PtdIns is an important precursor for inositol-containing lipids, including polyphosphoinositides, which have multiple essential functions in eukaryotic cells. It was previously proposed that different regulatory functions of inositol-containing lipids may be performed by independent lipid pools; however, it remains unclear how such subcellular pools are established and maintained. In the present paper, a previously uncharacterized Arabidopsis gene product with similarity to the known Arabidopsis PIS (PtdIns synthase), PIS1, is shown to be an active enzyme, PIS2, capable of producing PtdIns in vitro. PIS1 and PIS2 diverged slightly in substrate preferences for CDP-DAG [cytidinediphospho-DAG (diacylglycerol)] species differing in fatty acid composition, PIS2 preferring unsaturated substrates in vitro. Transient expression of fluorescently tagged PIS1 or PIS2 in onion epidermal cells indicates localization of both enzymes in the ER (endoplasmic reticulum) and, possibly, Golgi, as was reported previously for fungal and mammalian homologues. Constitutive ectopic overexpression of PIS1 or PIS2 in Arabidopsis plants resulted in elevated levels of PtdIns in leaves. PIS2-overexpressors additionally exhibited significantly elevated levels of PtdIns(4)P and PtdIns(4,5)P2, whereas polyphosphoinositides were not elevated in plants overexpressing PIS1. In contrast, PIS1-overexpressors contained significantly elevated levels of DAG and PtdEtn (phosphatidylethanolamine), an effect not observed in plants overexpressing PIS2. Biochemical analysis of transgenic plants with regards to fatty acids associated with relevant lipids indicates that lipids increasing with PIS1 overexpression were enriched in saturated or monounsaturated fatty acids, whereas lipids increasing with PIS2 overexpression, including polyphosphoinositides, contained more unsaturated fatty acids. The results indicate that PtdIns populations originating from different PIS isoforms may enter alternative routes of metabolic conversion, possibly based on specificity and immediate metabolic context of the biosynthetic enzymes.
The complex interplay of physiological processes in eukaryotic organisms, including plants, relies in part upon the regulation by inositol-containing lipids such as polyphosphoinositides [1–4]. As regulatory functions of polyphosphoinositides require tight spatiotemporal co-ordination [5,6], it has been proposed for a range of eukaryotic model systems that alternative regulatory functions of inositol-containing lipids are performed by discrete lipid pools of different metabolic origin [5–11]. For instance, we have previously demonstrated that stress-induced and constitutive polyphosphoinositides of Arabidopsis plants differ in the composition of their respective associated fatty acids, indicating an origin of these lipid pools from precursors differing in fatty composition . So far, it is unresolved how independent pools of inositol-containing lipids can be established and maintained.
The key biosynthetic precursor of inositol-containing lipids is PtdIns, formed by PIS (PtdIns synthase, EC 22.214.171.124) catalysing the transfer of Ins on to an activated CDP-DAG [cytidinediphospho-DAG (diacylglycerol)] . PISs from different biological sources have previously been characterized, including examples from mammals [13,14], fungi  and plants [16–18]. A characteristic property of reported PIS sequences is a CDP-alcohol phosphatidyltransferase domain and the possession of four hydrophobic membrane-spanning domains, identifying PISs as membrane-integral proteins. A BLAST search (blastp) with the amino acid sequence of the Arabidopsis PIS, PIS1, against publicly accessible genomic databases indicates the presence of more than one PIS isoform in most eukaryotic organisms, including plants, fungi and mammals, suggesting that distinct PIS isoforms may contribute to the establishment of different physiological pools of inositol lipids. Besides the known PIS1 gene locus (At1g68000), the Arabidopsis genome contains at least one more gene (At4g38570, PIS2) with similarity to PIS-coding sequences, which has not been functionally characterized. Because mammalian and fungal PISs have been reported to reside at the interface of the ER (endoplasmic reticulum) and Golgi stacks [19,20], and plant PIS activity has been found associated with microsomal fractions , PtdIns production is thought to take place in the ER and/or the Golgi, although precise localization studies have not been performed on plant enzymes. From multiple reports of polyphosphoinositide-formation in cellular compartments other than the ER or Golgi [4,9,21–24], the question arises as to how PtdIns is channelled to those other cellular locations in order to provide the PtdIns necessary for polyphosphoinositide production. Several mechanisms have been proposed by which PtdIns may be distributed throughout the cell, and possible explanations include the action of PtdIns-specific lipid-transfer proteins [25,26], a distribution by targeted vesicle delivery, or a combination of both.
