Improved Inhibitors of Glucosylceramide Synthase*

Previous work has led to the identification of inhibitors of glucosylceramide synthase, the enzyme catalyzing the first glycosylation step in the synthesis of glucosylceramide-based glycosphingolipids. These inhibitors have two identified sites of action: the inhibition of glucosylceramide synthase, resulting in the depletion of cellular glycosphingolipids, and the inhibition of 1-O-acylceramide synthase, resulting in the elevation of cell ceramide levels. A new series of glucosylceramide synthase inhibitors based on substitutions in the phenyl ring of a parent compound, 1-phenyl-2-palmitoylamino-3-pyrrolidino-1-propanol (P4), was made. For substitutions of single functional groups, the potency of these inhibitors in blocking glucosylceramide synthase was primarily dependent upon the hydrophobic and electronic properties of the substituents. An exponential relationship was found between the IC50 of each inhibitor and the sum of derived hydrophobic (π) and electronic (ς) parameters. This relationship demonstrated that substitutions that increased the electron-donating characteristics and decreased the lipophilic characteristics of the homologues enhanced the potency of these compounds in blocking glucosylceramide formation. A novel compound was subsequently designed and observed to be even more active in blocking glucosylceramide formation. This compound,d-threo-4′-hydroxy-P4, inhibited glucosylceramide synthase at an IC50 of 90 nm. In addition, a series of dioxane substitutions was designed and tested. These included 3′,4′-methylenedioxyphenyl-, 3′,4′-ethylenedioxyphenyl-, and 3′4′-trimethylenedioxyphenyl-substituted homologues.d-threo-3′,4′-Ethylenedioxy-P4-inhibited glucosylceramide synthase was comparably active to thep-hydroxy homologue. 4′-Hydroxy-P4 and ethylenedioxy-P4 blocked glucosylceramide synthase activity at concentrations that had little effect on 1-O-acylceramide synthase activity. These novel inhibitors resulted in the inhibition of glycosphingolipid synthesis in cultured cells at concentrations that did not significantly raise intracellular ceramide levels or inhibit cell growth.

Previous work has led to the identification of inhibitors of glucosylceramide synthase, the enzyme catalyzing the first glycosylation step in the synthesis of glucosylceramide-based glycosphingolipids. These inhibitors have two identified sites of action: the inhibition of glucosylceramide synthase, resulting in the depletion of cellular glycosphingolipids, and the inhibition of 1-O-acylceramide synthase, resulting in the elevation of cell ceramide levels. A new series of glucosylceramide synthase inhibitors based on substitutions in the phenyl ring of a parent compound, 1-phenyl-2-palmitoylamino-3-pyrrolidino-1-propanol (P4), was made. For substitutions of single functional groups, the potency of these inhibitors in blocking glucosylceramide synthase was primarily dependent upon the hydrophobic and electronic properties of the substituents. An exponential relationship was found between the IC 50 of each inhibitor and the sum of derived hydrophobic () and electronic () parameters. This relationship demonstrated that substitutions that increased the electron-donating characteristics and decreased the lipophilic characteristics of the homologues enhanced the potency of these compounds in blocking glucosylceramide formation. A novel compound was subsequently designed and observed to be even more active in blocking glucosylceramide formation. This compound, D-threo-4-hydroxy-P4, inhibited glucosylceramide synthase at an IC 50 of 90 nM. In addition, a series of dioxane substitutions was designed and tested. These included 3,4-methylenedioxyphenyl-, 3,4-ethylenedioxyphenyl-, and 34-trimethylenedioxyphenyl-substituted homologues. D-threo-3,4-Ethylenedioxy-P4-inhibited glucosylceramide synthase was comparably active to the p-hydroxy homologue. 4-Hydroxy-P4 and ethylenedioxy-P4 blocked glucosylceramide synthase activity at concentrations that had little effect on 1-O-acylceramide synthase activity. These novel inhibitors resulted in the inhibition of glycosphingolipid synthesis in cultured cells at concentrations that did not significantly raise intracellular ceramide levels or inhibit cell growth.
