Acylation of Naturally Occurring and Synthetic 1-Deoxysphinganines by Ceramide Synthase

Fumonisin B1 (FB1) is the predominant member of a family of mycotoxins produced byFusarium moniliforme (Sheldon) and related fungi. Certain foods also contain the aminopentol backbone (AP1) that is formed upon base hydrolysis of the ester-linked tricarballylic acids of FB1. Both FB1 and, to a lesser extent, AP1 inhibit ceramide synthase due to structural similarities between fumonisins (as 1-deoxy-analogs of sphinganine) and sphingoid bases. To explore these structure-function relationships further, erythro- and threo-2-amino, 3-hydroxy- (and 3, 5-dihydroxy-) octadecanes were prepared by highly stereoselective syntheses. All of these analogs inhibit the acylation of sphingoid bases by ceramide synthase, and are themselves acylated with V max/K m of 40–125 for the erythro-isomers (compared with approximately 250 for d-erythro-sphinganine) and 4–6 for the threo-isomers. Ceramide synthase also acylates AP1 (but not FB1, under the conditions tested) to N-palmitoyl-AP1 (PAP1) with a V max/K m of approximately 1. The toxicity of PAP1 was evaluated using HT29 cells, a human colonic cell line. PAP1 was at least 10 times more toxic than FB1 or AP1 and caused sphinganine accumulation as an inhibitor of ceramide synthase. These studies demonstrate that: the 1-hydroxyl group is not required for sphingoid bases to be acylated; both erythro- andthreo-isomers are acylated with the highest apparentV max/K m for theerythro-analogs; and AP1 is acylated to PAP1, a new category of ceramide synthase inhibitor as well as a toxic metabolite that may play a role in the diseases caused by fumonisins.

Fumonisin B 1 (FB 1 ) is the predominant member of a family of mycotoxins produced by Fusarium moniliforme (Sheldon) and related fungi. Certain foods also contain the aminopentol backbone (AP 1 ) that is formed upon base hydrolysis of the ester-linked tricarballylic acids of FB 1 . Both FB 1 and, to a lesser extent, AP 1 inhibit ceramide synthase due to structural similarities between fumonisins (as 1-deoxy-analogs of sphinganine) and sphingoid bases. To explore these structure-function relationships further, erythro-and threo-2-amino, 3-hydroxy-(and 3, 5-dihydroxy-) octadecanes were prepared by highly stereoselective syntheses. All of these analogs inhibit the acylation of sphingoid bases by ceramide synthase, and are themselves acylated with V max /K m of 40 -125 for the erythro-isomers (compared with approximately 250 for D-erythro-sphinganine) and 4 -6 for the threo-isomers. Ceramide synthase also acylates AP 1 (but not FB 1 , under the conditions tested) to N-palmitoyl-AP 1 (PAP 1 ) with a V max /K m of approximately 1. The toxicity of PAP 1 was evaluated using HT29 cells, a human colonic cell line. PAP 1 was at least 10 times more toxic than FB 1 or AP 1 and caused sphinganine accumulation as an inhibitor of ceramide synthase. These studies demonstrate that: the 1-hydroxyl group is not required for sphingoid bases to be acylated; both erythro-and threo-isomers are acylated with the highest apparent V max /K m for the erythro-analogs; and AP 1 is acylated to PAP 1 , a new category of ceramide synthase inhibitor as well as a toxic metabolite that may play a role in the diseases caused by fumonisins.
Fumonisins are mycotoxins produced by Fusarium moniliforme (Sheldon) and related fungi that are common contaminants of maize and certain other foods (1). Fumonisin B 1 (FB 1 ) 1 is comprised of a long chain aminopentol (AP 1 ) with two of the side chain hydroxyls esterified to tricarballylic acids ( Fig. 1) (2,3). Although numerous fumonisins have been characterized (1)(2)(3)(4), FB 1 is usually the most abundant in contaminated food, except when corn has been treated with base to produce masa flour for tortillas, which hydrolyzes FB 1 to AP 1 (1).
Fumonisins are responsible for at least two diseases, equine leukoencephalomalacia and porcine pulmonary edema (5,6). Studies with these and other animals have uncovered a wide spectrum of toxicologic effects, which include hepatotoxicity, nephrotoxicity, neurotoxicity, developmental toxicity, and immunosuppression (and immunostimulation under some conditions) (for reviews, see Refs. 1, 7, and 8). Fumonisins are also hepatocarcinogenic in rats (1) and have been implicated in esophageal cancer in humans in South Africa and China (9 -11).
