Extraplastidial cytidinediphosphate diacylglycerol synthase activity is required for vegetative development in Arabidopsis thaliana


For correspondence (e-mail yonghong.zhou@bio1.rwth-aachen.de).


Cytidinediphosphate diacylglycerol synthase (CDS) catalyzes the activation of phosphatidic acid to cytidinediphosphate (CDP)-diacylglycerol, a central intermediate in glycerolipid biosynthesis in prokaryotic and eukaryotic organisms. Cytidinediphosphate-diacylglycerol is the precursor to phosphatidylinositol, phosphatidylglycerol (PG) and cardiolipin of eukaryotic phospholipids that are essential for various cellular functions. Isoforms of CDS are located in plastids, mitochondria and the endomembrane system of plants and are encoded by five genes in Arabidopsis. Two genes have previously been shown to code for the plastidial isoforms which are indispensable for the biosynthesis of plastidial PG, and thus biogenesis and function of thylakoid membranes. Here we have focused on the extraplastidial CDS isoforms, encoded by CDS1 and CDS2 which are constitutively expressed contrary to CDS3. We provide evidence that these closely related CDS genes code for membrane proteins located in the endoplasmic reticulum and possess very similar enzymatic properties. Development and analysis of Arabidopsis mutants lacking either one or both CDS1 and CDS2 genes clearly shows that these two genes have redundant functions. As reflected in the seedling lethal phenotype of the cds1cds2 double mutant, plant cells require at least one catalytically active microsomal CDS isoform for cell division and expansion. According to the altered glycerolipid composition of the double mutant in comparison with wild-type seedlings, it is likely that the drastic decrease in the level of phosphatidylinositol and the increase in phosphatidic acid cause defects in cell division and expansion.


Cytidinediphosphate diacylglycerol synthase (CDS) is a ubiquitous, membrane-bound enzyme catalyzing the transfer of a cytidyl group from cytidine triphosphate (CTP) to phosphatidic acid (PA), giving rise to cytidine diphosphate diacylglycerol (CDP-DAG), the important branch point intermediate in the glycerolipid biosynthesis of prokaryotic and eukaryotic organisms. In eukaryotes CDS competes with PA phosphatase for the substrate. Unlike PA phosphatase, which produces diacylglycerol (DAG) for the biosynthesis of the major membrane and storage lipids, CDS channels its reaction product into the biosynthesis of anionic phospholipids, which are minor but essential components in the various membrane systems. Hence, the partitioning of PA among CDP-DAG and DAG is an important regulatory point in glycerolipid metabolism in eukaryotes.

In eukaryotes CDS has been shown to be an integral membrane protein of the inner mitochondrial membrane and the microsomes, especially the endoplasmic reticulum (ER), while plants possess CDS activity in the inner envelope membrane of plastids as well (Dowhan, 1997; Lykidis and Jackowski, 2000; Frentzen, 2004). Within the mitochondria, CDP-DAG can be channeled into the biosynthesis of phosphatidylglycerol (PG) and cardiolipin (CL) (Schlame et al., 2000), while plastidial CDS activity is required for the biosynthesis of plastidial PG (Andrews and Mudd, 1985; Haselier et al., 2010). On the other hand, microsomal CDS activity provides CDP-DAG for the biosynthesis of phosphatidylinositol (PI) and its polyphosphorylated derivatives (Lykidis and Jackowski, 2000; Löfke et al., 2008; Gillaspy, 2011). Plants also utilize microsomal CDP-DAG for biosynthesis of PG (Kleppinger-Sparace and Moore, 1985; Müller and Frentzen, 2001) and certain species, like wheat, for biosynthesis of phosphatidylserine (PS) (Delhaize et al., 1999; Vance and Steenbergen, 2005).

Yeast cells have been shown to possess a single CDS gene that is essential for cell viability (Shen et al., 1996). In Drosophila melanogaster, alternative splicing causes the formation of two isoforms, of which CDS2 expressed in the fly retina serves as a key regulator of phototransduction (Wu et al., 1995). Vertebrates possess two CDS genes encoding a house-keeping and a tissue-specific isoform, which are both located in the ER (Lykidis et al., 1997; Inglis-Broadgate et al., 2005). In vertebrates CDS activity was found to be a key regulator in phosphoinositide-mediated signaling pathways by controlling the levels of phosphoinositides (Wu et al., 1995; Volta et al., 1999; Pan et al., 2012). The CDS isoforms of plants are encoded by a small gene family that comprises five members in Arabidopsis thaliana (Beisson et al., 2003). These genes, termed CDS1 to CDS5, were all functionally expressed in yeast cells and rescued the lethal phenotype of the mutant strain carrying a disrupted CDS1 gene. The CDS1, CDS2 and CDS3 protein sequences are very similar to each other, and each protein possesses the same pattern of predicted transmembrane domains (Haselier et al., 2010). They resemble the CDS proteins of vertebrates (Wu et al., 1995; Volta et al., 1999; Pan et al., 2012) but differ in sequence and structure from the closely related CDS4 and CDS5 proteins, which show highest sequence similarity to cyanobacterial proteins (Sato et al., 2000). The proteins CDS4 and CDS5 have been shown to represent the plastidial isoforms located in the envelope membranes of plastids (Froehlich et al., 2003; Haselier et al., 2010). Analysis of Arabidopsis cds4cds5 mutants has provided clear evidence for the importance of plastidial PG for photoautotrophic growth (Haselier et al., 2010), in line with results obtained with an Arabidopsis mutant deficient or defective in the plastidial phosphatidyl glycerophosphate synthase gene (Hagio et al., 2002; Xu et al., 2002, 2006; Babiychuk et al., 2003). Unlike CDS4 and CDS5, the other three CDS genes of Arabidopsis probably encode extraplastidial isoforms, of which CDS3 differs in its expression pattern from the other genes (Figure S1). While CDS1 and CDS2, like CDS4 and CDS5, are constitutively expressed, CDS3 is only expressed during flower and seed development and showed its highest transcript levels in stamens and mature pollens (Winter et al., 2007).

