FVT-1 is a mammalian 3-ketodihydrosphingosine reductase with an active site that faces the cytosolic side of the endoplasmic reticulum membrane.

Sphingolipids are essential membrane components of eukaryotic cells. Their synthesis is initiated with the condensation of l-serine with palmitoyl-CoA, producing 3-ketodihydrosphingosine (KDS), followed by a reduction to dihydrosphingosine by KDS reductase. Until now, only yeast TSC10 has been identified as a KDS reductase gene. Here, we provide evidence that the human FVT-1 (hFVT-1) and mouse FVT-1 (mFVT-1) are functional mammalian KDS reductases. The forced expression of hFVT-1 or mFVT-1 in TSC10-null yeast cells suppressed growth defects, and hFVT-1 overproduced in cultured cells exhibited KDS reductase activity in vitro. Moreover, purified recombinant hFVT-1 protein exhibited NADPH-dependent KDS reductase activity. The identification of the FVT-1 genes enabled us to characterize the mammalian KDS reductase at the molecular level. Northern blot analyses demonstrated that both hFVT-1 and mFVT-1 mRNAs are ubiquitously expressed, suggesting that FVT-1 is a major KDS reductase. We also found the presence of hFVT-1 variants, which were differentially expressed among tissues. Immunofluorescence microscopic analysis revealed that hFVT-1 is localized at the endoplasmic reticulum. Moreover, a proteinase K digestion assay revealed that the large hydrophilic domain of hFVT-1, which contains putative active site residues, faces the cytosol. These results suggest that KDS is converted to dihydrosphingosine in the cytosolic side of the endoplasmic reticulum membrane. Moreover, the topology studies provide insight into the spatial organization of the sphingolipid biosynthetic pathway.

Sphingolipids are essential lipid components of the eukaryotic membrane. Recent studies have demonstrated that sphingolipids, together with cholesterol, form lipid microdomains or rafts that play important roles in membrane traffic and cell signaling (1). Moreover, some sphingolipid metabolites such as ceramide, sphingosine, and sphingosine 1-phosphate function as bioactive lipid molecules that regulate cell proliferation, differentiation, apoptosis, and cell migration (2).
Sphingolipids consist of a hydrophobic ceramide backbone and a hydrophilic polar head group. Mammalian sphingolipids include sphingomyelin, with phosphocholine as its head group, and hundreds of glycosphingolipids with carbohydrate head groups. Biosynthesis of these sphingolipids begins with the condensation of palmitoyl-CoA and L-serine by serine palmitoyltransferase, producing 3-ketodihydrosphingosine (KDS) 1 (3,4). The carbonyl group at the C-3 position of KDS is then reduced by KDS reductase to generate dihydrosphingosine (DHS) (3)(4)(5). Subsequently, DHS is N-acylated by ceramide (dihydroceramide) synthase to form dihydroceramide. A double bond is then formed between the C-4 and C-5 positions of dihydroceramide by dihydroceramide desaturase, creating ceramide (3,4). This biosynthetic pathway is well conserved, even in lower eukaryotes like the yeast Saccharomyces cerevisiae, with the exception of the final step (4). In the yeast, dihydroceramide desaturase does not exist; hence sphingosinetype ceramide is not produced. Instead, dihydroceramide or phytoceramide serves as the hydrophobic backbone of the sphingolipids, with phytosphingosine (PHS) being generated by hydroxylation at the C-4 position of DHS by C-4 hydroxylase (6,7).
