7-O-acetyl-GD3 in human T-lymphocytes is detected by a specific T-cell-activating monoclonal antibody.

The monoclonal antibody U5, which is a potent inducer of proliferation in human T-cells, was found to bind to an alkali-sensitive derivative of ganglioside GD3. Using immunochemical and spectroscopic methods, the structure of the U5 antigen was determined as 7-O-acetyl-GD3. The antibody U5 did not react with 9-O-acetyl-GD3 and bound severalfold more stronger to 7-O-acetyl-GD3 than to GD3. U5 is the first antibody known to detect preferentially 7-O-acetyl-GD3. Flow cytometric analysis showed that each major class of human leukocytes contained a significant fraction of cells binding the U5 antibody.

Gangliosides are sialic acid-containing glycosphingolipids (GSLs) 1 consisting of an oligosaccharide chain attached to a lipid core structure. They are plasma membrane constituents of all mammalian cells. Recently, we showed that normal human leukocytes contain disialogangliosides with an 9-O-acetyl group on their terminal sialic acid (1). Not only was there a very restricted surface expression of this GSL on human blood cells (2), but it was also found to be the first surface marker for helper cells within the CD8 positive T-cell population (3). These findings suggested that slight modifications of cell surface molecules, such as O-acetylation, might suffice to define new functional subpopulations of leukocytes. This hypothesis is in accordance with observations that the pattern of glycolipids expressed on human hematopoietic cells is cell type-specific (4 -6). Our studies also indicated that O-acetylated disialogangliosides other than the 9-O-acetylated forms were present on human cells (1). During the Fifth Workshop and Conference on Human Leukocyte Differentiation Antigens (Boston, 1993), we presented evidence that a monoclonal antibody (mAb), U5, bound strongly to an alkali labile form of G D3 , which was different from 9-O-acetyl-G D3 and, furthermore, that antibodies specific for 9-O-acetyl-G D3 failed to bind to this labile G D3 derivative. This, taken together with the observation that binding of mAb U5 to human CD4 ϩ and CD8 ϩ cells induced a strong T-cell proliferation, which was accompanied by up-regulation of antigen expression (7), stimulated our interest in characterizing the structure of the U5 antigen. In this report, we describe the purification of the U5 antigen and identify it as the ganglioside 7-O-acetyl-G D3 . In addition, the distribution of the U5 antigen on human blood cells is analyzed.
Purification of the U5 Antigen, of 9-O-Acetyl-G D3 and of G D3 from Bovine Buttermilk-1460 g of buttermilk powder, obtained by freeze drying 7.5 liters of buttermilk, was suspended in 7.5 liters of chloroform/methanol (2:1) (v/v) and stirred for 1 h at ambient temperature. The suspension was filtered under reduced pressure through a Buchner funnel. The residue was reextracted twice as above. The extracts were combined, and the solvent was evaporated in vacuo at a maximum of 25°C. The evaporated lipid extract was dialyzed for 3 days at 4°C against several changes of water. The desalted extract was lyophilized and dissolved in chloroform/methanol/water (30:60:8) (v/v/v). After removal of insoluble material, the extract was pumped onto a 3.5 ϫ 20-cm column filled with DEAE-Sepharose (acetate form) in three separate runs. Elution was performed with 1 liter of chloroform/methanol/water (30:60:8), 1 liter of methanol, 1 liter of 20 mM, 2 liters of 50 mM, and finally with 1 liter of 150 mM ammonium acetate in methanol, respectively. The 50 mM ammonium acetate eluates contained all of the U5 antigen, the 9-O-acetyl-G D3 , and the G D3 , as shown by immunostaining of thin-layer chromatograms using the mAbs U5, M-T32, and R24, respectively (see below). The 50 mM ammonium acetate eluates were pooled, concentrated by evaporation, dialyzed, lyophilized, dissolved in chloroform/methanol (85:15) (v/v), and pumped onto a HPLC column (16 ϫ 500 mm) filled with LiChrosorb Si 60 5-m particles (Merck). Elution was performed using a linear