Characterization of Ceramide Synthesis

Ceramide (N-acylsphingosine) biosynthesis has been proposed to involve introduction of the 4,5-trans-double bond of sphingosine after synthesis of dihydroceramide (i.e. N-acylsphinganine). For the first time, the in vitro conversion of dihydroceramide to ceramide has been demonstrated using rat liver microsomes andN-[1-14C]octanoyl-d-erythro-sphinganine (st-H2Cer) and either NADH or NADPH as co-substrate; the apparent K m values for st-H2Cer and NADH were 340 and 120 μm, respectively. Molecular oxygen is required for enzymatic activity, and cyanide, divalent copper, as well as antibodies raised against cytochrome b 5are inhibitory, which suggests that this enzyme should be named dihydroceramide desaturase based on these similarities with the mechanism of Δ9-desaturase (stearoyl-CoA desaturase). Factors that influenced the activity of dihydroceramide desaturase include the alkyl chain length of the sphingoid base (in the order C18 > C12 > C8) and fatty acid (C8 > C18); the stereochemistry of the sphingoid base (d-erythro- >l-threo-dihydroceramides); the nature of the headgroup, with the highest activity with dihydroceramide, but some (∼20%) activity with dihydrosphingomyelin (activity was not detected with dihydroglucosylceramide, however); and the ability to utilize alternative reductants (ascorbic acid could substitute for a reduced pyridine nucleotide, but was inhibitory at higher concentrations). Dihydroceramide desaturase was inhibited by dithiothreitol, which suggests that it might be possible to alter ceramide synthesis by varying the thiol status of hepatocytes. Consistent with this hypothesis, when rat hepatocytes were cultured in varying concentrations of N-acetylcysteine (5 and 10 mm), there was a decrease in the relative incorporation of [14C]serine into [14C]ceramide. These studies have conclusively established the pathway of ceramide synthesis via desaturation of dihydroceramide and have uncovered several properties of this reaction that warrant further consideration for their relevance to both sphingolipid metabolism and signaling.

Over the last few years, it has become clear that ceramide plays important roles in the metabolism of cells. It is involved as a second messenger in what has become known as the sphingomyelin cycle (reviewed in Ref. 1) and, to name but a few, serves as a mediator of cellular senescence (2), apoptosis, and differentiation in many cell types (3). It was also observed that not only ceramide derived from the sphingomyelin pool found in the plasma membrane but also that from de novo synthesis is involved in the cellular responses to inducers of apoptosis (4) and differentiation (5). These results suggest an important role of the anabolic pathway of ceramides in signal transduction. It is therefore very likely that sphingolipid biosynthesis is tightly regulated. Taking into consideration that dihydroceramide does not mimic the effects of ceramide (6,7), one possible site of regulation might be dihydroceramide desaturase. In addition, ceramide is also involved in the transport of glycosylphosphatidylinositol-anchored proteins, for example in yeast (8).
The biosynthesis of the lipid anchor of all glycosphingolipids and sphingomyelin starts with the condensation of L-serine and palmitoyl-CoA catalyzed by the pyridoxal phosphate-dependent serine palmitoyltransferase (EC 2.3.1.50). Its product, 3-dehydrosphinganine, is immediately reduced by the NADPH-dependent 3-dehydrosphinganine reductase yielding D-erythrosphinganine (see Ref. 9, and references therein). Whether this sphingoid base is first desaturated to form sphingosine and then acylated to yield ceramide (reviewed in Ref. 9) or first acylated and then desaturated as proposed by several authors (9 -12) remained unclear for some time, although experimental results strongly favored the latter pathway. The discovery of fumonisin B 1 as a potent inhibitor of the N-acylation of sphingoid bases (12,13), helped us and others to demonstrate that, indeed, dihydroceramide is an intermediate in sphingolipid biosynthesis in many cell types (14,15), whereas sphingosine is not. Thus, D-erythro-sphinganine is first acylated and subsequent introduction of the 4,5-double bond by the dihydroceramide desaturase leads to the formation of ceramide.
The first three enzymes involved in sphingolipid biosynthesis are well characterized, and their subcellular localization and topology were found to be the cytosolic face of the endoplasmic reticulum (16,17). We describe here for the first time an in vitro assay for dihydroceramide desaturase activity, which gave us the possibility to characterize this enzyme and study its substrate specificity in vitro.
Chicken anti-rat cytochrome b 5 antibodies and chicken control IgG were kindly provided by Dr. John B. Schenkman (Department of Pharmacology, University of Connecticut, Farmington, CT). The activity and specificity against rat liver microsomes were shown by Western blot analysis using horseradish peroxidase-coupled anti-chicken IgG as secondary antibody. The anti-cytochrome b 5 antibodies bound exclusively to cytochrome b 5 , while treating the blot with the chicken control IgG showed no detectable staining.

