A Conserved Cysteine Motif Is Critical for Rice Ceramide Kinase Activity and Function

Background Ceramide kinase (CERK) is a key regulator of cell survival in dicotyledonous plants and animals. Much less is known about the roles of CERK and ceramides in mediating cellular processes in monocot plants. Here, we report the characterization of a ceramide kinase, OsCERK, from rice (Oryza sativa spp. Japonica cv. Nipponbare) and investigate the effects of ceramides on rice cell viability. Principal Findings OsCERK can complement the Arabidopsis CERK mutant acd5. Recombinant OsCERK has ceramide kinase activity with Michaelis-Menten kinetics and optimal activity at 7.0 pH and 40°C. Mg2+ activates OsCERK in a concentration-dependent manner. Importantly, a CXXXCXXC motif, conserved in all ceramide kinases and important for the activity of the human enzyme, is critical for OsCERK enzyme activity and in planta function. In a rice protoplast system, inhibition of CERK leads to cell death and the ratio of added ceramide and ceramide-1-phosphate, CERK's substrate and product, respectively, influences cell survival. Ceramide-induced rice cell death has apoptotic features and is an active process that requires both de novo protein synthesis and phosphorylation, respectively. Finally, mitochondria membrane potential loss previously associated with ceramide-induced cell death in Arabidopsis was also found in rice, but it occurred with different timing. Conclusions OsCERK is a bona fide ceramide kinase with a functionally and evolutionarily conserved Cys-rich motif that plays an important role in modulating cell fate in plants. The vital function of the conserved motif in both human and rice CERKs suggests that the biochemical mechanism of CERKs is similar in animals and plants. Furthermore, ceramides induce cell death with similar features in monocot and dicot plants.


Introduction
Sphingolipids are key structural components of membranes and important signal molecules for cell growth, cell death, embryogenesis and development [1,2,3]. Sphingolipid metabolism and functions have been intensively researched in animals and yeast. Sphingolipids and their metabolic intermediates, including ceramide, sphingosine, sphingosine-1-phosphate, and ceramide-1-phosphate (C1P), are significant bioactive molecules in a variety of biological processes [4]. Ceramide, which is the central core lipid in the metabolism of sphingolipids, mediates cell cycle events, differentiation, senescence, necrosis, proliferation and apoptosis [5]. Sphingosine, a metabolite of ceramide, exerts pleiotropic effects on protein kinases and other targets in animals, including regulating the actin cytoskeleton, endocytosis, cell cycle and apoptosis [3], whereas sphingosine-1-phosphate is intimately involved in cell motility, cell survival, cell proliferation and inflammation [1,3]. C1P also has important roles in mediating cell survival, cell proliferation and inflammation [1,3]. Moreover, the dynamic balance between levels of ceramides and C1P is critical for cell survival in the model dicot plant Arabidopsis thaliana and in human cells [1,6]. In Arabidopsis, loss of ceramide kinase (CERK) activity due to mutation of the ACCELERATED CELL DEATH 5 (ACD5) gene results in spontaneous cell death late in development that correlates with the accumulation of ceramides [6].
Ceramide is a critical sphingolipid metabolite involved in the regulation of programmed cell death (PCD) in plants [6,28,29]. Exogenous ceramides, the substrate of CERK, regulate the effects of the fungal toxin AAL in tomato leaves [30]. Arabidopsis acd5 mutant plants show excessive cell death and permit enhanced pathogen growth when infected by Pseudomonas syringae, indicating that ACD5 plays an important role in controlling disease susceptibility [31]. Moreover, short chain ceramides reduce mitochondrial membrane potential and cause subsequent cytochrome c release in Arabidopsis [32].
In monocots, little is known about the roles of ceramides and the enzymes of the sphingolipid pathway. In this paper, we report the cloning of a ceramide kinase in rice and demonstrate that OsCERK is a bona fide ceramide kinase that can completely rescue acd5 mutant phenotypes, including spontaneous cell death, susceptibility to bacterial infection, reduced CERK activity and elevated PR1 expression. We show that a conserved CXXXCXXC motif found in all CERKs is vital to OsCERK function. Finally, we show that ceramide-induced cell death features and antagonism by C1P are similar in rice and Arabidopsis, indicating a conserved role for ceramide and ceramide kinase in monocots and dicots.