gradient from chloroform/methanol/water (v/v/v) (82.6:16.4:1) to (40:50:10) in 400 min at a flow rate of 2 ml/min. Fractions were collected every 2 min. Fractions 112-128 contained 9-O-acetyl-G D3 , fractions 129 -134 contained the U5 antigen, fractions 135-140 contained a mixture of the U5 antigen and G D3 , and fractions 141-159 contained G D3 . Fractions 129 -134 were pooled and further purified on an analytical 4 ϫ 250 mm Partisil 5-m silica HPLC column (Whatman Ltd., Maidstone, United Kingdom) using a gradient from chloroform/methanol/water (v/v/v) (82.6:16.4:1) to (40:52:8) in 200 min at 1 ml/min. Fractions were collected every minute. Fractions 36 -40 contained the U5 antigen together with an impurity that showed a somewhat higher chromatographic mobility than the U5 antigen. After a final HPLC separation using the Partisil column and a gradient from chloroform/methanol/water (v/v/v) (82.6:16.4:1) to (50:45:5) in 200 min at 1 ml/min (200 fractions), apparently pure U5 antigen (about 87 g) was found in the fractions 57-65. Fractions 54 -56 and 66 -75 also * 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.
Preparation of the Disialoganglioside Fraction from Unseparated Human Leukocytes-The disialogangliosides were prepared exactly as has been described previously (1).
Quantitation-For quantitation of TLC-separated gangliosides, the HPTLC plates were sprayed with resorcinol/HCl, and then covered with a glass plate and heated at 95°C for 30 min. Densitometric measurements were made in transmission mode at 580 nm (10) using a Shimadzu dual wavelength TLC Scanner CS9001 PC (Shimadzu, Dü sseldorf, FRG). 0.5-3.0 g of G D3 was used for calibration.
TLC Immunostaining-Ganglioside antigens separated on HPTLC plates were detected by immunostaining using the method of Bethke et al. (11) with modifications (12).
Isomerization of the U5 Antigen to 9-O-Acetyl-G D3 -Dried U5 antigen in a glass test tube was incubated in 20 mM aqueous ammonia for 30 min at ambient temperature (14) followed by lyophilization. U5 antigen on HPTLC plates immobilized with polyisobutylmethacrylate (13) was isomerized by incubation of the plate in 0.1 M glycine-NaOH buffer, pH 10.0 for 2 h at 37°C (1). The plates were then washed 3 times for 5 min with phosphate-buffered saline (PBS), pH 7.3. Isomerized U5 antigen was visualized by immunostaining using mAb UM4D4 as described above.
O-Deacetylation of Gangliosides-Alkaline hydrolysis of gangliosides after separation on HPTLC plates was performed by incubating the plates for 17 h at ambient temperature in a chamber with an atmosphere saturated with 13.3 N aqueous ammonia. The ammonia-treated plates were dried for at least 3 h in vacuo in the presence of P 2 O 5 . In vitro O-deacetylation was done by treating the dried sample for 1 h at 37°C with 1 ml of 13.3 N aqueous ammonia followed by evaporation of the ammonia in vacuo at 30°C.
Release and Characterization of O-Acetylated Sialic Acids-Sialic acids were released enzymatically from gangliosides, purified, and analyzed by HPLC as described previously (15,14). To confirm the identity of 7-O-acetyl-5-N-acetylneuraminic acid, which was not easily separable from 5-N-acetylneuraminic acid, an aliquot of the released acid was treated with 20 mM aqueous ammonia as described previously (14). The expected rearrangement product 9-O-acetyl-5-N-acetylneuraminic acid was identified using an authentic standard, which was a kind gift from Dr. R. Schauer, Kiel, FRG.
Electrospray Mass Spectrometry-A Finnigan MAT TSQ 700 triple quadrupol mass spectrometer equipped with a Finnigan electrospray ion source (Finnigan MAT Corp., San Jose, CA) was used. The Oacetylated G D3 derivative was dissolved in methanol and injected at a flow rate of 2 l/min into the electrospray chamber. A voltage of 4.5 kV was applied to the electrospray needle. For collision-induced decomposition experiments, the doubly charged parent ions were selectively transmitted by the first mass analyzer and directed into the collision cell (argon was used as collision gas) with a kinetic energy set at ϩ29 eV.
