Glycosphingolipids Covalently Linked to Agarose Gel or Glass Beads USE OF THE COMPOUNDS FOR PURIFICATION OF ANTIBODIES DIRECTED AGAINST GLOBOSIDE AND

Abstract Oxidative ozonolysis of the olefinic bond of the sphingosine moiety of either globoside (βGalNAc1→3αGal1→4βGal1→ 4βGlc→ceramide), or of the methyl ester of hematoside (αN-glycolylneuraminyl2→3βGal1→4βGlc→ceramide), and coupling of the carboxyl-bearing product to aminoethylagarose or to amino group-bearing glass beads in the presence of a carbodiimide resulted in globoside or hematoside covalently linked to agarose or glass beads. These compounds were used for purification of anti-glycosphingolipid antibodies from serum of immunized rabbits. The antibodies bound to the substrate were released by 1 m sodium iodide and their immunological properties were studied. Anti-globoside is directed toward the terminal β-(N-acetyl)galactosaminosyl-(1→3)α-galactopyranosyl structure, while anti-hematoside is directed predominantly toward the sialosyl residue of hematoside.


GENESA YOGEESWARAN, AND SEN-ITIROH HAKOMORI
From the Departments of Pathobiology and Microbiology, University of Washington, Seattle, Washington 98195 SUMMARY Oxidative ozonolysis of the olefinic bond of the sphingosine moiety of either globoside (PGalNAcl-+3crGall+4/3Gall--) 4PGlc+ceramide), or of the methyl ester of hematoside (aN-glycolylneuraminy12+3~Gall~4~Glc~ceramide), and coupling of the carboxyl-bearing product to aminoethylagarose or to amino group-bearing glass beads in the presence of a carbodiimide resulted in globoside or hematoside covalently linked to agarose or glass beads.
These compounds were used for purification of anti-glycosphingolipid antibodies from serum of immunized rabbits.
The antibodies bound to the substrate were released by 1 M sodium iodide and their immunological properties were studied. Anti-globoside is dikected toward the terminal fl-(N-acetyl)galactosaminosyl-(1-+3)a-galactopyranosyl structure, while anti-hematoside is directed predominantly toward the sialosyl residue of hematoside.
Glycosphingolipids are components of plasma membranes (2-6), whose carbohydrate moieties are exposed to the outer environment (7). In view of their changes in carbohydrate com-pc&ion on cell to cell contact (8,9), during the mitotic cell cycle,' and ori malignant transformation (10, ll), they may play some role in control of cellular interaction and proliferation (12). Some of them may serve as sites for ion transport (13,14) and specific sites for binding exogenous protein such as bacterial "exotoxins" (15)(16)(17) and lectins (18,19). Lectins have been used with increasing popularity to study various cell surface properties related to carbohydrates (20,21) ; nevertheless, most of their specificities are ill defined, and they show various kinds of "toxicity" to animal cells. They are not simple sugar-binding reagents but have multiple binding sites which could induce a series of alterations in membrane properties (see under "Discussion"). If antibodies directed against a single glycosphingolipid with defined carbohydrate structure were purified, they would be useful reagents to study various phases of surface function related to specified carbohydrate structure.
This paper describes methods for preparation of globoside and hematoside covalently linked to agarose gel (Sepharose) or to glass beads for affinity columns which enable purification of antibodies directed against those glycosphingolipids.
In addition, the immunological properties of the purified antibodies are examined.

Preparation
of N-Glycolylneuraminyllactosylceramide-acyl-glass Beads Complex and Evidence for This Linkage-The final product described in the preceding section was suspended in 10 ml bf chloroform-methanol (2:1), and 2 ml of 0.5yo sodium methoxide in methanol was added. After 30 min at room temperature, the beads were filtered and washed with 50 ml of chloroform-methanol (2:l). followed bv washing with 100 ml of water. then with 50 ml , of phosphate buffer, pH 7.0, and were stored in this buffer after addition of 2 mM of sodium azide.