Cloning and phylogenetic analysis of OsCERK
Rice harbors a single gene on chromosome 2 (at locus Os02g0656200) predicted to encode a ceramide kinase. The predicted amino acid sequence of this putative rice ceramide kinase (OsCERK) has 48% sequence identity and 64% sequence similarity to the Arabidopsis ceramide kinase called ACCELERATED CELL DEATH 5 (ACD5 [6]) and 27.2% sequence identity and 40.8% sequence similarity to CERK from humans (hCERK, Genbank accession number NP_073603 [8]. The full length OsCERK is made up of 13 exons, 104 bp 59-UTR and 447 bp 39-UTR ( Figure 1A). An ORF within the OsCERK gene encodes a polypeptide of 607 amino acid residues, with a predicted molecular mass of 67.29 KDa and a pI of 6.19.
Database searches indicated that proteins with homology to CERK are ubiquitous in multicellular organisms including vertebrates, invertebrates and plants. OsCERK clusters with the ceramide kinase of Arabidopsis thaliana and possible ceramide kinase orthologues in Populus trichocarpa and Vitis vinifera ( Figure 1B). Moreover, putative CERKs clearly cluster into vertebrate, invertebrate and plant lineages, respectively. CLUSTAL W alignment of the amino acid sequences of OsCERK with putative CERKs from various organisms showed that the predicted OsCERK protein has a conserved CERK-specific motif CXXXCXXC that is critical for the enzymatic activity of hCERK [12] (Figure 1C), and five conserved domains (C1-C5) previously identified in human CERK ( Figure S1). Gaps in the alignment between OsCERK (or ACD5) and hCERK due to sequence divergence may correspond to regions important for substrate recognition and regulatory mechanisms specific to plants and animals, respectively. A search using the SMART online bioinformatics tool revealed a diacyl glycerol kinase (DAGK) domain in residues 139-357 of OsCERK and in 164-366 of ACD5 as reported before [6].
Steady-state levels of OsCERK transcripts were highest in leaves, followed by the stems, with almost no expression in roots when compared with leaves ( Figure 1D). Fresh rice leaf extracts harbored CERK enzyme activity, phosphorylating C6-NBDceramide to give C6-NBD-ceramide-1-phosphate, as did recombinant OsCERK and rACD5, respectively ( Figure 1E, F). Almost no CERK activity was detected in the root extract ( Figure 1E), which is consistent with the low levels of OsCERK transcripts ( Figure 1D).

Biochemical characterization of recombinant OsCERK
Purified His-tagged OsCERK had an optimum reaction temperature of 40uC and an optimum pH value of 7, respectively ( Figure 2A, B). With synthetic C6-NBD-ceramide as substrate, the reaction catalyzed by OsCERK showed classical Michaelis-Menten kinetics ( Figure 2C, D), with an apparent K m of 4.1 mM and an apparent V max of 4.74 nmole 21 min 21 mg 21 protein for D-erythro-C6-ceramide ( Figure 2E). The K m and V max for ATP was 35.36 mM and 2.58 nmole 21 min 21 mg 21 protein, respectively ( Figure 2F). To identify the role of divalent ions for CERK activity, the effect of Ca 2+ and Mg 2+ on OsCERK activity was determined with C6-NBD-ceramide as substrate. In the absence of Mg 2+ , Ca 2+ at ,200 mM produced no change in CERK activity ( Figure 2G), though the activity was increased at 2 mM. In the absence of Ca 2+ , the activity of CERK was increased by Mg 2+ in a concentration-dependent manner ( Figure 2H), which closely resembled the properties of hCERK [8]. Moreover, the addition of Ca 2+ did not boost the reaction rate when 15 mM MgCl 2 was present in the reaction system (data not shown). Thus, unlike ACD5 [6] and hCERK [8], which can be activated by Ca 2+ , OsCERK activity was mainly activated by Mg 2+ , and to a smaller extent by Ca 2+ .

