Regulation and Traffic of Ceramide 1-Phosphate Produced by Ceramide Kinase

Ceramide 1-phosphate (C1P) has been characterized as a sphingolipid that participates in cell signaling. Although C1P synthesis is thought to occur via phosphorylation of ceramide by ceramide kinase (CerK), the processes that regulate C1P formation and fate remain largely unknown. In this study we analyzed bone marrow-derived macrophages (BMDM) from CerK-null mice (Cerk-/-) and found significant levels of C1P, suggesting that previously unrecognized pathways may also lead to C1P formation. After these experiments we used an overexpression system, BMDM from Cerk-/- mice, and short-chain fluorescent ceramides to trace CerK-dependent formation of C1P. Because the ceramide analogs can also be converted to glucosylceramide (GlcCer) and sphingomyelin (SM), they allowed us to directly compare all three metabolites. We found that C1P produced by CerK is turned over rapidly when serum is removed or upon calcium chelation, whereas GlcCer and SM are stable under these conditions. We further demonstrated that ceramide must be transported to the Golgi complex to be phosphorylated by CerK. Inhibition of the ceramide transfer protein slowed down SM formation without decreasing C1P, suggesting an alternate route of ceramide transport. Other experiments indicated that, like GlcCer and SM, C1P traffics along the secretory pathway to reach the plasma membrane. Furthermore, in BMDM C1P was secreted more readily than was GlcCer or SM. Altogether, our results indicate that CerK is essential to C1P formation via phosphorylation of Cer, providing the first insights into mechanisms underlying ceramide access to CerK and C1P trafficking as well as clarifying C1P as a signaling entity.

Sphingolipids, a class of lipids found in all eukaryotic cell membranes, also function as signaling molecules. De novo synthesis of sphingolipids occurs through a well documented pathway beginning with condensation of serine and palmitoyl-CoA and eventually leading to ceramide (Cer), 3 which is a central intermediate (1,2). Pathways that regulate Cer levels prompt intense scientific interest because of the role of Cer in many cellular mechanisms including apoptosis and differentiation as well as implications for disease (1)(2)(3)(4)(5)(6)(7). Cer metabolism to more complex sphingolipids is important to understand because it impacts Cer levels and can produce bioactive lipids (2)(3)(4)(5). Cer is converted to glucosylceramide (GlcCer) by a glucosylceramide synthase and to sphingomyelin (SM) by sphingomyelin synthases (SMS) (1). Recently, a ceramide kinase (CerK) was identified that can phosphorylate Cer to ceramide 1-phosphate (C1P) (8).
Although C1P was first described about 20 years ago (9,10), regulation of its synthesis and fate has not yet been studied. Synthesis is believed to occur mainly through the phosphorylation of Cer by CerK. C1P is classified as a signaling lipid based on various reports showing that C1P participates in phagocytosis (11), eicosanoid synthesis (12,13), plays a role in mast cell degranulation (14), controls apoptosis in bone marrow-derived macrophages (15), mobilizes calcium in GH4C1 pituitary cells (16) and Jurkat T-cells (17), stimulates dopamine release in PC12 cells (18), and activates cell proliferation and survival (19,20).
In the present work we provide the first evidence that CerK is not the only source of C1P. After making this discovery using cells isolated from Cerk Ϫ/Ϫ 4 mice, we focused on refining our understanding of C1P by differentiating CerK-mediated phosphorylation of Cer from other mechanisms that produce C1P. To do so we used short-chain fluorescent ceramides as precision tools to study only CerK-dependent C1P formation. Importantly, these tools have enabled us to show for the first time that C1P is a short-lived metabolite compared with GlcCer and SM, strongly supporting the hypothesis that C1P is a signaling messenger whose turnover is tightly regulated. Furthermore, because of the unique properties of the fluorescent ceramide analogs used here, we have been able to probe other important aspects of C1P metabolism. We found, for example, that Cer is not phosphorylated at the endoplasmic reticulum (ER) but must be transported to the Golgi complex for phos-* 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. ER, endoplasmic reticulum; NBD-Cer, N-(7-(4-nitrobenzo-2-oxa-l,3-diazole))-6aminocaproyl-D-erythrosphingosine; GlcCer, glucosylceramide; SM, sphingomyelin; SMS, SM synthase; TRB-Cer, N-((4-(4,4-difluoro-5-(2-thienyl)-4bora-3a,4a-diaza-s-indacene-3-yl)phenoxy)acetyl)sphingosine; HPA-12, N-(hydroxy-1-hydroxymethyl-3-phenylpropyl) dodecanamide; CER, Cer transfer protein; DMEM, Dulbecco's modified Eagle's medium; FCS, fetal calf serum; HBSS, Hanks' balanced salt solution; WT, wild type. 4 The generation of Cerk Ϫ/Ϫ will be described elsewhere. phorylation by CerK. Trafficking of Cer to CerK, however, does not appear to rely on the same pathway as that of SM synthesis because pharmacological inhibition of the Cer transfer protein (CERT) did not slow down C1P synthesis. Finally, we observed that C1P readily traffics from the Golgi network to the plasma membrane, as do GlcCer and SM (21).
