Ceramide Kinase, a Novel Lipid Kinase

Ceramide-1-phosphate is a sphingolipid metabolite that has been implicated in membrane fusion of brain synaptic vesicles and neutrophil phagolysosome formation. Ceramide-1-phosphate can be produced by ATP-dependent ceramide kinase activity, although little is known of this enzyme because it has not yet been highly purified or cloned. Based on sequence homology to sphingosine kinase type 1, we have now cloned a related lipid kinase, human ceramide kinase (hCERK). hCERK encodes a protein of 537 amino acids that has a catalytic region with a high degree of similarity to the diacylglycerol kinase catalytic domain. hCERK also has a putative N-myristoylation site on its NH2 terminus followed by a pleckstrin homology domain. Membrane but not cytosolic fractions from HEK293 cells transiently transfected with a hCERK expression vector readily phosphorylated ceramide but not sphingosine or other sphingoid bases, diacylglycerol or phosphatidylinositol. This activity was clearly distinguished from those of bacterial or human diacylglycerol kinases. With natural ceramide as a substrate, the enzyme had a pH optimum of 6.0–7.5 and showed Michaelis-Menten kinetics, with K m values of 187 and 32 μm for ceramide and ATP, respectively. Northern blot analysis revealed that hCERK mRNA expression was high in the brain, heart, skeletal muscle, kidney, and liver. A BLAST search analysis using the hCERK sequence revealed that putative ceramide kinases (CERKs) exist widely in diverse multicellular organisms including plants, nematodes, insects, and vertebrates. Phylogenetic analysis revealed that CERKs are a new class of lipid kinases that are clearly distinct from sphingosine and diacylglycerol kinases. Cloning of CERK should provide new molecular tools to investigate the physiological functions of ceramide-1-phosphate.

Sphingolipids and their metabolic products, including ceramide, sphingosine, and sphingosine-1-phosphate (S1P), 1 have emerged as important signaling molecules in a variety of biological processes (1)(2)(3)(4). Ceramide has been implicated as one of the components regulating apoptotic responses to stress, particularly those initiated by the mitochondria (1,4,5). Ceramide has been reported as a regulator of several protein kinases and phosphatases, including ceramide-activated protein kinase (6,7), protein phosphatase (8), and protein kinase C (9,10), and may transduce signals through the regulation of these enzymes. Sphingosine, a further metabolite of ceramide, might also play a role in mitochondria-mediated apoptosis (11)(12)(13). Sphingosine inhibits several protein kinases, including protein kinase C (14) and Ca 2ϩ -calmodulin-dependent kinase II (15). S1P, produced by phosphorylation of sphingosine, regulates many biological processes, including mitogenesis, differentiation, migration, and suppression of apoptosis (3, 16 -18). S1P can bind to five members of the G-protein-coupled receptor family known as sphingosine-1-phosphate receptors to elicit diverse biological responses (19,20). Analysis of sphingosine-1-phosphate receptor 1 knockout mice revealed that S1P is essential for normal vascular development (21) and may be implicated in pathophysiological disease states such as angiogenesis, cancer, and inflammation.
Although termination of the second messenger actions of ceramide is considered to be mediated by ceramidase, which hydrolyzes it to form sphingosine, it was suggested more than a decade ago that ceramide can also be phosphorylated to ceramide-1-phosphate (22). Ceramide-1-phosphate was reported as the product of calcium-stimulated lipid kinase in brain synaptic vesicles (22), and it was suggested that the generation of ceramide-1-phosphate may play a role in regulating the secretion of neurotransmitters by increasing the fusibility of the vesicle membranes (22). Elegant studies demonstrate the existence of ceramide kinase activity in HL-60 cells that is functionally and physically separable from that of diacylglycerol kinase (23). In these cells, ceramide-1-phosphate was shown to be specifically derived from ceramide released from sphingomyelin and not from glycosphingolipids (24). Ceramide is also generated during the phagocytosis of antibodycoated erythrocytes by polymorphonuclear leukocytes by activation of sphingomyelinase. Recently, it was shown that ceramide thus formed can be converted to ceramide-1-phosphate through activation of ceramide kinase. Hence, it was suggested that ceramide-1-phosphate may promote phagolysosome formation (25). On the other hand, ceramide-1-phosphate has also been reported to have mitogenic effects. The addition of both natural and synthetic ceramide-1-phosphate to fibroblasts stimulated DNA synthesis (26,27). Whereas it is possible that this effect could result from the conversion of ceramide-1-phosphate to S1P, ceramide-1-phosphate is not a substrate * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AB079066 and AB079067.