Complementation of ceramide kinase-deficient acd5 mutant by OsCERK
To assess the in planta function of OsCERK, we characterized T3 homozygous lines of the Arabidopsis acd5 mutant harboring a cDNA of OsCERK. Several phenotypes of acd5 were complemented by OsCERK, including spontaneous cell death and elevated PR1 transcript levels late in development (Figure 3 A-D), and enhanced disease susceptibility to virulent Pseudomonas syringae strain DG3 of younger (21 day old) plants that lack cell death ( Figure 3E). Thus, the OsCERK transgene complemented multiple phenotypes of acd5.
Importantly, in plant extracts, CERK activity was partially restored in acd5+OsCERK plants ( Figure 3F, G). We noticed that an extra band appeared with all plant extracts that was more alignment of CERK from various organisms showing the highly conserved CXXXCXXC motif. (D) RT-PCR expression analysis of OsCERK in leaves, stem and roots of 2-week-old seedlings. b-actin was used as a control. (E) CERK activity of rice extracts. Crude plant extracts made from leaves and roots of rice, respectively, were incubated at 40uC for 30 min as described in Materials and methods. The reaction mixture (1 ml) was resolved on a TLC plate. Purified OsCERK recombinant protein was used as the positive control (+CK). For the negative control (2CK) no extract was added. (F) Comparison of recombinant OsCERK and rACD5. The recombinant protein (1.0 mg) was added in 100 ml reaction system, and incubated at 40uC for 30 min. -CK, no recombinant protein was added. Arrows in (E and F) indicated the formed product C6-NBD-ceramide-1-phosphate (C6-NBD-Cer-1P). doi:10.1371/journal.pone.0018079.g001 intense with acd5+OsCERK plant extracts. We speculate that OsCERK overexpression activates other lipid kinases and/or the C6 ceramide substrate was incorporated into larger lipids. Taken together, the acd5 phenotypes were rescued by the OsCERK transgene.
Effects of site-directed mutations in a conserved C-rich motif on OsCERK activity and function To discern the role of the CXXXCXXC motif conserved among all CERK species, we individually mutated each of the three cysteine residues in OsCERK, and assayed CERK activity using purified recombinant enzymes. The three mutant proteins displayed 25% (C461A), 64% (C458A) or 74% (C454A) residual activity, respectively ( Figure 4A, B).
Because yeast lack a clear ceramide kinase homolog [8], it seemed possible this would be a good surrogate system to assess the effect of mutations on OsCERK function. Indeed, FLAG-tagged wild-type OsCERK expressed under galactose control dramatically suppressed yeast growth ( Figure 4C, D), presumably due to a lipid imbalance conferred by OsCERK. In contrast, two mutated versions (C461A and C458A) still accumulated in yeast ( Figure 4C), but had a reduced effect on growth ( Figure 4D). We also attempted to assess the effects of the C454A variant on yeast growth, but the protein failed to accumulate. Based on the effects of mutation of C461 or C458, the conserved CXXXCXXC motif is important for OsCERK activity and impacts its function.
To test the consequences of disrupting the CXXXCXXC motif in planta, we transformed the site-directed OsCERK mutants into acd5. More than 5 independent lines for each site-directed mutation of OsCERK were obtained. At least 50 plants from each of 2 independent lines were used for phenotypic analysis of transgenic acd5 harboring each site-mutated OsCERK. Lines with the same construct showed similar phenotypes; representative lines are shown in Figure 4. As shown in Figure 4E, the OsCERK gene was stably expressed in transgenic acd5 plants; no OsCERK expression was detected in acd5 control plants. Importantly, acd5 plants transformed with OsCERK harboring mutations in the conserved motif still showed some spontaneous cell death and elevated PR1 expression late in development ( Figure 4F, 36 d). Similar to acd5, leaves taken from transgenic plants prior to spontaneous cell death showed no PR1 expression ( Figure 4E line acd5+C458A2).
Interestingly, the onset and extent of cell death in acd5 carrying C461A OsCERK at 36 days and thereafter was similar to that seen with acd5 plants. However, acd5 carrying C454A or C458A OsCERK, respectively, showed less cell death than acd5 at 36 days. Thus, the mutated versions of OsCERK showed varying degrees of defects in complementing the acd5 cell death phenotype. Base on all above data, the highly conserved CXXXCXXC motif and particularly C461 is vital for OsCERK activity and function.

