Production of Ceramides Causes Apoptosis during Early Neural Differentiation in Vitro *

To investigate signal transduction pathways leading to apoptosis during the early phase of neurogenesis, we employed PCC7-Mz1 cells, which cease to proliferate and begin to differentiate into a stable pattern of neurons, astroglial cells, and fibroblasts upon incubation with retinoic acid (RA). As part of lineage determination, a sizable fraction of RA-treated cultures die by apoptosis. Applying natural long-chain C16-ceramides as well as membrane-permeable C2/C6-ceramide analogs caused apoptosis, whereas the biologically nonactive C2-dihydroceramide did not. Treating PCC7-Mz1 stem cells with a neutral sphingomyelinase or with the ceramidase inhibitorN-oleoylethanolamine elevated the endogenous ceramide levels and concomitantly induced apoptosis. Addition of RA caused an increase in ceramide levels within 3–5 h, which reached a maximum (up to 3.5-fold of control) between days 1 and 3 of differentiation. Differentiated PCC7-Mz1 cells did not respond with ceramide formation and apoptosis to RA treatment. The acidic sphingomyelinase contributed only weakly and the neutral Mg2+-dependent and Mg2+-independent sphingomyelinases not at all to the RA-mediated production of ceramides. However, ceramide increase was sensitive to the ceramide synthase inhibitor fumonisin B1, suggesting a crucial role for the de novo synthesis pathway. Enzymatic assays revealed that ceramide synthase activity remained unaltered, whereas serine palmitoyltransferase (SPT), a key enzyme in ceramide synthesis, was activated ∼2.5-fold by RA treatment. Activation of SPT seemed to be mediated via a post-translational mechanism because levels of the mRNAs coding for the two SPT subunits were unaffected. Expression of marker proteins shows that ceramide regulates apoptosis, rather than differentiation, during early neural differentiation.

Neural cells die at all developmental stages and for many different reasons. Successful removal of cells with minimum disturbance of the surrounding tissue is achieved by a process called programmed cell death or apoptosis. In contrast to the striking heterogeneity of cell death induction pathways, the execution of the death program often involves identical molecules and is associated with characteristic morphological and biochemical changes (1). It has been reported that, during brain development, 20 -80% of all neurons are eliminated by apoptosis (2). The reasons for this massive neuronal cell death are discussed in the context of, for example, correction of erroneous projections, creation of pathways for axonal outgrowth, numerical limitations imposed by mechanisms involving successive cell doubling, and transient functions of the eliminated neurons (reviewed in Refs. 2 and 3). Although some of these hypotheses have received experimental support in individual systems, none appear to be generally applicable, and it is unlikely that a single explanation exists. Neurons seem to be produced in excess to allow competition for contacts with their cellular partners and thus adjust their numbers to provide sufficient enervation of their targets, a process called the "neurotrophic strategy" (reviewed in Ref. 4).
During recent years, a novel type of neural programmed cell death has become evident. In several systems, it was demonstrated that neuronal cells die very early during neurogenesis and, in some cases, well before the period of target contact (5,6). During chick development, cells die in the early neural tube (embryonic days 2-3) shortly after becoming post-mitotic or even during the cell cycle (7). Preventing this programmed cell death by administering caspase inhibitors blocks neural tube closure (8). Also, widespread programmed cell death was discovered to commence after embryonic day 10 in the murine cerebral cortex (5). Interestingly, the majority of dying cells were found within zones of proliferating cells rather than in regions of post-mitotic cells. At present, neither the exact biological function nor the specific mechanisms regulating this new form of neural cell death during pattern formation are known. A satisfactory biochemical analysis of this cell death during early neurogenesis is hampered in vivo due to the lack of specific markers for dying cells and of sufficient material, to problems of drug application, and to the fact that development of neural cells is not synchronized. To evade these limitations, it was desirable to establish and to employ an appropriate in vitro cell system.
We reported earlier that the teratocarcinoma cell line PCC7-Mz1 adequately mimics early steps of neural development (9 -11), including the processes of determination, differentiation, and apoptosis (12). Following incubation with all-trans-retinoic acid (RA), 1 PCC7-Mz1 cells cease proliferation and differenti-ate, over a period of several days, into a tissue-like pattern of neurons, astroglial cells, and fibroblasts (9,10). The embryonic carcinoma cell line PCC7-Mz1 may therefore be considered as a clone of pluripotent cells of neuroectodermal origin and serve as a model of neural development. We recently reported that apoptosis may be part of early neurogenesis, i.e. when PCC7-Mz1 stem cells become committed to their specific cell lineages (12). Within 24 h of RA treatment, a considerable fraction (ϳ23%) of the PCC7-Mz1 culture detaches and dies, whereas the remaining cells of the culture begin to differentiate along their specific lineages (neuronal, astroglial, and fibroblast). Apoptosis was characterized by nuclear condensation, intranucleosomal degradation of genomic DNA, activation of caspases, and formation of apoptotic bodies. Neurotrophic factors were ineffective in preventing this cell death, as was also shown for early cell death in the avian cervical cord (6), implying that cell autonomous mechanisms are responsible. Therefore, apoptosis may be instructed by the gene regulatory program that controls cell lineage determination and the formation of patterns of neuroectodermal derivatives. This notion is supported by the finding that Bcl-2, a negative regulator of apoptosis, becomes up-regulated in PCC7-Mz1 cells within a few days of RA treatment. Increase in Bcl-2 expression accompanies neural differentiation and is conversely related to the extent of apoptosis, which rapidly declines once the neuronal network starts to emerge in PCC7-Mz1 cultures, i.e. 2-3 days after treatment with RA. Hence, expression of Bcl-2 seems to provide protection from apoptosis for PCC7-Mz1 cells that have already achieved a certain level of differentiation. The intracellular signaling pathways leading to apoptosis in PCC7-Mz1 cells depend on caspases, but do not seem to involve conventional or novel members of the protein kinase C gene family (12).
Here we study the role of ceramide (N-acyl-erythro-sphingosine), which has been shown to act as a key signaling molecule in a distinct signal transduction pathway, in the sphingomyelin pathway or cycle, consisting of hydrolysis and resynthesis (reviewed in Refs. 13 and 14). The important regulatory function of ceramide in cell proliferation (15), cell differentiation (16), cell growth arrest (17), and cell death (18,19) is well documented. Extracellular agents such as the cytokines tumor necrosis factor ␣, nerve growth factor, interferon, interleukin-1-6, and CD95-L (all binding to cell-surface receptors) and vitamin D 3 and dexamethasone (acting via nuclear receptors) were shown to initiate rapid ceramide generation by activation of sphingomyelinases (13,14,17). One of the most interesting and intriguing properties of ceramide is the induction of apoptosis in cells in culture and in vivo (20). Ceramide generation was measured under various conditions that are known to lead to apoptosis, and ceramide levels were shown to increase in cells exposed to ionizing radiation (21), withdrawal of serum (22), UV light radiation, heat shock (23), chemotherapeutic agents such as daunorubicin (24), and oxidative stress (19). It was demonstrated that lymphoblasts from patients with Niemann-Pick disease, an inherited deficiency of acid sphingomyelinase (SMase), and acid SMase-deficient mice show defects in, but not total loss of, the apoptotic response (25).
