Synthetic soluble analogs of galactosylceramide (GalCer) bind to the V3 domain of HIV-1 gp120 and inhibit HIV-1-induced fusion and entry.

Galactosylceramide (GalCer) is an alternative receptor allowing human immunodeficiency virus (HIV)-1 entry into CD4-negative cells of neural and colonic origin. Several lines of evidence suggest that this glycosphingolipid recognizes the V3 region of HIV-1 surface envelope glycoprotein gp120. Since the V3 loop plays a key role in the fusion process driven by HIV-1, we decided to synthesize soluble analogs of GalCer with the aim to develop a new class of anti-HIV-1 agents that could neutralize HIV-1 infection through masking of the V3 loop. We describe a short route, in three steps, for the synthesis of soluble analogs of GalCer, using unprotected lactose as the starting sugar. The analogs were prescreened in an assay based on the interaction between a V3 loop-derived synthetic peptide and [3H]suramin, a polysulfonyl compound displaying high affinity for the V3 loop. One of the soluble analogs, i.e. CA52(n15), strongly inhibited the binding of [3H]suramin to the V3 peptide, with an IC50 of 1.2 μM. This molecule was also able to inhibit [3H]suramin binding to recombinant gp120 with similar activity. Using a competition enzyme-linked immunosorbent assay with highly specific anti-gp120 monoclonal antibodies, the region recognized by CA52(n15) could be mapped to amino acids 318-323, which corresponds to the highly conserved consensus motif GPGRAF. Interestingly, the region recognized by suramin, i.e. IQRGP-R-F, was partially overlapping this motif. CA52(n15) was able to inhibit HIV-1-induced cell fusion as well as HIV-1 entry into both CD4+ and CD4−/GalCer+ cells. A structure-activity relationship study showed that: (i) the antiviral activity of soluble analogs of GalCer correlates with V3 loop binding, and (ii) the hydrophobic moiety of the molecule plays an important role in this activity. Taken together, these data show that synthetic analogs of GalCer can inhibit HIV-1 entry into both CD4− and CD4+ cells through masking of the V3 loop.

The third variable region of the HIV-1 1 surface glycoprotein gp120 (V3 loop) appears to play a key role in HIV-1 infection and pathogenesis (1,2). This domain is the major immunodominant epitope for the generation of neutralizing antibodies (3) and is essential for virus infectivity and tropism (4 -6). A current idea is that the V3 loop may be involved in the postbinding events necessary for viral entry into the cells (7,8), probably by interacting with secondary receptors (2,9). Due to the high variability of the V3 loop sequence, the neutralizing activity of anti-V3 antibodies is generally restricted to one or, in the best cases, to a few related HIV-1 isolates, which renders vaccinal strategies particularly puzzling (10,11). However, since the V3 domain of most HIV-1 strains contains several well conserved basic amino acid residues (12), it can bind to a wide variety of anionic compounds, including sulfated polysaccharides, heparin, and suramin, which are efficient inhibitors of HIV-1 infection in vitro (13)(14)(15). Moreover, the V3 loop is also involved in the recognition of galactosylceramide (GalCer), a glycosphingolipid allowing HIV-1 entry into CD4-negative cells of neural and colonic origin (16 -19). These data prompted us to use soluble analogs of the GalCer receptor as potential inhibitors of HIV-1 infection. The synthetic scheme of these molecules was based on the original use of unprotected lactose, a low cost disaccharide, as the starting sugar (20).
We report here the synthesis and characterization of such soluble GalCer analogs that can block HIV-1-induced fusion as well as entry into both CD4 Ϫ and CD4 ϩ cells. These analogs have been first evaluated for their ability to inhibit the binding of [ 3 H]suramin to SPC3, a synthetic peptide displaying eight V3 consensus motifs (GPGRAF) radially branched on uncharged poly-Lys core matrix (15). This prescreening assay proved useful to select those analogs that recognized the V3 loop, and a good correlation was found between the anti-HIV-1 activity of a given analog and its affinity for the V3 loop. These data show that synthetic soluble analogs of glycosphingolipids may represent a new class of anti-HIV-1 drugs that could be obtained at a large scale with a low cost.

