Arabidopsis mutants of sphingolipid fatty acid α-hydroxylases accumulate ceramides and salicylates


Author for correspondence:

Ivo Feussner

Tel: +49 551 39 5743



  • In Arabidopsis, the fatty acid moiety of sphingolipids is mainly α-hydroxylated. The consequences of a reduction in this modification were analysed.
  • Mutants of both Fatty Acid Hydroxylase genes (AtFAH1 and AtFAH2) were analysed for sphingolipid profiles. To elucidate further consequences of the mutations, metabolic analyses were performed and the influence on pathogen defence was determined.
  • Ceramide and glucosylceramide profiles of double-mutant plants showed a reduction in sphingolipids with α-hydroxylated fatty acid moieties, and an accumulation of sphingolipids without these moieties. In addition, the free trihydroxylated long-chain bases and ceramides were increased by five- and ten-fold, respectively, whereas the amount of glucosylceramides was decreased by 25%. Metabolite analysis of the double mutant revealed salicylates as enriched metabolites. Infection experiments supported the metabolic changes, as the double mutant showed an enhanced disease-resistant phenotype for infection with the obligate biotrophic pathogen Golovinomyces cichoracearum.
  • In summary, these results suggest that fatty acid hydroxylation of ceramides is important for the biosynthesis of complex sphingolipids. Its absence leads to the accumulation of long-chain bases and ceramides as their precursors. This increases salicylate levels and resistance towards obligate biotrophic fungal pathogens, confirming a role of sphingolipids in salicylic acid-dependent defence reactions.


Sphingolipids are important structural components of membranes of eukaryotic cells. In addition to this role, they are involved in different physiological processes, such as cell polarity, programmed cell death (PCD), guard cell closure and different stress responses (Ng et al., 2001; Coursol et al., 2003; Liang et al., 2003; Shi et al., 2007; Peer et al., 2010; Markham et al., 2011).

Plant sphingolipids can be divided into four different groups: free long-chain bases (LCBs), ceramides, glucosylceramides and glycosyl inositolphosphorylceramides (GIPCs; Pata et al., 2010; Berkey et al., 2012). Sphingolipid biosynthesis starts with the formation of LCBs from a saturated acyl-CoA and the amino acid serine (Sperling & Heinz, 2003). In Arabidopsis, LCBs derive only from palmitoyl-CoA and therefore have a chain length of 18 carbon (C) atoms (Markham et al., 2006). The simplest form is sphinganine, which is hydroxylated at C1 and C3 and has an amino group at C2. Phytosphingosine has an additional hydroxyl group at C4. LCBs are the basic building block of the more complex ceramides. The ceramides are formed by ceramide synthase which catalyses an N-acylation of the LCB (Markham et al., 2011; Ternes et al., 2011). The fatty acid can have various chain lengths from C16 to C26 and can be α-hydroxylated and/or desaturated (Markham et al., 2006). As an additional modification, the LCB portion may be desaturated at positions Δ4 and Δ8, respectively (Sperling et al., 1998; Michaelson et al., 2009). The attachment of a polar head group to the ceramide forms complex sphingolipids (Berkey et al., 2012). Depending on the head group, glucosylceramide and GIPC are the products. The different modifications in the molecules, such as chain length modification, hydroxylation and degrees of desaturation, yield a great structural diversity and complexity, with at least 168 species in Arabidopsis (Markham & Jaworski, 2007).

The importance of the different modifications for the metabolic flow within the biosynthetic pathway has been shown in different mutants: Defects in both LCB Δ8-desaturases, AtSLD1 and AtSLD2, lead to a 50% reduction in glucosylceramides and an increase in GIPCs. As a consequence, double-mutant plants show altered responses to prolonged stress (Chen et al., 2012). A more severe phenotype was described for mutants lacking the LCB C4 hydroxylase. These plants are severely dwarfed and development is blocked in the vegetative stage (Chen et al., 2008). In these plants, the total amount of sphingolipids is increased, in particular species with C16 fatty acids. In addition, for the LCB Δ4-desaturase, a role in the sorting of LCBs for the synthesis of glucosylceramides has been shown (Michaelson et al., 2009). All these modifications occur in the LCBs. Little is known so far about the modifications in the fatty acid moiety in plants.

The Arabidopsis genome harbours two Fatty Acid Hydroxylase genes (AtFAH1 and AtFAH2). Mitchell & Martin (1997) were the first to describe the enzyme AtFAH1, which is a homologue to the yeast FAH1 enzyme, but lacks the cytochrome b5 (Cytb5) domain. However, the Arabidopsis gene has been shown to be able to complement yeast mutants lacking FAH1 activity. An explanation for this successful complementation has been provided recently, as AtFAH1 and AtFAH2 have been shown to interact in plant cells with Cytb5 (Nagano et al., 2009). In addition, it has been shown that the cell death suppressor AtBI-1 can interact with Cytb5, and larger amounts of α-hydroxylated fatty acids can be detected in AtBI-1-overexpressing mutants. Because of this observation, it was suggested that AtBI-1 may regulate PCD by interaction with either both or only one AtFAH, resulting in changes in the amount of α-hydroxylated ceramides. Indeed, this link between an elevated level of α-hydroxylated ceramides and the suppression of PCD supported earlier findings (Townley et al., 2005). Here, elevated levels of the precursor molecules, the nonhydroxylated ceramides, induced PCD in Arabidopsis cells, whereas α-hydroxylated ceramides did not.