Recently, two isoforms of mouse and human sphingosine kinase have been cloned, mouse SPK1 (28), hSPK1 (29), mouse SPK2, and hSPK2 (30). These kinases possess five evolutionarily conserved domains. These domains are likely involved in enzymatic catalysis and/or the determination of substrate specificity. The similarity of the reactions catalyzed by sphingosine and ceramide kinases suggested that ceramide kinases might contain similar conserved domains. Homology searches against a comprehensive nonredundant data base revealed that several similar expressed sequence tags (ESTs) at NCBI had significant homology to the conserved domains of sphingosine kinases yet also had substantial sequence differences. This enabled us to clone another putative lipid kinase. Characterization of this expressed kinase revealed that it has bona fide specific ceramide-phosphorylating activity. This is the first report of the cloning, functional characterization, and tissue distribution of a ceramide kinase (CERK).

EXPERIMENTAL PROCEDURES
Materials-Natural ceramide and diacylglycerol were from Sigma. Other ceramides were from BIOMOL Research Laboratory Inc. (Plymouth Meeting, PA). Sphingosine and S1P were from Matreya Inc. (Pleasant Gap, PA). N,N-Dimethylsphingosine was from Calbiochem. All other lipids were purchased from Avanti Polar Lipids (Birmingham, AL). [␥-32 P]ATP (6000 Ci/mmol) was purchased from Amersham Biosciences. Restriction enzymes were from Takara. Poly(A) ϩ RNA blots of multiple human tissues were purchased from CLONTECH. Lipo-fectAMINE PLUS and LipofectAMINE were from Invitrogen. F-12509A and B-5354c were prepared as described previously (31)(32)(33).
cDNA Cloning of hCERK-BLAST searches of the NCBI EST data base identified two human EST clones (GenBank TM accession numbers AA355581 and H55199) with significant homologies to the conserved domains of sphingosine kinase 1 (28) yet with substantial sequence differences. To obtain a full-length cDNA corresponding to these EST clones, a cDNA library was screened with a labeled PCR fragment prepared with Marathon-Ready TM human leukemia cDNA (CLON-TECH) as a template and LA Taq polymerase (Takara) as follows: 94°C for 2 min; 30 cycles of 94°C for 1 min, 60°C for 1 min, and 72°C for 2 min; and 72°C for 5 min with primers (sense, 5Ј-TCACCACTGACAT-CATCGTTACTGAACATGCTA-3Ј; antisense, 5Ј-CAACGATATGCAGC-GCCGAGGTTTCTGCGTCGCTG-3Ј). The resulting 300-bp PCR product was cloned into pCR2.1, sequenced, and then labeled with [␣-32 P ]dCTP by random priming (Invitrogen) and used to screen a Jurkat cell line cDNA Lamda Zap II library (Stratagene). Approximately 5 ϫ 10 5 bacteriophage plaques were screened under stringent conditions of hybridization, and several positive clones were obtained. The largest clone was sequenced using the BigDye Terminator Cycle sequencing system (Applied Biosystems) and a DNA sequencer (ABI3700).
cDNA Cloning of mCERK-We also identified a mouse EST clone (GenBank TM accession number R74736) with high similarity to hCERK. 5Ј-RACE and 3Ј-RACE (Invitrogen) were performed to obtain a full-length cDNA corresponding to this EST clone. Poly(A) ϩ RNA was isolated from mouse brain using a Quick Prep mRNA purification kit (Amersham Biosciences). First-strand cDNA was synthesized at 50°C for 80 min with 10 g of mouse brain poly(A) ϩ RNA using a target antisense primer designed from the sequence of R74736 (m-GSP1, 5Ј-ACGATGCCATCATAGC-3Ј) and SuperScript II reverse transcriptase (Invitrogen). Two consecutive PCR reactions using this cDNA as a template and LA Taq polymerase (Takara) were carried out as follows: (a) first PCR, 94°C for 2 min; 30 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 4 min; and 72°C for 10 min with 5Ј-RACE Abridged Anchor Primer (5Ј-GGCCACGCGTCGACTAGTACGGGIIGGGIIGGG-IIG-3Ј) and the target-specific antisense primer m-GSP2 (5Ј-TGATCT-CGTATAAAGTCTCCTTGGC-3Ј); and (b) second PCR, the same conditions as described for the first PCR, except that the annealing temperature was 65°C, with Abridged Universal Amplification Primer (5Ј-GGCCACGCGTCGACTAGTAC-3Ј) and m-GSP3 (5Ј-CGTAGTGAT-GGAAGCCAAGGTAAAC-3Ј). PCR products were cloned into pCR2.1 and sequenced.