The diverse effects of these mycotoxins appear to be due to inhibition of ceramide synthase, the enzyme that catalyzes the acylation of sphinganine, sphingosine, and other sphingoid bases (12,13). Inhibition has been shown in vitro with intact cells and for animals exposed to fumonisins (7) and is manifested by up to several hundredfold increases in cellular levels of sphinganine (and sometimes other sphingoid bases) (8,14,15). This elevation in sphinganine, a highly bioactive compound, initiates a cascade of cellular alterations that are thought to be largely responsible for the toxicity and carcinogenicity of this mycotoxin (7, 16 -19).
Removal of the tricarballylic acids groups from FB 1 reduces the potency of inhibition of ceramide synthase by approximately 10-fold (13,19). This observation, plus the kinetic pattern for inhibition (20), suggests that ceramide synthase recognizes both the AP 1 moiety (as a sphingoid base analog) and the tricarballylic acids side chains (presumably, as analogs of the phosphate groups of fatty acyl-CoA) (13). The toxicity of AP 1 for cells in culture mirrors this reduced potency (19); however, feeding studies have found that AP 1 causes hepatic and renal lesions in rats that are indistinguishable from those caused by FB 1 (21,22), which suggests that the mechanism(s) of action of AP 1 versus FB 1 are not yet fully understood.
To characterize further the properties of the AP 1 backbone, a series of 2-amino-3-hydroxy-(and 3, 5-dihydroxy-) octadecanes have been synthesized, and this report describes the somewhat unexpected finding that these analogs not only inhibit ceramide synthase but also serve as alternative substrates. Because these findings indicated that the aminopentol backbone of fumonisins may be acylated by ceramide synthase, the formation of N-palmitoyl-AP 1 (PAP 1 ) was also characterized. Finally, PAP 1 was found to be highly toxic for HT29 cells, apparently as a new category of ceramide synthase inhibitor.

EXPERIMENTAL PROCEDURES
Materials-The synthetic fumonisin and sphingoid base analogs were prepared by methods that will be described in another publication. 2 FB 1 was isolated from culture material, and AP 1 was synthesized from FB 1 by basic hydrolysis, as described by Wilson et al. (23). Radiolabeled compounds were purchased from Amersham Pharmacia Biotech ([1- 14
Synthesis of N-Palmitoyl-FB 1 -FB 1 (1 mg, 1.38 mol) and 4.4 mg of palmitic anhydride (6.92 mol) were dissolved in 1 ml of methanol (MeOH):CHCl 3 (1:1, v/v) and kept at room temperature for up to 48 h. The reaction was checked by TLC using silica 60 TLC plates (Merck, EM Separations, Gibbstown, NJ) and CHCl 3 :MeOH:H 2 O (90:20:0.5) as the developing solvent. To detect fumonisins, the plates were sprayed with a 0.5% p-anisaldehyde solution in MeOH:acetic acid:sulfuric acid (85:10:5), heated to 100°C for 10 min, and visually examined. The clean-up was carried out by extraction with n-butanol as described for the ceramide synthase assay of FB 1 (see below). Results: yield 40 -50%; Synthesis of N-Palmitoyl-AP 1 (PAP 1 )-AP 1 (1.1 mg, 2.7 mol) and 5.3 mg of palmitic anhydride (10.7 mol) were dissolved in 1.5 ml of MeOH: CHCl 3 (1:1) and kept at room temperature for up to 48 h. The reaction was checked by TLC as described above. The reaction mixture (0.5 ml) was applied to a SiO 2 SPE column (500 mg capacity; Waters, Milford, MA), which was preconditioned with CHCl 3 . The minicolumn was washed with 6 ml of CHCl 3 followed by 6 ml of CHCl 3 /MeOH (95:5). Assays of Ceramide Synthase-Ceramide synthase was assayed as described by Merrill and Wang (24) by following either acylation of [ 3 H]sphingosine to ceramide (for inhibition studies) or acylation of the sphingoid base analog using a radiolabeled fatty acyl-CoA. For the studies with FB 1 or AP 1 , the reaction mixture (totalling 100 L) contained 25 mM potassium phosphate buffer, pH 7.4, 0.5 mM dithiothreitol, 50 M [1-14 C]palmitoyl-CoA, varying concentrations of FB 1 or AP 1 (1-10 M), and 150 g of microsomal protein. The enzymatic reaction mixture was incubated in a shaking water bath for 15 min at 37°C. In the case of AP 1 as substrate, the reaction was stopped by the addition of 1 ml of MeOH and 0.5 ml of CHCl 3 . For lipid extraction, 1 ml of CHCl 3 and synthetic PAP 1 (ϳ25 g, as carrier) was added. The reaction mixture was vortexed, approximately 3 ml of slightly basic water was added, and the mixture was vortexed again. The CHCl 3 layer was washed, dried, and evaporated as described above. In the case of FB 1 as substrate, the reaction was stopped by the addition of 1 ml of CHCl 3 , then, ϳ25 g of synthetic N-palmitoyl-FB 1 was added as carrier, followed by 2 ml of slightly basic water. The reaction mixture was vortexed, centrifuged, and the CHCl 3 layer was again washed with slightly basic water, dried, evaporated, and analyzed by TLC. The aqueous phase was acidified (to pH 4 -5) by adding 0.1 N HCl, and extracted two times with 1 ml of n-butanol. The combined butanol layers were dried over Na 2 SO 4 , evaporated to dryness using a Speed Vac (Savant Instruments, Farmingdale, NY) and analyzed by TLC.
