Brefeldin A induced inhibition of de novo globo- and neolacto-series glycolipid core chain biosynthesis in human cells. Evidence for an effect on beta 1-->4galactosyltransferase activity.

De novo synthesis of neolacto-series glycolipids has been studied in human cell lines via metabolic labeling of ceramide with [3H]serine. Intense labeling of ceramide mono- and dihexoside glycolipids occurred with labeling of progressively longer chain derivatives with increasing time. Most of the label was recovered in neutral glycolipids with about 5% of the total labeling in the ganglioside fraction. Experiments done using cell treatment with 2.5 micrograms/ml brefeldin A resulted in a stimulation in the total amount of labeling, accumulation of a neutral glycolipid identified as Lc3 due to inhibited transfer of the neolacto-series core chain terminal beta-Gal residue, and a corresponding inhibition of labeling of longer chain neutral glycolipids in all cell lines. Brefeldin A also blocked synthesis of the globo-series precursor, Gb3, longer chain sialylated structures such as IV3NeuAcnLc4, but not de novo GM3 synthesis. Brefeldin A treatment had no effect on cellular beta 1-->3N-acetylglucosaminyl-, beta 1-->4galactosyl-, or alpha 1-->3fucosyltransferase specific activities, nor was it inhibitory in beta 1-->4galactosyltransferase assays in vitro. The results describe brefeldin A-induced blocks in globo- and neolacto-series glycolipid biosynthesis, consistent with differential localization of enzymes in intracellular membranes. In particular, the results suggest that the beta 1-->4galactosyltransferase in these cells is either not redistributed by brefeldin A or is otherwise rendered nonfunctional.

* This investigation was supported by Research Career Development Award KO4 CA01343, Grant CA41521 from the National Cancer Institute-National Institutes of Health, and funds provided by the Quest for Truth Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $ T o whom correspondence and reprint requests should be addressed The Pacific Northwest Research Foundation, 720 Broadway, Seattle, WA 98122.
The abbreviations used are: ER, endoplasmic reticulum; PBS, phosphate-buffered saline, 8.1 mM Na2HP04, 1.5 mM KH2P04, 137 mM NaCI, 2.7 mM KCl, pH 7.4; BFA, brefeldin A; HPTLC, high performance thin layer chromatography; glycolipids are designated according to the recommendations of the IUPAC Nomenclature Committee, but the suffix OseCer is omitted (39). The major glycolipid structures are lactosylceramide, Gal@l4Glc@l+lCer; Lc3, G l c N A c @ l + 3 G a l~l 4 G 1 c~l~l C e r ; nLc4, Gal@l+4GlcNAc@l+ 3Gal@l4Glc@1+1Cer; nLcs, GlcNAc@l+3Gal~l4G1cNAc~l+ 3Gal@l4Glc@l+lCer; II13FucnLc4, Gal~14(Fucal+3)GlcNAc@1 +3Gal@l+4Glc@l-+lCer; IV2FucnLc,, Fuca1+2Gal@1+4GlcNAc~1 membranes (1)(2)(3)(4). This has also been demonstrated for enzymes associated with glycolipid biosynthesis. Recent evidence reported for ganglioside biosynthesis in rat liver indicated that many glycosyltransferases are localized in sub-Golgi fractions in the order in which they act (5, 6). Similar results were obtained in density gradient studies of enzymes involved in lacto-series type 1 and 2 glycolipid biosynthesis in Colo 205 and SW403 cells (7). These studies suggest that a considerable degree of regulation of glycosyltransferase function in the cell occurs by virtue of their membrane localization. Thus, cellular regulation of both enzyme activity and the partitioning of common precursors between enzymes that otherwise would compete randomly for these intermediates bears a great impact on the nature of cell surface carbohydrate antigens expressed. One example of this influence is the observation of almost exclusive expression of lacto-series type 1 chain-based antigens in Colo 205 cells. These cells also express high levels of enzyme activities capable of type 2 chain-based antigen expression (7, 8). Alterations in these cellular properties may have significant impact on expression of, for example, tumor-associated antigens.
