Up-regulation of Neutral Glycosphingolipid Synthesis upon Long Term Inhibition of Ceramide Synthesis by Fumonisin B1 *

In a previous study we observed that long term (5 days) incubation with fumonisin B1 (FB1), an inhibitor of acylation of sphingoid long chain bases to (dihydro)ceramide, resulted in morphological and biochemical changes in 3T3 fibroblasts (Meivar-Levy, I., Sabanay, H., Bershadsky, A. D., and Futerman, A. H. (1997) J. Biol. Chem. 272, 1558–1564). Among these were changes in the profile of synthesis of sphingolipids (SLs) and glycosphingolipids (GSLs). Whereas [3H]globotriaosylceramide ([3H]Gb3) comprised 1.9% of the total [3H]SLs and [3H]GSLs synthesized in control cells, it comprised 16.5% in FB1-treated cells. We now demonstrate by in vitro analysis that inhibition of ceramide synthesis by FB1 for 5 days results in up-regulation of the activities of three enzymes in the pathway of Gb3 synthesis, namely glucosylceramide, lactosylceramide, and Gb3 synthases; up-regulation is due to an increase inV max, with no change in K m values toward lipid substrates. Moreover, molecular analysis (reverse transcriptase-polymerase chain reaction) of glucosylceramide synthase indicated that this enzyme is up-regulated at the transcriptional level. No changes in either the V max orK m values of sphingomyelin or of GM3synthase were detected after FB1 treatment. Analysis of SL and GSL synthesis in cultured cells using [4,5-3H]sphinganine as a metabolic precursor demonstrated that at low substrate concentrations, Gb3 synthesis is favored over GM3 synthesis and glucosylceramide synthesis is favored over sphingomyelin synthesis, whereas the opposite is true at high substrate concentrations. These data demonstrate that GSL synthesis and in particular Gb3 synthesis are tightly regulated in fibroblasts, presumably so as to maintain constant levels of Gb3 on the cell surface.

Sphingolipids (SLs) 1 and glycosphingolipids (GSLs) are ubiquitous and essential components of eukaryotic cell membranes (1). Significant variation exists in the types and levels of both acidic and neutral GSLs between different cells, and although most attention has been paid to the sialic-acidic containing GSLs, the gangliosides (2), neutral GSLs are also found at relatively high levels in a number of tissues (1,3). Among these is globotriaosylceramide (Gal␣1-4Gal␤1-4Glc␤1-ceramide; Gb 3 ). Gb 3 is expressed in many types of human blood cells, including erythrocytes, lymphocytes, and platelets, and its ex-pression is elevated in some lymphomas (4,5). Sequential changes in the expression of Gb 3 and of other GSLs occurs during B cell differentiation due to sequential activation of the corresponding glycosyltransferases (6). No single function has been ascribed to Gb 3 , although it has been implicated as a differentiation antigen for B lymphocytes, as the P k blood group antigen (7), and as a marker for apoptosis of germinal center B-cells (8). In addition, although not a physiological function, Gb 3 acts as the cell surface receptor for Shiga toxin (9 -11).
In a recent study, we demonstrated that upon inhibition of SL and GSL synthesis by fumonisin B 1 (FB 1 ), an inhibitor of (dihydro)ceramide synthesis (12), levels of Gb 3 synthesis were not reduced (13). Upon incubation of 3T3 fibroblasts for 5 days with 20 M FB 1 , the incorporation of [4, H]sphinganine (a precursor of the synthesis of all SLs and GSLs (14) 16.5% in FB 1 -treated cells (13).
In the current study, we examine the molecular mechanisms underlying the differences in profiles of SL and GSL synthesis after long term inhibition of ceramide synthesis by FB 1 . By in vitro analyses, we demonstrate that three glycosyltransferases in the metabolic pathway leading to Gb 3 synthesis are upregulated. In addition, in vivo analysis demonstrates that differences in K m values also result in a preference for Gb 3 synthesis when ceramide is synthesized at low levels. That fibroblasts specifically maintain cellular Gb 3 levels, in preference over other GSLs and SLs, adds weight to the idea that Gb 3 plays one or more essential, although as yet unidentified, physiological functions in fibroblasts and presumably in other cells. culture dishes for in vivo experiments, both at densities of ϳ5 ϫ 10 4 cells/ml of medium. Medium was not changed before harvesting.
GlcCer Synthase (UDP-Glucose:N-Acylsphingosine ␤1-1-Glucosyltransferase)-Cells were washed three times with phosphate-buffered saline, harvested using a rubber policeman, and homogenized in a hand-held Potter-Elvehjem homogenizer in 1 ml of 25 mM KCl and 50 mM Tris (pH 7.4) (TK buffer). Homogenates were used fresh. GlcCer synthesis was assayed exactly as described (17). The standard reaction mixture contained 50 g of protein (determined according to Ref. 18), 5 mM UDP-Glc, 10 M [ 14 C]hexanoyl Cer, 10 mM MnCl 2 , in total volume of 500 l of TK buffer. The reaction was terminated after 30 min at 37°C by addition of 1.5 ml chloroform/methanol (1:2 v/v). Lipids were extracted (19) and separated on TLC using chloroform/methanol/water (65:25:4 v/v/v) as developing solvent. TLC plates were exposed to a 14 C-sensitive imaging plate, lipids were recovered from the plates by scraping, and radioactivity was determined by liquid scintillation counting.
SM Synthase (N-Acylsphingosine:Phosphatidylcholine Phosphocholinetransferase)-SM synthase (20) was assayed as described for GlcCer synthase, except that 100 g of protein was used for a 2-h incubation and no UDP-Glc or MnCl 2 were added to the reaction mixture.