Preparation of Rat Liver Microsomes
All solutions were prepared a day before use and stored at 0 -4°C. Male Wistar rats (50 -55 days old, 200 -250 g) were starved for 12 h and killed by decapitation. The livers were excised and washed twice in ice-cold sucrose/phosphate buffer (0.25 M sucrose, 100 mM NaH 2 PO 4 / Na 2 HPO 4 , pH 7.4). All subsequent procedures were carried out at 0 -4°C. Individual livers were weighed, minced with a scalpel, added to the same sucrose/phosphate buffer (0.5 g of liver/ml) and homogenized with 5 up-and-down strokes at 600 rpm in a Braun glass homogenizer with a loose-fitting Teflon pestle. The homogenate was centrifuged at 680 ϫ g for 10 min. The supernatant was saved, and the pellet was resuspended and centrifuged the same way as described above. The combined supernatants were centrifuged at 10,000 ϫ g for 10 min and the resulting supernatant at 105,000 ϫ g for 1 h. The pellet was resuspended in alkaline phosphate buffer (100 mM NaH 2 PO 4 /Na 2 HPO 4 , pH 8.0) (1 g of liver/ml). After centrifugation at 105,000 ϫ g for 1 h, the pellet finally was resuspended in phosphate buffer (100 mM NaH 2 PO 4 / Na 2 HPO 4 , pH 7.4) (2.5 g of liver/ml), frozen in aliquots in liquid nitrogen, and stored at Ϫ80°C.

Preparation of Silica Gel/Sodium Borate Thin Layer Chromatography Plates
Defatted glass plates were each coated with a mixture of 30 ml of double-distilled water, 0.8 g of Na 2 B 4 O 7 ⅐10H 2 O, and 14 g of Silica gel G 60 and were allowed to dry for 2 days at room temperature (18).

Preparation of 14 C-Labeled Substrates
The labeled substrates were synthesized by activation of 1-[ 14 C]fatty acid with N-hydroxysuccinimide and dicyclohexylcarbodiimide for subsequent acylation of the respective sphingoid base as described previously (19). The crude product mixtures were evaporated under a stream of nitrogen. Purification first involved separation by TLC with CHCl 3 / CH 3 OH/H 2 O (80:10:1, v/v) for Cer and H 2 Cer, CHCl 3 /CH 3 OH/2 N ammonia (65:25:4, v/v) for st-GlcCer and CHCl 3 /CH 3 OH/2 N ammonia (60:35:8, v/v) for st-SM as the developing systems. After detection by radioactive scanning, the spots were scraped from the TLC plate and re-extracted with CHCl 3 /CH 3 OH (2:1, v/v). Separation from free fatty acid was achieved by TLC with diethyl ether/CH 3 OH (99:2, v/v) as the developing system, scraping, and re-extraction with CHCl 3 /CH 3 OH (2:1, v/v).

Solubilization of the Lipid Substrate
Method A: BSA⅐Substrate Complexes-Defatted BSA⅐H 2 Cer complexes were prepared as described for other lipids (20), omitting dialysis to remove the ethanol (10%, v/v). The solution contained 30 nmol of BSA, 15 nmol of st-H 2 Cer, 10 l of ethanol, and 90 l of phosphate buffer (100 mM NaH 2 PO 4 /Na 2 HPO 4 , pH 7.4).
Method B: Solubilization Using Zwitterionic Detergent CHAPS-A solution containing 15 nmol of substrate was evaporated under a stream of nitrogen in a 1.5 ml test tube. A solution of 1.1 mg of CHAPS in 10 l of phosphate buffer (100 mM NaH 2 PO 4 /Na 2 HPO 4 , pH 7.4) was added, thoroughly mixed, and sonicated for 3 min.