In continuing our research on the CDS genes of Arabidopsis, we have focused on the extraplastidial CDS isoforms. We provide evidence that these isoforms are microsomal proteins with very similar enzymatic properties. Moreover, we present data indicating the essential role of CDS1 and CDS2 in plant cell division and expansion.


Enzymatic properties of CDS1, CDS2 and CDS3

In plant cells CDS activity was detected in plastidial, mitochondrial and microsomal membranes (Andrews and Mudd, 1985; Kleppinger-Sparace and Moore, 1985). CDS4 and CDS5 of Arabidopsis have been shown to encode the plastidial isoforms (Haselier et al., 2010) while CDS1 was claimed to represent a microsomal enzyme (Kopka et al., 1997). To determine the enzymatic properties of the extraplastidial isoforms CDS1, CDS2 and CDS3 and to compare them with the plastidial ones, we performed in vitro enzyme assays with radioactively labeled CTP and unlabeled PA using mitochondrial membranes from yeast cells expressing CDS1 or CDS2 or CDS3. As described before, mitochondrial membranes gave several-fold higher CDS activities than microsomal membranes, regardless of whether a plastidial or an extraplastidial CDS gene was expressed in yeast cells (Haselier et al., 2010). Figure 1 shows that CDS1 and CDS2 were most active at pH 8 while CDS3 had a pH optimum at about pH 7.5, similar to the plastidial CDS enzymes. The CDS isoforms of Arabidopsis were found to require Mg2+ for activity that was stimulated by Triton X-100. Unlike the plastidial enzymes, which were most active with 20 mm Mg2+ and 1% (w/v) Triton X-100 (Haselier et al., 2010), the extraplastidial enzymes reached highest activity at 2–5 mm Mg2+ and 0.5% detergent (Figure 1b). Activities of the extraplastidial isoforms as a function of the substrate concentrations gave the highest activities at 2 mm CTP and 0.4 mm PA in contrast to CDS4 and CDS5 for which several-fold higher concentrations (30 mm CTP, 2 mm PA) were optimal. Similar to CDS4 and CDS5, the extraplastidial isoforms were most active with the 18:1/18:1 species of PA and their activities decreased in the order of 16:0/18:1 PA to 16:0/16:0 PA (Figure 1d–f). Hence, the various isoforms encoded by the five Arabidopsis genes have similar properties, except that the extraplastidial enzymes reached substrate saturation at considerably lower concentrations than the plastidial isoforms.

Figure 1.

Enzymatic properties of CDS1, CDS2 and CDS3 heterologously expressed in yeast mitochondrial membrane. Cytidinediphosphate diacylglycerol synthase (CDS) activity as a function of (a) the pH value of the reaction mixture, (b) the Mg2+ concentration and (c) cytidine triphosphate (CTP) concentration (CDS1, diamonds; CDS2, squares; CDS3, triangles). (d–f) Activity of the given CDS enzymes as a function of the phosphatidic acid (PA) species: 18:1/18:1, crosses; 16:0/18:1, bars; 16:0/16:0, circles.

CDS1, CDS2 and CDS3 are ER membrane proteins

As reported before, the extraplastidial CDS proteins of Arabidopsis lack typical N-terminal targeting sequences so that different prediction programs for subcellular localization give inconsistent results (Heazlewood et al., 2005, 2007; Haselier et al., 2010). To gain insight into the subcellular localization of CDS isoforms experimentally, they were expressed as GFP fusion proteins in plant cells. To this end the open reading frames of the CDS genes were fused to the 5′ end of the enhanced GFP (eGFP) sequence and transiently expressed under the control of cauliflower mosaic virus 35S promoter in tobacco leaf epidermal cells. These cells were found to be better suited for the expression of CDS fusion proteins than the previously used BY2 cells (Haselier et al., 2010). For co-localization experiments, agrobacteria containing certain GFP and RFP constructs were mixed and infiltrated into tobacco leaves. As demonstrated in Figure 2(a,b), both CDS1-GFP and CDS2.1-GFP exhibited co-localization with SEKDEL-RFP, an ER marker in tobacco leaves (Maclean et al., 2007). In addition, the N-terminal peptide fusions CDS1_174-GFP and CDS2.1_174-GFP, in which five of the eight transmembrane domains were removed, showed co-localization with the ER marker as well (Figure S2a,b). On the other hand, none of the GFP fusion proteins gave signal patterns typical of chloroplast or mitochondria (Figure 2c,d). Moreover, CDS3-GFP expression provided results similar to CDS1 and CDS2 but weaker fluorescent signals (Figure S2c,d). This also holds true with regard to the CDS2.3 fusion protein having an N-terminal region that is 24 amino acids longer than in CDS2.1 (Haselier et al., 2010). Based on these data, it is likely that the extraplastidial CDS enzymes represent ER membrane proteins and the N-terminal regions covering the transmembrane domains of the proteins can serve as signal peptides which direct CDS1, CDS2.1 and CDS3 to the ER membrane. However, we cannot exclude the possibility that the proteins are located not only in the ER but also at least to a certain extent in mitochondrial membranes. Mitochondrial GFP signals might escape detection because they are rather week or located next to ER structures.

Figure 2.

