Transcriptomic and lipidomic profiles of glycerolipids during Arabidopsis flower development

Authors

  • Yuki Nakamura,

    Corresponding author
    1. Institute of Plant and Microbial Biology, Academia Sinica, Nankang, Taipei, Taiwan
    2. PRESTO, Japan Science and Technology Agency, Saitama, Japan
    3. Department of Biochemistry, Yong Loo Lin School of Medicine, National University of Singapore, Singapore city, Singapore
    4. Temasek Life Sciences Laboratory, 1 Research Link, National University of Singapore, Singapore city, Singapore
    Search for more papers by this author
  • Norman Z. W. Teo,

    1. Department of Biochemistry, Yong Loo Lin School of Medicine, National University of Singapore, Singapore city, Singapore
    2. Temasek Life Sciences Laboratory, 1 Research Link, National University of Singapore, Singapore city, Singapore
    3. Department of Biological Sciences, National University of Singapore, Singapore city, 117543 Singapore
    Search for more papers by this author
  • Guanghou Shui,

    1. Department of Biochemistry, Yong Loo Lin School of Medicine, National University of Singapore, Singapore city, Singapore
    2. State Key Laboratory of Molecular Developmental Biology, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author
  • Christine H. L. Chua,

    1. Temasek Life Sciences Laboratory, 1 Research Link, National University of Singapore, Singapore city, Singapore
    Search for more papers by this author
  • Wei-Fun Cheong,

    1. Department of Biochemistry, Yong Loo Lin School of Medicine, National University of Singapore, Singapore city, Singapore
    Search for more papers by this author
  • Sriram Parameswaran,

    1. Temasek Life Sciences Laboratory, 1 Research Link, National University of Singapore, Singapore city, Singapore
    Search for more papers by this author
  • Ryota Koizumi,

    1. Department of Biological Sciences, Tokyo Institute of Technology, Midori-ku, Yokohama, Japan
    Search for more papers by this author
  • Hiroyuki Ohta,

    1. Center for Biological Resources and Informatics, Tokyo Institute of Technology, Midori-ku, Yokohama, Japan
    2. Earth-Life Science Institute, Tokyo Institute of Technology, Meguro-ku, Tokyo, Japan
    Search for more papers by this author
  • Markus R. Wenk,

    1. Department of Biochemistry, Yong Loo Lin School of Medicine, National University of Singapore, Singapore city, Singapore
    2. Department of Biological Sciences, National University of Singapore, Singapore city, 117543 Singapore
    3. Swiss Tropical and Public Health Institute, University of Basel, Basel, Switzerland
    Search for more papers by this author
  • Toshiro Ito

    Corresponding author
    1. Temasek Life Sciences Laboratory, 1 Research Link, National University of Singapore, Singapore city, Singapore
    2. Department of Biological Sciences, National University of Singapore, Singapore city, 117543 Singapore
    Search for more papers by this author

Summary

  • Flower glycerolipids are the yet-to-be discovered frontier of the lipidome. Although ample evidence suggests important roles for glycerolipids in flower development, stage-specific lipid profiling in tiny Arabidopsis flowers is challenging. Here, we utilized a transgenic system to synchronize flower development in Arabidopsis.
  • The transgenic plant PAP1::AP1-GR ap1-1 cal-5 showed synchronized flower development upon dexamethasone treatment, which enabled massive harvesting of floral samples of homogenous developmental stages for glycerolipid profiling.
  • Glycerolipid profiling revealed a decrease in concentrations of phospholipids involved in signaling during the early development stages, such as phosphatidic acid and phosphatidylinositol, and a marked increase in concentrations of nonphosphorous galactolipids during the late stage. Moreover, in the midstage, phosphatidylinositol 4,5-bisphosphate concentration was increased transiently, which suggests the stimulation of the phosphoinositide metabolism. Accompanying transcriptomic profiling of relevant glycerolipid metabolic genes revealed simultaneous induction of multiple phosphoinositide biosynthetic genes associated with the increased phosphatidylinositol 4,5-bisphosphate concentration, with a high degree of differential expression patterns for genes encoding other glycerolipid-metabolic genes. The phosphatidic acid phosphatase mutant pah1 pah2 showed flower developmental defect, suggesting a role for phosphatidic acid in flower development.
  • Our concurrent profiling of glycerolipids and relevant metabolic gene expression revealed distinct metabolic pathways stimulated at different stages of flower development in Arabidopsis.

Introduction

Flower development is a highly coordinated event throughout the life cycle of seed plants with the aim of successful fertilization and propagation of the subsequent generation. In Arabidopsis and other model plants, genetic study has contributed significantly to unraveling how the dramatic event of flower development is finely regulated (Causier et al., 2010). To further understand the molecular basis of the developmental process, a systems approach demands -omics study.

Because the Arabidopsis inflorescence consists of tiny flowers in different developmental stages, harvesting a sufficient amount of flower samples in a homogenous floral stage is challenging even for RNA/DNA extraction. In transcriptomic studies, efforts have involved synchronizing flower development by means of transgenic technology (Gomez-Mena et al., 2005). In Arabidopsis, double knockout of APETALA1 and CAULIFLOWER (ap1 cal) causes arrest of flower development at the initial step, thus leading to an inflorescence with a cluster of overproliferated nascent floral buds (Kempin et al., 1995). Expressing a gene for the initial step of flower development can be a trigger to release this developmental arrest, which leads to a synchronized development of the ‘cauliflower’ flower. Gomez-Mena et al. (2005) used AGAMOUS (AG) as a trigger to create a transgenic line, 35S::AG-GR ap1-1 cal-1, showing successful release of the arrest to reveal a transcriptional network controlled by AG (Gomez-Mena et al., 2005). To further study the initial stage of flower development that 35S::AG-GR ap1-1 cal-1 skips, Wellmer et al. (2006) created 35S::AP1-GR ap1-1 cal-1 to demonstrate a high degree of differential gene expression during the early stages of flower development. In line with the systems understanding, we still lack information on downstream events associated with the revealed gene expression pattern, that is, metabolomic or proteomic study with the synchronized flower system following the transcriptomic studies highlighted earlier (Bellaire et al., 2014).