One hundred milligrams of this compound were methanolyzed in 0.5 N methanolic HCl at 70" for 18 hours, and the methanolyzed products were determined by gas chromatography as the trimethylsilyl derivatives. Column-Anti-globoside serum was mixed with 56 volume of saturated ammonium sulfate at 4" and allowed to stir for 1 hour. The susnension was centrifuged at 30,000 X g for 30 min and the precipitate was dissolved in PBS followed by dialysis against three changes of PBS over 24 hours. One-half milliliter of anti-globoside solution, containing 26.2 mg of protein, was applied to a 2.5-ml column of adsorbant. If raw serum were used for the purification, heating at 56" for 30 min was performed to inactivate complement. After a 5-min incubation period, the column was eluted with 15 ml of PBS. Six milliliters of 1 M sodium iodide were used to elute the bound antibody, followed by three washes with 3 ml of PBS. The resulting solution was dialyzed against PBS (three changes) and reduced in volume by reverse dialysis against polyethylene glycol 20,000 molecular weight.
PBS was used to wash the column for recycling for the next application of anti-globoside. If the column were to be stored, 0.002 M sodium azide was added. Absorption of anti-N-glycolylhematoside antibody was performed by the same procedure as described above.
It is recommended, however, that anti-bovine serum albumin present in the crude antihematoside antisera be removed by bovine serum albumin at the optimal ratio of precipitation.
Forty microliters of the antisera which had been treated with bovine serum albumin and which contained 2.6 mg of protein were applied onto 2.5 ml of the glass beads column matrix linked with N-glycolylhematoside.
Immunological Assays-Activities of anti-globoside and antihematoside were titrated by hemagglutination on "microtiter plates" (Cooke Engineering Co., Alexandria, Va.) with a 1% suspension of trypsinized human erythrocytes for detection of antigloboside and with a 1% human erythrocytes passively coated with N-glycolylhematoside for detection of anti-hematoside, respectively.
Coating of hematoside on human erythrocytes was carried ou,t according to the procedure of Marcus and Cass (36). Settling patterns of erythrocjtes were read after 1 hour at 2 & 37". These agglutina'tions were inhibitable bv nloboside and hematoside butnot by other glycolipids as previoisly described (33). Reactivities of antibodies were also determined by microcomplement fixation employing reagents as described by Levine (37) and performed on "microtiter plates" according to the procedure as described by Lennette (38). Antigen glycolipids were mixed with 5 parts each of cholesterol and lecithin in chloroformmethanol (2:l).
After drying under a stream of nitrogen, the mixtures were dissolved in ethanol and 10 ~1 of the appropriate concentration were dissolved in 1 ml of hot gelatin-Verona1 buffer (37). This stock solution, 20 to 100 pg per ml, was diluted with buffer before use.
Ouchterlony gel immunodiffusion tests were performed on microscope slides with plastic templates according to the procedure of Sharples and LoGrippo (39). For detection of anti-bovine serum albumin, tanned rabbit erythrocytes coated with bovine serum albumin were used (40). Protein was determined by the method of Lowry et al. (41).

Analysis of Products
after Oxidative Ozonolysis--After ozonolysis and degradation with hydrogen peroxide, the N-acetylated methanalysis products from globoside were examined by gas-liquid chromatography and mass spectrometry.

Si(CH3)3-O-CH2-CH-CH-COOCH3
CH&O!JH bSi(CHa)a The molar mass response (flame ionization detector) of Peak A indicated a yield of 30% over-all in the degradative scheme, and correcting for the lower mass response (0.6 to 0.8) seen for aminecontaining compounds in the flame ionization detector, the yield was estimated at 35 to 45yo for the degradation product from globoside.
A 70 e.v. mass spectrum of Peak A is shown in Fig. 3. The expected molecular ion would be at m/e (mass to charge ratio) 335. Loss of methyl gave a large ion at m/e 320, and further loss of 59 from 1,2 cleavage yielded m/e 261. The ion product of 1,2 cleavage only was at m/e 230. Additional ions confirming the structure were m/e 232, from 3,4 scission, and m/e 174 from 2,3 cleavage with charge retention on the 3,4 fragment.
m/e 73 is the base peak as expected for trimethylsilyl derivatives and m/e 43 and 45 indicate the presence of an acetyl function.