Subcellular localization of CDS1 and CDS2 transiently expressed as GFP fusion proteins in tobacco leaf epidermal cells. (a) Expression of CDS1_GFP and the endoplasmic reticulum marker SEKDEL_RFP (RFP, red fluorescent protein). (b) CDS2_GFP and SEKDEL_RFP. (c) CDS1_GFP and chloroplast marker cTP_RFP. (d) CDS2_GFP counter stained with Mito Tracker Red. First column, GFP signal; second column, RFP and Mito Tracker signal, respectively; third column, merge of first and second columns. Bars in a, b and c are equal to 50 μm, in d to 10 μm. cTP, chloroplast-transit peptide sequence of Rubisco small-subunit from Solanum tuberosum.

Characterization of cds1 and cds2 single mutants

Based on the microarray data (Winter et al., 2007), CDS1 and CDS2 are constitutively expressed in Arabidopsis, while CDS3 is only expressed in certain plant structures like stamens and mature pollens (Figure S1). To study the biological functions of the extraplastidial CDS isoforms during the vegetative growth phase of Arabidopsis, we have thus focused on CDS1 and CDS2. First, T-DNA insertion mutants were analyzed. Homozygous mutant lines were identified by PCR using gene-specific and T-DNA left border primers. The positions of the insertion sites were confirmed by PCR amplification and sequencing. These studies showed that the T-DNA in cds1 (SALK_088268) is situated in the 5′ untranslated region of At1 g62430, 25 bp upstream of the start codon. The cds2 mutant (SALK_011704) contains a T-DNA in the intron of the 5′ untranslated region at position –169. In addition, a region of 40 bp 5′ of the insertion site is missing (Figure S3). Analysis of the RT-PCR products of wild-type (WT), cds1 and cds2 leaves showed that CDS1 transcripts in the cds1 mutant and CDS2 transcripts in the cds2 mutant were undetectable, while the expression levels of the other CDS genes were very similar to WT. In line with microarray data, CDS3 transcripts were undetectable in leaf tissue (Figure S3).

Homozygous mutants of both cds1 and cds2 showed no obvious alterations of growth and development. They were able to germinate on soil or MS medium with or without sucrose. Mutant plants also developed flowers, siliques and seeds like the WT. In addition, lipid analysis gave no indication that the lipid metabolism of the single cds mutants was altered. Hence single mutations in CDS genes have no obvious effect on plant physiology, suggesting that a defect in CDS1 can largely be compensated by CDS2 and vice versa.

The cds1cds2 double mutant is seedling lethal

To directly demonstrate that CDS1 and CDS2 have redundant functions, double mutants were created by crossing homozygous cds1 and cds2 mutant plants. The F1 progeny were genotyped by PCR and mutant plants heterozygous for the two CDS genes were selfed. In the F2 progeny some seedlings developed distinctly more slowly than the others. To investigate whether these tiny seedlings represent homozygous double mutants, heterozygous mutant plants (cds1+/−cds2−/−) of the F2 generation were selfed. Abnormal tiny seedlings were also discovered among the F3 progeny at a ratio of about 1–3 (101 abnormal and 334 normal seedlings). Genotype analysis based on PCR indicated that all of the abnormal seedlings represented homozygous cds1cds2 double mutants.

The cds1cds2 double mutants germinated on plates with MS medium including Gamborg B5 vitamins (MSG) with or without sucrose and expanded cotyledons, but grew slowly and had smaller cotyledons and shorter hypocotyls and radicles than WT seedlings (Figure 3a). More critically, cds1cds2 mutants failed to develop true leaves and died about 2 weeks after germination, independently of sucrose supply. The inability to grow after germination was reflected in very high amounts of starch in the chloroplasts of cotyledons from double mutant seedlings, which was not observed in WT seedlings (Figure 3b). On the other hand, chlorophyll autofluorescence of cotyledons from cds1cds2 double mutants was as strong as that of WT seedlings. In that way, the cds1cds2 mutant differed from the thylakoid-defective cds4cds5 mutant (Haselier et al., 2010), which showed a rather weak autofluorescence signal and needed externally added sucrose for growth (Figure S4). These data combined with the green color of cotyledons indicated that the thylakoid membranes of cds1cds2 double mutant plants may be intact, and this was supported by ultrastructural analyses (see below).

Figure 3.

The phenotype of double mutant cds1cds2 seedlings. (a) Phenotype of 5-day-old seedlings of homozygous cds1cds2 mutant (two in the middle) and heterozygous mutant (left and right). Bar = 2 mm. (b) Lugol's solution-based starch staining of 7-day-old cds1cds2 (left) and wild type (WT) (right) seedling. Bar = 2 mm. (c) Complementation assay of cds1cds2 by expression of the CDS1 or CDS2 cDNA. Seedlings were grown on MSG (MS medium including Gamborg B5 vitamins) plates containing 2% sucrose and 10 μm β-estradiol and MSG containing 10 μm β-estradiol was added every day for 2 weeks; a, cds1cds2 transformed with an estradiol-inducible CDS2 cDNA (cds1cds2+XVE::CDS2); b, WT; c, cds1cds2 mutants; d, cds1cd s2+XVE::CDS1. Bar = 5 mm.

For a complementation experiment, double mutant plants homozygous for cds2 and heterozygous for cds1 were transformed with the CDS1 or CDS2 cDNA under the control of the XVE-Olex-46 promoter system, under which gene expression is strongly induced by estradiol. As depicted in Figure 3(c), homozygous double mutants expressing a chimeric CDS gene developed and grew like WT plants while the seedlings of cds1cds2 without complementation construct stopped growing, became yellowish and died. Hence, the seedling-lethal phenotype of cds1cds2 double mutants can be rescued by inducing the expression of a functional CDS1 or CDS2 gene.