In short, using a dual gain-of-function/loss-of function strategy (by overexpressing CerK and by working with Cerk Ϫ/Ϫ cells) as well as sensitive and specific fluorescent ceramide analogs, we explain here for the first time how CerK-dependent C1P is formed, regulated, and trafficked compared with GlcCer and SM.
N-(Hydroxy-1-hydroxymethyl-3-phenylpropyl) dodecanamide (HPA-12) and its inactive enantiomer mix were kindly donated by Dr. K. Hanada (National Institute of Infectious Diseases, Tokyo). PIK93 was synthesized at Novartis Institutes for BioMedical Research. All other reagents were from Sigma unless otherwise stated.
Cell Culture and Treatments-COS-1 cells were cultured in DMEM supplemented with 10% FCS at 37°C, 5% CO 2 in a humidified atmosphere. A COS-CerK cell line was generated as previously described (24). For bone marrow-derived macrophage (BMDM) preparation, wild type and Cerk Ϫ/Ϫ Balb/c 4 mice were sacrificed via cervical dislocation. Femurs were extracted and washed twice in phosphate-buffered saline. Ends of the bones were cut, and bone marrow was extruded with RPMI, 10% heat-inactivated FCS and antibiotics (culture medium) using a 25-gauge syringe. Cells were filtered through a 70-m cell strainer, and the filter was washed with culture medium. Cells were spun down at 150 ϫ g for 10 min at room temperature. Cells were re-suspended in culture medium and seeded in 10-cm dishes incubated at 37°C, 5% CO 2 in a humidified atmosphere. After 24 h non-adherent cells were extracted and seeded in multi-well culture plates. Cells differentiated to macrophages with 40 ng/ml recombinant mouse macrophage colony stimulation factor for 7 days (media were changed after 3 days). For treatment with transport inhibitors, cells were grown to confluence in multi-well culture plates before incubation with inhibitors or cell culture medium as a control for 10 min at 37°C. Cells were then incubated with 5 M DMB-C5 Cer at 4°C for 30 min before being warmed to 37°C.
Fluorescence Microscopy-COS-1 cells were seeded into 4-chambered coverslips (LabTEK TM II; Nalge Nunc) at 5 ϫ 10 4 cells per chamber. 24 h later cells were cooled down to 4°C for 15 min, then washed with cold 1% FCS-containing DMEM medium and incubated for 30 min with 5 M fluorescent lipid. Cells were washed again with cold 1% FCS-containing medium and either imaged immediately or further incubated in the presence of ceramide transport inhibitors for 15 min at 4°C before incubation at 25°C for various time points and imaging. For fixation, cell monolayers were incubated for 30 min at 25°C in Hanks' balanced salt solution (HBSS) (without calcium and magnesium) containing 3% paraformaldehyde. Monolayers were then washed 5 times for 5 min in cold HBSS. 5 M fluorescent lipid was applied to fixed cells in FCS-free cold DMEM and incubated on ice for 30 min. Monolayers were finally washed 4 times for 30 min at 25°C in HBSS containing 3.4 mg/ml bovine serum albumin before imaging. Labeling of the endoplasmic reticulum with the ER-Tracker reagent (0.1 g/ml) was done on live cells as described by the manufacturer (Molecular Probes). Live-cell fluorescence microscopy was performed on an inverted microscope Axiovert 200M equipped with a high resolution microscopy camera AxioCam MRc (Zeiss) and an objective Plan-Neofluar 32x/1.30 DIC using the following filter set: excitation 470/emission 525 nm for green fluorescence (NBD-Cer, DMB-Cer, and ER-Tracker) and 546/ 590 nm for red fluorescence (TRB-Cer).