The role of this lipid second messenger in both differentiation and RA-induced apoptosis of neural cells is so far only poorly understood. Therefore, we analyzed, by utilizing the PCC7-Mz1 system, the significance of ceramide in early neural determination. We show that ceramide induces apoptosis in a timeand dose-dependent manner and, furthermore, that RA induces an elevation of cellular ceramide levels in PCC7-Mz1 stem cells by de novo synthesis. Ceramide production was independent of sphingomyelinase and ceramide synthase action, but was due to activation of serine palmitoyltransferase. Interestingly, ceramide causes only apoptosis, but not neural determination or differentiation. Furthermore, we isolated subclones of PCC7-Mz1, which were unable to produce ceramides upon RA treatment and therefore did not undergo programmed cell death in response to RA.

MATERIALS AND METHODS
Cell Culture-The mouse embryonic carcinoma cell line PCC7-Mz1 is a subclone of the PCC7-S-AzaR 1 (clone 1009) cell line. Culture conditions, growth characteristics, and the RA-induced differentiation pattern have been described previously (9,10,12).
PCC7-Mz1 cells were grown in plastic tissue culture flasks in Dulbecco's modified Eagle's medium (Flow, Meckenheim, Germany) supplemented with 12.5% fetal calf serum (batch 148, Roche Molecular Biochemicals, Mannheim, Germany) at 37°C in humidified air and 10% CO 2 . RA-resistant cells were selected by culturing PCC7-Mz1 cells in the presence of 0.1 M RA. Proliferating cell colonies were isolated by ring cloning, and individual clones (R-clones) were established. Stock cultures of R-clones were cultivated in the presence of 0.01 M RA, which was omitted for the experimental cultures.
Prior to induction of differentiation, PCC7 cells were seeded at a density of 1.75 ϫ 10 4 cells/cm 2 in plastic culture dishes. For differentiation, cultures were treated with 0.1 M RA (final concentration; Sigma, Munich, Germany) 1 day after plating. 24 h later, the culture medium was replaced by Dulbecco's modified Eagle's medium supplemented with 12.5% fetal calf serum, 0.1 M RA, and 1 mM dibutyryl cAMP (Roche Molecular Biochemicals).
Treatment with fumonisin B 1 and desipramine (Sigma) was carried out 1 h before adding RA. C 16 -ceramide (Sigma) was applied according to the method described by Ji et al. (26). Briefly, C 16 -ceramide was dissolved in ethanol/dodecane (98:2, v/v; Sigma), thoroughly mixed with the medium. and added to PCC7-Mz1 cells plated the day before in a 96-well plate (5 ϫ 10 4 cells/well). The final concentration of the solvent was Ͻ0.5% ethanol and 0.01% dodecane.
Cell Viability Assay-For quantification of the degree of cell death in cell culture, we employed the viability assay based on the reduction of tetrazolium salt to formazan by mitochondrial dehydrogenase activity (27). The assay was performed in 96-well microtiter plates (Falcon, Heidelberg, Germany) as described previously (12), but WST-1 (Roche Molecular Biochemicals) was used instead of 3-(4,5-dimethylthiazol-2yl)-2,5-diphenyltetrazolium bromide. The light absorbance at 405 nm of the medium including all factors but without cells was determined and subtracted from the absorption readings with cells. At least eight wells/ sample point were analyzed. Each experiment was repeated three times.
Genomic DNA Analysis-Cells were cultivated in 6-cm dishes, and genomic DNA was isolated as described previously (12). Briefly, cells in the medium were collected by centrifugation (2000 ϫ g, 10 min, 4°C), and the pellet was resuspended in 20 l of cell lysis buffer (0.5% Triton X-100, 20 mM EDTA, and 5 mM Tris-HCl, pH 8). Adherent cells were washed once with ice-cold phosphate-buffered saline (PBS), detached with a rubber policeman, and centrifuged. This cell pellet was also resuspended in 20 l of cell lysis buffer. The lysed cells were immediately incubated with 0.5 mg/ml proteinase K (Sigma) and then with 0.5 mg/ml RNase A (Roche Molecular Biochemicals) at 50°C for 1 h each. The reactions were stopped by heat treatment at 70°C for 10 min. The samples were kept at 56°C, and 15 l of prewarmed sample buffer (1% low-melt agarose, 10 mM EDTA, 0.25% bromphenol blue, and 40% sucrose) was added. The samples were then loaded onto a 1.4% agarose gel. After electrophoresis, the DNA band pattern was visualized under UV light using ethidium bromide.
Determination of Ceramide Levels-Cellular ceramide levels were estimated by the diacylglycerol kinase assay and by employing charring densitometry of TLC plates.
Diacylglycerol Kinase Assay-Total cellular ceramide levels were quantified by the diacylglycerol (DAG) kinase assay as 32 P incorporated upon phosphorylation of ceramide to ceramide 1-phosphate by diacylglycerol kinase (30). PCC7-Mz1 cells were plated in 25-cm tissue culture dishes at a density of 2 ϫ 10 4 cells/cm 2 . The following day, cultures were treated with various factors; and at the indicated times, cells were washed twice with ice-cold PBS and harvested in glass tubes. After centrifugation (1000 ϫ g, 5 min, 4°C), lipids were extracted with 1 ml of chloroform/methanol/hydrochloric acid (1 N; 100:100:1, v/v/v), 170 l of buffered saline solution (135 mM NaCl, 4.5 mM KCl, 1.5 mM CaCl 2 , 0.5 mM MgCl 2 , 5.6 mM glucose, and 10 mM Hepes, pH 7.2), and 30 l of 100 mM EDTA. The lipids of the organic phase were transferred to a new glass tube and dried under a stream of N 2 . Lipid extracts were then subjected to mild alkaline hydrolysis (0.1 M KOH in methanol for 1 h at 37°C) to remove glycerophospholipids. 500 l of chloroform, 270 l of buffered saline solution, and 30 l of 100 mM EDTA were added. After drying the organic phase with N 2 , in vitro phosphorylation of extracted ceramide was performed as described by the manufacturer (RPN 200 kit, Amersham Pharmacia Biotech). 1 Ci of [␥-32 P]ATP (4000 Ci/mmol; ICN) was used to start the reaction. After 30 min at room temperature, the reaction was stopped by extraction of lipids with 1 ml of chloroform/ methanol/hydrochloric acid (1 N; 100:100:1, v/v/v), 170 l of buffered saline solution, and 30 l of 100 mM EDTA. The lower organic phase was dried under N 2 . The samples were resuspended in 30 l of chloroform/methanol (95:5, v/v) and spotted on Silica Gel 60 high performance thin-layer chromatography (HPTLC) plates (Merck, Darmstadt, Germany), which were activated in a saturated atmosphere of acetone shortly before usage. Ceramide 1-phosphate was resolved by TLC using chloroform/methanol/acetic acid (75:25:5, v/v/v) as solvent and migrated as a single spot at R F ϭ 0.25. Linearity of the assay was established using purified C 16 -ceramide (Sigma). Phosphorylation was detected and quantified by a phosphoimager (Bio-Rad 250) using the program Molecular Analyst.