EXPERIMENTAL PROCEDURES
Materials-SPC3, i.e. (GPGRAF) 8 -(K) 4 -(K) 2 -K-␤A, was generously provided by M. Mollard (Eurethics, Paris, France). The peptide was * This work was supported by the Fondation pour la Recherche Médicale (Sidaction). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  synthesized according to Tam (21) purified to homogeneity, and the amino acid analysis of the purified peptide agreed with the deduced amino acid ratios. The other peptides used in this study were obtained as described previously (22). [ 3 H]Suramin (49 Ci/mmol) was purchased from Isotopchim (Ganagobie-Peyruis, France). Polyvinyl chloride multiwell plates (number 3911) were from Falcon-Becton Dickinson (Le Pont de Claix, France). Immulon 1 multiwell plates were from Poly Labo (Strasbourg, France). Tissue culture media and fetal calf serum were from BioWhittaker (Gagny, France). Recombinant gp120 was from Intracel Corp. (London). Polyclonal sheep anti-gp120 directed to the C-terminal domain, amino acids 497-511 (reference D7326) were from Aalto Bioreagents (Dublin, Ireland). Mouse monoclonal anti-gp120 antibodies F5, 110-A, and 110-H were generously given by F. Traincard and J. C. Mazié (Institut Pasteur, Paris, France).
Chemical Synthesis of Soluble GalCer Analog CA52(n15)-The synthetic scheme of analog CA52(n15) is illustrated in Fig. 4. The intermediate 2 is obtained in two stages from lactose and the corresponding -amino acid, after reduction with sodium borohydride according to a procedure described elsewhere (20,23). Briefly, to a solution of 11aminoundecanoic sodium carboxylate (22 mmol) in methanol (60 ml) is added 13.7 mmol of lactose monohydrate dissolved in 30 ml of water. The mixture is stirred 12 h at room temperature, then heated at 55°C for 6 h. Product 1 is not isolated. Sodium borohydride (15 mmol) is immediately added at room temperature by little portions to the reaction mixture and stirred for 12 h. The residual solution is evaporated in vacuo. The crude mixture is purified by chromatography on silica gel (230 -400 mesh) eluting with chloroform/methanol/ammoniac (5/3.5/ 1.5). After lyophilization pure product is obtained (3.4 g, 45% on two steps). The soluble analog of GalCer is then obtained by the following procedure. To a solution of compound 2 (3.28 mmol) in 250 ml of N,N-dimethylformamide, and 3.28 mmol of triethylamine is added the acylating agent 3,3-hexadecanoylthiazolidine-2-thione (6.40 mmol) prepared as described previously (20,23). The mixture is stirred for 4 days at 60°C. After evaporation to dryness, the crude product is purified by chromatography on silica gel (CHCl 3 /MeOH/NH 3 ,H 2 O ϭ 6/3/1). 1.2 g of pure product CA52(n15) is obtained (50% yield). The structures of compounds and purities were checked by 1 H, 13 C NMR, mass spectroscopy (fast atom bombardment), and microanalysis.
[ 3 H]Suramin Solid Phase Assay-100 l of SPC3 at the indicated concentration were incubated in polyvinyl chloride 96-well plates overnight at 4°C. The wells were washed three times with 200 l of phosphate-buffered saline (PBS) and subsequently treated with PBS containing 1% gelatin for 90 min at 37°C to reduce nonspecific binding. The plates were then incubated with 100 l of [ 3 H]suramin (1 Ci/ml). After 1 h at 37°C, the plates were washed five times with 200 l of PBS, each well was individualized, and the radioactivity was determined in a ␤ scintillation counter (Beckman, Marseille, France). In some experiments, nonrelevant multibranched peptides (15) were used as negative controls.