Dihydroceramide Desaturase Assay
A microsomal suspension 2 containing 600 g of protein and a solution of 15 nmol of lipid substrate were added in a test tube, brought to a volume of 300 l with phosphate buffer (100 mM NaH 2 PO 4 /Na 2 HPO 4 , pH 7.4), and thoroughly mixed. All solutions were kept and mixed at 4°C. After a preincubation time of 5 min at 37°C, 1 mol of NADH in 30 l of phosphate buffer was added and the reaction mixture was incubated for 60 min at 37°C.
The reaction was terminated by addition of 200 l of CHCl 3 /CH 3 OH (83:17, v/v) on ice. Lipid extraction was achieved by addition of 343 l of CH 3 OH and 22 l of CHCl 3 , and vigorously mixing for 20 min. Phases were separated by centrifugation, and the lower phase was collected. The extraction procedure was repeated twice with 200 l of CHCl 3 / CH 3 OH (83:17, v/v), each.
The combined organic phases were evaporated under a steam of N 2 and dissolved in 30 l of CHCl 3 /CH 3 OH (1:1, v/v). The lipid mixture was separated by TLC using silica gel/sodium borate plates. The following were used as developing systems: CHCl 3 /CH 3 OH (9:1, v/v) for Cer/ H 2 Cer, CHCl 3 /CH 3 OH/2 N ammonia (65:25:4, v/v) for st-GlcH 2 Cer/Glc-Cer, and CHCl 3 /CH 3 OH/2 N ammonia (60: TLC was done twice to achieve a better separation of radiolabeled compounds. The lipids were visualized by autoradiography, and the R F values for st-H 2 Cer and st-Cer were 0.7 Ϯ 0.1 and 0.6 Ϯ 0.1, respectively (depending on the homogenity of the plates; however, a ⌬R F of at least 0.1 was always observed). The lipids were recovered from the plate by scraping, and the radioactivities determined by liquid scintillation counting in a Packard 1900CA Tri-Carb analyzer using Ultima Gold (Packard, Frankfurt, Germany) as the scintillation liquid, after allowing the samples to stand at room temperatur for 1 h. A background value of radioactivity, which was subtracted from each corresponding value, was obtained by scraping an area of about the same size in each lane of the plate containing no detectable radioactivity.

Sphingolipid Biosynthesis by Intact Hepatocytes
Rat hepatocytes were isolated and placed in primary culture, and sphingolipid biosynthesis from [ 14 C]serine was analyzed as described previously (22,23). Briefly, to each dish (approximately 1 mg of protein/ dish) was added 1 ml of Dulbecco's modified Eagle's medium containing 10 Ci of [ 14 C]serine (for a total serine concentration of 1 mM) and 0, 5, or 10 M N-acetylcysteine (or sodium acetate). After incubation for 12 h, the lipids were extracted and acid-hydrolyzed to release the radiolabeled sphinganine and sphingosine backbones, and the radiolabeled sphingoid bases were quantified by thin layer chromatography and scintillation counting (22,23).

Miscellaneous Procedures
Protein concentrations were measured by the method of Bradford (24) using bovine serum albumin as standard protein.

Presentation of Data
All data presented are means of at least duplicate determinations.