In previously reported work, a series of PDMP homologues and analogues was synthesized (4). Replacing the decanoyl moiety with a palmitoyl moiety enhanced the effectiveness of PDMP. In addition, replacing the morpholino ring with a pyrrolidino ring, forming DL-threo-1-phenyl-2-palmitoylamino-3pyrrolidino-1-propanol (DL-threo-P4), also enhanced the inhibitory activity. It was also noted that the DL-threo-P4 derivative possessing a p-methoxy substituent on the phenyl group increased the inhibitory activity further. This latter observation led to the present study, an evaluation of other phenyl group substitutions in which the phenyl group of the P4 compound was modified by various electron-donating or -withdrawing groups. As expected, only the D-threo-enantiomers among P4 or P4 derivatives specifically inhibited the enzyme activity. The potency of these compounds in inhibiting GlcCer synthase was quantitatively related to the hydrophobic and electronic properties of the phenyl group substitutions of single substituents. This association resulted in the design of a new PDMP homologue (4Ј-hydoxy-P4) that was significantly more potent than those studied to date.

Materials
The acetophenones and amines were from Aldrich, Lancaster Synthesis Inc., and Maybridge Chemical Co. Silica gel for column chromatography (70 -230 mesh ASTM) and silica gel thin-layer chromatography plates were purchased from Merck. The reagents and their sources included non-hydroxy fatty acid ceramide from bovine brain and delipidated bovine serum albumin from Sigma, dioleoylphosphatidylcholine from Avanti, DL-dithiothreitol from Calbiochem, and uridine diphosphate-[1-3 H]glucose from NEN Life Science Products. Octanoylsphingosine was prepared as described previously (7).

General Synthesis of Inhibitors
The aromatic inhibitors were synthesized by the Mannich reaction from 2-N-acylaminoacetophenone, paraformaldehyde, and pyrrolidine, followed by reduction with sodium borohydride as described previously (2,4). The reaction produced a mixture of four isomers, due to the presence of two asymmetric carbons. For syntheses in which phenylsubstituted starting materials were used, the chloroacetophenone, methoxyacetophenone, methylenedioxyacetophenone, and methylacetophenone were brominated and converted to the primary amine. Brominations of the methoxyacetophenone, dimethoxyacetophenone, and 3Ј,4Ј-(methylenedioxy)acetophenone were performed in chloroform at room temperature, and the products were recrystallized from ethyl acetate and hexane.

Synthesis of 1-(4Ј-Hydroxy)phenyl-2-palmitoylamino-3-pyrrolidino-1-propanol
4Ј-Benzyloxyacetophenone Formation (Product 2)-4Ј-Hydroxyacetophenone (compound 1; 13.62 g, 100 mmol), benzyl bromide (17.1 g, 100 mmol), and cesium carbonate (35.83 g, 100 mmol) were added to tetrahydrofuran at room temperature and stirred overnight. The product was concentrated to dryness and recrystallized from ether and hexane to yield 15 g of 4Ј-benzyloxyacetophenone, which appeared as a white powder. An R F of 0.42 was observed when resolved by thin-layer chromatography using methylene chloride. 1  Bromination of 4Ј-Benzyloxyacetophenone (Product 3)-Bromine (80 mmol) was added dropwise over 5 min to a stirred solution of 4Јbenzyloxyacetophenone (70 mmol) in 40 ml of chloroform. This mixture was stirred for an additional 5 min and quenched with saturated sodium bicarbonate in water until the pH reached 7. The organic layers were combined, dried over MgSO 4 , and concentrated to dryness. The crude mixture was purified over a silica gel column and eluted with methylene chloride to yield 2-bromo-4Ј-benyloxyacetophenone. An R F of 0.62 was observed when resolved by thin-layer chromatography using methylene chloride. 