Most of the enzymes involved in sphingolipid biosynthesis have been identified in recent years, and the contribution of yeast genetic approaches in these studies has been quite significant. LCB1 and LCB2 were identified as genes required for the serine palmitoyltransferase activity in yeast (8 -11). Subsequent studies demonstrated that Lcb1p and Lcb2p are subunits of serine palmitoyltransferase (11). By their homology, mammalian LCB1 and LCB2 were identified and found to possess serine palmitoyltransferase activity (12)(13)(14). Similarly, Lag1p and Lac1p share homology and have overlapping functions in dihydroceramide/phytoceramide synthesis in yeast (15,16). Their homologs, UOG1, TRH1, and TRH4, recently were demonstrated to be involved in dihydroceramide/ceramide synthesis in mammals (17)(18)(19). Interestingly, Sur2p, a C-4 hydroxylase found in yeast, shares the same desaturase/hydroxylase superfamily as DES1, which is a mammalian dihydroceramide desaturase (6,7,20).
The yeast gene TSC10 was identified in a screening for temperature-sensitive suppressors of the Ca 2ϩ -sensitive ⌬csg2 mutant (5). This mutant exhibits a defect in the synthesis of mannosylinositol phosphorylceramide, one of three myo-inositol-containing sphingolipids in yeast (21). The temperature-sensitive tsc10 mutants accumulated KDS and lacked KDS reductase activity (5). Moreover, purified recombinant Tsc10p catalyzed the conversion of KDS to DHS in an NADPH-dependent manner, indicating that Tsc10p is a KDS reductase (5).
Although Tsc10p was identified in 1998 (5), experimental evidence identifying its mammalian homolog has remained lacking thus far. In the present study, we demonstrate that the KDS reductase in mammals is encoded by the follicular lymphoma variant translocation-1 (FVT-1) gene, a gene identified previously in follicular lymphoma as being juxtaposed to an immunoglobulin J segment by chromosome translocation (22). With the identity of this mammalian gene and its product in hand, we are now at last able to establish the characteristics for this mammalian KDS reductase and examine its role in sphingolipid synthesis. Northern blotting demonstrated that FVT-1 is expressed throughout the body in all tissues examined, although its expression varied among tissues, suggesting that FVT-1 is a ubiquitous KDS reductase. Although FVT-1 had been predicted to be a secreted protein (22), our immunofluorescence microscopic analysis showed that FVT-1 is localized in the endoplasmic reticulum (ER). Moreover, we also reveal that the domain encompassing its active site is exposed to the cytosol, indicating that sphingolipid synthesis proceeds on the cytosolic leaflet of the ER membrane, at least up to the production of DHS.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-Human embryonic kidney (HEK293T) cells and HeLa cells were grown in Dulbecco's modified Eagle's medium D6429 or D6046, respectively (Sigma) containing 10% fetal calf serum and supplemented with 100 units/ml penicillin and 100 g/ml streptomycin. HEK293T cells were grown on collagen-coated dishes. Transfection was performed using LipofectAMINE Plus TM Reagent (Invitrogen).
Plasmids-The pAK80 plasmid is a yeast cloning vector for TDH3 promoter-dependent expression (24). The pAK157 plasmid, a derivative of pAK80, was constructed to produce an N-terminally FLAG-tagged fusion protein. For construction of the pAK287 plasmid (FLAG-TSC10), the TSC10 gene was amplified using yeast genomic DNA and primers 5Ј-GGATCCATGAAGTTTACGTTAGAAGACC-3Ј and 5Ј-TGAGAGTG-GATGTGCTCTCAGTTGG-3Ј. The amplified fragment was cloned into pGEM-T Easy to generate pAK270. The pAK287 plasmid was produced by cloning the 1.0-kb BamHI-NotI fragment of pAK270 into the BamHI-NotI site of pAK157.
The mouse FVT-1 (mFVT-1) cDNA was amplified by PCR using total cDNA prepared from the mouse teratocarcinoma cell line F9 and primers mFVT1-F1 (5Ј-TCGGCAGAGATGTTGCTGTTGGCCG-3Ј) and mFVT1-R1 (5Ј-GGGGCAAGGTTTAGGCAGTTTTGTC-3Ј). The amplified fragment was cloned into pGEM-T Easy to create pAK565. The obtained nucleotide sequence has been deposited with GenBank TM with accession number AY634684. For expression in yeast, pAK572 was constructed by cloning the 1.1-kb NotI-NotI fragment of pAK565 into the NotI site of pAK80.