Brefeldin A (BFA) has been reported to block protein transport from the ER to Golgi and cause a redistribution of the cts/medial/truns-Golgi membranes (9-12), but not the trans-Golgi network (13), to an intermediate compartment which utilizes a microtubule-dependent pathway to the ER. These studies have relied largely on ultrastructural analysis. BFA action has also been extended to biochemical effects such as synthesis of oligosaccharide moieties of ganglio-series glycolipids (14,15) and N-linked oligosaccharides (16). Brefeldin A was found to inhibit conversion of de novo synthesized GM, to GMZ, indicating that GM2 synthetase localizes to a compartment functionally distinct from the ER which is late in the intracellular trafficking pathway (14,15). Its effect on N-linked oligosaccharides gave rise to incomplete complextype chains with a high degree of terminal GlcNAc residues caused by diminished amounts of P-Gal substitutions. This has a further effect on other terminal modifications such as fucosylation, sialylation, or transfer of terminal a-Gal residues (16).
This suggests that part of the Pl+ 4galactosyltransferase activity remains functionally distinct from the ER (or from enzymes catalyzing the preceding reactions of neolacto core chain synthesis) as a result of BFA treatment.

Methods
Growth of Cells-Cell lines NCI-H69, HCT-15, PC9, and U937 were grown in RPMI 1640 medium supplemented with 10% fetal calf serum. HL-60 cells were grown in RPMI medium supplemented with 20% fetal calf serum. The cells were harvested and passed every 6 to 10 days.
PHJSerine Metabolic Labeling of Cultured Cells-Nonadherent lines (NCI-H69, HL-60, PC9, and U937), 1-2-ml packed cell volume, were harvested by centrifugation, washed once with serine-free Dulbecco's modified Eagle's medium, followed by incubation with fresh serine-free Dulbecco's modified Eagle's medium for 1 h at 37 'C. The cells were again centrifuged and placed in 130 ml of serine-free Dulbecco's modified Eagle's medium containing 150 pCi of [3H]serine in a tissue culture flask. Twenty-ml aliquots were collected at 2-, 4-, 8-, and 12-h intervals. At 12 h, the remaining cells in each flask were harvested, washed once with RPMI 1640, and replaced in 100 ml of RPMI with no [3H]serine label. Fifty ml was collected for the 24-and 48-h time points. Pellets harvested at each of the time intervals (200-400 pl of packed cells) were washed once with PBS, suspended in 5 ml of isopropyl alcoho1:hexane:water (55: 25:20) and stored at -20 "C.
Confluent 15-cm plates of the adherent HCT-15 cell line were subjected to the same pre-and postlabeling conditions as the nonadherent lines. The cells were [3H]serine-labeled in 20 ml of medium containing 1 pCi of [3H]serine per ml of medium. Two plates were harvested per time point. Pellets were stored in isopropyl alcohol:hexane:water (55:25:20) at -20 "C.
In some experiments, labeling was also conducted in the presence of medium containing 2.5 pg/ml brefeldin A (BFA). In these experiments, the cells (2 to 4 ml packed cell volume) were incubated with the serine-free, BFA-containing medium for 3 h prior to the addition of the [3H]serine label. After labeling, the cells were harvested as before.
In order to verify the distribution of the 3H label in the glycolipid fraction, total glycolipids were hydrolyzed with endoglycoceramidase, and the recovery of 3H in the oligosaccharide and ceramide portions was determined. In all cases under the conditions used in these experiments, no detectable label was found in oligosaccharide. Rather, only 3H labeling of ceramide occurred.
Enzyme Assays /314Galactosyltransferase-~l4Galactosyltransferase activity was determined in reaction mixtures containing 2.5 pmol of HEPES buffer, pH 7.0, 30 pg of LcB, 100 pg of taurodeoxycholate, 1 pmol of MnC12, 15 nmol of UDP-["Clgalactose (15,000 cpm/nmol), 0.5 pmol of galactonolactone, and 0.1 mg of protein in a total volume of 0.1 ml. The reaction was conducted for 1 h at 37 "C and stopped by the addition of 6 pmol of EDTA and 100 pl of CHC13:CH30H (2:l). The entire reaction mixture was streaked onto a 4-cm wide strip of Whatman No. 3MM paper and chromatographed with water overnight. The labeled product remaining at the origin was quantitated in a liquid scintillation counter. One unit of activity is defined as transfer of 1 nmol of galactose per h under the conditions of the assay.
al4/4Fucosyltransferme-The fucosyltransferase activity was determined in reaction mixtures containing 2.5 pmol of HEPES buffer, pH 7.2, 30 pg of nLcr, 100 pg of taurodeoxycholate, 1 pmol of MnCIZ, 0.5 pmol of CDP-choline, 15 nmol of GDP-['4C]fucose (15,000 cpm/nmol), and 15-150 pg of protein in a total volume of 0.1 ml. The reaction mixture was incubated for 2 h at 37 "C, terminated, and quantitated as described above. One unit of activity is defined as the transfer of 1 nmol of fucose per h under the conditions of the assay.