1  2-Amino-4Ј-benzyloxyacetophenone HCl Formation (Product 4)-Hexamethylenetetramine (methenamine; 3.8 g, 23 mmol) was added to a stirred solution of 2-bromo-4Ј-benyloxyacetophenone (6.8 g, 23 mmol) in 100 ml of chloroform. After 4 h, the crystalline adduct was filtered and washed with chloroform. The product was dried and heated with 150 ml of methanol and 8 ml of concentrated HCl in an oil bath at 85°C for 3 h. Upon cooling, the precipitated hydrochloride salt (2.5 g) was removed by filtration. The filtrate was left at Ϫ20°C overnight, and additional product (2.1 g) was isolated. The yield was 4.6 g (82.6% 2-Palmitoylamino-4Ј-benyloxyacetophenone Formation (Product 5)-Sodium acetate (50% in water, 29 ml) was added in three portions to a stirred solution of 2-amino-4Ј-benzyloxyacetophenone HCl (4.6 g, 17 mmol) and tetrahydrofuran (200 ml). Palmitoyl chloride (19 mmol) in tetrahydrofuran (25 ml) was added dropwise over 20 min, yielding a dark brown solution. The mixture was stirred overnight at room temperature. The aqueous fraction was removed by use of a separatory funnel, and chloroform/methanol (2:1, 150 ml) was added to the organic layer, which was then washed with water (50 ml). The yellow aqueous layer was extracted once with chloroform (50 ml). The organic solutions were then pooled and rotoevaporated until near dryness. The residue was redissolved in chloroform (100 ml) and crystallized by the addition of hexane (400 ml). The flask was cooled to 4°C for 2 h. The crystals were filtered, washed with cold hexane, and dried in a fume hood overnight. The product yield was 3.79 g (8 mmol). An R F of 0.21 was 1-(4Ј-Benzyloxy)phenyl-2-palmitoylamino-3-pyrrolidino-1-propanol Formation (Products 6 and 7)-2-Palmitoylamino-4Ј-benyloxyacetophenone (3.79 g, 8.0 mmol), paraformaldehyde (0.25 g, 2.7 mmol, equivalent to 8.1 mmol of formaldehyde), pyrrolidine (0.96 ml, 11.4 mmol), and ethanol (70 ml) were stirred under nitrogen. Concentrated HCl (0.26 ml) was added through the condenser, and the mixture was heated to reflux for 16 h. The resultant brown solution was cooled on ice, and then sodium borohydride (1.3 g, 34 mmol) was added in three portions. The mixture was stirred at room temperature overnight, and the product was dried in a solvent evaporator. The residue was redissolved in dichloromethane (130 ml) and hydrolyzed with 3 N HCl (pH ϳ4). The aqueous layer was twice extracted with dichloromethane (50 ml). The organic layers were pooled, washed twice with water (30 ml) and twice with saturated sodium chloride (30 ml), and dried over anhydrous magnesium sulfate. The dichloromethane solution was rotoevaporated to a semisolid and purified by use of a silica rotor using a solvent consisting of 10% methanol in dichloromethane. This yielded a mixture of DL-threo-and DL-erythro-enantiomers (2.53 g, 4.2 mmol). R F values of 0.43 for the erythro-diastereomers and 0.36 for the threo-diastereomers were observed when resolved by thin-layer chromatography using methanol/methylene chloride (1:9). [M ϩ H] ϩ : 565 for C 36 H 56 N 2 O 3 .
2-Palmitoylamino-3Ј, 4Ј-(ethylenedioxy)acetophenone-Aminoacetophenone HCl (7.1 g, 31 mmol) and tetrahydrofuran (300 ml) were placed in a 1-liter three-neck round-bottom flask with a large stir bar. Sodium acetate (50% in water, 31 ml) was added in three portions to this suspension. Palmitoyl chloride (31 ml, 10% excess, 36 mmol) in tetrahydrofuran (25 ml) was then added dropwise over 20 min to yield a dark brown solution. This mixture was stirred for an additional 2 h at room temperature. The resultant mixture was poured into a separatory funnel to remove the aqueous solution. Chloroform/methanol (2:1, 150 ml) was then added to the organic layer and washed with water (50 ml). The yellow aqueous layer was extracted once with chloroform (50 ml). The organic solutions were pooled and rotoevaporated until almost dry. The residue was redissolved in chloroform (100 ml) and crystallized by the addition of hexane (400 ml). The flask was then cooled to 4°C for 2 h. The crystals were filtered and washed with cold hexane until they were almost white and then dried in a fume hood overnight. The yield of the product was 27 mmol (11.6 g).