Polar glycerolipids (hereafter glycerolipids) consist of a distinct set of metabolites that serve as membrane constituents (e.g. major phospholipids, galactolipids and a sulfolipid) and signaling molecules (e.g. phosphatidic acid (PA), phosphatidylinositol (PI) and phosphoinositides). Distinct biochemical features of glycerolipid metabolism are shown in Petunia flowers. Digalactosyldiacylglycerol (DGDG) is the major glycolipid in flowers, and metabolism in diacylglycerol (DAG) production reveals floral organ-specificity (Nakamura et al., 2003; Nakamura & Ohta, 2007). In Arabidopsis, knockout studies have provided ample evidence of glycerolipid biosynthesis involved in flower development. For example, a knockout mutant of glycerol 3-phosphate acyltransferase 1 (GPAT1) that catalyzes the first step of the extraplastidic Kennedy pathway results in aberrant pollen structure (Zheng et al., 2003). Knockout of plastidic lysophosphatidate acyltransferase 1 (LPAT1) induces embryonic lethality (Kim & Huang, 2004; Yu et al., 2004). A leaky mutant of CTP:phosphorylethanolamine cytidylyltransferase 1 (PECT1) affects flower development, fertility and embryonic development (Mizoi et al., 2006). Knockout of phosphatidylserine synthase 1 (PSS1) confers a male gametophytic defect (Yamaoka et al., 2011). This evidence suggests the critical roles of glycerolipids in multiple stages of flower development.

Here, to understand glycerolipid profiles during flower development, we developed an Arabidopsis transgenic line with synchronized flower development. With this system, glycerolipid profiling by a lipidomics platform revealed a marked decrease in PA and PI concentrations in the early developmental stages and a subsequent increase in galactolipid concentrations, mainly monogalactosyldiacylglycerol (MGDG), in the later stages. Together with the concomitant transcriptomic profiling of glycerolipid metabolic genes by quantitative reverse transcription polymerase chain reaction (qRT-PCR), our analyses reveal distinct patterns of glycerolipid metabolism at different stages of flower development in Arabidopsis.

Materials and Methods

Plant growth and treatment conditions

Arabidopsis thaliana (L.) Heynh plants were grown under continuous light (150 μmol m−2 s−1). The pah1 pah2 mutant plants were grown at 22°C, and pAP1::AP1-GR ap1-1 cal-5 plants were grown at 18°C to avoid precocious initiation of flower development. Dexamethasone (DEX) was applied to the inflorescences of pAP1::AP1-GR ap1-1 cal-5 by dipping in an aqueous solution (pH 7.0) containing 1 μM DEX and 0.015% (v/v) Silwet L-77. Construction of pAP1::AP1-GR ap1-1 cal-5 was described previously (Bellaire et al., 2014).

Lipid extraction

Inflorescences of pAP1::AP1-GR ap1-1 cal-5 were harvested, immediately frozen in liquid nitrogen, and kept at −80°C until lipid extraction. Before lipid extraction, frozen tissues were incubated in hot (75°C) isopropanol containing 0.05% (v/v) butylated hydroxytoluene (cat. no. B1378; Sigma-Aldrich) for 15 min. Total lipid was extracted from c. 500 μl (c. 10 mg DW) frozen tissue as previously described (Bligh & Dyer, 1959).

Lipid analysis by high-performance liquid chromatography (HPLC)/mass spectrometry

An Agilent HPLC system coupled with an Applied Biosystems Triple Quadrupole/Ion Trap mass spectrometer (4000Qtrap; Applied Biosystems, Foster City, CA, USA) was used to quantify individual polar lipids (phospholipids). Polar phospholipid species were quantified using targeted lipidomic approaches as previously described (Fei et al., 2008). Briefly, from product and precursor ion analysis of head groups, multiple reaction monitoring transitions were set up to quantify various polar lipids. Individual lipid concentrations were quantified by normalizing to the corresponding spiked internal standards. Dimyristoyl phosphatidylcholine (28:0-PC) was used to quantify endogenous PC, dimyristoyl phosphatidylethanolamine (28:0-PE) was used for endogenous PE, dimyristoyl phosphatidylserine (28:0-PS) was used for endogenous PS, dimyristoyl phosphatidylglycerol (28:0-PG) was used for endogenous PG, and dimyristoyl phosphatidic acid (28:0-PA) was used for endogenous PA (all of these standards were from Avanti Polar Lipids (Alabaster, AL, USA)). Dioctanoyl PI (16:0-PI; Echelon Biosciences, Salt Lake City, UT, USA) was used to quantify PI species.

Galactolipids, MGDG and DGDG were analyzed using an Agilent 1100 HPLC system coupled with an Applied Biosystems 4000 QTrap mass spectrometer (W.-F. Cheong et al., unpublished). The HPLC system is made up of an Agilent 1100 binary pump, an Agilent 1100 thermo sampler and an Agilent 1100 column oven. A Kinetex 2.6 u C18 100A column (i.d. 4.6 × 100 mm; Phenomenex, Torrance, CA, USA) was used to perform lipid separation with a mobile phase containing chloroform : methanol : 2% 50 mM sodium acetate at flow rate of 180 μl min−1 for 25 min. The MS instrument was operated in positive electronspray ionization (ESI) mode with a capillary voltage of 5000 V, capillary temperature of 300°C and the collision energy ranged from 70 to 75 V. Product ion scan was performed using this approach to generate specific multiple reaction monitoring (MRM) transitions for the individual species of MGDG and DGDG lipids. Individual lipid concentrations were quantified according to spiked purified standards, MGDG and DGDG, from Matreya LLC (Pleasant Gap, PA, USA).

Quantification of phosphatidylinositol 4-phosphate (PI4P) and phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2)

Phosphatidylinositol 4-phosphate (PI4P) and PI(4,5)P2 were quantitatively analyzed as previously described (Nasuhoglu et al., 2002) with a Dionex Ion Chromatography 3000 system (Dionex, Sunnyvale, CA, USA). Lipid extracts were deacylated by incubation with 0.5 ml methylamine reagent (MeOH : 40% methylamine in water : 1-butanol : water (47 : 36 : 9 : 8, v/v)) at 50°C for 45 min. The aqueous phase was dried, resuspended in 0.5 ml of 1-butanol : petroleum ether : ethyl formate (20 : 40 : 1, v/v), and extracted twice with an equal volume of water. Aqueous extracts were dried, resuspended in water, and subjected to anion-exchange HPLC on an Ionpac AS11-HC column (Dionex). Negatively charged glycerol head groups were eluted with a 1.5–86 mM KOH gradient and detected online by suppressed conductivity 75 in a Dionex ion chromatography system equipped with an ASRS-ultra II self-regenerating suppressor (Dionex). Individual peaks of head groups were identified by matching their standard retention times and peak areas were calculated using Chromeleon software (Dionex). Lipid concentrations were calculated with deacylated anionic phospholipids such as dipalmitoyl PI4P and dipalmitoyl PI(4,5)P2 (both from Echelon Biosciences) used as standards.