Conversion of Dihydroceramide to Ceramide by Rat Liver
Microsomes-The in vitro assay described under "Experimental Procedures" allowed us to characterize the enzymatic reaction that is responsible for converting dihydroceramide to ceramide. Examination of the lipid extracts from these assays by TLC (Fig. 1, lane 1) revealed that [ 14 C]st-Cer was produced from [ 14 C]st-H 2 Cer, and that these compounds accounted for at least 90% of the radiolabel. The identity of these compounds was confirmed by fast atom bombardment mass spectroscopy of the radioactive material that was recovered from the regions of the TLC plate marked st-Cer and st-H 2 Cer in Fig. 1. Characteristic peaks were observed at m/z 426 for MH ϩ , m/z 408 for MH ϩ -H 2 O, and m/z 448 for MNa ϩ , which were identical to those obtained from synthetic st-Cer. In addition, when the extracted st-Cer was reduced by H 2 /Pd(OAc) 2 (1.5 torr, 1 h) in methanol, it was converted to st-H 2 Cer. These procedures were applied to all the substrates used in this study, and the respective enzyme products were obtained.
The remaining radioactive products seen in Fig. 1 were identified as [1-14 C]octanoic acid (from cleavage of either the substrate or product by ceramidase, which is known to be present in microsomal preparations), and uncharacterized polar product(s) (which are likely to include [ 14 C]sphingomyelin). These products were not formed when the enzyme was denatured prior to use in the assay (Fig. 1, lane 2).
Thin layer chromatography of the products using CHCl 3 / CH 3 OH (9:1, v/v) as the developing system (which separates the cis and trans compounds by approximately 0.15 R F ) revealed that the double bond introduced by this enzyme had a trans-or (E)-configuration, which is the same stereochemistry as for ceramides that are produced in vivo.
Double Bond Formation Has the Characteristics of a Desaturase Rather than a Dehydrogenase-The conversion of st-H 2 Cer to st-Cer required a reductant; the apparent K m for NADH was 120 M, and NADPH worked equally well (data not shown). Ascorbic acid, another water-soluble electron donor, was able to substitute for a pyridine nucleotide in the assay; however, ascorbate also inhibited the activity with an apparent IC 50 of 10 M.
The requirement for an electron donor (rather than acceptor) suggests that this reaction is catalyzed by a desaturase rather than a dehydrogenase. If so, molecular oxygen would be required as the electron acceptor in the desaturation process, and an oxygen scavenger, such as pyrogallol (1,2,3-trihydroxybenzene), should be inhibitory. As shown in Fig. 2, replacing the air in the assay tube with argon caused a small reduction in enzymatic activity (the lack of complete inhibition is not surprising because the solutions were not purged of residual oxygen); moreover, the conversion of st-H 2 Cer to st-Cer was strongly inhibited by adding increasing concentrations of pyrogallol, with Ͼ90% inhibition at 30 mM (no activity was detected at concentrations higher than 30 mM). The inhibition by pyrogallol was less potent if the assays were conducted under an oxygen atmosphere rather than argon; thus, it can be concluded that molecular oxygen is required for enzymatic introduction of the double bond into st-H 2 Cer.
The characteristics of this activity (presence in microsomes, utilization of a reduced pyridine nucleotide, and requirement for oxygen) are similar to the microsomal stearoyl-coenzyme A desaturase (⌬ 9 desaturase, reviewed in Ref. 25). This enzyme is part of a multi-enzyme complex, which consists of NADHcytochrome b 5 reductase, cytochrome b 5 , and ⌬ 9 desaturase. The knowledge about this system and its inhibitors gave us the opportunity to study further the likelihood that conversion of st-H 2 Cer to st-Cer is catalyzed by an analogous desaturase. First, cyanide is known to inhibit the ⌬ 9 desaturase activity, because it is bound more tightly than O 2 , and we found that NaCN also inhibits the in vitro conversion of st-H 2 Cer to st-Cer with an IC 50 of 100 M. Second, divalent copper blocks electron transport to molecular oxygen in ⌬ 9 desaturase (by inhibiting the formation of intermediate superoxide anions from O 2 ) (26) and was inhibitory for dihydroceramide desaturation as well (with an IC 50 of 31 M). Third, antibodies raised against cytochrome b 5 have been found to inhibit plasmalogen desaturase utilizing the cytochrome b 5 electron transport system (27). Likewise, when rat liver microsomes were incubated for 30 min with different amounts of anti-cytochrome b 5 antibodies prior to the desaturase assay, formation of ceramide was inhibited by 48% up to 82% compared with controls when varying the dilution of the antibody serum from 1:1000 to undiluted, respectively (four different dilutions; data not shown).