PtdIns can be converted into other lipid intermediates by several alternative pathways. The inositol headgroup can be successively phosphorylated at the D3, D4 or D5 positions by PI kinases (PtdIns kinases) and PIP kinases (PtdIns phosphate kinases), giving rise to a variety of polyphosphoinositides, many of which with reported regulatory effects on physiology [1,3,5]. Alternatively, PtdIns can be degraded by PLD (phospholipase D) to PtdOH (phosphatidic acid), which is a central biosynthetic intermediate of phospholipid metabolism and can be transformed into any other glycerophospholipid [27–29]. In addition to the pathways mentioned, a number of other lipid-modifying enzymes, including the phospholipases PLC (phospholipase C), PLA (phospholipase A) 1 and PLA2, and inositolphosphoceramide synthase, may convert PtdIns into a variety of other products . In this context we aimed to address the question as to whether PtdIns produced by different Arabidopsis PIS isoforms may enter discernible alternative routes of further metabolic conversion.
In the present paper we show that the recombinant PIS2 protein is catalytically active as a PIS in vitro and that Arabidopsis thus contains at least two PIS isoforms. In the course of experiments for in vitro characterization, subtle preferences of PIS1 and PIS2 for saturated and unsaturated CDP-DAG substrates respectively, were observed. PIS1 and PIS2 are both localized in the ER with only very slight differences in localization between the isoforms. Increased PtdIns levels with overexpression of PIS1 or of PIS2 in transgenic Arabidopsis plants were correlated to either increased DAG and PtdEtn levels (PIS1) or to increased levels of polyphosphoinositides (PIS2). The results suggest a role for PIS isoforms in the channelling of PtdIns towards different pools of PtdIns-derived lipids.
The cDNA sequences encoding Arabidopsis PIS1 and PIS2 were amplified from cDNA prepared from flowers and siliques using the primer combinations 5′-GATCGAGTCGACATGGCTAAAAAGGAGAGAC-3′ (primer A)/5′-GATCGAGCGGCCGCTCAAGGCTTCTGCTGCTTCTCTATA-3′ and 5′-GATCGAGTCGACATGGCTAAACAGAGACCGGCG-3′ (primer B)/5′-GATCGAGCGGCCGCTCAAGGCTTCTTATGCTGTTTCTC-3′ respectively. For constructs to be C-terminally tagged by EYFP (enhanced yellow fluorescent protein), PIS1 and PIS2 cDNA fragments were amplified using the primer combinations: primer A/5′-GATCGAGCGGCCGCTAGGCTTCTGCTGCTTCTCTATA-3′ and primer B/5′-GATCGAGCGGCCGCTAGGCTTCTTATGCTGTTTCTC-3′ respectively, leaving out the stop codons and introducing one additional nucleotide base for in-frame-fusion with the tag. The cDNAs encoding the EYFP and CFP (cyan fluorescent protein) tags were amplified from plasmids carrying the authentic clones provided by Dr Martin Fulda (Göttingen), in both cases using the primer combination 5′-GATCGCGGCCGCCCATGGTGAGCAAGGGCGAG-3′/5′-GATCGATATCTTACTTGTACAGCTCGTCCATG-3′. All amplifications were carried out using Phusion High-Fidelity DNA polymerase (Finnzymes) and a PCR protocol of 2 min denaturing at 95 °C followed by 36 cycles of 30 s denaturing at 95 °C, 30 s annealing at 60 °C and 1 min elongation at 72 °C, with a terminal elongation step for 5 min at 72 °C. Amplicons were cloned into the pGEM-T-Easy vector (Promega) according to the manufacturer's instructions and sequenced. PIS1 and PIS2 cDNA clones were ligated as SalI/NotI fragments into the bacterial expression vector pET28b (Novagen). For plant expression, the plasmid pUC18 (Fermentas) was modified by the addition of a multiple cloning site and gateway cassette of the pENTR2b-vector (Invitrogen), yielding the plasmid pUC18-ENTR. The cDNAs encoding EYFP or CFP were introduced into pUC18-ENTR as NotI/EcoRV fragments, yielding the plasmids pUC18-ENTR-EYFP and pUC18-ENTR-CFP respectively. PIS1 and PIS2 cDNA clones were ligated as SalI/NotI fragments first into pUC18-ENTR; PIS1 and PIS2 cDNA fragments lacking the stop codon were ligated as SalI/NotI fragments into pUC18-ENTR-EYFP or into pUC18-ENTR-CFP. Subsequently, cDNAs were ligated from the pUC18-derived vectors into a pCAMBIA plasmid (pCAMBIA3300–0GC) modified to contain a Gateway cassette (Invitrogen) using Gateway technology (Invitrogen), according to the manufacturer's instructions.