RNA extraction, cDNA synthesis and qRT-PCR

Total RNA was extracted from the floral samples with the RNeasy Plant Mini Kit (Qiagen), and cDNA was synthesized with the SuperScript III First-Strand reverse transcriptase kit (Invitrogen). Specific primers were designed as listed in Supporting Information Table S1 and confirmed to be specific to the target gene by dissociation analysis. qRT-PCR involved the 7900 HT Fast Real Time PCR System (Applied Biosystems) with Power SYBR Green PCR Master Mix. For each qRT-PCR run, triple technical replicates were prepared and results were averaged. Data in figures were further averaged by three biological replicates of samples. Error bars are indicated for genes with > twofold change in expression. For other genes, SD values were < 10%. The phosphoinositide-biosynthetic gene nomenclature is as defined previously (Mueller-Roeber & Pical, 2002). PIPK10 (At4g01190) and PIPK11 (At1g01460) are according to Perera et al. (2005) and PIPLC6 (At2g40116), PIPLC8 (At3g47220) and PIPLC9 (At3g47290) are according to Hunt et al. (2004).

Results and Discussion

Observation of a transgenic line that synchronizes flower development

The pAP1::AP1-GR ap1-1 cal-5 transgenic plants were modified from a previously used design (35S::AP1-GR ap1-1 cal-1) by using the native promoter of AP1 (pAP1) to reflect its endogenous spatiotemporal expression properties. We transformed it in the ap1-1 cal-5 genetic background, because cal-5 and ap1-1 are both in the Ler ecotype background while cal-1 is in the WS ecotype background (Kempin et al., 1995; Ferrandiz et al., 2000). AP1-GR is an AP1 gene with a carboxyterminal fusion to the steroid-binding domain of the rat glucocorticoid receptor GR (Lloyd et al., 1994). Under normal conditions, the AP1-GR fusion protein is bound to Hsp90 and remains in the cytoplasm, where a nuclear protein AP1 is inactive. Upon treatment with DEX, AP1-GR is released from Hsp90 and translocates at the nucleus to induce expression of its target genes (Fig. S1). With a single DEX application (1 μM), our pAP1::AP1-GR ap1-1 cal-5 line showed highly synchronized flower development until anthesis (Fig. 1a). When plants were grown under optimal conditions (18°C, 150 μmol m−2 s−1 continuous light), inflorescences reached the fully opened flower stage (stage 13; Smyth et al., 1990) at 14 d after DEX treatment. Lipid composition was not affected by DEX treatment in the flowers of ap1-1 cal-5 in the absence of pAP1::AP1-GR (Fig. S2).

Figure 1.

Developmental stage-specific glycerolipid profiling in Arabidopsis thaliana flowers. (a) Synchronized flower development of dexamethasone (DEX)-treated pAP1::AP1-GR ap1-1 cal-5 flowers. Photographs indicate time (d) after DEX treatment (e.g. D2, 2 d after DEX treatment). Bars, 5 mm. (b) Profiling of glycerolipid classes during flower development. pAP1::AP1-GR ap1-1 cal-5 flowers were harvested at different times after DEX treatment, and total lipids were extracted and analyzed as described in the 'Materials and Methods' section. Data are means ± SD (= 4). Asterisks indicate significance (< 0.01) from the previous time point. PI, phosphatidylinositol; PE, phosphatidylethanolamine; PA, phosphatidic acid; PG, phosphatidylglycerol; PS, phosphatidylserine; SQDG, sulfoquinovosyldiacylglycerol; PC, phosphatidylcholine; MGDG, monogalactosyldiacylglycerol; DGDG, digalactosyldiacylglycerol.

Glycerolipid profiles during flower development

We performed developmental stage-specific glycerolipid profiling using this Arabidopsis pAP1::AP1-GR ap1-1 cal-5 system. The most dramatic change was in the profile of PA, which showed a transient increase (12–16 mol%) at day 2 after DEX treatment (hereafter D2), when ectopic inflorescence meristems were converted to floral meristems, followed by a rapid decrease (from 16 to 4 mol%; Figs 1b, 2). Concomitantly, the concentration of PC moderately increased at the midstage and that of MGDG increased steeply at the late stages. Other glycerolipid profiles were fairly stable, although some changes occurred at D2. The temporary concomitant PA decrease and MGDG increase suggests that PA is probably dephosphorylated by PA phosphatase to yield DAG, which is then converted to MGDG by MGDG synthase (MGD).

Figure 2.

Profiling of molecular species of glycerolipid classes of Arabidopsis thaliana flowers in Fig. 1. pAP1::AP1-GR ap1-1 cal-5 flowers were harvested at different times after dexamethasone (DEX) treatment (e.g. D2, 2 d after DEX treatment), and total lipid was extracted and analyzed as described in the 'Materials and Methods' section. Data are means ± SD (= 4). Asterisks indicate significance (< 0.01) from the previous time point. See Fig. 1(a) for developmental stages of each time after DEX treatment. PI, phosphatidylinositol; PE, phosphatidylethanolamine; PA, phosphatidic acid; PG, phosphatidylglycerol; PS, phosphatidylserine; SQDG, sulfoquinovosyldiacylglycerol; PC, phosphatidylcholine; MGDG, monogalactosyldiacylglycerol; DGDG, digalactosyldiacylglycerol.

Acyl profile of polar glycerolpid classes during flower development

To further detail the glycerolipid profiles, we analyzed different molecular species of phospholipid and nonphosphorous glycerolipid classes during flower development (Fig. 2). Here, we discuss each profile by lipid class.

