Characterization of Ceramide Synthase 2

Ceramide is an important lipid signaling molecule and a key intermediate in sphingolipid biosynthesis. Recent studies have implied a previously unappreciated role for the ceramide N-acyl chain length, inasmuch as ceramides containing specific fatty acids appear to play defined roles in cell physiology. The discovery of a family of mammalian ceramide synthases (CerS), each of which utilizes a restricted subset of acyl-CoAs for ceramide synthesis, strengthens this notion. We now report the characterization of mammalian CerS2. qPCR analysis reveals that CerS2 mRNA is found at the highest level of all CerS and has the broadest tissue distribution. CerS2 has a remarkable acyl-CoA specificity, showing no activity using C16:0-CoA and very low activity using C18:0, rather utilizing longer acyl-chain CoAs (C20–C26) for ceramide synthesis. There is a good correlation between CerS2 mRNA levels and levels of ceramide and sphingomyelin containing long acyl chains, at least in tissues where CerS2 mRNA is expressed at high levels. Interestingly, the activity of CerS2 can be regulated by another bioactive sphingolipid, sphingosine 1-phosphate (S1P), via interaction of S1P with two residues that are part of an S1P receptor-like motif found only in CerS2. These findings provide insight into the biochemical basis for the ceramide N-acyl chain composition of cells, and also reveal a novel and potentially important interplay between two bioactive sphingolipids that could be relevant to the regulation of sphingolipid metabolism and the opposing functions that these lipids play in signaling pathways.

The past decade has seen an upsurge of interest in sphingolipids (SLs), 2 due largely to the extraordinary number of complex species that have been found in eukaryotes (1), as well as the involvement of the lipid backbones in signaling pathways as both first and second messengers (2)(3)(4). Indeed, ceramide (3) and sphingosine 1-phosphate (S1P) (5) appear to play opposing roles in cell proliferation, migration, and survival, which underscores how the balance of the levels of these two lipids has ramifications for diverse pathological and pathophysiological processes (6 -8).
In mammals, ceramide is synthesized by a family of six enzymes, ceramide synthases (CerS) 1-6 (9), each of which uses a relatively restricted subset of fatty acyl-CoAs for N-acylation (10 -12) of the sphingoid long chain base. Thus, CerS1 and CerS5, which are the best characterized CerS proteins, synthesize C18-and C16-ceramide, respectively (10,11,13,14), whereas CerS2 and -3 appear to have a broader specificity (15). The existence of these six CerS genes in mammals implies an important and largely unexplored role for ceramides containing specific fatty acids in cell physiology (9). One possibility is that different tissues contain ceramides with defined fatty acids, necessitating the presence of specific CerS in specific tissues for their synthesis. However, with the exception of an early study by semi-quantitative reverse transcription-PCR (11), little is known about CerS tissue distribution.
This study describes the characterization of CerS2, which has received relatively little attention. We demonstrate that CerS2 mRNA occurs at much higher levels than most other CerS, has the broadest tissue distribution, and synthesizes ceramides containing mainly C20 -C26 fatty acids, with little or no synthesis of C16-and C18-ceramides. Moreover, CerS2 activity is inhibited by S1P via interaction of S1P with an S1P receptorlike motif found only in CerS2. This unique link between S1P and a key enzyme of ceramide metabolism might be of significance to understanding the interplay between these two lipids in metabolic and in signaling pathways.

EXPERIMENTAL PROCEDURES
Materials-D-erythro- [4, H]Sphinganine was synthesized as described (16). S1P was from Sigma-Aldrich or from Avanti Polar Lipids (Alabaster, AL); fatty acyl-CoAs and the internal standards for liquid chromatography electrospray ionizationtandem mass spectrometry (LC ESI MS/MS) were also from Avanti. An anti-protein disulfide isomerase antibody was from Stressgen (Victoria, BC, Canada), and an anti-hemagglutinin (HA) antibody was from Santa Cruz Biotechnology (Santa Cruz, CA). Horseradish peroxidase was from Jackson Laboratories (Bar Harbor, MA). Pfu polymerase was from Promega (Madison, WI) or from Stratagene (La Jolla, CA). TaqMan TM was from Applied Biosystems (Foster City, CA). A PerfectPure RNA Kit was from 5Prime (Gaithersburg, MD). A Reverse-iT first strand synthesis kit was from Thermo Scientific (Epsom, UK). Silica gel 60 TLC plates were from Merck. All solvents were of analytical grade and were purchased from Biolab (Jerusalem, Israel).
