within 120 -240 mn, depending on the extent of pression variation.

RESULTS
Interaction of V3 Loop Peptide and HIV-1 gp120 with Gal-Cer-The interaction of the multibranched V3 loop peptide and recombinant gp120 with monolayers of GalCer was followed by measurement of the surface pressure increase in a constant area set up. As shown in Fig. 1, the addition of either gp120 or the V3 peptide under a GalCer-HFA monolayer of 12 mN/m resulted in a maximal pressure increase of 9.2 and 15.6 mN/m, respectively. However, the V3 peptide was less efficient since the concentration required to obtain the maximal effect was 30 nM instead of 0.6 nM for gp120. Similarly, the half-maximal response was obtained with 4 nM V3 peptide and 0.25 nM gp120. To investigate the lipid specificity of the penetration process, experiments were performed with GalCer-HFA, GalCer-NFA, GluCer, ceramide-HFA and ceramide-NFA (Fig. 2). In these studies, the increase in surface pressure (⌬⌸) caused by penetration of the monolayer by the added component was measured as a function of initial surface pressure of the monolayer. The underlying idea is that lipid monolayers show decreased compressibility with increasing surface pressure, so that ⌬⌸ is expected to decrease as the initial surface pressure of the monolayer increases (37). Under these conditions, Fig. 2A shows that gp120 interacts specifically with monolayers of GalCer-HFA but not of GalCer-NFA. The specificity of the recombinant glycoprotein for GalCer-HFA was further demonstated by the lack of penetration of monolayers of GluCer or ceramide. In contrast, the V3 peptide (100 nM) could interact with monolayers of various lipid composition, yet with different efficiency (Fig. 2B). Indeed, for an initial pressure of 30 mN/m, the variation of surface pressure was 10, 5, 5.5, 0.9, and 0 mN/m for GalCer-HFA, Cer-HFA, GalCer-NFA, GluCer, and Cer-NFA, respectively. These data suggest that both the ␣-hydroxyl group on the ceramide moiety and galactose are required for an optimal interaction with the V3 peptide.
Interaction of HIV-1 and HIV-2 Envelope Glycoproteins with GalCer Monolayers-To further extend these studies originally performed with recombinant HIV-1 gp120, natural surface envelope glycoproteins from various isolates of HIV-1 and HIV-2 (Table I) were purified by affinity chromatography and assayed on GalCer-HFA monolayers. The biological activity of the purified glycoproteins was demonstrated by their ability to bind to CD4 in an ELISA assay (data not shown). As shown in Fig. 3, from the five glycoproteins studied, three could interact with a monolayer of GalCer-HFA, i.e. gp120 from HIV-1(LAI) and HIV-1(NDK) and gp105 from HIV-2(ROD). In contrast, the gp120 from HIV-1(89.6) and HIV-1(SEN) did not react with GalCer-HFA. These data show that the penetration into a GalCer monolayer is a property restricted to a subset of HIV surface envelope glycoproteins. Interestingly, there is a perfect correlation between this property and the ability of the correponding virus to infect the CD4 Ϫ /GalCer ϩ epithelial intestinal cell line HT-29 since the cells could be infected by HIV-1(LAI), HIV-1(NDK), and HIV-2(ROD), but not by HIV-1(89.6) and HIV-1(SEN) ( Table I).
Interaction of V3 Loop Peptide and HIV-1 gp120 with Gangliosides-Ganglioside GM3 is one of the main glycosphingolipids expressed by human macrophages and CD4 ϩ lymphocytes (38). The recent report that this ganglioside was recognized by the V3 loop peptide used in the present study (24) raised the interesting possibility that GM3 could represent an alternative binding site for HIV-1 gp120. As shown in Fig. 4, both the V3 peptide and recombinant HIV-1 gp120 could interact with a monolayer of GM3. The reaction was dose-dependent, with a maximal pressure increase of 16.3 mN/m for 140 nM peptide and 13.2 mN/m for 5 nM gp120. The concentration required for the half-maximal effect was 24 and 2.5 nM, respectively, for the peptide and gp120, confirming the highest efficiency of gp120 in this assay. In addition to GM3, the V3 peptide could interact with a monolayer of GD3, a ganglioside expressed on the surface of human CD4 ϩ lymphocytes but absent from most T-cell lines and macrophages (38). As shown in Fig. 5, the peptide showed, however, a marked preference for GM3. For an initial surface pressure of 30 mN/m, the variation of surface pressure induced by 140 nM of peptide was 12.7 and 6.3 mN/m for GM3 and GD3, respectively. In contrast, gp120 was much more specific and recognized almost exclusively GM3.