PA

All of the five major species (34:3, 34:2, 36:5, 36:4 and 36:3) found in flowers showed a transient increase at D2, followed by a rapid decrease in subsequent developmental stages. The extent of the decrease differed between C34 and C36 species; the level of 36:4 and 36:3 rapidly decreased at D4-6, and that of 34:3 and 34:2 was high before decreasing after D6. This temporary high level of C16 fatty acid-containing species may cause transient compositional changes in the PA pool at D2–4, which may affect PA signaling. In particular, the profiles for 36:4 and 36:3 at D0–4 coincided well with those of PC and PE, so the transient increase of these PA species may be derived from that of PC or PE without remodeling acyl moieties, that is either by phospholipase D (PLD) or nonspecific phospholipase C (NPC) together with diacylglycerol kinase (DGK). It should be noted that all phospholipid classes increased 36:3 and 36:4 composition, whereas 36:6 in galactolipids decreased. This suggests high recycling of phospholipids via PA but limited synthesis of galactolipids at this time.

PC

Levels of two C34 species (34:3 and 34:2) remained stable, but that of C36 species increased gradually at the later stages. Because the PC class contains mostly C36 species, at the later stages, it is enriched with polyunsaturated C18-containing fatty acid.

PE

In contrast to PC, the PE class exhibits mostly C34 species. The profile of the major species resembled that of the PC class after D6, with a difference found at earlier stages: C34 species retained similar levels from D0 to D2, but 36:4 and 36:3 showed a transient increase at D2, as did the corresponding species in the PC class. This transient increase coincided with the profile of PA, which suggests that the PLD pathway may contribute to the PA increase at D2.

PG

We found a differential profile among the five molecular species of PG analyzed (32:1, 32:0, 34:4, 34:3 and 34:2): a concomitant decrease in 32:0 and increase in 34:4. Because 32:0 represents the de novo synthesis of PG and 34:4 corresponds to 18:3/t16:1-PG, a representative plastidic species, this profile suggests compositional changes of PG within the plastids by the action of plastidic fatty acid desaturases (FAD4, 6, 7 and 8) or decreased concentration of extraplastidic PG (e.g. mitochondrial PG), which is abundant in 16:0 (Li-Beisson et al., 2010).

PI

Two major PI species (34:3 and 34:2) decreased during Arabidopsis flower development. This decrease can be associated with repression of PI biosynthesis or stimulation of phosphoinositide biosynthesis that uses PI as an initial substrate. During flower development, both PIS1 and PIS2 showed stable transcript abundances (Fig. 3d); the genes are responsible for PI biosynthesis (Collin et al., 1999; Xue et al., 2000; Justin et al., 2003; Lofke et al., 2008). The finding suggests the utilization of PI for phosphoinositide metabolism (described further in Figs 4, 5).

Figure 3.

Gene expression profiles of characterized or putative glycerolipid biosynthetic genes during flower development of Arabidopsis thaliana. Data are means ± SD fold change of transcript abundances relative to that at D0, averaged by triple biological replicates. Asterisks indicate significance (< 0.01) from the previous time point. (a) Phosphatidylserine (PS) biosynthetic genes; (b) phosphatidylcholine (PC) and phosphatidylethanolamine (PE) biosynthetic genes except choline kinases; (c) choline kinases; (d) phosphatidylglycerol (PG) and phosphatidylinositol (PI) biosynthetic genes; (e) nonphosphorous glycerolipid biosynthetic genes; (f) glycerol 3-phosphate acyltransferase (GPAT) genes except GPAT2 and GPAT5, (g) GPAT2 and GPAT5 genes; (h) lysophosphatidic acid acyltransferase (LPAT) genes; (i) phosphatidic acid phosphatase genes including lipid phosphate phosphatase (LPP) and phosphatidate phosphohydrolase (PAH); (j) nonspecific phospholipase C (NPC) genes; (k) phospholipase D (PLD) genes; (l) DGK4 and DGK6 genes. Note that the y-axis is logarithmic in (g) and (l). DGK, diacylglycerol kinase.

Figure 4.

Profiling of phosphatidylinositol 4-phosphate (PI4P) and phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2) during flower development of Arabidopsis thaliana. pAP1::AP1-GR ap1-1 cal-5 flowers were harvested at different times after dexamethasone (DEX) treatment (e.g. D2, 2 d after DEX treatment), and total lipid was extracted and analyzed as described in the 'Materials and Methods' section. Data are means ± SD (= 4). Asterisks indicate significance (< 0.01) from the previous time point. See Fig. 1 for developmental stages of each time after DEX treatment.

Figure 5.

Gene expression profiles of characterized or putative phosphoinositide kinases and phosphoinositide-specific phospholipase C (PIPLC) genes during flower development of Arabidopsis thaliana. Data are means ± SD fold change of transcript abundance relative to that at D0, averaged by triple biological replicates. Asterisks indicate significance (< 0.01) from the previous time point. (a) Type III PI 4-kinase (PI4K) genes; (b) type II PI4K genes; (c) PI4P 5-kinase (PIP5K) genes and PI4P kinase (PIPK) genes; (d) PIPLC genes except PIPLC6; (e) PIPLC6 gene. See Fig. 1(a) for developmental stages of each time after dexamethasone (DEX) treatment.

PS

In agreement with the previous analysis that flower PS contains very long-chain fatty acids (VLCFAs) ranging from C20 to C24 (Yamaoka et al., 2011), we detected C38 and C40 species in addition to C34 and C36 species. Although the total PS concentration was stable during flower development, we observed a slight decrease for species containing VLCFAs. Consistent with PC and PE, 36:4 and 36:3 showed a transient increase at D2, which suggests an involvement of these PS species in PA metabolism.

MGDG

Monogalactosyldiacylglycerol content was markedly increased at the later stages of flower development (Fig. 1b). Of the two major species, 34:6, representing prokaryotic MGDG (18:3/16:3-MGDG), showed a significant increase starting at D6–8 (Fig. 2), while the increase in 36:6 species (18:3/18:3-MGDG) was not obvious. Because MGD1 catalyzes the prokaryotic pathway of MGDG biosynthesis and contributes critically to the MGDG concentrations (Kobayashi et al., 2007), it suggests that MGD1 mainly contributes to the increase in MGDG at this stage.