Preparation of Soluble and Membrane Fractions-HEK293T cells transfected with the indicated plasmids were washed twice with phosphate-buffered saline, suspended in buffer A (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 10% glycerol, 1 mM dithiothreitol, 1 mM EDTA, 1ϫ protease inhibitor mixture (Complete TM , Roche Diagnostics), and 1 mM phenylmethylsulfonyl fluoride (PMSF)), and sonicated. After centrifugation at 300 ϫ g for 3 min at 4°C, the resulting supernatant was used as a total fraction. For preparing soluble and membrane fractions, the total fraction was subjected to centrifugation at 100,000 ϫ g for 1 h at 4°C.
In Vitro KDS Reductase Assay-The reaction mixture contained 50 mM Tris-HCl (pH 7.5), 10% glycerol, 150 mM NaCl, 1 mM EDTA, 10 mM 2-mercaptoethanol, 1ϫ protease inhibitor mixture, 1 mM PMSF, 0.2% Triton X-100, 100 M NADPH, 50 M KDS, and purified MBP-⌬N-hFVT-1 protein. When total membrane fractions were used instead of the purified protein, 1 mM NADPH and 100 M KDS were included. All reactions were performed by incubating the samples at 37°C. The reactions were terminated by adding 1 M NH 4 OH, and the oxidation of NADPH was monitored by measuring the decreased absorbance at 340 nm. Alternatively, for lipid extractions the reactions were terminated by 3.75 volumes of chloroform/methanol (1:2, v/v), followed by 1.25 volumes of chloroform and 1.25 volumes of 1% KCl. Phases were separated by centrifugation, and the organic phase was recovered, dried, and suspended in chloroform/methanol (2:1, v/v). The lipids were resolved by TLC on Silica Gel 60 high performance TLC plates (Merck) with chloroform, methanol, 2 N NH 4 OH (40:10:1, v/v). KDS and DHS were visualized by spraying the plates with 0.3% ninhydrin in 1-butanol/acetic acid (100:3, v/v) and heating at 180°C for 1 min. KDS and DHS were purchased from Matreya, Inc. (Pleasant Gap, PA) and Sigma, respectively.
Northern Blotting-Poly(A) ϩ RNA blots containing 1 g of RNA from human tissues (Clontech, BD Biosciences, Palo Alto, CA) and total RNA blots containing 20 g of RNA from mouse tissues (Seegene Inc., Seoul, Korea) were used. The hFVT-1 fragments were amplified using the pAK567 plasmid and primers hFVT1-F1 and hFVT1-R1, whereas the mFVT-1 fragments were amplified using pAK565 and primers mFVT1-F1 and mFVT1-R1. These fragments were labeled with [ 32 P]dCTP by random priming using the random primer DNA labeling kit, version 2 (TAKARA Bio Inc., Shiga, Japan), to generate their probes. Hybridization was carried out in ExpressHyb buffer (Clontech, BD Biosciences) for 2 h at 68°C.
Proteinase K Digestion Assay-HEK293T cells transfected with pCEpuro hFVT-1 were washed with phosphate-buffered saline twice, suspended in buffer D (20 mM HEPES/NaOH (pH 7.5), 0.25 M sucrose, 1 mM dithiothreitol, 1ϫ protease inhibitor mixture, and 1 mM PMSF), and lysed using an electric Potter homogenizer for 10 strokes. After removal of cell debris by centrifugation at 300 ϫ g for 3 min at 4°C, the supernatant was subjected to ultracentrifugation at 100,000 ϫ g for 1 h at 4°C. The resulting pellet was suspended in buffer D (without the protease inhibitor mixture or PMSF) and treated with 0.5 mg/ml proteinase K at 4°C for 2 h in the presence or absence of 1% Triton X-100. After termination of the digestion with 1 mM PMSF, total proteins were precipitated with 5% trichloroacetic acid on ice for 20 min. Protein precipitates were washed with 5% trichloroacetic acid and with acetone and then suspended in SDS sample buffer (62.5 mM Tris-HCl (pH 6.8), 2% SDS, 10% glycerol, and a trace amount of bromphenol blue) containing 1 mM PMSF, separated by SDS-PAGE, and subjected to immunoblotting.