ELISA-The indicated amounts of gangliosides, dissolved in 50 l of methanol, were aliquoted into a Polysorb (Nunc, Wiesbaden-Biebrich, FRG) microtiter plate. The samples (up to 30 ng) were dried at 30°C for about 3 h. Nonspecific binding sites were saturated with PBS containing 5% (w/v) bovine serum albumin for 1 h at ambient temperature. The plate was incubated with 50 l/well purified antibody (5 g/ml) overnight at 4°C. Following four washes with PBS, the plate was incubated for 2 h at 37°C with 100 l/well of an alkaline phosphatase-conjugated goat anti-mouse IgG antibody (Sigma) diluted 1:2000 in PBS containing 1% bovine serum albumin. After washing 3 times with PBS and 2 times with 0.1 M glycine-NaOH buffer, pH 10.0, containing 1 mM MgCl 2 and 1 mM ZnCl 2 , the plate was developed with 0.8 mg/ml p-nitrophenylphosphate (Sigma) in the glycine-NaOH buffer. Absorbance was read at 405 nm in an SLT Spectra III ELISA reader (SLT, Crailsheim, FRG) after 1 h.
Isolation of Leukocyte Cell Populations-CD4 ϩ and CD8 ϩ T-cells were prepared by a combination of "panning" technique and complement-mediated lysis. Flow cytometric analysis showed that CD4 ϩ and CD8 ϩ T-cells comprised 90% of the respective preparations, with not more than 5% CD16 ϩ NK cells and less than 2% CD8 ϩ /CD4 ϩ T-cells, B-cells, and monocytes.
CD3 ϩ T-cells were obtained by depletion of monocytes and B-cells followed by complement mediated lysis of CD16 ϩ NK cells.
Monocytes were purified to Ͼ90% by adherence to tissue culture dishes (Greiner, Solingen, FRG) for 60 min at 37°C.
Granulocytes were prepared from the erythrocyte layer obtained by Ficoll-Hypaque centrifugation of peripheral blood mononuclear cells. Cells were suspended in 20 ml of PBS, pH 7.2, and 5 ml of dextran 250 solution (5% (w/v) in physiological saline) were added. After 20 min of incubation at room temperature, the supernatant cells were removed and washed in 40 ml of PBS (150 ϫ g, 7 min, without brake). Remaining erythrocytes were lysed using a solution containing 0.82% NH 4 Cl, 0.1% KHCO 3 , 0.1 mM EDTA, pH 7.27. The preparations consisted of 90% granulocytes as shown by flow cytometry.
Immunoprecipitation-Purified CD3 ϩ T-cells (5 ϫ 10 7 ) were washed 3 times with 50 mM Tris-HCl, pH 8.0, containing 0.15 M NaCl and 5 mM EDTA and solubilized in 1.4 ml of this buffer containing, in addition, 0.5% Nonidet P-40, 1% phenylmethanesulfonyl fluoride, and 10.5% aprotinin. The suspension was sonicated for 2 min and incubated 15 min on ice. After centrifugation (4000 ϫ g, 4°C, 15 min) the supernatant was precleared by stirring for 30 min at 4°C with 100 l of Pansorbin pellet (Pharmacia Biotech Inc.), followed by incubation with 20 g of mAb U5/800 l of supernatant for 1 h under the same conditions. 500 l of this cell lysate/antibody mixture were incubated under rotation with 50 l of protein A-Sepharose (Pharmacia) for 3 h or overnight at 4°C. The protein A-Sepharose-antibody-antigen complex was washed 3 times with 50 mM Tris-HCl, pH 8.0, containing 0.15 M NaCl and 5 mM EDTA buffer by successive centrifugation (1250 ϫ g, 3 min) and resuspension, and then incubated with 30 l of lysis buffer (62.5 mM Tris, 2% SDS, 10% glycerol, pH 6.8) for 30 min in an ultrasonic bath. The suspension was then centrifuged (4000 ϫ g, 4°C, 15 min), and the resulting supernatant was shock frozen in liquid nitrogen.
