Demonstration of direct glycosylation of nondegradable glucosylceramide analogs in cultured cells.

After uptake by various cells (human skin fibroblasts, rat neuroblastoma B 104, human neuroblastoma SHSY5Y, murine cerebellar cells), a radioactive and a fluorescent analog of a nondegradable glucosylceramide with sulfur in the glycosidic link were glycosylated to a cell-specific pattern of glycolipid analogs. These results, for the first time, show that a glucosylceramide analog can be conveyed from the plasma membrane of cultured cells to those Golgi compartments that function in the early glycosylation steps of glycolipids. This observation is further confirmed by the fact that the cationic ionophore monensin, known to impede membrane flow from proximal to distal Golgi cisternae, inhibited the formation of complex ganglioside analogs but not those of lactosylceramide, sialyl lactosylceramide (GM3), and disialyl lactosylceramide (GD3).

After uptake by various cells (human skin fibroblasts, rat neuroblastoma B 104, human neuroblastoma SHSY5Y, murine cerebellar cells), a radioactive and a fluorescent analog of a nondegradable glucosylceramide with sulfur in the glycosidic link were glycosylated to a cell-specific pattern of glycolipid analogs. These results, for the first time, show that a glucosylceramide analog can be conveyed from the plasma membrane of cultured cells to those Golgi compartments that function in the early glycosylation steps of glycolipids. This observation is further confirmed by the fact that the cationic ionophore monensin, known to impede membrane flow from proximal to distal Golgi cisternae, inhibited the formation of complex ganglioside analogs but not those of lactosylceramide, sialyl lactosylceramide (G M3 ), and disialyl lactosylceramide (G D3 ).
On the cell surface of vertebrate cells, glycosphingolipids (GSL) 1 form cell-specific patterns which change specifically with cell differentiation, morphogenesis, and oncogenic transformation (for review see Refs. 1 and 2). Although being sur-mised to be the place for the functional role of GSL, the plasma membrane is not the site of their metabolism. Rather, GSL are synthesized in the Golgi complex by sequential addition of monosaccharide units to ceramide and are degraded in lysosomes by the sequential removal of glycosyl residues starting from the nonreducing end. Thus, the maintenance of a balanced glycolipid profile requires a stringent control of metabolism and transport of GSL (for review see Refs. 3 and 4).
We have observed labeled and cell-specific glycosylation products when a radioactive or fluorescent analog of glucosylceramide had been administered to cultured cells. These products (analogs of globosides and gangliosides) could have been formed in the Golgi complex by direct glycosylation of the labeled glucosylceramide analogs or from the labeled ceramide analogs; the latter resulting from deglucosylation, most likely in lysosomes, of the former. Therefore, we have studied the metabolism of glucosylceramide analogs that contain sulfur in the glycosidic bond ( Fig. 1) and are thus resistant to enzymatic deglucosylation. We will present data showing unambiguously that a direct glycosylation of these analogs of glucosylceramide takes place in various cultured cells leading to a cell-specific pattern of labeled glycolipid analogs.
Preparation of the Glucosylthioceramide⅐BSA Complexes-An aliquot of a stock solution of the desired glucosylthioceramide (100 nmol) in methanol was dried, first under a stream of nitrogen and then in vacuo. The dried lipid was dissolved by first adding 20 l of ethanol and then 1 ml of DMEM containing 7 mg of defatted BSA under vigorous stirring. The resulting clear solution was diluted with 9 ml of DMEM to yield a 10 M lipid-BSA complex.
Cell Culture-Monolayer cultures of human skin fibroblasts obtained from biopsy of a male infant and rat neuroblastoma cells B 104 were cultured as described (11). Human neuroblastoma-derived SHSY5Y cells were grown as described (12). All cells were grown at 37°C in a water-saturated atmosphere of 5% CO 2 in air. Cultures were seeded with 7 ϫ 10 5 fibroblasts or B 104 in 25-cm 2 tissue culture flasks. The cultures were then grown for 1 to 2 days. For human neuroblastoma cultures, 10 5 SHSY5Y cells were seeded and grown for 4 days prior to use. Murine cerebellar cells were prepared essentially as described (13) and were plated onto poly(L-lysine)-coated 35-mm-diameter Petri * This work was supported by Deutsche Forschungsgemeinschaft Grant SFB 284/B5. 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.