pl4N-Acetylglucosaminyltransferase-Assays of N-acetylglucosaminyltransferase activity were performed in reaction mixtures containing 2.5 pmol of HEPES buffer, pH 7.2,40 pg of lactosylceramide, 150 pg of Triton CF-54,0.5 pmol of MnCl2, 0.5 pmol of CDP-choline, 50 nmol of UDP-['4C]N-acetylglucosamine (5000 cpm/nmol), and 200-400 pg of protein in a total volume of 0.05 ml. The reaction mixture was incubated for 2 h at 37 "C, terminated, and quantitated as described above. One unit of activity is defined as the transfer of 1 pmol of GlcNAc per h under the conditions of the assay. Fractionation of Cells Cells, 1 to 2 ml packed volume, with and without BFA treatment and after metabolic labeling with [3H]serine as above were fractionated for glycolipid and glycosyltransferase analysis of Golgi-derived membranes. All steps were conducted at 0-4 "C. An aliquot of each 1-to 2-ml cell pellet was reserved for analysis of whole cells. The cells were suspended in 40 ml of a relaxation buffer composed of 10 mM HEPES buffer, pH 7.2, 100 mM KCl, 3 mM NaC1, 1 mM ATP, 3.5 mM MgClz and disrupted at 500 p.s.i. for 20 min in a nitrogen cavitation bomb (Parr Instrument Co., Moline, IL). The disrupted cells were centrifuged at 3000 X g for 10 min, and the resulting supernatant fraction was centrifuged at 27,000 X g for 30 min. The resulting pellet was suspended in 25 ml of a buffer composed of 50 mM HEPES buffer, pH 7.2,l mM EDTA, 0.5 M sucrose by one stroke of a Potter-Elvehjem homogenizer and layered onto a 15-ml cushion composed of 50 mM HEPES buffer, pH 7.2, 1 mM EDTA, 1.2 M sucrose. This was then centrifuged at 94,000 X g for 90 min. The Golgi membrane-enriched fraction was isolated at the interface of the 0.5 and 1.2 M sucrose layers. This fraction was diluted with 5 volumes of water and isolated by centrifugation at 27,000 X g for 30 min. Aliquots of each fraction were used for assay of enzyme activity and glycolipid extraction and analysis.

Extraction of Glycolipids from Cells and Cell Subfractions
Glycolipids were isolated from [3H]serine-labeled packed cells or cell subfractions by extraction with 10 ml of isopropyl alco-h0l:hexane:water (55: 25:20). The cells were sonicated in a bath sonicator for 10 min, followed by centrifugation at 1500 X g for 10 min. The insoluble pellet was re-extracted with 10 ml of the same solvent followed by centrifugation. The combined supernatant fractions were concentrated to near dryness and transferred to Spectrapor 3 membrane tubing (Spectrum Medical Industries, Los Angeles, CA) and dialyzed extensively against water. The solution was removed from the dialysis bag, concentrated to dryness, and acetylated with 2 ml of pyridine and 1 ml of acetic anhydride. The acetylated glycolipid fraction was obtained by chromatography on a Florisil column (23) and deacetylated with NaOMe. The deacetylated glycolipid fraction, now depleted of labeled phosphatidylserine, was dissolved in a solvent composed of CHC13:CH30H:H20 (30608) and subjected to chromatography on DEAE-Sephadex A-25 according to the method of Yu and Ledeen (24) to separate neutral glycolipids from gangliosides. The resultant total neutral glycolipid and ganglioside fractions after extensive dialysis with water were dissolved in CHC13:CH30H (2:l). An amount of glycolipid corresponding to 2 mg of dried cell residue was spotted for TLC or used for solid phase radioimmunoassays as indicated.

TLC Zmmunostaining of Glycolipids
Immunostaining of glycolipids, separated on HPTLC plates, was performed using the procedure of Magnani et al. (25) as modified by Kannagi et al. (20). Glycolipids were separated on an HPTLC plate (J. T. Baker Co.) using solvent systems composed of CHCI3:CH30H:H20 (563810) or CHCI3:CH30H:H20 (60409) containing 0.02% CaC12-2H20. After development, the plate was dried and soaked for 2 h in 5% bovine serum albumin in PBS to block nonspecific antibody binding. The plate was then incubated in culture supernatants of the derived monoclonal antibodies overnight, followed by sequential incubations with 1:lOOO diluted rabbit anti-mouse Ig antibody solution and with 9-Protein A solution. After extensive washes with PBS between each step and after '251-Protein A treatment, the plate was dried, and labeled bands were detected by autoradiography.