DGDG

We observed only a slight increase in the level of DGDG species (Fig. 2), despite the marked increase in MGDG. This finding suggests that DGDG biosynthesis is not much influenced by increased MGDG during flower development.

Sulfoquinovosyldiacylglycerol (SQDG)

Major SQDG species (34:3, 34:2 and 36:6) showed a transient decrease during the midstage of flower development. The later increase in 34:3 species corresponds with increases in 34:6-MGDG and 34:4-PG species in stimulation of plastidic lipid biosynthesis. A high level of SQDG at the initial stage of flower development is intriguing, with no further supportive data for the increase in SQDG in this developmental stage.

Transcriptional profile of genes involved in glycerolipid metabolism

To correlate the glycerolipid profiles analyzed with the expression of relevant metabolic genes, we profiled the transcript abundances of such genes by qRT-PCR using the specific primers in Table S1. We compared our qRT-PCR data with microarray dataset publicly available for the later stages of wildtype Arabidopsis flower development (Schmid et al., 2005). We found fairly good agreement between them, suggesting that our data for earlier stages of flower development also reflects that of wildtype flowers. Here, we grouped genes by the metabolic pathways of lipid class to which they contribute or probably contribute.

PS biosynthesis

Phosphatidylserine is synthesized from PE by PSS1 and metabolized to PE by PS decarboxylase (PSD). PSS1 is the predominant isogene in charge of PS biosynthesis because gene knockout leaves no detectable PS and confers severe growth defect and microspore development (Yamaoka et al., 2011). A study of three isoforms of PSD, PSD1, PSD2 and PSD3, showed that triple knockout resulted in no detectable PSD activity and reduced PE content in mitochondria, although the main contributors to the conversion of PS to PE in other organelles remain elusive (Nerlich et al., 2007). The transcriptional profile of these genes showed rather stable expression during flower development, but a decrease in PSD1 level at later stages (Fig. 3a). This finding agrees with the microarray data of wildtype flowers in which PSD1 expression is higher at stage 9 than at later stages, and increase in the level of PS-containing 34:2 and 36:4 species in later flower development (Fig. 2). The decrease in level of PS-containing VLCFA species may not be explained by PSD, although an increase in level of VLCFA species in psd1 psd2 psd3 flowers (Nerlich et al., 2007) suggests that PSD activity may be involved in the metabolic fate of VLCFA-containing PS.

PC and PE biosynthesis

Among the genes analyzed for PC and PE biosynthesis, phosphorylethanolamine N-methyltransferase2 (NMT2) showed a transient three-fold increase in level at D6 (Fig. 3b) and choline kinase1 (CK1) showed a threefold increase in level at later stages of flower development (Fig. 3c). NMT2 catalyzes the second and third steps of phosphor-base methylation (BeGora et al., 2010), and its knockout specifically affects the 34:3 species of PE. NASCArray data show a marked increase in NMT2 expression between stages 9 and 12 of flower development. Because only this species showed a transient increase in level at D6 within PE molecular species, NMT2 may be involved in this temporary change in PE quality at the midstage of flower development. However, this change is not critical for plant viability because the nmt2 mutant showed no reproductive defect (BeGora et al., 2010). Functional study of CK1 has not been reported; however, the increase in its level coincides with the increase in level of C36 species of PC (Fig. 2), for a possible involvement of CK1 in this change.

Anionic phospholipid biosynthesis

Genes involved in PG and PI biosynthesis were analyzed (Fig. 3d). Cytidine diphosphate diacylglycerol (CDP-DAG) synthase (CDS) catalyzes the conversion of PA to CDP-DAG; phosphatidylglycerol synthase (PGPS) catalyzes the conversion of CDP-DAG to PG phosphate (PGP), which is subsequently converted to PG; and PI synthase (PIS) converts CDP-DAG to PI. Among these genes, only CDS3 showed an increase in level, by nine-fold, at the midstage and later flower development stages (Fig. 3d). Among the five CDS genes, the activity of CDS3 was lowest in yeast, although knockout study has not been reported (Haselier et al., 2010). Although CDS4 and CDS5 are localized at chloroplasts and contribute to plastidic lipid biosynthesis (Haselier et al., 2010), an association of increased abundance of plastidic species of PG (34:4; Fig. 2) and CDS3 transcripts suggests an involvement in PG biosynthesis. PIS1 and PIS2 showed stable transcript abundances, although PI decreased during flower development (Figs 1b, 2). This finding suggests utilization of PI for phosphoinositide metabolism.

Nonphosphorous glycerolipid biosynthesis

Biosynthesis of MGDG, DGDG and SQDG occurs exclusively at plastids. A transient decrease in transcripts of SQD1 and SQD2 (Fig. 3e) agrees well with SQDG profiles at the early stages of flower development (Fig. 2), which suggests that this profile is probably controlled transcriptionally. DGD1 and DGD2 showed a stable expression pattern, which agrees well with DGDG profiles. MGD2 showed a sixfold increase in level at later stages, which agrees with previous studies showing flower-specific expression of MGD2 and strong promoter MGD2-GUS staining at stamens and developing pollen tubes (Awai et al., 2001; Kobayashi et al., 2004). However, knockout of MGD2 did not affect flower development, although the MGDG concentration in the mgd2 mutant flower has not been reported (Kobayashi et al., 2009). Because the concentrations of prokaryotic MGDG were increased exclusively during development (Fig. 2), this profile indicates that MGD1 mainly contributed to this increase. Contribution of MGD1 to the reproductive process remains elusive, because the knockout mutant of MGD1 cannot reach this developmental stage owing to a severe growth defect at germination (Kobayashi et al., 2007). Active expression of MGD1 at stamens was shown by GUS staining (Kobayashi et al., 2004).