D-threo-1-(3Ј,4Ј-Ethylenedioxy)phenyl-2-palmitoylamino-3-pyrrolidino-
1-propanol-Palmitoylaminoacetophenone (11.6 g, 27 mmol), paraformaldehyde (0.81 g, 9 mmol), pyrrolidine (3.6 ml, 42 mmol), and ethanol (250 ml) were added to a 500-ml round-flask under nitrogen flow. Concentrated HCl (0.8 ml) was added to this mixture through the reflux condenser, and the mixture was refluxed for 16 h. The brown solution was cooled in an ice bath. Sodium borohydride (2.28 g, 60 mmol) was added in three portions. This mixture was stirred at room temperature for 3 h and then rotoevaporated. The residue was dissolved in 130 ml of dichloromethane, and the borate complex was hydrolyzed with HCl (3 N) until the pH was ϳ4. The aqueous layer was extracted twice with 50 ml of dichloromethane. The organic layers were pooled, washed twice with H 2 O (30 ml) and saturated NaCl (30 ml), and dried over anhydrous MgSO 4 . The dichloromethane solution was rotoevaporated to a viscous oil, which was purified by use of a Chromatotron with a solvent consisting of 10% methanol in dichloromethane to obtain a mixture of DL-threo-and erythroenantiomers (2.24 g, 4 mmol).

GlcCer Synthase Activity
The enzyme activity was measured as described previously (8). Madin-Darby canine kidney (MDCK) cell homogenates (120 g of protein) were incubated with uridine diphosphate-[ 3 H]glucose (100,000 cpm) and liposomes consisting of 85 g of octanoylsphingosine, 570 g of dioleoylphosphatidylcholine, and 100 g of sodium sulfatide in a 200-l reaction mixture and kept for 1 h at 37°C. P4 and P4 derivatives dissolved in dimethyl sulfoxide was dispersed into the reaction mixture after adding the liposomes. The final concentration of dimethyl sulfoxide was Ͻ1%. At this concentration, there was no inhibition of the enzyme activity.

Cell Culture and Lipid Extraction
MDCK cells (5 ϫ 10 5 ) were seeded into 10-cm dishes containing 8 ml of serum-free supplemented Dulbecco's modified Eagle's medium (9). After 24 h, the medium was replaced with 8 ml of the medium containing 0, 11.3, 113, or 1130 nM D-threo-P4, D-threo-3Ј,4Ј-ethylenedioxy-P4, or D-threo-4Ј-hydroxy-P4. The GlcCer synthase inhibitors were added to the medium as a 1:1 molecular complex with delipidated bovine serum albumin (4,6). The cells were incubated for 24 or 48 h with the inhibitors. After the incubation, the cells were washed twice with 8 ml of cold phosphate-buffered saline and fixed with 2 ml of cold methanol. The fixed cells were scraped and transferred to a glass tube. An additional 1 ml of methanol was used to recover the remaining cells in the dish.
Three ml of chloroform were added to the tube and briefly sonicated using a water bath sonicator. After centrifugation at 800 ϫ g for 5 min, the supernatant was transferred into another glass tube. The residues were reextracted with chloroform/methanol (1:1). After the centrifugation, the resultant supernatant was combined with the first one. The residues were air-evaporated and kept for protein analysis by the bicinchoninic acid method. NaCl (0.9%) was added to the supernatant, and the ratio of chloroform/methanol/water was adjusted to 1:1:1. After centrifugation at 2000 ϫ g for 5 min, the upper layer was discarded. The lower layer was washed with methanol/water (1:1) in an volume equal to that of the lower layer. The resultant lower layer was transferred into a glass tube and dried under a stream of nitrogen gas. A portion of the lipid was used for lipid phosphate determination (10). The remainder was analyzed using high performance thin-layer chromatography.

Resolution of PDMP Homologues by Chiral Chromatography-
The structures of the parent compound, D-threo-P4, and the phenyl-substituted homologues including the new dioxysubstituted and 4Ј-hydroxy-P4 homologues are shown in Fig. 2. Initially, the effect of each P4 isomer separated by chiral chromatography on GlcCer synthase activity was determined (Fig.  3A). Four peaks were observed for the chiral separation of P4. Peaks 1 and 2 represent the erythro-diastereomers, and peaks 3 and 4 represent the threo-diastereomers as determined by a sequential separation of the P4 mixture by reverse-phase chromatography followed by the chiral separation. The enzyme activity was specifically inhibited by the fourth peak, the Dthreo-isomer (Fig. 3B). This specificity for the D-threo-enantiomer was consistent with the previous results observed in PDMP and PDMP homologues (2)(3)(4). The IC 50 of D-threo-P4 was 0.5 M for GlcCer synthase activity measured in the MDCK cell homogenates.

FIG. 5. Effects of D-threo-4-hydroxy-P4 compared with D-threop-methoxy-P4 on GlcCer synthase activity.