Despite these similarities, an iron(II)-chelator (bathophenanthroline sulfonate) that inhibits ⌬ 9 desaturase by removing non-heme bound and catalytically important iron(II) (28) had no effect on dihydroceramide desaturation. Thus, dihydroceramide desaturase and ⌬ 9 desaturase are likely to have similar electron transport system(s), but are clearly distinct enzymes.
Utilization of Different Dihydroceramides by This Desaturase-A Lineweaver-Burk plot of the activity with varying concentrations of N-octanoyl-D-erythro-C 18 -sphinganine (st-H 2 Cer) (and 20 mM NADH) produced an apparent K m for this substrate of 340 M. Other dihydroceramide analogs that differed in chain length, stereochemistry, and type of headgroup were also analyzed (Fig. 3). 3 The in vitro activity decreased as the chain length of the amide-linked fatty acid of the H 2 Cer was increased (i.e. 18/8 Ͼ 18/12 Ͼ 18/18), presumably due to the greater insolubility of the longer chain homologs. No desaturation of the unmodified sphingoid base (sphinganine) was detected.
Somewhat surprisingly, dihydrosphingomyelin (DE-18/8-H 2 SM) was a relatively good substrate for the desaturase, yielding 20% of the activity with DE-18/8-H 2 Cer. In contrast, st-GlcH 2 Cer was not desaturated at a detectable rate in vitro, although there was considerable hydrolysis of GlcH 2 Cer by the microsomes (50 -60% of the total radioactive substrate added) and Cer was produced, presumably due to desaturation of the st-H 2 Cer. To exclude the possibility that GlcCer was formed from GlcH 2 Cer, but was not detected because it was hydrolyzed, the hydrolysis was inhibited with conduritol ␤-epoxide (a potent ␤-glucosidase inhibitor that reduced the hydrolysis to Ͻ10% of the total radioactivity), and no desaturation of st-GlcH 2 Cer was detected.
The conclusion that H 2 SM is a substrate for the desaturase is more definitive than the lack of activity with st-GlcH 2 Cer, because the latter might be affected by the physical state of the substrate or another artifact of the in vitro assay conditions. Nonetheless, these observations are consistent with studies that have been conducted in intact cells. Smith and Merrill (29) have reported that de novo sphingolipid biosynthesis from [ 14 C]serine by J774 cells initially produces large amounts of complex sphingolipids with a [ 14 C]sphinganine backbone, and which are converted to the more typical sphingosine-containing species over time. Another study of the metabolism of NBDlabeled sphingolipids by HT-29 cells 4 has found that NBD-H 2 Cer is initially incorporated into Glc-NBD-H 2 Cer, but direct desaturation of Glc-NBD-H 2 Cer was not observed. Thus, as was seen on our study of the desaturase activity in vitro, desaturation of GlcH 2 Cer actually reflected hydrolysis of the glycoconjugate to the (NBD-)H 2 Cer, which was oxidized and (in the case of the studies of the NBD-sphingolipids in intact cells) the glycosyl headgroup was added.
The addition of the product of the desaturase reaction, st-Cer, to the in vitro assay mixture (with DE-18/8-H 2 Cer as the substrate) was found to inhibit the activity of the desaturase with an IC 50 of 140 M. Future studies should determine whether this product inhibition is competitive or represents allosteric regulation of the desaturase.
Other Characteristics of Dihydroceramide Desaturase Activity in Vitro-The activity of dihydroceramide desaturase is relatively stable: approximately 50% of the activity was lost after 1 year of storage at Ϫ80°C, or 7 days at 4°C, or 3 h at 37°C. When the assays were conducted at varying temperature, activity could be detected at 5°C (about 8% of maximal activity) and steadily increased with the incubation temperature until a maximum of 43°C (the half-maximal enzymatic activities were at 27°C and 47°C); above 60°C, no activity was detected (data not shown).
Optimal activity was obtained over a fairly wide pH range (pH 6.5-9) (data not shown). A much lower activity was obtained at pH Ͻ 6.5; however, this may reflect a buffer effect (at least in part) because there was a large drop in activity of about 80% at pH 6 when the assay was conducted with citrate buffer instead of phosphate buffer. Nevertheless, further decrease of citrate buffer pH value (down to pH 4) resulted in further reduction of activity. When the assays were conducted at neutral pH, the same activities were found when HEPES/NaOH, MOPS/KOH, NaH 2 PO 4 /Na 2 HPO 4 , or glycine/NaOH were used over a range of concentrations (50 -300 mM); however, the activity using Tris/HCl was reduced by approximately 30%.