In contrast to WT and cds4cds5 mutant tissues, which were easily induced to form calluses, we failed to develop callus of cds1cds2 double mutant regardless whether cotyledons, hypocotyls or radicles were transferred on callus-inducing medium (Figure S5). Just cds1cds2 containing an inducible construct with a functional CDS1 or CDS2 gene was able to form callus when CDS gene expression was induced by estradiol (Figure 4). If such a callus was cut into two pieces and transferred to callus-inducing medium with or without estradiol, it only survived on media with estradiol while in the absence of estradiol callus stopped growing and died, unlike WT calluses (Figure 4). These data indicate that CDS1 or CDS2 functions are indispensible for callus formation and growth.

Figure 4.

Callus development and growth requires a functional CDS1 or CDS2 gene.(a), (b) Cotyledons of wild type (WT) (a), cds1cds2 (b), cds1cds2+XVE::CDS1 (c) and cds1cds2+XVE::CDS2 (d) were transferred onto callus-inducing medium without (a) and with (b) β-estradiol, respectively, and incubated at 26°C in the dark for 2 weeks. Bar = 1 mm. (c), (d) Callus of WT (a), cds1cds2+XVE::CDS1 (c) and cds1cds2+XVE::CDS2 (d) were cut and transferred onto callus-inducing medium without (c) and with β-estradiol (d), respectively, and incubated at 26°C in the dark for 3 weeks. Bar = 2 mm.

Ultrastructure of mesophyll cells in cds1cds2 double mutant

Ultrastructural comparison of mesophyll cells from cotyledons shown in Figure 5 demonstrated that mesophyll cells of the double mutant had a smaller cell size than WT seedlings. In addition, vacuoles in the double mutant cells were relatively small so that the cytoplasm with organelles took up more space in mutant than in WT cells. This finding points to a defect in cell expansion of the double mutant resulting in strong suppression of growth (Figure 3). Chloroplasts in the double mutant were slightly smaller and accumulated much more starch than WT chloroplasts. They appeared to have a typical thylakoid structure, but due to the starch grains accumulating in the organelles thylakoid membranes were compressed (Figure 5c–f). We observed no obvious difference in the structures of mitochondria and ER between double mutant and WT (Figure 5e–h).

Figure 5.

Ultrastructure of wild type (WT) and cds1cds2 cells of cotyledons and radicles. Transmission electron microscopy images of mesophyll cells from cotyledons (a–f) and cortex cells from radicles (g, h) from 5-day-old WT (a, c, e and g) and cds1cds2 seedlings (b, d, f and h). (a), (b) Survey of mesophyll cells. (c), (d) Typical structure of chloroplasts. (e), (f) Mitochondria and part of the chloroplasts. (g), (h) Mitochondria and membranes of the endoplasmic reticulum. Bar = 5 μm in (a) and (b). Bar = 1 μm in (c) and (d). Bar = 0.5 μm in (d–h).

Biosynthesis of membrane lipids in cds1cds2

The biosynthesis of phospholipids in Arabidopsis seedlings and calluses was studied using in vivo [33P] phosphoric acid labeling. As shown in Figure 6, the labeling of PI and PG, the products of CDS-catalyzed reactions, was markedly decreased in double mutant seedlings. Labeling of PI was about one-third of that of WT and PG was less than 50% compared with that of WT seedlings. However, the labeling of CL, the downstream biosynthesis product of mitochondrial PG, was maintained at low levels very similar to WT. Labeling of PC, the major phospholipid of eukaryotes, mainly generated by an aminoalcohol phosphotransferase activity in eukaryotes, was increased from 28% in WT to 46% in double mutant tissue. On the other hand, the percentage of PE in double mutant seedlings was almost equal to that in WT seedlings although Arabidopsis predominantly produced PE by aminoalcohol phosphotransferase activity as well (Figure 6; Nerlich et al., 2007). Phosphatidic acid, the lipophilic substrate of the CDS reaction, accumulated in the double mutant while it was hardly detectable in WT seedlings. These results underline the importance of CDS activity for phospholipid metabolism. The analysis of 33P labeling experiments and of steady-state lipid composition of calluses revealed similar changes in phospholipid biosynthesis (Figures S6 and S7). The differences in labeling patterns of WT and double mutant calluses were, however, less pronounced, probably because the calluses still contained some residual estradiol and thus some CDS activity after transfer from estradiol-containing to inductor-free medium.

Figure 6.

The phospholipid labeling pattern of cds1cds2 seedlings differs from that of wild-type (WT) seedlings. (a) Lipids extracted from [33P] phosphate-treated seedlings of WT and cds1cds2 were separated by thin layer chromatography; labeled phospholipids were visualized by phosphoimaging, identified by co-chromatography with lipid standards and quantified by scintillation counting [phosphatidic acid (PA), phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylglycerol (PG), phosphatidylinositol (PI), cardiolipin (CL)]. *Separation of PI and phosphatidylserine (PS) in chloroform/triethylamine/ethanol/water (30:35:30:7) on 1.8% boric acid-treated TLC plates verified that labeled PI was not contaminated with labeled PS. (b) Labeling of the corresponding phospholipids from WT (black bars) and cds1cds2 seedlings (white bars) in per cent of total labeling. Data were calculated from five biological replicates and standard deviations are given.

Membrane lipid composition of cds1cds2 mutant seedlings

Membrane lipid compositions of Arabidopsis cds1cds2 double mutant seedlings were analyzed by nanospray ionization tandem mass spectrometry (nanospray-MS/MS) in comparison with WT and the cds4cds5 double mutant. Determination of the glycerolipid compositions showed considerable differences in the patterns of lipid classes in double mutant and WT seedlings (Figure 7). Galactolipids, the major membrane lipids of chloroplasts, amount to 27% monogalactosyldiacylglycerol (MGDG) and 15% digalactosyldiacylglycerol (DGDG) in WT seedlings, but to distinctly lower levels in the total glycerolipids of mutant seedlings. Unlike the galactolipids, the level of the anionic glycolipid sulfoquinovosyldiacylglycerol (SQDG) was only slightly affected in the mutants. Interestingly, analyses of the galactolipid species gave an increased level of the prokaryotic 34:X species (primarily 34:6 MGDG and 36:3 DGDG) and a decreased level of the eukaryotic 36:6 species not only in the cds4cds5 but also in the cds1cds2 seedlings. This was confirmed by sn2 lyso-lipid analysis which showed a strong increase of 16:3 lyso-MGDG and 16:0 lyso-DGDG and a severe decrease of 18:3 lyso-DGDG in cds4cds5 (Figure 8).