Effects of CERK inhibitors on OsCERK activity and cell death
In order to discern whether OsCERK activity is required for cell viability, we investigated the effect of reducing OsCERK activity with recently commercialized CERK inhibitors, such as N,N-dimethylsphingosine (DMS), which partially inhibits hCERK activity [8], and the ceramide kinase inhibitor K1 (K1), a specific noncompetitive CERK inhibitor [33]. As shown in Figure 5, OsCERK activity was inhibited by K1 and DMS in a concentration dependent manner, and K1 showed stronger inhibition than DMS ( Figure 5A, B, C), which is consistent with the inhibitory effect on hCERK. OsCERK activity decreased by about 70% in 100 mM DMS and 50 mM K1. K1 caused greater than 90% inhibition of OsCERK activity at 200 mM concentration, demonstrating that K1 can almost completely abolish OsCERK activity.
To examine the effect of K1 on cell viability, rice and Arabidopsis protoplasts were treated with K1. In rice protoplasts treated with different concentrations of K1, cell death occurred in a dose-and time-dependent manner ( Figure 5D). In Arabidopsis protoplasts, K1 induced cell death much less in acd5 than in wild-type plants ( Figure 5E). These data are consistent with K1 specifically affecting CERK activity, which is crucial for the maintenance of cell viability.

Characterization of ceramide-induced PCD in rice
To assess whether ceramides, the substrates of OsCERK, can induce cell death in rice, we isolated rice protoplasts and treated them with ceramides. We used short-chain ceramides because of the difficulty in ensuring that longer chain ceramides efficiently enter plant cells. In animal cells, short chain ceramides are converted into longer ceramides in vivo [34]; it seems likely that similar events occur in plants. As shown in Figure S2, both C2and C6-ceramides induced cell death in a time-dependent manner, whereas C2-and C6-dihydroceramide, the biosynthetic precursors of ceramides, had almost no effect at the same concentration as ceramides, similar to what has been reported in animal and Arabidopsis cells [2,6]. C6-ceramide and C2 similarly induced rice protoplast cell death. However, the concentration of C6 that induced cell death was 2 times higher than that of C2 ( Figure S2A, B), suggesting that it is more difficult for C6 than C2 to enter protoplasts. Cell death appeared to be programmed, since 12 h after treatment with C6-ceramide, a clear DNA ladder was seen ( Figure S2D) and nuclear fragmentation characteristic of apoptotic cells was seen by TUNEL staining ( Figure S2C) [32,35].
Cell death induced by C6 ceramide in rice involved new protein synthesis and protein phosphorylation. This was evidenced by the inhibition of each process using cycloheximide or the serine/ threonine kinase inhibitor K252a, respectively, which suppressed cell death to varying extents ( Figure 6A, B). Treatment with the inhibitors alone at the concentrations used did not affect viability of protoplasts. Additionally, as was found in Arabidopsis, the presence of C2-1-phosphate partially blocked cell death induced by C2 ceramide in rice ( Figure 6C), supporting the idea that the balance between ceramides and their phosphorylated derivatives modulates cell death in plants.
As in Arabidopsis, rice protoplasts showed functional changes in mitochondria after ceramide treatment. C6 ceramide, but not C6-DHC, induced a time-dependent loss in mitochondrial membrane    (Figure 7). Protoplasts treated with carbonyl cyanide mchlorophenylhydrazone (CCCP), as a positive control, showed dramatic Dy m loss after 30 min treatment. The results of treatment with C2 were similar to C6 (data not shown). Although Dy m loss was seen in rice and Arabidopsis after ceramide treatment [32], the timing of Dy m loss in rice did not precede cell death onset as was previously observed in Arabidopsis [32]. Thus, there may be differences in the relative importance of Dy m loss in rice and Arabidopsis.