Analysis of the Lipid Constituents of the Immunoprecipitates-The immunoprecipitates were analyzed as described previously (16).
Proliferation Assays-Separated T-cells were incubated in triplicate in microwell culture plates (Nunclon 1-67008, Nunc, Roskilde, Denmark) in the presence or absence of the anti-G D3 antibodies. The antibodies were used at final concentrations from 100 to 1.56 g/ml. Phytohaemagglutinin (HA17, Wellcome Diagnostics, Dartford, UK) was applied at 0.5 g/ml. The cultures were pulsed with 0.5 Ci/well [ 3 H]thymidine (Dupont NEN) for 18 h on day 5 of incubation, and incorporated radioactivity was measured in a Betaplate Counter (Pharmacia).

Different Proliferative Responses of Human T Cells Induced by
Binding of the mAbs R24, U5, and E11-Previous reports had suggested that mAb R24 could induce T-cell proliferation (17). When we compared two different anti-G D3 antibodies, R24 and E11, with the putative anti-G D3 mAb U5 in an effort to corroborate and extend those findings, we found that there were extreme variations in the capacities of these mAbs to induce T-cell growth as assayed by thymidine incorporation (Fig. 1). The mAbs R24 and U5 stimulated CD4 ϩ and CD8 ϩ T-cell growth without addition of exogenous cytokines, whereas E11 did not. Antibody U5 always induced higher levels of proliferation at significantly lower antibody concentrations than R24. This finding correlated with the antibody reactivity measured by flow cytometry. mAb U5 stained 30 -70% of the T-cell subpopulations; R24 stained 10 -30%. E11 bound to only 4% of the T-cells, and thus did not differ from the mouse IgG 3 subclass control antibody. The similar affinity constants of the antibodies R24 (2 ϫ 10 7 liters/mol) and U5 (1.6 ϫ 10 7 liters/ mol) for the ganglioside G D3 (18) could not explain these differences. The data thus suggested that the mAb U5 might recognize an additional ganglioside on human T-cells.
Differences in the Ganglioside Immunostaining Patterns of the mAbs E11, R24, and U5-We next compared the binding of the antibodies E11, R24, and U5 with glycosphingolipids in the disialoganglioside fraction obtained from unseparated human leukocytes (Fig. 2). mAb R24 (Fig. 2, lane A) and mAb E11 (not shown) both detected a double band with the chromatographic mobility of standard G D3 (Fig. 2, S1, the reference contained only the upper band of the doublet). mAb U5 not only recognized the same double band but recognized in addition an unknown glycolipid migrating somewhat faster (Fig. 2, lane B,  arrow). The unknown GSL was not identical to any of the antigens recognized by mAb UM4D4 (Fig. 2C), which binds strictly to 9-O-acetylated G D3 (Fig. 2, S2, only the upper band was present in the reference material) and to similar gangliosides possessing a terminal 9-O-acetylated disialosyl group (1). Our interest focused on the differences in the antigen specificities of the two mAbs, which induced T-cell proliferation, R24 and U5. As a working hypothesis, we proposed that the functional difference between the two mAbs could be ascribed to the recognition by mAb U5 of the additional GSL band indicated by an arrow in Fig. 2, lane B. We therefore undertook the isolation and characterization of this ganglioside.