Analyses of the Ceramide Synthase Assay Products by TLC-Samples were dissolved in MeOH:CHCl 3 (1:2) and spotted onto type 60 silica TLC plates that were developed using CHCl 3 :MeOH (90:10, solvent 1) for PAP 1 , diethylether:MeOH (99:1, solvent 2) for ceramides, CHCl 3 : MeOH:H 2 O:acetic acid (90:20:0.5:1, solvent 3) for N-palmitoyl-FB 1 . After the plates were developed and air dried, the products were visualized by spraying with p-anisaldehyde solution, and the radiolabeled regions of the plate were detected using a Bioscan System 200 Image Scanner (Bioscan Inc., Washington, DC). The regions that migrated coincident with the standards were collected and counted using a detergent-containing scintillation mixture for 10 min in a Beckmann LS 7000 scintillation counter (Beckman Instruments, Palo Alto, CA).
Cell Culture-HT29 cells (ATCC, Rockville, MD) were seeded at 1 ϫ 10 5 cells per 100-mm dish (Corning, Cambridge, MA), and grown in 8 ml of Dulbecco's modified Eagle medium (Sigma) supplemented with 10% fetal calf serum (HyClone, Logan, UT), 3.5 g/liter glucose, and 10 ml/liter of a solution of 6.1 mg/ml penicillin G and 10 mg/ml streptomycin (Sigma) in an incubator at 37°C and a humidified atmosphere of 5% CO 2 . The medium was changed 24 h after seeding, then the cells were incubated with Dulbecco's modified Eagle's medium containing PAP 1 (solubilized in 10 l of ethanol:dodecane, 98:2; addition of this volume of ethanol:dodecane had negligible effects on cell viability). After 24 h, the cells were harvested with 0.25% trypsin in 0.05% EDTA, and viable cells were determined by trypan blue exclusion.
To determine the effects of fumonisins on cellular sphinganine and ceramides, the cells were treated in the same manner, and the lipids were extracted and analyzed by HPLC as described by Schmelz et al. (19).
Statistical Analyses-Analyses were conducted in triplicate unless otherwise noted, and the statistical significance of differences between groups was evaluated by the Student's t test using the InStat statistical package (GraphPad Software). The kinetic parameters were derived from linear regression analyses of the data.

Inhibition of Ceramide Synthase by Fumonisin Analogs-All
of the fumonisin analogs inhibited the acylation of D-erythro-[ 3 H]sphingosine (Fig. 2), but were less potent than FB 1 . These results agree with the previous demonstration that FB 1 is more potent than AP 1 which, like these analogs, lacks the tricarballylic acid side chains (13). Upon more detailed analyses of the products of the assays, we discovered that the analogs also served as substrates; therefore, the kinetic parameters of acylation of these compounds were analyzed (Fig. 3). Of the six fumonisin analogs tested, XXa had the highest apparent V max , which was comparable with the naturally occurring D-erythrosphinganine and -sphingosine (13,25). Similarly, the other erythro-analogs (Va and XXb) were acylated reasonably well, and the threo-compounds (Vb, XIVa and XIVb) were also acylated, but with much lower apparent V max . Therefore, ceramide synthase does not require a 1-hydroxyl group on the sphingoid base substrate and as has been seen in analyses of the stereoisomers of sphingosine and sphinganine (25) will accommodate both threo-and erythro-stereoisomers, although the latter are better substrates. The ability of this enzyme to acylate these fumonisin analogs raised the possibility that FB 1 or AP 1 might also be acylated, therefore, this was investigated. Assays of Ceramide Synthase with AP 1 and FB 1 -Since PAP 1 is similar to natural ceramides in polarity, the products were extracted using CHCl 3 :MeOH (24); however, N-palmitoyl-FB 1 bears two tricarballylic acids groups (with four charged carboxyl groups) and required a different extraction protocol. Briefly, the reaction mixture was extracted with CHCl 3 to remove nonpolar lipids, then the aqueous phase was acidified with HCl and extracted with n-butanol to obtain N-palmitoyl-FB 1 (for details see "Experimental Procedures"). The success of the extraction was monitored by TLC and MS, and most of the carrier N-palmitoyl-FB 1 was found in the n-butanol layer (attempts to purify N-palmitoyl-FB 1 by other approaches, such as using reversed phase or anion exchange cartridges, were not successful).