FIG. 1. In vitro analysis of SM and GlcCer synthases.
A, GlcCer synthase was assayed in vitro using 50 g of protein from homogenates obtained from control (f) or FB 1 -treated fibroblasts (Ⅺ). SM synthase was assayed in cells treated with (E) or without (q) FB 1 using 100 g of protein. Each point is the mean Ϯ S.E. (n ϭ 6); the error bar is smaller than the size of the symbol in cases were no error bars can be seen. B, TLC plate showing a typical experiment analyzing GlcCer and SM synthases. The two left-hand lanes show the activity of GlcCer synthase, and the two right-hand lanes show SM synthase, both assayed after 5 days of incubation with FB 1 . Con, control; FB 1 , FB 1 -treated. In Vivo Analysis of SL Metabolism [4, H]Sphinganine was synthesized by reduction of D-erythrosphingosine with NaB[ 3 ]H 4 (10 Ci/mmol) (14,23,24). After 24 h of incubation with [4, H]sphinganine, cells were washed with phosphate-buffered saline, removed by scraping with a rubber policeman, and centrifuged (15,000 ϫ g av , 30 min, 4°C). Protein was determined (18), and [ 3 H]SLs/[ 3 H]GSLs were extracted and analyzed exactly as described (13). Upon metabolism of [4, H]dihydroceramide to [ 3 H]ceramide, we assume that 50% of the 3 H radioactivity is lost due to dehydrogenation of the 4,5-double bond; this was taken into account when quantifying [ 3 H]GSL synthesis as described previously (14).