Figure 7.

Glycerolipid composition and molecular species composition of wild-type (WT), cds1cds2 and cds4cds5 seedlings. Lipids were extracted and analyzed from 5-day-old seedlings of WT (black bars), cds1cds2 (white bars) and cds4cds5 (gray bars). Data are given in mole per cent and molecular species in total acyl carbons:total double bonds. MGDG, monogalactosyldiacylglycerol; DGDG, digalactosyldiacylglycerol; PA, phosphatidic acid; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PI, phosphatidylinositol.

Figure 8.

Fatty acid composition at the sn-2 position of the glycerolipids. Lipids extracted from 5-day-old seedlings of wild type (WT, black bars), cds1cds2 (white bars) and cds4cds5 (gray bars), were digested with Rhizopus arrhizus lipase to remove the fatty acid at the sn-1 position, and the lyso-lipids were analyzed by a quadrupole time-of-flight tandem mass spectrometer Data were calculated from three to five biological replicates and standard deviations are given. MGDG, monogalactosyldiacylglycerol; DGDG, digalactosyldiacylglycerol; PG, phosphatidylglycerol; PE, phosphatidylethanolamine; PC, phosphatidylcholine.

Unlike galactolipids, the levels of the sum of PC and PE, the major membrane lipids of the extraplastidial membranes of eukaryotic cells, rose from 36 to 50% in the cds1cds2 mutant. Not only the molecular species composition of PC and PE, but also the alterations in species composition between WT and the mutants were very similar, namely a slight increase in 34:X and a decrease in 36:X species in cds1cds2 (Figure 7). In line with the biosynthetic origin of PC and PE from the eukaryotic pathway, sn2 lyso-PC and lyso-PE in WT and mutants almost exclusively consisted of 18:X species (Figure 8). Phosphatidic acid, the central intermediate of glycerolipid biosynthesis, amounted to less than 1% of the total glycerolipids of WT and cds4cds5 mutant. The PA level was eight-fold higher in the cds1cds2 mutant. In addition, the PA lipid species composition was very similar to those of PC and PE, in particular with regard to the 34:X and 36:X species (Figure 7). These data are consistent with the origin of PA from the eukaryotic pathway at the ER, like PE and PC.

Phosphatidylserine, an anionic, very low abundance membrane lipid formed by a base-exchange mechanism in Arabidopsis (Yamaoka et al., 2011), comprised about 1% of the total glycerolipids of WT and mutant seedlings (Figure 7) and was undetectable in phosphate labeling experiments (Figure 6). Unlike PS, the anionic phospholipids PI and PG are synthesized in CDP-diacylglycerol-dependent reactions and both phospholipids showed reduced labeling patterns in mutant seedlings (Figure 6). The level of PI in total glycerolipid extracts was clearly reduced in cds1cds2 but increased in cds4cds5 seedlings, while its molecular species composition showed only slight alterations. On the other hand, the PG level of the cds1cds2 mutant was hardly affected, unlike that of the cds4cds5 mutant lacking plastidial PG synthesis. Both double mutants showed considerably altered PG species compositions. While the level of the typical plastidial 34:4 PG species carrying 18:3 at position 1 and 16:1Δ3t at position 2 (Devaiah et al., 2006) was strongly reduced, that of 34:3 was increased in both mutants (Figure 7). Lyso-PG analysis revealed that the biosynthesis of PG with sn2 16:1 Δ3t and 16:0 was blocked in the cds4cds5 mutant, while the content of sn2 16:1 Δ3t was reduced with a concomitant increase in 16:0 in PG of cds1cds2 (Figure 8).


Arabidopsis possesses five CDS genes, termed CDS1 to CDS5, of which CDS4 and CDS5 have been shown to encode plastidial isoforms. In this study we have focused on the extraplastidial CDS genes. According to a comparison of the genes and encoded proteins, it is likely that CDS1 and CDS2 derive from the same ancestral gene as CDS3 and code for proteins with characteristics typical of CDS from other eukaryotes (Kopka et al., 1997; Haselier et al., 2010). All three proteins were functionally expressed in yeast membranes and displayed very similar enzymatic properties (Figure 1). In comparison to the plastidial CDS isoforms of Arabidopsis, the extraplastidial ones, however, showed saturation kinetics at distinctly lower substrate concentrations. When CDS proteins were expressed as GFP fusions, they co-localized with an ER marker, suggesting that they are integral membrane proteins of the ER. We cannot exclude the possibility that any of the CDS proteins are located in both the ER and mitochondria. This, however, appears unlikely because none of the GFP fusion proteins gave signal patterns typical of mitochondria and plastids (Figure 2).

Arabidopsis mutants lacking either CDS1 or CDS2 showed no obvious phenotype compared with the WT. On the other hand, the homozygous cds1cds2 double mutant was found to be seedling lethal. It failed to develop true leaves and died 2–3 weeks after germination. We showed that the seedling lethal phenotype of the cds1cds2 double mutant can be rescued by inducing expression of a functional copy of CDS1 or CDS2. The growth defect of cds1cds2 tissue was reflected in its inability to form calluses. Division and extension of cds1cds2 cells were only observed in transformed double mutant tissue when expression of a functional CDS1 or CDS2 gene was induced. Ultrastructure analyses of cotyledons from double mutant seedlings revealed minor differences in comparison with WT seedlings (Figure 5). Apart from smaller cells and vacuoles, numerous starch grains were visible in double mutant chloroplasts. Starch accumulation might point towards a reduced export efficiency of photosynthetic carbon from the chloroplasts.

