CERT Mediates Intermembrane Transfer of Various Molecular Species of Ceramides*

Ceramide produced at the endoplasmic reticulum is transported to the Golgi apparatus for conversion to sphingomyelin. The main pathway of endoplasmic reticulum-to-Golgi transport of ceramide is mediated by CERT, a cytosolic 68-kDa protein, in a nonvesicular manner. CERT contains a domain that catalyzes the intermembrane transfer of natural C16-ceramide. In this study, we examined the ligand specificity of CERT in detail by using a cell-free assay system for intermembrane transfer of lipids. CERT did not mediate the transfer of sphingosine or sphingomyelin at all. The activity of CERT to transfer saturated and unsaturated diacylglycerols, which structurally resemble ceramide, was 5–10% of the activity toward C16-ceramide. Among four stereoisomers of C16-ceramide, CERT specifically recognized the natural d-erythro isomer. CERT efficiently transferred ceramides having C14, C16, C18, and C20 chains, but not longer acyl chains, and also mediated efficient transfer of C16-dihydroceramide and C16-phyto-ceramide. Binding assays showed that CERT also recognizes short chain fluorescent analogs of ceramide with a stoichiometry of 1:1. Moreover, (1R,3R)-N-(3-hydroxy-1-hydroxymethyl-3-phenylpropyl)dodecamide, which inhibited the CERT-dependent pathway of ceramide trafficking in intact cells, was found to be an antagonist of the CERT protein. These results indicate that CERT can mediate transfer of various types of ceramides that naturally exist and their close relatives.

The intracellular transport of lipids from the sites of their synthesis to their appropriate destinations must occur, because various steps in lipid biosynthesis occur in different intracellular compartments. The trafficking of integral membrane proteins in eukaryotic cells is mediated by transport vesicles, which load the desired set of proteins and deliver them to the correct organelles. By contrast, many types of lipid synthesized in the endoplasmic reticulum (ER) 1 have been suggested to be sorted to other organelles by nonvesicular mechanisms, although some lipid flux routes such as the endocytosis of plasma membrane lipids occur by vesicle-mediated mechanisms (1)(2)(3). In mammalian cells, ceramide is synthesized at the ER and translocated to the Golgi compartment for conversion to sphingomyelin (4). There are at least two pathways by which ceramide is transported from the ER to the Golgi site for the synthesis of sphingomyelin: an ATP-and cytosol-dependent major pathway and an ATP-or cytosol-independent (or less dependent) minor pathway (5)(6)(7). The major pathway is impaired in a Chinese hamster ovary (CHO) mutant cell line, LY-A, without any deficiency in the ER-to-Golgi transport of proteins (5)(6)(7). We have identified CERT as a factor defective in LY-A cells by functional rescue experiments, and we have shown that CERT mediates the ATP-dependent pathway of ER-to-Golgi trafficking of ceramide in a nonvesicular manner (8).
CERT is a tripartite cytosolic protein ϳ600 amino acids in length (8 -10). The amino-terminal region of ϳ120 amino acids is a phosphatidylinositol 4-phosphate (PtdIns4P)-binding pleckstrin homology domain, which can target the Golgi apparatus (11). The next region of ϳ250 amino acids (referred to as the middle region) contains coiled-coil motifs (9), which might play a role in homo-or hetero-oligomerization, and a motif that may participate in association with the ER (12). The carboxyl terminus of ϳ230 amino acids is a steroidogenic acute regulatory protein (StAR)-related lipid transfer (START) domain.
START domains were initially recognized as putative lipidbinding domains of ϳ210 amino acid residues, which exist in various types of proteins implicated in intracellular lipid transport, lipid metabolism, and signal transduction (13,14). Although more than 200 proteins have been nominated so far as proteins having START domains in data bases (for example, see smart.embl-heidelberg.de/), only a few have been experimentally shown to bind or transfer specific lipids. For example, StAR and MLN64 proteins recognize cholesterol (14 -17), and phosphatidylcholine (PtdCho)-transfer protein is capable of intermembrane transfer of PtdCho in vitro (18,19). The silkworm Bombyx mori larvae produce a carotenoid-binding START domain (20). We demonstrated previously that the START domain of CERT can efficiently extract natural long chain C 16 ceramide but not other types of lipids, including sphingosine, sphingomyelin, PtdCho, and cholesterol, from phospholipid bilayers (8). We have also shown that the START domain of CERT greatly facilitated intermembrane transfer of C 16 -cer-amide in a cell-free system (8). However, many details as to the substrate specificity of the START domain of CERT remain undetermined, although various molecular species of ceramide exist in mammalian cells.