Overexpression of hCERK and Preparation of Lysates-To construct an expression vector containing the full-length open reading frame, the largest clone was digested with SacII, filled with T4 DNA polymerase, and then subcloned into the SmaI site of pCR3.1 (Invitrogen). HEK293 cells (ATCC CRL-1573) cultured as described previously (28) were seeded at 5 ϫ 10 6 cells/well in poly-L-lysine-coated 100-mm cell culture dishes. After 24 h, cells were transfected with 4 g of vector alone or with vector containing the hCERK construct (pCR.3.1-hCERK), 30 l/ dish LipofectAMINE reagent, and 20 l/dish PLUS reagent according to the manufacturer's protocol. After 3 days, the cells were harvested and lysed by sonication in lysis buffer (20 mM MOPS (pH 7.2), 2 mM EGTA, 1 mM dithiothreitol, complete TM protease inhibitor (Roche Diagnostics), and 10% glycerol). In some experiments, cell lysates were fractionated into cytosol and membrane fractions by centrifugation at 100,000 ϫ g for 60 min at 4°C.
Other Enzyme Assays-Sphingosine kinase and DGK activity was measured exactly as described by Kohama et al. (28) and Bajjalieh and Batchelor (34), respectively, with slight modifications. In brief, the DGK assay contained 50 g of diacylglycerol, 2 mM EDTA, and 5 mM MgCl 2 . Phosphorylated lipids were separated by TLC and quantified with an imaging analyzer (BAS2000; Fuji Film).
Northern Blotting Analysis-Poly(A) ϩ RNA blots containing 2 g poly(A) ϩ RNA/lane from multiple adult human and mouse tissues and mouse embryos were purchased from CLONTECH. Blots were hybridized with the 1.4-kb Aor51HI fragment of pCR3.1-hCERK and the 0.48-kb PvuII fragment of pCR2.1-mCERK-3Ј-RACE after gel purification and labeling with [␣-32 P]dCTP. Hybridization was carried out in ExpressHyb buffer (CLONTECH) at 65°C overnight according to the manufacturer's protocol. Blots were reprobed with ␤-actin (CLON-TECH) as a loading control. Bands were quantified using an imaging analyzer (BAS2000; Fuji Film).
Sequence Alignment and Phylogenetic Analysis-Protein sequences corresponding to known lipid kinases were obtained from the Gen-Bank TM using BLAST searches (35). The sequences of putative sphingosine kinase and CERK homologues were also obtained by BLAST searches against NCBI EST and genomic databases. Some putative kinase sequences were assembled from two or more sequence fragments to form the kinase consensus sequence before alignment. The amino acid sequences of the open reading frames were initially aligned using ClustalW version 1.7 (36). These alignments were then reconciled and further adjusted to minimize insertion/deletion events. Only regions of unambiguous alignments were used in the subsequent phylogenetic analyses. Phylogenetic analyses were performed using the unweighted pair group method with the arithmetic mean of the GENETYX-WIN program,

RESULTS AND DISCUSSION
Cloning of cDNA Encoding CERK-Blast searches of the EST data base identified several ESTs that displayed significant homology to the sphingosine kinase 1 sequences (28 -30). Specific primers were designed from the sequences of these ESTs and used to screen a human Jurkat cell line cDNA lambda ZAP II library. Three positive clones, designated pBK-5, pBK-29, and pBK-33, were isolated; all had identical sequences at overlapping regions (data not shown). Among them, clone pBK-33 was found to contain a complete open reading frame. The complete sequence contains 4459 bp consisting of a 123-bp 5Ј-untranslated region, a 2725-bp 3Ј-noncoding region, and a 1611-bp open reading frame. The deduced 537-amino acid protein (Fig. 1) has the following putative posttranslational modification sites: 4 N-glycosylation sites, 15 phosphorylation sites, 5 prenylation sites, and 2 amidation sites. The gene is located on chromosome 22q13 and consists of 13 exons.
We also cloned the mouse homologue by 5Ј-RACE and 3Ј-RACE using the sequence of EST R74737, which contained a 1593-bp open reading frame. The deduced 531-amino acid protein ( Fig. 1) has 4 N-glycosylation sites, 15 phosphorylation sites, 3 prenylation sites, and 1 amidation site.