Coupling of Spingolipid Derivatives to Solid Subslrate-[3H]-Globoside was prepared by the galactose oxidase-sodium borohydride method of Suzuki (32) to a specific activity of 8 X 10' cpm per mg. After coupling of the globoside derivative to Sepharose according to Fig. 1, 1.62 X lo8 cpm per ml or 21 rg per ml (16 nM) were bound.
The amount of hematoside bound to the glass bead column (Fig. 2) was determined by methanolysis of the product and comparison with an internal standard of mannitol on GLC. Four hundred ten nanomoles of N-glycolylhematoside were bound per g of the glass bead support.
The carboxyl group of sialosyl residue of hematoside attached to the column was titrated with 0.01 N sodium hydroxide under nitrogen.
The titration curve showed a sigmoid between pH 6 and 9. Absorption and El&ion of Anti-Glycolipid Antibody lo Sepharose or Glass Bead Complex-Results from the elution of a globoside-Sepharose column are shown in Fig. 4 (lop).
The passed fractions contained 25.8 mg of protein (98.5yo) and 12% of the anti-globoside activity, as demonstrated by agglutination of trypsinized human red cells (33). The fractions eluted with sodium iodide contained 0.42 mg of protein and 83yo of the anti-globoside activity.
In a control experiment with a column of aminoethyl-Sepharose, all anti-globoside activity was eluted with the breakthrough fraction. Anti-N-glycolylhematoside was purified on the N-glycolylhematoside-glass beads column in the same manner as described for anti-globoside Fig. 4  FIG. 3. The mass spectrum of a peak obtained from gas chromatography of the methanolysis products of globoside derivative. The compound was postulated-to be 2,4-dihydroxy-3-aminobutvric acid and is identified as the 2.4-di-O-trimethvlsilvl-3-acetamidomethyl ester. Recording is irom a Finnigin {uadrupole GLC-mass spectrometer model 3000. The ionizing electron energy was 70 e.v., the molecular separator and ion source temperatures were 250". The molecular weight of the expected product is 335. TMS, trimethylsilyl. FIG. 4. Top, elution pattern of anti-globoside from globoside-Sepharose column. Rabbit anti-globoside serum containing 26.2 mg of protein was applied to a 2.5 ml column of globoside-Sepharose, 5-min incubation time was allowed after application of the crude antibody, and the column was eluted with 15 ml of PBS, followed by elution with 1 M sodium iodide.
Three-milliliter fractions were collected. Activity was measured by agglutination of a 1% trypsinized human erythrocyte suspension at 6". ---, elution pattern of protein; --, hemagglutination unit per rg of protein.
The unit member is the reciprocal of the maximal dilution of antibody that causes obvious hemagglutination at 4". Protein was determined by the Lowry method.
Botlom, elution pattern of anti-N-glycolylhematoside from a hematoside-glass bead column. Column volume was 2.5 ml; antibody was allowed to incubate with column for 5 min and 3-ml fractions of PBS were collected.
Bound antibody was removed with 1 M sodium iodide. Agglutination was passive with N-glycolylhematoside absorbed on human erythrocytes. with N-glycolylhematoside (36) was increased 50 to 100 fold after purification through the hematoside-glass beads column. However, the amino-glass bead itself has asmall, but measurable degree of capacity to absorb the antibody by nonspecific interaction. The degree of such capacity was practically negligible after the glass beads were coupled to hematoside, and the anti-hematoside antibody eluted from the hematoside-glass beads column by sodium iodide had no reaction with bovine serum albumin and showed high specificity to hematoside.
SpeciJicities of Anti-Globoside and Anti-Hematoside-Since bovine serum albumin was used as a carrier of globoside to produce anti-globoside, the immune sera contained large amounts-of antibovine serum albumin.