In most types of mammalian cells, the hydrophobic moiety of complex sphingolipids is mainly composed of ceramide but also includes dihydroceramide and phytoceramide at low levels (21)(22)(23). Notably, dihydrosphingomyelin is abundant in human lens membranes (24,25). Moreover, the length of the amido acyl chain of the ceramide moiety is diverse; C 16 -C 26 acyl chains are observed in natural sphingomyelin. Notably, C 18 -and C 24:1ceramide is predominant for sphingomyelin in the brain (21,26,27), whereas C 16 -ceramide is predominant in many other tissues (22, 28 -30). Such structural diversity in the ceramide moiety may affect the nature of membranes where complex sphingolipids are abundant. The physiological importance of the diversity of the ceramide structure has also been recognized, based on differences in bio-modulation activity between ceramide and dihydroceramide (31) or between natural long chain ceramide and unnatural short chain ceramide (32). Hence, it should be of biological significance to determine whether CERT can catalyze the intermembrane transfer of various species of ceramide and its relatives in addition to C 16 -ceramide.
In the present study, we show that CERT is capable of mediating the intermembrane transfer of various types of ceramides that naturally exist in its START domain-dependent manner. In addition, we show that an inhibitor of ER-to-Golgi transport of ceramide is an antagonist of CERT.
Preparation of C 16 (34). In our standard method, 400 nmol of [ 14 C]palmitic acid (55 mCi/mmol) and 2 mol of appropriate sphingoid base (D-erythro-sphingosine, D-threo-sphingosine, L-erythrosphingosine, L-threo-sphingosine, D-erythro-dihydrosphingosine, or Dribo-phytosphingosine) dissolved in ethanol were mixed in a screwcapped Pyrex glass tube and dried under nitrogen gas stream. After adding 2 ml of buffer A (50 mM sodium phosphate buffer (pH 7.0) containing 0.2% Triton X-100) to the dried lipids, the tube was sonicated with a bath-type sonicator for 5 min. 2 ml of SCDase (1 milli-unit/ml in buffer A) was then added to the tube, and the mixture was incubated at 37°C for ϳ16 h. Lipids were extracted from the mixture by the method of Bligh and Dyer (35), dried, dissolved in 400 l of chloroform/methanol (19:1, by volume), and subjected to TLC with a solvent system of benzene, diethyl ether, ethyl acetate, methanol, 25% (w/v) ammonia (65:7.5:7.5:25:0.75, v/v). After detection of radioactive lipids separated on the TLC plate with a BAS1800 image analyzer (Fuji Film), gel containing a desired radioactive lipid was collected from the TLC plate by scraping. By this image analysis, the efficiency of conversion of the radiolabeled starting material to its ceramide product by the SCDase reaction was also estimated. For extraction of the radioactive lipid by the method of Bligh and Dyer (35), 1.9 ml of a solvent of 0.1 M KCl/chloroform/methanol (0.8:1:2, v/v) was added to the collected gel and mixed with a swirling mixer. After addition of 0.5 ml of chloroform and 0.5 ml of 0.1 M KCl to the mixture for phase separation, the mixture was centrifuged (1000 ϫ g, 3 min), and the lower organic phase was collected as a lipid extract. Then the extract was dried under a nitrogen gas stream, dissolved in 1 ml of chloroform/methanol (19:1,v/v), and stored at Ϫ20°C. The radioactivities of synthesized radioactive lipids were determined by liquid scintillation counting. We assumed that specific activities of the synthesized C 16 -[ 14 C]ceramide and its isomers were identical to the specific activity (55 mCi/mmol) of the precursor Purification of Recombinant CERT and CERT⌬ST-His 6 -tagged recombinant human CERT and its START domain-deleted mutant CERT⌬ST were purified as described previously (8,36).