Real-time qPCR-Tissues were harvested from 6-to 8-weekold mice; females were used for all tissues except prostate and testis. RNA was isolated using a PerfectPure RNA kit according to manufacturer's instructions, which included a DNase step. cDNA synthesis was performed using a Reverse-iT first strand synthesis kit using random decamers with 30-min incubation at 42°C and then at 47°C. cDNA generation demonstrated equivalent efficiency of synthesis with input RNA ranging from 15 to 500 ng per reaction (supplemental Fig. S1). 100 ng of total RNA was used to determine expression levels of mouse CerS mRNA, using TaqMan TM analysis and a 7300 Sequence Detection System (Applied Biosystems). Relative CerS expression levels were determined in all tissues as compared with brain; quantitative analysis was assessed in brain by comparison to a standard curve generated via dilution of expression plasmids for each gene. To control for variability of RNA input, all PCR reactions were normalized to the amount of hypoxanthine guanine phosphoribosyltransferase 1 mRNA. Primer/probe sets for CerS1 (Mm00433562_m1), CerS2 (Mm00504086_m1), CerS4 (Mm01212479_m1), CerS5 (Mm00510996_g1), and CerS6 (Mm00556165_m1), and for hypoxanthine guanine phosphoribosyltransferase 1 (Mm00446968_m1) were pre-validated sets obtained from Applied Biosystems. CerS3 was custom ordered from Applied Biosystems and designed to span exon 2 and exon 3 (Locus DQ646881). Oligonucleotides used for generation of qPCR templates are given in Table 1.
Short Interfering RNA (siRNA)-siRNAs were subcloned into the pSUPER vector according to the manufacturer's instruc-tions (OligoEngine, Seattle, WA). Two siRNA targets were chosen for CerS2: siCerS2i, 5Ј-AAGCAGGTGGAAGTAGAGCT-TTT-3Ј, and siCerS2ii, 5Ј-AAGCCAGCTGGAGATTCACA-TTT-3Ј. The sequences were chosen because they recognize all known CerS2 isoforms (9) and do not recognize other CerS genes. CerS2 knockdown was accomplished by transfecting Hek cells with the pSUPER vector using the calcium phosphate method. After various times of incubation, total RNA was extracted using the RNeasy Mini Kit (Qiagen). Reverse transcription was performed using the EZ-First strand cDNA synthesis kit (Biological Industries, Beit Haemek, Israel), and PCR was performed using the primers listed in Table 1.
Cell Culture and Transfection-Human embryonic kidney cells (Hek 293) were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 100 IU/ml penicillin, and 100 g/ml streptomycin. Hek 293 cells were transfected with human CerS genes using the calcium phosphate method (0.25 g of plasmid per cm 2 of culture dish), which gave ϳ90% transfection efficiency. Human CerS genes (17) in a pCMV vector, with an N terminus FLAG tag, were obtained from Dr. Richard Kolesnick (Sloan Kettering Institute). CerS2 was subsequently subcloned into a pcDNA3 vector containing an HA tag, which was located at the C terminus (CerS2-HA). Site-directed mutagenesis of CerS2-HA was performed using 2-Step PCR and Touch-up cycling conditions.
ESI-MS/MS-SL analyses by ESI-MS/MS were conducted using a PE-Sciex API 3000 triple quadrupole mass spectrometer and an ABI 4000 quadrupole-linear ion trap mass spectrometer as described previously (10,11,18,19). Hek 293 cells were transfected with pcDNA, human CerS2, or siRNA and after 36 h, harvested by trypsinization, collected by centrifugation, washed twice with ice-cold phosphate-buffered saline, and lyophilized. The samples were spiked with an SL internal standard mixture (Avanti Polar Lipids) then extracted and analyzed by LC ESI-MS/MS (10,11,18,19). For tissue analyses, the tis-  Fig. S1). b Primers A and D were used for amplification of the N terminus of the R230 construct and primers B and C for the C terminus. Primers A and F were used for amplification of the N terminus of the R325 construct and primers B and E for the C terminus. In the second step, the two fragments were annealed using an additional touch-up PCR step. Annealing temperatures of 54°C were used for the first 5 cycles, and a temperature of 62°C for the next 30 cycles.