Discussion
We showed here that OsCERK both biochemically and genetically functions as a bona fide ceramide kinase that is important for plant cell survival and that ceramides modulated by OsCERK can induce programmed cell death in rice. Rice ceramide kinase requires a CERK-specific CXXXCXXC motif for its function in planta and its enzymatic activity in vitro. This suggests an evolutionarily conserved and key function for the CXXXCXXC motif in CERK enzymes.

Novel biochemical properties of OsCERK
OsCERK activity is stimulated by Mg 2+ , but only modestly Ca 2+ . This contrasts with the properties of human CERK, which shows Ca 2+ dependence and is stimulated by Mg 2+ in the micromolar concentration range [8]. The differences in the biochemical properties of the human and rice enzymes might be linked to the domains present in hCERK versus OsCERK and/or their respective structures. hCERK contains a calcium/calmodulin binding motif of the 1-8-14 type B [8] that may modulate calcium/calmodulin binding to CERKs and control their function; this motif is absent from the plant enzymes. Stimulation of CERK activity in plants may need a higher concentration calcium than the human enzyme; indeed, ACD5 activity was significantly stimulated by 100 mM Ca 2+ [6]. OsCERK was most active in the neutral pH range from 6.5 to 7.5, with optimal activity at pH 7.0. This pH dependence is quite similar to hCERK and is different from ACD5 (optimal pH 8.2) [6]. The K m with D-erythro-C6-NBD-ceramide was 4.1 mM, much lower than the previous reported K m values for CERK from other organisms. The K m for ATP, 36.35 mM was similar to hCERK. The activity may vary with substrate, substrate delivery [36], reaction system and detection methods of activity. Notably, OsCERK has a temperature optimum of 40uC, higher than ACD5 or hCERK. This high temperature optimum may be related to fact that rice can grow at higher temperatures.

The CXXXCXXC motif is important for OsCERK activity and function
The highly conserved CXXXCXXC motif of CERK, between the C4 and C5 domains, is essential for hCERK enzyme activity and cellular localization [12]. However, the role of the cysteine motif for the full in vivo function(s) of hCERK has not been reported. Our site-directed mutagenesis results indicate that each of the three cysteines contributes to the enzyme activity of OsCERK, with the mutation C461A that removes the third cysteine showing the strongest effect. In planta, each of the three cysteines tested were important, since none of the site-directed mutants could fully rescue of the cell death phenotype of acd5 plants. In the surrogate yeast system, we also confirmed that C458A and C461A did not affect yeast growth the way the wildtype protein did. Thus, this conserved C-rich motif is also important for the in planta function of OsCERK.
In hCERK, the first cysteine in the motif is most important for its enzymatic activity [12], whereas in OsCERK, mutation of the third cysteine had the most dramatic effect on activity and in planta function. The reason for this difference is not clear. Whether the motif also impacts the localization of OsCERK in plants awaits further determination. Interestingly, an identical cysteine motif has been characterized in the radical S-adenosylmethionine super family of enzymes, where the motif provides a metal binding site that is essential for activity [37]. Whether CERK is also regulated by metal coordination will require further research.