The preferred substrates of the α-hydroxylases are most probably ceramides and not free fatty acids, although data supporting this assumption are lacking (Sperling & Heinz, 2003; Warnecke & Heinz, 2003).

To show that both AtFAH enzymes are capable of forming α-hydroxylated ceramides, and to further clarify the function of the α-hydroxylation of the fatty acid moiety in plant sphingolipids, T-DNA insertion mutants of both genes were analysed. In the mutants, changes in sphingolipid species and other metabolites were determined, and their behaviour in development and in response to fungal infection was analysed.

Materials and Methods

Plant growth and cultivation

Plant lines used in the experiments were all in the Arabidopsis thaliana (L.) Heynh Col 0 background. Mutant lines of AtFAH1 and AtFAH2 were named fah1 and fah2. All lines were obtained from the Nottingham Arabidopsis Stock Centre (Loughborough, UK). Plants were grown on either soil or Murashige and Skoog (MS) medium (Duchefa Biochemie, Haarlem, the Netherlands). For soil-grown plants, seeds were sown on steamed (8 h, 80°C) soil and stratified for 2 d in the dark at 4°C. Plants were grown under long-day conditions (16 h light : 8 h dark) at 22°C, 60% humidity and a light intensity of 120–150 μmol m−2 s−1.

For sterile growth on half-strength MS medium, seeds were first sterilized and afterwards plated onto Petri dishes with solid half-strength MS medium with 1% sucrose. After stratification for 2 d at 4°C in the dark, plants were grown under long-day conditions.

Reverse transcriptase-polymerase chain reaction (RT-PCR)

Total RNA was extracted from 100 mg of rosette leaf material using the TRIZOL method (Chomczynski & Mackey, 1995). RNA was treated with DNase and then reverse transcribed by RevertAid™ H Minus Reverse Transcriptase with oligo(dT) primer according to the manufacturer's instructions (MBI Fermentas, St. Leon Rot, Germany). One microlitre of cDNA was used for PCR with ExTakara Taq polymerase (Takara Bio Inc., Madison, WI, USA) according to the manufacturer's instructions. Primers for PCR are shown in Supporting Information Table S1.

Complementation of fah1×fah2

cDNA of the AtFAH2 gene was amplified with Phusion Polymerase (Finnzymes, Espoo, Finland) using the primers shown in Table S2. First, the gene was transferred into the pUC18-Entry vector and then integrated between a p35S Cauliflower mosaic virus (CaMV) promoter and a 35S polyA terminator in Gateway-compatible pCAMBIA1300.GS vector suitable for hygromycin selection (Hornung et al., 2005).

Homozygous plants of the fah1×fah2 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 selected as described by Harrison et al. (2006).

Promoter-β-glucuronidase (GUS) assay

GUSplus DNA (Cambia, Brisbane, Qld, Australia) was amplified with Phusion Polymerase and cloned into the pUC18-Entry vector. For AtFAH1 and AtFAH2 promoter cloning, the region of 1400 bp in front of the ATG and the first two exons were amplified using the primers shown in Table S2. Each AtFAH promoter was cloned in front of the GUS gene in the pUC18-Entry vector and the construct was transferred into the pCambia3300.GC vector (with BASTA selection gene), which was used for A. tumefaciens transformation. BASTA-resistant plants were selected and GUS staining was performed as described previously (Jefferson et al., 1987).

Analysis of ceramides, glucosylceramides and LCBs

Analyses were performed as described previously (Ternes et al., 2011).

Metabolite fingerprinting

For metabolite fingerprinting, plant material (three biological replicates per condition) was extracted using a two-phase extraction with methyl-tert-butylether (MTBE), according to Matyash et al. (2008). The analysis was performed twice for each extract by ultra-performance liquid chromatography (UPLC, ACQUITY UPLC System; Waters Corporation, Milford, MA, USA) coupled with a photodiode array detector (UPLC eLambda 800 nm; Waters Corporation) and with an orthogonal time-of-flight mass spectrometer (TOF-MS, LCT Premier; Waters Corporation).

For LC, an ACQUITY UPLC BEH SHIELD RP18 column (1 × 100 mm, 1.7 μm particle size; Waters Corporation) was used at a temperature of 40°C, flow rate of 0.2 ml min−1 and with a binary gradient of solvent A (water : formic acid (100 : 0.1, v/v)) and solvent B (acetonitrile : formic acid (100 : 0.1, v/v)). The following gradient was applied for the analysis of the polar phase: 0–0.5 min, 10% solvent B; 0.5–3 min, from 10% to 28% solvent B; 3–8 min, from 28% to 95% solvent B; 8–10 min, 95% solvent B; 10–14 min, 10% solvent B. The following gradient was applied for the analysis of the nonpolar phase: 0–0.5 min, 46% solvent B; 0.5–5.5 min, 46–99% solvent B; 5.5–10 min, 100% solvent B; 10–13 min, 46% solvent B.