Intermembrane Ceramide Transfer Assay-Intermembrane transfer of ceramide and its relatives was assayed in a cell-free system that we briefly described previously (8). Here we describe the method in more detail.
On the day of the lipid transfer assay, purified recombinant CERT and CERT⌬ST in 10 mM Tris-HCl buffer (pH 7.4) containing 250 mM sucrose buffer were diluted to 1 nmol/ml (71.8 and 45.7 g/ml for the His 6 -tagged CERT and CERT⌬ST, respectively) with buffer C (20 mM Hepes-Na buffer (pH 7.4) containing 50 mM NaCl, and 1 mM EDTA). Donor vesicles per assay consist of 32 nmol of egg PtdCho, 8 nmol of egg PtdEtn, 4 nmol of porcine lactosylceramide, and 0.5 nmol of radioactive ceramide (27.5 nCi for [ 14 C]ceramides or 10 nCi for [ 3 H]ceramides). Acceptor vesicles per assay consist of 320 nmol of egg PtdCho and 80 nmol of egg PtdEtn. Note that the excess amount of acceptor vesicles to donor vesicles is crucial to minimize donor-to-donor transfer of ceramide, which interferes with the donor-to-acceptor transfer reaction. When necessary, donor vesicles additionally contain [ 3 H]dipalmitoyl-PtdCho (125 nCi per assay) as a nonexchangeable lipid marker. According to the number of assays, appropriate amounts of lipids dissolved in organic solvents were mixed in a polypropylene tube (Eppendorf) and dried under a nitrogen gas stream. After addition of buffer C, phospholipid vesicles were prepared by sonication with a probe-type sonicator (model UP-50H, Dr. Hielscher GmbH, Teltow, Germany) at 80% output and 50% cycle for 10 min in a water bath at room temperature. The volume of buffer C that should be added at this step was 20 l per assay in donor vesicles and 60 l per assay in acceptor vesicles (note that at least 200 l of the buffer is required for the sonication step). To remove lipid aggregates, the sonicated samples were centrifuged at 20,000 ϫ g for 30 min at 4°C, and the supernatant fraction was collected as small vesicles. The radioactivity of the supernatant was determined by liquid scintillation counter for assessing the recovery yields after pre-centrifugation. In some cases, the recovery of lipids in the supernatant fraction was also assessed by the lipid phosphorous quantification method (37). Both assessments showed that over 90% of lipids were reproducibly recovered in the supernatant fraction. The prepared small vesicles were used for intermembrane ceramide transfer assay as follows. In typical experiments, 18 l of buffer C, 60 l of acceptor vesicles, and 2 l of recombinant CERT or CERT⌬ST (1 nmol/ml in buffer C) were mixed in a 1.5-ml polypropylene tube. Then 20 l of donor vesicles was added to the tube to start the ceramide transfer reaction. After tapping the tube quickly, the reaction mixture was incubated for 10 min at 37°C. For mock incubation, buffer C as the vehicle buffer was added in place of the recombinant protein. To stop the reaction, 30 l of R. communis agglutinin (2.5 mg/ml in phosphate-buffered saline) was added to the reaction mixture and mixed by pipetting. The agglutinin selectively aggregates donor vesicles by cross-linking of the terminal galactose residue of lactosylceramide embedded in donor vesicles. The mixture was chilled on ice for 10 min and centrifuged (20,000 ϫ g, 3 min, 4°C) to precipitate agglutinated donor vesicles. Then 115 l of the supernatant fluid was carefully retrieved, and the radioactivity of the supernatant was measured in 2 ml of ACS-II® (Amersham Biosciences) by liquid scintillation counting. To remove the radioactivity due to incomplete precipitation of donor vesicles, the radioactivity from the mock incubation without any CERT recombinants was subtracted from the radioactivity of each sample. When the effects of HPA-12 and its derivatives on ceramide transfer activity were examined, several modifications were made. Specifically, in preparation of donor vesicles, the amount of C 16 -[ 14 C]cer-amide added was reduced from 0.5 to 0.1 nmol. Then 913 l of buffer C, 20 l of donor vesicles, 1-5 l of 3 mM drugs and the vehicle dimethyl sulfoxide (the final concentration of dimethyl sulfoxide was adjusted to 0.5%), and 2 l of recombinant CERT or CERT⌬ST (1 nmol/ml) were mixed in a 1.5-ml tube. After preincubation of the mixture for 5 min at 37°C, 60 l of acceptor vesicles was added to the mixture and incubated for 30 min at 37°C. Then after addition of 30 l of R. communis agglutinin (2.5 mg/ml), the chilled mixture was centrifuged (20,000 ϫ g, 3 min, 4°C), and 960 l of the supernatant fluid was retrieved for liquid scintillation counting.