Heterologous expression in Escherichia coli
PIS1- and PIS2-pET28b constructs were heterologously expressed in E. coli C43(DE3) cells optimized for the accumulation of integral membrane proteins . Cells were transformed by heat-shock treatment . Expression cultures (400 ml) were grown in Luria–Bertani medium  in 1 litre Erlenmeyer flasks at 37 °C with shaking at 200 rev./min. Protein expression was induced by addition of 1 mM isopropyl-β-D-thiogalactoside at an attenuance (600 nm) of 0.8. Cultures were harvested 4 h after induction, and cells were disrupted on ice by three 30 s pulses of ultrasound using a Sonifier Cell disruptor B15 (Branson) at 50% performance. The homogenate was centrifuged for 10 min at 840 g at 4 °C, and the cleared supernatants were subjected to ultracentrifugation at 24000 rev./min in a Beckmann SW28 rotor for 60 min at 4 °C. Membrane pellets were resuspended in 1 ml of 20 mM Tris/HCl, 1 mM EDTA, 8 mM 2-mercaptoethanol, pH 8.0, containing 20% (v/v) glycerol. Protein contents were determined by the Bradford method (Bio-Rad) with BSA as a standard .
Synthesis of CDP-DAG substrates
CDP-DAG substrates that were not commercially available were synthesized from PtdOH by chemical modification via a lipid-morpholidate as previously described . PtdOH-species used for chemical conversion were obtained from Avanti (1,2-dipalmitoyl-glycerin-3-phosphate, 1,2-dioleoyl-glycerin-3-phosphate and 1,2-dilinoleoyl-glycerin-3-phosphate) or Sigma (1,2-diheptadecanoyl-glycerin-3-phosphate). In addition to synthetic lipids, PtdOH was extracted from 50 g of green leaf tissue , purified by TLC and used for conversion into CDP-DAG. CDP-DAG species obtained co-migrated with authentic CDP-DAG standards (Avanti) and were purified by TLC. Samples were dissolved in CHCl3, dried under streaming nitrogen and diluted in CHCl3 to a lipid content of approx. 0.3 mg/ml.
PIS activity tests
The activity of recombinant PIS1 and PIS2 proteins was detected in bacterial membrane pellets by monitoring the production of PtdIns from CDP-DAG (Avanti) and non-labelled D-myo-inositol (Sigma). Before addition of soluble components, CDP-DAG was dissolved in ethanol and dried into glass reaction tubes under streaming nitrogen. The assay mixture contained 50 mM Tris/HCl, pH 8.0, 20 mM MgCl2, 1.5 mM D-myo-inositol, 0.9 mM CDP-DAG and 1% (v/v) Triton X-100 in a reaction volume of 200 μl, as previously described . In experiments to determine the fatty acid compositions of CDP-DAG substrates and PtdIns formed, CDP-DAG substrates were used at a concentration of 0.2 mM. Reactions were started by the addition of 20 μg of bacterial membranes and incubated for 30 min at room temperature (22 °C). Lipid reaction products were extracted , and the organic phase was washed with 1% (w/v) NaCl solution and dried under streaming nitrogen. Samples were subjected to TLC using a developing solvent containing methylacetate, 2-propanol, CHCl3, CH3OH and 0.25% (w/v) aqueous KCl solution at volume ratios of 25:25:25:10:9 . Lipids were visualized by submerging dried chromatography plates in an aqueous solution of 10% (w/v) CuSO4 (Sigma) containing 8% (v/v) H3PO4 (Sigma) and subsequent heating to 180 °C. Lipid products were identified according to co-migration with authentic standards (Avanti). Alternatively to destructive visualization, lipids were reisolated from TLC plates and analysed for their fatty acid composition as described below.