The TOF-MS was operated in negative as well as positive electrospray ionization (ESI) mode in W optics and with a mass resolution larger than 10 000. Data were acquired by MassLynx 4.1 software (Waters Corporation) in centroided format within the dynamic range enhancement mode over a mass range of m/z 85–1200 with scan duration of 0.5 s and an interscan delay of 0.1 s. The capillary and cone voltages were maintained at 2700 V and 30 V, respectively, and the desolvation and source temperatures at 350 and 80°C, respectively. Nitrogen was used as cone (30 l h−1) and desolvation (800 l h−1) gas. All analyses were monitored using leucine-enkephaline ([M + H]+ 556.2771 or [M + H] 554.2615) as well as its 13C isotopomer ([M + H]+ 557.2799 or [M + H] 555.2615; Sigma-Aldrich, Deisenheim, Germany) as lock spray reference compounds. The raw mass spectrometry data of all samples were processed using MarkerLynx Application Manager 4.1 for MassLynx software (Waters Corporation), resulting in four datasets.

For further data processing, and for clustering and visualization, the toolbox MarVis (MarkerVisualization, was used (Kaever et al., 2009, 2012). First, an ANOVA test, combined with a false discovery rate (FDR) control test (Benjamini & Hochberg, 1995), was performed to select a subset of high-quality marker candidates with FDR < 0.01 (dataset of nonpolar analysis) and 10−6 (dataset of polar analysis). Adduct correction was performed according to the following ionization rules: positive ionization: [M + H]+, [M + Na]+, [M + NH4]+; negative ionization: [M − H], [M + CH2O2 − H], [M + CH2O2 + Na − 2H]. Datasets of the positive and negative ionization mode were combined, which resulted in one dataset for the polar phase containing 1649 marker candidates and one dataset for the nonpolar phase containing 2039 marker candidates. The combined datasets were used for cluster analysis by means of one-dimensional self-organizing maps (1D-SOMs). Before clustering, the replicate sample intensities per condition and marker candidate were aggregated using the average value, and the resulting profiles were normalized to Euclidean unit length. Marker candidates were annotated by matching the corrected masses to exact compound masses in the following public databases using a mass tolerance of 5 mDa: AraCyc (, MetaCyc (, KEGG (, KNApSAcK ( and LIPIDMAPS (

The identity of dihydroxybenzoic acid (DHBA) glucoside and xyloside, as well as the glucosides of hydroxyindole-3-carboxylic acid and hydroxyindole-3-carboxaldehyde, has been confirmed by MS/MS analysis. Samples were analysed by LC 1290 Infinity (Agilent Technologies, Santa Clara, CA, USA) coupled with a 6540 UHD Accurate-Mass Q-TOF LC MS instrument with Agilent Jet Stream Technology as ESI source (Agilent Technologies). For LC, an ACQUITY UPLC BEH SHIELD RP18 column (2.1 × 100 mm, 1.7 μm particle size; Waters Corporation) was used at a temperature of 40°C, flow rate of 0.5 ml min−1 and with a comparable solvent system and gradient as applied for UPLC analysis. The Q-TOF MS instrument was operated with a detection frequency of 2 GHz in the targeted MS/MS mode. The following source conditions were used: gas temperature, 250°C; drying gas flow, 8 l min−1; nebulizer pressure, 35 psi; sheath gas temperature, 300°C; sheath gas flow, 8 l min−1; capillary voltage, 3000 V; nozzle voltage, 200 V; fragmenter voltage, 100 V. For exact mass measurement, the reference mass corrections with trifluoroacetic acid ([M − H] 112.98559) and HP-921 ([M + CH2O2 − H] 966.00073) as reference compounds were used. Data were acquired by MassHunter Workstation Acquisition software B.04.00 (Agilent Technologies). MassHunter Qualitative Analysis B.05.00 (Agilent Technologies) was used for data analysis.

Determination of multiple phytohormones by high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS)

Two hundred milligrams (FW) of plant material were extracted as described for metabolite fingerprinting analysis. Analysis was performed on an Agilent 1100 HPLC system (Agilent) coupled to an Applied Biosystems 3200 hybrid triple quadrupole⁄linear ion trap mass spectrometer (Life Technologies, Carlsbad, CA, USA). Phytohormones were quantified by multiple reaction monitoring (MRM) in negative ion mode as described by Ternes et al. (2011).

Infection with powdery mildew

Golovinomyces cichoracearum strain UCSC1 was maintained on hypersusceptible Arabidopsis pen2pad4sag101 mutant plants. Plants were inoculated between 4 and 5 wk of age by gently brushing the leaves of diseased plants and healthy plants together to pass the conidia (asexual spores). The disease phenotype was scored 14 d post-inoculation (dpi).

Infection with Verticillium longisporum

Plants were grown in a 1 : 1 sand : soil mixture for 20 d. For infection, roots were gently uprooted, washed and incubated for 35 min in spore solution ((3–4) × 105 ml−1) or in tap water (controls), and then transferred into soil. Infection was documented at 21 dpi.

Verticillium longisporum was grown in potato dextrose broth medium supplemented with streptomycin. Cultures were inoculated with spores and cultivated for 2 wk on a shaker. At 3–4 d before infection, the fungal biomass was transferred into Czapek Dox medium to induce spore production. Spores were harvested by draining the culture through miracloth followed by centrifugation. Spores were washed twice with sterile tap water before they were diluted to the appropriate concentration for infection.

Leaf area measurements

Photographs of each plant were taken from above and the projected leaf area was determined by a software program provided by datInf (Tübingen, Germany).