Purification and Immunological Properties of the U5 Anti-gen-The overall concentration of disialogangliosides in human leukocytes is very low (about 122 g of lipid-bound sialic acid in 10 12 unseparated leukocytes (1)), making this a poor source for the purification of the U5 antigen. We therefore searched for an alternative source of the antigen and detected it in buttermilk, which has been reported to contain several disialogangliosides of the G D3 type (19). The purification of the U5 antigen from bovine buttermilk was achieved by ion-exchange chromatography and three consecutive fractionations on HPLC silica columns as described under "Experimental Procedures." Two different methods were used to identify the U5 antigen in the course of the purification. The first was direct TLC immunostaining using the mAb U5 as already shown in Fig. 2. However, because of the cross-reactivity of this antibody with the non-O-acetylated ganglioside G D3 (Fig. 2, lane B) and because the upper G D3 band and the U5 antigen migrated very close together, it was, especially in the presence of the large amounts of G D3 found in bovine buttermilk, often difficult to distinguish between these two GSLs using mAb U5. For this reason, we developed a second method to identify the U5 antigen, which took advantage of the alkali-induced (pH 10) rearrangement of the U5 antigen to 9-O-acetyl-G D3 , an antigen that could be detected with mAb UM4D4 (Fig. 3). In the left half of Fig. 3 the characteristic binding patterns of the 9-O-acetylated disialogangliosides from bovine buttermilk (lane a) and from unseparated human leukocytes (lane b) are shown. After treatment of the plate at pH 10, one new major band appeared in both lanes (Fig. 3, ϩ panel, large arrows). As shown below, this major band originated from the U5 antigen. The band could clearly be distinguished from 9-O-acetyl-G D3 because of its different chromatographic mobility. Using both methods, the U5 antigen could be distinguished with certainty from both 9-O-acetyl-G D3 (by U5 staining) and G D3 (by staining with 9-O-acetyl-G D3specific antibodies after the alkali-induced rearrangement).
The changes in the U5 antigen upon mild and strong alkali treatment in vitro are shown in Fig. 4. Reaction products were separated by TLC and analyzed by immunostaining with different antibodies and by the nonspecific detection of all GSL antigens using DIG staining (13). U5 antigen (Fig. 4A, lane 1) purified from bovine buttermilk was not detectable by the strictly 9-O-acetyl-G D3 -specific mAb (15) UM4D4 (Fig. 4B,   FIG. 1. Mitogenic effect of anti G D3 antibodies on human CD4 ؉ and CD8 ؉ T-cells. The proliferative response of CD4 ϩ and CD8 ϩ T-cells, separated as described previously (7), was measured after stimulation with mAb U5 or R24 at final concentrations between 100 and 1.56 g/ml. The S.D. ranged within 15%. The values of phytohaemagglutinin stimulation (0.5 g/ml), mAb E11 (100 g/ml), and medium control for CD4 ϩ T-cells were 152,930, 189, and 286 cpm, respectively; for CD8 ϩ T-cells 90,411, 49, and 65 cpm, respectively. lane 1) but could be detected by DIG staining (Fig. 4C, lane 1). Treatment of the U5 antigen with mild alkali (20 mM aqueous ammonia, 30 min, 22°C) (Fig. 4, lanes 2) abolished binding of mAb U5 (Fig. 4A, lane 2), but the rearranged product was now recognized by mAb UM4D4 and showed the mobility of 9-Oacetyl-G D3 (Fig. 4B, lane 2). This change could also be seen in the DIG stain (Fig. 4C, lane 2, band I). Also visible in the same lane was another band (II) of minor intensity that had the same mobility as reference ganglioside G D3 (Fig. 4C, lane 4). Treatment of the U5 antigen with strong alkali (13.3 N ammonium hydroxide, 1 h, 37°C) resulted in a single product with the mobility of G D3 (Fig. 4C, lane 3).
