A Convenient Oxidation of Natural Glycosphingolipids to Their “Ceramide Acids” for Neoglycoconjugation

A new method to cleave the double bond of sphingolipids has been developed. Using limited concentrations of KMnO4 and an excess of NaIO4, in a neutral aqueous tert-butanol solvent system gave nearly quantitative yields of the oxidized product. A variety of natural glycosphingolipids (GSLs): GlcC, GalC, SGC, LC, Gb3C, Gb4C, Gg4C, Gb5C, and GM1C, gave the corresponding acids: 2-hydroxy-3-(N-acyl)-4-(O-glycosyl)-oxybutyric acids, i.e. “glycosyl ceramide acids” (GSL·CCOOH) in excellent yields (80–90%). Deacyl GSLs (dGSLs) were oxidized to acids containing the oligosaccharides devoid of hydrocarbon chains, i.e. “ceramide oligosaccharides” (dGSL·NRR1 CCOOH, where R = R1 = H; R = H, R1 = CH3CO; or R = R1 = Me). The efficacy of this method was demonstrated by transforming natural GSLs: GlcC, GalC, GalS, SGC, LC, Gb3C, and Gb4C into neoglycoproteins via coupling glycosyl ceramide acids (except GalS, which was coupled directly) to bovine serum albumin (BSA). Mass spectroscopic analysis of GalC-BSA conjugates, (GalC·CONH) n BSA and (GalS·NHCO) n BSA gave a value of 9 ± 1 and 16 ± 2 for n. Neoglycoconjugates derived from GlcC, GalC (type I and II and the behenic analog), SGC, LC, and Gb3C were recognized by the recombinant human immunodeficiency virus coat protein gp120 (rgp120). The GalS conjugate showed significantly reduced binding, and the Gb4C conjugate showed no binding. Thus, rgp120/GSL-BSA interaction requires a terminal galactose and/or glucose residue. TerminalN-acetylgalactosamine containing GSLs are not bound. The ceramide acid conjugates provide a more effective scaffold for presentation of glycone for rgp120 binding than those derived from dGSLs. The retention of receptor specificity of the glycoconjugates was validated by retention of the expected binding specificity of VT1 and VT2e for Gb3C and Gb4C conjugates, respectively. These studies open a new vista in the generation of glycoconjugates from GSLs and further emphasize the role of aglycone in glycolipid recognition.

The carboxylate group of the oxidized GSLs can be coupled to an amino matrix where the immobilized glycan can be used in the affinity purification of antibodies (30,33), glycosyl hydrolases, and transferases (34). Also, the oxidized GSLs can be coupled to proteins or tailor-made polymers to yield neoglycoproteins or multivalent, high affinity glycopolymers. Although several synthetic schemes have been contrived for the generation of such multivalent glycoforms, they inevitably require the synthesis of oligosaccharide monomers containing suitable functional groups for polymerization and/or coupling (35)(36)(37)(38). The availability of a facile method to cleavage the sphingosine double bond of a GSL or dGSL should circumvent the necessity of synthesizing such a glycan precursor. The availability of such carboxylate/amino containing derivatives provides a robust route to transform natural GSLs into various neoglycoconjugates.
Synthesis of ceramide acids from natural GSLs was investigated by Hakomori and co-workers using systems such as ozone in dichloromethane (30), KMnO 4 -crown ether complex in benzene (31) and KMnO 4 in acetone (32). Our attempts to use KMnO 4 -crown ether-benzene system to oxidize dGSLs (GalS or Gb 3 S) 2 gave the ceramide oligosaccharides in very low yields, * This work was supported by Canadian Medical Research Council Grant MT13073 and National Institutes of Health Grant R01 DK52098. 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 U.S.C. Section 1734 solely to indicate this fact.
Considering our particular need to oxidize dGSLs and the limited availability of natural GSLs, we required a microscale (Ͻ0.5 mg) procedure that gave the ceramide acid or ceramide oligosaccharide as the major (single) product. This study elaborates an oxidation method that fulfills these criteria.