Lipid Extraction and Thin Layer Chromatography (TLC)-Cells were rinsed with HBSS buffer supplemented with 10 mM EDTA. Lipids were then extracted with 200 l of methanol, 200 l of chloroform, and 150 l of HBSS/EDTA. Samples were vortexed and spun in a tabletop centrifuge. The organic phase was then taken and dried using a Thermo SPD SpeedVac. Dried samples (corresponding to total lipid extracts from equivalent number of cells) were dissolved in 10 l of methanol:chloroform 1:1 and analyzed on Silica Gel 60 HPTLC plates (Merck) using butanol:acetic acid:water, 3:1:1, as the mobile phase. TLC plates were dried and imaged using Fuji film LAS3000 intelligent dark box in SYBR Green fluorescent light. Quantitative analysis of TLC plates was performed using the ImageJ software obtained from http://rsb.info.nih.gov/ij/download.html.
Radiolabeling and Two-dimensional TLC Analysis-Cells were seeded to a 6-well plate. On day 7 cells were washed and preincubated for 6 h in phosphate-free medium (MEM Earl) supplemented with 10% FCS (dialyzed against Tris-buffered saline). 15 min before the end of the preincubation period, inhibitors were added. Medium was then replaced by fresh medium containing 200 Ci/ml [ 33 P]orthophosphate (GE healthcare) as well as the indicated concentrations of inhibitor or an equivalent dilution of vehicle (Me 2 SO). After 2 h of incubation, the medium was removed, and cells were washed twice with phosphate-buffered saline, scraped with 0.5 ml of cold methanol, and transferred to Eppendorf vials. After the addition of 0.5 ml of chloroform and 0.45 ml of 0.5 N HCl/2 M NaCl, samples were vortexed, and after a short centrifugation the chloroform phase was analyzed by two-dimensional TLC (Silica 60 HPTLC plates, Merck) with chloroform methanol, 25% NH 4 OH, 60:35:8, for the first dimension and chloroform/methanol/acetone/acetic acid/water, 10:4:3:2:1, for the second dimension. An acetone wash step was included before the second dimension.

Determination of C1P Levels by Liquid Chromatography/ Electrospray Ionization/Tandem Mass Spectroscopy-Cells
(1-2 ϫ 10 6 /sample) were lysed by freeze/thawing in 400 l of phosphate-buffered saline. Two hundred l were transferred into a glass tube, and 300 l of 2 M KCl, 4% HCl as well as 1875 l of methanol/chloroform 2:1 was added. Samples were spiked with an internal standard (C8-C1P, Avanti Polar Lipids). After vortexing and sonication, 675 l of chloroform was, added, and after further vortexing, 500 l of water was added. The sample was finally centrifuged for 10 min at 2600 rpm at 4°C. The organic phase was dried in an autosampler vial. Samples were dissolved in mobile phase and subjected to high performance liquid chromatography (HPLC 1100; Agilent), performed as described in Merrill et al. (25) with some modifications. In short, a Supelco Discovery C 18 HS column (2.1 ϫ 50 mm, 3-m particle size) was used and eluted with a gradient (eluent A, 5 mM ammonium formiate ϩ 1% formic acid in methanol/water 70/30 ϩ 2% tetrahydrofuran; eluent B, 5 mM ammonium formiate ϩ 1% formic acid in methanol ϩ 2% tetrahydrofuran; 70 -100% B in 1.8 min) at a flow of 400 l/min at 60°C. Electrospray ionization with tandem mass spectroscopy using an API 4000 QTrap instrument (MDS Sciex) was employed to detect C1P with positive ionization. The optimal collision energy for C16-C1P was ϩ51 V and for C8-C1P was ϩ39 V; the multiple reaction-monitoring transitions followed were m/z 618.6/264.1 and 506.4/264.2, respectively. This procedure allows for detection of C1P formed by CerK. Alternative protocols have been reported where KOH is used at the extraction step. However, if the reaction is not adequately neutralized afterward, SM is readily hydrolyzed into C1P during the subsequent dry-down step, leading to overestimation of C1P levels (26).

RESULTS
Unique Role for CerK in the Production of C1P from Cer-We selected BMDM for use in this study because we found these primary cells have significant CerK activity. Thus, we were surprised to find that C16-C1P, a major subspecies of C1P, was not abolished in BMDM from Cerk Ϫ/Ϫ mice (Fig. 1A). Labeling of BMDM using [ 33 P]orthophosphate clearly showed that BMDM from Cerk Ϫ/Ϫ mice cannot phosphorylate Cer to produce C1P (Fig. 1B). These results indicate that other sources for C1P formation exist and demonstrate the critical role of CerK in the de novo phosphorylation of Cer.