sues were obtained from C57BL/6 mice at 7 weeks of age, homogenized as described above, and aliquots were lyophilized. Samples corresponding to 1 mg of lyophilized tissue were spiked with the SL internal standard mixture, extracted, and analyzed by LC ESI MS/MS. CerS Assay-CerS activity was assayed as described previously (10,11,14) using Hek 293 cell homogenates and 0.25 Ci of [4, H]sphinganine/15 M sphinganine/20 M defatted-bovine serum albumin/50 M fatty acyl-CoA for 20 min at 37°C (17). Different amounts of protein were used for homogenates obtained from cells transfected with different CerS protein in order that the time of the reaction was linear with respect to protein (17), and different acyl-CoAs were used in accordance with the substrate specificity of each CerS (10 -12, 14, 15) (CerS1, 100 g of protein, C18-CoA; CerS2, 150 g of protein, C22-CoA; CerS3, 200 g of protein, C24-CoA; CerS4, 200 g of protein, C20-CoA; CerS5 and CerS6, 50 g of protein, C16-CoA). For inhibition assays, S1P, dissolved in methanol, was added to the reaction mix prior to addition of substrates.
Immunofluorescence-The intracellular localization of human CerS2-HA was performed by confocal laser scanning microscopy as described for CerS4 and -5 (11), using Mito-Tracker Deep Red as a mitochondrial marker and protein disulfide isomerase as an endoplasmic reticulum marker.

CerS Expression in Mouse
Tissues-An early study examining the tissue distribution of CerS mRNA by semi-quantitative reverse transcription-PCR suggested that each CerS has a somewhat unique tissue distribution, with CerS2 (trh3) mRNA the most ubiquitously expressed (11). We have now established robust reaction conditions for real-time quantitative PCR (qPCR), in which the reactions are linear over 8 to 10 orders of magnitude and are Ͼ99% efficient (supplemental Fig. S2). Using these conditions, we analyzed CerS mRNA levels in 14 mouse tissues.
In agreement with the earlier study (11), CerS2 is ubiquitously expressed. However, due to the linearity and sensitivity of qPCR, we now demonstrate that CerS2 mRNA expression levels are significantly higher than those of the other five CerS genes, in some cases as much as an order of magnitude higher (Fig. 1). Highest CerS2 expression (30 -40 molecules RNA/pg of total RNA) was detected in liver and kidney (Fig. 1), with lower levels (ϳ5 molecules RNA/pg of total RNA) in most other tissues. In contrast, CerS1 and -3 were expressed mainly in brain and skeletal muscle, and in skin and testis (15), respectively, and were virtually undetectable in other tissues. CerS4 was expressed at the highest levels in skin, leukocytes, heart, and liver and, in the other tissues, was expressed at levels of 1-2 molecules/pg of total RNA (Fig. 1). CerS5 and -6 were expressed in most tissues, with expression levels of 1-3 molecules/pg of total RNA (Fig. 1). CerS2 expression tended to be low in tissues expressing highest levels of CerS1 or CerS3.
To confirm the acyl-CoA specificity of CerS2, cells were transfected with two different siRNAs in a pSUPER vector, and compared with cells incubated with the pSUPER vector alone. CerS2 mRNA levels were significantly reduced 24 h after transfection with siCerS2ii and 72 h after transfection with siCerS2i (Fig. 3A). CerS activity, using C22:0-acyl-CoA as substrate, revealed a more rapid loss of activity after transfection with FIGURE 1. CerS2 is ubiquitously expressed and is highly abundant in mouse tissues. Tissues from 6-to 8-week-old C57Bl/6 mice, except skin which was obtained from new born pups, were harvested. Total RNA was isolated, and equal amounts were used to prepare cDNA, which was then used to determine gene expression using Taqman TM primer and probes in real-time PCR reactions. Results are means Ϯ S.E. from 3-4 individual tissue samples. In A, CerS expression levels are plotted using the same y-axis scale as for CerS2. In B, each panel is scaled to the expression level of each particular CerS, except for CerS2 and -3, whose expression levels can be clearly seen from A and are therefore not shown. In, intestine; Ln, lymph node; Sp, spleen; Th, thymus; Bm, bone marrow; Sk, skin; Br, brain; Le, leukocytes; He, heart; Li, liver; Sm, skeletal muscle; Pr, prostate; Te, testis; and Ki, kidney.