Solid Phase Binding Assays
Glycolipids were deposited on 96-well vinyl plates in solutions containing 3 pg of cholesterol, 5 pg of phosphatidylcholine, and total glycolipid corresponding to a known amount of dried cell residue per ml of absolute ethanol. The glycolipids were serially diluted in ethanol containing cholesterol and phosphatidylcholine alone. Aliquots of 50 pl each were dispensed into each well and allowed to air-dry. The plates were blocked with PBS containing 5% bovine serum albumin for 2 h followed by incubation with antibody-containing culture supernatant for 18 h. The plates were washed extensively with PBS followed by incubation with 1:500 diluted rabbit anti-mouse Ig (ICN Immunobiologicals) for 1 h. The plates were again extensively washed with PBS and incubated with 1251-Protein A (90,000 cpm/well) for 1 FIG. 1. Thin layer chromatographic analysis of time-dependent ['Hlserine labeling of neutral glycolipids extracted from cultured cells. Total neutral glycolipids were labeled and extracted as described under "Experimental Procedures." Lane 1, total neutral glycolipids revealed by orcinol spray; lanes 2-7,   ; panel E, U937. An amount of glycolipid corresponding to 2 mg of dried extracted cell residue was spotted for TLC in each case. The HPTLC plates were developed in a solvent composed of CHC13:CH30HH20, 60409, containing 0.2% CaC12.2H20. Fluorography of the 3H-labeled glycolipids was conducted after spraying the plate with EN3HANCE (Du Pont-New England Nuclear). The TLC mobility of the following glycolipid standards is indicated by GlcCer ( a ) , lactosylceramide (b), and n k ( c ) . A h. The plates were washed again with PBS, and the amount of in each well was determined in a y counter.

RESULTS
Glycolipid biosynthesis proceeds via a stepwise elongation of oligosaccharide chains linked to ceramide. The nature of the products produced is a function of the glycosyltransferases expressed in a given cell which are distributed in an orderly fashion in the ER and Golgi membranes. In order to study de mu0 synthesis of neolacto-series glycolipid chains in expressing cells, metabolic labeling with [3H]serine was conducted taking advantage of the [3H]serine-specific labeling of the ceramide moiety.
Time Course of [3HJSerine Metabolic Labeling of Cultured Cells-Uptake of [3H]serine into neutral glycolipids and gangliosides was studied using labeling times varying from 2 to 12 h, and the fate of the label was followed up to 36 h later as described under "Experimental Procedures." Fig. 1 shows the time-dependent [3H]serine labeling profile of neutral glycolipids isolated from a variety of transformed human cell lines, each of which shares the common property of expressing neolacto-series glycolipids. Intense labeling of glucosyl and galactosyl derivatives of ceramide is observed with each cell line, even at relatively short labeling times. In general, a progressive and stepwise labeling of slower migrating, longer chain glycolipids occurs for all cell lines tested as the [3H]serine labeling time is increased. After 12 h, the 3H label was removed, and the fate of the labeled glycolipids was followed for an additional 36 h. In most cases, the extent of labeling of the longer chain glycolipids increased further during this time, suggesting a relatively slow turnover of the oligosaccharide chains. This was not the case for HCT-15 cells, where labeling of longer chain structures was more rapidly depleted after removal of the [3H]serine.
In contrast to the relatively efficient labeling of neutral glycolipids by [3H]serine, only weak labeling of the ganglioside fraction was observed. The extent of ganglioside labeling was approximately 5% of that found for neutral glycolipids for all cell lines tested. Most of this label was present in GM3, although labeling of bands co-migrating with IV3NeuAcnLc4 was observed, particularly for NCI-H69 cells (see below and Fig. 5). Despite the weak labeling by [3H]serine, each cell line was found to express a significant chemical quantity of ganglioside structures, suggesting a slow turnover in the cell of these components.