Binding Assay of C 5 -DMB-Ceramide, C 6 -NBD-Ceramide, and LPS Alexa Fluor® 488 to CERT-On the day of the assay, frozen stocks of recombinant CERT and CERT⌬ST dissolved in 10 mM Tris-HCl (pH 7.4) containing 150 mM NaCl (Tris-buffered saline; TBS) were thawed and centrifuged (20,000 ϫ g, 4°C, 10 min) to remove aggregates. The supernatant fraction was retrieved, and its protein concentration was determined. For each binding assay, 400 pmol of His 6 -tagged CERT or CERT⌬ST (28.7 or 18.3 g, respectively) in 118 l of TBS was mixed in a polypropylene tube (Eppendorf). For a negative control, TBS was added in place of the recombinant proteins. After adding 2 l of 0.1 mM ethanolic stock solution of C 5 -DMB-ceramide or C 6 -NBD-ceramide and 60 l of TBS to the tube, the mixture was incubated at 37°C for 30 min for the binding reaction. Then 60 l of 50% (v/v) slurry of TALON metal affinity resin pre-equilibrated with buffer B was added to the binding reaction mixture and incubated for 10 min at room temperature with rotary shaking. After centrifugation (20,000 ϫ g, 10 s), the supernatant was retrieved as the "unbound fraction." For washing, the resin was suspended in 150 l of buffer B containing 10 mM imidazole and precipitated, and the supernatant was retrieved as the "wash fraction." This washing step was repeated. The TALON-bound protein was then eluted by incubation with 150 l of buffer B containing 250 mM imidazole for 5 min at room temperature with occasional tapping. After centrifugation (20,000 ϫ g, 10 s), the supernatant was retrieved as the "elute fraction." A 3.75-fold volume of chloroform/methanol (1:2, v/v) was then added to each retrieved fraction, mixed, and centrifuged (20,000 ϫ g, 10 s). In addition, to retrieve fluorophores that were nonspecifically bound to the resin and tube, 170 l of TBS and 750 l of chloroform/methanol (1:2, v/v) were added to the tube containing the resin used, mixed, and centrifuged (20,000 ϫ g, 10 s). The supernatant was retrieved as the "residual fraction." The DMB (excitation at 480 nm; emission at 515 nm) and NBD (excitation at 470 nm; emission at 530 nm) fluorophores in these fractions were quantified with a fluorescence spectrophotometer (model F-3000, Hitachi, Tokyo, Japan). When the binding stoichiometry of CERT and ceramide was analyzed, some modifications were made, because the amount of ceramide must be in excess to that of CERT for this analysis. Briefly, various concentrations of C 5 -DMB-ceramide were mixed with 40 pmol of recombinant CERT or CERT⌬ST and 15 l of 50% slurry of TALON metal affinity resin, and then the mixture (the volume of which was 135 l) was incubated at 37°C for 30 min. The amounts of the recombinant proteins distributed to the elute fraction were estimated by densitometric analysis after a portion of the fraction was subjected to SDS-PAGE and Coomassie Blue® staining, using calibration patterns made with known amounts of CERT and CERT⌬ST. The amount of C 5 -DMB-ceramide in the elute fraction was quantified as described above.