Prescreening Assay: Inhibition of [ 3 H]Suramin Binding to SPC3-The prescreening assay is based on the potential ability of synthetic GalCer analogs to compete with [ 3 H]suramin for binding to SPC3. In this assay, 100 l of SPC3 (5 M) were deposed in polyvinyl chloride 96-well plates as described above. The indicated concentration of each drug was added in competition during [ 3 H]suramin incubation. The plates were then washed and processed for radioactivity measurements exactly as for the [ 3 H]suramin solid phase assay.
Binding of [ 3 H]Suramin to Recombinant gp120 -100 l of recombinant gp120 (2.5 g/ml) were incubated in polyvinyl chloride 96-well plates overnight at 4°C. After three washes in PBS, the plates were saturated with PBS 1% gelatin for 90 min at 37°C. [ 3 H]Suramin (1 Ci/ml) was then added in either the absence or presence of synthetic GalCer analogs. The binding of [ 3 H]suramin to recombinant gp120 was measured by counting the radioactivity as described above.
Recombinant gp120 ELISA-100 l of anti-gp120 polyclonal sheep antibodies directed to the C-terminal domain of gp120 (10 g/ml in 5 mM Tris-HCl, pH 7.4, 150 mM NaCl) were incubated in Immulon 1 multiwell plates overnight at 4°C. The plates were washed three times with 25 mM Tris-HCl, pH 7.6, 144 mM NaCl and saturated with 5 mM Tris-HCl, pH 7.4, 150 mM NaCl, low fat dry milk 2% (saturation buffer) for 120 min at room temperature. Recombinant gp120 (0.2 g/ml) was then added to the wells and incubated for 120 min at room temperature. The plates were washed three times and then incubated with the indicated anti-gp120 monoclonal antibody at a concentration of 5 g/ml in the saturation buffer for 120 min at room temperature. After three washes, the plates were incubated with peroxidase-conjugated rabbit anti-mouse IgG for 60 min at room temperature. The plates were then rinsed and incubated with o-phenylenediamine. The reaction was stopped by addition of 4 N H 2 SO 4 , and the optical density was read at 490 nm with a Biotek EL311 (OSI, Les Ulis, France) spectrophotometer. When indicated, the inhibitors (suramin and CA52) were present 15 min before and during the incubation with anti-V3 mAbs.
Binding of Anti-V3 mAb to V3 Peptides-The V3 peptides were coated on Immulon 1 multiwell plates for 1 h at 37°C. Nonspecific binding sites were saturated with PBS, containing 2% bovine serum albumin for 1 h at 37°C. After washing, the peptides were incubated with the anti-gp120 mAb F5 (2.5 g/ml) for 1 h at 37°C. The wells were then washed, and the bound mAb was revealed with peroxidase-conjugated rabbit anti-mouse antibodies (1:1000) using o-phenylenediamine as substrate. The absorbance was measured at 490 nm as described above.
ELISA of Lipids-Stock solution of lipids were prepared in chloroform/methanol (1:1, v/v) at a concentration of 1 mg/ml. The indicated amounts of lipids were then resuspended in chloroform/ethanol (1:20, v/v) and allowed to adsorb on Immulon 1 multiwell plates by evaporation of the solvent under a chemical hood. After saturation of the plates with PBS, containing 2% bovine serum albumin (1 h at 37°C), the lipids were probed with the anti-GalCer R-mAb (18) and processed for ELISA as described above.
Cell Culture-Human peripheral blood mononuclear cells (PBMC) were obtained from healthy donors, activated with phytohemagglutinin, and cultured in RPMI 1640 medium containing 10% fetal calf serum and interleukin 2 as described (22). The human T-lymphoblastoid C8166 cell line was grown in RPMI 1640 containing 10% fetal calf serum. The human colon epithelial cell line HT-29 was cultured in Dulbecco's modified Eagle's medium/F-12 with 10% fetal calf serum.