Incubation of Cells with Radioactive or Fluorescent Glucosylthioceramide for Biochemical Studies-Cells were washed in DMEM and incubated with the [ 14 C]C 8 -Glc-S-Cer⅐BSA complex (10 M) or the NBD-C 8 -Glc-S-Cer⅐BSA complex (10 M) in DMEM for 2 h at 37°C, washed with DMEM, and further incubated for 20 h at 37°C in DMEM containing 0.3% heat-inactivated fetal bovine serum or heat-inactivated horse serum in the case of murine cerebellar cells. For inactivation, the sera were heated for 30 min at 56°C. In cases where monensin was used, this drug was present at a 1 M concentration at all times of incubation. The incubation media were saved, and the cells were washed with PBS, harvested with a rubber policeman, and centrifuged at 2,000 ϫ g for 10 min. For protein determination (15), the pellet was suspended in 0.4 ml of H 2 O, and aliquots of 5 l were assayed. The lipids were extracted with 4 ml of chloroform/methanol (1:1, by volume) for 3 h at 38°C. The lipid extracts and the media were desalted according to Ref. 16. Total fluorescent lipids of cell extracts and media were determined by measurements of their relative fluorescence compared to a fluorescent standard containing a known amount of NBD-lipid. All measurements were made in chloroform/methanol (1:1, by volume) on a Shimadzu RF 5000 spectrophotofluorometer ( ex ϭ 475 nm; em ϭ 530 nm) (Shimadzu, Duisburg, Germany). Total radioactive lipids of both cells and media were determined by liquid scintillation counting of aliquots. The combined lipid extracts of cells and media were analyzed by TLC using chloroform/methanol/15 mM calcium chloride (60:35:8, by volume) as the developing solvent. Individual NBD-lipids separated by TLC were determined with a TLC fluorescence scanner Shimadzu CS-910 with on-line integrator C-R4A; measurements were done in the reflection mode ( ex ϭ 475 nm; em ϭ 530 nm). Individual radioactive lipids were determined using a Fuji BAS 1000 Bio Imaging analyzer (Raytest, Pforzheim, Germany). In addition, TLC plates were exposed to x-ray film (Kodak X-Omat XAR-5) and photographed on Polaroid film (type 667) under UV light for radioactive and fluorescent lipids, respectively.
Oxidation of Glucosylthioceramide Analogs to the Sulfoxides and Reduction to Starting Materials-The material of the thick band just below [ 14 C]C 8 -Glc-S-Cer (confer to x 1 of Figs. 2 and 4) was isolated from TLC plates and quantified. For the reduction with trimethylsilyl iodide according to Ref. 17, the material (about 10 nmol) was dissolved in acetonitrile and mixed with 15 nmol of trimethylsilyl chloride and 30 nmol of sodium iodide. The mixture was stirred at room temperature under argon for 20 min. TLC analysis (Fig. 6) showed that this material could be reduced to [ 14 C]C 8 -Glc-S-Cer again and therefore represented the corresponding sulfoxide. Further proof was obtained as follows: NBD-C 8 -Glc-S-Cer and [ 14 C]C 8 -Glc-S-Cer were oxidized as described for sulfur compounds (18) with modifications. Briefly, 200 nmol of the thiolipids were oxidized overnight at 38°C in 120 l of 0.1 M methanolic hydrogen peroxide. The oxidized products were characterized by TLC (Fig. 6) and FAB-MS.
Identification of Labeled Metabolites-Glycosylation products were scraped from TLC plates and analyzed as follows. The radioactive analog of globotriaosylceramide from fibroblast extract was characterized by FAB-MS and identified by its degradation to the radioactive analog of lactosylceramide using ␣-galactosidase (6). The G M3 and Lac-Cer analogs of B 104 cell extracts were characterized by FAB-MS. Other glycolipid analogs of total lipid extracts were degraded by two subse-quent actions of glycohydrolases (fibroblasts: ␤-hexosaminidases and ␣-galactosidase, SHSY5Y cells: ␤-hexosaminidases and sialidase, and murine cerebellar cells: sialidase and G M1 -␤-galactosidase) as outlined in Fig. 4. Briefly, extracted lipids were dissolved in a total volume of 330 l of 85 mM sodium citrate buffer, pH 4.3, containing 2 mM sodium taurodeoxycholate in addition to the respective enzymes. For ␤-hexosaminidases, sialidase, or ␣and ␤-galactosidase, 17,000, 1,300, 185, or 17 picokatals, respectively, were used. To assays containing ␤-hexosaminidases G M2 activator protein (1 unit according to Ref. 19) was also added. After 10 h at 37°C, each assay was split and one-half was treated further with the second enzyme as outlined in Fig. 4. The lipid extract of B 104 cells was treated with sialidase as described above for murine cerebellar cells.
