The role of sphingolipids in the maintenance of fibroblast morphology. The inhibition of protrusional activity, cell spreading, and cytokinesis induced by fumonisin B1 can be reversed by ganglioside GM3.

Previous studies demonstrated that inhibition of sphingolipid synthesis by the mycotoxin fumonisin B1 (FB1) disrupts axonal growth in cultured hippocampal neurons (Harel, R., and Futerman, A. H. (1993) J. Biol. Chem. 268, 14476-14481) by affecting the formation or stabilization of axonal branches (Schwarz, A., Rapaport, E., Hirschberg, K., and Futerman, A.H. (1995) J. Biol. Chem. 270, 10990-10998). We now demonstrate that long term incubation with FB1 affects fibroblast morphology and proliferation. Incubation of Swiss 3T3 cells with FB1 resulted in a decrease in synthesis of ganglioside GM3, the major glycosphingolipid in 3T3 fibroblasts and of sphingomyelin. The projected cell area of FB1-treated cells was ∼45% less than control cells. FB1 had no affect on the organization of microtubules or intermediate filaments, but fewer actin-rich stress fibers were observed, and there was a loss of actin-rich lamellipodia at the leading edge. Three other processes involving the actin cytoskeleton, cytokinesis, microvilli formation, and the formation of long processes induced by protein kinase inhibitors, were all disrupted by FB1. All the effects of FB1 on cell morphology could be reversed by addition of ganglioside GM3 even in the presence of FB1, whereas the bioactive intermediates, sphinganine, sphingosine, and ceramide, were without effect. Finally, FB1 blocked cell proliferation and DNA synthesis in a reversible manner, although ganglioside GM3 could not reverse the effects of FB1 on cell proliferation. Together, these data suggest that ongoing sphingolipid synthesis is required for the assembly of both new membrane and of the underlying cytoskeleton.

Sphingolipids (SLs) 1 are almost ubiquitous components of eukaryotic cell membranes where they play a variety of roles (1)(2)(3). An important approach to defining the precise roles of SLs is to inhibit their synthesis, by either genetic approaches, such as the production of mutants defective in SL synthesis (4), or by specific chemical inhibitors (5,6). Unfortunately, little information is available using the first approach due to the lack of success in purifying the enzymes of SL synthesis, and the small number of mutants that have been obtained. However, specific inhibitors of SL synthesis have become available recently (5), including the mycotoxin, fumonisin B 1 (FB 1 ) (7). FB 1 inhibits acylation of the sphingoid long chain bases sphinganine (dihydrosphingosine) and sphingosine to dihydroceramide and ceramide, respectively.
A number of studies have examined the effects of FB 1 on the growth of cultured cells in attempts to determine the cellular basis for the diseases associated with FB 1 (reviewed in Merrill et al. (8)). It has been shown that FB 1 stimulates DNA synthesis in confluent cultures of fibroblasts (9), but inhibits the proliferation of renal epithelial cells (10) and the growth of Saccharomyces cerevisiae (11). In cultured hippocampal neurons, FB 1 disrupts axonal growth by affecting the formation or stabilization of axonal branches (12,13) and disrupts dendrite growth in cerebellar Purkinje neurons (14). Together, these studies suggest that SL synthesis is required for cell growth and morphogenesis.
To further study the roles of SLs in cell morphogenesis, we have now analyzed the effects of FB 1 on Swiss 3T3 fibroblasts cultured at subconfluent densities. We previously demonstrated that the delivery of new membrane to the leading edge of these cells is required for pseudopodial activity and for directional migration (15), and we now show that long term incubation with FB 1 causes profound changes in a number of morphological processes associated with the actin cytoskeleton. Remarkably, all the effects of FB 1 on cell morphology could be reversed by addition of low concentrations of ganglioside GM 3 . These results suggest that SLs may be involved in the assembly of both new membrane and of the underlying cytoskeleton.
Cell Culture-Swiss mouse 3T3 cells were cultured in Dulbecco's modified medium containing 10% calf serum, and maintained in a water-saturated atmosphere of 5% CO 2 . Cells were dissociated with trypsin/EDTA and plated in either 60-mm culture dishes for biochemical experiments, or on glass coverslips for morphological analysis, both at densities of ϳ2 ϫ 10 4 cells/ml of medium.