Inhibition of Syncytia Formation-Syncytia formation was observed in C8166 cell cultures at day 7 postinfection with HIV-1(LAI) at a multiplicity of infection (m.o.i.) of 0.001 tissue culture infectious dose 50% (TCID 50 ) per cell. Synthetic GalCer analogs were preincubated with the virus in 96-well culture plates in a total volume of 100 l. After 30 min at 37°C, 100 l of C8166 cells (10 6 cells/ml) were added to the wells. After 7 days of culture, the presence of syncytia was observed under a phase contrast Leica inverted microscope equipped with an integrated camera exposure system (OSI).
Inhibition of HIV-1 Infection-The synthetic analogs of GalCer were tested for their ability to inhibit HIV-1 infection in two cell targets: the CD4 Ϫ /GalCer ϩ human colon epithelial cell line HT-29 and normal human PBMC. In both cases, HIV-1 was preincubated with the indicated concentration of the synthetic analog for 30 min at 37°C. The mixture was then either deposed in six-well plates onto exponentially growing HT-29 cells (m.o.i. of 0.1 TCID 50 per cell) or used to resuspend a cell pellet of PBMC (m.o.i. of 0.01 TCID 50 per cell). The level of infection was analyzed 7 days postinfection by measuring the concentration of p24 antigen in the cell-free culture supernatants as described elsewhere (15,24). An antigen capture assay (DuPont, Les Ulis, France) was used for p24 determinations. The viruses used in these experiments were the prototype HIV-1(LAI) or the african HIV-1(NDK) isolates, as described previously (18,19,22,24).
Toxicity Assay-The effects of synthetic analogs of GalCer on the proliferation and viability of PBMC were studied in a colorimetric assay utilizing the tetrazolium salt XTT, i.e. sodium 3Ј-[-1-[(phenylamino)carbonyl]-3,4-tetrazolium]-bis(4-methoxy-6-nitro)benzenesulfonic acid hydrate (25). In these experiments, the cells were cultured in phenol red-free RPMI 1640 to reduce the blank values. Briefly, XTT was dissolved in prewarmed medium at a concentration of 1 mg/ml. Immediately before use, phenazine methosulfate (PMS) was added to the XTT solution (final concentration of PMS, 125 M). 25 l of XTT/PMS solution was added to 100 l of culture giving a final concentration of 0.2 mg/ml XTT and 25 M PMS. After incubation at 37°C for 4 -8 h, the optical density was determined using a test wavelength of 450 nm and a reference wavelength of 650 nm, subtracting blank control values with medium alone.

RESULTS
Recognition of SPC3 by an Anti-gp120 mAb Directed to the V3 Loop-A first set of experiments was designed with the aim to assess that the multibranched peptide SPC3 was representative of the tip of the V3 loop as exposed in native gp120. Using an ELISA detection method, we analyzed the reactivity of SPC3 with F5, an anti-gp120 mAb directed to the GPGRAFVT motif of the HIV-1(LAI) isolate (15). As shown in Fig. 1, this anti-V3 mAb bound to SPC3 with high specificity. In contrast, the F5 mAb did not recognize the monomeric GPGRAF motif nor SPC3 derivatives with lower valencies, i.e. the four-branched (GPGRAF) 4 -(K) 2 -K-␤A and the 2-branched (GPGRAF) 2 -K-␤A synthetic peptides. These data suggested that the presentation of the GPGRAF motif in the eightbranched peptide SPC3, but not in its derivatives with lower valencies, was reminiscent of the physiological situation. Thus, SPC3 was used throughout this study as a model for the tip of the V3 loop.