HPTLC for Ceramide Detection and Charring Densitometry-Total lipids were isolated according to the method of Bligh and Dyer (31). For separation of ceramide from other neutral lipids, dried samples were resolved in chloroform/methanol (9:1, v/v) and spotted onto Silica Gel 60 HPTLC plates. Purified C 16 -and C 18 -ceramides were loaded as standards. The TLC plates were placed in an equilibrated chamber; and for ceramide analysis, the solvent system dichloromethane/methanol/acetic acid (100:2:5, v/v/v) was used. The TLC separation was followed by visualization and quantification of phospholipids by charring of the plates with cupric reagent as described previously (32). The plates were dried for 10 min at 180°C, cooled down to room temperature, and exposed to a solution of 10% copper(II) sulfate in 8% aqueous phosphoric acid for 15 s. After drying for 2 min at 110°C, charring was performed at 176°C for ϳ10 min. For quantification, the charred TLC plates were surveyed by two-dimensional laser scanning densitometry (Personal Densitometer, Molecular Dynamics, Inc.), and calculations were performed using the PC-BAS TINA program (Raytest).
Quantitation of Sphingomyelinase Activity in Vitro-Sphingomyelinase activity was assessed in vitro by the enzymatic micellar assay. PCC7-Mz1 cultures grown in 10-cm dishes were rinsed twice with ice-cold PBS, and cells were lysed in 150 l of lysis buffer (20 mM Tris-HCl, pH 7.4, 0.2% Triton X-100, 1 mM EDTA, 0.5 mM dithiothreitol, 1 mM ATP, 100 g/ml leupeptin, 100 g/ml aprotinin, 10 mM benzamidine, and 2 mM phenylmethylsulfonyl fluoride) by repeated squeezing through a needle. Concentrations of soluble proteins in extracts were estimated with the BCA method, and 10 g of protein was used for the selective determination of neutral Mg 2ϩ -dependent and Mg 2ϩ -independent and acid sphingomyelinase activities. The buffer for neutral Mg 2ϩ -dependent sphingomyelinase was 10 mM MgCl 2 , 0.2% Triton X-100, and 0.2 M Tris-HCl, pH 7.4; that for neutral Mg 2ϩ -independent was 0.2% Triton X-100 and 0.2 M Tris-HCl, pH 7.4; and that for acidic sphingomyelinase was 0.2% Triton X-100 and 0.2 M sodium acetate, pH 5.0. The reaction took place in a volume of 50 l and was started by adding 0.05 Ci of [methylcholine-14 C]sphingomyelin with a specific activity of 52 Ci/mmol (ICN) and a 350 g/ml final concentration of unlabeled sphingomyelin from bovine brain (Sigma). As a positive control, isolated neutral sphingomyelinase from Staphylococcus aureus (Calbiochem) was utilized instead of cell extracts. After incubation for 1 h at 37°C, the reaction was stopped by addition of 800 l of chloroform/methanol (2:1, v/v) and 250 l of H 2 O. After phase separation by centrifugation, radioactivity in 200 l of the aqueous phase was measured by scintillation counting using 5 ml of scintillation fluid (Zinsser, Frankfurt, Germany).
Ceramide Synthase Assay-1 day after plating, PCC7-Mz1 cells were washed three times with ice-cold PBS, scraped off the culture dish with a rubber spatula, and concentrated by centrifugation. Microsomal membranes were prepared as described (33). Briefly, cells were homogenized in 25 mM Hepes, pH 7.4, 5 mM EGTA, 50 mM NaF, and 10 g/ml each leupeptin and soybean trypsin inhibitor, and lysates were pelleted at 800 ϫ g for 5 min. The post-nuclear supernatant was centrifuged at 250,000 ϫ g for 35 min, and the microsomal membrane pellet was resuspended in 1.0 ml of homogenization buffer. For assaying ceramide synthase (sphinganine N-acyltransferase) activity, sphinganine and palmitoyl-CoA were used as substrates (33). The reaction mixture (in a total volume of 1. Serine Palmitoyltransferase Activity-The enzymatic activity of serine palmitoyltransferase was measured using [ 14 C]serine and palmitoyl-CoA as substrates (34,35). The assay mixture contained 0.1 M Hepes, pH 7.4, 5 mM dithiothreitol, 10 mM EDTA, 50 M pyridoxal 5Ј-phosphate, 1.2 mM L-[ 14 C]serine (1.6 Ci), 0.15 mM palmitoyl-CoA, and 100 -150 g of cell protein in a total volume of 100 l. After incubation for 10 min at 37°C, reactions were terminated by addition of chloroform/methanol (5:3, v/v). The lipids were extracted by phase separation and applied to a TLC plate, which was developed with chloroform, methanol, and 2 M NH 4 OH (40:10:1, v/v/v). The radiolabeled 3-ketosphinganine was quantified by phosphoimaging.
Northern Blot Analysis-For RNA preparation, cells were seeded at a density of 1.75 ϫ 10 4 cells/cm 2 in plastic dishes. At the indicated times after treatment with 0.1 M RA, cells were washed with ice-cold PBS and harvested. Total RNA was extracted using the RNeasy Total RNA kit (QIAGEN, Hilden, Germany) according to the manufacturer's instructions. 10 g of total cellular RNA was separated on 37% formaldehyde-containing 1.2% agarose gels and transferred onto Biodyne B membranes (Pall, Hamburg). After prehybridization, filters were incubated with cDNA fragments labeled by random priming in the presence of [␣-32 P]dCTP. The probes were a 807-base pair AvaI fragment of mLCB1 and a 739-base pair HindIII-PstI fragment of mLCB2 (36). The final washing conditions were 0.2ϫ SSC and 0.2% SDS at 60°C for 45 min (37). Blots were exposed to Kodak X-AR5 films with intensifier screens at Ϫ70°C for 2 days.

Ceramide-induced Apoptosis in PCC7-Mz1
Cells-Incubation of PCC7-Mz1 stem cells with the classical morphogen RA causes both apoptosis and neural differentiation in a dose-and time-dependent manner (12). Within 24 h, RA (0.1 M) induced the formation of apoptotic bodies, the DNA of which showed typical DNA laddering (Fig. 1A). The cleavage of genomic DNA into multimeric fragments of ϳ180 -200 base pairs is a hall-mark of apoptotic, but not necrotic, cell death. At the same time, expression of neuronal marker proteins like GAP-43 is induced in the adherent, differentiating PCC7-Mz1 cells (12,28). These findings establish that RA-induced apoptosis is due to the activation of a cellular differentiation program rather than to a cytotoxic effect.