Lipid extraction and biochemical analysis
Plant material was ground under liquid nitrogen to a fine powder. Phospholipids were extracted from powdered plant material by using an acidic extraction protocol . Lipids were separated by TLC on silica gel plates (Merck) using developing solvents for optimal resolution: for phosphoinositides and PtdOH, CHCl3/CH3OH/NH4OH/H2O (57:50:4:11 by vol.) ; for PtdEtn, acetone/toluol/H2O (91:30:7 by vol.) ; for isolating PtdIns, the same developing solvent was used as described for the PIS assay. Lanes with lipid standards (5 μg) run in parallel to biological samples or to lipids formed in in vitro assays were cut and lipids were visualized as described above. Unstained lipids that were located on the remaining parts of the TLC plates according to standard migration, were scraped, redissolved in their respective developing solvents and dried under nitrogen flow. Lipids were transmethylated , and fatty acid methyl esters were dissolved in acetonitrile and analysed using a GC6890 gas chromatograph with flame ionization detection (Agilent) fitted with a 30 m×250 μm DB-23 capillary column (Agilent). Helium flowed as a carrier gas at 1 ml·min−1. Samples were injected at 220 °C. After 1 min at 150 °C, the oven temperature was raised to 200 °C at a rate of 8 °C·min−1, then to 250 °C at 25 °C·min−1, and then kept at 250 °C for 6 min. Fatty acids were identified according to authentic standards and by their characteristic mass spectrometric fragmentation patterns (results not shown), and quantified according to internal tri-pentadecanoic acid standards of known concentration. Owing to limiting material in samples representing isolated minor lipids, fatty acids of low abundance may be absent from fatty acid patterns.
Transient gene expression by particle bombardment
Epidermal sections (1 mm) of quartered onion (Allium cepa) bulbs were placed on moist Whatman paper and were transformed by bombardment with plasmid-coated 1 μm gold particles using a helium-driven particle accelerator (PDS-1000/He; Bio-Rad) with 1350 psi (1 lbf/in2 = 6.9 kPa) rupture discs at a vacuum of 28 inches of mercury. Gold-particles (1.25 mg) were coated with 3–7 μg of plasmid DNA of the pCAMBIA-vectors also used for plant transformation. After bombardment, onion sections were incubated in covered petri dishes for 24 h at room temperature in a closed moist styrofoam box.
Images of onion epidermal peels were recorded using a Zeiss LSM 510 meta confocal microscope. EYFP was excited at 514 nm and imaged using an HFT 405/514/633 nm MBS (major beam splitter) and a 530–600 nm band pass filter; CFP was excited at 405 nm and imaged using an HFT 405/514/633 nm MBS and a 470–500 nm band pass filter. In co-expression experiments, CFP and EYFP were synchronously excited at 405 nm and 514 nm respectively, and imaged using an HFT 405/514/633 nm and NFT 515 nm MBSs and 470–500 nm and 530–600 nm band pass filters respectively.
Plant culture and transformation
Arabidopsis thaliana (L.), ecotype Columbia-0 (col-0) plants were grown on soil in growth chambers (York) at 22–25 °C under a regime of 8 h exposure to 130–150 μmol of photons·m−1·s−1 and 16 h darkness at approx. 60% humidity. Recombinant DNA constructs were introduced into Arabidopsis plants by Agrobacterium tumefaciens-mediated transformation, using A. tumefaciens strain EHA105  and the floral dip method . Independent transformants were selected according to resistance against aerosolic glufosinate-ammonium (Bayer). All analyses were carried out on 6-week-old plants. Plant growth was estimated according to leaf area of plants of the same age. Leaf areas were determined using Blattfläche 126.96.36.199 software (DatInf GmBH) according to the manufacturer's instructions.