Immunofluorescence and Fluorescence Microscopy-Microtubule distribution was examined using a monoclonal anti-␣-tubulin antibody (clone DM1A, Sigma), intermediate filaments were examined using a polyclonal anti-vimentin antibody (provided by Dr. Benny Geiger, Department of Molecular Cell Biology, Weizmann Institute of Science), and actin distribution was examined using TRITC-conjugated phalloidin. Cells were observed using Plan Apochromat 40ϫ/1.3 n.a., 63ϫ/1.4 n.a., and 100ϫ/1.3 n.a. oil objectives of a Zeiss Axiovert 35 microscope with an appropriate filter for rhodamine fluorescence. Cells labeled with DAPI were examined using a Plan Neofluar 20 ϫ 0.5 n.a. objective of a Zeiss Axiophot microscope with a filter for DAPI fluorescence.
Analysis of Cell Morphology-Projected cell area (i.e. the area occupied by the cell on the substrate) was determined after labeling cells with TRITC-conjugated phalloidin, and capturing images of labeled cells via an Applitec MSV-700L CCD camera to a Macintosh 840AV computer and using NIH imaging software. Dispersion and elongation indices of cell outlines were calculated according to Dunn and Brown (15,21). Cell outlines were identified and analyzed using software provided by Dr. Z. Kam, Department of Molecular Cell Biology, Weizmann Institute of Science.
Scanning Electron Microscopy-Fibroblasts, grown on 13-mm glass coverslips, were fixed in glutaraldehyde (2%, w/v) and paraformaldehyde (3%, w/v), followed by postfixation in 1% (w/v) osmium tetroxide in 0.15 M Na-cacodylate, pH 7.4 (22). Cells were rinsed in distilled water, treated with 1% tannic acid, rinsed again, treated with 2% (w/v) uranyl acetate in distilled water for 30 min, and then rinsed again. Specimens were dehydrated in a graded series of acetone, and critical point drying was performed using a Pelco critical point dryer. Specimens were coated with gold using a S150 Sputter Coater (Edwards) and examined using a Jeol JSM 6400 scanning electron microscope.

FB 1 Inhibits SL Synthesis in 3T3
Fibroblasts-To determine the extent of inhibition of SL synthesis by FB 1 in Swiss 3T3 fibroblasts, various concentrations of FB 1 were added to the medium at the time of plating. A dose-dependent inhibition of [ 3 H]SL synthesis was observed (Fig. 1A), with significant inhibition obtained using 20 M FB 1 . Using this concentration, [ 3 H]SL synthesis was inhibited to a similar extent when analyzed on each of the first 5-6 days in culture, and even short incubations with FB 1 (i.e. 1 h) resulted in similar levels of inhibition. When expressed as a function of the amount of protein per culture dish on day 5, the incorporation of [4, H]sphinganine into total [ 3 H]SLs was inhibited by Ͼ70% at 20 M FB 1 and above (Fig. 1B). The extent of inhibition appeared less when expressed as incorporation of [4,5-3 H]sphinganine per g protein (compare Fig. 1, A and B), since FB 1 -treated cells did not proliferate (see below), and the amount of protein per dish was much lower in FB 1 -treated cells, particularly at high concentrations of FB 1 (20  Cell Spreading Is Reduced by FB 1 Treatment-Inhibition of SL synthesis by FB 1 (20 M) for 5 days caused profound changes in cell morphology. Whereas control cells displayed typical fibroblast morphology, with a leading edge and a trailing edge ( Fig. 2A), FB 1 -treated cells were less well spread on the substrate and displayed reduced pseudopodial activity (Fig.  2B). After incubation with FB 1 for 5 days, the projected cell area was reduced by ϳ45% (Table I), and there was a significant decrease in the morphometric index (21) of dispersion (dispersion index of control cells was 0.91 Ϯ 0.09, and of FB 1treated cells was 0.55 Ϯ 0.09, n ϭ 40), but not of elongation (elongation index of control cells was 1.23 Ϯ 0.1 and of FB 1treated cells was 1.10 Ϯ 0.08, n ϭ 40). Elongation is a measure of the extent to which shape must be compressed along its longitudinal axis in order to minimize its difference from a circle, while dispersion is invariant to stretching, compressing, or shearing the shape in any direction; both of these indices are equal to zero for a circle. Elongation is considered as a measure of cell bipolarity, whereas dispersion is a measure of multipolarity (21). Thus, the formation of protrusions is reduced in FB 1 -treated cells, while cells remain elongated to a similar extent to control cells.