Consequently, the different phenotype of the single and double mutants and the ability of one functional CDS1 or CDS2 gene to complement the cds1cds2 double mutant provide strong evidence that CDS1 and CDS2 have redundant functions, analogous to CDS4 and CDS5 (Haselier et al., 2010). However, CDS4 and CDS5 supply the plastidial PG pathway with substrate while PI and its phosphorylated derivatives are the main reaction products of the microsomal pathway which is under control of the CDS1 and CDS2 activities. Additionally, in plants the microsomal pathway forms PG as well (Moore, 1982; Müller and Frentzen, 2001). In line with that, phosphate labeling experiments with WT and double mutant seedlings resulted in reduced labeling of the anionic membrane lipids PI and PG, while cardiolipin, the mitochondrial marker lipid synthesized via CDP-diacylglycerol-dependent reactions in the inner mitochondrial membrane (Griebau and Frentzen, 1994), was not affected. The labeling pattern corresponds to the identification of CDS1 and CDS2 as integral membrane enzymes of the ER. Lack of microsomal CDS activities caused an increase in the level of the CDS substrate PA that was in part directed into PC biosynthesis, giving rise to a higher PC level in the mutant than in the WT (Figure 6).

Analyses of the polar membrane lipid composition of seedlings from WT, cds4cds5 and cds1cds2 double mutants support and extend the results of the labeling experiments (Figures 7 and 8). The defect in thylakoid development of the cds4cds5 mutant was reflected in a drastic reduction in the galactolipids and plastidial PG as reported before (Haselier et al., 2010). Seedlings of the cds1cds2 mutant showed similar but less pronounced reductions in the proportion of galactolipids. Since ultrastructure analyses of cds1cds2 seedlings provided no clear evidence for defects in thylakoid development, it is likely that overall chloroplast physiology is affected in cds1cds2. This is also supported by the numerous starch grains that accumulate in the double mutant (Figure 5). Interestingly, the two different mutants had similar alterations in their galactolipid species composition, namely an increased level of the prokaryotic 18:3/16:3 (34:6) MGDG species and 18:3/16:0 (34:3) DGDG species synthesized from plastidial diacylglycerol and a decreased level of the eukaryotic 18:3/18:3 (36:6) species derived from the ER. In the cds4cds5 mutant this can be explained by a re-directing prokaryotic PA from PG synthesis into galactolipids (Figure 7) and this was supported by positional analysis (Figure 8). These altered species compositions were less pronounced in plastids of cds1cds2 mutants and mechanisms different from those in cds4cds5 plastids have to be considered, such as reduced lipid transport from the ER to plastids.

Inactivation of the plastidial CDS enzymes was clearly reflected in a drastically reduced PG level in cds4cds5 seedlings (Figures 7 and 8). On the other hand, phosphate labeling experiments revealed a defect in microsomal PG synthesis in cds1cds2 seedlings (Figure 6) whereas the steady-state PG level was hardly reduced, presumably because of the relatively low level of extraplastidial PG and the different sensitivities of the utilized methods. Interestingly nanospray-MS/MS analysis revealed considerably altered PG species compositions in both mutants. In the cds4cds5 mutant lacking plastidial CDS activity, a drastic reduction in prokaryotic PG (34:4, i.e. 18:3-16:1Δ3t) was found. In the cds1cds2 mutant lacking microsomal CDS activity, PG 34:4 was also reduced. However, this reduction does not reflect a general downregulation of prokaryotic PG synthesis, but rather a decreased degree of desaturation of 16:0 at the sn2 position (Figure 8). Thus, the alteration in PG species in the cds1cds2 mutant can be explained by reduced FAD4 activity (Gao et al., 2009) (Figures 7 and 8).

Unlike PG, PI is exclusively synthesized at the ER (Löfke et al., 2008). Its level was markedly reduced in both phosphate labeling experiments and glycerolipid analysis of the cds1cds2 seedlings (Figures 6 and 7). Phosphatidylinositol not only serves as a membrane lipid but its phosphorylated derivates also play a pivotal role in nearly all cellular processes by serving as signaling molecules that mediate temporal and spatial information to downstream effectors (Xue et al., 2009; Munnik and Vermeer, 2010; Gillaspy, 2011). Therefore it is likely that the defect in PI biosynthesis is decisive in causing the seedling lethal phenotype of the cds1cds2 mutant. In addition, the increase in PA might contribute to the observed phenotype of the mutant since it serves as a central intermediate in glycerolipid biosynthesis and as second messenger (Xu et al., 2005). Unexpectedly, the PI level of the homozygous cds1cds2 mutant was drastically reduced but not abolished (Figures 7 and 8) whereas the level of CL was not affected (Figure 7), showing that the mutant can provide both ER and mitochondria with CDP-DAG independently of CDS1 and CDS2 activity. In this regard CDS3 cannot play a role because it is not expressed in leaves of WT or CDS mutant plants (Figure S1 and Y. Zhou and M. Frentzen, unpublished results). Further experiments that will elucidate whether Arabidopsis possesses further CDS genes unrelated to known CDS sequences or lipid transfer mechanisms between plastids and ER or mitochondria have not yet described in plants.

Experimental procedures

Plant materials and growth conditions

The T-DNA insertion mutants of cds1 and cds2 (SALK_088268 and SALK_011704, Alonso et al., 2003) were obtained from the European Arabidopsis Stock Centre (http://arabidopsis.info/). Seeds were surface sterilized and sown on 0.8% agar-solidified MSG medium (Duchefa, http://www.duchefa-biochemie.nl/) containing 2% sucrose. After stratification in the dark at 4°C for 2 days, seeds were germinated by incubation in short-day conditions (8-h light/16-h dark) at 22°C. Two-week-old seedlings were transferred to potting soil. To induce flowering, plants were cultivated in long-day conditions (16-h light, 23°C/8-h dark, 21°C).