Arabidopsis contains two distinct PIS isoforms capable of generating PtdIns
A search of publicly accessible databases [30,45] suggested the gene locus At4g38570 encodes a previously uncharacterized PIS isoform in Arabidopsis in addition to the known PIS1 encoded by the locus At1g68000. The amino acid sequence of the deduced PIS2 protein indicates strong similarity to other PISs from plants, fungi and mammals (Figure 1). According to published transcript-array data accessible through the Genevestigator portal , the PIS1 and PIS2 genes are ubiquitously expressed at modest levels throughout the plant, indicating no tissue-specific functions for either isoenzyme. Sequencing of independently cloned full length PIS2 cDNAs consistently indicated a mismatch against the published Arabidopsis genomic sequence at position 444 (guanosine instead of cytosine), resulting in the amino-acid exchange C148W, as indicated (Figure 1). As Cys148 is conserved in PISs from plants and fungi, but not in mammalian enzymes (Figure 1), at present the relevance of the polymorphism is unclear. In order to test the functionality of the PIS2 gene product, the cDNA was recombinantly expressed in E. coli. Production of the correct recombinant proteins was verified by immunodetection of the poly-His-tagged proteins, and PIS1 and PIS2 proteins were present only in pellet fractions representing insoluble membrane-proteins (results not shown). Protein levels of PIS1 and PIS2 preparations tested were adjusted according to immunodetection. Recombinant PIS2 (C148W) protein exhibited substantial activity, producing PtdIns from CDP-DAG and D-myo-inositol substrates (Figure 2). In side-by-side expression experiments, PIS2 activity was comparable with that of the previously reported PIS1 (Figure 2). Affinity-purification of poly-His-tagged recombinant enzymes was not successful, probably because of the inherent difficulties with reconstituting membrane-integral enzymes in an active form, and, therefore, no specific activity was calculated. Even so, the results presented establish PIS2 as an active PIS isoform in Arabidopsis.
Alignment of deduced amino acid sequences of PISs from different biological origins
PIS activity of recombinant Arabidopsis PIS1 and PIS2 proteins in vitro
PIS1 and PIS2 proteins have preferences for different CDP-DAG fatty acyl species
In order to test the substrate preferences of PIS1 and PIS2, catalytic activity was assayed in the presence of CDP-DAG species differing in fatty acid composition (Figure 3). CDP-DAG substrates tested included 1,2-dipalmitoyl-CDP-DAG; 1,2-diheptadecanoyl-CDP-DAG; 1,2-dioleoyl-CDP-DAG and 1,2-dilinoleoyl-CDP-DAG. In a first experiment, PIS activity was assayed under identical conditions against substrates presented individually (Figure 3A). Catalytic activity was consistently higher with substrates containing C18-fatty acids than with those containing medium-chain fatty acids. For PIS1, the highest activity was determined using 1,2-dioleoyl-CDP-DAG, whereas PIS2 was more active with 1,2-dilinoleoyl-CDP-DAG. In order to assess substrate preferences in a competition assay, activity tests were performed against a mixture of all four available synthetic CDP-DAG substrates (Figure 3B). The preference of PIS1 and PIS2 for substrates containing C18-fatty acids was confirmed. In the competitive assay, PIS1 preferentially converted 1,2-dioleoyl-CDP-DAG, whereas PIS2 preferred 1,2-dilinoleoyl-CDP-DAG. All CDP-DAG species tested so far were of synthetic origin and symmetrical with regard to the associated fatty acids. As biological lipids are not usually symmetrical and contain different fatty acids, CDP-DAG synthesized from PtdOH isolated from Arabidopsis leaves was also tested as a substrate for PIS1 and PIS2 (Figure 3C). The pattern of fatty acids associated with the ‘biological’ CDP-DAG input mixture in these experiments resembled that of structural phospholipids of Arabidopsis leaves [11,46,47]. After in vitro conversion of the CDP-DAG by PIS1 or PIS2, the fatty acid composition of CDP-DAG not converted (CDP-DAG residue) and that of newly formed PtdIns (PtdIns product) was determined. The biological CDP-DAG mixture was converted into PtdIns with high efficiency by PIS2, whereas conversion by PIS1 was only moderate (Figure 3C). PtdIns formed by PIS1 contained mainly palmitic acid and, to a lesser degree, linoleic acid. The CDP-DAG residue after PIS1 conversion contained high proportions of linoleic acid and linolenic acids, indicating that CDP-DAG species containing these fatty acids were not effectively converted by PIS1. In contrast, PtdIns formed by PIS2 from biological CDP-DAG contained predominantly palmitic acid, linoleic acid and linolenic acid, concomitant with a severe depletion of these fatty acids in the corresponding CDP-DAG residue. Together, the in vitro data suggest a preference of recombinant PIS2 for polyunsaturated CDP-DAG species.