siCerS2ii compared with siCerS2i. Examination of levels of ceramides by ESI-MS/MS after transfection with siCerS2ii ( Fig.  3C) was consistent with ESI-MS/MS data obtained after overexpression of CerS2 ( Fig. 2A). Thus, CerS2 has a completely different profile of use of acyl-CoAs than the other CerS, using mainly medium-to long-chain CoAs, but is unable to synthesize C16:0-and C18:0-ceramides.
Similar to other CerS for which the intracellular localization has been examined by immunofluorescence after overexpression (10,11), rather than by biochemical isolation of mitochondrial fractions (20), CerS2 is localized to the endoplasmic reticulum (Fig. 4), with no co-localization with a mitochondrial marker (21).
Relationship between CerS mRNA Expression and Ceramide N-Acyl Chain Composition-Although CerS2 mRNA is widely distributed and found at high levels in various tissues (Fig. 1), nothing is known about how this is related to levels of expression of the CerS2 protein or of the relative proportions of ceramides containing C20:0-C26:0-fatty acids, the subspecies synthesized by CerS2 (Figs. 2 and 3). Because there are no antibodies currently available to mouse CerS2, 3 we examined the ceramide subspecies distribution by LC ESI-MS/MS to determine if tissues that display high levels of CerS2 mRNA are 3 A commercial antibody is available for human CerS2 but unfortunately shows no cross-reactivity to mouse CerS2.   enriched in the corresponding ceramide subspecies in ceramides, sphingomyelin (SM), and monohexosylceramide (Hex-Cer) (Fig. 5). Comparison of the ceramide N-acyl chain distribution with that of the relative levels of expression of CerS mRNA (Fig. 5) reveals that the two tissues with highest CerS2 mRNA levels, kidney and liver, also have the highest proportions of C22-to C24-ceramides. Kidney also has high proportions of C22-C24 acyl chains in SM and HexCer (Fig. 5) as does liver, although the N-acyl chain composition of HexCer differs from that of ceramides and SM. For the other three tissues (brain, testis, and skeletal muscle), CerS2 is less prevalent than the mRNAs of the other CerS, and the proportions of C22-, C24-, and C24:1-ceramides and -SMs are correspondingly lower. Interestingly, for two of these tissues (brain and skeletal muscle), HexCer contains surprisingly high proportions of C22-C24-ceramides, suggesting that there are factors other than the relative amounts of the CerS mRNA that affect the subspecies distribution, particularly in downstream complex SLs and glyco-SLs. Inhibition of CerS2 Activity by S1P-A number of studies have proposed that S1P and/or SK may negatively regulate ceramide synthesis, and this interaction may be one of the potential mechanisms by which S1P mediates its pro-survival effects (22)(23)(24); however, there is no evidence showing direct modulation of ceramide synthesis by S1P. We therefore examined whether S1P can affect the activity of CerS proteins in vitro. S1P, at concentrations up to 25 M, had no effect on ceramide synthesis by CerS1, -3, -4, -5, and -6, but surprisingly, inhibited CerS2 activity (Fig. 6A) by a noncompetitive mode of inhibition (Fig.  6, B and C). The non-competitive inhibition suggested that S1P was not directly binding to the active site of CerS2 but rather at a regulatory site.
Using blocks analysis (Block Searcher, version 8/22/03.1) 4 for all CerS protein sequences (with the default parameters of the program) we found that a region of CerS2 has limited homology to two of seven blocks of the S1P receptors (Fig. 7) (25). The combined e-value of the two blocks was 0.86 (Fig. 7). No significant homology to S1P receptors was identified by blocks analysis in any other CerS proteins. Within these regions, two residues were identified as essential for S1P binding to the S1P 1 receptor (26), corresponding to residues Arg-230 and Arg-325 in CerS2 (Fig. 8A); although these residues are found in some other CerS, they were not identified by the blocks algorithm as being part of an S1P receptor block. Mutation of each of these residues alone had no effect on S1P inhibition of CerS2 activity, but mutation of both residues to alanine completely abolished the inhibitory effect of S1P on CerS2 (Fig. 8C), without affecting the basal activity of CerS2 (Fig. 8B). Thus, S1P can regulate CerS2 activity via a direct interaction with an S1P receptor-like motif found uniquely in CerS2.