RESULTS
Identification of FVT-1-To identify a mammalian KDS reductase, the GenBank TM data base was searched using the BLAST program for sequences similar to yeast Tsc10p. Although no apparent homologs could be found, various members of the short-chain dehydrogenase/reductase (SDR) family, to which Tsc10p belongs, exhibited moderate sequence similarities (ϳ20% identity). These included FVT-1, 11-␤-hydroxysteroid dehydrogenase type 3, some short-chain dehydrogenase/reductase 10 isoforms, 17-␤-hydroxysteroid dehydrogenase type 2, and sepiapterin reductase. Of these, we chose human FVT-1 (23.5% identity and 40.7% similarity with Tsc10p) as a candidate for being a mammalian KDS reductase because its amino acid length (332 amino acids) is similar to that of Tsc10p (320 amino acids) and because its function was unknown. Additionally, we found a putative mouse homolog of hFVT-1. The mFVT-1 sequence indicated 92.8% identity and 97.3% similarity to hFVT-1 and 20.8% identity and 39.5% similarity to Tsc10p (Fig. 1).
Both FVT-1 proteins contain a 25-amino acid extension at the N terminus, which is absent in Tsc10p. This N-terminal stretch is considerably hydrophobic and is predicted to function as a transmembrane segment (Fig. 1). In addition, two putative transmembrane segments are present in hFVT-1 and mFVT-1 at their C termini (Fig. 1). The large hydrophilic domain located between the putative first and second transmembrane segments of FVT-1 contains a region (corresponding to amino acid residues 35-269 of hFVT-1) that exhibits similarity to SDR family members. In the SDR family only a single Tyr residue is strictly conserved, and site-directed mutagenesis and x-ray crystallography have demonstrated that this Tyr residue is crucial to enzyme activity (27)(28)(29)(30)(31). A Lys residue four positions downstream is also conserved in most SDR members (32), so a catalytic mechanism based on the Tyr-X-X-X-Lys motif has been proposed (33,34). Tyr-186 and Lys-190 in hFVT-1 correspond to this motif. The same motif is found in mFVT-1 and Tsc10p (Fig. 1, dotted residues). SDR members require NAD(H) or NADP(H) as a coenzyme for their enzyme activity. A Gly-X-X-X-Gly-X-Gly segment, generally characteristic of the coenzyme-binding fold in dehydrogenases (32,34,35), is also present in FVT-1 and Tsc10p proteins (Fig. 1, dashed line). A single Lys residue is observed at the Ϫ3 position in the C termini of both hFVT-1 and mFVT-1. Lys is known to be important for ER membrane proteins to be retained at the ER, although typically ER retention signals are KKXX and KXKXX sequences (34). Tsc10p contains a KKXX type sequence at its C terminus (Fig. 1).
FVT-1 Is a KDS Reductase-First we examined whether ectopic expression of hFVT-1 or mFVT-1 in the TSC10-null yeast cells complemented the growth phenotype. Because sphingolipids are essential for cell viability, ⌬tsc10 cells cannot grow normally, unless the addition of DHS or PHS to the medium bypasses the requirement of de novo sphingolipid synthesis (5). KHY625 (⌬tsc10) cells bearing a vector plasmid could not grow on SC plates lacking uracil (SC Ϫ URA plates) in the absence of PHS but grew once it was added to the medium ( Fig.  2A). The introduction of pAK287, encoding N-terminally FLAG-tagged TSC10, allowed KHY625 cells to grow in the absence of PHS ( Fig. 2A). Moreover, both pAK572 and pAK574 plasmids, encoding the mFVT-1 and hFVT-1 genes, respectively, enabled the growth of KHY625 cells (Fig. 2A). These results indicate that hFVT-1 and mFVT-1 are functional homologs of Tsc10p.