To assess whether ceramide is involved in mediating the death signal originating from RA, we exposed cultures of PCC7-Mz1 stem cells to synthetic, short-chain, membrane-permeable ceramides. After 24 h of incubating PCC7-Mz1 cells with C 2 -ceramide (N-acetylsphingosine) and C 6 -ceramide (Nhexanoylsphingosine), 10 M each, genomic DNA of detached cells was isolated and analyzed by agarose gel electrophoresis. Both ceramides caused DNA laddering (Fig. 1A) in a dose-de-pendent manner, whereby C 6 -ceramide was slightly more efficient than C 2 -ceramide. Furthermore, C 2 /C 6 -ceramide activated within hours apoptosis-related cysteine/aspartatespecific proteases (caspases), demonstrated by cleavage of poly(ADP-ribose) polymerase, a typical substrate for this family of enzymes (data not shown). Thus, apoptotic cell death by ceramide was confirmed by DNA laddering and the activation of caspases. Ceramide application did, however, not induce expression of neuronal markers (data not shown).
Since DNA laddering is difficult to quantify, we employed a viability assay for precise estimation of the degree of cell death. The tetrazolium salt WST-1 becomes reduced to the purpleblue formazan by the dehydrogenase activity of the mitochondria of living cells (27). We added RA and C 6 -and C 2 -ceramides to PCC7-Mz1 cells for 24 h and evaluated the viability by the WST-1 assay. 10 M C 6 -and C 2 -ceramides caused a loss of viability of 23.1 Ϯ 5.2 and 20.6 Ϯ 4.3%, respectively, which was similar to 0.1 M RA (23.5 Ϯ 4.5%) (Fig. 1A, middle panel). The concentrations for the half-maximal effect (EC 50 ) of cell death were 17 Ϯ 2.5 and 30 Ϯ 4.3 M for C 6 -and C 2 -ceramides, respectively (n ϭ 3), and thus were relatively low compared with other cell systems (see Ref. 20). Adding C 6 -ceramide (10 M) and RA (0.1 M) together had a more than additive effect, with a loss of viability of 56.5 Ϯ 9.4% (n ϭ 3). These results imply a potential coupling of RA-and ceramide-activated apoptotic pathways.
Due to insolubility in aqueous solution, a mixture of C 16ceramides up to a final concentration of 100 M did not activate the suicide program in PCC7-Mz1 stem cells (Fig. 1A, right panel; and data not shown). However, when C 16 -ceramide (1 mM) was dispersed in a solvent mixture of ethanol and dodecane (98:2, v/v) prior to addition to the culture medium, a final concentration of 1 M C 16 -ceramide efficiently caused the death of 85% of the culture (EC 50 ϭ 0.5 Ϯ 0.2 M) (Fig. 1A, right  panel). Thus, natural long-chain ceramides, when delivered properly, are even more effective in induction of apoptosis than the short-chain C 2 /C 6 -ceramide analogs.
In addition to applying synthetic ceramides, we also used compounds from endogenous sources that caused an increase in cellular ceramide levels, which mimics a rather more physiological situation. Incubation of PCC7-Mz1 cells with the neutral sphingomyelinase from S. aureus, which cleaves sphingomyelin of the plasma membrane and thus increases the cellular ceramide level, was likewise able to induce apoptosis (EC 50 ϭ 1 Ϯ 0.25 units/ml; n ϭ 3) in PCC7-Mz1 stem cells. Thus, all ceramide-generating agents tested efficiently caused apoptosis.
To compare the kinetics of ceramide-and RA-induced cell death, PCC7-Mz1 stem cell cultures were treated with RA (0.1 M), C 2 -ceramide (10 M), C 6 -ceramide (10 M), and the biologically less active C 2 -dihydroceramide (10 M). Dihydroceramide is a naturally occurring ceramide that lacks the 4,5-trans double bond, but retains the stereochemical configuration; and the uptake and metabolism are very similar to those of D-erythroceramide (38). DNA of detached cells was analyzed after 3, 6, 12, 18, 24, and 48 h and compared with solvent-treated (0.05% Me 2 SO) control cultures (Fig. 1B). Both the C 2 -dihydroceramide-and solvent-treated cells showed only little, spontaneous DNA fragmentation, whereas the biologically active C 2 /C 6 -ceramides caused pronounced DNA fragmentation already after 18 h. It is noteworthy that, at that time point, RA-induced DNA degradation was hardly visible, but became obvious just a few hours later at ϳ24 h of RA incubation.
RA-induced Generation of Ceramide-We investigated whether RA, which acts by binding to nuclear receptors, fulfills some of its effects during early neurogenesis by generation of the second messenger ceramide. PCC7-Mz1 cells were incu- Left panel, PCC7-Mz1 stem cells were plated in 6-cm dishes at a density of 1.75 ϫ 10 4 cells/cm 2 . The following day, cultures remained untreated (control (Co)) or were treated with 0.1 M RA to induce differentiation and apoptosis (12). In parallel, cultures were incubated with 10 M membrane-permeable C 2 -ceramide (C2-Cer) and C 6 -ceramide (C6-Cer) as indicated. After 24 h, all detached cells in the supernatant of the dishes were collected by centrifugation. The genomic DNA was isolated, loaded, and separated on a 1.4% agarose gel and visualized after ethidium bromide staining by UV light. RA and C 6 -and C 2 -ceramides caused DNA laddering indicative of apoptotic cell death. The marker (M) was a mixture of phage DNA and pYH48 DNA digested with the restriction enzymes HindIII and AluI, respectively. Right panels, for quantification of cell death, PCC7-Mz1 cells were seeded in 96-well microtiter plates (5 ϫ 10 3 cells/well, i.e. 2 ϫ 10 3 cells/cm 2 ) 1 day before incubation with 0.1 M RA, 10 M C 6 -ceramide, and 10 M C 2 -ceramide as indicated. After 24 h, the light absorbance of the WST-1 assay reflecting viability of cultures was measured and is shown as a percentage of the signal obtained with cells having received the solvent (0.05% Me 2 SO (DMSO)) only (Х100%, control). The error bars represent means Ϯ S.E. of three independent experiments, with eight samples/ condition each. Viability was reduced to 77.4 Ϯ 8.8% after RA treatment, to 76.9 Ϯ 10.6% after C 6 -ceramide treatment, and to 79.5 Ϯ 11.5% after C 2 -ceramide treatment. Long-chain C 16 -ceramide applied directly to the culture medium at a final concentration of 1 M did not, because of its insolubility, cause cell death. However, when dispersed in a mixture of ethanol and dodecane, C 16 -ceramide (1 M) effectively caused loss of viability by 85% as depicted. B, time course of ceramide-induced DNA laddering. PCC7-Mz1 stem cells were incubated with RA (0.1 M); C 2 -ceramide, C 6 -ceramide, and C 2 -dihydroceramide (10 M each); and solvent (0.05% Me 2 SO) as indicated. Genomic DNA of detached cells was isolated at the time points depicted and loaded onto a 1.4% agarose gel. DNA laddering became detectable after 18 h of incubation with C 2and C 6 -ceramides and after 24 h with RA. Such a DNA laddering was not obvious when the biologically inactive C 2 -dihydroceramide or the solvent (Me 2 SO) for RA and ceramide was administered. DNA laddering was specifically induced by RA and C 2 /C 6 -ceramides. bated with RA (0.1 M) and solvent (0.05% Me 2 SO) for several time periods. At the indicated time points, cellular lipids were extracted, mildly hydrolyzed with alkaline, and phosphorylated with the diacylglycerol kinase in the presence of [␥-32 P]ATP; and phosphorylated ceramide was separated by TLC ( Fig. 2A, lower panel). The amount of phosphorylated ceramide was quantified using a phosphoimager and calculated per mg of protein. The examination revealed that, after a lag period of 3 h, the endogenous level of ceramide increased ϳ3 Ϯ 0.8-fold within 24 h. Ceramide levels remained at an elevated plateau for 3 days before they declined to near basal levels 5 days after RA-induced differentiation ( Fig. 2A, upper panel, OE). To prove that the drastic attenuation of ceramide levels within 1 day was due to RA treatment and not to unfavorable culturing conditions, PCC7-Mz1 cells were incubated with the corresponding amount of solvent (0.05% Me 2 SO) as a control. This administration had a minor effect and increased ceramide levels merely by 50% after 24 h ( Fig. 2A, upper panel, E). Therefore, we conclude that the rapid increase in cellular ceramide levels is specific for RA action and parallels the RA-induced processes of neural determination and differentiation.
