•The bioactive lipid ceramide is produced by the enzyme ceramide synthase, which exists in several isoforms in most eukaryotic organisms. Here, we investigated functional differences between the three ceramide synthase isoforms in Arabidopsis thaliana.
•The biochemical properties of the three ceramide synthases were investigated by comparing lipid profiles of yeast strains expressing LOH1, LOH2 or LOH3 with those of wild-type and loh1, loh2 and loh3 knockout plants. Expression profiles of the ceramide synthases and of the pathogenesis-related gene PR-1 were investigated by real-time PCR.
•Each ceramide synthase isoform showed a characteristic preference regarding acyl-CoA chain length as well as sphingoid base hydroxylation, which matches the pattern of ceramide and glucosylceramide species found in leaves. After extended culture under short-day conditions, loh1 plants showed spontaneous cell death accompanied by enhanced expression of PR-1. The levels of free trihydroxy sphingoid bases as well as ceramide and glucosylceramide species with C16 fatty acid were significantly elevated while species with C20–C28 fatty acids were reduced.
•These data suggest that spontaneous cell death in the loh1 line is triggered either by the accumulation of free trihydroxy sphingoid bases or ceramide species with C16 fatty acid.
Ceramide (Cer) forms the hydrophobic core of sphingolipid molecules and is a ‘bioactive lipid’ in the sense that ‘changes in lipid levels result in functional consequences’ (Hannun & Obeid, 2008). Many research efforts are now focusing on Cer because it is a central intermediate of sphingolipid metabolism at which the pathways of de novo biosynthesis, degradation and recycling are interconnected. Its endogenous levels depend on the activities of many biosynthetic and degradative enzymes such as Cer synthases, ceramidases, Cer kinase, Cer-1-phosphate phosphatase and different phospholipase C analogs, which define a complex metabolic network centered around Cer.
De novo biosynthesis of Cer starts with the production of sphingoid bases by the sequential action of serine palmitoyltransferase and 3-ketosphinganine reductase in the endoplasmic reticulum (ER). Cer synthase then links newly synthesized sphingoid bases and coenzyme A (CoA)-activated fatty acids to form Cer. This de novo pathway plays well-established roles in the heat-induced cell cycle arrest in yeast (Dickson, 2008) and in regulating stress responses and apoptosis in mammalian cells (Hannun & Obeid, 2008). In plants, there is also evidence that de novo biosynthesis contributes to the production of bioactive sphingolipid metabolites which trigger programmed cell death (Liang et al., 2003; Coupe et al., 2004; Townley et al., 2005; Shi et al., 2007; Takahashi et al., 2009).
The phylogenetic tree in Fig. 1 shows that most eukaryotic organisms have several Cer synthase homologs. This phylogenetic diversity of Cer syntheses has, in many cases, been linked to differences in their biochemical and physiological functions. For example, the six mammalian Cer synthases differ in their substrate specificities for acyl-CoAs of different chain lengths which are linked to specialized physiological functions (Sequences 28–32 and 34 in Fig. 1; reviewed in Levy & Futerman, 2010): C22-C24-Cer species produced by CerS2 are found in all tissues (Laviad et al., 2008). Phenotypes of CerS2 knockout mice include myelination defects and increased cell proliferation in the liver (Imgrund et al., 2009; Pewzner-Jung et al., 2010a,b). C16-Cer species produced by CerS5 and CerS6 play complex roles in the regulation of cell proliferation and apoptosis (Schiffmann et al., 2009; Mesicek et al., 2010; Senkal et al., 2010). Cer species containing very long-chain polyunsaturated fatty acids produced by CerS3 have important functions in the process of spermatogenesis and in the skin (reviewed in Sandhoff, 2010).
The two Cer synthases Lag1p and Lac1p from the yeast Saccharomyces cerevisiae (sequences 11 and 12) preferentially make C26-Cer species (Guillas et al., 2001; Schorling et al., 2001). Both share a high (73%) level of amino acid identity and can functionally complement each other since single mutants have no growth defect (Barz & Walter, 1999). Nevertheless, they are not completely redundant as only the transcription of the LAC1 gene is controlled by the pleiotropic drug resistance pathway (Kolaczkowski et al., 2004).
Most fungi have a second Cer synthase homolog that is phylogenetically and biochemically distinct from S. cervisiae Lag1p and Lac1p: Kluyveromyces lactis Lac1p and Pichia pastoris Bar1p (sequences 4 and 5) make C16- and C18-Cer species required for glucosylceramide (GlcCer) biosynthesis (Takakuwa et al., 2008; Ternes et al., 2011). Physiological functions have been investigated in the filamentous fungus Aspergillus nidulans, where deletion of BarA (sequence 3) confers resistance against heat-stable antifungal factor and causes a cell polarity defect (Li et al., 2006).