We next investigated the KDS reductase activity of hFVT-1 in vitro. HEK293T cells were transiently transfected with a pCE-puro hFVT-1 plasmid or a control vector plasmid. Cell lysates were prepared and subjected to immunoblotting using anti-hFVT-1 antibody. The hFVT-1 protein was detected in the lysates from pCE-puro hFVT-1 plasmid-transfected cells as a reductases. An alignment of the amino acid sequences of hFVT-1, mFVT-1, and Tsc10p was generated using the ClustalV (50) and BOX-SHADE programs (Institute for Animal Health, Surrey, United Kingdom). The corresponding GenBank TM accession numbers are CAA45197 (hFVT-1), AY634684 (mFVT-1), and CAA85228 (Tsc10p). Black boxes indicate identical residues, and gray boxes show amino acid similarity. The putative transmembrane segments are indicated by lines. Dotted residues indicate the putative active site motif, Tyr-X-X-X-Lys, which is conserved in the SDR family. The predicted Gly-X-X-X-Gly-X-Gly segment functioning in NAD(H) or NADP(H) binding is marked with a dashed line.
Cell fractionation by centrifugation at 100,000 ϫ g for 1 h demonstrated that most of the hFVT-1 was localized in the membrane fraction (Fig. 3B). Therefore, we performed an in vitro KDS reductase assay using the membrane fractions. In assays using membrane fractions (2.5 g of proteins) prepared from HEK293T cells transfected with the vector plasmid, no significant decrease in KDS levels was observed, and little DHS was produced (Fig. 3C, lane 6). In contrast, the membrane fractions (2.5 g of proteins) from HEK293T cells overproducing hFVT-1 converted all of the KDS to DHS (Fig. 3C, lane 8). This conversion was highly efficient and dose-dependent so that as little as 0.125 g of protein converted most of the KDS present (Fig. 3D). Thus, the expression of hFVT-1 generated cellular KDS reductase activity.
To exclude the possibility that FVT-1 is not a KDS reductase but an activator, we tried to purify recombinant hFVT-1 protein from E. coli, which does not contain sphingolipids or KDS reductase. The full-length hFVT-1 fused N-terminally to MBP (MBP-hFVT-1) failed to express in E. coli (data not shown). However, removal of the N-terminal hydrophobic stretch, which is not conserved in Tsc10p, allowed the expression of the truncated form of MBP-hFVT1 (MBP-⌬N-hFVT-1). This Nterminal truncation did not abrogate the function of hFVT-1 because FLAG-tagged ⌬N-hFVT-1 (FLAG-⌬N-hFVT-1) could enable the growth of ⌬tsc10 cells in the absence of PHS, similarly to its full-length control (FLAG-hFVT-1) (Fig. 2B). Once expressed in E. coli, MBP-⌬N-hFVT-1 was recovered in the membrane fraction. The crude membrane fraction was solubilized with Triton X-100, and MBP-⌬N-hFVT-1 was purified using an amylose resin. Based on Coomassie staining MBP-⌬N-hFVT-1 was purified to near homogeneity (Fig. 4A). Using the purified protein, we performed an in vitro KDS reductase assay. KDS was unchanged upon incubation with buffer (Fig.  4B, lane 4), whereas inclusion of the MBP-⌬N-hFVT-1 protein efficiently converted the KDS to DHS (Fig. 4B, lane 3). This reaction was completely dependent on NADPH (Fig. 4B, lane  1), as NADH could not substitute for NADPH (Fig. 4B, lane 2). Using different concentrations of KDS or NADPH (Fig. 5) we estimated K m values for the reductase activity of the purified protein to be 3 and 28 M for KDS and NADPH, respectively. These results indicated that hFVT-1 itself possesses an NADPHdependent KDS reductase activity.