Specificity of [ 3 H]Suramin
Binding to SPC3-The binding of [ 3 H]suramin to the multibranched V3 peptide SPC3 was analyzed in a solid phase radioassay. In this assay, the peptide was coated on polyvinyl chloride 96-well plates, and [ 3 H]suramin was used as a revealing agent. As shown in Fig. 2, the binding of [ 3 H]suramin to SPC3 was dose-dependent and saturable. The binding specificity was further demonstrated by using SPC3 derivatives with either a shorter or a different motif (GPGR, GPGRA, GPGKTL, or GPGQAF). [ 3 H]Suramin did not bind to any of these peptides (Ref. 15 and data not shown), indicating that [ 3 H]suramin recognized the entire V3 consensus motif GPGRAF, with special emphasis to the Arg (R) and Phe (F) residues. Moreover, [ 3 H]suramin could not detect SPC3 derivatives with lower valencies, in agreement with the data of Fig. 1. This raised the interesting possibility to use the [ 3 H]suramin solid phase radioassay as a prescreening test for a rapid and low-cost identification of a new class of anti-HIV drugs susceptible to bind to the consensus V3 loop motif. Among the myriad of potential V3-binding compounds (heparin, sulfated sugars, neutralizing antibodies, etc . . . ), we decided to focus on soluble analogs of GalCer, an alternative receptor recognized by the V3 loop of HIV-1 gp120 (15,24,26).
Synthesis of GalCer Analogs-The synthetic analogs of Gal-Cer developed in this study were obtained in only three stages from lactose. The chemical structure of CA52(n15), the prototype of this new series of analogs, is shown in Fig. 3. Other analogs were obtained by varying the number of carbon atoms in the lateral hydrophobic chain (n ϭ 15 for CA52(n15)). The synthetic scheme of CA52(n15) is illustrated in Fig. 4. The feature of note is the use of unprotected lactose as starting sugar, a low cost disaccharide consisting of a galactose and glucose units linked by a 1-4 junction in the ␤ configuration. Thus, opening of the glucose ring by reductive amination in the first two stages of the synthesis leads to the GalCer-like unit. Another point is the terminal carboxylate function of this new compound leading to the good solubility of the substrate in aqueous media. For this reason, the molecules synthesized according to this scheme can be considered as true soluble analogs of GalCer. Indeed, the synthetic analog CA52(n15) was fully recognized by the anti-GalCer R-mAb in an ELISA assay (Fig. 5). was the only analog able to totally prevent the binding of [ 3 H]suramin. This derivative interfered with SPC3 recognition in a dose-dependent manner, with a 50% inhibitory concentration (IC 50 ) of 0.7 M (0.6 g/ml). The activity of CA52(n15) was highly specific, since a more polar derivative with only 8 carbon atoms in the hydrophobic chain, i.e. CA50(n8) did not affect the binding of [ 3 H]suramin to SPC3. Moreover, when the length of the carbon chain was intermediary between the fully active CA52(n15) and the nonactive CA50(n8) analogs, the corresponding molecule (i.e. CA49(n11), with 11 carbon atoms) showed a weak, yet nonnegligible activity, with an IC 50 of 9.4 M (7 g/ml). These data underscored the high specificity of the [ 3 H]suramin solid phase assay, which could discriminate between closely related GalCer analogs.

Inhibition of [ 3 H]Suramin
Binding to gp120 by CA52(n15)-Two experiments were conducted in order to ensure that CA52(n15) actually recognized the V3 loop motif GPGRAF. In the first one, CA52(n15) was preincubated with SPC3 before the plate was rinsed and then exposed to [ 3 H]suramin. Under these conditions, the IC 50 of CA52(n15) was not changed, which strongly suggests that the analog binds to SPC3 and not to  [ 3 H]suramin (data not shown). In the second one, CA52(n15) was tested for its ability to interfere with the binding of [ 3 H]suramin to recombinant gp120. In this case, the plates were coated with gp120, and the binding of [ 3 H]suramin was done in the presence of various concentrations of CA52(n15). The results in Fig. 7 show that CA52(n15) was indeed able to inhibit the binding of [ 3 H]suramin to the V3 domain of recombinant gp120 in a dose-dependent fashion, with an IC 50 of 2.4 M (2 g/ml).