DISCUSSION
Since the initial molecular identification of CerS1 (uog1) (10) as a ceramide synthase, considerable effort has been invested in 4 Blocks are multiply aligned ungapped segments corresponding to the most highly conserved regions of proteins.  The left-hand panel shows the percent distribution of CerS mRNA in the five tissues analyzed, taken from the data used in Fig. 1. The three right-hand panels show the percent distribution of acyl chains in ceramide, SM, and HexCer for the same tissues. The pie charts are colorcoded according to the specific CerS and the ceramide species that they synthesize, as shown in the legend. For simplicity, only the main ceramide species synthesized by each CerS are shown. Thus, C22, C24, and C24:1 are shown for CerS2 with C20 (which is only a minor species) excluded; the different ceramides made by CerS2 are indicated in the figure. For CerS4, only C20 is shown. Data are from three triplicate analyses from two animals. Levels of ceramides (in pmol/mg of protein) for each tissue are given in supplemental Table S1.
characterizing the six members of this mammalian gene family.
Most studies to date have focused on the acyl-CoA specificity of the proteins (10 -12, 15). However, with the exception of one in vitro ceramide synthase assay reported in Mizutani et al. (12), CerS2 has not been well characterized. We now rectify this situation and report the surprising results that CerS2 mRNA is highly abundant, that CerS2 displays a remarkable fatty acyl-CoA specificity, showing essentially no activity with C16-CoA, very low or no activity with C18-CoA, and little selectivity among longer chain fatty acyl-CoAs, and provocatively, that CerS2 can be regulated by S1P. Together, these results suggest important differences in both the biology of CerS, and in their biochemical modes of regulation. The distinctiveness of CerS2 from other CerS can also be ascertained from study of its genomic organization (Table 2). CerS2 has a compact gene size, a low number of introns, short 5Ј-and 3Ј-UTRs, with a large percentage of surrounding chromosomal sequence containing CpG and Alu elements, and contains a low percentage of LINE-1s. Further, CerS2 is located within chromosomal regions that are replicated early within the cell cycle. 5 These genomic features are characteristic of a "housekeeping" gene; indeed, predictive analysis performed using multiple parameters (27)(28)(29) supports the possibility that the CerS2 gene is a genuine housekeeping gene (Table 2). Interestingly, no other CerS genes display these characteristics.
Classically, a housekeeping gene is defined as a gene with consistent levels of expression from tissue to tissue and that encodes a protein that is generally involved in routine cellular metabolism. However, the advent of qPCR and microarray analysis has indicated that housekeeping genes can display considerable variation in expression levels in tissues and can respond to stimuli (30,31).
The fact that CerS2 fulfils many of the genomic criteria of a housekeeping gene does not necessarily mean that it is a housekeeping ceramide synthase. However, some of the intracellular signaling functions of ceramides have been ascribed to C16ceramide (reviewed in Ref. 9), which, conspicuously, CerS2 does not synthesize. The broad tissue distribution of CerS2 might imply that ceramides synthesized by CerS2 are intermediates in SL metabolism in most cells, whereas other CerS, which are found at lower expression levels and which synthesize a more restricted subset of ceramide acyl chains (i.e. CerS5, which synthesizes C16-ceramide (11), and CerS1, which synthesizes C18-ceramide (10)), may be involved in ceramide syn-5 I. Simon, Hebrew University of Jerusalem, Israel, personal communication.  . Identification of a putative S1P binding site. Blocks analysis identified two S1P receptor-like motifs in CerS2, designated IPB00406C and IPB00406F. Sequences are presented as a logo, which is a representation of the conservation of amino acids at each position; the height of each letter represents the dominance of that particular residue at that position. thesis in specific tissues under specific physiological conditions, or after certain stimuli, such as in apoptosis or cell proliferation (13,32). In this regard, the regulation of CerS2 by S1P, via the S1P receptor-like motif reported in this study, adds to the interest surrounding the precise function of each CerS, particularly CerS2.