Tissue Distribution of FVT-1-To investigate tissue-specific expression patterns of mFVT-1 and hFVT-1, we performed high stringency Northern blot analysis using mRNA extracted from different tissues. A predominant 2.2-kb mFVT-1 mRNA was detected in every tissue examined, although its level varied (Fig. 6A). The highest expression was observed in placenta. Lung, kidney, stomach, and small intestine expressed high levels of mFVT-1 mRNA, whereas only low levels of expression were detected in heart, spleen, and skeletal muscle (Fig. 6A). Similarly, hFVT-1 mRNA was expressed ubiquitously in the tissues examined (Fig. 6B). In contrast to mFVT-1, hFVT-1 mRNA was detected as two bands (2.7 and 2.5 kb) (Fig. 6B). In most tissues expression of the 2.5-kb mRNA was higher than that of the 2.7-kb mRNA (Fig. 6B). However, the 2.7-kb mRNA was the major species in placenta, and nearly equal levels of the two mRNAs were observed in liver (Fig. 6B). Moreover, the expression profile of hFVT-1 was different from that of mFVT-1. The levels of the hFVT-1 mRNAs were the highest in skeletal muscle and heart and lowest in the colon, thymus, and peripheral blood leukocytes (Fig. 6B). A 1.3-kb hFVT-1 transcript, possibly another splice variant, was detected in the heart and skeletal muscle tissues (Fig. 6B).
FVT-1 Is Localized in the ER-To investigate the subcellular localization of hFVT-1, indirect immunofluorescence microscopic analysis was performed using HeLa cells transfected with a pCE-puro hFVT-1 plasmid for 24 h. hFVT-1 was detected as a reticular structure with a pattern similar to that of ER (Fig. 7A). Therefore, we applied double staining of the cells with the anti-FVT-1 antibody and an antibody directed to the ER retention signal of the ER luminal protein KDEL. The staining pattern of hFVT-1 showed a complete overlap with the ER marker KDEL (Fig. 7B), indicating that hFVT-1 is an ER-resident protein.
The Active Site of FVT-1 Faces the Cytosolic Side of the ER-Both FVT-1 proteins contain a large hydrophilic domain between the putative first and second transmembrane-spanning segments. This domain exhibits similarity to domains of other members in the SDR family and contains the putative NADPH binding site and the active site motif Tyr-X-X-X-Lys. Determination of the membrane topology of FVT-1 is important for identifying on which side of the ER membrane DHS is produced. For this purpose, intact organelles were prepared from HEK293T cells transfected with pCE-puro hFVT-1 and were treated with proteinase K. Calnexin is a type I ER membrane protein with an N-terminal large domain that is exposed to the lumen of the ER (36). We chose calnexin as a control to assess the proper orientation of the isolated ER. Protease treatment reportedly results in the production of C-terminally truncated forms of calnexin (calnexin-⌬C) (36). Accordingly, upon treatment with proteinase K, calnexin was converted to calnexin-⌬C, which reacted with an antibody raised against the N-terminal region of calnexin (amino acid residues 1-70) (Fig.  8A, lane 5). This band disappeared after disrupting the ER membrane with Triton X-100 (Fig. 8A, lane 6), indicating that calnexin-⌬C is protected by the ER membrane. On the other

FVT-1 Is a 3-Ketodihydrosphingosine Reductase
hand, the FVT-1 protein was not protected against cleavage by proteinase K in the absence of Triton X-100 (Fig. 8A, lane 2). Because the anti-hFVT-1 antibody was raised against an epitope between residues 26 and 292 in the middle region of the hFVT-1 protein, this result suggests that the large hydrophilic domain of FVT-1 is exposed to the cytosolic side of the ER membrane. DISCUSSION During the last decade, most of the mammalian genes responsible for sphingolipid biosynthesis have been identified (12-14, 17-20, 37, 38). However, the identification of the gene for KDS reductase, which is involved in the second step of de novo sphingolipid synthesis, has remained a missing link. We searched for a mammalian gene encoding a product similar to Tsc10p and found FVT-1 as a functional mammalian KDS reductase. First, in ⌬tsc10 yeast cells forced expression of either human or mouse isoforms of FVT-1 restored the growth defect in the absence of PHS ( Fig. 2A). Second, expression of hFVT-1 conferred KDS reductase activity to HEK293T cells (Fig. 3C). Third, purified recombinant MBP-⌬N-hFVT-1 exhibited unequivocal KDS reductase activity (Fig. 4B). Together, these results provide strong support for the hypothesis that FVT-1 is a mammalian KDS reductase.