An indication of nuritv of the affinitvretained anti-globoside was the lack of anti-bovine serum albumin in the purified fraction as shown by Ouchterlony immunodiffusion precipitation lines in Fig. 5 (top). Rabbit erythrocytes conjugated with bovine serum albumin by tannic acid (40) gave a negative reaction to agglutination by the purified antibody. Quantitative microcomplement fixation with anti-globoside detected some cross-reaction of glycolipids structurally related to globoside.
Cytolipin R, a close homolog differing only in the position of internal a-galactosyl-&galactose linkage (30), gave cross-reaction with anti-globoside, which was, under the observed indication, indistinguishable from the reaction with globoside. At very high levels of antigen, both Forssman glycolipid and aGall-+4pGall-+4pGlc-ceramide gave a low degree of cross-reaction, at least NO-fold less than globoside.
Ceramide lactoside and lacto-N-neotetraosylceramide gave a negative reaction. These results are summarized in Fig. 6 (left).
The purified anti-globoside contained a predominance of IgG and a small amount of IgM, as shown by immunoprecipitin lines with goat anti-rabbit IgG and IgM. Reactivity of the purified anti-hematoside antibody was tested by complement fixation and Ouchterlony immunodiffusion with N-acetylhematoside, N-glycolylhematoside, N-glycolyl-O-acetylhematoside,a N-glycolylhematoside methyl ester, lactosylceramide, and gangliosides of htiman brain.
Cross-reactivities for N-glycolyl-O-acetyl-and N-glycolylhematoside methyl ester but not N-acetylhematoside were demonstrated by both complement fixation (Fig. 6, right) and immunodiffusion (Fig. 5). Ceramide lactoside and other gangliosides of brain gave completely negative results.
Inhibition of passive agglutination with N-glycolylhematoside-coated erythrocytes was tested with lactosylceramide, CTH, N-acetylhematosides, mixed human brain gangliosides, and N-glycolylhematosides (CTH and lactosylceramides were mixed with brain gangliosides for solubility).
Only N-glycolylhematoside inhibited this agglutination. Ouchterlony gel immunodiffusion confirms this reactivity as shown in Fig. 5 (lower right).
Methyl ester of N-glycolylhematoside showed, however, only a faint precipitation line while it gave a strong complement fixation.
The agglutinating property of the purified anti-globoside showed an enormous temperature sensitivity, agglutinating trypsinized human erythrocytes very strongly at 4", to a lesser extent at room temperature, and very weakly at 37". Such temperature response was not observed in agglutination caused by anti-hematoside.

DISCUSSION
The results clearly show that either neutral glycosphingolipids or gangliosides can be covalently linked to solid substrata carry- Determination was performed on "microtiter plates" according to the modified procedure by Lennette (38) using sodium barbiturate 6.7 mM buffer, ~1% 7.0, containing 0.1% gelatin and 0.15 mM calcium and 0.5 mM magnesium as described by Levine (37). Ordinate, reciprocals of antibody dilution, the "dilution 1" contained 3 pg of antibody protein per well. Abscissa, reciprocals of the dilution of glycolipid antigen; the "dilution 1" contained 63 ng of glycolipid and 315 ng of each cholesterol and lecithin as auxiliary lipids (see text). The quantity of titrated complement and the amount of hemolysinsensitized sheep erythrocytes added per well was the same as described by Levine (37). Antibody and antigen controls were used in each determination.
After performing the reaction, the contents of the wells that showed inhibited hemolysis were transferred in small test tubes, centrifuged, and the optical density of the supernatant fraction was read against the content of wells that showed complete hemolysis.
Thus, a series of fixation curves can be depicted. ing amino groups.
It is expected that such a covalent complex with other macromolecules or with cellular membrane could be a useful antigen to produce antibodies against glycolipid in much more effective ways than the conventional method (33). The complex could be a useful material for studying various cell surface phenomena, e.g. to test cell surface '(trans" glycosylation as proposed by Roseman (43) and Roth (44) ; the complexes could also be useful for affinity purification of specific glycosyltransferases and hydrolases.
As a first step in a series of experiments, use of such complexes for purification of antibodies directed against globoside and hematoside is shown.