Can specialized biochemical or physiological functions of individual Cer synthase homologs also be found in plants? In tomato, a dominant allele of the Cer synthase gene asc confers resistance to the inhibitors Alternaria alternata (AAL) toxin and fumonisin B1 (Brandwagt et al., 2000). The observation that the pattern of Cer species in the Asc-1 line changes upon AAL toxin treatment indicates that there might be additional AAL toxin-sensitive Cer synthase isoforms with different substrate specificities (Spassieva et al., 2002). All three Cer synthase homologs from Arabidopsis thaliana (sequences 16, 17 and 23) are localized in the ER (Marion et al., 2008), as are the Cer synthases from mammals and yeast (Barz & Walter, 1999; Venkataraman et al., 2002; Riebeling et al., 2003; Mizutani et al., 2005).
To address this question directly, we investigated the biochemical functions of the three A. thaliana Cer synthases by comparing the molecular species of Cer produced in yeast strains expressing one plant enzyme as their only Cer synthase to the pattern of Cer and GlcCer species found in leaves from wild-type A. thaliana plants and T-DNA insertion lines in which one of the Cer synthases is inactivated. In addition, a first physiological function was identified in one T-DNA insertion line which shows spontaneous cell death after extended cultivation under short-day conditions.
Materials and Methods
Saccharomyces cerevisiae strains expressing LOH1, LOH2, and LOH3
The open reading frames (ORFs) of the three A. thaliana (L.) Heynh. Cer synthases were amplified by PCR and ligated into a Gateway-compatible (Life Technologies, Carlsbad, CA, USA) version of the integrative expression vector YIplac204GPD (Ternes et al., 2011). The cloning strategy was adopted from Marion et al. (2008). See the Supporting Information, Methods S1 for details.
The resulting plasmids were linearized with BstXI and used to transform cells of the S. cerevisiae strain WBY616-LAG1 (RH6602; Kageyama-Yahara & Riezman, 2006). Positive transformants were selected on CSM-Trp− plates and checked by PCR. The complementing plasmid pRS416-LAG1 (Kageyama-Yahara & Riezman, 2006) was counterselected by streaking on CSM-Trp− plates containing 1 mg/ml 5-fluoroorotic acid. The resulting yeast strains WBY616-LOH1, WBY616-LOH2 and WBY616-LOH3 express a FLAG-tagged version of one of the three A. thaliana Cer synthases as their only Cer synthase. Expression of FLAG-tagged LOH1, LOH2 or LOH3 and absence of the complementing c-Myc-tagged S. cerevisiae Lag1p were confirmed by western blotting with anti-FLAG and anti-c-Myc antibodies (Fig. 2a).
Preparation of yeast microsomes
Cells from a 200 ml culture of the S. cerevisiae strain WBY616-LAG1 and from 400 ml cultures of the strains WBY616-LOH1, WBY616-LOH2 and WBY616-LOH3 grown in liquid yeast extract peptone dextrose (YPD) medium were harvested by centrifugation, resuspended in 1 ml of Lysis Buffer (20 mM N-2-hydroxyethylpiperazine-N ′-2-ethanesulfonic acid (HEPES)/KOH, pH 7.4; 25 mM KCl; 2 mM MgCl2; 250 mM sorbitol) containing Proteinase Inhibitor Cocktail (Sigma-Aldrich) and broken by bead-bashing at 4°C for 1 h. Cell debris was removed by centrifuging at 1000 g and 4°C for 10 min. The supernatant was loaded on a 60% (w : w) sucrose cushion and centrifuged at 100 000 g for 1 h. The microsomes were collected from the interphase, snap-frozen in liquid nitrogen and stored at −80°C.
In vitro Cer assay
The in vitro Cer synthase reaction was performed with [3H] sphinganine and C16-(palmitoyl)-CoA as substrates essentially as described in Hirschberg et al. (1993) and Lahiri et al. (2007). The reaction products were separated by thin layer chromatography (TLC) and detected by phosphoimaging. See Methods S1 for details.
Drop dilution test
The S. cerevisiae strains WBY616-LAG1, WBY616-LOH1, WBY616-LOH2 and WBY616-LOH3 were grown in liquid YPD medium until stationary phase. The cultures were diluted with water to OD600 = 1.40 ± 0.02. From this, a series of 10-fold dilutions was prepared. A 5 μl sample of each dilution was applied to a YPD plate and incubated at 30°C for 2 d. The image was acquired using a flatbed scanner (Canon, Tokyo, Japan).
S. cerevisiae strains WBY616-LAG1, WBY616-LOH1, WBY616-LOH2 and WBY616-LOH3 were grown in liquid YPD medium. Growth was stopped by adding trichloroacetic acid to 5% (w : v) final concentration and incubating on ice for 15 min. The cells were harvested by centrifugation, washed twice with ice-cold PBS and stored at −20°C.
A. thaliana lines Col-0, loh1, loh2 and loh3 were grown in soil under short-day conditions (8 h light, 16 h dark). Rosette leaves were harvested at an age of 6 wk, snap-frozen in liquid nitrogen, ground with pestle and mortal, and stored at −80°C until extraction.