Substitution on GlcCer Synthase Activity-In previous work, the addition of a p-methoxy group to DL-threo-P4 was found to enhance the effect of the inhibitor on the enzyme activity (4). This improved efficacy was confirmed for the D-threo-enantiomer. The enzyme activity was potently inhibited by D-threop-methoxy-P4 (IC 50 ϭ 0.2 M) (Fig. 3C). Chiral chromatography of the four p-methoxy-P4 enantiomers failed to resolve completely to base line each enantiomer (Fig. 3A). A slight inhibition of the enzyme activity by p-methoxy-P4 in a combined mixture of D-erythro-and L-threo-enantiomers (peaks 2 and 3) was observed; this might have been due to contamination of the D-threo-isomer (peak 4) into these fractions, although this is not apparent on the separation shown.
Next, a series of D-threo-P4 derivatives containing other single substituents on the phenyl group was investigated. The potency of these phenyl-substituted compounds as GlcCer synthase inhibitors was inferior to that of D-threo-P4 or p-methoxy-D-threo-P4 (Table I). For many drugs, the effect of aromatic substitutions on the biological activity has been characterized and is often predictive. Generally, the IC 50 can be described by the following equation (11): log(1/IC 50 ) ϭ a (hydrophobic parameter ()) ϩ b (electronic parameter ()) ϩ c (steric parameter) ϩ d (another descriptor) ϩ e, where a-e are the regression coefficients.
The hydrophobic effect () is described by the equation ϭ log P X Ϫ log P H , where P X is the partition coefficient of the substituted derivative and P H is that of the parent compound, measured as the distribution between octanol and water. The electronic substituent parameter () was originally developed by Hammett (12) and is expressed as ϭ log K X Ϫ log K H , where K X and K H are the ionization constants for a para-or meta-substituted derivative and benzoic acid, respectively. Positive values represent electron-withdrawing properties, and negative values represent electron-donating properties.
The potency of D-threo-P4 and P4 derivatives as inhibitors is mainly dependent upon two factors (hydrophobic and electronic properties) of a substituent of the phenyl group (Table I). Surprisingly, a linear relationship was observed between log(IC 50 ) and ϩ (Fig. 4). These findings suggest that the more negative the value of ϩ , the more potent are D-threo-P4 derivatives made as GlcCer synthase inhibitors.
The p-Hydroxy-substituted Homologue Is a Significantly Better GlcCer Synthase Inhibitor-The strong association between ϩ and GlcCer synthase inhibition suggested that a still more potent inhibitor could be produced by increasing the electron-donating properties and decreasing the lipophilic properties of the phenyl group substituent. A predictably negative ϩ value would be observed for the p-hydroxy homologue. This compound was synthesized (see "Experimental Procedures"), and the D-threo-enantiomer was isolated by chiral chromatography. An IC 50 of 90 nM for GlcCer synthase inhibi-tion was observed (Fig. 5), suggesting that the p-hydroxy homologue was twice as active as the p-methoxy compound. Moreover, the linear relationship between log(IC 50 ) and ϩ was preserved (Fig. 4).
b These values were derived from Fig. 3. For other compounds, the same analytical approach as shown in Fig. 3 was carried out to obtain the IC 50 values (data not shown). methoxy-P4, as an inhibitor (Fig. 6). As the parameters m and p (the Hammett constants for the meta-and para-substitutions, respectively) and for a single methoxy substituent are 0.12, Ϫ0.27, and Ϫ0.02, respectively (11), the value of ϩ for D-threo-dimethoxy-P4 is presumed to be negative. Therefore, based only on the electronic substituent parameters, D-threodimethoxy-P4 deviates quite far from the correlation observed in Fig. 4. This may be due to a repulsion between two methoxy groups in the dimethoxy-P4 molecule that induces a steric effect that was negligible in mono-substituted D-threo-P4 derivatives studied in Fig. 4. GlcCer synthase is thought to possess a domain that interacts with D-threo-PDMP and PDMP homologues and that modulates the enzyme activity (2,6). The steric effect generated by an additional methoxy group may affect one or more of these interaction domains. As a result, the potency of the dimethoxy homologue as a GlcCer synthase inhibitor was markedly decreased.