Activity of Recombinant Ceramide Kinase-Because of the high similarity to the diacylglycerol kinases catalytic domain, we examined the kinase activity of lysates prepared from transiently transfected HEK293 cells with various sphingolipid substrates. In comparison with human DGK␥, bacterial DGK, and human SPK1, these lysates were enriched in specific ATPdependent ceramide-phosphorylating activity and did not significantly catalyze the phosphorylation of diacylglycerol or sphingosine ( Fig. 2A). As expected, human DGK␥ only phosphorylated diacylglycerol and not ceramide or sphingosine, whereas hSPK1 only phosphorylated sphingosine as reported previously (29). In agreement with previous results (37), bacterial DGK phosphorylated both diacylglycerol and ceramide; however, diacylglycerol was a better substrate than ceramide. These results suggest that this new lipid kinase activity is distinct from other known lipid kinases and is highly specific for ceramide. Hence, we have cloned a bona fide CERK.
Ceramide-phosphorylating activity was previously shown to be membrane-associated in the brain and in leukocytes (22,23,25,34). However, a Kyte-Doolittle hydropathy plot did not indicate the presence of hydrophobic membrane-spanning regions. Thus, it was of interest to examine the subcellular distribution of recombinant CERK. As shown in Fig. 2B, vectortransfected HEK293 cells have barely detectable cytosolic or membrane-associated ATP-dependent ceramide-phosphorylating activity. After transfection with pCR3.1-hCERK, ceramide-phosphorylating activity was markedly increased by Ͼ180-fold in the membrane fraction, whereas the cytosolic activity in the hCERK transfectants was only slightly increased. It is possible that association of hCERK with membranes may be regulated by N-myristoylation of the protein after translation or due to interactions of its PH domain with membrane-bound phosphoinositides (see below).
Characteristics of Recombinant hCERK-Because natural ceramide is a mixture of several molecular species with different acyl chains, it was of interest to analyze acyl chain length substrate specificity of recombinant hCERK in more detail. Among the ceramide analogues examined, C8-ceramide was the best substrate, followed by C16-ceramide, natural ceramide, and C6-ceramide (Fig. 3). A long-chain ceramide such as C16-ceramide was phosphorylated almost equally to that of natural ceramide, but a very short-chain ceramide analogue, C2-ceramide, was a poor substrate compared with natural ceramide. The dihydroceramides were relatively poor substrates compared with the corresponding unsaturated ceramide. However, significant activity was exhibited toward C8dihydroceramide. Importantly, hCERK did not have significant activity toward sphingosine, D,L-threo-dihydrosphingosine, N,N-dimethylsphingosine, or phyto-sphingosine at all, suggesting high specificity for the ceramide structure (Fig. 3).
hCERK was most active in the neutral pH range from pH 6 to pH 8, with optimal activity at pH 6.5, and the activity decreased markedly at pH values below and above this range (Fig. 4A). This pH dependence is quite similar to the ceramidephosphorylating activity of leukocyte and HL-60 cells reported previously (23,25). Because both brain and leukocyte ceramide kinase activities are calcium-dependent (22,25), we next assessed calcium and magnesium dependence of hCERK. hCERK activity was increased by calcium in a concentration-dependent manner at concentrations as low as 10 Ϫ7 M and reached a maximum at about 500 M (Fig. 4B). Magnesium also activated hCERK in a similar dose-dependent manner, although to a lesser extent than calcium, in agreement with previous studies on the brain and leukocyte activities (22,23).
With natural ceramide as substrate, typical Michaelis-Menten kinetics were observed for recombinant hCERK (Fig. 5A). The K m value for ceramide was 187 M, almost identical to the K m value previously found for brain ceramide-phosphorylating activity (22). Surprisingly, the K m for ATP was only 32 M, which is much lower than the value reported for sphingosine kinase (28).
Molecular Structure of CERK-CERKs contain the five conserved domains (C1 to C5) previously identified in the sphingosine kinases (28,38). However, CERKs have additional conserved regions that are not homologous to sphingosine kinase. Moreover, even within the sphingosine kinase homologous regions, CERKs show some variability. The consensus sequence in the sphingosine kinase C1 region, *LVL*NP*GG*GKA, is changed to L*VFINPXGG*GXG in CERK. Other consensus sequences in the conserved regions are changed in CERK as follows: C2, *SGDGL*HEV*NGL**R is changed to VGGDG-XFXE*LXGX*XR; C3, *P*G**PCGSGNALAXSVXH is changed to *R*GIIPAGSTDX*XXXXXX; and C5, X*VGDE is changed to WNXGDE (Fig. 1). It is possible that these changes may be important for substrate recognition as well as regulatory mechanisms specific to CERKs.