Testing of the Incubation Media for Glycosyltransferase Activities-In a total assay volume of 100 l of DMEM, pH 7.4, containing 50 nmol each of CMP-NeuAc, UDP-Gal, UDP-GalNAc, as well as 6 l of heatinactivated fetal bovine serum or horse serum, 5 nmol of [ 14 C]C 8 -Glc-S-Cer was incubated for 22 h at 37°C. After the addition of methanol (50 l), salts and nucleotide sugars were removed according to Ref. 16. The lipids were analyzed by TLC (Fig. 3).
Miscellaneous Procedures-Fast atom bombardment mass spectrometry (FAB-MS) was carried out as described (20) using a ZAB HF instrument (VG Analytical, Manchester, UK). Protein was quantified as described (15) using BSA as standard.

Uptake of Labeled Glucosylthioceramides by Cultured
Cells-We first examined the uptake of [ 14 C]C 8 -Glc-S-Cer and NBD-C 8 -Glc-S-Cer using equal concentrations of radioactive lipid⅐BSA and fluorescent lipid⅐BSA complexes. All cell types were incubated for 2 h at 37°C with a 10 M concentration of either complex, washed, and further incubated for 20 h in DMEM devoid of the labeled lipid⅐BSA complexes, but containing 0.3% heat-inactivated fetal bovine serum or heat-inactivated horse serum in the case of murine cerebellar cells. During the 20-h incubation, some protein(s) of the sera extracted a considerable percentage of the glucosylthioceramide analogs and their metabolites from the plasma membrane of cells. Therefore, the labeled lipids were isolated from the incubation media, quantified, and combined with the lipid extract of the corresponding cells. The amount of cell-associated radiolabeled and fluorescent lipids is shown in Table I. Except for fibroblasts, similar values were obtained for both labeled lipids demonstrating that the fluorescent label had no drastic influence on the uptake rate. There was, however, a much higher percentage of fluorescent glycolipids than radioactive glycolipids that were extracted into the incubation medium. During the 20-h incubation, roughly 25% and 50% of the radiocarbon and fluorescent-labeled lipids, respectively, were extracted from the various cells.
Glycosylation of Radioactive Glucosylthioceramide-We next examined the possible metabolism of the radioactive glucosylceramide analog in four different cell types. Fig. 2 shows that  each cell type displayed a different glycosylation pattern. From comparison of the product bands to labeled glycolipid analogs with similar structures (reference lane) and from metabolic labeling of the endogenous glycolipid pattern of each cell type, it became immediately obvious that the undegradable radioactive glucosylceramide analog had been glycosylated to yield a cell-type specific pattern of glycolipid analogs. Thus, in fibroblasts, obviously those of globosides and G M3 were predominantly formed (Fig. 2, lane 2). In rat neuroblastoma B 104 cells, high amounts of the analog of G M3 and smaller amounts of lactosylceramide were produced (Fig. 2, lane 4), whereas in murine cerebellar cells, labeled ganglioside analogs of G M3 , G D3 , G M2 , G M1 , G D1a , G D1b , and G T1b were present (Fig. 2, lane  8). These data agreed perfectly with the endogenous pattern obtained by metabolic labeling of the glycolipids of these cells (14). In human neuroblastoma SHSY5Y cells, the analogs of G M2 and G D2 were the prevalent glycosylation products formed (Fig. 2, lane 6). This result again complies with the endogenous glycolipid pattern (12). We could also observe that the glycosylation products found in the incubation media were the same as those found in the respective cell pellets. This is demonstrated for fibroblasts in Fig. 3 (lanes 3 and 4). These patterns comply well with that of Fig. 2 (lane 2). In addition, we have shown that the incubation media containing 0.3% heat-inactivated fetal bovine serum were devoid of any detectable glycosyltransferase activity (Fig. 3, lane 2). The same holds for heat-inactivated horse serum (data not shown).