The increase in ceramide levels by RA treatment was confirmed by a method alternative to the DAG kinase assay, viz. ceramide was visualized by charring densitometry (32,39). Lipids of PCC7-Mz1 cells treated for 24 h with RA (0.1 M) or with solvent (0.05% Me 2 SO) were isolated, separated by HPTLC, and visualized by charring the plates (Fig. 2B, right  panel). The various lipids were identified with the help of standards run in parallel, and their relative amounts were determined by laser scanning of the TLC plate. The level of cholesterol, which remained constant, served as endogenous reference. The analysis revealed an increase in ceramide levels after a 24-h RA treatment to 175 Ϯ 50% compared with control cells (Fig. 2B, left panel). The amount of ceramide increased from 3.0 to 5.23% of total lipids. Accordingly, we confirmed, by a different method, that RA treatment induces ceramide generation in PCC7-Mz1 stem cells.
We addressed the question whether elevated cellular ceramide levels are a consequence of the apoptotic process or a regulator of the process itself. Since DNA fragmentation, as shown in Fig. 1, is a quite late event in the course of apoptosis, an earlier biochemical event, i.e. activation of caspase-3, was investigated. Characteristic for caspase-3 activation is the cleavage of the 32-kDa proenzyme into fragments of 11 and 20 kDa, which can easily be detected by Western blot analysis. Fig. 2C demonstrates that active caspase-3 was not detectable during the first 8 h of RA treatment, i.e. during the phase of commitment to apoptosis (12) and when ceramide levels start to rise ( Fig. 2A). As a control, we investigated extracts of cells incubated for 24 h with RA, which clearly expressed the 20-kDa fragment (indicated on Fig. 2C). We conclude that ceramide production takes place during the initial phase of RA-induced apoptosis.
RA-induced de Novo Synthesis of Ceramide-Since the cellular level of ceramide appears to determine the degree of apoptosis, we tried to further increase the RA-induced increase in ceramide levels by inhibiting the catabolism of ceramide. Treating PCC7-Mz1 cells for 24 h with 500 M N-oleoylethanolamine (nOE), which blocks the ceramide-hydrolyzing enzyme ceramidase, enhanced the endogenous level of ceramide by ϳ3.5-fold compared with untreated control cells (Fig. 3A) and efficiently prompted apoptosis (Fig. 3B). Thus, the effects of nOE on ceramide levels and apoptosis were similar to those of 0.1 M RA. Furthermore, treating PCC7-Mz1 cells with nOE and RA together had an enhancing effect neither on ceramide levels nor on apoptosis in comparison with the efficacy of each compound alone (Fig. 3, A and B). Therefore, it was conceivable that RA, as nOE, may augment the ceramide level by reducing the turnover of ceramide.

FIG. 2. Determination of ceramide levels after RA treatment.
A, DAG kinase assay. To measure the endogenous level of ceramide, 2 ϫ 10 4 PCC7-Mz1 cells/cm 2 were plated in 25-cm culture dishes. Untreated cells (day 0) and cells treated for 3, 6, 9, 12, 15, 18, 21, 24, 48, 72, and 120 h with 0.1 M RA were harvested. As a control, cells were treated with only the solvent (0.05% Me 2 SO). Lipids were extracted and, after mild alkaline hydrolysis, subjected to phosphorylation by the DAG kinase in the presence of [␥-32 P]ATP as described under "Materials and Methods." The resulting ceramide 1-phosphate was separated by TLC and detected by autoradiography as shown in the lower panel. The start and the position of the phosphorylated ceramide (ceramide 1-phosphate) are marked on the right. For estimation of cellular ceramide content, the TLC plates were analyzed using a Bio-Rad phosphoimager. The radioactivity of ceramide 1-phosphate was calculated per mg of cellular protein and set as 1 for the PCC7-Mz1 stem cells (day 0). The upper panel shows that, during the first 24 h of RA treatment, the ceramide level increased by 3.5-fold, remained at a high level until day 3, and was attenuated by day 5. The results depicted show the means of four independent experiments, each performed in duplicate. B, charring densitometry. After treatment of PCC7-Mz1 cells for 24 h with RA, phospholipids were extracted, separated by HPTLC, visualized by charring, and analyzed by two-dimensional laser scanning (right panel). The positions of the lipids are marked on the right. Densitometric scanning (left panel) revealed an increase in C 16 -and C 18 -ceramides to 170% in RA-treated cells (RA) compared with untreated control cells (Co). Data are from a single experiment representative of two independent experiments performed in duplicate. C, analysis of caspase-3 activation. PCC7-Mz1 cultures remained untreated (0 h) or were treated with RA for the indicated time periods (h). Then, all cells of the culture were harvested and extracted; and 25 g of protein was separated by 12.5% SDS-polyacrylamide gel electrophoresis, followed by caspase-3 Western blotting as described under "Materials and Methods." Only cells treated for 24 h with RA displayed the active 20-kDa fragment of caspase-3.