Effect of Brefeldin A on Glycolipid Synthesis-The disruption of normal Golgi function induced by BFA provides a means to evaluate the stepwise biosynthesis of neolacto-series antigens in Golgi membranes. Fig. 2 shows the effect of BFA on the [3H]serine labeling profiles of neutral glycolipids using labeling times established above to yield significant labeling of slower migrating glycolipids (generally 12 or 24 h, see Fig.  1). As previously observed, de mu0 synthesis of 3H-labeled glycolipids corresponding to total cellular glycolipids, as revealed by orcinol staining, occurred in the absence of BFA for each cell line. Parallel cells treated with 2.5 pg/ml BFA, as described under "Experimental Procedures," yielded remarkably different results. For each cell line, BFA-treated cells were characterized by relatively weak or absent 3H labeling of slower migrating glycolipids, as compared to untreated cells, along with the appearance of a strongly 3H-labeled band migrating as a ceramide-trisaccharide. This band was uniformly accumulated in all cell lines. Despite the de mu0 expression of this glycolipid as a consequence of short term BFA treatment (ie. during the [3H]serine labeling period prior to cell harvest), a significant chemical quantity was accumulated, as indicated by orcinol spray (see, in particular, results for NCI-H69 and HCT-15 cells). TLC immunostaining of neutral glycolipids from untreated and BFA-treated cells indicated strong staining of the BFA-induced ceramide-trisaccharide band with the anti-Lc3 TE5 antibody. These results also revealed that the L C 3 component in most cell lines was undetectable in glycolipids from untreated cells. In two cases where detectable amounts occurred (PC9 and HL-60 cells), the staining intensity was greatly stimulated after BFA treatment. Further TLC immunostain analysis with antibodies specific for longer chain derivatives of neolacto-series core structures indicated that many of the bands labeled with [3H]serine in the absence of BFA were composed of a1+2-and/or al+3-fucosylated structures (IV2FucnLc4, II13FucnLc4, or II131V2FucznLc4). Although these species were also present in whole glycolipid extracts from BFA-treated cells, their lack of [3H]serine labeling indicated these structures do not represent de novo-synthesized glycolipids in the presence of BFA. In fact, the BFA biosynthetic block after synthesis of Lc3 was highly effective in all cell lines tested. Only in instances where the highest extent of 3H labeling of Lc3 occurred was there further elongation, presumably to yield nLc4 and nLc5, based upon TLC immunostaining results. A quantitative comparison of LC3 expression in untreated and BFA-treated cells by solid phase radioimmunoassay is shown in Fig. 3. These data show the effect of glycolipid titration by serial dilution on binding of excess specific antibodies. Depending on the cell line, an increase of from 4-to 20-fold higher Lc3 expression occurred based upon titration curves with antibody TE5 using glycolipids derived from kBFAtreated cells. As an internal control, little difference was observed between the treatments with 1B2 reactivity. These results are reflective of intense biosynthesis of glycolipids in the presence of BFA, leading to high accumulations of Lc3 with only weak further modification.
BFA-induced inhibition of trafficking and secretion out of the Golgi has been documented (9-13), including effects on sphingolipid distribution (14,15). In order to confirm that the localization of neolacto-series glycolipid precursor structures after BFA treatment behaved similarly, the composition of de nouo-synthesized glycolipids from isolated Golgi fractions was compared with that of whole cells after [3H]serine labeling in the presence or absence of BFA. Solid phase immunoassays shown in Fig. 4 indicate the presence of neolacto-series glycolipids from whole cells and Golgi fractions in two representative cell lines, PC9 and HCT-15. Based upon these glycolipid titration curves, the characteristic increase in TE5 reactivity (anti-Lc3) in the presence of BFA is observed both in total cellular glycolipids and in Golgi-derived glycolipid fractions. Control assays also indicated significant expression of neolacto-series core chain and LeY antigen in all fractions. Thus, the results obtained with Golgi-derived glycolipid fractions correspond in each case to the results from whole cells. These data are consistent with the presence and enrichment of glycolipids (both short and longer chain) in Golgi-derived fractions regardless of BFA treatment. In addition, pl-3Nacetylglucosaminyltransferase and pl-wigalactosyltransferase activities were highly expressed in both crude cell homogenates and Golgi-enriched fractions from each cell line regardless of BFA exposure (results not shown). Further, fluorographs of [3H]serine-labeled, de nouo-synthesized glycolipids after TLC separation indicated BFA-induced accumulations of Lc3 in both whole cells and Golgi fractions.
Moreover, very similar labeling profiles for total cellular or Golgi-derived glycolipids for both BFA-treated and untreated cell fractions were obtained (results not shown). These results indicate that neolacto-series glycolipid precursors behave similarly to other Golgi membrane components upon BFA treatment, leading presumably to accumulations of biosynthetic intermediates in redistributed Golgi membrane compartments.