To illustrate the biological potential of the method, we took advantage of the binding specificity of gp120 of HIV for GSLs (39 -41).
Solvents were dried by storing over activated (ϳ120°C for 16 h) molecular sieves. Crown ether (10 g) was recrystallized from a hexane (4 -5 ml) solution at Ϫ20°C, washed with cold (Ϫ20°C) hexanes (1 ml) and dried at 40°C under a stream of N 2 . BSA (99%, essentially fatty acid-free) was purchased from Sigma and further purified on a hydroxyapatite column (42). Recombinant gp120 (rgp120, 1 mg/ml in TBS; strain: LAV, baculovirus expression system) was purchased from Protein Sciences Corp. (Meriden, CT), and rabbit anti-gp120 polyclonal antibody (IgG, 1 mg/ml in phosphate-buffered saline) from Immuno Diagnostics, Inc. (Bedford, MA). Human sera from HIV patients containing anti-gp120 antibodies was a gift from Dr. S. Read, Division of Infectious Disease, HSC. LC, Gb 3 C, and Gb 4 C were purified from human kidney (43), Gb 5 C was purified from sheep blood (44), and GM 1 C was purified from bovine brain (45) as detailed previously. GalC, GlcC, and SGC were purchased from Sigma. Gg 4 C was prepared by acid hydrolysis of GM 1 C with 1 M acetic acid at 80°C for 1 h (46). Deacylated derivatives Gb 3 S and GalS (phychosine) were prepared by saponification of Gb 3 C and GalC at 102°C with 1 M methanolic NaOH for 3 h (44).

Synthesis of Dimethylated Derivatives Gb 3 S⅐NNMe 2 , GalS⅐NNMe 2
Formaldehyde (40 l of 37% H 2 CO solution, 500 mol) and a methanolic solution of NaCNBH 3 (100 l, 0.3 M stock) were added to a solution of dGSL (1 mg, 2 mol for GalS or 1.25 mol for Gb 3 S) in methanol (0.5 ml) (47). After stirring the reaction mixture for 16 h at room temperature (25°C), methanol was removed under N 2 and the remaining solid was then dissolved, by sonication, in 5 ml of distilled water. The resulting suspension was passed through a C-18 reverse phase cartridge, washed with 20 ml of water, and eluted with 20 ml of methanol. The estimated yield was Ͼ90% by TLC. Methylated compounds have reduced mobility on TLC; the R F for GalS and
The imidazole derivative (e.g. imidazole-Tca, 20 l, 10 mol) was added to a suspension of GalS (3 mg, 6 mol) in DCM (3 ml), and the mixture was stirred at room temperature. Monitoring the reaction by

Synthesis of Gal O C and Gal E C Homologues
To a solution of GalS (2 mg, 4 mol) in dry pyridine (2 ml), an excess of the anhydride (approximately 5 mg, 9 or 8 mol for oleic or erucic anhydrides, respectively) was added and stirred at 37°C for 18 h. Pyridine was removed under N 2 , and the residue was treated with 1 M methanolic NaOH (2 ml) for 5 h at 25°C, neutralized with 1 M HCl (2 ml), diluted with water (5 ml), extracted with Et 2 O (three times, 5 ml each), and the combined extracts were dried. Dried crude material was dissolved in DCM (1 ml) and loaded onto a silica column (0.5 ϫ 10 cm in C:M; 98:2). The free fatty acids were eluted with DCE: iso PrOH; 85:15 (20 ml), and the product was eluted with C:M:H 2 O; 80:20:2 (six 4-ml fractions). The estimated yield by TLC was Ͼ95%.

Synthesis of Peracetylated Derivatives
Method A-Method A was suitable for natural, NAc, and NNMe 2 derivatives. A mixture of 1:2 acetic anhydride:pyridine was added to a dried sample of GSL or dGSL (final concentration of 1 mg/ml) and stirred at 37°C for 2 h. The reactions were monitored by TLC (DCE: iso PrOH, 80:15) and, upon completion, dried under N 2 .