As shown in Fig. 4, incubation of rat liver microsomes with 10 M AP 1 and [1-14 C]palmitoyl-CoA produced a radiolabeled product that migrated at the same position on the chromatoplate (R f ϳ0.2) as the reference PAP 1 (radiolabel was not found in this region of the chromatoplate when AP 1 was omitted from the assay). Co-migration of this labeled product and the PAP 1 standard on TLC was also obtained using several other solvent systems (not shown). To confirm that this product was PAP 1 , the assay mixture was extracted and the product analyzed by HPLC-ES-MS/MS using a reversed phase column and a water/ acetonitrile gradient. We were able to detect the protonated molecular ion in the single ion monitoring mode at m/z 644. 8 [MϩH] ϩ . Since only low amounts of PAP 1 were produced in these enzymatic syntheses, we were not able to obtain daughter ion spectra; however, the m/z 644.8 [MϩH] ϩ signal showed the same HPLC retention time as the reference compound, which is strong evidence for the identity of this product.
The concentration dependence for PAP 1 formation fits a simple Michaelis-Menten relationship (Fig. 5). The apparent K m was 3.4 M and V max was 4.0 pmol/min/mg microsomal protein for AP 1 (Fig. 5). Thus, the V max /K m for AP 1 is approximately 1, which is on the same order of magnitude as for the threofumonisin analogs (Vb, XIVa and b).
When 10 M FB 1 was incubated with rat liver microsomes and [1-14 C]palmitoyl-CoA, no radiolabel was detected in the region of the N-palmitoyl-FB 1 standard, therefore, FB 1 was not acylated under these assay conditions.
Effects of PAP 1 on HT29 Cells-To determine whether PAP 1 is cytotoxic, this compound was incubated with HT29 cells, a human colonic cell line that is sensitive to FB 1 and AP 1 (19). All of the tested concentrations of PAP 1 (1-50 M) caused a significant reduction in the number of viable cells within 24 h (cell death was noted as early as 4 h after addition of 50 M PAP 1 ) (Fig. 6). In other studies (19), we have shown that 50 M FB 1 and AP 1 reduced cell number by 50 and 32%, respectively, after 24 h. 3 Therefore, PAP 1 appears to be considerably more toxic for HT29 cells than either of the parent fumonisin precursors.
Since the toxicity of fumonisins has been associated with the accumulation of sphinganine, the effects of PAP 1 on free sphinganine and ceramide were determined (Fig. 7). The elevation in sphinganine induced by 1 M PAP 1 (17- no fumonisin control) was the same as that for 50 M AP 1 , and the elevation induced by 5 M PAP 1 (148-fold versus the no fumonisin control) was approximately twice the fold increase induced by 50 M FB 1 (68-fold). None of these treatments altered the ceramide levels of the cells over this period (Fig. 7), which is consistent with the sphinganine accumulation arising from inhibition of ceramide synthase, rather than activation of a ceramidase.
Inhibition of Ceramide Synthase in Vitro by PAP 1 and N-Palmitoyl-XIVa-As predicted by these findings with HT29 cells, PAP 1 (and the N-palmitoyl derivative of the synthetic analog that most resembles AP 1 , XIVa) inhibited ceramide synthase in vitro, as shown in Fig. 8. Both caused ϳ50% inhibition at ϳ9 M; for comparison, FB 1 caused 50% inhibition at 6 M in parallel assays conducted with this enzyme preparation and reagents (data not shown).