TLC immunostaining results for total neutral glycolipids, using the anti-Gb3 1A4E10 antibody, indicated intense expression of this antigen in both PC9 and U937 cells.
[3H]Serine labeling of this glycolipid band was also halted in the presence of BFA in both cell lines. This indicates that BFA-induced redistribution of Golgi membranes also segregated the al+ 4galactosyltransferase for biosynthesis of Gb3, such that it did not have access to the de nouo-synthesized lactosylceramide precursor. Thus, enzymes responsible for globo-series glycolipids would presumably be localized in compartments not redistributed by BFA.
As indicated above, weak [3H]serine labeling of gangliosides was observed for each cell line tested. As shown in Fig. 5, stronger labeling was observed in ganglioside fractions from cells labeled in the presence of BFA. In addition, labeling of longer chain gangliosides was virtually abolished by BFA treatment. These results indicate that after BFA-induced redistribution, the a2+3sialyltransferase responsible for GM3 synthesis continued to have access to de nouo-synthesized lactosylceramide. These results are consistent with previous localization of the GM3 synthetase to the cis-Golgi (5) and the BFA-induced block in synthesis of longer chain a2-Ssialyltransferase acceptors described above.
A summary of the distribution of [3H]serine label in glycolipids isolated from untreated and BFA-treated cells is shown in Table I.
Effect of BFA Treatment on Glycosyltransferase Activities of the Neolacto-series Pathway-The above results indicate a marked effect of BFA on neolacto-series glycolipid synthesis. To exclude the possibility of this being due to changes in levels of glycosyltransferase activities, the specific activities of pl-wigalactosyltransferase, pl+3N-acetylglucosaminyltransferase, and a1+3fucosyltransferase were determined for all cell lines under both treatment conditions. These results are shown in Table 11. Although the enzyme levels varied between cell lines, there was no difference in the specific activity of any enzyme tested as a consequence of BFA treatment. Similarly, no evidence was obtained for differences in hydrolytic activities as a consequence of BFA treatment. The ['4C]galactose from in vitro UDP-['4C]galacto~e labeling of endogenous glycolipids of isolated Golgi membranes was very stable under both conditions. Similar results were observed after either N-acetylglucosamine or fucose transfer (results not shown).
In order to account for BFA-induced accumulation of Lc3, the effect of BFA as a direct inhibitor of pl+ 4galactosyltransferase activity was tested using a crude cell homogenate of HCT-15 cells. These results are shown in Table 111. The effect of BFA on in vitro enzyme activity was tested using BFA concentrations 20-and 40-fold higher than that used to induce its physiological effect on the cells in culture. Further, conditions for enzyme activation were varied to include detergent, phospholipid, or no additions in case one or more of these components interfered with an inhibitory property. The results indicate that BFA had no effect on in vitro p1+4galactosyltransferase activity in any of the assay conditions tested. Thus, accumulation of Lc3 as a result of BFA treatment Fold Glycolipid Dilution  FIG. 4. Solid phase radioimmunoassays of total neutral glycolipids derived from whole cells or isolated Golgi membranes after treatment with or without 2.5 pg/ml BFA for 12 h. Glycolipids derived from 380 pg of dried extracted whole cell residue or 88 pg of dried extracted Golgi membrane residue per ml of solution were serially diluted and assayed as described under "Experimental Procedures." Panels A, B, and C show results from whole cell glycolipids with antibodies TE5, 1B2, and AH6, respectively. Panels D, E, and F show results from isolated Golgi membranes with antibodies TE5, 1B2, and AH6, respectively. Data are given for the following cell lines and conditions, +--+, PC9 cells without BFA; *--*, PC9 cells with BFA; 0--0, HCT-15 cells without BFA x--x, HCT-15 cells with BFA.
does not correlate with altered levels of biosynthetic enzymes either in vivo or in vitro. Instead, the data are consistent either with BFA-induced membrane changes effecting efficient flow of biosynthetic intermediates between enzymes or with changes in functional levels of glycosyltransferase activity, presumably due to partitioning between BFA-sensitive and -resistant compartments. In the latter context, this is particularly noteworthy in view of the comparative levels of p1-3N-acetylglucosaminyltransferase and pl-wigalactosyltransferase and their function in biosynthesis of neolactoseries glycolipids. For all cell lines tested, the pl+ 4galactosyltransferase activity was between 24-and 144-fold higher than that of ~1+3N-acetylglucosaminyltransferase.