Substrate preferences of recombinant PIS1 and PIS2 for different CDP-DAG species
PIS1 and PIS2 proteins localize in the ER of onion epidermal cells
In a previous study, plant PIS activity was reported to associate with microsomal fractions , and because mammalian and fungal PISs have been reported to reside in both ER and Golgi membranes [19,20], experiments were initiated to test the subcellular localization of Arabidopsis PIS1 and PIS2 proteins. In-frame fusions of PIS1 or PIS2 with C-terminally attached EYFP or CFP tags were transiently expressed in onion epidermal cells, and the fluorescence distribution of the PIS1–EYFP and PIS2–CFP proteins were synchronously monitored by laser-scanning confocal microscopy (Figure 4). PIS1–EYFP and PIS2–CFP both exhibited clear ER localization (Figure 4A). Differences in the localization of PIS1–EYFP and PIS2–CFP were restricted to small punctate structures of limited occurrence (Figure 4A, arrowheads) where PIS1 was observed in the centre and PIS2 at the periphery (Figure 4B). The distribution of PIS1 and PIS2 was evaluated in relation to that of the Golgi-associated signal peptide of the rat sialyl-S-transferase Sial . When PIS1–EYFP or PIS2–EYFP were coexpressed with CFP–Sial, a limited degree of overlap was observed, as shown for PIS1–EYFP in Figure 4(C). The combined localization data indicate that PIS1 and PIS2 are present in the ER, including tubular domains in close proximity to Golgi particles, and exhibit slight but reproducible differences in localization patterns.
Subcellular distribution of Arabidopsis PIS1 and PIS2 transiently expressed in onion epidermal cells
Constitutive ectopic overexpression of PIS1 or PIS2 in Arabidopsis affects plant growth
In order to test whether the metabolism of polyphosphoinositides or other phospholipids could be disturbed by overproduction of PtdIns in plants, PIS1 and PIS2 were individually overexpressed in transgenic Arabidopsis plants under the constitutive cauliflower mosaic virus 35S promoter. Successful transformation was confirmed by PCR-based genotyping of glufosinate-ammonium-resistant transformants and detection of the 35S-PIS1 or 35S-PIS2 transgene (results not shown). The T2-progeny of resistant plants exhibited severely diminished growth (Figure 5A), indicating a substantial influence of transgene expression on physiology. Evaluation of plant size by automated leaf-area detection indicates a significant difference (P<0.01) in plant size compared with wild-type plants grown in parallel (Figure 5B).
Influence of constitutive ectopic PIS overexpression on Arabidopsis growth
Increased formation of PtdIns and PtdOH with ectopic overexpression of PIS1 or PIS2 in Arabidopsis plants
Of the transgenic Arabidopsis lines expressing either PIS1 or PIS2, three lines each were chosen for detailed biochemical analysis of leaf lipids. The first question asked was whether PIS overexpression was correlated with increased accumulation of PtdIns in the transgenic plants. GC-based quantification of fatty acids associated with PtdIns  indicated that PtdIns levels were significantly (P<0.01) increased in all PIS-overexpressing lines tested (Figure 6A). This result contrasts somewhat with previous results indicating that the levels of PtdIns are tightly controlled and kept at a constant level with either metabolic disturbance or with applied stress treatments . When the levels of PtdOH as a potential breakdown-product of PtdIns were tested in the PIS-overexpressing lines, PtdOH-levels were also found to be significantly increased (P<0.01) over those detected in wild-type plants (Figure 6B), suggesting the action of regulatory mechanisms to limit PtdIns accumulation.
Increased levels of PtdIns and PtdOH in Arabidopsis plants constitutively overexpressing PIS1 or PIS2
Alternative metabolic fates of PtdIns produced by PIS1 or PIS2
When other lipids were included in the analysis of PIS1- or PIS2-overexpressors, it was found that PIS1 and PIS2 exerted different effects on downstream metabolites (Figure 7). Whereas overexpression of PIS1 resulted in a significant increase in the levels of DAG and PtdEtn (phosphatidylethanolamine) in all cases, no such increases were observed with expression of PIS2 (Figures 7A and 7B). In contrast, the levels of the polyphosphoinositides, PtdIns4P and PtdIns(4,5)P2, were significantly (P<0.01) increased in plants overexpressing PIS2, but not in those overexpressing PIS1 (Figures 7C and 7D). The lipid analyses performed indicate that, despite equivalent accumulation of PtdIns with overexpression of PIS1 or PIS2 (Figure 6A), different lipid metabolites were affected by PtdIns overproduction through each PIS isoform.