Distinguishing between Inhibition of GlcCer Synthase and 1-O-Acylceramide Synthase-Prior studies on PDMP and related homologues revealed that both the threo-and erythrodiastereomers were capable of increasing cell ceramide and inhibiting cell growth despite the observation that only the D-threo-enantiomers blocked GlcCer synthase (4). An alternative pathway for ceramide metabolism was subsequently iden-tified (the acylation of ceramide at the 1-hydroxyl position) that was blocked by both threo-and erythro-diastereomers of PDMP. The specificities of D-threo-P4, D-threo-3Ј,4Ј-ethylenedioxy-P4, and D-threo-4Ј-hydroxy-P4 for GlcCer synthase were studied by assaying the transacylase. Although there was a significant difference in activity among D-threo-3Ј,4Ј-ethylenedioxy-P4, D-threo-4Ј-hydroxy-P4, and D-threo-P4 in inhibiting GlcCer synthase, the D-threo-enantiomers of all three compounds demonstrated comparable activity in blocking 1-O-acylceramide synthase (Fig. 7).
To determine whether inhibition of 1-O-acylceramide synthase was the basis for inhibitor-mediated ceramide accumulation, the ceramide and diradylglycerol levels of MDCK cells treated with D-threo-P4, D-threo-3Ј,4Ј-ethylenedioxy-P4, and D-threo-4Ј-hydroxy-P4 were measured (Table II). Significant increases in both ceramide and diradylglycerol occurred only in cells treated with inhibitor concentrations in excess of 1 M. This was ϳ30-fold lower than the concentration required for inhibition of the 1-O-acylceramide synthase assayed in the cellular homogenates. This disparity in concentration effects most likely reflects the ability of the more potent homologues to accumulate within intact cells (6).
Effects of D-threo-P4, D-threo-4Ј-Hydroxy-P4, and D-threo-Ethylenedioxy-P4 on GlcCer Synthesis and Cell Growth-To confirm the cellular specificity of D-threo-3Ј,4Ј-ethylenedioxy-P4 and D-threo-4Ј-hydroxy-P4 as compared with Dthreo-P4, MDCK cells were treated with different concentrations of the inhibitors. Approximately 66 and 78% of the GlcCer was lost from the cells treated with 11.3 nM D-threo-4Ј-hydroxy-P4 and D-threo-ethylenedioxy-P4, respectively (Fig. 8, B and C). By contrast, only 27% depletion of GlcCer occurred in cells exposed to D-threo-P4 (Fig. 8A). A low level of GlcCer persisted in the cells treated with a 113 or 1130 nM concentration of either compound. This may be due to the contribution, by degradation, of more highly glycosylated sphingolipids or the existence of another GlcCer synthase that is insensitive to the inhibitor.
On the other hand, there was little difference in the total protein content between untreated cells and those treated with 11.3 or 113 nM D-threo-4Ј-hydroxy-P4 and D-threo-ethylenedioxy-P4 (Fig. 8, B and C). A significant decrease in total protein was observed in the cells treated with a 1130 nM concentration of either P4 homologue. In addition, the level of ceramide in the cells treated with 1130 nM D-threo-ethylenedioxy-P4 and D-threo-4Ј-hydroxy-P4 was two times higher than that measured in the untreated cells (Table II). There was no change in ceramide or diradylglycerol levels in cells treated with 11.3 and 113 nM concentrations of either compound. Similar patterns for GlcCer levels and protein content were observed after 48-h incubations (data not shown).
The phospholipid content was unaffected at the lower concentrations of either D-threo-ethylenedioxy-P4 or D-threo-4Ј-hydroxy-P4. The ratios of cell protein to cellular phospholipid phosphate (g of protein/nmol of phosphate) were 4.94 Ϯ 0. respectively, suggesting that the changes in GlcCer content were truly related to inhibition of GlcCer synthase activity. These results strongly indicate that two new inhibitors, D-threo-4Ј-hydroxy-P4 and D-threo-3,4ethylenedioxy-P4, are able to potently and specifically inhibit GlcCer synthesis in intact cells at low nM concentrations without any inhibition of cell growth. MDCK cells (5 ϫ 10 5 ) were seeded into a 10-cm dish and incubated for 24 h. Following the incubation, the cells were treated for 24 or 48 h with or without P4 or the phenyl-substituted homologues. Both ceramide and diradylglycerol contents were determined by the method of Preiss et al. (17). GlcCer content was measured densitometrically from TLCseparated glycosphingolipids by a video camera and use of NIH Image 1.49 as previously described (9). For D-threo-ethylenedioxy-P4, only two determinations were made.