Acyltransferases in the Kennedy pathway

Glycerol 3-phosphate acyltransferase catalyzes the initial step of de novo glycerolipid biosynthesis, and hence serves as a committed step of glycerolipid biosynthesis. The sole isoform of plastidic GPAT (ACT1) showed a stable expression pattern during flower development, but several extraplastidic GPATs showed increased levels at mid- and/or late stages (Fig. 3f). GPAT2 and 7 showed transient increases in level at D6–8, by 20-fold and threefold, respectively (Fig. 3f,g). In the later stages, GPAT1, 3, 5 and 6 showed marked increases in level, by five-, four-, eight- and 140-fold, respectively (Fig. 3f,g). An increase in GPAT1 transcripts at the later stage agrees with previous findings that GPAT1 is important in tapetum and anther development (Zheng et al., 2003; Yang et al., 2012). Which isoforms are the main extraplastidic GPATs is still unclear (Yang et al., 2012). Moreover, some are involved in surface lipid biosynthesis. For example, GPAT5 is involved in suberin biosynthesis and GPAT6 is a bifunctional enzyme coding for phosphatase activity to produce monoacylglycerol (Beisson et al., 2007; Li-Beisson et al., 2009). Surface lipid biosynthesis may be stimulated in the later stages, but we did not profile these in this study.

Lysophosphatidic acid acyltransferase catalyzes the second acylation to lysophosphatidic acid (LPA) to produce PA. The only plastidic isoform, LPAT1, showed stable expression, as did ACT1 (Fig. 3f,h). Among four extraplastidic LPAT-coding genes, LPAT3 showed a transient increase in level at mid- and later stages (Fig. 3h). This profile agrees with the initiation and maturation of pollen development, which is consistent with the study of LPAT3 suggesting roles in pollen development (Kim et al., 2005). The profile of PA showed a continuous decrease at this stage (Fig. 2), which suggests that PA synthesized by LPAT3 can be rapidly converted to other glycerolipids or that LPAT3 specifically produces PA in pollen that cannot be detected by harvesting entire flowers for lipid extraction.

PA phosphatases (PAPs)

Arabidopsis has two types of PAPs, a membrane-bound type named lipid phosphate phosphatase (LPP) and a soluble type termed phosphatidate phosphohydrolase (PAH; Nakamura et al., 2007, 2009b). Despite a continuous decrease in PA during flower development (Fig. 2), most PAP genes showed stable expression profiles (Fig. 3i). LPPα4 showed an 18-fold increase in level at D6, which suggests involvement in reproductive organ development. However, knockout of LPPγ, which showed a stable expression pattern, has a lethal effect, probably the result of a male gametophytic defect (Nakamura et al., 2007); the knockout phenotype of LPPα4 is not known. Overall, decrease in PA during the flower development is unlikely to be controlled by the transcriptional level of PAP genes.

NPC

Nonspecific phospholipase C hydrolyzes PC and other major membrane phospholipids to produce DAG and the corresponding polar head group (Nakamura et al., 2005). Among six isoforms, NPC4 and NPC5 levels were increased eight- and 55-fold, respectively, at D6 (Fig. 3j). These NPCs are involved in phospholipid hydrolysis for galactolipid biosynthesis during phosphate starvation (Nakamura et al., 2005; Gaude et al., 2008). Fig. 2 shows an increase in galactolipid (particularly MGDG) from D6 onward, so these NPCs may contribute to this increase. Indeed, some phospholipid species showed a decrease in level when these NPCs were induced.

PLD

Phosopholipase D hydrolyzes primary phospholipids to produce PA. Of the 12 PLD isoforms in Arabidopsis, we could not design a specific primer set for PLDγ3 because of high homology with PLDγ2 and PLDγ1. As shown in Fig. 3(k), PLDζ2 was induced transiently at D8 by 4.5-fold, and PLDα2 and PLDβ1 were induced at later stages by four- to fivefold. PLDζ2 is induced by phosphate starvation and functions in parallel with NPC4 and NPC5 in galactolipid biosynthesis (Cruz-Ramirez et al., 2006). Although induction of PLDζ2 was slightly later than that of NPC4 and NPC5, these enzymes may cooperate in the increase in galactolipid biosynthesis in the later stages of flower development. Although PLDβ1 is involved in the defense response (Zhao et al., 2013), the roles of PLDα2 and PLDβ1 in flowers are unknown. These PLDs may contribute to galactolipid synthesis at later stages or function in local PA signaling for floral maturation.

DGK

Diacylglycerol kinase phosphorylates DAG to produce PA (Katagiri et al., 1996). Among seven DGK isoforms in Arabidopsis, five showed stable expression (Fig. S3), although DGK4 and DGK6 showed an increase, by 38- and 800-fold, respectively, at D6 (Fig. 3l). Induction of DGK anticipates PA production, although we found no increase in PA at this developmental stage (Figs 2, 3). Neither DGK4 nor DGK6 have been functionally characterized. Because high functional redundancy is reported among DGK isoforms, these DGKs may work together for PA signaling in addition to the PLD above in the midstage of flower development.

Phosphoinositide profiles during flower development

The lipidomic profiling of PI showed a continuous decrease in expression throughout Arabidopsis flower development (Figs 1b, 2), although genes involved in PI synthesis (PIS1 and PIS2) showed stable expression (Fig. 3d). Phosphoinositide metabolism involving PI as an initial substrate may be stimulated during flower development. The major pathway of phosphoinositide metabolism is phosphorylation of PI by PI 4-kinase (PI4K) to produce PI4 phosphate (PI4P), which is further phosphorylated by PI phosphates 5-kinase (PIP5K) to produce PI(4,5)P2. Eventually, this bis-phosphorylated substrate is hydrolyzed by phosphoinositide-specific phospholipase C (PIPLC) to produce inositol 1,4,5-trisphosphate (IP3) and DAG (Boss & Im, 2012). IP3 and its phosphorylated derivatives trigger intracellular signaling such as Ca2+ efflux (Im et al., 2010), whereas DAG is converted to PA by DGK for a distinct signaling function (Arisz et al., 2009). In this context, we profiled two phosphoinositides, PI4P and PI(4,5)P2, according to the stages shown in Fig. 1(a). PI4P level showed a slight transient increase at D4–6, and PI(4,5)P2 showed a clear transient increase at D10 (Fig. 4). Thus, phosphoinositide metabolism may be induced at the midstage of flower development, and a decrease in PI concentration is likely for stimulation of phosphoinositide metabolism.