From these experiments it was concluded that the U5 antigen was an O-acetylated derivative of ganglioside G D3 different from 9-O-acetyl-G D3 . Because the conditions of the in vitro mild alkali treatment were the same as those used by Diaz et al. (14) to achieve an intramolecular migration of O-acetyl groups from the 7-to the 9-position of the O-acetylated sialic acid, we predicted that the U5 antigen should be identical with or closely related to 7-O-acetyl-G D3 . HPLC analysis of sialic acids released by Arthrobacter ureafaciens sialidase treatment of the purified U5 antigen showed the presence of a small peak as a shoulder eluting before the 5-N-acetylneuraminic acid main peak (Fig. 5B). This shoulder (R Neu5Ac ϭ 0.95) has been reported to be 7-O-acetyl-5-N-acetylneuraminic acid (20). Although the 7-O-acetyl-5-N-acetylneuraminic acid (Neu5,7Ac 2 ) derivative could only be partially separated from the 5-Nacetylneuraminic acid originating from the penultimate sialic acid residue (elution times in this system were 57 versus 60 min for 7-O-acetylated and unsubstituted 5-N-acetylneuraminic acid, respectively, Fig. 5B, arrow), its disappearance concomitant with the appearance of 9-O-acetylneuraminic acid (standard in Fig. 5A) after treatment with 20 mM aqueous ammonia was a further indication of the identity of the shoulder at 57 min as the 7-O-acetylated product (Fig. 5C). The fact that the peak areas of Neu5Ac (from the penultimate sialic acid residue) and of Neu5,9Ac 2 were not equal is most likely the result of some overall de-O-acetylation occurring during the induction of acetyl group migration.
Another indication of the 7 position of the O-acetyl group came from periodate oxidation experiments combined with DIG staining. This latter method is dependent on periodate oxidation of cis diol groups for the formation of the digoxigenin hydrazones, which are then manifested immunologically. Periodate attack of the non-O-acetylated ganglioside G D3 can only take place in the exocyclic glycerol-like side chain of the terminal sialic acid residue (21). The presence of an O-acetyl substitution in this side chain in either the 9 or 8 position would prevent an attack by mild periodate and subsequent DIG staining. For 7-O-acetyl-G D3 , a cleavage between the terminal (C-9) and the subterminal (C-8) carbon atoms in the sialic acid side chain could be expected with attendant detectability by DIG staining. An in situ periodate oxidation followed by DIG labeling with unsubstituted G D3 (lane 1), with the U5 antigen (lane 2), and with 9-O-acetylated G D3 (lane 3) is shown in Fig. 6. In panel A, G D3 and the U5 antigen but not the 9-O-acetyl-G D3 were oxidized by periodate as shown by DIG staining. In panels B, C, and D, control stains with the mAbs U5, UM4D4 after alkali-induced rearrangement, and UM4D4 without alkali-induced rearrangement, respectively, are shown. Thus, the detectability of the U5 antigen by DIG staining also suggested that the O-acetyl group was located in the 7 position.
Mass Spectrometric Analysis of the U5 Antigen-Structural analysis of the U5 reactive ganglioside from bovine buttermilk was also performed using negative ion electrospray mass spectrometry (Fig. 7). A cluster of six intense doubly charged molecular ions was detected at m/z 770.6, 777,4, 783.9, 791.1, 798.0, and 805.1, whereas the corresponding singly charged species were detectable but of rather low abundance (Fig. 7A). These molecular ions suggest the presence of a series of gangliosides incorporating a homogeneous carbohydrate moiety, i.e. a monoacetylated tetrasaccharide of the composition Ac-NeuAc-NeuAc-Hex 2 linked to a heterogeneous ceramide portion with C19-C24 fatty acids bound to a C18 sphingosine. These assumptions were confirmed by tandem mass spectrometric-experiments. After collision-induced decomposition of the doubly charged parent ion at m/z 791, the spectrum depicted in Fig. 7B  Only the latter signals shifted by the expected mass increment when a different parent ion was decomposed, confirming the assignments. These results confirmed the structure of the U5 antigen as a terminally O-acetylated derivative of ganglioside G D3 .
Quantitative Binding of mAbs U5, R24, and E11 to G D3 and 7-O-acetyl-G D3 -As shown above, mAb U5 did not bind to 9-Oacetyl-G D3 (Fig. 2, lane B). However, the antibody bound to some extent to G D3 . We therefore compared quantitatively binding of the mAbs U5, R24, and E11 to 7-O-acetyl-G D3 and G D3 in an ELISA assay (Fig. 8). The affinity of mAb U5 for 7-O-acetyl-G D3 was severalfold higher than that of the two other mAbs for this antigen, whereas all three mAbs had relatively low affinities for G D3 . The high binding affinity and specificity of mAb U5 for 7-O-acetylated-versus nonacetylated G D3 classifies this mAb as the first with a preferential specificity for 7-O-acetyl-G D3 .