Method B-Method B was suitable for the preparation of NTca and NTfa derivatives. A mixture of 2:1 trifluoroacetic anhydride:glacial acetic acid was added to a dried sample of N-trihaloacetyl GSL derivative (final concentration of 1 mg/ml) and stirred at 25°C. The reactions were monitored every 30 min by TLC (DCE: iso PrOH, 80:15) and, upon completion, dried under N 2 .
The peracetylated crude material was dissolved in DCE (1 ml) and loaded on a silica column (for 3 mg, 0.5 ϫ 5 cm in DCE) and eluted with DCM:M (25:Y), where Y was varied from 100 l, in 100-l increments (for each solvent composition, six 4-ml fractions were collected). The mobility of the peracetylated derivatives during column chromatography varies significantly with small changes in the activity of silica gel. Concomitant changes to the polarity of the eluent may be necessary. Purity (important for the oxidation to proceed smoothly) of the product was checked by TLC, where the plates were developed with I 2 vapor. bovine brain. GSLs with specific acyl chains are denoted: Gal B C, Gal O C, Gal E C, and L 26 C, where superscripts denote behenic, oleic, erucic, or a C26-fatty acid-containing ceramide. Notations following the chem point indicate modifications to the ceramide or the sphingosine: galactosyl N-acetylsphingosine (GalS⅐NAc); galactosyl N,N-dimethyl sphingosine (GalS⅐NNMe 2 ); galactosyl N-trifluoroacetylsphingosine (GalS⅐NTfa); galactosyl N-trichloroacetylsphingosine (GalS⅐NTca). Glycosyl ceramide acids are denoted GSL⅐ C COOH, where superscript "c" indicates ceramide acid, e.g. GalC⅐ C COOH, GM 1 ⅐ C COOH, etc. Glycosyl ceramide oligosaccharides are denoted dGSL⅐ C COOH, e.g. GalS⅐NAc C COOH, Gb 3 S⅐NNMe 2 C COOH, etc. Dicarboxylic acids (from the oxidation of double bonds on acyl chain and sphingosine), e.g. GalC⅐ C COOH n COOH, where n is the length of the acyl chain. Acetylated derivatives are denoted with (OAc) n , where n ϭ number of OH groups on glycan ϩ1 for sphingosine, e.g. GalC(OAc) 5 , SGC(OAc) 4 , GalC(OAc) 5  . When clean precursors were used, catalytic regeneration of KMnO 4 proceeded smoothly until the reaction was terminated. Otherwise, the purple color diminished with concomitant formation of MnO 2 . In such cases, additional aliquots (5 l) of KMnO 4 solution was added. The reaction was quenched by the addition of 1.5 ml of quenching solution and 1 ml of water, and the resulting colorless solution was extracted with Et 2 O (three times, 5 ml each). Occasionally (due to insufficient quenching) the ether extract turned yellow, and in such cases, the combined extracts were washed with 1 ml of quenching solution. The ether extract was washed with water (two times, 1 ml each) and dried under N 2 at 25°C. Residual water in the crude product was removed by adding absolute EtOH (1 to 2 ml) and evaporating under N 2 ( Figs. 1 and 2).
For some peracetylated ceramide oligosaccharides, e.g. those with hydrophobic substitutions like Tfa or Tca (e.g. (OAc) 5 GalS⅐NTfa C C-OOH), the work-up procedure described above is applicable. However, the ceramide oligosaccharides with smaller hydrophobic substitutions such as acetyl or dimethyl (e.g. (OAc) 5 GalS⅐NR C COOH, R ϭ Ac or Me 2 ) and charged ceramide acids ((OAc) 4 SGC⅐ C COOH), partitioning into Et 2 O was inefficient. In such cases, the reaction was quenched by adding an excess of solid NaHSO 3 (50 mg). The colorless (occasionally pale yellow) suspension was dried on a rotary evaporator and extracted (three times, 5-7 ml) with C:M:H 2 O; 80:20:2 and the combined extracts was passed through a silica column (0.5 ϫ 4 cm in C:M:H 2 O; 80:20:2) to remove most of the salts (Fig. 2).