Verticillium longisporum DNA quantification

The preparation and densitometric analysis of the DNA standard were performed according to Brandfass & Karlovsky (2006), and real-time PCR analysis was performed according to Eynck et al. (2007).


Double-mutant line fah1×fah2 shows a growth phenotype

Two genes have been suggested to hydroxylate the fatty acid moiety of ceramides in ArabidopsisAtFAH1 and AtFAH2 (Mitchell & Martin, 1997; Nagano et al., 2009). For AtFAH1, only one T-DNA insertion line in the promoter of the gene was available in the Col 0 background, and consequently some residual transcript was detected by RT-PCR (Fig. 1a). For the AtFAH2 gene, one T-DNA line with an insertion in the fifth exon was available. Here, RT-PCR data revealed a complete knockout for AtFAH2 (Fig. 1a). Both lines showed no visible changes in their phenotypes. However, under long-day conditions, the double mutant showed c. 25% reduced leaf and root growth (Fig. 1b–e). In contrast with the growth phenotype, no differences in the flowering time point could be detected (Fig. 1f). To confirm that the phenotype is caused by the insertions in the AtFAH genes, the double mutant was complemented with the AtFAH2 gene under the control of the 35S promoter. Complemented lines with high AtFAH2 expression showed growth like the wild-type, whereas low expression resulted in the double-mutant phenotype (Fig. S1).

Figure 1.

The double mutant of the T-DNA insertion lines of AtFAH1 and AtFAH2 in Arabidopsis shows a growth phenotype. (a) The transcript levels of the AtFAH1 and AtFAH2 genes in Col 0, fah1, fah2 and fah1×fah2 were analysed by semiquantitative reverse transcriptase-polymerase chain reaction (RT-PCR) of leaf material. (b) Root growth of Col 0 und fah1×fah2 plants grown vertically on Murashige and Skoog agar plates (bar, 1 cm). (c) Photographs of the single- and double-mutant plants grown for 31 d under long-day conditions. (d) Leaf area quantification of plants grown for 24 d under long-day conditions on soil. The data represent the mean values of 16 plants from three independent experiments ± SD. (e) Root growth quantification of plants grown for 14 d under long-day conditions on vertical Murashige and Skoog agar plates. The mean values of 72 plants per line from three independent experiments are shown ± SD. (f) Determination of the flowering time point of plants grown under long-day conditions. The numbers of leaves were determined when the inflorescence stem of the plants had a height of 1 cm. The mean values of 24 plants per line from two independent experiments are shown ± SD. Asterisks indicate significant differences between control and infected plants according to Student's t-test (***,  0.001).

AtFAH gene expression

To compare the expression of both AtFAH genes in different tissues, promoter : GUS fusions were generated and introduced into Col 0 Arabidopsis plants by A. tumefaciens-mediated transformation. Expression of both genes showed a similar pattern in the different tissues (Fig. 2). In 14-d-old seedlings, expression was visible in the whole seedling – in the roots and in the shoots. In addition, in young rosette leaves of 28-d-old plants, expression was detectable, but not in old leaves of the same plant. Moreover, in floral buds including the young part of the inflorescence stem and the young leaves, at the stem blue staining was detected. In the open flower, expression was restricted to the pollen.

Figure 2.

AtFAH1 and AtFAH2 are expressed in young growing tissue and in the pollen grains. Localization of β-glucuronidase (GUS) activity in Arabidopsis plants transformed with AtFAH promoter–GUS fusions. Transformed plants were grown under long-day conditions on soil, except for the seedlings which were grown on Murashige and Skoog agar plates. Staining was performed with three independent plant lines of each construct with comparable results (bars, 2 mm).

In summary, the expression of both genes was detected in young growing tissue and in the pollen. In comparison, the Arabidopsis eFP browser ( confirms the expression of both genes in the pollen, which is c. 10–30 times higher than in the leaves of the plants. The similar expression of both genes suggests redundancy of both genes, also seen in the growth phenotype.

Comparison of the expression data from the eFP browser of different sphingolipid genes revealed that the AtFAH genes show a similar expression pattern to sphingolipid-C4-hydroxylase 2 (SBH2) and the glucosylceramide synthase (GCS; Fig. S4). All show high expression in the pollen and minor expression in the roots, leaves and floral tissue,

Sphingolipid profiles in the mutant lines

Ceramides, glucosylceramides and GIPCs are the main sphingolipid species in Arabidopsis. With the extraction method and UPLC-ESI-TOF-MS analysis used in this study, only ceramides and glucosylceramides were measurable (Ternes et al., 2011; Fig. 3). These metabolites harbour hydroxyl groups, which can be detected in different positions in the molecule. Dihydroxy-ceramides harbour two hydroxyl groups at positions C1 and C2 of the LCB. Trihydroxy-α-ceramides contain an additional hydroxyl group in the fatty acid moiety at the α-position, whereas, in the trihydroxy-phytoceramides, the third hydroxyl group is in the C4 position of the LCB. Tetrahydroxy-ceramides contain both additional hydroxyl groups. It was expected that, in the double mutant, the trihydroxy-α- and tetrahydroxy-ceramides, and glucosylceramides, would be reduced.

Figure 3.