Gene expression profiles of phosphoinositide metabolic genes during flower development

To investigate whether the transient increase in levels of PI4P and PI(4,5)P2 coincides with induced gene expression involved in the metabolism of these phosphoinositides, we used qRT-PCR analysis for transcript profiling of 13 PI4K, nine PIP5K, two PIPK and nine PIPLC isoforms. We detected no transcript of PI4Kγ10 in any of the floral samples analyzed, so it was excluded from the study. Interestingly, some of the genes were induced concomitantly at D8 (Fig. 5), almost when PI(4,5)P2 level increases significantly during flower development. These genes include PI4Kα2, PI4Kγ1, PI4Kγ7, PI4Kγ8 and PI4Kγ9 for PI4K-coding genes; PIP5K3 and PIPK11 for PIPK-coding genes; and PIPLC1, PIPLC3, PIPLC5, PIPLC6 and PIPLC7 for PIPLC-coding genes. Involvement of kinases in flower development is partially known. For example, PIPK11 is one of the 150 genes specifically expressed in pollen, and is important in pollen tube function (Becker et al., 2003; Ischebeck et al., 2011). It is intriguing that PIP5K3 expression was found to be transiently increased during flower development, as this enzyme has previously been reported to be only expressed in roots, specifically root hairs (Kusano et al., 2008; Stenzel et al., 2008). This suggests a possible function of PIP5K3 in pollen tubes, which are highly similar to root hairs in structure. Involvement of PI4Kγ subfamily in phosphoinositide metabolism remains elusive as these enzymes have not been shown to harbor PI4K activity, but rather seem to be protein kinases (Galvao et al., 2008; Liu et al., 2013). By contrast, involvement of specific PIPLC isoforms in flower development is still unknown in Arabidopsis, although these isoforms encode functional enzymes (Hirayama et al., 1995; Sanchez & Chua, 2001; Xu et al., 2004). The simultaneous increase in the level of multiple genes at D8 suggests a coordinated transcriptional control that achieves a transient increase in PI(4,5)P2 level for a signaling function possibly required in flower development.

Flower developmental defect found in the pah1 pah2 mutant

To demonstrate the function of galactolipids, which changes during flower development, we observed the flower phenotype of lipid biosynthetic gene mutants. As shown in Figs 1(b) and 2, PA rapidly decreased in the early stages of flower development. Because the pah1 pah2 mutant was previously reported to be deficient in major PA phosphatase activity and to retain higher PA concentrations (Nakamura et al., 2009b), we observed flower phenotype. We noted that pah1 pah2 flowers showed weak terminal flower phenotypes (Fig. 6a), which is caused by precocious termination of inflorescence meristem activity (Shannon & Meeks-Wagner, 1991; Alvarez et al., 1992). In addition, fusion of the floral organ was occasionally observed. Fig. 6(b) showed a chimeric organ between petals and stamens in addition to four petals. A magnified observation by scanning electron microscopy revealed that petal-like tissue was attached to the anther (Fig. 6c). Fused pistils (Fig. 6d) or fused stamens (Fig. 6e) were often found as well. Furthermore, abnormal phyllotaxis was observed in the inflorescence (Fig. 6f). We found significantly increased PA concentrations in flowers of pah1 pah2 (Fig. 6g). Organ fusion and abnormal phyllotaxis are commonly found in mutants defective in organ boundary formation (Aida & Tasaka, 2006). Together with terminal flower formation, our observation suggests that rapid decrease of PA in the early stage of flower development may facilitate proper organogenesis through meristem maintenance and organ boundary formation.

Figure 6.

Flower developmental defect in pah1 pah2 mutant of Arabidopsis thaliana. (a) A terminal flower in the pah1 pah2 mutant. Bar, 5 mm. (b) A flower producing a chimeric organ between petals and stamens (indicated by red arrow) in addition to four petals. Bar, 1 mm. (c) Scanning electron microscopy image of a chimeric organ between petals and stamens found in the pah1 pah2 flower. An, anther; fl, filament; pe, petal; po, pollen. Bar, 100 μm. (d) Fused pistils in pah1 pah2. Bar, 1 mm. (e) A pah1 pah2 flower containing fused stamen filaments indicated by red arrow. Bar, 500 μm. (f) Abnormal flower phyllotaxis found in the inflorescence of pah1 pah2. Unusual positioning of pedicels is indicated by red arrows. Bar, 10 mm. (g) Increased phosphatidic acid (PA) concentration in floral buds of pah1 pah2. Data are means ± SD (= 4). WT, wildtype.

Distinct glycerolipid metabolism stimulated at different stages of flower development

By using a transgenic tool to synchronize flower development, our concurrent profiling of glycerolipids and relevant metabolic gene expression during flower development has highlighted the stimulation of distinct metabolic pathways at different developmental stages. Here, to summarize our findings, we separated the 2 wk of flower development into three stages: early stage (D0–4), midstage (D6–8) and late stage (D10–14; Fig. 7).

Figure 7.

Glycerolipid metabolism suggested to be stimulated at the early stage, midstage and late stage of flower development in Arabidopsis thaliana. Increased glycerolipid classes at each stage are highlighted with an orange background. Genes with expression induced at the respective stages are shown in blue at the corresponding metabolic steps at which these gene products catalyze or are likely to catalyze. CDP-DAG, cytidin diphospho-diacylglycerol; CDS, cytidin diphospho-diacylglycerol synthase; Cho, choline; CK, choline kinase; DAG, sn-1,2-diacylglycerol; DGK, diacylglycerol kinase; G3P, glycerol 3-phosphate; GPAT, glycerol 3-phosphate acyltransferase; LPA, lysophosphatidic acid; LPAT, lysophosphatidic acid acyltransferase; LPP, lipid phosphate phosphatase; MGD, monogalactosyldiacylglycerol synthase; MGDG, monogalactosyldiacylglycerol; NMT, phosphor-base N-methyltransferase; NPC, nonspecific phospholipase C; P-Etn, phosphorylethanolamine; P-Cho, phosphorylcholine; PA, phosphatidic acid; PAH, phosphatidic acid phosphohydrolase; PC, phosphatidylcholine; PI, phosphatidylinositol; PI4K, phosphatidylinositol 4-kinase; PI4P, phosphatidylinositol 4-phosphate; PI(4,5)P2, phosphatidylinositol 4,5-bisphosphate; PIPK, phosphatidylinositol 4-phosphate 5-kinase; PIPLC, phosphoinositide-specific phospholipase C; PLD, phospholipase D.