Binding of LPS Alexa Fluor® 488 to recombinant proteins was assayed by essentially the same procedures except that the fluorescent intensity of LPS Alexa Fluor® 488 (excitation at 488 nm; emission at 538 nm) distributed to each fraction was measured without organic solvent extraction.
Effect of (1R,3R)-HPA-12 on Metabolic Labeling of Lipids with [ 14 C]Serine in CHO Cells-LY-A, a CHO-K1-derived mutant cell line, is defective in the trafficking of ceramide from the endoplasmic reticulum to the Golgi apparatus because of a mutation in the CERT gene (6,8).
The LY-A2 cell line is a stable transformant of LY-A expressing the ecotropic retrovirus receptor, and the LY-A2/hCERT cell line is a stable transformant of LY-A2 with a retroviral vector for expression of human CERT cDNA (8). For a concentrated stock, (1R,3R)-HPA-12 was dissolved in dimethyl sulfoxide at 10 mM. Cells were seeded at a density of 1.0 ϫ 10 6 per 6-cm dish in 5 ml of Ham's F-12 medium supplemented with 10% newborn bovine serum, penicillin G (100 units/ml), and streptomycin sulfate (100 g/ml) and cultured for ϳ16 h at 33°C in a 5% CO 2 atmosphere. After two washes with 2 ml of serum-free Ham's F-12 medium, the cells were incubated in 1.5 ml of Nutridoma medium (Ham's F-12 medium supplemented with 1% Nutridoma-SP (Roche Diagnostics) and 25 g/ml gentamicin) containing 1 M (1R,3R)-HPA-12 or the vehicle dimethyl sulfoxide for 15 min on ice and, after the addition of L-[U-14 C]serine (0.75 Ci) to the medium, were incubated for 2 h at 33°C. The metabolically labeled lipids were then analyzed as described previously (38).
Determination of Protein Concentration-Protein concentrations were determined using the bicinchoninic acid protein assay kit (Pierce) with bovine serum albumin as the standard.

Intermembrane Transfer of Lipids by CERT-We
showed previously (8) that CERT efficiently extracts ceramide, but not nonceramide lipids, from phospholipid vesicles. In addition, we showed that CERT greatly facilitates intermembrane transfer of natural long chain C 16 -ceramide (8). However, the lipid substrate specificity of CERT-mediated intermembrane transfer remained unexplored. Thus, we tested the substrate specificity of CERT-mediated intermembrane transfer in a cell-free assay system. Because lipid transfer between artificial membranes might be nonspecifically enhanced by proteins, we used the START domain-deleted CERT⌬ST recombinant in control assays to assess the START domain-dependent transfer of lipids accurately. The lipid transfer assays showed that CERT catalyzes the efficient intermembrane transfer of ceramide, but not sphingosine, sphingomyelin, PtdCho, or cholesterol (Fig. 2), in a START domain-dependent manner. This pattern was consistent with the substrate specificity of the lipid extracting activity of CERT (8). The activity to transfer dioleoylglycerol and dipalmitoylglycerol was ϳ5 and ϳ10%, respectively, of the activity toward ceramide (Fig. 2), raising the possibility that CERT has the potential to transfer diacylglycerol in intact cells (see "Discussion").
Stereochemical Specificity of CERT-mediated Transfer of C 16 -Ceramide-Because ceramide has two chiral carbon atoms at positions C-2 and C-3 of the sphingosine backbone, there can be four stereochemical isomers of C 16 -ceramide, among which D-erythro is the natural configuration (Fig. 1). To examine the stereochemical selectivity of ceramide recognition by CERT, we synthesized the four isomers of C 16 -[ 14 C]ceramide by SCDasecatalyzed in vitro N-palmitoylation of sphingosines. The four isomers could be prepared in radioactively pure forms, although production yields of the unnatural isomers were much less than the yield of the natural isomer (Fig. 1).