Ceramide, glucosylceramide and long-chain base (LCB) pools in wild-type and α-hydroxylase mutant Arabidopsis plants. Col 0 (white bars), fah1 (light grey bars), fah2 (dark grey bars) and fah1×fah2 (black bars) plants were grown for 35 d under long-day conditions. (a) Ceramides and glucosylceramides were extracted from leaf material by chloroform : methanol extraction, solid phase extraction separation and alkaline hydrolysis, and analysed by ultra-performance liquid chromatography-electrospray ionization-time of flight-mass spectrometry (UPLC/ESI-TOF-MS). The mean values of six biological replicates from two independent experiments are shown. (b) Total ceramides and glucosylceramides are shown as the sum of the mean values of (a). (c) LCB analysis was performed by UPLC/ESI-TOF-MS with three biological replicates.

In general, the analysis showed that α-hydroxylated ceramides could be detected, especially in the glucosylceramide pool and only in trace amounts in the ceramide pool (Fig. 3), as described previously (Pata et al., 2010). For the glucosylceramide analysis, the changes in the double mutant were as expected (Fig. 3a). The amount of trihydroxy-α-ceramide was much lower in the double mutant (0.1 nmol g−1 FW) than in the wild-type (10 nmol g−1 FW). Similarly, the amount of the tetrahydroxy species was significantly lower (10 nmol g−1 FW) than in wild-type plants (40 nmol g−1 FW). In marked contrast, the dihydroxy- and trihydroxy-phytoglucosylceramides were 10-fold higher in the double mutant relative to the wild-type.

With respect to the ceramide pool of the double-mutant plants, the trihydroxy-phytoceramides were 10 times higher than in wild-type plants (40 nmol g−1 FW vs 4 nmol g−1 FW; Fig. 3b). Trihydroxy-α-ceramides were only detectable in trace amounts and below the limit of detection in the double mutants. However, the tetrahydroxy-ceramide amount in the double mutant was enriched from 3 nmol (wild-type) to 8 nmol g−1 FW.

Comparison of the different chain lengths of the fatty acid moieties within the ceramides suggested distinct substrate specificities of the AtFAHs. Compared with AtFAH1, AtFAH2 appears to preferentially hydroxylate C16 fatty acid moieties (Fig. 4), because species with the C16 moiety were more strongly reduced in the fah2 than in the fah1 mutant. The same was observed for the glucosylceramides (Fig. 4). This observation is in agreement with recent published data of Nagano et al. (2012), who showed that AtFAH2 has a substrate preference for C16 fatty acids, whereas AtFAH1 favours very-long-chain fatty acids.

Figure 4.

Acyl chain length distribution of ceramide and glucosylceramide species of wild-type and α-hydroxylase mutant Arabidopsis plants. The data are from the same analysis as in Fig. 3. Ceramide species are shown according to their fatty acid chain length and hydroxylation status. The mean values of six biological replicates from two independent experiments are shown. Cer, ceramide without α-hydroxyl group; hCer, ceramide with α-hydroxyl group; Glc-Cer, glucosylceramide with Cer; Glc-hCer, glucosylceramide with hCer.

In summary, these data show that the α-hydroxylation of fatty acids linked to ceramides regulates the partitioning of sphingolipids into the different sphingolipid pools, as a reduction in α-hydroxylation activity leads to a reduced glucosylceramide pool, whereas ceramides and LCBs strongly accumulate in double-mutant plants (Fig. 3).

Metabolic consequences of changes in the sphingolipid pools

To elucidate whether the observed changes in the sphingolipid pools have consequences on the overall metabolism of the plants, and to explain the growth phenotype, nontargeted metabolic fingerprinting was performed with fah1×fah2 double-mutant and wild-type plants. Three different plant ages (14, 24 and 35 d) were chosen to analyse the differences in a plant age-dependent time course; 14-d-old plants were analysed because no differences in the phenotype of the double mutant compared with the wild-type could be detected at this time point. In 24-d-old plants, the leaf area and shape started to differ, and, in 35-d-old plants, the differences were obvious (Fig. 1).

For metabolite analysis, a two-phase extraction system was used, resulting in polar and nonpolar extraction phases. As a quality control of the method, we first analysed the lipid-harbouring nonpolar (MTBE) phase and compared the results of this experiment with those from the targeted measurements described in the previous section. The intensity profiles of 2039 marker candidates (Table S5), which were filtered according to an FDR of 0.01, were selected for clustering by means of 1D-SOM. Clustering and visualization by 1D-SOMs allows an overview to be made of the metabolic situation within this sample set of six conditions. Most of the clusters (1–3, 7–8 and 13–14) indicate a correlation of the included marker candidates with the age of the plants (Fig. S2). Specific differences between fah1×fah2 double-mutant and wild-type plants were detected in three of the 15 clusters. These clusters summarize the intensity profiles of eight ceramides and seven glucosylceramides. They represent 10 of 12 major species detected in the targeted analysis and also show comparable fold changes as in the targeted analysis between the mutant and wild-type plants (Table S3).