Deprotection of the ceramide acids or the ceramide oligosaccharides was carried out by treating dried material (0.

Mass Spectroscopic Analyses
The ES spectra were recorded on a Sciex API III spectrometer, FAB on a VG ZAB-SE and MALDI-TOF on a Voyager-Elite spectrometer (sinapinic acid matrix, linear mode, delayed extraction) using standard conditions.

Synthesis and Analysis of BSA Glycoconjugates
To maximize the number of glycosyl units coupled to BSA, reactions were carried out with 1:30 mol ratio of BSA to glycosyl ceramide acid. Prior to coupling, peracetylated ceramide acids were deprotected with Et 3 N solution (see oxidation protocol) and dried under N 2 . Residual acetate was removed by adding an acidic aqueous ethanolic solution (to 1 mg of ceramide acid was added 1 ml of 0.01 M HCl in 9:1, EtOH:H 2 O) and dried on a rotary evaporator. During coupling, a dried sample of ceramide acid (1 mg of deprotected acid) or dGSL was added a solution of BSA (1 mg/ml in 40 mM phosphate buffer, pH 7.4) and NHS (approximately 2 mg, 20 nmol) and the pH of the resulting mixture was adjusted (with 0.1 M NaOH) to between 7 and 8. EDAC was then added in three portions (total of 6 mg, 30 nmol) at 2-h intervals and stirred at room temperature for 12 h. Low molecular weight species (reagents, etc.) were removed by washing (three cycles) in a Centricon-30 (which had been first washed with 1% BSA and then with water). During each cycle, the reaction mixture was concentrated to 0.5 ml and was diluted with 2 ml of 40 mM phosphate buffer (Fig. 7).

Ligand Binding
Neoglyoconjugates were adsorbed on to nitrocellulose membranes either directly or transferred after SDS-PAGE. The membranes were blocked with 2.5% milk powder in TBS 100 (10 mM TBS, 100 mM NaCl) for 1 h at room temperature, rinsed three times with TBS 100 , and incubated with rgp120 (1 g/ml) in TBS 50 (10 mM TBS, 50 mM NaCl) for 3 h at room temperature. The blots were washed as above with TBS 100 and incubated with human HIV serum (1:50) in 2.5% milk powder in TBS 100 or rabbit anti-rgp120 (1:500) in TBS 100 for 3 h. After rinsing as above, blots were incubated with the secondary antibody (1:1000) in TBS 100 for 45 min. Finally, the blots were rinsed as above and the binding was visualized by treating with 4-chloro-1-naphthol (48) (Fig. 8).
The verotoxin binding assays were performed under conditions similar to the TLC overlay assay in which conjugates adsorbed to the nitrocellulose membrane were substituted for lipids on a TLC plate (48) (Fig. 8).  Comparison of lanes 7-8 and 9 -10 shows that, when prematurely quenched, the reaction contains only three components: ceramide oligosaccharide (at the origin), unreacted precursor (closer to the solvent front in 8, but closer to the origin in 10), and the intermediate.

RESULTS AND DISCUSSION
Oxidation Method-The chemical steps involved in the oxidation of natural GSLs is depicted in Scheme 1. This method provides the means to transform a natural GSL into a variety of neoglycoconjugates. Our objective was to develop a protocol that can be carried out with microgram quantities of GSLs, using readily available reagents. We reasoned that a procedure using "off the shelf" KMnO 4 could be readily implemented, as compared with ozonolysis. The oxidation step in Scheme 1 is based on the principles developed by Lemieux and von Rudloff (49); however, it is modified for GSL substrates. Oxidation of protected GSLs and dGSLs using limited amounts of KMnO 4 , with an excess of NaIO 4 in an aqueous tert-butanol solvent system gave high yields (Ͼ80%) of 2-hydroxy-3-(N-acyl)-4-(Oglycosyl)-oxybutyric acid (glycosyl ceramide acid). In the case of dGSLs, the acids generated by oxidation are referred to as ceramide oligosaccharides, to emphasize the formation of oligosaccharide free of hydrocarbon chains. Synthesis of such oligosaccharides from hexose precursors would require highly specialized synthetic organic chemistry expertise and considerable time.