Detection of 7-O-Acetyl-G D3 in Human T-Cells-Direct evidence for the presence of 7-O-acetyl-G D3 in human T-cells was obtained by analysis of the lipids extracted from mAb U5 immunoprecipitates of purified human T-cells (Fig. 9). Lipid extracts from total leukocytes (A) or T-cell immunoprecipitates (B) were separated on TLC plates and detected by immunostaining with mAb UM4D4 before and after alkali treatment. The band indicated by the large arrow originated from 7-Oacetyl-G D3 as inferred from the facts that it could only be detected after pH 10 treatment of the lipid extract and that it migrated between the positions of 9-O-acetyl-G D3 and 9-O-acetyl-DSPG.
Expression of U5 Positive Gangliosides by Different Human Leukocyte Populations-The distribution of the U5 antigen in different leukocyte populations as determined by flow cytometry is shown in Table I. Although our findings suggest that this antigen serves as a receptor for the functional activation of T-cells, it is not a specific marker for them, as surface expression of U5 antigen was also found in a significant fraction of the cells in all other classes of leukocytes analyzed.

DISCUSSION
In this study, we have identified the target antigen of the mAb U5 as 7-O-acetyl-G D3 and have shown that this GSL is present in the disialoganglioside fraction of human leukocytes, where it was heretofore unknown. 7-O-Acetyl-G D3 has recently been identified in bovine buttermilk and in melanoma cells of hamsters and humans (19,22,23). Its occurrence in normal human leukocytes may have been overlooked for two reasons. First, this antigen shows a migration on TLC very similar to that of unsubstituted G D3 ; second, the classical method for the detection of alkali labile GSL, a characteristic decrease in their TLC mobility upon ammonia treatment (24), failed in this case since there is essentially no difference in the mobilities of G D3 and 7-O-acetyl-G D3 .
In human leukocytes, O-acetylated sialic acid residues are ubiquitous components of disialogangliosides. We found in previous work that a majority of the disialogangliosides from human leukocytes were O-acetylated and identified the major component as 9-O-acetylated G D3 , and two minor components as 9-O-acetyl-G D3 analogs containing in addition one and two lactosamine disaccharide units (1). We also showed that treatment of the disialogangliosides from unseparated leukocytes with mild alkali caused a considerable increase in the amount of 9-O-acetylated gangliosides (1). This suggested the presence of unknown O-acetylated forms of the gangliosides that had rearranged to the 9-O-acetates during mild alkali treatment, a supposition that we have now confirmed with the identification of the U5 antigen as 7-O-acetyl-G D3 .
Theoretically, the O-acetyl group of the U5 antigen could also be located at the 8 position. However, HPLC separation of enzymatically released sialic acid showed the characteristic shoulder of the 7-O-acetylated derivative (the position of the 8-O-acetylated N-acetylneuraminic acid is not known in this HPLC system because of the extreme lability of this molecule). In addition, the U5 antigen was susceptible to mild periodate, which could only be expected for the 7-O-acetyl derivative. The presence of unsubstituted G D3 in our purified antigen could be excluded by mass spectrometry. It was not possible to quantitate the proportion of 7-O-acetylated, 9-O-acetylated, and nonacetylated forms of disialogangliosides originally present in human leukocytes or in purified T-cells since it could not be excluded that the O-acetylated gangliosides were partially deacetylated during purification. Indeed, it is conceivable that the non-O-acetylated disialogangliosides originate entirely through deacetylation during purification.