Mapping of the CA52(n15) Binding Site on gp120 -Taken together, these data strongly suggested that CA52(n15) recognized a region of the gp120 V3 loop closely related to the consensus motif GPGRAF. In order to further define the region of the V3 loop involved in CA52(n15) binding, we used two anti-gp120 mAbs that recognize two overlapping epitopes at the tip of the V3 loop: IQRGP, recognized by mAb 110-A, and GPGRAFVTI, recognized by mAb 110-H. In these mapping experiments, gp120 was oriented on Immulon 1 microplates precoated with sheep polyclonal antibodies directed to the Cterminal part of gp120. Under these conditions, anti-V3 mAbs had their epitope readily accessible on the recombinant glycoprotein (27). The soluble GalCer analog CA52(n15) inhibited the binding of both 110-A and 110-H mAbs, with an IC 50 of 61.7 and 49.3 M, respectively (Table I). In contrast, suramin did not affect the binding of mAb 110-H (IC 50 Ͼ 500 M), while it inhibited the binding of mAb 110-A with an IC 50 of 106 M; Table I). The putative binding sites for suramin and CA52(n15) at the tip of the V3 loop are indicated in Fig. 8 (see "Discussion" for comments).
Anti-HIV-1 Activity of CA52(n15)-The antiviral activity of CA52(n15) was first evaluated in HT-29, a CD4 Ϫ cell line expressing high levels of the GalCer receptor. In these experiments, the virus (HIV-1(NDK) at a m.o.i. of 0.1 TCID 50 per cell) was preincubated with various concentrations of CA52(n15) and then exposed to exponentially growing HT-29 cells cultures in six-well plates. After 2 h of infection, the cells were treated with trypsin to inactivate residual inoculum (19), and the level of infection was analyzed 7 days later by measuring the production of HIV-1 p24 antigen in the cell-free culture supernatant. The results in Fig. 9 show that CA52(n15) treatment induced a dose-dependent inhibition of HIV-1 infection, with an IC 50 of 47 M (38 g/ml). Thus, CA52(n15) is an inhibitor of the GalCer-dependent pathway of infection.
Since the V3 loop is also involved in the fusion process between the HIV-1 particle and the plasma membrane of CD4 ϩ cells, the activity of CA52(n15) as a fusion inhibitor was evaluated in a syncytium-forming assay. In this test, human Tlymphoblastoid cell were acutely infected with HIV-1(LAI) and cultured for 7 days. At this time, the formation of syncytia was

TABLE I
Binding of anti-V3 mAbs to gp120: effect of suramin and CA52(n15) Immulon 1 96-well plates were first treated with sheep polyclonal antibodies directed to the C-terminal part of gp120. After saturation, recombinant gp120 (0.2 g/ml) was added to the wells and incubated for 120 min at room temperature. After washing, gp120 was probed with the anti-V3 mAbs 110-H or 110-A with or without suramin or CA52(n15) at various concentrations ranging from 10 to 500 g/ml. The binding of anti-V3 mAbs was revealed with peroxidase-conjugated rabbit anti-mouse IgG and o-phenylenediamine as substrate. The reaction was stopped with H 2 SO 4 and the optical density was measured at 490 nm. The results are expressed as the mean of four separate experiments. evident in the infected cultures (Fig. 10A), especially when compared with noninfected ones (Fig. 10B). Pretreatment of the virus with 123 M CA52(n15) (100 g/ml) was sufficient to abrogate the formation of syncytia, as shown in Fig. 10C. In contrast, the GalCer analog CA50(n8), which did not recognize the V3 loop ( Fig. 6), had no activity in this test, even when present overall the culture time at a concentration of 284 M (200 g/ml) (Fig. 10D). Finally, analog CA49(n11), which displayed some activity in the [ 3 H]suramin solid phase radioassay (Fig. 6), could block the formation of syncytia at a concentration of 201 M (150 g/ml) (Fig. 10E), but was not active at lower concentrations (Fig. 10F). Indeed, some syncytia could still be observed in cultures treated with 167 M (125 g/ml) CA49-(n11) (not shown). Then, the anti-HIV-1 activity of CA52(n15) was studied in normal human PBMC, which represent the natural cellular targets for HIV-1. As shown in Fig. 