The effects of FB 1 on morphology were completely reversed 24 h after its removal from the medium, or by addition of exogenous ganglioside GM 3 to the medium for 24 h, even in the presence of FB 1 (Table I); GM 3 itself had no effect on projected cell area. Neither of the bioactive intermediates, sphinganine or sphingosine, which both accumulate upon long term treat-ment with FB 1 (8,23), had any effect on morphology even for up to 5 days incubation, and even after multiple additions (Table  I). Short acyl chain analogues of ceramide, whose level is depleted upon FB 1 treatment (8), and a short acyl chain analogue of dihydroceramide were also unable to reverse the effects of FB 1 (Table I).
Since in fibroblasts, short acyl chain analogues of ceramide are metabolized mainly to GlcCer and SM (not shown; see also Meivar-Levy et al. (24)), and since sphinganine, sphingosine, and ceramides do not significantly affect cell morphology, these data together demonstrate that the ability of GM 3 to restore cell morphology is due to depletion of an essential higher order glycosphingolipid, probably GM 3 , and not due to accumulation or depletion of bioactive intermediates. This is supported by observations that exogenously added GM 3 was metabolized to only a limited extent by fibroblasts during a 24-h incubation (see also Chigorno et al. (25)  ination of the distribution of cytoskeletal elements. FB 1 had no apparent effect on the radial distribution of microtubules or of intermediate filaments (not shown), but changes in the actin cytoskeleton were observed, with far fewer actin-rich "stress fibers," and loss of actin-rich lamellipodia at the leading edge (compare Fig. 3, A with C and E).
In addition to its roles in organization of the leading edge in interphase cells, the actin cytoskeleton also plays an important role in cell division, particularly in the formation of the contractile ring during cytokinesis. Incubation with FB 1 interfered with cytokinesis as demonstrated by the appearance of binuclear cells (Fig. 3, B, D, and F). The percent of binuclear cells was 3-4% in control cells, but 11-12% in FB 1 -treated cells (Fig.  4). Addition of GM 3 on day 4 in culture partially reversed the effects of FB 1 on cytokinesis (Fig. 4), but neither sphingosine or sphinganine had any effect (not shown).

FB 1 Interferes with Protrusional Activity and Microvilli
Formation-A number of protein kinase inhibitors, such as staurosporine and H-7, induce changes in fibroblast morphology (26,27). Upon incubation with H-7, fibroblasts acquire numerous long, thin processes (Fig. 5A) that result from the inability of cells to retract their trailing edge (26); at the leading edge, lamellipodial activity is unaffected and may even be enhanced (Fig. 5A). Pretreatment with FB 1 (20 M) for 5 days significantly decreased the number of long processes formed upon H-7 or staurosporine-treatment (Fig. 5), and GM 3 almost completely restored the ability of staurosporine to induce long processes (Fig. 5).
Fibroblasts were also incubated with mitomycin C, a drug that inhibits DNA replication and arrests cells in the S phase of the cell cycle. Since other biosynthetic processes are not affected, including synthesis of new membrane and new cytoskeletal components, cells continue to grow and acquire much larger sizes than untreated cells. The projected area of cells treated with 0.1 g/ml mitomycin C for 5 days was 4.3-fold greater than that of untreated cells (Table II), and mitomycin C-treated cells displayed a large number of small microvilli at the cell surface that were generally located near the center of the cell (Fig. 6A). The projected area of FB 1 -treated cells was only 2.4-fold greater after incubation with mitomycin C (Table  II), and FB 1 -treated cells displayed far fewer microvilli (Fig.  6B). The addition of GM 3 on day 4 to cells that had been treated with both mitomycin C and FB 1 resulted in a significant increase in projected cell area (Table II), demonstrating that GM 3 is able to partially restore the inhibitory effects of FB 1 on cell spreading after mitomycin C treatment.
Cell Proliferation and DNA Synthesis Is Blocked by FB 1 -Little cell proliferation was observed for up to 10 days in the presence of 20 M FB 1 , although cells remained viable throughout this period. The number of cells in untreated cultures increased by ϳ23-fold during the first 7 days in culture, but only increased by 2.6-fold in FB 1 -treated cultures (Fig. 7A). The block in cell proliferation was reversible, since cell number increased after removal of FB 1 on day 4, and attained values similar to untreated cells (Fig. 7A). The block of cell proliferation could not be explained solely by inhibition of cytokinesis, since only 11-12% of the cells were binuclear after FB 1 -treat-ment (see Fig. 4). Indeed, analysis of [ 3 H]thymidine incorporation demonstrated that DNA synthesis was also significantly inhibited by FB 1 -treatment (Fig. 7B). The effects of FB 1 on DNA synthesis (Fig. 7B) could be reversed by removing FB 1 from the medium, but in contrast to its effects on cell morphology, addition of GM 3 had no effect on either cell proliferation (not shown) or on [ 3 H]thymidine incorporation (Fig. 7B). Addition of either sphingosine or sphinganine directly to the culture medium had no effect on cell proliferation, either in the absence or presence of FB 1 (not shown).