Early stage

The highlight of the early stage was the high concentration of PA: a transient increase at D2 (by 17 mol%) of total polar glycerolipids, followed by a rapid decrease (Fig. 1b). This unusually high concentrations of PA were previously reported in pistils of Petunia flowers (Nakamura & Ohta, 2007). Given that PA concentrations in vegetative tissues are minimal (e.g. 7-wk-old leaves contain 0.7 mol% PA; Welti et al., 2002), this high PA content is probably established during the floral transition. Indeed, a small increase in transcripts of GPAT2 and LPAT3, encoding activity for de novo glycerolipid biosynthesis (Fig. 3f–h), suggests slight transcriptional induction of the Kennedy pathway at the early stage. We still do not know why PA concentration is so high in flowers (Nakamura & Ohta, 2007). This PA pool could be used for active signaling; however, it could also be a pool for galactolipid biosynthesis given the quantitatively equivalent changes in mol%. Here, PAP activity to hydrolyze PA may be stimulated post-transcriptionally because no PAP isoforms showed induced gene expression at this stage (Fig. 3i). PAH1 and PAH2 play major roles in PA hydrolysis (Nakamura et al., 2009b). The Saccharomyces cerevisiae ortholog of PAH1 and PAH2, Pah1p, requires phosphorylation/dephosphorylation in regulating enzyme activity and controls expression of phospholipid biosynthetic genes (O'Hara et al., 2006). Likewise, floral PA produced by PAH1 and PAH2 may be involved in gene regulations.

Midstage

The profile becomes more complex at this stage. The increase in PC can be achieved by stimulation of the PC biosynthetic pathway, as indicated by the induction of NMT2 expression (Fig. 3b). However, the induction of acyltransferase expression (GPAT2, GPAT7 and LPAT3; Fig. 3f,g) implies stimulation of de novo glycerolipid biosynthesis. The induction of PLDζ2, NPC4, NPC5, LPPα4 and DGK4 (Fig. 3i–l) stimulates turnover of the PC–PA–DAG triangle, for fine-tuning the expression of these metabolites. The transient increase in levels of PI4P and PI(4,5)P2 (Fig. 5), another highlight of the midstage, is rather straightforward, being controlled transcriptionally, which is followed by simultaneous induction of a set of kinases and phospholipases. Phosphoinositide signaling may be required at the midstage, probably to coordinate the onset of reproductive organ development.

Late stage

Multiple genes are expressed at this stage, but an obvious lipid profile is the increase in prokaryotic species of MGDG (Fig. 2). Apparently, this increase involves MGD1, as knocking out of MGD1, but not MGD2 and MGD3, affects the prokaryotic species of MGDG (Kobayashi et al., 2007, 2009). Stimulation of galactolipid biosynthesis during flower development was found previously in Petunia (Nakamura et al., 2003). However, in Petunia, the increase in DGDG exceeds that of MGDG, for DGDG is the major glycolipid (Nakamura et al., 2003). The increase in MGDG concentration may be explained in part by the development of sepals that contain chloroplasts. However, because pistils in Petunia have higher galactolipid biosynthetic activity than do leaves, developing carpels in Arabidopsis flower may produce a large amount of MGDG. Alternatively, galactolipid accumulation in the developing pollen tube may explain the finding (Nakamura et al., 2009a; Botte et al., 2011).

Our concurrent profiling of glycerolipids and relevant metabolic gene expression revealed distinct metabolic pathways stimulated at different stages of flower development in Arabidopsis. The most dramatic change observed in Fig. 1(b) – a transient increase and subsequent rapid decrease in PA in the early stages; a moderate increase in PC at the midstage; and a steep increase in MGDG in the later stages – suggests that PA could be dephosphorylated to DAG by PA phosphatase, and that this is utilized as a substrate for PC and MGDG at later stages (Nakamura et al., 2009b; Eastmond et al., 2010). At the midstage, reproductive organs are massively developed, and thus these membrane lipids may be required as a constituent of such organs. It will be interesting to understand why MGDG is increased at later stages. MGDG is an essential chloroplast membrane lipid for photosynthetic function; however, its role in the later stages of flower development is unknown. Previously, MGDG biosynthetic activity in the pistils of Petunia flowers was shown to be higher than that in leaves (Nakamura et al., 2003). Pistils may contain some chloroplasts, as judged from the light green color of the organ; however, the higher MGDG biosynthetic activity in pistils suggests that MGDG is required for pistil development that is not associated with known photosynthetic function. Apart from these major glycerolipids, we found dynamic changes in the profiles of minor glycerolipids, such as phosphoinositides, even in the earlier stages. Direct binding of phosphoinositides to key proteins in meristem maintenance is known. For example, POLTERGEIST (POL) and PLL1, which are components of the CLAVATA3 (CLV3)/WUSCHEL (WUS) pathway and are essential for maintenance of meristems, bind to some phosphoinositide species for stimulation (Gagne & Clark, 2010). In addition, the Arabidopsis homolog of trithorax (ATX1), which regulates AGAMOUS (AG) required for the initiation of reproductive organ development, binds phosphatidylinositol 5-phosphate (Alvarez-Venegas et al., 2006). It is possible that the dynamic change in phosphoinositide profile is related to regulatory function in meristem fate determination. A number of functional studies are anticipated on the basis of our current profiling.

Acknowledgements

The authors thank Siou-Ting Gan for technical assistance, Temasek Life Sciences Laboratory for resources, facilities, and technical assistance, and Arabidopsis TAIR (http://arabidopsis.org) for information and materials. This work was supported by research grants to T.I. from the Temasek Life Sciences Laboratory, the National Research Foundation of Singapore, under its Competitive Research Programme (CRP Award NRFCRP001-108); a grant to Y.N. and T.I. from PRESTO, Japan Science and Technology Agency, 4-1-8 Honcho Kawaguchi, Saitama, Japan; a grant to Y.N. from the Institute of Plant and Microbial Biology, Academia Sinica, Taipei; grants to M.R.W. from the Singapore National Research Foundation under CRP Award No. 2007-04, the Biomedical Research Council of Singapore (R-183-000-211-305), the National Medical Research Council (R-183-000-224-213), and the SystemsX.ch RTD project LipidX. Y.N. was supported by the Japanese Society for the Promotion of Science (JSPS) Postdoctoral Fellowship for Research Abroad.

Ancillary