11, the soluble GalCer analog was able to block the infection of PBMC at 246 M (200 g/ml), and this inhibition was not associated with any toxicity as evidenced by the XTT assay. The IC 50 was 108 M (88 g/ml), in agreement with the data obtained with HT-29 and C8166 cells. DISCUSSION The aim of the present study was to synthesize soluble analogs of GalCer that were expected to bind to the V3 loop of HIV-1 gp120 and thus to block HIV-1 infection. The rationale for this strategy was based on previous observations suggesting that the GalCer receptor, putatively used by HIV-1 to infect neural cells (16,17) and colon cells (18,19), was recognized by the V3 domain of gp120: (i) anti-V3 antibodies block gp120 binding to GalCer (26); (ii) these antibodies inhibit HIV-1 infection of CD4 Ϫ /GalCer ϩ human colon epithelial HT-29 cells (15,26); (iii) synthetic multimeric peptide constructs of the V3 consensus sequence (GPGRAF) bind to GalCer and prevent HIV-1 entry into HT-29 cells (24); (iv) the V3 loop is a common genetic determinant controlling HIV-1 tropism for neural SKNMC cells (17,28) and HT-29 cells (29), as demonstrated by using chimeric proviral clones. Although these data strongly suggested the involvement of the V3 loop in GalCer recognition, direct evidences for a physical interaction between GalCer and the V3 loop of gp120 were lacking. Thus, special attention was devoted to the demonstration that the soluble GalCer analogs described in this study actually bound to the V3 loop. This was done by two distinct methods. First, we developed a quantitative V3 binding assay based on the interaction between suramin and the V3 loop of gp120 (15). In this assay, the V3 loop was mimicked by SPC3, a synthetic multibranched V3 peptide that has been recently characterized by our group as an inhibitor of HIV-1 infection in both CD4 Ϫ and CD4 ϩ cells (30). Synthetic analog CA52(n15) was a potent inhibitor of recombinant gp120, which occurs in the V3 region (31). Based on these data, one could conclude that CA52(n15) binds to the V3 motif GPGRAF harbored by both SPC3 and gp120 and thus competitively inhibits the binding of [ 3 H]suramin to the same domain. To refine this mapping, we used two anti-V3 mAbs recognizing closely related and partially overlapping epitopes of the V3 loop: IQRGP (mAb 110-A) and GPGRAFVTI (110-H). CA52(n15) inhibited the binding of both antibodies to recombinant gp120, while suramin interfered with mAb 110-A binding exclusively. However, one cannot simply conclude from these data that suramin binds preferentially to the IQRGP sequence and CA52(n15) to GPGRAFVTI. Indeed, our results demonstrate that both suramin and CA52(n15) bind to the GPGRAFbearing multibranched peptide SPC3. Moreover, recent data showed that [ 3 H]suramin did not bind to SPC3 derivatives exhibiting a shorter motif (i.e. GPGR and GPGRA), which underscores the importance of the C-terminal Phe residue (F) in the GPGRAF motif of SPC3. In addition, acetylation of the N-terminal Gly residue (G) of the motif abolished the binding of [ 3 H]suramin (data not shown). Thus, it is likely that the suramin binding site on the V3 loop involves the Gly-Pro doublet (GP) and the downstream Phe (F), although it is probable that the two Arg residues (R) may stabilize the interaction with negatively charged sulfonyl groups of suramin. Thus we propose that suramin binds essentially to the discontinuous motif IQRGP-R-F (Fig. 8). As for CA52(n15), its binding site was assigned to the GPGRAF linear sequence, based on: (i) the ability of both CA52 (this study) and natural GalCer (15) to bind to SPC3 and (ii) the fact that CA52(n15) competes with the binding of mAbs 110-H and 110-A to recombinant gp120 (binding to the GPGRAF motif should totally or partially mask the epitopes of 110-H and 110-A mAbs, respectively). This tentative mapping of suramin and CA52(n15) to partially overlapping binding sites on gp120 could conciliate the finding that both drugs bind to SPC3 while recognizing slightly different regions of the V3 loop as revealed by competition ELISA.