DISCUSSION
Since the original description of an inhibitor of GlcCer synthesis by Vunnam and Radin (13), the pharmacological blockade of glycosphingolipid synthesis has proven to be a valuable approach to understanding the metabolism and function of glycosphingolipids. Previous work with the parent GlcCer synthase inhibitor (PDMP) identified two concurrent effects in cells. These included the time-dependent depletion of all Gl-cCer-based glycosphingolipids and the accumulation of ceramide. Originally, the ceramide accumulation was believed to be the result of substrate accumulation. However, it was discovered that the erythro-enantiomers of pyrrolidino-substituted compounds raised cell ceramide independent of GlcCer depletion (4). In addition, homologues with aliphatic substitutions depleted GlcCer with minimal effects on ceramide levels. The current data with the new phenyl-substituted inhibitors are consistent with this observation. No growth inhibition was observed at low nM concentrations of either D-threo-ethylenedioxy-P4 or D-threo-4Ј-hydroxy-P4, but was observed at higher concentrations at which ceramide levels increase. These findings suggested a second site of action for these inhibitors that is independent of the inhibition of GlcCer synthase.
The search for another site of inhibition of ceramide metabolism in the presence of either threo-or erythro-diastereomers of P4 led to the identification of a novel pathway for ceramide metabolism, the acylation of ceramide at the 1-hydroxyl position. The formation of this lipid is catalyzed by a novel phospholipase A 2 . In the presence of ceramide as an acceptor, 1-Oacylceramide synthase can transacylate ceramide utilizing the sn-2-fatty acid of phosphatidylethanolamine or phosphatidylcholine (14). This transacylase has recently been purified (15). By using the erythro-diastereomers of inhibitors that increase ceramide to the exclusion of blocking GlcCer formation, it has been determined that the growth inhibitory effects of these homologues are mediated by ceramide accumulation and not GlcCer depletion.
The dissociation of GlcCer depletion from ceramide accumulation is an important and necessary finding if one is to con-FIG. 8. Effects of P4 and related homologues on GlcCer synthesis and cell growth. MDCK cells were treated for 24 h with or without different concentrations of D-threo-P4 (A), D-threo-4Ј-hydroxy-P4 (B), and D-threo-ethylenedioxy-P4 (C). The total protein amount in each sample was determined by the bicinchoninic acid method (dotted lines). In the GlcCer analysis, lipid samples and standard lipids were applied to the same high performance TLC plate pretreated with borate and developed in a solvent consisting of chloroform/methanol/water (63:24:4). The level of GlcCer was estimated from a standard curve obtained using a computerized image scanner (solid lines). The values were normalized on the basis of the phospholipid content. Each point shown represents the average values from three dishes, with error bars corresponding to 1 S.D. sider the development of GlcCer synthase inhibitors as potential drugs for inherited glycosphingolipid storage diseases. Ideally, such a drug should exhibit little or no cellular toxicity. The growth inhibitory and proapoptotic effects of ceramide should ideally be eliminated. Recently, support for the concept of treating sphingolipid storage disorders by inhibition of Gl-cCer synthase was reported by the reversal of the Tay-Sachs phenotype in knockout mice treated with a structurally unrelated inhibitor of GlcCer synthesis, N-butyldeoxynojirimycin (16). This inhibitor is significantly less potent and less specific than the compounds characterized in the present report.
Previous refinements of the parent structure of PDMP have resulted in compounds with greater activity against GlcCer synthase. However, these substitutions of the fatty acyl chain and cyclic amine moieties were designed empirically. In the present study, single phenyl substitutions permitted the analysis of inhibitor activity based on constants derived by Hansch and others many years ago (18). Parameters of both lipophilicity and electronegativity were observed to be predictive of inhibitory activity. Specifically, the design and synthesis of D-threo-4Ј-hydroxy-P4 yielded the most potent GlcCer synthase inhibitor to date. These parameters, however, could be applied only to simple substitutions and were inadequate to explain the structure activity profile of more complex ring structures. Nevertheless, the further application of these principles may result in newer glycolipid synthase inhibitors with even greater activity and specificity.