For the analysis of Cer and GlcCer, lipids were extracted from c. 0.5 g of yeast cells or 1 g of ground leaf tissue (Folch et al., 1957). The pattern of Cer and GlcCer species was analysed by ultra performance liquid chromatography/electrospray ionization time-of-flight mass spectrometry (UPLC/ESI-TOF-MS) as described in Ternes et al. (2011). For the analysis of free sphingoid bases, lipids were extracted from 0.25 g of ground leaf tissue similar to Method IV from Markham et al. (2006). Free sphingoid bases were separated on an ACQUITY UPLC system coupled to an LCT Premier ESI-TOF-MS analyser (Waters, Milford, MA, USA). Mass spectra in the range 280–580 Da with a mass resolution of > 104 were acquired by ESI-TOF-MS in positive ionization mode with dynamic range enhancement. We have checked that the instrument response is linear in the required range (data not shown). See Methods S1 for details.
Approx. 200 mg (FW) of the same plant material as for the lipid analysis were extracted, similar to the procedure described for lipids in Matyash et al. (2008). Analysis was performed on an Agilent 1100 HPLC system (Agilent, Santa Clara, CA, USA) coupled to an Applied Biosystems 3200 hybrid triple quadrupole/linear ion trap mass spectrometer (Life Technologies, Carlsbad, CA, USA). Phytohormones were identified by multiple reaction monitoring (MRM) in negative ion mode. See Methods S1 for details.
SALK-Lines with T-DNA insertions in LOH1, LOH2 and LOH3
Seeds of the T-DNA insertion lines SALK_069253 (loh1), SALK_018608C (loh2), and SALK_150849 (loh3) were obtained from the Nottingham Arabidopsis Stock Centre (Loughborough, Great Britain). Homozygous plants were identified by PCR analysis of genomic DNA with the primers listed in Table S1. The exact location of the T-DNA left border was determined by direct sequencing of the PCR products.
The presence of transcripts from LOH1, LOH2, and LOH3 in cDNA prepared from leaves of the T-DNA insertion lines was tested by PCR using primers SALK_number-cDNA-LP and SALK_number-cDNA-RP (Table S1). The primers were designed so that the point of T-DNA insertion lies between the primer binding sites and that one primer of each pair bridges two adjacent exons.
Complementation of the loh1 line
The LOH1 ORF was transferred from the GFP-LOH1 expression construct (Marion et al., 2008) first to pDONR221 (Life Technologies, Carlsbad, CA, USA), and then to a Gateway-compatible pCAMBIA3300.1 vector suitable for Basta selection (Hornung et al., 2005). The final construct, which comprised the LOH1 ORF under the control of the ubiquitous 35S promotor, was checked by sequencing.
Homozygous plants of the loh1 line were transformed with this construct by Agrobacterium tumefaciens-mediated transformation using the floral dip method (Clough & Bent, 1998). Seeds of the transformed plants were germinated on soil and subjected to Basta selection. The presence of LOH1 transcripts in cDNA prepared from leaves was detected by semiquantitative PCR with primers SALK_69253-cDNA-LP and SALK_69253-cDNA-RP. Seeds from the plants with the highest transcript levels were collected and used for the complementation experiments.
cDNA was prepared from the 6th-oldest leaf of plants grown in soil under short-day conditions. Transcript levels were quantified by real time PCR with the QuantiTect Primer Assays listed in Table S2 (Qiagen) as described in Mosblech et al. (2008) using ACT2 for normalization.
Trypan blue staining
Plants were grown in soil under short-day conditions. Cells undergoing accelerated cell death were stained with Trypan blue as described in Rate et al. (1999). Images were acquired using an SZX12 binocular equipped with a ColorView II digital camera (Olympus, Tokyo, Japan) using analysis docu software (Soft-Imaging-Systems, Münster, Germany). Pictures of whole plants were taken with a PowerShot G7 digital camera (Canon, Tokyo, Japan).
Results and Discussion
Phylogenetic analysis of plant Cer synthases
A. thaliana has three Cer synthase homologs that are part of a plant-specific branch in Fig. 1. LOH1 (At3g25540) and LOH3 (At1g13580) are highly similar (77% identity) and must originate from a recent gene duplication because the respective genes in other plant species have diverged independently. The genetic diversity within this branch is evident in both dicotyledonous (A. thaliana, Populus trichocarpa, Vitis vinifera) and monocotyledonous plants (Oryza sativa) (Fig. S1). LOH2 (At3g19260) is less similar (≈ 45% identity) to LOH1 and LOH3. It has diverged from the last two very early in plant evolution because it forms a separate phylogenetic branch already evident in the moss Physcomitrella patens (Rensing et al., 2008).