Additionally, ceramides can be produced by two major pathways, viz. either by de novo synthesis or by hydrolysis of membranous sphingomyelin (14). To discriminate between these two pathways, the use of specific inhibitors has proven to be very helpful. Fumonisins belong to a recently discovered group of mycotoxins structurally related to sphinganine and sphingosine (40). In rat hepatocytes, fumonisin B 1 (FB 1 ) was shown to block a crucial step of de novo ceramide synthesis, catalyzed by sphingosine N-acyltransferase (ceramide synthase) (41). Treating PCC7-Mz1 cells with 25 M FB 1 for 24 h by its own did not alter the cellular ceramide level (Fig. 3A) or cause DNA fragmentation (data not shown) or alteration of viability (Fig. 3B). Adding FB 1 1 h before induction of neural differentiation by RA completely blocked the RA-induced increase in ceramide levels within 24 h, as was shown by the DAG kinase assay (Fig. 3A). The same result was independently obtained when applying the HPTLC charring method for ceramide detection (data not shown). Remarkably, the number of cells dying was drastically reduced, and the viability increased dose-dependently from 79% (only RA) to 91% (10 M FB 1 plus RA; data not shown) and 97% (25 M FB 1 plus RA) of untreated control cells (Fig. 3B). These results demonstrate that RA caused a significant elevation in ceramide levels arising from de novo biosynthesis and that interruption of this pathway efficiently prevents RA-induced apoptosis. Accordingly, ceramide production seems largely to mediate the RA-induced death signal during early neurogenesis.
Sphingomyelinase Activity-Another potential source of cel-lular ceramide is the hydrolysis of membranous sphingomyelin, which has been shown to be used by a variety of cytokines and stress factors (for review, see Ref. 20). Since FB 1 did not completely blocked the RA-induced increase in cellular ceramide levels (Fig. 3A), we investigated whether the action of SMases contributed likewise to the rise in ceramide levels. Therefore, we treated PCC7-Mz1 cells for 24 h with RA, prepared cell extracts, and measured the activities of the three types of SMase (neutral magnesium-dependent, neutral magnesiumindependent, and acidic) by the hydrolysis of [choline-methyl- 14 C]sphingomyelin using the appropriate buffers. The two neutral SMases (Mg 2ϩ -dependent and Mg 2ϩ -independent) showed little activity in untreated and RA-treated PCC7-Mz1 cells compared with the acid SMase, which largely contributed to total SMase activity (Fig. 4A). Its activity increased slightly within 5 h, but increased to 150 Ϯ 15% of control levels 24 h after applying RA (n ϭ 4). To evaluate to which extent the acid SMase contributed to the RA-induced cell death, the activity of acid SMase was specifically blocked by adding desipramine (0.1-10 M) (30) 1 h before applying RA (0.1 M) for 24 h. Concentrations Ͼ10 M proved to be cytotoxic for PCC7-Mz1 stem cells (Fig. 4B), but desipramine concentrations up to 5 M had no significant effect on cell viability and RA-induced apoptosis. However, 1 and 5 M desipramine efficiently reduced the cellular activity of the acidic sphingomyelinase to 56 and 29%, respectively, of untreated controls and completely inhibited the increase in acidic SMase activity by RA (Fig. 4C). Furthermore, we used extracts of cells treated with FB 1 as described (Fig. 3) and measured acidic sphingomyelinase activity. FB 1 up to concentrations of 25 M had no effect on the acidic sphingomyelinase reaction (data not shown). Since activation of acidic sphingomyelinase was not observed 5 h after RA treatment (Fig. 4A), when the cellular level of ceramide was already drastically elevated (Fig. 2), and because desipramine blocked acidic SMase, but not apoptosis, we conclude that the increase in ceramide levels leading to apoptosis of RA-treated PCC7-Mz1 cells is due to an activation of de novo ceramide synthesis (Fig. 3) and not to an activation of acid sphingomyelinase.
Ceramide Synthase and Serine Palmitoyltransferase Activities-We investigated the effect of RA on enzymes of the de novo ceramide synthesis pathway and whether FB 1 indeed obstructs ceramide synthase activity. Therefore, PCC7-Mz1 cells remained untreated or were incubated with RA in the presence or absence of FB 1 . After 24 h, cells were harvested, and proteins were extracted and analyzed for ceramide synthase activity (Fig. 5A). The levels of in vitro produced 14 Clabeled dihydroceramide were independent of treating the cells with RA and indicate that the increase in ceramide levels after RA treatment was not due to an augmentation of ceramide synthase activity. As anticipated, fumonisin B 1 (25 M) efficiently attenuated ceramide synthase action in the presence and absence of RA (Fig. 5A).
The long-chain base sphinganine originates from the condensation of serine and palmitoyl-CoA to 3-oxosphinganine (3ketosphinganine, 3-dehydrosphinganine) catalyzed by the key enzyme of sphingolipid biosynthesis, serine palmitoyltransferase (SPT). To evaluate the effect of RA on SPT, we analyzed SPT activity in extracts of untreated or RA-treated PCC7-Mz1 cells. SPT activity in extracts was measured by adding the radioactive SPT substrate [ 14 C]serine. Lipids were then extracted and separated by TLC, and the synthesized 3-ketosphinganine was detected by fluorography. Fig. 5B shows that SPT activity was 2.1-fold (100 g of protein) and 3.2-fold (150 g of protein) enhanced in extracts of RA-treated cells. Taken together, these results (Figs. 3 and 4) demonstrate that RA- induced changes in ceramide levels are mediated by an increase in de novo synthesis caused by enhanced SPT activity.
Expression of Serine Palmitoyltransferase-Since RA acts, after binding to nuclear receptors, directly on the transcription of many genes containing hormone-responsive elements (42,43), we studied whether the increase in de novo ceramide synthesis is due to the up-regulation of mRNAs coding for enzymes in this pathway. Murine SPT consists of two subunits, the catalytically active mLCB2 and the regulatory mLCB1, which are highly expressed in brain (36). Total RNA was isolated from PCC7-Mz1 stem cells and from cultures treated for 1 and 2 days with RA and was subjected to Northern blot hybridization analysis. The radioactively labeled mLCB1 (Fig.  6A) and mLCB2 (Fig. 6B) cDNA probes detected transcripts of 2.9 and 2.3 kilobases in length, respectively. Both mRNAs were constitutively expressed at a constant ratio and to similar degrees. Therefore, transcription of these SPT genes does not seem to be regulated by RA receptors.