Current hypotheses for the formation of C1P focus on phosphorylation of Cer by lipid kinases such as CerK. However, the inability of a recently identified CerK-like protein to phosphorylate ceramide (27) has renewed interest in finding alternate pathways of C1P formation; in particular, via acylation of the lysosphingolipid sphingosine 1-phosphate. To investigate this potential pathway, we applied DMB (dimethyl-BODIPY)-labeled sphingosine to WT and Cerk Ϫ/Ϫ BMDM; DMB-sphingosine is a fluorescent sphingosine analog, which has been validated as a substrate for intracellular sphingosine kinases (28). Although we detected DMBsphingosine 1-phosphate, we found no evidence of DMB-C1P under these conditions (Fig. 1C). We also evaluated DMB-dehydrosphingosine as a substrate because dehydrosphingosine is an anabolite intermediate of the de novo sphingolipid synthesis pathway that can also be phosphorylated by sphingosine kinases (29). DMB-dehydrosphingosine was indeed converted to its phosphorylated counterpart; however, we detected no further acylation (Fig. 1C). These results suggest that acylation of sphingosine 1-phosphate (or DH-sphingosine 1-phosphate) is not likely to be a physiological route for the production of C1P (DH-C1P) (Fig. 1D). The rest of our experiments in this present study focused exclusively on C1P that originates from CerK-catalyzed phosphorylation of Cer.
In previous studies we used a short chain [nitrobenzo-2-oxa-1,3-diazole]-labeled ceramide (NBD-Cer) to assess CerK activity in cell-based assays (27). In most cell lines NBD-Cer is readily converted to GlcCer and SM, but C1P is barely detectable unless CerK is overexpressed, as shown in Fig. 2A.
In this study we establish that this assay is entirely selective for CerK because no phosphorylation of NDB-Cer occurs in Cerk Ϫ/Ϫ BMDM ( Fig. 2A).The sensitivity of this TLC-based readout is well in the range of the standard in vitro CerK activity assay (compare Fig. 2, A and B). Moreover, fluorescent ceramides and cell-based assays enabled us to simultaneously analyze, along C1P, the two major Cer metabolites, GlcCer and SM. NBD-Cer is known to be a marker of the trans-Golgi network (30), where the respective enzymes, including CerK (31), can be found. Competition for substrate is illustrated in Fig. 2A; in the presence of CerK, the levels of NBD-GlcCer and NBD-SM diminish, as does the residual cellular pool of NBD-Cer. Because of the specificity, sensi-tivity, and multiplicity of readouts, we have selected this experimental paradigm in our present study.

C1P Produced by CerK Is a Shortlived Sphingolipid Metabolite-If
C1P can act as a signaling molecule, its turnover should be regulated. To analyze serum dependence in C1P formation, we serum-starved BMDM for 2 h and then re-introduced serum in some BMDM, incubating cells with NBD-Cer for 2 h. We found that without serum, little C1P was formed, but when serum was re-introduced to the medium, C1P levels sharply increased (Fig.  3A). The positive correlation with serum was stronger for C1P formation (ϩ328%) than for GlcCer (ϩ103%) or for SM (ϩ17%). Therefore, steady-state levels of C1P are more dependent on serum than are those of GlcCer and SM. When cells were first incubated for 1 h with NBD-Cer in serum-containing medium and then incubated in a serum-free medium, C1P levels decreased 70%, although levels of other metabolites did not change (Fig.  3B). These results indicate that C1P is more labile than GlcCer and SM and reinforce dependence of CerK/C1P on factors present in serum. CerK is activated by calcium ions (8), and a Ca 2ϩ -calmodulin binding domain was recently characterized in the C-terminal part of the protein (32). In vitro, CerK activity is dependent on divalent ions, particularly calcium ions (33), and no CerK activity is found in the presence of saturating amounts of chelators such as EDTA (supplemental Fig. 2). To study the dependence of CerK on calcium ions, we used an NBD-Cer cell-based assay and the intracellular calcium chelator BAPTA (1,2-bis(2aminophenoxy)ethane-N,N,NЈ,NЈ-tetraacetic acid). BAPTA decreased C1P formation almost 100% (Fig. 3C) while decreasing GlcCer and SM synthesis 70 and 30%, respectively (Fig. 3C). When BAPTA was added to a serum-containing medium for 1 h after a prior incubation with NBD-Cer, C1P levels decreased more than 90% compared with controls; levels of the other two metabolites, however, were largely maintained (Fig. 3D). These results from a cell-based assay demonstrate the total dependence of CerK on Ca 2ϩ ions and the remarkably short half-life of C1P compared with GlcCer and SM.