Because LOH2 belongs to a highly conserved branch distinct from LOH1 and LOH3, functional differences can be expected between LOH1 and LOH3 on one side and LOH2 on the other. LOH1 and LOH3 themselves may be redundant because of their high sequence identity and their recent evolutionary origin. Conversely, the high genetic diversity in the branch leading to LOH1 and LOH3 may point to an evolutionary pressure favoring the presence of multiple Cer synthase isoforms in each organism.
LOH1, LOH2 and LOH3 are Cer synthases
To determine whether the Cer synthase homologs LOH1, LOH2 and LOH3 are indeed active as Cer synthases, FLAG-tagged versions of these proteins were expressed under the control of the constitutive GPD1 promotor in an S. cerevisiae lag1Δlac1Δ strain lacking endogenous Cer synthase activity. A control strain expresses c-Myc-tagged S. cerevisiae Lag1p under the control of its endogenous promotor (Kageyama-Yahara & Riezman, 2006). Expression was confirmed by western blotting of microsomal preparations (Fig. 2a).
Cer synthase activity could be detected in vitro in microsomes prepared from the strains expressing LOH1, LOH2 or LOH3 but not Lag1p. [4,5-3H] sphinganine and unlabeled C16-(palmitoyl)-CoA were used as substrates (Fig. 2a). Cer synthase activity was undetectable in the control strain because Lag1p does not accept palmitoyl-CoA but instead prefers very long-chain acyl-CoAs (Guillas et al., 2001; Cerantola et al., 2007).
In addition to showing Cer synthase activity in vitro, LOH1, LOH2 and LOH3 complemented the growth defect that was observed in the uncomplemented lag1Δlac1Δ strain (Barz & Walter, 1999) (Fig. 2b). Together, these results show that A. thaliana LOH1, LOH2 and LOH3 are biochemically active as Cer synthases in vitro and can fulfill the physiological requirements for Cer biosynthesis when expressed in yeast.
LOH1, LOH2 and LOH3 produce Cer species with distinct chain length distributions and hydroxylation patterns
To investigate whether the three A. thaliana isoforms prefer different acyl-CoA chain lengths, Cer species produced by the S. cerevisiae strains expressing LOH1, LOH2, or LOH3 and by the control strain were analysed by UPLC/MS-TOF. Molecular species of Cer were identified based on their exact masses and chromatographic retention times and quantified relative to an internal standard. The full list of Cer species that were investigated is shown in Table S3.
The chain length distributions of the Cer species produced in these yeast strains are shown in Fig. 3. Because fragmentation of the Cer molecule is not possible with the MS method used for the analysis, only the sum of carbons in the sphingoid base and the fatty acid can be determined directly. As the yeast strains used produce sphingolipids with both C18 and C20 sphingoid bases (Fig. S2), the chain length of the fatty acid is estimated by subtracting 18 or 20 from the total number of carbons in the following text and in Figs 3, 4.
The control strain produced almost exclusively (> 90%) Cer species containing C24–C28 fatty acids. This matches previously published data on the preferred acyl-CoA chain length of S. cerevisiae Lag1p (Guillas et al., 2001; Schorling et al., 2001). The yeast strains expressing LOH1 or LOH3 produce Cer species with a very broad chain length distribution, indicating that they can accept acyl-CoAs with chain lengths ranging from 12 to 28. The small maxima seen at the chain lengths c. 18 and 26 correspond to the dominant acyl-CoA species found in yeast (M. Scharnewski & M. Fulda, unpublished). By contrast, most Cer species (> 80%) produced in the yeast strain expressing LOH2 have a C14–C18 fatty acid, with 16 most likely being the predominant chain length.
The sphingoid base in yeast sphingolipids can be hydroxylated at the C4-position, resulting in a trihydroxy sphingoid base, while the fatty acid can be hydroxylated at the α-carbon (reviewed in Lester & Dickson, 1993). More than 90% of the Cer species produced in the strains expressing LOH1 or LOH3 or in the control strain contain trihydroxy sphingoid bases regardless of their acyl chain length (Fig. 4a). By contrast, in the strain expressing LOH2 only approx. one-third of the Cer species with C12–C20 fatty acids contain trihydroxy sphingoid bases. These account for the majority of Cer species in this strain (Fig. 3) and therefore most likely reflect the substrate preference of LOH2. Interestingly, the Cer species with very long-chain fatty acids are > 99% hydroxylated. These species only represent a minor fraction, suggesting that they are produced independently of LOH2, for example, by a reverse ceramidase reaction. No significant differences were observed between the strains expressing LOH1, LOH2 and LOH3 regarding α-hydroxylation of the fatty acid (Fig. 4b).
In conclusion, LOH2 differs from LOH1 and LOH3 in two respects. First, LOH2 seems to prefer acyl-CoAs with chain lengths c. 16 carbons, while LOH1 and LOH3 accept a wide range of acyl-CoAs. In this respect, LOH2 resembles mammalian CerS5 (Cerantola et al., 2007). Second, LOH2 accepts both dihydroxy and trihydroxy sphingoid bases while LOH1 and LOH3 seem to prefer trihydroxy sphingoid bases. While in the current literature, the substrate specificities of Cer synthases are usually discussed with respect to the preferred acyl-CoA chain length, these results suggest that structural features of the sphingoid base such as hydroxylation or desaturation also have to be taken into account.