Identification of Glycosylation Products-The structures of some glycosylation products, i.e. globotriaosylceramide analog from fibroblasts and lactosylceramide, as well as G M3 analogs from rat neuroblastoma B 104 cells, have been confirmed by FAB-MS. The structures of the other glycolipid analogs have been corroborated by sequential enzymatic degradation (Fig.  4). Thus, treatment of the glycosylation products of SHSY5Y cells with hexosaminidases resulted in a shift in both main bands to bands of analogs of G D3 and G M3 (Fig. 4, lane 6). Subsequent sialidase treatment further shifted these bands to labeled lactosylceramide analog with the same property as an authentic synthesized lactosylceramide analog (reference lane), whereas the G M1 analog remained undegraded (Fig. 4,  lane 7). Similarly, sialidase treatment of glycosylation products of murine cerebellar cells led to an increase in the analogs of G M1 and lactosylceramide (Fig. 4, lane 8) which on subsequent ␤-galactosidase treatment were converted mostly to the G M2 analog and glucosylthioceramide, respectively (Fig. 4, lane 9). When human fibroblasts were treated with hexosaminidases, the bands of G M3 and globotriaosylceramide analogs remained, whereas the bands of G M2 and globotetraosylceramide analogs diminished (Fig. 4, lane 2). Subsequent hydrolysis with ␣-galactosidase (Fig. 4, lane 3) converted the latter only slightly into the corresponding lactosylceramide. Therefore, the putative globotriaosylceramide analog was isolated from the TLC plate and treated with ␣-galactosidase again. Fig. 5 demonstrates that about 50% of this lipid was hydrolyzed to lactosylceramide analog, confirming the data obtained by FAB-MS. Sialidase treatment of the lipid extract of B 104 cells diminished the analog of G M3 while increasing the corresponding lactosylceramide. These results proved the structures of the glycosylation products of [ 14 C]C 8 -Glc-S-Cer.
The thick band just below glucosylthioceramide in all lipid extracts (denoted by x 1 in Figs. 2, 3, and 4) is formed predominantly in the process of cell incubation. After isolation from the TLC plate, its structure was shown by FAB-MS to be the oxidation product, i.e. the sulfoxide of [ 14 C]C 8 -Glc-S-Cer. When treated with trimethylsilyl iodide according to Ref. 17, the sulfoxide was reduced to [ 14 C]C 8 -Glc-S-Cer (Fig. 6, lane 3). This sulfoxide could also be obtained by treating glucosylthioceramide with methanolic hydrogen peroxide. Under this condition, the corresponding sulfone is also produced. Both products could be separated by TLC (Fig. 6, lane 4 and 5), and their structure proven by FAB-MS. The faint bands seen between [ 14 C]C 8 -Glc-S-Cer and its sulfoxide in Fig. 4 (denoted by y in lanes 6 and 7) may well represent this sulfone. The sulfoxide of [ 14 C]C 8 -Glc-S-Cer also seems to be prone to glycosylation as most clearly demonstrated by the prominent band just below the G M3 analog for the B 104 cell extract (denoted by x 4 in Fig. 2, lane 4).

Effect of Monensin on the Glycosylation of [ 14 C]C 8 -Glc-S-
Cer-The fact that [ 14 C]C 8 -Glc-S-Cer, after incorporation into various cells, became glycosylated to yield a cell-type specific pattern of glycolipid analogs strongly suggested that this exogenous glucosylthioceramide enters the glycolipid biosynthetic pathway exactly as for endogenous glucosylceramide. To further substantiate this notion, we have examined the effect of the ionophore monensin on the glycosylation pattern of [ 14 C]C 8 -Glc-S-Cer. It has been shown before that this drug impedes vesicular membrane flow between the proximal and distal regions of the Golgi apparatus (21), thus dissecting early and late glycosylation steps in glycolipid biosynthesis (14,22). Indeed, when the incorporation and incubation was performed in the presence of monensin a drastic simplification of the labeled lipid pattern was observed (Fig. 2, lanes 3, 7, and 9). The complex glycolipids in contrast to G M3 , G D3 , and lactosylceramide analogs were highly reduced. Glycosylation of Fluorescent Glucosylthioceramide-When cells were treated with NBD-C 8 -Glc-S-Cer the degree of its glycosylation was much less than that for [ 14 C]C 8 -Glc-S-Cer (Table II). The patterns of the fluorescent glycosylation products that were found in the cell pellets and the respective incubation media were, however, almost the same as for the radioactive ones. In fibroblasts, however, more fluorescent G M3 than globotriaosylceramide analog was synthesized (Fig. 7,  lane 2). Again, in the presence of monensin, the patterns were changed. This is best to be observed for SHSY5Y cells (Fig. 7,  lanes 6 and 7) and murine cerebellar cells (Fig. 7, lanes 8 and  9). For both cell types, no fluorescent ganglioside analogs but for G M3 were produced in detectable amounts in the presence of this drug.
The band closest to the fluorescent glucosylthioceramide (x 1 in Fig. 7) has been identified by FAB-MS and chemical reduction as the corresponding sulfoxide.