Ceramides affect cell viability and Dy m in rice
CERK phosphorylates ceramide to yield C1P, a molecule with interesting signaling properties [38]. Ceramides, as the substrates of CERK, are closely involved in PCD, and affect many cellular events in animals [2]. Ceramide also activates programmed cell death in Arabidopsis [6] and rice. As was observed in Arabidopsis [6], ceramide-induced cell death in rice is also suppressed by the presence of C1P. This suppressive effect could result from direct competition of C1P with ceramide in affecting the same cellular target or through modulation by ceramide phosphate of signaling. We favor the latter hypothesis, since protein phosphorylation, a modification often associated with signal transduction, is important for ceramide-induced cell death in rice and C1P has signaling roles in animals.
In Arabidopsis, a mitochondrial permeability transition, just like in animal apoptosis, is one of the early events in PCD induced by C2ceramide [32]. In rice protoplasts Dy m loss was also observed during PCD induced by short-chain ceramides, however, the membrane potential loss appears simultaneously with PCD, suggesting that in rice Dy m loss may not be essential for PCD induced by ceramide.
In conclusion, a ceramide kinase from rice was cloned and characterized. The ceramide kinase inhibitors, which were characterized on CERK from animals, almost completely blocked OsCERK activity and induced cell death in both rice and Arabidopsis. By site-directed mutagenesis of the conserved CXX XCXXC motif in organisms, three highly conserved cysteines were shown to be vital for CERK activity and function. Since CERK is one of the key enzymes in sphingolipid metabolism, our data offers important information for understanding the conserved sphingolipid functions in animals and plants. It will be interesting to pursue the further role of CERK in plant developmental processes during environmental stress.   dark cycle at 22uC under 60% relative humidity. Pseudomonas syringae pv. maculicola strain DG3 (virulent) was inoculated as described previously [31,39].
For RT-PCR analysis, total RNA samples and Reverse-Transcription were done as described above. The cDNA contents in the different reverse transcription mixtures were normalized by amplifying the transcripts of b-actin using the primer 59-CCATTGGTGCTGAGCGTT-39 and 59-CCCGCAGCTTCC ATTCCTA-39 in rice. Evaluation of the transcript levels of OsCERK in the different tissue was carried out using the normalized cDNA mixtures and the primer 59-TGGCTCAACG-GATGCTATT-39 and 59-ACGACCCTTGGACCAGACC-39. For Arabidopsis expression analysis, the primer 59-GTCACTTGCT GCCTTCGTTT-39 and 59-CTTCTTTCGTGCTCATTTTG C-39 were for tubulin amplification as an internal control, the PR1 expression was amplified using the primer 59-GTGCT CTTGTTCTTCCCTCG-39 and 59-AGCCTTCTCGCTAA CCCACA-39.

Plasmid construction of yeast expression, spot assay and western blots
The wild-type and site-mutated coding regions of OsCERK were introduced into Sal I and Xba I digested pYES2+Flag expression vector (reconstructed from pYES2 vector). In the spot assay, yeast cells grown in SC-glucose medium for 1 day were adjusted to A600 of 0.1. Five microliters of five-fold serial dilutions of each yeast culture was spotted on to SC-glucose or SC-galactose medium and incubated for 2 days (glucose medium) or 3 days (galactose medium) at 30uC. For western blots, yeast cells cultured in liquid galactose medium were collected by centrifugation at 0 h and 15 h, respectively. Cell lysates were prepared using acid-washed glass beads (Sigma-Aldrich, St. Louis, MO, USA). Protein concentration was determined with a Protein assay kit (Bio-Rad) using IgG as the standard. Proteins (10-20 mg) were separated by SDS-PAGE (10%, v/v). Separated proteins were electrophoretically transferred to a nitrocellulose filter membrane (0.45 mm, Bio-Rad). Nonspecific binding was blocked with 5% (w/v) skim milk in TBST buffer (25 mM

Phylogenetic analysis
Based on analysis of sequence alignment described in Methods S1, the phylogenetic tree and executed bootstrap test were calculated and decided by the Neighbor-Joining method from MEGA 4.0 [40].