The biochemical functions of the three A. thaliana Cer synthases match the phylogenetic relationships between them (Figs. 1, S1). LOH1 and LOH3, which are closely related, have very similar biochemical properties and are likely to be redundant in this respect. LOH2 differs from LOH1 and LOH3 in its biochemical properties and is more distantly related. Interestingly, Cer synthases with similar biochemical properties to those of LOH2 are found in both animals and fungi, but these enzymes are in different branches of the phylogenetic tree shown in Fig. 1 (e.g. Bar1p from fungi, sequences 1–5, and mammalian CerS5, sequence 28).
Characterization of the T-DNA insertion lines loh1, loh2 and loh3
To test whether specialized physiological functions underlie the genetic diversification observed in the evolutionary branch leading to LOH1 and LOH3 (Figs 1, S1), we obtained homozygous plants from the T-DNA insertion lines loh1 (SALK_69253), loh2 (SALK_18608C), and loh3 (SALK_150849). When the study was initiated, these were the only SALK lines available where the insertion was annotated to be in an exon and which were tested positive by PCR screening of seedlings. The exact point of insertion was determined by sequencing of the T-DNA left border–genome junction (Fig. S3a). To test whether the T-DNA insertion lines are complete knockouts, the presence of transcripts from LOH1, LOH2 and LOH3 was checked for by PCR analysis of cDNA prepared from leaves. No transcript of the gene into which the T-DNA is inserted was detectable in leaves of any of the three insertion lines while the other two genes were still expressed (Fig. S3b). In conclusion, all three T-DNA lines can be regarded as complete knockouts. Of the three possible double knockouts, loh1loh3 is embryonically lethal while the other two show no obvious phenotype under standard conditions (Markham et al., 2011).
The loh1 line shows spontaneous cell death at advanced age
After 8–10 wk in short-day conditions (8 h light, 16 h dark), plants of the loh1 line suddenly (within a few days) developed yellow lesions, starting with the oldest leaves (Fig. 5a). Formation of lesions started at the margins of the leaves and proceeded towards the center. Before this point, loh1 plants were phenotypically indistinguishable from Col-0 or the other T-DNA insertion lines. Simultaneously with the first lesions, Trypan blue staining showed the spontaneous appearance of dead cells (Fig. 5b). These appeared as single cells or small cell clusters spread over the entire leaf surface. While areas of highest density of Trypan blue-positive cells corresponded to macroscopic lesion areas, many Trypan blue-positive cells were also detected in areas of green leaf tissue. This phenotype, which was only apparent under short-day conditions, strongly resembles the so-called initiation class of lesion-mimic mutants (Lorrain et al., 2003).
Complementation of the loh1 knockout by expression of a P35S–LOH1 construct drastically reduced the severity of the phenotype: whereas in the loh1 line, nearly all plants showed strong spontaneous cell death (yellow lesions covering a major part of most older leaves and leaf size visibly reduced), in the complemented line only 50% of the plants still showed a strong phenotype, 29% showed a weak phenotype (yellow lesions covering only a minor part of a few leaves and leaf size not reduced) and 21% showed no phenotype at all. Trypan blue-positive cells could still be detected in complemented plants showing no phenotype, but their density was very low (Fig. 5). It is important to note that the complemented line is a segregating T1 generation, therefore 25% of the plants would be expected to have a strong phenotype anyway.
In conclusion, the extent of spontaneous cell death was strongly reduced in the loh1 line complemented with the P35S-LOH1 construct. The phenotype is thus very likely caused by the inactivation of the LOH1 gene. The fact that the phenotype is not completely reversed could be caused by differences in the expression level or the spatial/temporal expression pattern between the 35S promotor and the endogenous promotor.
Time-course of LOH1, LOH2 and LOH3 transcript levels
LOH1 and LOH3 share a high level (77%) of amino acid identity, have a close phylogenetic relationship (Figs 1, S1), rescue the lag1Δlac1Δ yeast strain to the same extent (Fig. 2b), produce Cer species with a very similar chain length distribution (Fig. 3) and prefer trihydroxy sphingoid bases (Fig. 4a). Previously, LOH1 and LOH3 showed very similar staining patterns when their subcellular localization was investigated in A. thaliana seedlings (Marion et al., 2008). In order to find out why only the loh1 line shows spontaneous cell death, we investigated whether there were any differences in expression level or in the time-course of gene expression between LOH1 and LOH3.