In this latter regard, overexpression of SK1 in NIH 3T3 cells leads to a reduction in total cellular ceramide levels (32); because the expression profile of CerS genes is not known in NIH 3T3 cells, it is not possible to conclude from these data that elevated S1P levels directly affect CerS2 activity. However, after overexpression of SK1 in Hek cells, followed by serum starvation and addition of exogenous sphinganine, a significant reduction was seen in levels of ceramides containing all fatty acids examined, including those synthesized by CerS2 (i.e. C20 -C26), but also C16 and C18 (22). CerS2 is found at high levels in Hek cells, 6 which may be partly responsible for the changes in C20-C26-ceramide levels; the reason for the concomitant change in C16-and C18-ceramide levels is unclear, but may be due to regulation of other CerS (i.e. CerS1 and CerS5) by other, currently unknown, mechanisms.
Indeed, our data imply that there are multiple ways to regulate CerS activity and expression, and to regulate levels of ceramides containing specific acyl chains. Thus, although there is a relatively good correlation between CerS2 mRNA expression and levels of C20 -26-ceramides, C20 -26-SM and C20 -26-HexCer in kidney, there is a remarkable difference between the acyl chain composition of SM and HexCer in liver, even though both tissues contain high levels C20 -26-ceramide, and high levels of CerS2 expression. Likewise, levels of C16-ceramide, C16-HexCer, and C16-SM are high in testis, even though expression of CerS3 mRNA is highest in this tissue; CerS3 synthesizes mainly long chain ceramides, but perhaps also C16ceramide (15). C18-ceramide and C18-SM levels are high in brain and skeletal muscle, which contains correspondingly high levels of CerS1 mRNA (33), but HexCer contains mainly longer acyl chain ceramides in these tissues. This clearly demonstrates multiple levels of regulation of ceramide biosynthesis, which most likely also includes the availability of the co-substrate, fatty acyl-CoAs, subspecies-selective trafficking (because CERT, a protein involved in ceramide transport from the endoplasmic reticulum for SM synthesis in the Golgi apparatus (34), has a defined specificity toward different acyl chain ceramides (35)), and factors such as the degree to which ceramide remodeling may occur in at the Golgi apparatus, where ceramidases 6 R. Erez and A. H. Futerman, unpublished observations. FIGURE 8. S1P inhibits CerS2 via an S1P receptor-like block. A, residues that are part of the S1P receptor blocks in CerS2 are shown; mutated residues are underlined. B, activity of CerS2 and CerS2 mutant proteins. Results show a typical experiment, which was repeated four times with similar results. Western blot analysis (not shown) also revealed that there was no change in expression levels. C, effect of S1P on CerS2 activity and on activity of mutated proteins. Results are means Ϯ S.E. for three individual experiments performed in duplicate and are expressed as percent inhibition of basal activity.

TABLE 2 CerS2 has genomic features characteristic of a housekeeping gene
Gene size and length of introns were calculated by examination of each gene using the UCSC genome browser. The presence or absence of Line-1 and Alu elements were analyzed in chromosomal regions 100,000 bp upstream and downstream using the RepeatMasker program that screens DNA for interspersed repeats and low complexity sequences. CpG elements in the same region were analyzed using the CpGreport program, which searches for regions of 50 bp or longer with Ͼ50% GC content. Characterization of Ceramide Synthase 2 FEBRUARY 29, 2008 • VOLUME 283 • NUMBER 9

Gene
JOURNAL OF BIOLOGICAL CHEMISTRY 5683 have been found (36). It is also possible that there is channeling of intermediates into specific ceramides and downstream SLs, in analogy to findings with CerS1 (10), in which C18-ceramide was channeled into neutral glyco-SLs compared with acidic glyco-SLs (gangliosides) and SM. In summary, our results indicate significant and unexpected biological and biochemical differences between the CerS genes and proteins, and imply a unique role for CerS2. The intriguing observation that CerS2 activity can be regulated by S1P paves the way for a series of biochemical studies to attempt to unravel the cross-talk between these two important lipid signaling molecules.