DISCUSSION
In this paper we present, for the first time, an unambiguous proof for direct glycosylation of glucosylceramide analogs in various cultured cell types. In human fibroblasts, rat neuroblastoma B 104 cells, and murine cerebellar cells, the glycosylation pattern obtained was identical with the pattern that is obtained by metabolic labeling of these cells (14). Human neuroblastoma SHSY5Y cells also yielded a glycosylation pattern which completely agreed with the endogenous glycolipid pattern (12). The amount of the fluorescent glycosylation products was, however, much less than that of the radioactive anabolites (Table II) indicating that the fluorescent tag somehow inter- FIG. 4. Enzymatic degradation of cell lipid extracts with glycohydrolases. Lipid extracts obtained as for Fig. 2 were treated with the enzymes as described under "Experimental Procedures." For a second degradation step, one-half of the assay mixture was treated with a second enzyme for an additional 10 h. Thereafter, the degradation products were separated by TLC and visualized as for Fig. 2. In the right-hand margin, the mobilities of [ 14 C]C 8 -Glc-S-Cer and of its glycosylation products are denoted by the standard abbreviation for GSL as outlined in Footnote 1. feres with glycosylation and/or transport. This is most obvious for fibroblasts when comparing the patterns of labeled analogs showing that more fluorescent G M3 than fluorescent globotriaosylceramide was formed (compare Fig. 7, lane 2, to Fig. 2,  lane 2). In contrast, more of the radioactive analog of globotriaosylceramide than that of G M3 was synthesized, thus exactly mimicking the endogenous glycolipid pattern. The NBD group may render the fluorescent analogs more polar. This may explain the high degree of fluorescent lipids extracted into the incubation medium and may also explain why less fluorescent than radioactive glycosylation products have been produced.
Owing to the hydrophilic glucosyl head group, no spontaneous diffusion across the plasma membrane can occur, and the internalization of the labeled analogs at 37°C very likely takes place by vesicular membrane flow. Hence we have to assume that vesicles carrying these glucosylceramide analogs eventually fuse with Golgi cisternae (or other compartments active in glycosylation) that contain the enzymes for lactosylceramide, G M3 , and G D3 synthesis. We do not yet know from where these vesicles are derived and if endosomes and/or lysosomes are intermediate stations in this transport. It has been shown clearly, however, that these glucosylceramide derivatives participate in the biosynthetic pathway of the respective cellspecific glycosphingolipids.
For glycosphingolipids, the Golgi apparatus is the well accepted site for their biosynthesis. It is believed that lactosylceramide and gangliosides G M3 and G D3 are synthesized presumably in proximal Golgi cisternae, and that the more complex glycosphingolipids are synthesized in distal Golgi membranes (23). The membrane flow from proximal to distal Golgi membranes is inhibited by monensin (for review, see Ref. 24). Thus, from our experiments with monensin, we infer that the glucosylthioceramides were conveyed to Golgi cisternae that function in the early glycosylation steps in glycolipid biosynthesis, i.e. the formation of lactosylceramide and gangliosides G M3 and G D3 . The formation of analogs of lactosylceramide and G M3 was not impaired by monensin whereas the production of the complex glycosylation products was severely decreased.
Our findings lend support to observations made by others who found indication for a glycosylation of glucosylceramide in cultured fibroblasts derived from patients with Gaucher disease (25). It has to be kept in mind, however, that these fibroblasts may contain enough residual glucocerebrosidase activity (26) to degrade an appreciable amount of the exogenously supplied glucosylceramide and that its labeled degradation product may have been used for de novo synthesis of glucosylceramide and more complex glycolipids. Also, the mode of uptake of the exogenously supplied glucosylceramide via liposomes may have directed this lipid to cellular compartments active in another nonlysosomal glucocerebrosidase that is not deficient in Gaucher disease (27) or active in unspecific ␤-glucosidases that fall within the normal range of activity in patients with all forms of Gaucher disease (28).
An important question is whether native glucosylceramide and perhaps other glycosphingolipid molecules with long acyl chains would participate in the transport process we have observed for the labeled glucosylthioceramides. If so, would this have implications in terms of regulation of glycolipid biosynthesis? To address this question, we intend to perform studies employing undegradable analogs of glucosylceramide and other glycosphingolipids carrying long acyl chains in their ceramide portion. Even if the short chain glucosylceramide analogs should not mimic the endocytotic pathway of the endogenous glycolipids, they are valuable tools for revealing membrane transport that otherwise may be overlooked and that may be important for cellular events. Further experiments are needed to clarify this point.