RESULTS
Five major SLs and GSLs are synthesized by 3T3 fibroblasts, namely SM, GlcCer, LacCer, Gb 3 , and GM 3 (13). Using [4, H]sphinganine as a precursor of SL and GSL synthesis, we previously observed that residual levels of synthesis of each lipid differed after 5 days incubation with FB 1 (20 M) (13), and unexpectedly, Gb 3 synthesis was not inhibited to any extent. No differences in residual levels of SL or GSL synthesis were observed after incubation with FB 1 for short times (i.e. 1 h). 3 We have now systematically determined the activity of the five enzymes responsible for the synthesis of each lipid by in vitro analyses and analyzed whether differences in K m values between the different enzymes might account for the change in the profile of SL and GSL synthesis.
In Vitro Analysis of SL Synthesis-Cultured fibroblasts were incubated with FB 1 (20 M) for various times, removed from the culture dishes, and homogenized, and SM and GlcCer synthases were assayed in vitro using a short acyl chain radioactive analogue of ceramide, [ 14 C]hexanoyl Cer (15,17,20). After 3 h of incubation of fibroblasts with FB 1 , there was no difference in the activity of GlcCer synthase compared with control cells, but after 1 day, a 30% increase was detected (Fig. 1). As time of incubation with FB 1 increased, the activity of GlcCer synthase increased to a maximum of ϳ3-fold higher in cells incubated with FB 1 for 5 days compared with untreated cells (Fig. 1). Addition of FB 1 had no effect on GlcCer synthase activity or on the activity of any of the other enzymes assayed below, when added directly to the reaction mixture. The activity of SM synthase was much lower than GlcCer synthase in vitro, and no change in the level of SM synthase was detected even after 5 days of incubation with FB 1 (Fig. 1).
To determine whether the increase in activity of GlcCer synthase was due to changes in the V max or in the K m of the reaction, assays were performed using conditions in which the initial reaction rate (V o ) was linear with respect to time and to protein concentration and was not limited by substrate availability. After 5 days incubation with FB 1 , the V max of GlcCer synthase was 333 pmol/min/mg of protein, compared with 167 pmol/min/mg of protein in control cells (Fig. 2), but the K m values with respect to [ 14 C]hexanoyl Cer were unchanged (6.3 M in FB 1 -treated cells and 6.5 M in control cells (Fig. 2)). No change was detected in either the V max or K m value of SM synthase after 5 days of incubation with FB 1 (not shown).
Because the cDNA encoding GlcCer synthase has recently been isolated (29), we analyzed the molecular mechanisms by which GlcCer synthase activity was regulated. RT-PCR analysis revealed that GlcCer synthase mRNA expression was increased in cells incubated with FB 1 for 3 (not shown) or 5 days (Fig. 3

) but not for 3 h (not shown). This demonstrates that
GlcCer synthase is up-regulated at the transcriptional level upon long term incubation with FB 1 .
We next examined the activity of LacCer synthase using [ 14 C]hexanoyl GlcCer as substrate. The V max of LacCer synthase also increased upon incubation with FB 1 for 5 days, from 143 fmol/min/mg of protein in control cells to 250 fmol/min/mg of protein in FB 1 -treated cells (Fig. 4). The K m values with respect to [ 14 C]hexanoyl GlcCer were unchanged (1.25 M in FB 1 -treated cells and 1.57 M in control cells (Fig. 4)). Simi- larly, the activity of Gb 3 synthase was 2.7-fold higher in FB 1treated versus control cells (Fig. 5), 4 but in contrast, there was no increase in the V max of GM 3 synthase after FB 1 treatment for 5 days (Fig. 6), although a small but statistically insignificant reduction in the activity of GM 3 synthase was observed after FB 1 treatment.
Incubation with FB 1 causes depletion of ceramide from the synthetic pathway but also accumulation of sphinganine (12). Incubation of cultured cells directly with sphinganine (10 M added each day for 5 days) had no effect on enzyme activity measured in vitro (GlcCer, SM, and LacCer synthases), and co-incubation of FB 1 together with sphinganine (10 M added each day for 5 days) did not change the extent of enzyme up-regulation induced by FB 1 (not shown). These results suggest that the effects observed are due to inhibition of ceramide synthesis rather than accumulation of sphinganine.

In Vivo Analysis of [ 3 H]GSL and [ 3 H]SM Synthesis-
The data presented above demonstrate that long term inhibition (5 days) of ceramide synthesis results in up-regulation of the activity of three glycosyltransferases that use ceramide or downstream metabolites of ceramide as substrate, GlcCer, Lac-Cer, and Gb 3 synthases, but does not affect SM or GM 3 synthases. We next examined whether flux to different branches of the SL and GSL synthesis pathway (see Fig. 8 (Fig. 7B). These results demonstrate that the amount of substrate shunted to one or the other branch of the synthesis pathway varies depending on substrate concentration, with a preference for Gb 3 versus GM 3 synthesis at low substrate levels.
Analysis of GSL Turnover-To determine whether GSL turnover is also affected upon long term incubation with FB 1 , we analyzed acid glucosylceramidase activity in an in vitro assay 4 Assays for Gb 3 synthesis were not performed at saturating levels of UDP[ 14 C]Gal because the low specific activity of this enzyme precluded detection of activity when UDP[ 14 C]Gal was diluted with UDP-Gal.