In lipid transfer assays using the synthesized stereochemical isomers of C 16 -[ 14 C]ceramide, CERT catalyzed the efficient transfer of D-erythro C 16 -ceramide, but not the unnatural Lerythro-, D-threo-, or L-threo-types (Fig. 3A). Thus, CERT recognizes only the natural isomer among the four stereochemical isomers of C 16 -ceramide.
Recognition of Dihydroceramide and Phytoceramide by CERT-Although ceramide is the predominant hydrophobic backbone of complex sphingolipids in mammalian cells, some sphingolipids also contains dihydroceramide and phytoceramide (21)(22)(23)(24)(25). In vitro lipid transfer assays using synthesized radioactive substrates demonstrated that CERT is capable of catalyzing the intermembrane transfer of C 16-dihydroceramide and C 16 -phytoceramide with ϳ40% efficiency of the C 16 -ceramide transfer (Fig. 3B).
CERT-mediated Transfer of Ceramides Having Fatty Acyl Chains of Various Lengths-To examine the effects of different acyl chain lengths of ceramide on CERT-mediated lipid transfer, we prepared D-erythro-[ 3 H]ceramides having C 14 -, C 16 -, C 18 -, C 20 -, C 22 -, and C 24 -saturated acyl chains and also a C 24:1monounsaturated acyl chain (Fig. 1). In vitro lipid transfer assays showed that the efficiency of the CERT-mediated transfer of ceramide is dependent on its acyl chain length (Fig. 4). Among the molecular species having different acyl chain lengths, C 14 -, C 16 -, C 18 -, C 20 -ceramides were similarly effective. When compared with the amount of C 16 -ceramide transferred, the transfer efficiency of C 22 -and C 24:1 -ceramide was ϳ40%. The transfer of C 24 -ceramide was less (Fig. 4).
There might be the possibility that the same amount of different acyl chain ceramides was not incorporated into donor vesicles, thereby resulting in differences in the transfer effi- ciency. To rule out this possibility, we performed another control experiment. Donor vesicle preparations containing different acyl chain [ 3 H]ceramides were centrifuged at 20,000 ϫ g for 3 min in the presence or absence of Ricinus communis lectin. Regardless of the differences in the acyl chain lengths, most (Ͼ99%) of the radioactivity added to each preparation was precipitated in the presence of the lectin, whereas none (Ͻ1%) of the radioactivity was precipitated in the absence of the lectin. These results indicated that nearly 100% of these radioactive ceramides added to vesicle preparations were actually incorporated into donor vesicles.

CERT Recognizes Fluorescent Short Chain Analogs of Ceramide-
The fluorescent analogs of ceramide C 5 -DMB-ceramide and C 6 -NBD-ceramide have been widely used as probes mimicking natural ceramide in intact cells and in cell-free systems (6,39,40). Because these short chain fluorescent analogs of ceramide spontaneously transfer between membranes (6,40,41), it was difficult to determine accurately the CERT-dependent transfer of C 5 -DMB-ceramide and C 6 -NBD-ceramide in our cell-free transfer assay system. Therefore, to examine if CERT recognized these lipids, we performed a binding assay. CERT could clearly bind both C 5 -DMB-ceramide and C 6 -NBD-ceramide (Fig. 5, A and B).
We were also interested in testing if CERT recognizes the endotoxin LPS, because LPS has a moiety that may be structurally similar to ceramide (42). However, we detected no binding of a fluorophore-conjugated LPS to CERT (Fig. 5C).
Binding Stoichiometry of CERT and Ceramide-We next attempted to determine a binding stoichiometry of CERT and ceramide. For this, we used C 5 -DMB-ceramide as a ceramide ligand, because binding assays at various concentrations of C 5 -DMB-ceramide were feasible under liposome-free conditions. The molar ratio of C 5 -DMB-ceramide bound to CERT in the presence of large excess C 5 -DMB-ceramide was estimated to be about 0.8 (Fig. 5D). These results most likely indicated that the binding stoichiometry of CERT and ceramide is 1:1. The binding assays also suggested that the apparent dissociation constant between CERT and C 5 -DMB-ceramide was about 200 nM (Fig. 5D).