We found that FVT-1 is localized at the ER (Fig. 7), which is consistent with previous results that sphingolipid biosynthesis, up to the point of ceramide formation, takes place in the ER (39).
FVT-1 appears to be ubiquitous and somewhat conserved because the mouse and human genes share sequence homology and, to some extent, expression patterns. Although it is possible that multiple KDS reductases exist in mammalian cells, the ubiquitous expression of both hFVT-1 and mFVT-1 (Fig. 6) suggests that FVT-1 is a major KDS reductase. Interestingly, two predominant FVT-1 mRNA isoforms (2.5 and 2.7 kb) were detected in human tissues. Expression of the 2.5-kb mRNA was higher than that of the 2.7-kb form in most tissues. However, in the liver their levels were nearly equal, whereas in the placenta the 2.7-kb form was the major species. We found that the cDNA in the GenBank TM with accession number AK025120 comprises a full-length hFVT-1 cDNA. This cDNA is composed of 2044 bp in addition to the poly(A) sequence. Using the BLAST program, we compared the sequence of AK025120 with other human expressed sequence tag clones containing hFVT-1 cDNA and found that most of the expressed sequence tag clones, including AI754522, BQ010432, and BQ020986, have a 3Ј-untranslated region identical to that of AK025120. However, some expressed sequence tag clones, such as BM995782, BQ182543, and CA438131, contain an extra 100 bp in the 3Ј-untranslated region. Therefore, the two mRNA species detected by Northern blotting (Fig. 6) may reflect these differences in the length of the 3Ј-untranslated region.
FVT-1 was originally identified in a study searching for genes involved in tumor progression (22). Variant t(2;18)(p11; q21) chromosome translocation, noted in follicular lymphomas, induced the expression of the BCL-2 proto-oncogene, although rearrangement of BCL-2 was not observed by Southern blot analysis (40). Subsequent study revealed that in this translocation, a J segment is juxtaposed to the FVT-1 gene, which is localized 10 kb upstream of the BCL-2 locus (22). FVT-1 was weakly expressed in all of the normal hematopoietic tissues tested, whereas high levels of expression were observed in some T-cell malignancies (22). At the present time, any involvement of FVT-1 in tumorigenesis remains unclear.