The analysis of the polar extraction phase leads to a dataset of 1649 high-quality marker candidates (FDR < 1 × 10−6; Table S6). Most of these marker candidates, represented by cluster 1–5 and 8–11 of the 1D-SOM, were again correlated with the age of the plants (Fig. 5a). Strong differences between fah1×fah2 double-mutant and wild-type plants were found in three of the 12 clusters. The most obvious markers within these clusters were related to salicylic acid (SA) and its derivatives. SA, which was detected in cluster 2, accumulated in young mutant plants (14 d) in which phenotypic differences were not yet visible (Fig. 5b). In 35-d-old plants, this accumulation was reduced to nearly wild-type levels. By contrast, SA glucoside (SAG), detected in cluster 7, showed stronger enrichment in older plants (Fig. 5). In addition, the structurally related DHBA glucoside and xyloside accumulated in the fah1×fah2 plants and showed preferential enrichment in the older mutant plants (Bartsch et al., 2010). Next to these SA-related compounds, hydroxylated indole derivatives (hydroxyindole-3-carboxylic acid glucoside and hydroxyindole-3-carboxaldehyde glucoside) accumulated within the fah1×fah2 mutant plants (Table S4).

Figure 5.

Metabolite fingerprinting of fah1×fah2 mutant leaves reveals salicylic acid (SA)-derived metabolites as markers detected in the polar extraction phase. Arabidopsis plants were grown under long-day conditions for 14, 24 and 35 d. Leaf material was extracted by two-phase partitioning and analysed by ultra-performance liquid chromatography-electrospray ionization-time of flight-mass spectrometry (UPLC/ESI-TOF-MS). (a) Metabolite-based clustering by means of a one-dimensional self-organizing maps (1D-SOMs) of 1649 marker candidates from the polar extraction phase (false discovery rate (FDR) < 1 × 106) in 12 clusters is shown. Each cluster is represented by the average intensity profile (prototype) of the associated marker candidates (see colour map on right-hand side). The width of a prototype corresponds to the number of marker candidates in the cluster. Clusters 2, 7 and 12 (red frames) represent markers with increased intensities in fah1×fah2 mutant plants. (b) Box plots of prominent markers which accumulate in the fah1×fah2 mutant plants. For the analysis, four biological replicates per treatment were measured twice. d, days, fxf, fah1×fah2 mutant plants.

To confirm the results of the nontargeted approach on the SA-related compounds, phytohormone levels in 24- and 35-d-old wild-type and fah1×fah2 mutant plants were determined by targeted quantitative analysis (Figs 6, S3). For SA and SAG, this analysis confirmed the data of the nontargeted approach. SA was enriched from 1.1 nmol g−1 FW in the wild-type to 3.5 nmol g−1 FW in the 24-d-old mutant plants (Fig. 6a), whereas no enrichment was detectable in 35-d-old plants (Figs 6a, S3). For SAG, the enrichment was equal in both growth stages – the amount in the double mutant was 4.3 times higher (65–69 nmol g−1 FW) than in wild-type plants. In addition, for DHBA, the enrichment was detectable at both time points. In 24-d-old mutant plants, it was 4.5 times higher (7.8 nmol g−1 FW) than in wild-type plants. In addition to these SA-derived phytohormones, an enrichment of the known stress markers indole carboxylic acid and raphanusamic acid was detected. Both showed a strong enrichment in 24-d-old plants, but not in 35-d-old plants (Figs 6a, S3). For the other phytohormones analysed, such as auxin, abscisic acid and the jasmonic acid (JA)–isoleucine conjugate, no significant differences were detected relative to wild-type plants (Figs 6b, S3). In addition, the expression of SA- and JA-responsive genes was determined to support the metabolite data (Fig. 6c). Indeed, the SA-responsive genes PR1 and PR2 were strongly induced in the double mutant, whereas the JA-responsive gene PDF1.2 was not induced.

Figure 6.

Phytohormone and gene expression analysis in the fah1×fah2 mutant confirms enrichment of salicylic acid (SA) and derivatives in double-mutant plants. (a, b) Analysis of phytohormones and stress markers. Arabidopsis plants were grown for 24 d under long-day conditions. Leaf material of Col 0 (open bars) and fah1×fah2 (closed bars) was extracted with methyl-tert-butylether (MTBE) and phytohormone levels were determined by high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS). The mean values of three biological replicates are shown ± SD. Asterisks indicate significant differences between control and infected plants according to Student's t-test (*,  0.05; **,  0.01; ***,  0.001). (c) Semiquantitative reverse transcriptase-polymerase chain reaction (RT-PCR) of PR1, PR2 and PDF1.2 expression in Col 0 and the fah1×fah2 mutant. RT-PCR was performed with three biological replicates of each line with comparable results.

Together, these results suggest that fah1×fah2 double-mutant plants show constitutively elevated levels of salicylates which may explain their reduced growth.

Infection with powdery mildew and V. longisporum

Because the fah1×fah2 double mutant displayed enrichment of salicylates, stress-related metabolites, infections with two different fungal pathogens were performed: the obligate biotrophic leaf pathogen G. cichoracearum and the hemibiotrophic root pathogen V. longisporum (Klosterman et al., 2009; Lipka et al., 2010).

For powdery mildew infection experiments, the enhanced disease resistance1 (edr1) mutant (Frye & Innes, 1998) was included as control for a resistant plant and Col 0 wild-type and pen2/pad4/sag101 as control for susceptible and hypersusceptible plants, respectively (Lipka et al., 2005).

The photographs of the single leaves shown in Fig. 7 allow a direct comparison of the fungal colonization success. As expected, wild-type plants were normally susceptible, edr1 mutants were fully resistant and pen2/pad4/sag101 triple mutants were strongly infected. The infection phenotype of both single fah mutants was comparable with that of the wild-type, whereas double-mutant plants did not show any macroscopic disease symptoms and were comparable with the resistant edr1 mutant.