Interaction of HIV-1 and HIV-2 Envelope Glycoproteins with GM3 Monolayers-Since recombinant HIV-1 gp120 showed a high reactivity with monolayers of GM3, experiments were then conducted to study the interaction of purified HIV-1 and HIV-2 surface envelope glycoproteins with GM3 monolayers. As shown in Fig. 6, all five glycoproteins tested at a concentration of 5 nM could recognize GM3, which is in marked contrast with the data obtained with GalCer (compare Figs. 3 and 6). DISCUSSION In the present report, we have studied the interaction of HIV-1 and HIV-2 surface envelope glycoproteins with monolayers of glycosphingolipids at the air-water interface. One of the most important results of this study is the clear-cut confirmation that HIV-1 gp120 interacts preferentially with GalCer-HFA and not with GalCer-NFA, GluCer, or ceramides. Indeed, addition of HIV-1 gp120 in the aqueous phase underneath a monolayer of GalCer-HFA induced a marked increase of the surface pressure, and the effect was gradually decreased as the initial pressure of the monolayer increased. From a physical point of view, these data are interpretated as an evidence for ligand insertion into the lipid monolayer (39). In the case of gp120, the initial event is probably a primary interaction with galactose, since the glycoprotein does not bind to GluCer or ceramides (40) and does not affect the surface pressure of monolayers formed by these lipids. Then, secondary interactions with the ceramide moiety of GalCer may lead to partial insertion of gp120 into the monomolecular film. These data definitely identify GalCer-HFA as the only form of GalCer able to act as a receptor for HIV-1. In this respect, gp120 is far more specific that the anti-GalCer antibody R-mAb (15) which recognizes both forms of GalCer as well as several related galactolipids (41). However, GalCer-HFA is the main GalCer species expressed by the intestinal cell line HT-29, which is consistent with the inhibitory activity of the R-mAb in infection assays (15,18,19). The key role of GalCer-HFA in the infection of HT-29 cells is further demonstrated by our study with the glycoproteins purified from different HIV-1 and HIV-2 isolates. As shown in Table I, recognition of GalCer-HFA is restricted to a subset of HIV-1 and HIV-2 isolates that can infect CD4 Ϫ / GalCer ϩ cells.
Another important result of this study is the identification of GM3, one major ganglioside of human macrophages and CD4 ϩ lymphocytes (38), as a potential binding site for both HIV-1 and HIV-2. Indeed, all five of the HIV-1/HIV-2 surface envelope glycoproteins tested could interact with a GM3 monolayer. Moreover, the binding of recombinant HIV-1 gp120 to GM3 extracted from human PBMC was demonstrated by ELISA (data not shown). Taken together, these data suggest that GM3 may potentiate the adsorption of human immunodeficiency viruses to the surface of CD4 ϩ cells. Interestingly, recent data suggest that GM3 is closely associated with CD4 in plasma membrane microdomains of human CD4 ϩ lymphocytes (42). Thus, it is tempting to speculate that GM3 molecules may participate to the early steps of the fusion process through multiple interactions with both CD4 and gp120. In absence of CD4 expression, GM3 ϩ cell lines may also be infectable, providing that they express a coreceptor allowing HIV-1 entry. This is the case for rhabdomyosarcoma RD cells, which coexpress GM3 2 and CXCR4 (43). However, the rate of infection of these cells is low compared with GM3 ϩ /CD4 ϩ cells (19), suggesting that GM3 may function preferentially in synergy with CD4.
Originally, this study was motivated by our recent observation that a V3 loop-derived synthetic peptide inhibited HIV-1 and HIV-2 infection in both CD4 Ϫ and CD4 ϩ cells through interaction with GalCer and GM3/GD3, respectively (24). These data, obtained by a solid phase radioassay, were fully 2 D. Hammache, unpublished data.  confirmed by the monolayer approach. However, the specificity of the peptide for glycosphingolipids appeared to be much weaker than the surface envelope glycoproteins analyzed in the present study. For instance, the peptide could interact with ceramide-HFA and GD3 monolayers, whereas none of the five viral glycoproteins tested could. The high flexibility of the peptide in aqueous solution is certainly responsible for this striking loss of specificity. Moreover, the conformation of the V3 loop in native gp120 depends on fine interactions with several domains of the glycoprotein (44). Thus, the mean conformation of the synthetic peptide may differ slightly from the V3 loop in its natural intramolecular environment. Yet it should be noted that the peptide has retained the ability to interact preferentially with GalCer-HFA and GM3 monolayers. This may suggest that the V3 loop of HIV-1 and HIV-2 surface envelope glycoproteins is involved in the recognition of cell surface glycosphingolipids. Considering the V3 loop amino acid sequences of the glycoproteins tested in this study (Table I), it is likely that the structural basis of GalCer recognition is conformational rather than sequential. Indeed, the GPGRAF motif is found in both HIV-1 LAI and 89. 6 isolates, yet only the gp120 from LAI recognizes GalCer. In the same way, the V3 loops from HIV-1(NDK) and HIV-2(ROD) do not contain the GPGRAF sequence but do interact with GalCer. According to crystallographic studies and molecular modeling, in all isolates sequenced so far, the tip of the V3 loop may adopt a typical double-turn conformation that is probably critical for viral infectivity (45). These data suggest that the conformational changes generated by sequence variability of the V3 loop respect the binding site for GM3. In contrast, only a subset of HIV-1 and HIV-2 isolates have a V3 loop able to recognize GalCer, in agreement with infection studies.