Addition of various cations (CaCl 2 at 0.1-100 M, MgCl 2 at 1 M to 1 mM, MnCl 2 at 1 M to 1 mM, FeCl 2 at 1-100 M, KCl at 10 -150 mM, NaCl at 10 -100 mM, Na 2 SO 4 at 10 -100 mM, and EDTA at 10 -100 mM) had no effect on the activity, whereas ZnCl 2 (at 1 mM) decreased the activity by 25% (data not shown). Thus, the enzyme is tolerant to many inorganic cations. In contrast, the activity was inhibited 81% by addition of 1 mM dithiothreitol (DTT), a common reagent used to protect protein thiol groups from oxidation. This may explain why this reaction has not been observed in previous studies since 10 mM DTT is commonly used in preparing microsomes. Once this observation was made, our protocol omitted DTT from the procedure. A possible explanation for this effect is that dihydroceramide desaturase may have one (or more) disulfide bonds that is (are) required for protein stability and/or catalytic activity.
Conversion of Dihydroceramide to Ceramide by Intact Rat Hepatocytes-Previous studies have shown that radiolabeled serine is incorporated into Cer and SM by intact rat hepatocytes with little accumulation of H 2 Cer (22,23); therefore, the desaturase reaction is not rate-limiting in these cells (at least, under the conditions used). Since dihydroceramide desaturase is inhibited by DTT, we examined whether elevation of cellular thiols would suppress the formation of ceramide(s) by cultured rat hepatocytes. These experiments were conducted by incubating the hepatocytes with increasing concentrations of Nacetyl-L-cysteine, which is taken up by hepatocytes and deacetylated to increase the levels of intracellular thiols (glutathione and protein thiols) (30 -33). As shown in Fig. 4, Nacetyl-L-cysteine did not cause the accumulation of label in H 2 Cer, but reduced the amount of label in sphingosine-containing sphingolipids by approximately half. Since dihydroceramides have been found to accumulate in several other cell systems (9,29), it would be interesting to know if this is due to a difference in the levels of dihydroceramide desaturase per se or its modulation by factors such as the cellular thiol status.
Conclusion-For the first time, it has become possible to demonstrate the in vitro conversion of dihydroceramide to ceramide, to assign the name dihydroceramide desaturase to the enzyme that catalyzes this reaction, and to delineate the pathway for introduction of the 4,5-trans-double bond of sphingosine as shown in Fig. 5 (i.e. at the level of dihydroceramide and dihydrosphingomyelin). This not only fills a missing "gap" in the pathway of sphingolipid biosynthesis, it also enables more sophisticated studies of how cells regulate the formation of ceramide de novo. Control of this step of sphingolipid metabolism is clearly important because ceramide affects cell growth, differentiation, diverse cell behaviors, and programmed cell death, and induction of ceramide synthesis by some chemotherapeutic agents (such as daunorubicin) has been proposed to mediate their toxicity (3). Since dihydroceramides are much less potent than ceramides in many signaling events (34), the finding that dihydrosphingomyelin can serve as a substrate for the desaturase suggests that cells could minimize the amount of free ceramide that is produced during membrane biogenesis by utilizing the so far hypothetical pathway dihydroceramide 3 dihydrosphingomyelin 3 sphingomyelin. The existence of this pathway in vivo also depends on the topology of the enzymes involved. Sphingomyelin formation has been assigned to the luminal side of the Golgi membranes (35), whereas desaturation of dihydroceramide presumably occurs at ER membranes with so far unknown topology.
The sensitivity of dihydroceramide desaturase to thiols is also intriguing because thiols protect against cell injury by many agents (30 -33), whereas depletion of glutathione can lead to cell death. The assay system described in this paper will allow further characterization of this important enzyme and elucidation of its role in cell regulation.
FIG. 5. The pathway for de novo sphingolipid biosynthesis. Sphingolipid biosynthesis begins with the condensation of serine and palmitoyl-CoA followed by reduction of 3-ketosphinganine, and acylation of sphinganine to dihydroceramide. Introduction of the 4,5-transdouble bond of sphingosine into dihydroceramide (or dihydrosphingomyelin) is catalyzed by dihydroceramide desaturase, as shown. The diagram also illustrates how altering the cellular thiol state might affect signaling pathways that involve ceramide. It is by no means clear whether the step marked with an asterisk also occurs in vivo, because the subcellular site and topology of dihydroceramide desaturase is not known.