The results are summarized diagrammatically in Fig. 6.

DISCUSSION
Brefeldin A, a macrocyclic lactone isolated from fungi, has been found to promote the disassembly of the Golgi and cause it to mix with the ER. The bulk of the evidence indicates this process involves redistributing the cis/medial/truns-Golgi cisternae, but not the trans-Golgi network (9-13). BFA also inhibits membrane trafficking pathways (12) and association of Golgi binding proteins such as p coat protein (26) or ADPribosylation factors (27) to the Golgi. At present, there is no evidence to suggest that BFA is directly inhibitory to sugar nucleotide transport or of glycosyltransferase reactions. Because of this, BFA is a potent tool and has been used to analyze the sequential distribution of glycosyltransferase enzymes responsible for both glycoprotein (16) and glycolipid (14, 15, and this paper) oligosaccharide synthesis. Biosynthesis of the oligosaccharide chains of glycolipids occurs via stepwise addition of sugar residues giving rise to, in general, ganglio-, globo-, or lacto-series type 1 and 2 chainbased structures. Each of these glycolipid classes shares lactosylceramide (Gal/31+4Glc/?l+lCer) as a common biosynthetic precursor. The nature of the glycolipid series will be defined by the next sugar residue added. Transfer of a2-3 sialic acid or P1+4GalNAc is associated with ganglio-series synthesis, a 1 4 G a l for globo-series, and /31+3GlcNAc for lacto-series structures. These glycolipids are distributed in various cell types, and a given cell or cell type may express glycolipid structures of multiple series. Consequently, regu- latory control of glycolipid synthesis is dependent on the specific glycosyltransferases expressed in the cell and also on their organization in the membrane, which affects the access of one reaction product to the next transferase(s) downstream. Information relating to the membrane organization of these enzymes is of critical importance in order to understand the regulation of glycolipid synthesis and, in particular, division of common precursors between multiple glycosyltransferases.
In this report, we describe aspects of de nouo glycolipid biosynthesis based upon [3H]serine metabolic labeling of ceramide. Time-dependent labeling studies indicated the appearance of progressively longer chain glycolipids, consistent with the view of an ordered distribution of transferase enzymes. In this way, labeled pools of shorter chain structures accumulate and become progressively elongated to yield end stage structures. The pathway for synthesis of neolacto-series glycolipid core structures involves transfer of 614galactosyl residues to GlcCer, yielding lactosylceramide, followed by the addition of a /31+3-linked GlcNAc residue and, finally, a second @l+ 4galactosyl residue. In vitro evidence indicates that lactose synthetase (/314galactosyltransferase) may be able to transfer to both GlcCer and Lc3 (28). However, a distinct enzyme capable of efficient fl14galactosyl transfer to GlcCer has recently been reported from human kidney (38). The efficient synthesis and accumulation of LC3 in BFA-treated cells is consistent with this finding.
The rate-limiting step in this pathway is formation of LC3 (29), a product which is very efficiently converted to nLc4 by 614galactosyltransferase. This observation is fully expected given that 614galactosyltransferase specific activity in cells may be 20-to over 100-fold higher than that of the @1+3Nacetylglucosaminyltransferase. As a result, little of this intermediate is ordinarily present in total neutral glycolipid extracts. When it is found, evidence suggests that much of it is a hydrolytic product from turnover of longer chain Structures? In the presence of BFA, accumulations of de novo-synthesized Lc3 occurred and resulted in diminished amounts of longer chain glycolipids for all cell lines tested. This a surprising finding because it occurred in spite of the magnitude of the excess @14galactosyltransferase activity over that of the ~l+3N-acetylglucosaminyltransferase.