Figure 7.

Infection of Arabidopsis fah mutants with Golovinomyces cichoracearum. Plants were grown for 4 wk under short-day conditions and infected by dusting with fungal spores. The edr1 mutant was used as a control for a resistant plant and the pen2/pad4/sag101 mutant as a control for a hypersusceptible line. The photograph shows representative infected single leaves at 14 d post-infection. Infections were repeated three times with comparable results. edr, enhanced disease resistance; pad, phytoalexin deficient; pen, penetration mutant; sag, senescence-associated gene.

Infection with V. longisporum resulted in 20% stronger stunting of the infected leaves of the double mutant compared with the wild-type (Fig. 8). In addition, the V. longisporum DNA amount was increased slightly from 21 pg mg−1 FW to 36 pg mg−1 FW in these plants.

Figure 8.

Infection of the fah1×fah2 mutant with Verticillium longisporum. Arabidopsis Col 0 and fah1×fah2 plants were infected by root dip infection. (a) Photographs of infected and control plants at 21 d post-infection. (b) The relative leaf area was determined based on the photographs of the plants. The leaf area of infected plants was divided by the leaf area of control plants and the results of the mutant plants were set in relation to the value of the wild-type. The data represent the mean values of five independent experiments ± SD. (c) Determination of V. longisporum DNA in infected leaves. The mean values of 12 biological replicates from three independent experiments are shown ± SE. Asterisks indicate significant differences between control and infected plants according to Student's t-test (***,  0.001). VL, V. longisporum.

Together, the enrichment of SA and its derivatives in the double-mutant plants correlates with enhanced resistance to the obligate biotrophic leaf pathogen.


The aim of this work was to show that both AtFAH enzymes are capable of forming α-hydroxylated ceramides and to elucidate the function of the α-hydroxylation of the fatty acid moiety in plant sphingolipids. Because the involvement of the AtFAH1 and AtFAH2 genes in the synthesis of α-hydroxylated ceramides was not shown in planta until recently (Nagano et al., 2012), sphingolipid analysis of T-DNA insertion mutants of the two assigned α-hydroxylase genes in Arabidopsis was performed with single and double mutants (fah1×fah2) of the two genes. In addition, the impact of α-hydroxylated fatty acids within sphingolipids on plant growth and response to fungal infection was tested.

Changes in sphingolipid metabolism of the fah1×fah2 mutant

The fah1×fah2 double mutants showed strongly reduced amounts of ceramides and glucosylceramides with α-hydroxylated fatty acids (Fig. 3). This confirmed the recent data of Nagano et al. (2012), who showed that AtFAH single mutants have reduced amounts of α-hydroxylated fatty acids, and is also consistent with the detected localization of the genes at the endoplasmic reticulum, where ceramide synthesis takes place (Marion et al., 2008).

The fact that the amounts of ceramides and glucosylceramides with α-hydroxylated fatty acids decreased, but they were still present in minor amounts, could be explained by the residual expression of AtFAH1 in single and double mutants shown by semiquantitative RT-PCR analyses (Fig. 1). Surprisingly, the analysis revealed that tetrahydroxy-ceramides did not decrease in the double mutant, but rather increased. Although we may simply explain this observation by residual AtFAH1 activity, another possible explanation may be that there are more genes in the Arabidopsis genome that are capable of introducing hydroxyl groups at the α-position in fatty acids. In mammals, additional enzyme activities that can produce α-hydroxylated fatty acids outside of the nervous system in the absence of the AtFAH homologue FA2H have been described, but have not yet been identified: One possible candidate may be the PHYH gene product, which is an α-ketoglutarate-dependent acyl-CoA α-hydroxylase in the peroxisomal α-oxidation pathway (Hama, 2010). By analogy, possible candidates in Arabidopsis may be the α-dioxygenases 1 and 2 (Hamberg et al., 1999; Bannenberg et al., 2009). These enzymes oxygenate in vitro free fatty acids of different chain lengths, resulting in α-hydro(pero)xy fatty acids, aldehydes and fatty acids, shortened by one C atom (Hamberg et al., 1999). The observation that the Arabidopsis α-Dioxygenase 1 gene is involved in protection against oxidative stress and cell death may support this hypothesis (De Leon et al., 2002). However, the expression of both α-dioxygenase genes is not changed in the double mutant (Fig. S5), suggesting no compensatory effect of these genes. Sphingolipid analysis of α-dioxygenase 1 and α-dioxygenase 2 mutants, as well as crossing with the fah1×fah2 double mutant, will shed light on the involvement of the α-dioxygenase genes in ceramide-dependent α-hydroxylated fatty acid synthesis.

The analysis of ceramides and glucosylceramides showed a strong accumulation of trihydroxy-phytoceramides in the double mutant at the expense of species with α-hydroxylated fatty acids (Fig. 3). Although, in total, the amount of glucosylceramides decreased by c. 25%, the amount of ceramides instead increased by about 10-fold. This may be explained by the current model that α-hydroxylation occurs preferentially at the level of ceramides. This assumption is based on the fact that ceramide synthase from Tetrahymena was shown to be inhibited by α-hydroxylated fatty acids (Kaya et al., 1984). Therefore, the α-hydroxylation should take place after ceramide synthesis. In this case, ceramides would be synthesized in the fah1×fah2 mutant, but not further converted to their α-hydroxylated forms, and subsequently not further converted to α-hydroxylated glucosylceramides and α-hydroxylated inositolphosphorylceramides (IPCs).