To this end, the transcript levels of LOH1, LOH2 and LOH3 in leaves were followed by real-time PCR over a time-course from 4½- to 8½-wk-old plants grown in short-day conditions (Fig. 6a). While the transcript levels of LOH1, LOH2 and LOH3 remained roughly constant over time, LOH1 was expressed, on average, threefold more than LOH3 (dotted line in Fig. 6a). LOH2 showed an expression level intermediate between LOH1 and LOH3. The expression level of LOH1 was unchanged in the loh2 and loh3 lines compared with Col-0 and analogous observations were made for LOH2 and LOH3 (data not shown). genevestigator (https://www.genevestigator.com) confirms that while all three ceramide synthases are expressed ubiquitously throughout A. thaliana organs, expression of LOH1 is about threefold higher than that of LOH3. The fact that spontaneous cell death is only observed in the loh1 line might thus be explained by the higher expression level of LOH1 compared with LOH3.
Expression of the pathogenesis-related gene PR-1 is enhanced in the loh1 line but phytohormone levels are unchanged
To distinguish whether the spontaneous cell death in the loh1 line is a sign of premature senescence or instead represents a hypersensitive response-like reaction, the expression levels of the pathogenesis-related gene PR-1 and the senescence-associated gene SAG12 were followed over the same time-course as the LOH1, LOH2 and LOH3 expression described earlier (Fig. 6b). PR-1 is induced in both Col-0 and in the T-DNA insertion lines with increasing age (data not shown). This induction occurs earlier and stronger in the loh1 line than in the other lines. The maximum difference was observed in 7½ wk-old plants, where the expression of PR-1 was 160-fold stronger in the loh1 line than in Col-0 (Fig. 6b; logarithmic scale). SAG12 was not induced in any of the plant lines under these experimental conditions until the end of the time-course.
Interestingly, the levels of the phytohormones salicylic acid (SA), abscisic acid (ABA), jasmonates (JA) and indole-3-acetic acid (IAA) were essentially unchanged in the T-DNA insertion lines compared with Col-0 (Fig. S4). Gibberellin and cytokinin levels were below the detection limit. At the time-point when phytohormone levels were investigated (6 wk), amounts of PR-1 transcript were just beginning to increase (data not shown). Sphingolipid analysis, which was done with the same plant material as phytohormone analysis, already showed significant differences between loh1 and the other lines at this point (see Fig. 7 and the section ‘The balance between the two groups of Cer and GlcCer species is shifted in the T-DNA insertion lines’).
In conclusion, a strong induction of PR-1 with no concomitant increase in the expressing of SAG12 strongly suggests a hypersensitive response-like reaction. The spontaneous cell death in combination with the induction of pathogenesis-related genes strongly resembles the phenotype of several other mutants with perturbed sphingolipid metabolism (Cer kinase acd5, sphingolipid transfer protein acd11, sphingoid base hydroxylases sbh1 and sbh2, inositol phosphorylceramide (IPC) synthase erh1 and inositol phosphate synthase mips1) (Greenberg et al., 2000; Brodersen et al., 2002; Liang et al., 2003; Chen et al., 2008; Wang et al., 2008; Donahue et al., 2010). Together, our data and the published evidence point to the possibility that changes in the level of certain sphingolipid species can trigger pathogen response pathways, leading to the induction of cell death.
A. thaliana leaves contain two groups of Cer and GlcCer species which are distinguished by their acyl chain lengths
To determine whether changes in the level of certain sphingolipid species are responsible for the spontaneous induction of cell death in the loh1 line, we analysed the Cer and GlcCer composition of Col-0 and the T-DNA insertion lines by UPLC/MS-TOF. Unfortunately, this method is not suitable for analysing the major sphingolipid class in plants, (G)IPCs. Molecular species of Cer and GlcCer from leaves of 6-wk-old plants were identified based on their exact masses and retention times, and quantified relative to internal standards. The full list of Cer and GlcCer species which were investigated is shown in Table S3. Although the MS method used only allows the determination of the total number of backbone carbons, the acyl chain length of the Cer and GlcCer species can be easily inferred because A. thaliana contains only C18 sphingoid bases (Fig. S2; Imai et al., 2000; Sperling et al., 2005; Markham et al., 2006).
Fig. 7 shows the molecular species of Cer and GlcCer according to their acyl chain length. Two groups of Cer and GlcCer species can be distinguished: one group exclusively contains C16-Cer and GlcCer species, while the other has a broader chain length distribution ranging from C20 to C28 with a maximum at C24. The overall chain length distribution is very similar for Cer and GlcCer with the only difference being that the proportion of species containing a C16 fatty acid is significantly higher in GlcCer (7% ± 2% in Cer vs 19% ± 3% in GlcCer).
In conclusion, A. thaliana contains two groups of Cer and GlcCer species with a distinct acyl chain length distribution. The chain length of the C16-Cer and GlcCer species matches the Cer species produced by the S. cerevisiae strain expressing LOH2 (Fig. 3), suggesting that LOH2 is the enzyme responsible for the biosynthesis of Cer species with this chain length. Conversely, LOH1 and LOH3 seem to produce Cer species with a very broad chain length distribution when expressed in yeast, but it may be inferred from the data shown below that they are responsible for the biosynthesis of the group of C20–C28-Cer species in planta. This matches the conclusions drawn by Markham et al. (2011).