Transport of Cer to the Golgi Complex Allows for Phosphorylation by CerK-NBD-Cer is able to reach the Golgi membrane directly, even in fixed cells (30) (Fig. 4Aa). DMB-Cer (dimethyl BODIPY-Cer), an alternate fluorescent ceramide, accumulates at the ER if cells are fixed (Fig. 4Ab). In live cells, DMB-Cer traffics from the ER to the Golgi complex, relying on a cytosolic transport protein named CERT (34, 35) (Fig. 4A, c and e). Com- pared with NBD-Cer, DMB-Cer is poorly converted to GlcCer (Figs. 4B and 2A), although the reason why is not entirely clear (36). In contrast, SM is readily synthesized using DMB-Cer as substrate, which is delivered to SMS-1 by CERT (35). We then found that DMB-Cer can also serve as a substrate for recombinant CerK in COS-1 cells (Fig. 4B, left) and made similar observations about endogenous CerK in BMDM (Fig. 4B, right). We used a second BODIPY-based fluorescent ceramide, TRB-Cer; this analog accumulates at the ER as DMB-Cer but is barely transported to the Golgi complex (37) (Fig. 4Ad). We found TRB-Cer to be very poorly metabolized into any detectable species, particularly C1P (Fig. 4C). Altogether, these data establish that Cer must be transported to the Golgi complex to be converted into C1P.
Our observations suggest that CerK and SMS-1 might compete for ceramide provisioned by CERT-mediated transport. To test this hypothesis we used HPA-12, a ceramide analog synthesized by Hanada and co-workers (38) that is a direct antagonist of the CERT protein. In Chinese hamster ovary cells, the (1R,3R) HPA-12 enantiomer inhibited the transport of exogenous DMB-Cer and endogenous Cer for conversion to DMB-SM and SM, respectively (38). We observed that HPA-12 also inhibits DMB-Cer transport in COS-cells (Fig. 4Ae). Therefore, we investigated the effect of (1R,3R) HPA-12 treatment on DMB-Cer metabolism in COS-CerK cells and BMDM. In COS-CerK cells, pretreatment with 5 M (1R,3R) HPA-12 decreased the levels of SM synthesized from DMB-Cer 15 min after incubation at 37°C compared with untreated controls (Fig. 5A, top panel), findings consistent with data from Hanada and co-workers (38). Unexpectedly, the levels of C1P did not decrease with drug treatment but in fact increased (Fig. 5A, top  panel). In BMDM, 5 M (1R,3R) HPA-12 also decreased SM levels, whereas it increased DMB-C1P synthesis ϳ100% compared with controls (Fig. 5A, lower panel). In parallel to the active (1R,3R) HPA-12 compound, we tested a 5 M racemic mixture of (1S,3R) (3S,1R) enantiomers as an inactive control (38). This inactive compound had no effect on the amount of C1P and SM synthesized from DMB-C5 Cer in either COS-CerK cells or in BMDM compared with the untreated samples (Fig. 5A, right). These experiments show that pharmacological inhibition of CERT slows down SM formation but does not inhibit the production of C1P.
CERT bears a pleckstrin homology domain, which is known to interact with phosphatidylinositol 4-phosphate (35). Recently, phosphatidylinositol 4-kinase III␤ was shown to play a role in the transport of DMB-Cer to the Golgi complex in COS-cells (37). Inhibition of phosphatidylinositol 4-kinase III␤ by the selective inhibitor PIK93 negatively impacted ceramide transport and, consequently, SM synthesis (37). Therefore, we treated COS-CerK cells and BMDM with PIK93. As expected, metabolism of DMB-Cer to SM was greatly reduced (Fig. 5B). However, there was no concomitant decrease in C1P levels compared with control levels (Fig. 5B and not shown).
Using [ 33 P]orthophosphate labeling, we also checked for endogenous C1P formation in response to treatment with either HPA-12 or PIK93. Consistent with our observations using DMB-Cer, endogenous C1P formation was not decreased after CERT inhibition (Fig. 5C). Endogenous C16-C1P levels  MARCH 28, 2008 • VOLUME 283 • NUMBER 13

JOURNAL OF BIOLOGICAL CHEMISTRY 8521
were further verified by liquid chromatography followed by mass spectrometry and found to be unmodified after treatment with HPA-12 or PIK93 (not shown). Overall, these data show that inhibition of CERT, be it direct (with HPA-12) or indirect (with PIK93), has no effect on C1P synthesis, suggesting that CERT-mediated Cer transport is not critical for CerK to access its substrate.