A search of signaling domain sequences using the SMART search tool (39) revealed similarity in residues 132-278 of hCERK with the putative DGK catalytic domain, with an E value 0.000472. Residues 8 -126 of hCERK also showed similarity to the PH domain with an E value of 5.36814 (Figs. 1 and  6). The PH domain is thought to have several different functions and to participate in the regulatory roles of the proteins that contain it (40,41). For example, PH domain-containing proteins can bind to the ␤/␥ subunit of heterotrimeric G-proteins (42), to phosphatidylinositol-4,5-bisphosphate (43), and to phosphorylated tyrosine residues (40,44). The PH domain of CERK might not only be important for membrane anchoring but may also regulate its activity.
Interestingly, the deduced sequence of hCERK is more similar to hSPK2 (30) than hSPK1 (Fig. 6). Its sequence diverges considerably from hSPK1 in the center and at the NH 2 terminus. However, after amino acid 126 of hCERK, the sequences of all three kinases share a high degree of similarity. These sequences (amino acids 126 -342 for hCERK, 10 -223 for hSPK1, and 140 -353 for hSPK2) encompass the putative DGK or SPK catalytic domains. The predicted hCERK sequence of this portion shares 30.0% and 30.8% identity or 57.1% and 54.2% similarity with hSPK1 and hSPK2, respectively (Fig. 6). In the COOH-terminal portion of the proteins, there are also homologous regions from amino acids 385-537 for hCERK, 224 -384 for hSPK1, and 470 -618 for hSPK2. However, the similarity of this portion between hCERK and either hSPK is lower than that of SPK1 and SPK2. The deduced hCERK sequence of this portion shares 13.7% and 12.8% identity or 30.7% and 29.5% similarity with hSPK1 and hSPK2, respectively (Fig. 6).
A search of the domain structures of the hCERK sequence revealed a calcium/calmodulin binding motif (45) of the 1-8-14 type B ((F/I/L/V/W) XXXXXX (F/A/I/L/V/W) XXXXX (F/I/L/V/W) with a net charge of ϩ2 to ϩ4) spanning residues 422-435. This calcium/calmodulin binding motif contains a possible cAMP-dependent phosphorylation site ((R/K) XX (S/T)) and is conserved in human, mouse, and Drosophila CERK ( Fig. 1; data not shown). Because the net charge of the calcium/calmodulin binding motif is very important for its function, phosphorylation by cAMP-dependent protein kinase might modulate calcium/calmodulin binding to CERKs and their function. Because CERK is activated by calcium (Fig. 4B), it is possible that calcium/calmodulin may regulate CERK activity. However, the calcium/calmodulin complex and calmodulin antagonists such as trifluoperazine dimaleate or calmidazolium chloride showed no effect on CERK activity (data not shown). Similar results were obtained with crude brain extracts (22).
As shown in Fig. 1, hCERK contains many possible phosphorylation sites. Among them, the casein kinase II phosphorylation site ((S/T) XX (D/E)) at Ser 340 and the cAMP-dependent phosphorylation site at Ser 424 are conserved in mouse and Drosophila CERK. Among the five possible casein kinase II phosphorylation sites, three of them (Thr 179 , Ser 340 , and Thr 448 ) are also present in mice. Among the nine possible protein kinase C phosphorylation sites ((S/T) X (R/K)), six of them (Ser 72 , Thr 118 , Thr 127 , Ser 230 , Ser 300 , and Ser 340 ) are also present in mouse CERK. The putative protein kinase C phosphorylation site at Ser 300 is also found in the C4 domain of SPKs (28) of all multicellular organisms including plants, nematodes, insects, and vertebrates. Because SPK activity is thought to be regulated by protein kinase C, and because both SPK1 expression and CERK expression were increased by 12-O-tetradecanoylphorbol-13-acetate treatment (data not shown), it is feasible that the protein kinase C phosphorylation site at Ser 300 is important for regulating the activities of these kinases.