The acyl chain length distribution of the Cer species produced by LOH1 and LOH3 in A. thaliana leaves therefore seems to be more restricted than in the yeast strains expressing the respective enzymes. It is possible that additional factors restrict the access of LOH1 and LOH3 to acyl-CoAs with chain lengths ranging from C20 to C28, with C24 being the predominant chain length used in planta.
The two groups of Cer and GlcCer species are distinguished by the hydroxylation patterns of their sphingoid base and fatty acid
As in yeast, the sphingoid base in plant sphingolipids can be hydroxylated at the C4 position and the fatty acid can be hydroxylated at the α-carbon (reviewed in Lester & Dickson, 1993). The proportion of Cer and GlcCer species with trihydroxy sphingoid bases and/or α-hydroxylated fatty acids in Col-0 as well as in the T-DNA insertion lines is shown in Fig. 8. While species hydroxylated either at the sphingoid base or the fatty acid have the same molecular mass, they can be readily distinguished by their retention times.
C4-hydroxylation of the sphingoid base is an important feature distinguishing the two groups of Cer and GlcCer species with different acyl chain lengths: while C16-Cer and GlcCer species can have dihydroxy as well as trihydroxy sphingoid bases, C20–C28 species almost exclusively contain trihydroxy bases (Fig. 8a). Thus, C4-hydroxylation seems to be optional in C16 species, but obligatory in C20–C28 species. In support of this, simultaneous inactivation of the two sphingoid base hydroxylase genes SBH1 (At1g69640) and SBH2 (At1g14290) results in severely decreased proportions of C20–C28-Cer species (Chen et al., 2008). Plants of this double knockout line are severely dwarfed and their development does not progress from the vegetative to the reproductive phase.
The pattern of C4-hydroxylation found in leaves of Col-0 plants matches that found in the S. cerevisiae strains expressing LOH1, LOH2 or LOH3 (Fig. 4a). These similarities suggest that each of the three Cer synthases can be assigned to one of the two groups of Cer species found in leaves (see Fig. 9): the Cer species formed by the yeast strains expressing LOH1 or LOH3 contain exclusively trihydroxy sphingoid bases, as do the C20–C28-Cer species found in leaves. The Cer species in the yeast strain expressing LOH2 contain both dihydroxy and trihydroxy sphingoid bases linked to C14–C18 fatty acids. In planta, this corresponds to the group of Cer species with a dihydroxy or trihydroxy sphingoid base and a C16 fatty acid.
In contrast to sphingoid base C4-hydroxylation, the proportion of α-hydroxylated fatty acids in both Cer and GlcCer is largely independent of their acyl chain length (Fig. 8b). Instead, α-hydroxylation is a distinguishing feature between Cer and GlcCer: While Cer species of any chain length either contain a nonhydroxylated or an α-hydroxylated fatty acid, the fatty acid in GlcCer is always α-hydroxylated. This matches results obtained in the yeast P. pastoris (Ternes et al., 2011).
In addition to fatty acid α-hydroxylation, Cer and GlcCer are also distinguished by their level of desaturation (Notes S1 and Fig. S5). Sphingoid base desaturation (most likely at the Δ8 position) seems to be optional in Cer, but obligatory in GlcCer irrespective of the acyl chain length. Similar results were obtained regarding Δ4 desaturation in P. pastoris (Ternes et al., 2011).
The balance between the two groups of Cer and GlcCer species is shifted in the T-DNA insertion lines
Having characterized the structural properties of the two groups of Cer and GlcCer species, we then compared their levels in Col-0 with those in the T-DNA insertion lines (Fig. 7). The levels of C20–C28-Cer and GlcCer species are reduced in the loh1 and loh3 lines in comparison with Col-0, strongly suggesting that LOH1 and LOH3 are responsible for the biosynthesis of Cer species with these acyl chain lengths. The reduction in C20–C28-Cer and GlcCer species is stronger in the loh1 than in the loh3 line. This matches the real-time PCR data showing that the LOH1 gene is expressed at higher levels than LOH3 (Fig. 6a). Despite a significant redundancy between LOH1 and LOH3, LOH1 seems to be the major Cer synthase responsible for biosynthesis of C20–C28-Cer species.
In contrast to the relatively weak reduction in C20–C28 species in the loh1 and loh3 lines, C16-Cer and GlcCer species almost completely disappear in the loh2 line (P < 0.02%). Therefore, LOH2 seems to be the major enzyme responsible for the biosynthesis of the C16-Cer species. This agrees well with the chain length distribution of the Cer species produced by the yeast strain expressing LOH2 (Fig. 3).