Differentiating Cells Do Not Respond to RA and Ceramide-The finding that RA induced ceramide generation in PCC7-Mz1 cells and caused a sustained increase in the endogenous ceramide level for 3 days (Fig. 2) prompted us to examine whether RA and ceramide can still exert apoptotic effects on differentiating cells. PCC7-Mz1 stem cells (d0) and PCC7-Mz1 cells 1, 2, 5, and 7 days after induction of differentiation with RA were treated for 24 h either again with 0.1 M RA or with 10 M C 6 -ceramide or remained untreated as controls. The viability was determined by the WST-1 assay, and the results are shown in Fig. 7A. Whereas stem cell cultures lost ϳ20% of their viability by the RA and ceramide treatment, differentiating Cells were extracted, and ceramide synthase activity was analyzed by production of dihydrosphinganine as described under "Materials and Methods." As a negative control, the substrate sphinganine was omitted. Lipids were separated by TLC and visualized by autoradiography. The position of dihydroceramide is indicated. RA treatment did not affect the ceramide synthase activity, whereas fumonisin B 1 efficiently reduced its action. B, serine palmitoyltransferase activity. PCC7-Mz1 stem cells were cultured for 24 h in the absence (Ϫ) or presence (ϩ) of RA (0.1 M). Cells were then harvested, and serine palmitoyltransferase activity was determined as described under "Materials and Methods." A representative result of three different experiments run in duplicate each is depicted. Negative controls with heatinactivated cell protein (ϩ*) were run in parallel. Production of radioactively labeled 3-ketosphinganine was detected and quantified by scanning. There was a 2-3-fold increase in SPT activity upon RA treatment. cells at all time points (d1-d7) became completely resistant to both treatments. It is plausible that the ability to respond to ceramide is a prerequisite for RA-induced apoptosis.
To further investigate the link between RA-induced cell death and production of ceramide, we analyzed four independently isolated PCC7 subclones (RII/1, RII/2, RII/26, and RII/28) that are resistant to RA. These R-clones lost the potential to differentiate upon RA treatment, but retained a stem cell-like characteristic demonstrated by constant DNA synthesis, inability to express marker proteins, and failure to develop morphologically. 2 We investigated whether R-clones respond with apoptosis upon RA treatment. Neither a pronounced DNA laddering as seen in PCC7-Mz1 cells (Fig. 1A) nor loss of viability was observed upon an 24-h RA treatment of R-clones (Fig. 7B). These findings indicate that the capability to respond to RA with differentiation is connected with the ability to accomplish apoptosis. Addition of 10 M C 6 -ceramide, but not of 10 M dihydroceramide or 0.1 M RA, produced a pronounced DNA laddering in all R-clones (Fig. 6B). The EC 50 values of the four R-clones, evaluated by the WST-1 test, for induction of the programmed cell death were 23, 20, 17, and 22 M C 6 -ceramide and thus in the same order of magnitude as the original PCC7-Mz1 clone (17 Ϯ 2.5 M) (n ϭ 3). This means that R-clones, which neither fully differentiate nor die upon RA treatment, are still able to respond with apoptosis, but not with morphological differentiation (data not shown) upon exogenously applied ceramide. Hence, we investigated whether the R-clones were still capable of generating ceramide after RA treatment. Addition of 0.1 M RA to the R-clones did not produce a significant rise in the endogenous ceramide levels, whereas, in par-allel, the levels of PCC7-Mz1 control cells increased ϳ2.6-fold, as shown by the DAG kinase assay (Fig. 7B, right panel). Therefore, it is likely that the failure of R-clones to respond with ceramide production is responsible for their resistance to RA-induced apoptosis. Accordingly, the block of the RA-induced signal is upstream of ceramide generation, but downstream or at the level of the retinoic acid receptors and their targets.
Effect of Inhibition of Ceramide Synthesis on Neural Differentiation-It was shown in the neuroblastoma cell line Neuro2a that ceramide generation is involved in the regulation of RA-induced cell differentiation (44). We investigated the effect of inhibition of ceramide synthesis on RA-mediated neuronal determination by analyzing the induction of two early neuronal marker proteins, GAP-43 (28) and neuronal nClathrin (10). Parallel cultures of PCC7-Mz1 stem cells were treated for up to 3 days with 0.1 M RA in the presence or absence of 25 M FB 1 . Proteins were extracted, and the expression of GAP-43 and nClathrin was analyzed by Western blotting (Fig. 8). Whereas untreated cultures did not show any GAP-43 or nClathrin expression (d0), RA treatment induced an increase in the GAP-43 and nClathrin protein levels within 1 and 2 days, respectively, as described previously (28,10). This in- crease was not at all affected by the presence of the ceramide synthase inhibitor FB 1 (Fig. 8). Furthermore, microscopic examinations revealed that FB 1 prevented the RA-induced apoptosis, but did not sway morphological alterations that are typical for neuronal differentiation like formation of neurites (data not shown). We conclude that de novo synthesized ceramides are responsible for the induction of apoptosis; however, they do not change the determination and differentiation potential of the surviving neuronal cells.

DISCUSSION
The most common form of apoptosis in the developing nervous system (and consequently, the most well studied) involves developing neurons that are differentiated and have already established synaptic connections with afferent and efferent targets (2). More recently, studies have shown that apoptosis might occur much earlier during neurogenesis and affects individual neural cells as soon as hours after withdrawal from the cell cycle (5,45). Because of the historic emphasis on classic target-regulated neuronal cell death and the difficulty in characterizing and quantitating cell death among relative immature neurons or neuronal precursors, comparatively little is known regarding either the cellular and molecular mechanisms of this early programmed cell death or its biological significance.
Ceramide Mediates the Death Signal during Early Neural Differentiation-We have previously shown that treating proliferating stem cells of the teratocarcinoma cell line PCC7-Mz1 with physiological concentrations of retinoic acid promotes both development of neuroectodermal derivatives and apoptosis (9 -12). In the present study, this model system for neurogenesis served to dissect signal transduction pathways determining the fate of survival and death. We show that an increase in the cellular ceramide level is responsible for initiating early apoptosis. 1) Incubation of PCC7-Mz1 stem cells with membranepermeable C 2 -and C 6 -ceramides, but not with non-membranepermeable C 16 -ceramides or the biologically less active C 2dihydroceramide, caused apoptosis in a dose-and timedependent manner. 2) Apoptosis was triggered by adding bacterial sphingomyelinase to release ceramide from sphingomyelin of the cytoplasmic membrane and by applying the cer-amidase-specific inhibitor N-oleoylethanolamine. Both procedures caused an increase in the ceramide pool by utilizing endogenous sources. 3) PCC7-Mz1 stem cells, but not differentiating PCC7-Mz1 cells (Ͼd1), responded to both RA treatment and exogenously added C 2 /C 6 -ceramides in an apoptotic manner, implying that RA-and ceramide-activated pathways are interconnected. 4) C 2 /C 6 -ceramides caused significant DNA laddering after 18 h and thus several hours earlier than RA did (after 24 h), which may be due to a belated RA-mediated ceramide production. 5) Within a few hours of RA treatment, well before activation of caspases, the ceramide concentration steadily increased (up to 3.5-fold) and reached a maximal plateau after 24 h before declining after d3. These kinetics match perfectly the time course of apoptotic activity, which is high in the initial phase of neural differentiation, but ceases afterward when a stable neuronal network starts to develop (Ͼd3) (12). Additionally, we previously showed that at least 2 h of RA treatment are necessary to induce apoptosis (12). This period may be necessary to increase production of ceramide, which became obvious 3 h after RA treatment. 6) Inhibition of de novo synthesis of ceramide by blocking the ceramide synthase with fumonisin B 1 abrogated the RA-induced increase in ceramide levels and, concomitantly, apoptosis. However, inhibition of ceramide production did not affect neuronal differentiation.