Identification of the single and double mutants

The homozygous cds1 and cds2 single mutants were identified by PCR using the T-DNA-specific primer LBb1.3 and gene-specific primers (LP/RP_088268, LP/RP_011704; Table S1). The position of the T-DNA insertion in the genomic region was confirmed by DNA sequence analysis of the high-fidelity PCR products amplified by Phusion polymerase (New England Biolabs, https://www.neb.com/).

The homozygous cds1 and cds2 single mutants were used to generate double mutants by crossing. Heterozygous double mutant plants in the F1 progeny were identified by PCR and selfed to generate F2 progeny. Heterozygous double mutants possessing one functional CDS1 or CDS2 copy were identified and selfed again to obtain plants of the F3 generation. The homozygous double mutants were isolated from F3 progeny by PCR using specific primers for the cds1 and cds2 single mutants.

Plasmid construction

The cDNA clones of CDS1 (BX813816), CDS2 (pda06248), and CDS3 (BX826782) were obtained from the RIKEN BioResource Center (Seki et al., 2002, 2003; Sakurai et al., 2005; http://www.brc.riken.jp/inf/en/index.shtml) and Genoscope/Life Technologies (INRA, The French National Resources Center for Plant Genomics, Castelli et al., 2004; http://cnrgv.toulouse.inra.fr). The whole or part of the cDNAs were amplified by high-fidelity PCR using Phusion polymerase (New England Biolabs) and gene-specific primers (primers 8–20, Table S1), and were cloned into the entry vector pENTR/SD/D-TOPO (Invitrogen, http://www.invitrogen.com/). The Gateway LR recombination reaction (Invitrogen) was used for transferring cDNAs into destination vectors. For subcellular localization, cDNAs were transferred into destination vector pK7FWG2.0 (Karimi et al., 2002, 2007), in which the whole or part of the cDNAs were fused to the N-terminal of the eGFP gene. For functional expression studies in yeast, cDNAs were transferred into the galactose-inducible vector pYES-DEST52 (Invitrogen). For complementation of the cds1cds2 mutant, cDNAs were transferred into pMDC7, which conferred gene expression under control of an estradiol inducible promoter (Zuo et al., 2000; Curtis and Grossniklaus, 2003).

Plant transformation and identification

For in vivo functional study of CDS1 and CDS2, heterozygous double mutants possessing one functional CDS1 copy were transformed with pMDC7_CDS constructs using the floral dip method (Clough and Bent, 1998). In the F1 progeny, heterozygous double mutants containing one functional CDS1 and pMDC7_CDS construct were identified by PCR and selfed to obtain plants of the F2 generation, which were used for complementation assays and callus cultivation. Homozygous double mutants containing the pMDC7 construct were firstly selected according to the phenotype of homozygous double mutants on MSG plates without estradiol and secondly selected on MSG plates or callus-inducing medium with estradiol, according to WT-like growth, and confirmed by PCR.

Enzymatic assay

The yeast YBR029c cells containing pYES-DEST52_CDSs were obtained from Haselier et al. (2010). The yeast cells were cultivated and mitochondrial fractions were isolated from the cultures according to Zinser and Daum (1995) as described before (Haselier et al., 2010). The CDS activity in mitochondrial membrane was determined in 50 μl of reaction mixture containing 50 mm BTP [bis-TRIS-propane-HCl; TRIS, 2-amino-2-(hydroxymethyl)-1,3-propanediol], pH 8.0, 5 mm MgCl2, 0.5% Triton X-100, 0.2 mm 18:1/18:1 PA, 1.2 mm CTP with 3 × 105 d.p.m. [5-3H]CTP. The reaction was started by adding CTP, incubated at 30°C for 20 min, and subsequently the catalysis was stopped and lipids were extracted by adding 500 μl methanol:chloroform (1:1, v/v) and centrifugation at 106 g for 3 min. The reaction products were quantified by scintillation counting (LS 6500, Beckman, https://www.beckmancoulter.com/).

Subcellular localization of fluorescence proteins in tobacco leaf epidermal cells

Tobacco SR1 (Maliga et al., 1973) plants were grown in 16-h light, 23°C/8-h dark, 21°C and used for studying subcellular localization of fusion proteins using a rapid transient expression method according to Sparkes et al. (2006). Agrobacteria GV3101::pMP90RK containing GFP or RFP constructs were resuspended in infiltration medium (5 g L−1 d-glucose, 50 mm MES, 2 mm Na3PO4, 200 μm acetosyringone), infiltrated into tobacco leaves. After growth in the dark overnight and in 16-h light, 23°C/8-h dark, at 21°C for 48 h, the infiltrated areas of leaves were cut and observed using a laser confocal microscope (Leica, http://www.leica.com/). For the co-localization experiments, agrobacteria GV3101::pMP90RK containing the construct of pTRAkc-rbcs1-cTP or pTRAkc-ERH (Maclean et al., 2007), which show chloroplast and ER localization, respectively, were introduced into leaf epidermal cells together with CDS constructs.