DISCUSSION
Inhibition of ceramide synthesis by FB 1 causes a number of responses in various cells (8). We now demonstrate that the reduction in complex SL synthesis that occurs upon incubation of Swiss 3T3 cells with FB 1 results in major changes in the actin cytoskeleton and in processes related to, or dependent on, the actin cytoskeleton and that these effects can be reversed by addition of ganglioside GM 3 .
Characterization of the Biochemical Effects of FB 1 on Swiss 3T3 Fibroblasts-Of all the inhibitors tested (5), FB 1 has proved particularly useful in manipulating levels of SL synthesis (7,8,28). As a consequence of FB 1 treatment, sphingosine and sphinganine levels are elevated, and ceramide is depleted (7). All three of these molecules can themselves disrupt cell morphology and proliferation (10,29), and it is therefore essential when using FB 1 to distinguish between effects caused by changes in levels of these bioactive intermediates or effects caused by depletion of complex sphingolipids (8). The inability of sphingoid long chain bases and of ceramide, and the ability of GM 3 to reverse the effects of FB 1 even in the presence of FB 1 , strongly suggests that the effects we observed in sparse cultures of 3T3 cells are due to depletion of complex SLs, as appears to be the case for changes in cell morphology and growth upon FB 1 treatment of cultured hippocampal neurons (12,13). The major SLs in 3T3 cells are GM 3 and SM, and both of these lipids are depleted to a similar extent after incubation with FB 1 . Surprisingly, levels of the neutral glycosphingolipid, Gb 3 , are elevated after FB 1 treatment. In cultured cerebellar neurons (28), SM synthesis was more sensitive to FB 1 than glycosphingolipid synthesis. Since the activities of SM synthase and of glycosyltransferases were not directly affected by FB 1 , it was suggested that different enzymes in the metabolic pathway are sensitive to a different extent to reduction in the levels of their respective substrates (28). Our data suggest that Gb 3 synthase has a relatively high K m value, which renders it relatively insensitive to changes in levels of its substrate, lactosylceramide. The Ability of GM 3 to Reverse the Effects of FB 1 -Incubation with FB 1 affects a number of cellular processes, including cell spreading, microvilli formation, cytokinesis, formation of long processes, and disruption of DNA synthesis and cell proliferation. GM 3 restores the disruptive effects of FB 1 on cell morphology, but not on proliferation. Four pieces of evidence suggest that GM 3 itself is responsible for the restoration of cell morphology. (i) GM 3 is synthesized at far higher levels than either lactosylceramide or Gb 3 , rendering it more suitable to mediate interactions between the plasma membrane and the actin cytoskeleton (see below). Moreover, Gb 3 synthesis is not inhibited by FB 1 treatment. It should however be noted that no role has yet been ascribed to Gb 3 , and the differences in the sensitivity of Gb 3 and GM 3 synthesis to FB 1 -treatment may suggest that levels of Gb 3 must be maintained at a constant level in the cell for it to perform whatever function it may play. (ii) GM 3 is only degraded to a limited extent by fibroblasts (see also Chigorno et al. (25)). (iii) The lack of effects of short acyl chain analogues of ceramide (which are metabolized mainly to GlcCer and SM, but not to lactosylceramide, GM 3 and Gb 3 ), suggest that GlcCer cannot be responsible for the restoration of cell morphology. (iv) Neither SM or ceramide can be responsible for restoration since incubation with PDMP (which inhibits the synthesis of GlcCer and of higher order glycosphingolipids, but enhances SM synthesis (30) and elevates ceramide levels), results in a similar decrease in projected cell area to that observed with FB 1 , and addition of exogenous GM 3 together with PDMP restores normal cell morphology. 3 Although exogenous GM 3 could restore all of the disruptive effects of FB 1 on cell morphology, including cytokinesis, it was completely inactive in reversing the inhibition of cell proliferation and DNA synthesis. Other studies have also demonstrated inhibition of cell proliferation by FB 1 (10,11,31). Cell growth is also arrested in rabbit skin fibroblasts (32) and in Swiss 3T3 cells (33) by PDMP, with growth arrested at the G 1 /S and G 2 /M transitions in 3T3 cells (33). In contrast, FB 1 stimulates thymidine incorporation in 3T3 cells (9); however, the conditions used in this study (9) were completely different from those used in the current study, inasmuch as we examined cell proliferation and DNA synthesis during the logarithmic phase of growth, whereas Schroeder et al. (9) examined stimulation of DNA synthesis in quiescent cultures. Although we cannot draw any definitive conclusions about the mechanism(s) by which FB 1 inhibits cell proliferation in sparse cultures of 3T3 cells, the inability of GM 3 to restore proliferation demonstrates that GM 3 is not the limiting factor in the regulation of cell proliferation. Elucidation of the possible roles of bioactive sphingolipid intermediates in regulating proliferation (2) requires further study.