Our investigations with KMnO 4 as an oxidant for GSLs showed (discussed below) that maintaining a homogeneous reaction mixture was important. We found that a tert-BuOHwater solvent system satisfied this criterion, i.e. by dissolving both the ionic KMnO 4 and NaIO 4 and hydrophobic peracetylated GSLs. It is important that pure peracetylated GSLs are used in the oxidation (see "Experimental Procedures"). It was observed that impurities formed during peracetylation impede the catalytic regeneration of KMnO 4 , presumably by precipitating NaIO 4 . The quantities of the reagents used in the reaction have been adjusted such that they can accommodate at least a 50% variation in the concentration of GSLs. Therefore, the procedure can be directly employed for similar quantities of different GSLs. The minimum number of equivalents of NaIO 4 and KMnO 4 used with respect to the GSL are 10 and 0.75, respectively, and the measured pH of the initial reaction mixture was approximately 7. Although, in principle, catalytic amounts of KMnO 4 should have been adequate, using closer to 1 eq (i.e. in comparison to GSL precursor) decreased the reaction time significantly. Attempts to employ greater than 1 eq of KMnO 4 resulted in the formation of brown manganese dioxide precipitates, which, in turn, led to lower yields.
Prior to the development of this new procedure, our attempts to use the previously described KMnO 4 -crown ether-benzene system (31) gave very low yields. For example, oxidation of 1 mg of Gb 3 S gave undetectable product. Further investigations suggested that the heterogeneous nature of the reaction mixture, from excess solid KMnO 4 and MnO 2 precipitate, was an important factor contributing to low yields. Analysis of the precipitate demonstrated that it trapped ϳ25% of the starting material. Fig. 5 depicts HPTLCs of products obtained from a microscale oxidation of GalC(OAc) 5 using KMnO 4 -crown-ether and the KMnO 4 -NaIO 4 -tert-BuOH methods. Analyzing the product by HPTLC with a basic solvent system showed that the sample from the new method gave a strong band at the origin (due to the formation of the salt). The crown ether method gave no such band. Oxidation using KMnO 4 -acetone (32) system gave some product (approximately 30%) that showed similar characteristics on TLC analysis as the products from KMnO 4 -NaIO 4 -tert-BuOH method. Here again, the formation of MnO 2 precipitate might have led to the lowering of the yield, again suggesting the importance of maintaining a homogeneous reaction mixture.
Glycosyl Ceramide Acids- Fig. 1 shows the TLCs of the protected ceramide acids, GSL(OAc) n ⅐ C COOH of the natural GSLs (GalC, LC, Gb 3 C, Gb 4 C, Gg 4 C, Gb 5 C, and GM 1 C). For the higher homologue GSLs (Gb 4 C and above), the desired products are formed in almost quantitative yields. In the case of shorter chain GSL (Gb 3 C and shorter), TLC analysis (Fig. 2,   FIG. 3. TLC (nanosil plates, visualized with orcinol) showing the relative migration of deprotected ceramide acids, GSL⅐ C COOH, synthesized by treating the material used in Fig. 1 with lanes [7][8][9][10] shows only two additional bands with higher R F values. Of these bands, the one closer to the solvent front is the starting material and the one beneath, which usually appears as a doublet (diasteriomers), corresponds to an oxidation intermediate.