Inhibition of CERT-mediated Transfer of Ceramide by (1R,3R)-HPA-12 in Vitro-(1R,3R)-HPA-12, a chemically synthesized artificial compound, acts as a selective inhibitor of the transport of ceramide from the ER to the site of sphingomyelin synthesis (38). Because (1R,3R)-HPA-12 has structural similarity to D-erythro-ceramide, we hypothesized that CERT might be a target of (1R,3R)-HPA-12. To test this hypothesis, we examined whether (1R,3R)-HPA-12 inhibited CERT-mediated transfer of C 16 -D-erythro-ceramide in the cell-free assay system. As shown in Fig. 6, (1R,3R)-HPA-12 inhibited CERTmediated transfer of ceramide with a 50% inhibitory concentration of ϳ0.5 M. In contrast, its stereochemical isomers and methoxy derivatives, which are inactive as inhibitors of in vivo ceramide trafficking (33), did not affect the in vitro ceramide transfer even at 4 M (Fig. 6). Thus, the inhibition of CERTmediated intermembrane transfer of ceramide by (1R,3R)-HPA-12 was not due to possible nonspecific events such as drug-induced denaturing of proteins or lipids. Collectively, these results indicated that (1R,3R)-HPA-12 is an antagonist of CERT.
(1R,3R)-HPA-12 Inhibits CERT-mediated Trafficking of Ceramide in Intact CHO Cells-To see if (1R,3R)-HPA-12 really inhibits CERT-mediated trafficking of ceramide from the ER to the Golgi site for sphingomyelin synthesis, we examined the effect of the drug on de novo synthesis of sphingomyelin in various CHO cell lines. In wild-type CHO-K1 cells, (1R,3R)-HPA-12 inhibited de novo synthesis of sphingomyelin to the level seen in the drug-free LY-A2 cells, which have a mutation in the endogenous CERT gene (8) (Fig. 7). (1R,3R)-HPA-12 had no effects on sphingomyelin synthesis in mutant LY-A2 cells, consistent with our previous study (38). When LY-A2 cells were transfected with the human CERT cDNA, de novo synthesis of sphingomyelin was restored to the wild-type level (8), and the restored activity of sphingomyelin synthesis in CERT/LY-A2 cells was found to be again sensitive to (1R,3R)-HPA-12 (Fig.  7). These results confirmed that (1R,3R)-HPA-12 inhibits CERT-mediated trafficking of ceramide in intact CHO cells. DISCUSSION In the present study, we explored the substrate selectivity of CERT-mediated lipid transfer reactions in a cell-free assay system, and we showed that CERT is capable of mediating the efficient intermembrane transfer of various ceramide molecular species, including ceramide having C 14 -C 20 saturated acyl chains, C 16 -dihydrocermide, and C 16 -phytoceramide, that naturally exist in mammalian cells (Figs. 3 and 4). In mammalian tissues, "dihydrosphingomyelin" (phosphocholine dihydroceramide) widely exists in smaller amounts than sphingomyelin (21,25). Mammalian tissues might also have "phytosphingomyelin" (phosphocholine phytoceramide) in very small amounts (23,43). A homology search with publicly available tools and data bases predicts that mammals have no additional isoforms of CERT, except for a large splicing variant of CERT. 2 These results suggest that CERT and its splicing variant CERT L mediate the transport of various ceramide molecular species from the ER to the Golgi site, where sphingomyelin and its isoforms are synthesized.
Different members of the Lag1-related family have been suggested to regulate de novo synthesis of different molecular species of ceramide (44 -46). Notably, C 18 -ceramide synthesized in human embryonic kidney 293 cells overproducing UOG1, a Lag1-related family member, is selectively used for the synthesis of glucosylceramide, but not of sphingomyelin (44). The UOG1-dependent channeling of C 18 -ceramide to glucosylceramide synthesis is unlikely due to a possible selectivity of ceramide species by CERT, because CERT catalyzes the efficient transfer of C 18 -ceramide as well as C 16 -ceramide (Fig.  4). However, it remains unclear whether different interactions of different Lag1-related family members with CERT might affect destinations of ceramide species synthesized de novo.