These observations, taken together, begin to provide a more detailed view of the organization of glycosyltransferases in the Golgi and other subcellular compartments. BFA-induced redistribution of Golgi membranes presumably yielded segregated compartments which retain the capacity for glucosyl transfer to ceramide, 614galactosyl transfer to yield lactosylceramide, GM3 synthetase, and ~1+3~-acety~glucosaminyltransferase. Consequently, the major products under this condition were determined to be G M~, k 3 , and immediate precursors. Surprisingly, despite the huge excess of Dl+ 4galactosyltransferase over ~l+3N-acetylglucosaminyltransferase in these cells, only a small proportion of it may be active in this compartment. The slow conversion of labeled Lc3 to nLc, and subsequently to nLc5 is further indication that this /314galactosyltransferase enzyme pool must be very small (at least in terms of activity). These results are consistent with a previous report demonstrating increased terminal GlcNAc residues (and diminished amounts of &Gal residues) in N-linked oligosaccharides from bovine endothelial cells after BFA treatment (16). These observations, coupled with the normally efficient conversion of LC3 to longer derivatives, suggests the bulk of the fl14galactosyltransferase activity is functionally isolated from de nouo-synthesized precursors in separate compartments or is otherwise rendered nonfunctional by indirect effects from the BFA-induced disruption of the normal physiological environment. The latter possibility would presumably be peculiar to pl4galactosyltransferase since BFA did not block LC3 synthesis. Strous et al. (30) reported that a small amount of the total Dl+ 4galactosyltransferase was present and functional in the rough ER and cis-Golgi of HeLa cells. Such a distribution could account for the slow elongation of LC3 observed.   Components added to the reaction mixture to activate the enzyme reaction. Enzyme assays were otherwise conducted as described under "Experimental Procedures" in the presence of the indicated concentration of BFA. Phospholipids, when present, were at a final concentration of 250 pg/ml. The HCT-15 cell enzyme was used in this analysis using crude cell homogenates as the enzyme source.

W l "
" " -" " " It is unlikely that the effect observed in the presence of BFA is due to inhibition of enzyme activity or sugar nucleotide transport. No change in transferase activity occurred in cells cultured in the presence of BFA. Also, BFA did not directly inhibit /?1-4galactosyltransferase activity under a variety of in vitro assay conditions. Further, accumulation of significant amounts of Lc3 would require adequate UDPGal in the preceding reaction. No other evidence consistent with either of the above possible explanations has been reported. The simplest explanation, in keeping with recognized properties of BFA, is that the bulk of the fll+4galactosyltransferase activity is not redistributed to the ER by BFA. Multiple forms of this enzyme have been described in mammalian systems due to alternative initiation at two in-frame AUG codons (31, 32), and these forms have differing intracellular fates, establishing a Golgi and cell surface form of the enzyme (33). The BFA effect would presumably be restricted to the Golgi enzyme. Perhaps much of this enzyme in the cells tested is associated with the trans-Golgi network or with elements of the trans-Golgi which are poorly redistributed by BFA and thus straddle the block. On the other hand, pl4galactosyltransferase is reported to be a trans-Golgi marker (34, 35), and previous ultrastructural studies using antibodies specific for this enzyme showed that, a BFA-induced redistribution of it occurred (36, 37). Thus, the results point out potential inconsistencies between ultrastructural studies and biochemical function. This leaves open the possibility that although the enzyme is redistributed, it has little activity in situ by virtue of transient effects on the enzyme. It is also possible that BFA-induced changes in Golgi membranes causes a disruption of the physiological organization of many of these enzymes, effecting the efficient movement of the reaction product from one enzyme to the next enzyme downstream. This could also result in accumulation of biosynthetic intermediates. However, in the present case, it is surprising that this would occur for the acceptor of an enzyme present in such high comparative amounts in the cell. Further study will be required to resolve these questions.
The results presented further indicate that BFA blocks elongation of lactosylceramide to yield the globo-series precursor Gb, in cells which express al4galactosyltransferase. Earlier reports relating to ganglio-series chain biosynthesis indicated that BFA blocked elongation of de novo-synthesized GM3 to GM2 (14, 15). Thus, the fl14N-acetylgalactosaminyltransferase catalyzing this reaction is also trans to the BFA block. Consequently, synthesis of the first committed step in lacto-series chain synthesis (formation of Lc3) would be expected to occur in earlier Golgi cisternae than either of these other two reactions which define differing classes of glycolipids. A previous study (7) also provided evidence for expression of fl~+3N-acetylglucosaminyltransferase in early Golgi cisternae. These observations provide a basis for understanding the relative biosynthesis of core structures for differing glycolipid classes in terms of the fate of common precursors and the distribution of the respective biosynthetic enzymes in Golgi membranes.
The relative distribution of glycosyltransferases as defined by BFA sensitivity or resistance thus provides a basis for understanding membrane regulation of enzymes which compete for common oligosaccharide precursors (and which enzymes have greater access to these precursors). This is especially significant with respect to relative synthesis of lactoseries type 1 and 2 core chains which are expressed, for example, in colonic adenocarcinoma cells and form the basis for important tumor-associated carbohydrate antigens. In this instance, the Lc3 precursor is an acceptor for both pl+3-and p14galactosyltransferases. Studies focused on this topic are currently underway.