In the report in which hFVT-1 was first identified, hFVT-1 was predicted to be a secreted protein based on the presence of the N-terminal hydrophobic stretch (22), although its localiza-  4) or 0.5 mg/ml (lanes 2, 3, 5, and 6) proteinase K (Prot. K) at 4°C for 2 h in the absence (lanes 1, 2, 4, and 5) or presence (lanes 3 and 6) of 1% Triton X-100. Proteins were separated by SDS-PAGE, followed by immunoblotting with anti-hFVT-1 (lanes 1-3) or anti-calnexin (lanes 4 -6) antibodies. B, a model of the proposed topology of FVT-1 relative to the established topology of calnexin (36). FVT-1 is proposed to be an integral membrane protein with three transmembrane domains. The N terminus is located in the lumen of the ER, whereas the large hydrophilic domain containing the active site residues and the C terminus is located in the cytosol. tion was not examined. However, this N-terminal region lacks the features characteristic of signal sequences. Signal sequences are typically composed of a short positively charged N-terminal segment (n-region), a central hydrophobic segment (h-region), and a more polar C-terminal segment (c-region) (41). Neither human nor mouse FVT-1 has basic amino acids at its N terminus; thus both are lacking the n-region. The distribution of charged residues on either side of the hydrophobic core of signal sequences or transmembrane segments is important for determining the membrane orientation. The "positive-inside rule" indicates that flanking sequences with positive charges are predominantly localized to the cytosolic side of the membrane (42). Thus, basic amino acids in the n-region of signal sequences are important in orientating the signal sequence to N cyt /C exo (cytoplasmic N and exoplasmic C termini). Near the N-terminal hydrophobic stretch, hFVT-1 and mFVT-1 both contain a Lys residue, which is downstream to the hydrophobic core, but lack acidic amino acids. Thus, this hydrophobic stretch is predicted to be oriented N exo /C cyt , an inverse alignment to that of a signal sequence. Therefore, the N terminus does not appear to be a signal sequence but rather a transmembrane domain. Indeed, we obtained corroborating evidence that FVT-1 is an ER membrane protein (Fig. 7).
We propose a model for the topology of FVT-1 (Fig. 8B) in which FVT-1 has three transmembrane segments, with the N terminus facing the lumen of the ER and the large catalytic domain and C terminus located in the cytosol. This model is supported by experimental data and predictions. First, proteinase K digestion demonstrated that the large hydrophilic domain, which contains the active site residues, faces the cytosol (Fig. 8A). This result is consistent with the predicted N exo /C cyt orientation of the putative first transmembrane domain described above. Second, our results on the topology of Tsc10p are consistent with the topology model for FVT-1. FVT-1 and Tsc10p share a similar hydropathy profile, except that Tsc10p lacks the N-terminal putative transmembrane segment. Proteinase K digestion using N-terminally FLAG-tagged Tsc10p in conjunction with the results of a C-terminal reporter demonstrated that both the N terminus and C terminus of Tsc10p are exposed to the cytoplasm. 2 Finally, both Tsc10p and FVT-1 contain putative ER retention signals at their C termini. This signal is known to function in the cytosolic side of the ER membrane (43).
Previous biochemical analyses using mouse liver tissues suggested that sphingolipid synthesis is initiated in the cytoplasmic leaflet of the lipid bilayer of the ER membrane and proceeds to the point of dihydroceramide synthesis (39). This conclusion is based on the sensitivity of the sphingolipid biosynthetic activities to exogenous proteases and the membraneimpermeable compound 4,4Ј-diisothiocyanostilbene-2,2Ј-disulfonic acid (39). However, the possibility cannot be excluded that some membrane proteins present residues essential for activity on both sides of the membrane, and the precise topology can only be determined by detecting specific protein domains rather than enzymatic activity. Such topological study, which was performed recently, indicated that the active site of the serine palmitoyltransferase catalyzing the first step of sphingolipid biosynthesis is indeed exposed to the cytosolic side of the membrane (44). In this study, we have demonstrated that the domain containing the active site of the KDS reductase also faces the cytosol. Therefore, these results are consistent with the hypothesis that production of KDS and DHS takes place on the cytosolic side of the ER membrane. However, the side of the ER on which ceramide is synthesized remains unclear. DHS and PHS are generated in the luminal side of the ER mem-brane by the yeast long-chain base 1-phosphate phosphatase Lcb3p and are used as precursors of ceramide (45). Moreover, galactosylceramide and glucosylceramide are synthesized on the luminal side of the ER and on the cytosolic side of the Golgi apparatus, respectively, suggesting that their precursor, ceramide, may exist in both sides (46 -49). Thus, future topological studies are required to determine the precise orientation of ceramide synthase to reveal the complete trans-bilayer organization of the sphingolipid biosynthetic pathway.