DISCUSSION
The inhibition of ceramide synthase by FB 1 and other fumonisins has been attributed, in large part, to similarities between the AP 1 and the sphingoid base substrates for this enzyme (1, 7). Nonetheless, this analogy is not fully convincing because AP 1 differs from natural sphingoid bases in ways that are often important to enzyme-ligand interactions, namely the stereochemistry in the vicinity of the group to be modified (i.e. the 2-amino and 3-hydroxyl groups are threo for fumonisins and erythro for sphingoid bases), the absence of a hydroxyl group at position 1 (which might be important in binding, or orientation of the amino group for acylation), and the presence of an additional hydroxyl at position 5 (which might impose bulk or polarity in a key region of the enzyme active site). The structure-function analyses presented in this report alleviate these concerns by establishing that both erythro-and threo-1-deoxy, 5-hydroxy-sphinganines inhibit ceramide synthase and serve as reasonably good substrates.
These findings extend the recent evaluation of the stereoisomers of sphinganine and sphingosine as substrates for ceramide synthase (25), which ranked the stereoisomers (based on the apparent V max /K m ): D-erythro-(2S,3R) Ͼ D-threo-(2R,3R) Ͼ L-threo-(2S,3S) Ͼ L-erythro-(2R,3S). The V max /K m only differed by approximately 20-fold for naturally occurring D-erythrosphinganine versus L-threo-sphinganine 4 (Fig. 3) and arose from a 5-fold higher V max and a 4-fold lower K m . These results also agree with a much earlier finding by Stoffel and Bister (26)  that radiolabeled ceramides can be found in rat liver after intravenous injection of any of the radiolabeled stereoisomers of sphinganine. The reason(s) for this tolerance of multiple structural modifications are not known, 5 but a less restricted binding site might allow ceramide synthase(s) to accommodate the many different types of sphingoid bases that are found in nature (27). Because of this low degree of specificity, however, future studies should explore whether ceramide synthase can acylate other hydrophobic amines, including pharmaceuticals and toxins.
Compounds of particular interest are AP 1 and FB 1 , which are known inhibitors of ceramide synthase (13) but have not been analyzed as possible substrates. The acylation of AP 1 was similar to that for the threo-analogs; however, FB 1 was not detectably acylated, which is consistent with our suggestion (13,20) that the tricarballylic acid side chains of FB 1 occupy the acyl-CoA binding site and thereby block access of this co-substrate.
A somewhat unexpected finding of this study was that PAP 1 is highly cytotoxic, apparently as an inhibitor of ceramide synthase. This inhibition by N-acyl,1-deoxysphinganines represents a new category of ceramide synthase inhibitor. Thus, PAP 1 causes an increase in cellular sphinganine, which is cytotoxic for many cell types (29 -33) 6 and has been shown to mediate the toxicity of FB 1 for HT29 cells (19). Although PAP 1 might also be toxic as a ceramide analog because ceramides can induce apoptosis, this probably does not play a role in the toxicity of PAP 1 because the effects of N-acylsphingosines (ceramides) are usually stereoselective and require the 4,5-transdouble bond of the sphingosine backbone (28).
Feeding AP 1 to rats causes lesions in liver and kidney that are indistinguishable from those caused by FB 1 (21), and the organ-specific effects of feeding nixtamalized Fusarium moniliforme culture material (which contains AP 1 ) to rats are similar to those of the diet prepared from untreated (FB 1 -containing) culture material (21). AP 1 also appears to have the same liver cancer promoting activity as FB 1 (34). Heretofore, these in vivo effects of AP 1 have been somewhat puzzling because AP 1 is less potent than FB 1 as an inhibitor of ceramide synthase in vitro (13); however, if AP 1 is converted to an even more potent metabolite (PAP 1 ), this could account for this discrepancy. Further evidence for a similar mechanism of action of FB 1 and AP 1 is the close correlation between the toxicity of FB 1 or AP 1 for rats and elevations in sphinganine (22).
The toxicity of FB 1 , AP1 (19), and PAP 1 for HT29 cells, an intestinal cell line, implies that consumption of fumonisins may affect the gastrointestinal tract since a substantial portion of an administered dose is found in intestinal epithelial cells (35), excreta, bile, and gut contents as FB 1 , partially hydrolyzed FB 1 and AP 1 (36). Little is known about the intestinal uptake and fate of AP 1 ; however, sphingoid bases are rapidly taken up by intestinal cells (37); therefore, it is likely that AP 1 is taken up and converted to PAP 1 . This scenario might explain the recent report that human consumption of moldy maize (containing fumonisins) resulted in abdominal pain, borborygmi, and diarrhea (38).