An interesting feature of the reaction is the appearance of more than one band for each of the protected ceramide acids GSL(OAc) n ⅐ C COOH, which upon deprotection collapse to fewer bands (Fig. 3). We hypothesize that the multiple bands seen for GSL(OAc) n ⅐ C COOH are in part due to fatty acid heterogeneity and to the additional cleavage of double bonds present in unsaturated fatty acid chains. The heterogeneity in the fatty acid is apparently reflected in the R F to a higher degree than variation in the number of sugar residues. This hypothesis was tested by synthesizing GalC derivatives containing oleic (cis-9 octadecenoic, C18) or erucic (cis-13 docosenoic, C22) acyl chains. After oxidation each of these monounsaturated fatty acid containing homologues gave two products in approximately 6:1 ratio (Fig. 1B). The products from erucic homologue showed higher R F value than the products from the oleic homologue. On the assumption that the product obtained from the oxidation of the sphingosine double bond, (i.e. the ceramide acid) will have a greater hydrophobic character than the product from the oxidation of both double bonds (GalC(OAc) 5 ⅐ C COOH n COOH, n ϭ 9 or 13), the major band is assigned to the former case. This study also suggests that the oxidation has selectivity toward the sphingosine double bond.
Glycosyl Ceramide Oligosaccharides-Ceramide oligosaccharides, having two reactive groups for conjugation, enables the synthesis of more types of glycoconjugates than ceramide acids. Fig. 2 shows the TLCs of the protected ceramide oligosaccharides derived from GalS and Gb 3 S. The Tfa or the Tca groups can be removed to generate an amine function at an appropriate stage of glycoconjugate synthesis. Oxidation of dGSL(OAc) n precursors proceeded at a slightly faster rate (than GSL(OAc) n ) and gave good yields (80%) of the oligosaccharides. The oxidation was cleaner for precursors with NAc, NTfa, and NTca groups on sphingosine. The protected and deprotected oligosaccharides from dimethyl derivatives show two orcinol-positive bands (Figs. 2 and 4). We also found that treatment of GalS(OAc) 5 ⅐NTfa C COOH with Et 3 N/M/H 2 O for a long duration (18 h at 37°C) gave two orcinol-positive bands with R F values comparable to ceramide oligosaccharides (Fig. 4). However, only the lower band stained with ninhydrin, consistent with the ceramide oligosaccharide GalS⅐NH 2 C COOH. Absence of an amine function on the second compound suggests a partial Hoffmann-type deamination (50) side reaction during the deprotection procedure.
The positive ion FAB mass spectrum of ceramide acid LC⅐ C COOH is depicted in Fig. 6. Two types of ceramide acids can be identified; mass peaks at m/z ϭ 860 (MϩNa, M ϭ L 26 C⅐ C COOH), 846, 832, 818, and 804 correspond to ceramide acids having saturated fatty acids, and the peak at m/z ϭ 764 (n ϭ 18) has a hydroxylated fatty acid. Also, the spectrum shows a series of peaks resulting from the loss of a terminal sugar (loss of one glycal fragment, 162); e.g. peaks at m/e ϭ 684 arise from the molecular ion peak at 846 (MϩNa) and the one at m/e ϭ 602 is from the ion at m/e ϭ 764 (MϩNa).
Applications-We have shown that this oxidation method is reliable and can be used to transform small quantities of natural GSLs to novel bioactive glycosphingolipid derivatives. By way of preliminary illustration of this potential, we have coupled ceramide acids of GlcC, GalC (type I, type II, and the behenic homologue), GalS, SGC, LC, Gb 3 , and Gb 4 to BSA and studied their interaction with HIV coat protein rgp120 (40,51,52) and verotoxins (17,53). The pathway of HIV entry into T-lymphocytes is initiated by the binding of gp120 to the cellular CD4 receptor. This induces a conformational change in gp120 and causes a second interaction with a chemokine receptor. These events eventually lead to the fusion of viral and cellular membranes (54,55). However, GSLs, specifically GalC and SGC, have been implicated in the HIV infection of CD4negative tissues such as fibroblasts and neural and intestinal epithelial cells (40,51,56,57). A class of synthetic GalC mimics inhibited the interaction of rgp120 to suramin and inhibited the infection of HIV-1 in human peripheral blood mononuclear cells (58). We used verotoxin-GSL interactions as a means to verify the specificity of our glycoconjugates (discussed later) (17,53).