Regulation of Ceramide Production-Ceramide is produced in various cell types in response to stimulation by several agonists with outcomes as diverse as cell proliferation, differentiation, growth arrest, and apoptosis (13, 46 -48). However, reports describing links between RA signaling and ceramide production were rare. In GH4C1 cells, RA concentrations able to inhibit cell proliferation caused a significant and prolonged increase in cellular ceramide content as a result of enhanced sphingosine N-acylation (49). In a recent study, treatment of the neuroblastoma cell line Neuro2a with RA to induce differentiation prompted activation of sphingomyelinase and of the de novo ceramide synthesis pathway (44). Elevated levels of ceramide caused morphological differentiation of Neuro2a cells, i.e. formation of neurites, but not cell death, seemingly contrasting with our results. However, one has to bear in mind the status of differentiation of the Neuro2a neuroblastoma cells and of the PCC7-Mz1 embryonic carcinoma cells used here. Although neuroblastoma cells are already committed to become neurons, the PCC7-Mz1 stem cells are multipotent and able to differentiate into the various derivatives of the neuroectoderm (9,11). We show that the lipid messenger ceramide is involved in regulating the degree of apoptosis in the very early phase of neural determination. At later stages, also reflected by neuroblastoma cells, ceramide may play a mediator role in the maturation of neurons. In this context, it is noteworthy that exogenous C 2 /C 6 -ceramides are unable to promote cell death in PCC7-Mz1 cells when cell lineage determination is complete, i.e. after 1 day of RA treatment (11). This resistance may be in part due to the up-regulation of the anti-apoptotic Bcl-2 protein during neural differentiation (12). Indeed, Bcl-2 overexpression in PCC7-Mz1 stem cells provides resistance to both RA-and C 2 /C 6 -ceramide-induced apoptosis (data not shown), implying that Bcl-2 acts as a signal transduction component downstream of ceramide production, as was previously shown in leukemia cells (50).
There are several cellular modes of increasing ceramide levels. Ceramide concentrations may arise either by hydrolysis of membrane-inserted sphingomyelin or by de novo synthesis at the endoplasmic reticulum. RA increased slightly the activity of the acidic (but not the neutral) SMase in PCC7-Mz1 cells after only 24 h, yet blocking acid SMase in PCC7-Mz1 cells had no effect on RA-induced apoptosis. The RA-induced increase in the ceramide level in PCC7-Mz1 cells was nearly completely abolished by inhibiting de novo synthesis with fumonisin B 1 . Enzymatic assays revealed that the activity of the ceramide synthase was unaffected, but the activity of SPT, a rate-limiting enzyme in de novo ceramide biosynthesis (51), was enhanced (ϳ2-3-fold) in RA-incubated cultures. This increase matches very precisely the rise in cellular ceramide concentrations. We investigated whether RA, which binds nuclear receptors, controls the expression of the recently cloned serine palmitoyltransferase genes (36). However, RA does not exert its effect by increasing levels of the two SPT mRNAs during RA-induced neural differentiation. Therefore, we envisage an RA-mediated post-translational mechanism regulating the activity of this enzyme.
Another potential mechanism of regulating the sphingomyelin cycle is by inhibiting enzymes that metabolize ceramides like glycosylceramide synthase and ceramidase (sphingolipidceramide N-deacylase), which hydrolyzes ceramide and produces sphingosine. Indeed, ceramidase, which is highly expressed in brain (52), may be an additional target enzyme for the action of RA during early neurogenesis. This notion is based on the findings that N-oleoylethanolamine, a specific inhibitor of ceramidase, increased endogenous ceramide levels and induced apoptosis to a similar degree as RA. On the other hand, the finding that blockage of ceramidase still allowed RA-induced apoptosis excludes the possibility that sphingosine has to be formed and is the actual mediator of apoptosis. In recent years, it has been shown in other cell systems that sphingosine 1-phosphate, formed by phosphorylation of sphingosine, can inhibit apoptosis, whereas sphingosine promotes apoptosis (53). However, our data strongly favor ceramide itself to be the mediator of apoptosis in RA-treated PCC7-Mz1 cells.
Apoptosis during Early Neural Development-Apoptosis that occurs at very early stages in nervous system development may shape the gross morphology and/or create a permissive environment for the growth of axons (2). However, it is less clear why individual neural precursor cells die at early phases of their development. One reason for abolishing neural precursor cells is the adjustment of an optimal cell density for accomplishing neurogenesis. We showed that the degree of RA-induced cell death is indeed dependent on the cell density of PCC7-Mz1 cultures and that this cell death is not due to lack of neurotrophic factors (12). Another reason may be to eliminate unwanted precursor cells, i.e. cells with inappropriate phenotypes. Thereby, apoptosis may play a crucial role in pattern formation. Consistent with this possibility, dying cells are observed in four distinct regions of the caudal neural tube of the developing chick embryo (7). These dying cells either became recently post-mitotic or are even still within the cell cycle, and this programmed cell death coincides with cellular differentiation and cell lineage formation of the neural tube along the dorsal-ventral axis. This concept is reflected in our model system of neural differentiation presented here. RA induced ceramide production and apoptosis in PCC7-Mz1 cells only within the first 24 h, the period of determination. At later stages of development, when mainly neuronal maturation processes occur (10,11), cells became insensitive to both RA and ceramide treatment. This finding is in line with the observation that the cell lineage of a multipotent PCC7-Mz1 cell is determined within the first 24 h of RA treatment. Only within this time window can the fate of a cell be shifted by the presence of other cells or when the cells are plated on laminin-coated surfaces (11).
An analogous phenomenon was observed during the early development of the mouse limb. Apoptosis increases in sites of naturally occurring cell death upon treatment of the embryo with RA. This effect is under tight regulation in that there is a specific temporal sensitivity of the cells to RA (54). Whether RA induces ceramide production in the dying cells of the developing limb remains to be elucidated.
Here we describe that, during very early neural differentiation, apoptosis (but not neurogenesis) is caused by an increase in ceramide levels due to activation of SPT. The PCC7-Mz1 system will allow the function of this distinct cell death, which might serve to control cell density and/or lineage determination, to be revealed by cell biological, genetic, and biochemical methods.