The Relationship between GM 3 and the Actin Cytoskeleton-The most interesting conclusion from the current study is that GM 3 synthesis is required for a variety of processes that depend on the assembly of new membrane and of the underlying actin cytoskeleton. These processes are exemplified by the reduction in the number of microvilli in mitomycin C-treated cells (Fig. 6). Microvilli are surface extensions containing an actin core, and are found on the surface of many cells, particularly on cells that require a large surface area to function. Normal fibroblasts have few microvilli, but after mitomycin C-treatment, their levels are greatly increased, presumably since synthesis and assembly of actin and new membrane continues, even though cell division is arrested. The fact that FB 1 -treated cells have far fewer microvilli indicates that an intimate relationship exists between GM 3 and the actin cytoskeleton. Other examples of this relationship are illustrated by the reduction in projected cell area (Table I), pseudopodial activity (Fig. 2), the formation of long processes (Fig. 5), and the inhibition of cytokinesis (Figs. 3 and 4).
How might ganglioside GM 3 be related to the actin cytoskeleton? Since gangliosides are highly enriched in the plasma membrane (34), the effects described above are presumably due to alteration in a plasma membrane function. It is not known whether the effects depend on alterations in a physical property of the membrane, or alternatively on the assembly of new membrane, but it should be noted that all of the processes described above require membrane synthesis and assembly. For instance, protrusional activity at the leading edge can be blocked by treatments that inhibit the synthesis or delivery of new membrane (15,35). During the assembly of new membrane, actin-binding proteins such as ezrin, radixin, and moe- 3 The projected area of PDMP-treated (100 M) cells was 967 Ϯ 20 m 2 (n ϭ 3) compared to 1666 m 2 in untreated cells. The projected area of cells treated with PDMP (100 M) and GM 3 was 1562 Ϯ 63 (n ϭ 3) and 1794 Ϯ 7 (n ϭ 2) m 2 for 25 nM and 100 nM GM 3 respectively, and for cells treated with PDMP and GM 1 was 1081 Ϯ 12 (n ϭ 2) and 1165 Ϯ 4 (n ϭ 2) m 2 for 25 nM and 100 nM GM 1 .

TABLE II
The effects of FB 1 and GM 3 on the increase in cell area induced by mitomycin C Fibroblasts were treated 3-4 h after plating with mitomycin C (0.1 g/ml), with mitomycin C and FB 1 (20 M), or with mitomycin C and FB 1 on day 1 and GM 3 (25 nM) on day 4. Projected cell area was measured on days 5 and 6; the projected area of cells treated with FB 1 alone was 909 Ϯ 39 m. Each value represents the mean Ϯ S.E. of 3 independent experiments, for which at least 50 cells were analyzed.

Days in culture
Projected cell area after treatment with:

Control
Mitomycin sin (whose recruitment to the plasma membrane is required for microvilli formation) (36,37) should be recruited to the membrane, and it may be that GM 3 is involved in their recruitment. Consistent with GM 3 playing a role in the recruitment or activity of membrane proteins that direct actin assembly are studies showing that GM 3 is functionally associated with integrins (38). These and similar studies (39) focused mainly on the involvement of GM 3 in integrin-mediated cell adhesion. However, the data reported in the current study suggest that GM 3 plays a wider role since it appears necessary for mediating events associated with assembly of the actin cytoskeleton and of new membrane.