Deprotected ceramide acids were coupled to BSA using the carbodiimide-NHS system. A molar ratio of 30:1 of the ceramide acid to BSA was employed during coupling and the SDS-PAGE separation of the conjugates is shown in Fig. 7. Although the conjugates have a defined number (discussed later) of ceramide acids per BSA, they appear as diffused bands. Also, the gels show two sets of bands, presumably corresponding to monomeric (ϳ80 kDa) and dimeric conjugates (close to the gel front).
The number of galactosyl ceramide acid units per BSA was estimated using MALDI-TOF mass spectroscopic analysis (Fig.  6). Conjugates (GalC⅐ C CO) n BSA and (GalS⅐NH) n BSA gave broad peaks around 82 and 83 kDa, respectively. However, BSA treated with EDAC-NHS alone gave a peak at 75 kDa, (compared with 66 kDa for BSA), probably due to the coupling of small peptide impurities present in BSA. Therefore, 75 kDa was considered as the mass for uncoupled BSA. Shifts of 6.1 and 7.2 kDa for (GalC⅐ C CO) n BSA and (GalS⅐NH) n BSA conjugates correspond to a mean value of 9 and 16 for n, respectively. On the assumption that the broader peaks of the conjugates (as compared with the uncoupled BSA) result from variation in n, the value of n was calculated to be n ϭ 16 Ϯ 2 and n ϭ 9Ϯ1 for (GalS⅐NH) n BSA and (GalC⅐ C CO) n BSA, respectively. In the case of (GalC⅐ C CO) n BSA, a (weighted) average molecular weight was used for the ceramide acid GalC⅐ C COOH. The value of n obtained by this analysis for (GalS⅐NH) n BSA is consistent with a recent report, where 17 globo H antigens (monomers with an aldehydic spacer) were coupled to BSA by reductive amination (59). Investigating various coupling conditions showed that the amine function of dGSLs couples readily, whereas the coupling of ceramide acids is more difficult (data not shown).
A Western blot analysis, where the glycoconjugates were first separated by SDS-PAGE, transferred to nitrocellulose membrane and probed for rgp120 binding is shown in Fig. 7. Due to the diffuse nature of the monomeric conjugates, the binding profile of rgp120 is also diffuse, whereas the high molecular weight conjugates show stronger signals. Qualitatively, GalC, SGC, and LC conjugates bind rgp120, whereas GalS showed reduced binding, and Gb 4 C no binding. In the case of GalC⅐ C COOH and GalS conjugates, since the substitution per BSA is known, a quantitative interpretation of binding can be made. Although nearly twice as many GalS units were coupled per BSA than GalC⅐ C COOH, the ceramide acid conjugate is a better ligand for rgp120.
Dose responses for the binding of BSA conjugates to rgp120 by dot blot are shown in Fig. 8. Conjugates derived from GlcC, GalC, Gal OH C, Gal B C, SGC, LC, and Gb 3 C showed dose-dependent binding, whereas Gb 4 C did not bind. The binding profiles of the GalS conjugate consistently showed less binding than the ceramide acid conjugates. These findings differ significantly from the binding observed for intact GSLs adsorbed on TLC plates in which, consistent with the literature (39,40,56,57), only GalC and SGC bound rgp120. We propose that the presentation of carbohydrate is different in these two cases (i.e. the natural GSL on TLC versus neoglycoconjugate), modulated by the aglycone.
Some natural GSLs, such as bovine GalC, have a significant fraction of unsaturated fatty acids, which give rise to dicarboxylic acids after oxidation. For such species, coupling could occur at the distal acid group of the acyl chain or at the ceramide acid. To ensure that the coupled ceramide acid is recognized by rgp120, a GalC homologue containing only behenic acid was synthesized and transformed into the corresponding conjugate. The behenic conjugate showed a binding profile comparable to the conjugates derived from natural type I and type II GalC.