Differential effects of PIS1 or PIS2 overexpression on lipid metabolism
Lipid-backbone fatty acid composition indicates metabolic origin of downstream metabolites
In order to provide evidence for the metabolic origin of phospholipids exhibiting altered levels correlated with PIS1 or PIS2 overexpression (Figure 8), the flow of intermediates into alternative pathways must be traced. As [3H]-myo-inositol headgroup labelling is not feasible at a scale required for the detection of polyphosphoinositides in intact Arabidopsis tissue, and because the label would be lost from the part of the PtdIns molecule that enters pathways for conversion into other phospholipids including PtdEtn, metabolites were traced according to the fatty acid composition of their glycerolipid backbones, as described previously . The fatty acid composition of relevant phospholipids isolated from wild-type plants (Figure 8A) indicates a high proportion (50–60 mol%) of polyunsaturated fatty acids was associated with PtdIns, PtdOH and PtdEtn, whereas the polyphosphoinositides, PtdIns4P and PtdIns(4,5)P2, contained more saturated and monounsaturated fatty acids, as previously reported . With overexpression of PIS1, the fatty acid composition of PtdIns, PtdOH, DAG and PtdEtn shifted to contain less (30–45 mol%) polyunsaturated fatty acids and a greater proportion (55–65 mol%) of saturated and monounsaturated fatty acids (Figure 8B), whereas the fatty acid composition of polyphosphoinositides was not obviously altered. In contrast, overexpression of PIS2 resulted in an increased proportion of 55–75 mol% polyunsaturated fatty acids associated with all lipids tested, most remarkably in PtdIns4P and PtdIns(4,5)P2 (Figure 8C). The proportion of polyunsaturated fatty acids associated with PtdIns4P and PtdIns(4,5)P2 changed notably from only 25 and 15 mol% of the associated fatty acids in wild-type plants respectively, to 50 and 45 mol% respectively, in plants overexpressing PIS2 (compare Figure 8A with 8C). The results indicate that the degree of unsaturation of fatty acids associating with phospholipids affected by PIS overexpression changes differentially for PIS1 or PIS2. Changes in the degree of unsaturation of fatty acids associated with phospholipids resulting from overexpression of PIS1 or PIS2 are summarized in Supplementary Table S1 at http://www.BiochemJ.org/bj/413/bj4130115add.htm.
Differential effects of PIS1 or PIS2 overexpression on phospholipid fatty acid composition
The present study was initiated to explore the effects of increased PtdIns levels on plant lipid metabolism as a result of PIS overexpression. In addition to a known Arabidopsis PIS, PIS1, the previously uncharacterized PIS2 protein was included in this study and was demonstrated to have PIS activity in vitro (Figure 2). Activity tests were performed using crude bacterial extracts, because E. coli does not contain endogenous PIS activity . The lack of chaperones or post-translational modifications required for correct functionality of many eukaryotic proteins expressed recombinantly in E. coli did not obviously affect the catalytic activity of PIS1 or PIS2 (Figure 2), indicating that chaperones or modifications are not essential for PIS function. The presence of an unusual number of amino acid positions (>20 out of ~220 residues) conserved in PISs either as threonine or serine residues (Figure 1), some across organismic kingdoms, may indicate the presence of phosphorylation sites possibly regulating enzyme activity or localization in vivo. A detailed investigation of catalytic activity of PIS1 and PIS2 in vitro against CDP-DAG species differing in associated fatty acyl moieties indicated that PIS1 and PIS2 differ in their substrate preferences (Figure 3). Whereas PIS1 preferred CDP-DAG substrates containing a combination of palmitic acid and linoleic acid, PIS2 was most active with CDP-DAG substrates containing the polyunsaturated linoleic acid and linolenic acid. The observation that PIS1 exhibited substantially less activity than PIS2 against the biological CDP-DAG suggests that the endogenous PIS1 substrate may not have been represented in the mixture of CDP-DAG species tested.