Investigations into the existence and the properties of 7-Oacetyl-G D3 have been conducted primarily in two laboratories (19,22,23). However, their results concerning the general properties of this molecule differed in several points. The first matter of controversy is the stability of the antigen. Manzi et al. (23) found that 7-O-acetyl-G D3 was an extremely labile compound with a strong tendency to rearrange to the 9-O-isomer, which is in agreement with our present results. In contrast, Ren et al. (22) reported that the 7-O-isomer could be purified from hamster melanoma cells without extensive degradation. Second, Ren et al. (19) concluded that 7-and 9-O-acetyl-G D3 were practically indistinguishable because of their very similar physicochemical properties, whereas we found differences in the chromatographic mobilities of the two isomers that were sufficient to permit their separation. Third, Ren et al. (22) reported no difference in the binding of the 9-O-acetyl-G D3specific mAb "JONES" (25) to either 7-or 9-O-acetylated forms of G D3 , whereas Manzi et al. (23) found that 9-O-acetyl-G D3specific mAbs failed to bind to the 7-O-isomer. Our data (Fig. 4) support the specificities determined by Manzi et al. (23). Furthermore, in our hands mAb JONES also failed to bind to the 7-O-isomer. 2 The reasons for these discrepancies, which could only be resolved by an exchange of materials and antibodies, are at present unknown.
Immunoprecipitates made with mAb U5 from solubilized human T-cells were shown to contain 7-O-acetyl-G D3 (Fig. 9), which offered unequivocal proof of its presence in T-cells but did not distinguish between an intracellular and a cell surface distribution. The presence of 7-O-acetyl-G D3 on the cell surface should be a rather unexpected finding in view of its lability at physiological pH. The existence of 7-O-acetyl-G D3 in acidic compartments of human melanoma cells has been well documented (23), but it was presumed that the O-acetyl group underwent a rapid migration from the 7 to the 9 position following its translocation to the cell surface. The binding of mAb U5 to intact T-cells again does not prove the existence of 7-O-acetyl-G D3 on the cell surface since this binding could equally well have been caused by cross-reaction as a result of the expression of high concentrations of G D3 .
An argument in favor of a surface expression of the U5 antigen comes through inference, from the evidence that it is involved in mediating T-cell activation. Our interest in the characterization of the U5 antigen originated from the observation (Fig. 1) that the T-cell stimulatory capacity of mAb U5 was severalfold higher than that of mAb R24, although, as noted above, both bound to ganglioside G D3 with comparable affinities. This suggested that the primary antigen recognized by U5 and responsible for T-cell activation was different from G D3 . The U5 antigen as well as non-O-acetylated G D3 have also been implicated in earlier studies as activation molecules on T-cells (7). In contrast, eight different monoclonal antibodies specific for 9-O-acetylated derivatives of G D3 , tested under the auspices of the Fifth Workshop and Conference on Human Leukocyte Differentiation Antigens (26), were found to not induce T-cell proliferation (data not shown). Further detailed functional studies with a panel of related antibodies will be necessary to clarify and confirm the roles that these different disialogangliosides may or may not play in T-cell activation.
Whether or not gangliosides are directly involved in signal transduction from the cell surface is still largely unknown. An involvement of gangliosides in signaling through direct binding of G D1a , G D1b , G Q1b , G M1 , and G M2 to calmodulin and the calmodulin-dependent enzyme cyclic nucleotide phosphodiesterase has been demonstrated (27,28). Moreover, Hannun (29) and Yuan et al. (30) have suggested roles for the GSL metabolites ceramide and sphingosine 1-phosphate in the regulation of cell growth, differentiation and apoptosis (29,30). Recently, a tight and specific association of the signal transducing GPIlinked surface molecule CD59 with the ganglioside GM3 was described (16). Thus, as a working hypothesis for future investigations, it might be speculated that 7-O-acetyl-G D3 operates in a similar manner by forming a close association in a membrane microdomain with a T-cell-activating molecule such as CD2 or CD3. FIG. 9. Presence of 7-O-acetyl-G D3 in the U5 immunoprecipitate from human T-cells. The immunoprecipitate was prepared as described under "Experimental Procedures." Gangliosides were extracted from the precipitate as described previously (16). Immunostaining was performed using mAb UM4D4 before (Ϫ) and after (ϩ) pH 10 treatment.