C1P Traffics to the Plasma Membrane-After NBD-Cer is applied to cells, it is readily metabolized into NBD-GlcCer and NBD-SM (39), and these two metabolites traffic along the secretory pathway to finally reach the plasma membrane. At the plasma membrane, they can be removed by a so-called back-exchange process, whereby the lipid exits the external leaflet of the plasma membrane and moves into an acceptor in the medium, such as proteins (e.g. bovine serum albumin) and lipoproteins (40). This back-exchange is, therefore, an indicator of localization at the plasma membrane. No prior knowledge of this process exists for C1P. In our study, when we treated COS-CerK cells with 5 M NBD-Cer and incubated them at 37°C, we observed NBD-C1P in the medium in a manner consistent with NBD-GlcCer and NBD-SM (Fig. 6, A and B). This observation indicates that NBD-C1P has been transported to the plasma membrane and back-exchanged into the medium. To check for the dependence of this phenomenon on the secretory pathway, we incubated cells with the ionophore monensin, which blocks vesicular trafficking between the medial and the trans-Golgi network (41). We observed that, in the presence of 10 M monensin, all NBD-Cer metabolites including C1P accumulated in cells and decreased their release into the medium (Fig.  6A). NBD-C1P did not accumulate to the same extent as NBD-GlcCer or NBD-SM, which may indicate inhibition of CerK by monensin or result from the short half-life of C1P.
Trafficking along the secretory pathway is blocked at low temperatures (42). Thus, we analyzed the release of the various NBD-labeled metabolites during cell treatment and incubation at 18°C. When COS-CerK cells were incubated with NBD-Cer for 3 h at 18°C, metabolism to NBD-GlcCer, NBD-SM, and NBD-C1P was comparable with that obtained in cells incubated at 37°C. However, at 18°C there was no significant release of any NBD-Cer metabolites into the medium (Fig. 6B), thereby confirming the involvement of the secretory pathway in the trafficking of C1P to the plasma membrane.
Moreover, when WT BMDM were incubated with 5 M NBD-Cer, all detected metabolites were released into the medium (Fig. 6C, left). Remarkably, in these cells NBD-C1P was by far the most readily secreted NBD-Cer metabolite. Ratios of C1P secreted to C1P synthesized reached 40% after 60 min of incubation at 37°C without any lag time. During the same time period and after a slower onset, the ratios for NBD-SM and there is no more perinuclear staining; under similar conditions TRB-Cer remains in the ER as seen by co-staining with the ER-Tracker reagent (d). Kinetics of transport of DMB-Cer at 25°C (e). At t ϭ 0 the labeled ceramide is found at the plasma membrane, and at t ϭ 10 min the dye has reached the cytoplasm and starts appearing at the Golgi complex (red arrows). Enrichment at the Golgi complex increases at t ϭ 20 min and is almost complete at t ϭ 30 min. In the presence of 10 M HPA-12, DMB-Cer accumulation at the Golgi complex is slowed down. The red arrows indicate the diffuse labeling by DMB-Cer and the persistent perinuclear staining in the presence of this compound. B, metabolism of DMB-Cer in COS or COS-CerK cells (left) or BMDM from Cerk Ϫ/Ϫ or WT mice (right) after a 2-h incubation at 37°C with 5 M DMB-Cer; lipids were extracted and the whole extract was analyzed on TLC. GlcCer is poorly produced from DMB-Cer. However, if the DMB label is carried by the sphingosine chain, GlcCer can be made (cf. supplemental Fig. 3). C, TRB-Cer is not readily metabolized in either COS or COS-CerK cells. TLC analysis is after a 3-h incubation at 37°C with 10 M TRB-Cer.

DISCUSSION
CerK, a Major but Not Exclusive Pathway to C1P-Our study confirms that CerK is essential to the formation of C1P through the phosphorylation of Cer. Cells from Cerk Ϫ/Ϫ mice are unable to form C1P, as verified by three assays highly specific for CerK activity: (i) pulse labeling with [ 33 P]orthophosphate (Fig. 1B), (ii) NDB-Cer or DMB-Cer as substrates (Figs. 2A and 4B), and (iii) a standard in vitro kinase assay ( Fig. 2A). Our study also demonstrates for the first time that CerK is not the sole source of C1P, a discovery based on our measurements of significant residual levels of C1P in cells from Cerk Ϫ/Ϫ animals. To investigate other possible mechanisms of C1P synthesis, we used the fluorescent lysolipids sphingosine and sphinganine as precursors for their corresponding lysophospholipids and asked whether the latter could be further acylated to build C1P. Such activity, however, was not observed (Fig. 1C). Therefore, an alternate pathway leading to C1P has yet to be discovered, and it may be interesting to explore whether a type D sphingomyelinase could be implicated (Fig. 1D).