Expression of OsCERK and ACD5 in Escherichia coli and purification of recombinant protein
The full-length OsCERK coding sequence was subcloned into the E. coli expression vector pET28b (Novagen) and transformed into BL21 (DE3) with the forward primer 59-CGGGATCCGATG GAAGGCGGCGGCGAGGCGCTCT-39 (BamH I site) and reverse primer 59 -TCACTCGAGCTACACCTCCGGTCCT GACGCAAAG-39 (Xho I site). The transformed E.coli cultures were grown in LB medium for large-scale expression. Cells were grown at 37uC with shaking in LB medium. When the turbidity (A600) reached 0.5, IPTG (Merck) was added to 0.5 mM and growth was continued at 16uC overnight, then the cultures were harvested at 4uC. Lysates were analyzed by Western blots using anti-His tag antibody (Tiangen, Beijing, China). His-CERK was purified on Ni-NTA His-Bind resin (Novagen, USA) from 100 ml bacterial cultures, as described in the product manual. After purification, protein concentration was assayed with Easy Protein Quantitative Kit (Trans, Beijing, China). Purified enzyme was stored at 220uC after the addition of sterile glycerol to 50% (V/V). For ACD5, the coding region of acd5 was introduced into pET28b (Novagen) with forward primer 59-CGGGATCCGATGGAG GAAGGTCGTGACGACGAGT-39 (BamH I site) and reverse primer 59 -TCACTCGAGTTATATCTCTGGACCAGATGC GAAC-39 (Xho I site), and the expression system and conditions was same as OsCERK.
OsCERK activity assay and biochemical characterizations CERK activity was measured as previously described with modification [41]. Briefly, mixed micelles containing 1. The reaction mixture (1 ml) was spotted directly onto silica TLC plates (Gel60 F254: Merck, Darmstadt, Germany). The products were separated from the substrate by developing the plate in a solvent system comprising butanol/acetic acid/water (3:1:1, by volume). Fluorescence was detected with Typhoon Trio + scanner (GE Healthcare) in blue fluorescence mode. The C6-NBDceramide-1-phosphate products were identified by comparison with C6-NBD-ceramide (Avanti Polar lipids, Alabaster, AL, USA) and quantified using IQTL software (GE Healthcare). The specific activity was determined using standard curves with known concentrations of D-erythro-C6-NBD-ceramide. To test the effect of pH on the reaction rates, a series of reaction buffers containing 10 mM KCl, 15 mM MgCl 2 and 1 mM dithiothreitol were prepared with 50 mM citrate (pH 4.5, 5.5 or 6.0), 50 mM Hepes-HCl (pH 6.5 or 7.0) or 50 mM Tris-HCl (pH 7.5, 8.0, 8.5 or 9.0). The K m value was determined using the Lineweaver-Burk plot method.
For OsCERK activity of plant extracts, crude plant extracts were prepared as described previously [6]. The protein concentration was determined with Easy Protein Quantitative Kit (Trans, Beijing, China). CERK activity of rice lysates using C6-NBDceramide as a substrate was done with crude lysates using isotopelabeling method with modification [42]. Briefly, 50 ml of leaf lysates (60 to 120 mg of protein) in a total volume of 100 ml was reacted at 40uC for 30 min in the dark and terminated by adding 600 ml chloroform:methonal (1:1) and 265 ml of 1 M KCl in 20 mM MOPS buffer (pH 7.2). The reaction mix was analyzed by TLC developed with H 2 O:acetic acid:methanol:acetone:chloroform (5:10:15:20:50). The TLC plates were autoradiographed for 1-2 h in a phosphor imager cassette and scanned on a STORM 860 (GE Healthcare).

Molecular complementation of acd5 mutant
The full length OsCERK coding sequence was subcloned into pRT104 vector [43] with the forward primer 59-TCACTCGA-GATGGAAGGCGGCGGCGAGGCGCTCT-39 (Xho I site) and the reverse primer 59-CGGGATCCCTACACCTCCGGTCCT GACGCAAAG-39 (BamH I site). The plasmid pRT104:OsCERK was digested with Pst I, and cloned into the same sites of pCAMBIA1305.1 vector (CAMBIA). For OsCERK mutant constructs (C454A, C458A, C461A), the mutated constructs for prokaryotic expression were cut with BamH I and Xho I, and the site-mutated OsERK regions were introduced into BamH I and Sal I digested pCAMBIA13009 vector (the 35S promoter and NOS terminal were introduced into pCAMBIA1300). The obtained wild-type and mutant OsCERK constructs were transformed into Agrobacterium tumefaciens strain EHA105, which was used for transformation of acd5 mutant plants by the floral dip method [44]. The transformed progeny were selected using 1/2 MS plates (solidified with 0.8% tissue culture agar) supplemented with 25 mg/ml hygromycin (Newprobe, Beijing, China). After selfing the T0 plants, lines homozygous for the transgene were chosen for further study.