In contrast to the broad specificity of CERT for the ceramide substrate, CERT mediates no transfer of sphingosine, sphingo-myelin, cholesterol, and PdtCho (Fig. 2). Nevertheless, small but significant levels of CERT-mediated transfer of diacylglycerols were reproducibly observed (Fig. 2), consistent with our previous result showing that CERT could extract dioleoylglycerol from artificial membranes even at a much lower efficiency than C 16 -ceramide (8). When one molecule of sphingomyelin is newly synthesized by the PtdCho:ceramide phosphocholine transfer reaction catalyzed by sphingomyelin synthase, one molecule of PtdCho-derived diacylglycerol should be generated. The generated diacylglycerol might cause a feedback inhibition of sphingomyelin synthesis, because diacylglycerol can inhibit the activity of this enzyme in vitro (47). Although the metabolic fate of diacylglycerol generated during the de novo synthesis of sphingomyelin is unknown, a previous study (48) suggested that diacylglycerol generated during the resynthesis of sphingomyelin was rapidly metabolized to triacylglycerol in baby hamster kidney cells treated with extracellular sphingomyelinase. Triacylglycerol synthesis is likely to occur predominantly at the ER (49). Thus, it would be interesting to hypothesize that CERT transports ceramide from the ER to the Golgi and, in turn, transports diacylglycerol from the Golgi to the ER.
The intracellular redistribution of C 5 -DMB-ceramide from the ER to the Golgi region is impaired in LY-A cells having a mutation in the CERT gene and also in energy-poisoned wildtype CHO cells (6,8). Therefore, we have proposed that C 5 -DMB-ceramide may be a good probe for the CERT-mediated pathway of ceramide in cells (8). This proposal was further supported by the present study showing that CERT actually binds C 5 -DMB-ceramide (Fig. 5A). Binding assays with C 5 -DMB-ceramide allowed us to estimate the binding stoichiometry of CERT and ceramide to be 1:1 (Fig. 5D). This is consistent with the START domain of the PtdCho-transfer protein that harbors a single PtdCho molecule in the crystal structure of their complex (50).
Notably, although no clear impairments of the ER-to-Golgi redistribution of C 6 -NBD-ceramide were observed in LY-A cells nor in energy-poisoned cells (6), the cell-free binding assay showed that CERT is also capable of binding C 6 -NBD-ceramide (Fig. 5B). Previous studies with model membranes have shown that C 6 -NBD-ceramide undergoes spontaneous intermembrane transfer at a much faster rate than C 5 -DMB-ceramide (halftimes of transfer equilibration (t1 ⁄2 ) for C 6 -NBD-ceramide and C 5 -DMB-ceramide are ϳ0.4 and ϳ7 min, respectively) (40,41). The t1 ⁄2 of natural C 16 -ceramide has been estimated to be in the order of days (51), and CERT enhances the t1 ⁄2 of C 16 -ceramide to the order of minutes (7). Thus, the rapid spontaneous transfer of C 6 -NBD-ceramide likely masks the CERT-mediated transfer of C 6 -NBD-ceramide in intact cells, even when the latter process occurs.
(1R,3R)-HPA-12 is an inhibitor of an ATP-and cytosol-dependent transport of ceramide from the ER to the site of sphingomyelin synthesis (38). We also demonstrated here that (1R,3R)-HPA-12 inhibited CERT-mediated transfer of ceramide in a cell-free assay system (Fig. 6). HPA-12 derivatives incapable of inhibiting the CERT-dependent transport pathway in cells also did not affect the CERT-mediated intermembrane transfer of ceramide in the cell-free system (Fig. 6). Collectively, we conclude that (1R,3R)-HPA-12 is an antagonist of CERT. Bioinformatic studies have shown that the human genome encodes numerous proteins with putative lipid transfer domains such as START domains (13,14), Sec14-related domains (52), and oxysterol-binding protein-related domains (53) also candidates for new types of medicines. Our studies showing (1R,3R)-HPA-12 as an antagonist of a ceramide transfer protein will hopefully open the way to the development of antagonists of specific lipid transfer proteins.