Our work then focused on studying C1P produced by CerK. We used fluorescent ceramide analogs for three important reasons: (i) they can specifically probe C1P made by CerK (see above); (ii) they can be metabolized by CerK, glucosylceramide synthase, and SMS so that the fate of the three metabolites, C1P, GlcCer, and SM can be studied side by side, and (iii) their transport pathways could shed light on the mechanisms of synthesis of ceramide metabolites.
Probing CerK-dependent C1P Turnover and Transport with NBD-Cer-The first analog we used is a short-chain ceramide (C6) with a fluorescent NBD label (NBD-Cer). NBD-Cer is rap- idly taken up by cells and accumulates at the Golgi complex even after cell fixation (30) (Fig. 4Aa), i.e. no physiological transport is required. NBD-Cer is a good substrate for glucosylceramide synthase, SMS (39,40), and CerK ( Fig. 2A); furthermore, it is easily extracted from membranes so that the trafficking of NBD-Cer metabolites can be probed using a so-called "back exchange" procedure (40). In NBD-Cer-and cell-based assays, we have demonstrated competition between CerK, glucosylceramide synthase, and SMS for Cer access at the Golgi complex; C1P formation in COS-CerK and BMDM decreased the levels of the two other fluorescent ceramide metabolites in comparison to COS-1 cells or Cerk Ϫ/Ϫ cells ( Fig. 2A); this confirms that CerK can localize to the Golgi complex as we have shown previously (31). NBD-Cer was also used to identify the short half-life of C1P and to monitor trafficking to the plasma membrane.
C1P is believed to act as a signaling molecule based on the pioneering study by Chalfant and co-workers (12) showing that C1P mediates cytokine signaling in eicosanoid synthesis in A549 lung carcinoma cells. A key characteristic of a signaling intermediate is its transient appearance resulting from on and off signals, which are tightly regulated. We have confirmed this characteristic based on our comparative analyses of C1P, GlcCer, and SM, which are formed when incubated with NBD-Cer under regulated conditions such as reduced serum or calcium chelation. We found that NBD-C1P synthesis was more sensitive to serum compared with NBD-GlcCer and NBD-SM synthesis, as observed either directly (Fig. 3A) or indirectly (Fig.  3B). This suggests that, in the absence of serum, CerK may be less active or that C1P may be degraded more significantly than GlcCer and SM. To test these two possibilities, we used the intracellular calcium chelator BAPTA to inhibit CerK, because calcium ions are required for CerK activity in vitro (8, 33) (supplemental Fig. 2). During treatment with BAPTA, CerK and glucosylceramide synthase were strongly inhibited (reduced to 0% of activity and 30%, respectively, Fig. 3C). We used this paradigm to test the stability of NBD-Cer metabolites and found that NBD-C1P, but not NBD-GlcCer or NBD-SM, is rapidly turned over (Fig. 3D). This finding is the first evidence that C1P is a short-lived metabolite and strongly supports the role of C1P as a signaling molecule. We hypothesize that "on" signals may result from post-translational modifications of CerK (e.g. phosphorylation) that increase CerK activity and C1P formation. "Off" signals may trigger deactivation of CerK and/or dephosphorylation of C1P through the action of broad-specificity lipid phosphatases or a C1P phosphatase (43,44) or through degradation of C1P by other means. All these possibilities warrant further testing because our current understanding of the regulation of CerK/C1P is still limited.
After synthesis, NBD-C1P is trafficked out of the Golgi complex. Inhibition of intra-Golgi transport in COS-CerK cells decreased the amount of NBD-C1P that was available for backexchange from the plasma membrane. This decrease was paralleled by an increase of NBD-C1P inside the cell (Fig. 6A). At 18°C, when vesicular transport is blocked, no significant secretion of NBD-C1P was detected (Fig. 6B). These results demonstrate that C1P, like SM and GlcCer, is trafficked through the Golgi complex "en route to the plasma membrane" (40). In BMDM, the release of NBD-C1P into the medium was even more striking; NBD-C1P was secreted readily and more extensively than NBD-GlcCer or NBD-SM (Fig. 6C).