DISCUSSION
The data presented above demonstrate that although only a minor GSL compared with GM 3 , fibroblasts strive to maintain constant levels of Gb 3 . When levels of metabolic precursors are low, Gb 3 levels are preserved by two mechanisms (Fig. 8), namely (i) up-regulation of the activity of the three glycosyltransferases in the pathway of Gb 3 synthesis and (ii) shunting LacCer to Gb 3 synthesis rather than to GM 3 synthesis. Together, this results in maintenance of Gb 3 levels even when levels of metabolic precursors are low. In addition, GSL synthesis is preferred to SM synthesis when ceramide synthesis is low (Fig. 8) (see also Ref. 31). A similar dual mechanism of regulation has been suggested for GM 2 and GD 2 synthases in human cancer cell lines (32), where expression of GD 2 synthase is regulated by levels of its immediate precursor, GD 3 . Regulation of glycolipid expression by substrate availability has also been observed during development and oncogenesis (33,34).
Molecular analysis demonstrated that GlcCer synthase is up-regulated at the transcriptional level resulting in increased production of the glycosyltransferase rather than by posttranslational modification of existing enzyme. This is supported by the observation that relatively long times of incubation with FB 1 (Ͼ1 day) were required to detect changes in V max values. Similarly, GlcCer synthase is induced during keratinocyte differentiation (35) due to transcriptional up-regulation (36). Interestingly, upon long term incubation with FB 1 , the activity of three glycosyltransferases (GlcCer, LacCer, and Gb 3 synthases) is increased, presumably by coordinate regulation at the transcriptional level. At this time, it is not possible to determine whether Gb 3 synthase is also regulated at the transcriptional level because this enzyme has not been cloned, although a partial purification of enzyme activity was reported from rat liver (37).
We do not know whether up-regulation of the glycosyltransferase activities is a result of inhibition of ceramide synthesis

FIG. 7. Synthesis of [ 3 H]GSLs and [ 3 H]SM in cultured fibroblasts.
Ratios of synthesis of each lipid or lipid class are calculated from the total amount of synthesis as given in Table I (fmol/g  per se or of reduction of GM 3 and Gb 3 levels. It is well documented that cells respond in various ways to changes in ceramide levels, but most cellular responses are due to increased production of ceramide at the cell surface due to degradation of SM in ceramide-mediated signaling pathways (38) rather than ceramide depletion in the endoplasmic reticulum where it is synthesized (23,39). 5 Alternatively, fibroblasts may respond to changes in levels of the lipid products (GM 3 and Gb 3 ). Depletion of GM 3 in fibroblasts results in changes in the actin cytoskeleton and in a block in cell proliferation and DNA synthesis (13), and one of these effects could be an initial signaling event in glycosyltransferase up-regulation. Although we cannot distinguish between biochemical effects due to depletion of GM 3 compared with Gb 3 , it is nevertheless apparent that fibroblasts ascribe particular importance to maintaining Gb 3 levels.
That fibroblasts go to such length to maintain cellular Gb 3 levels implies that Gb 3 plays a key role in fibroblast function. Gb 3 is expressed in a restricted set of hematopoietic cells and is considered both a differentiation antigen (40) and more recently as a regulator of apoptosis during differentiation of the hematopoietic system (41). No single function has been ascribed to Gb 3 in fibroblasts. We assume that the function of Gb 3 is related to its localization at the cell surface and are testing the possibility that Gb 3 is found in GSL-enriched domains on the cell surface, as may be the case for both GM 3 (42) and GM 1 (43). Because a number of transducer molecules also appear to be found in these GSL-enriched domains, such as c-Src, Ras, Rho, and focal adhesion kinase, the presence of Gb 3 in these domains might suggest a regulatory role in signaling events related to cytoskeletal organization; indeed, we have recently shown that GM 3 mediates events associated with assembly of the actin cytoskeleton (13). This possible role has not been tested for Gb 3 , but if Gb 3 is also enriched in these domains, it will be of importance to determine the mode of interaction of these two structurally similar GSLs with both signaling molecules and with the underlying cytoskeleton.