Unexpectedly, the levels of C16-Cer and GlcCer species are significantly increased (Cer, 2.5-fold; GlcCer, 1.5-fold; P < 2%) in the loh1 but not in the loh3 line (Fig. 7). A similar increase in total sphingolipids containing a C16 fatty acid was observed by Markham et al. (2011). It is important to recall that LOH2 transcript levels in the loh1 line are unchanged. Instead, inactivation of LOH1 might lead to an accumulation of free sphingoid bases, which could be substrates for LOH2. Indeed, a roughly fivefold accumulation of free trihydroxy but not dihydroxy sphingoid bases is observed in the loh1 line (Fig. 10). As the C20–C28-Cer species, which LOH1 produces in Col-0, are nearly 100% C4-hydroxylated (Fig. 8a), the accumulation of free trihydroxy sphingoid bases may be caused by a decreased consumption for the biosynthesis of C20–C28-Cer species in the loh1 line. A comparison of the absolute quantities suggests that > 95% of these sphingoid bases may be converted into C16-Cer and GlcCer species: while free trihydroxy sphingoid bases are increased by 260 pmol relative the Col-0, C16-Cer is increased by 2230 pmol and C16-GlcCer by 3260 pmol. In support of this, the C16-Cer and GlcCer species are enriched in trihydroxy sphingoid bases in the loh1 line (Fig. 8a, top), suggesting that sphingoid bases normally linked to C20–C28 fatty acids are now being used for the biosynthesis of C16-Cer species. This compensation may occur either passively because more sphingoid bases are available as substrates for LOH2 or actively because a regulatory mechanism might exist to counteract the accumulation of free sphingoid bases.
These findings suggest that elevated amounts of either free trihydroxy sphingoid bases or C16-Cer species trigger the spontaneous cell death observed in the loh1 line. Given that the spontaneous cell death is only observed under short-day conditions, a future study should compare the time-course of C16-Cer and GlcCer build-up under short- and long-day conditions. Elevated amounts of C16-Cer species were also detected in two other A. thaliana mutants which show spontaneous cell death: In the A. thaliana erh1 mutant, Cer species with a trihydroxy sphingoid base and a nonhydroxylated C16 fatty acid are elevated to a higher extent than other Cer species (Wang et al., 2008). In this case, elevated amounts of C16-Cer are not accompanied by corresponding changes in free sphingoid bases. C16-Cer species with α-hydroxylated fatty acid are not elevated in erh1 while in the sbh1 sbh2 double mutant, similar to the loh1 line, C16-Cer species are elevated regardless of their hydroxylation status (Chen et al., 2008). In this double mutant, free sphingoid bases were approx. 60fold increased compared with Col-0. A similar effect to that found in the present study has also been observed in a CerS2 knockout mouse that cannot make C22-C24-Cer. The concentrations of C16-Cer are significantly elevated in this mouse and hepatocytes show an increased rate of apoptosis (Pewzner-Jung et al., 2010a,b). Also here, the increase in C16-Cer is accompanied by a several-fold increase in free sphingoid bases. Both Cer and free sphingoid bases have been shown to induce cell death in cultured plant cells (Townley et al., 2005; Shi et al., 2007; Lachaud et al., 2010). Future studies should aim at identifying the actual active lipid species in each context. In addition, it will be important to compare the changes observed in Cer and GlcCer with those in (G)IPCs. (G)IPCs are the major sphingolipid class in plants, but for technical reasons we were unable to measure them in the present study.
Coming back to the phylogenetic tree shown in Figs 1, S1 the present study has revealed first functional differences between the three Cer synthases from A. thaliana: LOH1 and LOH3 on one side and LOH2 on the other differ regarding their substrate preference for dihydroxy vs trihydroxy sphingoid bases as well as for acyl-CoAs with different chain lengths. These differences are reflected by their phylogenetic separation, which precedes the separation between mosses and vascular plants. LOH1 and LOH3 have redundant biochemical functions which are reflected by their high level of sequence similarity and recent evolutionary origin. Regarding their physiological functions, however, we could show that only the loh1 line shows spontaneous cell death. This may be reflected by the fact that LOH1 is expressed at higher levels and contributes more to Cer biosynthesis than LOH3. It will be interesting to see whether similar differences can be found in other plant species which have independently evolved several Cer synthase isoforms in the evolutionary branch leading to LOH1 and LOH3.
The authors thank Sharon Epstein (Université de Genève, Geneva, Switzerland) for help with the ceramide synthase assay and for discussions, Michael Scharnewski and Martin Fulda, (Georg-August-University, Göttingen, Germany) and Jean-Denis Faure (INRA Versailles-Grignon, Versailles, France) for sharing unpublished data and Cornelia Herrfurth, Sabine Freitag, and Sabrina Brodhun (Georg August University, Göttingen, Germany) for assisting with the analytical procedures and for confirming the identity of the sphingoid bases in Fig. S2. A plasmid containing GFP-LOH1 was kindly provided by Jean-Denis Faure. This work was supported by grants from the Deutsche Forschungsgemeinschaft to PT (TE 491/3-1) and IH (He 3424/1-5) and by a grant from the Swiss National Science Foundation to HR.