The validation of the [ 3 H]suramin solid phase radioassay as a prescreening test for V3 loop-targeted anti-HIV-1 compounds is emphasized by the virological data obtained with the soluble GalCer analogs. Indeed, CA52(n15), which was the more potent inhibitor in the prescreening assay, was also the most potent inhibitor of HIV-1 infection. CA52(n15) inhibited the formation of syncytia induced by HIV-1 in C8166 cultures, in agreement with the well established role of the V3 loop in HIV-1 fusion (7,8). This analog also inhibited the infection of HT-29 cells and PBMC through the GalCer and CD4 pathways, respectively, with similar IC 50 values. Taken together, these data are consistent with the putative mechanism of action of such compounds, i.e. neutralization of the virion through masking of the V3 loop of gp120. In addition, the synthetic analog CA52(n15) appeared to be a more potent inhibitor of HIV-1 infection than the authentic soluble form of GalCer, i.e. 3Ј-sulfo-GalCer (sulfatide). Indeed, recent data from our laboratory showed that sulfatide could inhibit the GalCer-dependent pathway of HIV-1 entry into HT-29 cells, but not the CD4-dependent pathway in PBMC. 2 The lower efficiency of sulfatide may be related to a decrease of concentration following the spontaneous transfer of this natural glycolipid from the aqueous medium to the plasma membrane (32).
Interestingly, the IC 50 of the antiviral effect of CA52(n15), i.e. 100 g/ml, was superior to the IC 50 determined by biochemical means (i.e. 1 g/ml), in agreement with the notion that the exposure of the V3 loop on the virion spikes is different from the one of monomeric gp120 (33). Similar data were obtained with anti-V3 mAbs (data not shown). However, the difference of activity between CA52(n15) and CA49(n11) in the biochemical assays was also found in the antiviral assays, especially when one compares the ability of each analog to inhibit the formation of syncytia in our fusion assay (Fig. 10). Finally, CA50(n8), which did not bind to the V3 loop, was also devoid of antiviral activity, at least over the range of concentration tested (up to 200 g/ml). In conclusion, the antiviral activity of the soluble analogs of GalCer was consistent with their activity in the prescreening assay.
The difference between the fully active CA52(n15), the nonactive CA50(n8), and the intermediary active CA49(n11) correlated with the level of hydrophobicity of the analog: the longer the hydrophobic moiety, the higher the biological activity. Yet analogs with more than 14 methylene residues were not active, suggesting that a narrow range of hydrophobicity is necessary to achieve the antiviral activity. The lack of activity of those analogs with a high degree of hydrophobicity could be tentatively explained by their ability to be incorporated in the plasma membrane, resulting in a significant decrease of concentration. This phenomenon has been reported previously for various glycolipids, including sulfatide (see above). Further studies will help to clarify this point.
In conclusion, we describe the synthesis of a new class of low-cost anti-HIV-1 compound that neutralize HIV-1 infection through masking of a highly conserved motif of the V3 loop. The soluble analogs of GalCer characterized in this study will constitute the basis for the design of second generations molecules, which will be selected for improved anti-HIV-1 activity. The prescreening assay described in this report will help to evaluate a wide number of molecules for their ability to bind to the V3 loop. Normal human PBMC were exposed to HIV-1(LAI) which had been treated previously with various concentrations of CA52(n15) for 30 min at 37°C. After 1 h of infection, the cells were rinsed and cultured in RPMI 1640 containing 10% FCS and interleukin 2. After 7 days, the amount of p24 antigen was measured in the cell-free supernatants (closed circles). The results are expressed as the mean of three separate experiments (ϮS.D.). The toxicity of CA52(n15) was assessed on PBMC that were treated with CA52(n15) under similar conditions except that the virus was omitted. Cell toxicity was analyzed using the XTT assay, and the optical density was measured as described under "Experimental Procedures" (open circles).