To authenticate the Gb 4 C and Gb 3 C conjugates, a dose-response dot blot binding analysis (Fig. 8) of Gb 4 C, Gb 3 C and GlcC BSA conjugates with verotoxins VT1 and VT2 e was performed. It is well established that VT1 is highly specific for Gb 3 C, whereas VT2 e binds to both Gb 3 C and Gb 4 C (60). Neither toxin binds to GlcC. The blots clearly showed that these conjugates bind to VTs with the expected selectivity, and their binding profiles are consistent with multiple binding sites of the toxin interacting with a multidentate ligand.
Three conclusions can be made from these studies involving rgp120. First, the binding specificity for the neoglycoconjugate is distinct from that for the free GSL. Neither GlcC or Gb 3 C are bound, but their BSA conjugates are efficiently recognized. The binding of Gb 3 -BSA is of particular interest in light of the role of Gb 3 in HIV-induced cell fusion (61), suggesting that the "molecular environment" may modulate Gb 3 binding. Second, glycoconjugates derived from ceramide acids of GalC are more effective receptors than those derived from the dGSL, GalS, indicating the utility of this oxidation procedure. This could be exploited in the design of soluble GSL mimetics. For example, ceramide acids having optimized acyl units could be coupled to dendrimers or to defined peptides. Third, our studies show that when GSLs are transformed into a neoglycoprotein scaffold, rgp120 recognizes epitopes with terminal galactose or glucose but does not bind ligands containing terminal N-acetylgalactosamine. This is consistent with the observed binding profile of the variable loop-derived (V3) synthetic peptide SPC3, which showed very low affinity toward terminal GalNAc containing Gb 4 or GM 2 but bound to SGC, LC, GM 3 , and GD 3 (51).
Our finding that GlcC-BSA is a good gp120 receptor is not explained by current concepts that govern the specificity of protein/carbohydrate interaction. The binding of both GalC and GlcC conjugates with rgp120 suggests that, in this context, the 4-OH of these hexoses may not play a direct role in binding. If we assume that rgp120 is a monomer and has only one carbohydrate binding domain, a possible explanation for the altered binding specificity might be carbohydrate-carbohydrate interaction within the GalC or GlcC conjugates. Since Gb 3 and LC BSAs are bound, it may be that the binding domain of rgp120 recognizes a linear disaccharide unit (i.e. 1-4 type arrangement) containing a Gal-Gal, Glc-Glc, or Gal-Glc sequence. In the case of Gb 3 C and LC conjugates, such arrangement is covalently established within each oligosaccharide unit. In GalC and GlcC conjugates, hydrogen bonding between the glycosidic oxygen of one sugar and the 4-OH (irrespective of whether axial or equatorial) of the second sugar could give a similar hexose-hexose arrangement, particularly along the edges of the pyranose rings (rather than the planes of the two rings). We observed that the rgp120 binding to GSL conjugates, particularly for the monohexose conjugates, was significantly better at lower salt concentration (50 mM versus 100 mM). Higher salt concentrations might disrupt such carbohydratecarbohydrate interactions mediated by direct hydrogen bonding.
In summary, we have developed a new procedure for the microscale oxidation of the sphingosine double bond, which gives a high yield of the desired carboxylic acid. This procedure will allow a new systematic approach to the generation of GSL mimics that can be used in the investigation of GSL-protein interactions and design of possible therapeutic agents.
FIG. 8. Dose response for rgp120 and verotoxin BSA neoglycoconjugate binding. The highest concentration for each conjugate is 2 g of protein (corresponding to less than 200 ng of glycosylceramide acid). All conjugates are numbered according to Fig. 7. Upper panel, dot blot binding of glycoconjugates (0.5 dilutions) and rgp120, where the bound rgp120 was detected by human or rabbit antisera. The rgp120 binding of conjugates 2-5 is quantitated in the inset by pixel integration. Data in solid lines are for human antiserum and, in the dotted lines, for rabbit antiserum. Open symbols, GalS conjugates; solid symbols, GalC⅐ C COOH conjugates. Lower panel, dose-response analysis involving glycoconjugates (0.2 dilutions) and verotoxins. Controls for goat anti-rabbit (GAR) and goat anti-human (GAH) serum alone showed no binding.