Probing Cer Access to CerK with TRB-Cer, DMB-Cer, HPA-12, and PIK93-We then used two BODIPY-labeled fluorescent ceramide analogs, TRB-Cer and DMB-Cer, to investigate the mechanisms of ceramide access to CerK. TRB-Cer accumulates at the ER and cannot be efficiently transported to the Golgi complex (37) (Fig. 4Ad). Using TRB-Cer, we have shown that Cer transport from the ER to the Golgi complex is required for C1P synthesis; we have also confirmed this requirement for SM and GlcCer synthesis (Fig. 4C). These observations are consistent with published immunohistochemistry data showing that CerK does not significantly localize to the ER (45). The second fluorescent ceramide, DMB-Cer, has been validated by many studies (36 -38, 46) as a tool to follow the synthesis of SM. The transport of DMB-Cer from the ER to the Golgi complex ( Fig.  4A, b, c, and e) relies, at least in part, on a recently identified ceramide transfer protein named CERT (35). When CERT is inhibited, the accumulation of SM is slowed down (38). GlcCer is poorly produced from DMB-Cer (36), and inhibiting CERT has a marginal impact on GlcCer synthesis (38). We report here that DMB-Cer can readily serve as a substrate for CerK (Fig.  4B), and therefore, we asked whether CERT would be as critical for C1P formation as it is for SM formation. Surprisingly, inhibition of CERT-dependent DMB-Cer transport to the Golgi complex either directly with HPA-12 (Fig. 4Ae) or indirectly with PIK93, a phosphatidylinositol 4-kinase III␤ inhibitor, slowed down synthesis of DMB-SM but did not inhibit DMB-C1P formation (Fig. 5, A and B). In fact, DMB-C1P levels increased during HPA-12 treatment in COS-CerK cells and in BMDM (Fig. 5A). An inactive enantiomer mix of HPA-12 had no effect on either DMB-SM or DMB-C1P formation (Fig. 5A, right). Similar observations were made when monitoring endogenous C1P formation (Fig. 5C), ruling out the possibility that the observed effects of these inhibitors of ceramide transport apply only to the use of artificial substrates. In summary, through a combination of direct and indirect pharmacological inhibition of CERT, it has been shown that the transport of Cer to CerK is not critically dependent on CERT, contrary to SMS-1, which relies on CERT to use Cer as a substrate (35,38). Moreover, the lack of effect of PIK93 on C1P formation also suggests that phosphatidylinositol 4-phosphate is not critical for the localization and function of CerK via its pleckstrin homology domain (47).
While our work was in progress, Lamour et al. (45) reported that CERT is a major provider of Cer for C1P synthesis, findings contrary to our results. Although we cannot account for all discrepancies between the two studies, some important differences may shed some light. First, the method of inhibition and the duration of the studies differ. Lamour et al. (45) used RNA interference to inhibit CERT, and therefore, readouts were taken after prolonged incubation times. In our study, we used short-term pharmacological inhibition of CERT, the effects of which can be analyzed within minutes. Short-term inhibition of CERT may not impact C1P formation, whereas this may occur when CERT is stably down-regulated. The inhibitors we used have been validated for transient use only (37-38) so our study did not address both short-and long-term responses. Nonetheless, working with pharmacological tools to inhibit CERT either directly or indirectly and using a close but inactive analog are clear advantages of our study. Second, because the cell types examined in each study were not the same, we might expect that the subcellular repartition of CerK would differ. In a recent report we found microtubules to be responsible for the rapid translocation of CerK between the Golgi complex and the plasma membrane (47). The differential localization of CerK may, therefore, depend on the cell type and its activation status, as depicted in Fig. 7. . CerK localization and C1P formation. The picture depicts current understanding of Cer metabolism with focus on CerK-catalyzed production of C1P. Cer at the ER can be transported on COP II vesicles or using the CERT protein to reach various locations and enzymes (48, 49) (anabolic pathway). CERT might play a role in providing Cer to CerK (45), but the present study suggests an alternative route(s). The topology of SMS-1 indicates an intraluminal active site, whereas the active site of glucosylceramide synthase (GCS) is on the cytosolic side (50 -52). CerK is myristoylated (31), supporting cytosolic orientation. CerK bears a pleckstrin homology (PH) domain important for localization and activity (8,31) and reported to bind phosphoinositides (47,53). Calmodulin (CaM) acts as a calcium sensor for CerK (32). CerK can traffic between the Golgi complex and the plasma membrane using microtubules (31,47) (dashed line with two arrowheads). At the plasma membrane (PM), CerK may be poised to act downstream sphingomyelinase activity to convert the produced Cer into C1P (catabolic pathway) (54).
In conclusion, our results identify some novel aspects of C1P biology. CerK-mediated phosphorylation of Cer is a major, but not an exclusive pathway to the formation of C1P. A regulation process controls the steady-state levels of C1P generated by CerK, thereby satisfying an essential criterion for signaling function. Moreover, CerK can build C1P at the Golgi complex by relying in part on a Cer transport mechanism that remains to be identified. Finally, C1P readily traffics along the secretory pathway to reach the plasma membrane, a finding consistent with reports that describe C1P in synaptic vesicles (9) and suggest its importance to processes such as phagocytosis (11).