Visualization of dead cells
Fresh tissue was boiled in lactophenol (10 ml of lactic acid, 10 ml of glycerol, 10 ml of liquid phenol, and 10 ml H 2 O) containing 10 mg trypan blue for 30 S and boiled in 95% ethanol:lacophenol (2:1) for 1 min, and then cleared in 50% ethanol. Photographs were taken with a fluorescence microscope (Axio Imager A1, Carl Zeiss).

Protoplast isolation and treatments
Rice protoplasts were isolated from 10-day-old seedling tissues as described previously with modifications [45]. Detailed method of this experiment is described in Methods S1. Protoplasts were isolated from leaves of 18-day-old Arabidopsis wild-type and acd5 plants as described previously [6,46].
Statistical analyses were performed using Statview statistical package 5.0.1 (SAS Institute, Cary, North Carolina, USA) for Macintosh. Fisher's protected least significant difference (PLSD), a post hoc multiple t-test, was used. Results were considered significant when P,0.05. Data are presented as means 6 SD (standard deviation). All experiments were repeated at least three times.

Measurement of the mitochondrial membrane potential (Dy m ) by flow cytometry
To measure Dy m , the mitochondrial probe 3,3-Dihexyloxacarbocyanine iodide [DiOC 6 (3), Molecular Probes, Eugene Oregon, USA] was used as described by [32]. Briefly, 5610 5 protoplasts were incubated with DiOC 6 (3) (5 nM) for 10 min in culture medium in the dark, at room temperature. As a control for Dy m loss, protoplasts were incubated 100 mM protonophore carbonyl cyanide m-chlorophenylhydrazone (CCCP, Sigma-Aldrich, ST. Louis, USA) for indicated times before flow analysis. The protoplasts were washed once with W5 medium, collected by centrifugation at 800 rpm/min for 3 min. Dyed protoplasts were then stored in the dark for analysis by flow cytometry (BD FACS Calibur) using excitation with a single 488 nm argon laser. At least 30000 events were collected per sample. Data acquisition and analysis were performed using the CellQuest software (BD FACS Calibur, Becton Dickinson) and Win MDI2.9 (Scripps Research Institute, San Diego, CA).

Accession numbers
The GenBank accession number for the nucleotide sequence of OsCERK is FJ765452. Other sequence data of the analyzed CERK proteins in this article can be found in the GenBank data library described in Methods S1. Figure S1 Sequence analysis of rice ceramide kinase (OsCERK). The alignment of amino acid sequences of putative CERKs was executed by using CLUSTAL W. The consensus amino acids are shaded with a 60% identity threshold, amino acids are denoted as colon with an 80% threshold, evolutionarily conserved amino acids are denoted with an asterisk. The conserved domains (C1 to C5, [8]) in SPKs (sphingosine kinase) are indicated by lines. The highly conserved CXXXCXXC motif that is essential for CERK function, is indicated by shaded rectangle. The DAGK catalytic domain depicted by the SMART search tool is indicated by shaded bars. . Letters indicate that values of viability differed in Fisher's PLSD test, a post hoc multiple t test (P,0.05). Control treatment was with 0.15% ethanol (the solvent for C2 or C6). (C) In situ detection of DNA fragmentation by TUNEL after C6-ceramide treatment. Rice protoplasts were treated with 0.15% ethanol and 100 mM C6-ceramide for 12 h and stained using the TUNEL method as described in Materials and Methods. The green signal (arrow) indicates a TUNELpositive nucleus. Protoplasts were stained for nuclei using DAPI. Scale bars represent 10 mM. (D) The DNA ladder induced by 100 mM C6-ceramide after 12 h treatment. (TIF)