A pure antibody directed against a specific carbohydrate structure occurring on natural cell membranes could be extremely useful in various phases of studies on functions of cell membranes that are related to carbohydrate structures. Lectins have been used with increasing interest and popularity for the same purpose in membrane studies; however, anti-carbohydrate antibodies have the following advantages over lectins: (a) monovalent species (Fab fraction) can be readily prepared. Lectins have been found to bear six to eight binding sites per molecule (45, 46) but it has not been clearly demonstrated that monovalent forms can be produced5; (b) Fab fraction or even 6 Simple trypsinization or papainization of some lectins such as concanavalin A did not yield monovalent lectin (Dr. Osawa, Faculty of Pharmaceutical Science, Tokyo University, private intact antibody in the absence of complement are not cytotoxic in contrast to lectins which are often seriously cytotosic (47) ; (c) specificities of antibodies are better defined and narrower than those of lectins, and the specificities of most lcctins as deduced from the inhibition studies by monosaccharides are not always reliable (48) ; (d) the specificity of an&hematoside as shown in the results of this study, is indeed predominantly directed towards the sialic acid moiety, while lectins with such specificity have not been described so far, although the receptor site for Heliz horfensis agglutination was reported to be destroyed by neuraminidase (49).
Results of studies on the reactivities of purified antibodies directed against globoside and hematoside indicated that: (a) specificity of anti-globoside was determined predominantly by the terminal N-acetylgalactosamine and to a lesser extent by the penultimate a-galactosyl residue, as CTH reacts to a small degree. The specificity resides partially on the internal a-galacto-syll+4P-galactosyl residue, since the reactivity with cytolipin R is not identical with that of globoside in partial agreement with Rapport et al. (50). No trace of reactivity was shown with lactosylceramide and "paragloboside" excluding the possibility of participation with either further internal structure or with alternative terminal structures; (b) specificity studies of anti-A-glycolylhematoside indicated that the antibody is clearly directed to the N-glycolylneuraminosyl residue.
Interestingly, the antibody was barely reactive or not at all reactive to N-acetylneuraminosyl hematoside, alt,hough it reacted to some extent with methyl ester of N-glycolylhematoside or to N-glycolyl-O-acetylhematoside.
The specificity, therefore, is indeed very much directed toward C5-N-glycolyl residue and probably to the adjacent structure of sialic acid.
Early work by Landsteiner established the method of coupling of haptens to proteins by diazotization and was later extended by Goebel and Avery (51) and Westphal and Feier (52) to coupling of p-aminobenzoyl-and p-aminophenylglycosides to proteins.
More recently, Himmelspach and Kleinhammer (53), Ashwell (54), and RI&-oom et al. (55) have described improved and expanded procedures for coupling of carbohydrates to polymers. Arnon et al. (56) and Taketomi and Yamakawa (57) have coupled glycosphingolipids to polymers by removal of the N-acyl moiety with alkali and subsequent carbodiimide reaction or preparation of aminophenyl and aminobenzoyl compounds for diazotization.
This technique is useful only for lower oligomers of glycosphingolipids because of the sensitivity of the carbohydrate chains to alkali degradation.
The presently described procedure should be useful for higher oligomers of glycosphingolipids, including those which contain sialic acid.
Recently, a ganglioside-Sepharose complex has been used for studies of interaction of bacterial exotoxin and animal cells by Cuatrecasas (58). In this compound, the carboxyl of sialyl group in monosialoganglioside was used for coupling to Sepharose. Such a complex may be useful for limited purposes, as this structural arrangement of gangliosides on Sepharose is not analogous to that on natural cell surfaces. Since the hematoside of the complex of hematoside-glass beads described in this paper is SO arranged that the carboxyl of the sialyl residue is free to the outer environment, the molecular arrangement of such a complex is closer to that of a natural cell surface.
A number of cell surface-mediated phenomena could be uniquely studiedwith the use of purified antibodies or their "Fab" communication) while some workers claimed to derivatize concanavalin A into monovalent form by trypsinization (59).