Tissue Distribution of CERK-Northern blot analysis was used to examine the expression of hCERK in different tissues (Fig. 7A) FIG. 8. Phylogenetic analysis of CERKs and SPKs. The phylogenetic tree was calculated by the unweighted pair group method with arithmetic mean method from the ClustalW multiple alignment of vertebrate and invertebrate kinase proteins. The vertebrate sequences used were SPK1 (human and mouse), SPK2 (human and mouse), and CERK (human and mouse). The invertebrate sequences used were SPK (D. melanogaster, C. elegans, O. sativa, S. cerevisiae (LCB4 and LCB5), and S. pombe) and genomic and EST sequences of putative CERKs from D. melanogaster (GenBank TM accession numbers AAF52040 and BF503060), C. elegans (GenBank TM accession number AAC67466), and O. sativa (GenBank TM accession numbers CAC39069 and AJ307662). Topologically similar trees were obtained with other phylogenetic analysis methods based on protein alignments. Branch lengths are scaled proportional to genetic distance, and the tree is rooted. ever, the level of expression was markedly variable and was highest in the heart, followed by brain, skeletal muscle, kidney, and liver, and was moderately high in peripheral blood leukocytes and thymus (Fig. 7A). High expression of cloned CERK in the brain and leukocytes supports the notion that it may play a role in neurotransmitter release in the brain (22) or in phagocytosis in polymorphonuclear leukocytes (25) and HL-60 cells (23). A similar expression pattern was observed in mouse tissues, except for skeletal muscle (Fig. 7B). During embryonic development, mCERK was expressed at high levels at embryonic day 7 and decreased rapidly thereafter, similar to the developmental pattern of mouse SPK1 expression (30). Its function in the heart is still a matter of speculation. Myocytes are rich in vesicles such as secretory granules, and CERK might regulate their functions.
Phylogenetic Analysis-Sphingosine kinases are widely expressed throughout evolution. Using BLAST searches of the EST and genomic data bases, we found full-length amino acid sequences of putative CERKs in the D. melanogaster, C. elegans, and O. sativa data bases. All of the CERKs are highly homologous at the amino acid sequence level but diverge somewhat at their NH 2 termini. The amino acid sequence identities of hCERK to mCERK, D. melanogaster CERK, C. elegans CERK, and O. sativa CERK are 84.4%, 35.3%, 30.0% and 32.7%, respectively, and similarities are 91.9%, 56.7%, 54.0% and 53.3%, respectively. However, although orthologues of mammalian SPKs in the yeast S. cerevisiae (LCB4 and LCB5) have been identified (46), no putative CERKs were identified in the yeast data base. Phylogenetic analysis (Fig. 8) clearly indicates that CERKs are a distinct class of lipid kinases that are distinguishable from sphingosine kinases. Moreover, CERKs and sphingosine kinases evolutionarily diverged after yeast, and both are found in diverse multicellular organisms including plants, nematodes, insects, and vertebrates.
Concluding Remarks-Because ceramide-1-phosphate is interconvertible with the bioactive sphingolipid metabolites, ceramide, sphingosine, and sphingosine-1-phosphate, it may also be involved in regulation of important physiological processes. Indeed, ceramide-1-phosphate participates in the process of vesicle fusion in neutrophils (25) and synapses (22) and has been implicated in cellular proliferation (26,27). The conversion of ceramide to ceramide-1-phosphate may also serve as a mechanism to terminate the proapoptotic actions of ceramide. In this regard, ceramide-1-phosphate appears to be specifically formed from ceramide derived from sphingomyelin hydrolysis and not from turnover of glycosphingolipids (24). However, progress in understanding the functions of ceramide-1-phosphate has been hampered because the enzymes regulating its metabolism have not been definitively identified or cloned. In this study, we have cloned and characterized the first mammalian CERK. Although CERK is somewhat homologous to SPK, especially in the previously identified conserved domains (28), it is highly specific for ceramide as substrate and does not catalyze phosphorylation of sphingosine or diacylglycerol. Moreover, the properties of recombinant CERK, including cellular localization, calcium dependence, and pH optimum, are very similar to those reported for crude synaptic and neutrophil ceramide-phosphorylating activities (22,23,25,34). Our findings substantiate the notion of another ceramide metabolic pathway: conversion to ceramide-1-phosphate (23). Furthermore, phylogenetic analysis strongly suggests that the CERKs are a new class of lipid kinases and are clearly distinct from SPKs and DGKs. The cloning of CERKs should add a new dimension to the study of sphingolipid function and metabolism.