Semi-quantitative reverse transcription–polymerase chain reaction (RT-PCR) analysis

Using the TRIzol method, DNase-treated total RNA was isolated from rosette leaves of 4-week-old plants of WT, cds1 and cds2. The synthesis of first-strand cDNA was performed using Moloney murine leukemia virus reverse transcriptase (Fermentas, http://www.thermoscientificbio.com/fermentas/) and oligo (dT)18 primer. The CDS1 to CDS5 transcripts were amplified by PCR using gene-specific primers (Table S1). In each PCR reaction, the following program was performed: initial denaturing at 95°C for 5 min, 30 cycles of 95°C for 30 sec, 56°C for 30 sec, 72°C for 45 sec, and final extension at 72°C for 5 min. The transcript level of Actin2 was analyzed as a control. The PCR for each of four biological replicates was done in triplicate. The PCR products were separated by 1% agarose gel electrophoresis and stained by ethidium bromide. The amount of products was determined by image lab software (Bio-Rad, http://www.bio-rad.com/).

Callus induction

The cotyledons, hypocotyls and radicles from 5-day-old seedlings of WT and homozygous double mutants of cds1cds2 and cds4cds5 were harvested and transferred to callus-inducing medium containing 3.16 g L−1 Gamborg B5 medium (Duchefa), 30 g L−1 sucrose, 8 g L−1 plant agar, 0.05 mg L−1 kinetin and 0.5 mg L−1 2,4-dichlorophenoxyacetic acid. For inducing expression of CDS in pMDC7 in the homozygous double mutant cds1cds2, 10 μm of β-estradiol was applied to callus-inducing medium. After incubation at 26°C in the dark for 3 weeks, the emerged calluses were further cultivated on callus-inducing medium under the same conditions, and transferred to fresh medium each week.

Microscopic analysis

The following seedlings for microscopy were germinated and grown on MSG plates for 5 days. To determine the starch in seedlings, seedlings of WT and cds1cds2 were harvested at hour 4 in an 8-h photoperiod, decolorized in 80% (v/v) ethanol at 80°C, stained with Lugol solution (Sigma, http://www.sigma-aldrich.com/) for 2 min, and briefly destained with distilled water. The bright field microscopy of starch-stained seedlings was done with a Keyence microscope. The WT and cds1cds2 seedlings were analyzed by transmission electron microscopy to study the ultrastructure of cotyledons and roots, as described before (Haselier et al., 2010). To analyze the red fluorescence of chlorophyll from WT, cds1cds2 and cds4cds5, cotyledons were harvested at hour 4 in an 8-h photoperiod and observed with a Leica fluorescence microscope. The fluorescence was excited at 450–490 nm and detected at 515 nm.

Phospholipid analysis

For in vivo phosphate labeling experiments, 5-day-old WT and cds1cds2 were incubated in MSG medium containing [33P] phosphoric acid for 18 h. Phospholipids were extracted according to Babiychuk et al. (2003), and separated by thin-layer chromatography in chloroform/methanol/acetic acid (65:25:8, v/v). Phosphatidylserine and PI in phospholipid extracts were separated on 1.8% boric acid-treated TLC plates with chloroform/triethylamine/ethanol/water (30:35:35:7). Radioactive phospholipids were visualized by phosphoimaging (Bioimager, Fujifilm, http://www.fujifilm.com/) and quantified by scintillation counting (LS 6500, Beckman).

Lipid profiling

Lipids were extracted from the following materials: leaves of 1-month-old WT, cds1 and cds2, seedlings of 5-day-old WT, cds1cds2 and cds4cds5, and calluses of WT and cds1cds2+XVE::CDS1 (Roughan et al., 1978). For glycerolipid quantification, an aliquot of the lipid sample was diluted in methanol/chloroform/300 mm ammonium acetate (665:300:35, v/v, Welti et al., 2002). Two internal phospholipid standards each were added for PC, PE, PG, PA and PS (Avanti Polar Lipids, http://avantilipids.com/) to the samples according to Welti et al. (2002). The lipid standards MGDG, DGDG, SQDG and PI (Larodan) were hydrogenated prior to addition to the sample according to Buseman et al. (2006). The samples were supplied to a Q-TOF (quadrupole time-of-flight) mass spectrometer (Q-TOF 6530; Agilent, http://www.home.agilent.com/) in methanol/chloroform/300 mm ammonium acetate (665:300:35, v/v) at a flow rate of 0.5 μl min−1 with a nanospray ion source (HPLC Chip/MS 1200 with infusion chip; Agilent). After ionization in the positive mode with a fragmentor voltage of 200 V, ammonium or proton adducts were selected in the quadrupole and fragmented in the collision cell (Welti et al., 2002). The amounts of glycerolipids were obtained after extraction of neutral loss or precursor ion scanning data from MS/MS experiments. After correction of isotopic overlap, lipid molecular species were quantified according to Devaiah et al. (2006).

Positional analysis of glycerolipids

The glycerolipids isolated from Arabidopsis seedlings were dissolved in chloroform and purified using silica columns (Strata 100 mg Silica columns, Phenomenex, http://www.phenomenex.com/). After washing the columns with chloroform, the galactolipids were eluted with acetone/isopropanol (1:1) and the phospholipids with methanol. The glycerolipids were dried and dissolved in ethanol. An aliquot of the glycerolipid fractions was mixed with digestion buffer (40 mm TRIS-HCl, 50 mm Na-borate, pH 7.2) and Rhizopus arrhizus lipase and incubated for 1 h at room temperature (Fischer et al., 1973; Williams et al., 1995). After extraction with chloroform/methanol (1:1) the lyso-glycerolipids (lyso-MGDG, lyso-DGDG, lyso-PC, lyso-PE and lyso-PG) were measured by Q-TOF mass spectrometry (see above).


This work was supported by a grant of the Deutsche Forschungsgemeinschaft, to MF. YZ was financially supported by the China Scholarship Council. The vector of pMDC7 was kindly provided by Professor Nam-Hai Chua (Rockefeller University, New York, USA).

Accession number

Sequence data and Arabidopsis seed stocks: CDS1, At1 g62430; CDS2, At4 g22340; CDS3, At4 g26770; cds1, SALK_088268; cds2, SALK_011704.