The UPLC-MS method used in our approach was not suitable for GIPC analysis in the plant material. As GIPCs are mainly hydroxylated and are the major sphingolipid class in Arabidopsis, it would also be interesting to analyse this sphingolipid class. As it is supposed that the hydroxylation takes place at the ceramide, a similar pattern as for the glucosylceramides can be expected. For the IPC synthase 2, a broad substrate specificity independent of the hydroxylation status of the ceramide has been shown (Wang et al., 2008), which suggests a shift from species with a hydroxylated fatty acid moiety to species with a nonhydroxylated fatty acid moiety in the fah1×fah2 mutant.

Metabolic changes and changes in plant defence in the double mutant

In fah1×fah2 plants, an age-dependent accumulation of SA and SAG was found (Figs 5, 6, S4). In addition, stress markers, such as raphanusamic acid and indoles, as well as DHBA derivatives, were enriched in these plants (Figs 6, S4). All these metabolites have been described in the literature to be enriched under infection conditions with different pathogens (Hagemeier et al., 2001; Bednarek et al., 2009; Bartsch et al., 2010; Iven et al., 2012). Therefore, their enrichment might be a consequence of changes in the SA status in the mutant plants. The accumulation of SA and its derivatives is common in mutant plants disturbed in sphingolipid metabolism, and the data presented in this study further support the observation that elevated ceramide levels may lead to an increase in salicylate levels (Greenberg et al., 2000; Brodersen et al., 2002; Wang et al., 2008). Like fah1×fah2 plants, the ceramide kinase mutant (acd5) and the IPC synthase mutant (erh1) have been described to accumulate ceramides, and subsequently show elevated amounts of SA and SA conjugates (Liang et al., 2003; Wang et al., 2008). However, in contrast with fah1×fah2 plants, both mutants show PCD. SA is an important inducer of the PCD-dependent hypersensitive response, and it has been shown that, after crossing these sphingolipid mutants with SA mutants, the lesion phenotype disappears (Greenberg et al., 2000; Wang et al., 2008). In addition, ceramides rather than hydroxy-ceramides have been described to be potent inducers of PCD (Townley et al., 2005). As the fah1×fah2 mutant also accumulates nonhydroxylated ceramides and salicylates, but shows no PCD, the connection between elevated ceramides, salicylates and the induction of PCD may be more complex than previously expected, or ceramides may not be involved directly in the induction of PCD.

In addition, to ceramides, elevated levels of LCBs were found in the fah1×fah2 mutant (Fig. 3). Moreover, the LCBs could be the reason for the enrichment of SA. The connection between LCBs and SA has been described in a recent model in which LCBs are involved in signal transduction leading to PCD via MPK6 and SA (Saucedo-García et al., 2011). Interestingly, the fah1×fah2 mutant shows only a moderate increase in LCB levels (Fig. 3). This may explain the disappearance of PCD in this mutant and renders the role of LCBs in inducing PCD, together with salicylates, more likely than the role of ceramides.

Salicylates were the strongest markers in the metabolite fingerprinting analysis in the double-mutant plants. This phenotype may be indicative of an enhanced disease resistance phenotype. To test this idea, the mutant plants were infected with G. cichoracearum and V. longisporum (Figs 7, 8). Although macroscopic powdery mildew infection symptoms in fah1 and fah2 single mutants were comparable with those in the wild-type, the double mutant showed an enhanced disease-resistant phenotype (Fig. 7). This is in agreement with the increase in salicylates in these mutants, because it is well established that salicylates are important for defence against biotrophic pathogens, including powdery mildew (Glazebrook, 2005). Enhanced resistance to powdery mildew has also been shown in the above-mentioned sphingolipid mutants acd5 and erh1 which accumulate SA (Wang et al., 2008), underlining the close connection between ceramides or LCB accumulation, enrichment in SA and pathogen defence.

Infection with V. longisporum resulted in 20% stronger stunting of the leaves, but the increase in the V. longisporum DNA amount was not significant (Fig. 8). This different reaction relative to powdery mildew infection is in agreement with the model that SA-mediated defence is especially effective against biotrophic fungi (Glazebrook, 2005). An increase in SAG and DHBA in V. longisporum-infected Arabidopsis plants was shown by Ralhan et al. (2012), but an impact on the infection itself could not be shown as the susceptibility of the SA mutants sid2-2 and nahG was not changed. Therefore, SA and derivatives do not seem to play a central role in the defence against V. longisporum.

In summary, AtFAH1 and AtFAH2 act as fatty acid α-hydroxylases of ceramides in Arabidopsis. This hydroxylation is important for the balance of the LCB, ceramide and glucosylceramide pools, which is subsequently important for SA metabolism. Metabolic changes in the double-mutant plants further promote resistance to G. cichoracearum infections.


We are grateful to Urs Benning, Sabrina Brodhun, Sabine Freitag and Alexandra Matei for excellent technical assistance and the members of the Verticillium research group for fruitful co-operation and discussions. This work was supported by the DFG Research Unit FOR546 Fe 446/6, Ka 1209/3 and the excellence initiative FL3 INST 186/822-1.