By monitoring the subcellular distribution of fluorescently tagged PIS1 and PIS2 proteins, ER localization is suggested for both plant enzymes (Figure 4). Both PIS1 and PIS2 associated with tubular ER structures in close proximity to Golgi particles (Figure 4C), consistent with reports of fungal and mammalian PISs which localize to the interface of ER and Golgi [19,20]. Although PIS1 and PIS2 generally co-localized in the ER, punctate ER patterns were regularly observed to which both PIS1 and PIS2 associated; however, with subtle differences in localization (Figure 4B). In this context it should be noted that algorithms for the prediction of subcellular targeting such as Predotar V1.03  or iPSORT  did not yield consistent targeting predictions for PIS1 and PIS2, despite substantial similarity of primary sequences (Figure 1). The interpretation of the visual information obtained by confocal microscopy was aided by the use of the CFP-tagged Golgi marker construct Sial , which is an established organellar marker for use in plants and has been used in previous studies . The results support a view of PtdIns biosynthesis at the interface of the lipid biosynthetic pathways in the ER and the secretory machinery of the Golgi. It is our working hypothesis that PISs act in subtly different metabolic contexts, and that the action of PIS1 and PIS2 may supply different pools of inositol-containing lipids with PtdIns.
Ectopic overexpression of PIS1 or the previously uncharacterized PIS2 resulted in significantly delayed growth of transgenic plants, suggesting that transgene expression was placing strain on metabolism or energy balance. It has been previously reported that the levels of PtdIns are not readily perturbed by stress treatments affecting the levels of other phospholipids, including PtdIns-derived lipids, suggesting PtdIns levels are tightly regulated . Increases in PtdIns levels in leaves of plants overexpressing PIS1 or PIS2 coincided with raised levels of PtdOH, the immediate PLD-breakdown product of PtdIns (Figure 6). The inhibitory effect of PIS overexpression on plant growth (Figure 5) may thus be due to the establishment of a futile cycle of increased PtdIns generation and breakdown and consequentially increased energy consumption. An alternative explanation is possible effects of altered polyphosphoinositide formation on signalling processes affecting plant growth or development [2,3]. In order to monitor downstream metabolic effects of increased PtdIns turnover in the transgenic plants, a range of candidate sinks for PtdIns-derived lipid backbones was analysed (Figure 7). Overexpression of PIS1 resulted in elevated DAG and PtdEtn levels in leaves of all transgenic lines tested (Figures 7A and 7B), whereas that of PIS2 consistently supported increased polyphosphoinositide production (Figures 7C and 7D). The hypothesis that PtdIns produced by PIS1 or PIS2 was channelled towards different routes of metabolic conversion was tested by tracing the glycerolipid backbones of phospholipid intermediates according to their fatty acid composition (Figure 8). The observation that PtdIns and other phospholipids increasing with PIS1 overexpression contained fatty acids with a lower degree of unsaturation, whereas PtdIns and phospholipids increasing with PIS2 overexpression contained fatty acids with a higher degree of unsaturation, suggests that the particular phospholipids may have their metabolic origin in distinct PtdIns populations formed by PIS1 or PIS2. The data are consistent with the results of the in vitro characterizations, indicating differences in substrate preferences of PIS1 and PIS2 for CDP-DAG species with particular fatty acid compositions. Influences of lipid-substrate fatty acid compositions on enzyme functionality have been previously demonstrated for other enzymes involved in the biosynthesis of inositol lipids, including invertebrate PIP kinases and PLC , and mammalian phosphoinositide phosphatases .
In contrast with the situation with PIS1, overexpression of PIS2 resulted in several-fold increased levels of PtdIns4P and PtdIns(4,5)P2 (Figures 7B and 7C), most probably as a consequence of increased substrate availability for PI kinases and PIP kinases. As the phosphorylation sequence leading to the formation of PtdIns(4,5)P2 is probably not ER-associated in plants, PtdIns originating from PIS2 activity and entering this sequence must be delivered for further metabolic conversion by a different route than PtdIns produced by PIS1. In summary, the results suggest that PIS1 and PIS2 may act in different metabolic contexts, resulting in PtdIns channelling towards different physiological pools. While it is clear that independent pools of inositol lipids are present in plant cells [8–11], the establishment of such pools may also be controlled by the distribution of lipids other than PtdIns. Future experiments will be aimed at elucidating the contribution of other phospholipid-biosynthetic enzymes in the formation of distinct lipid pools.
We thank Dr Ben Scheres (Utrecht University, The Netherlands) for the Golgi marker construct. We also thank the following individuals at Georg-August-University Göttingen: Dr Martin Fulda for plasmid DNA; Dr Oliver Einsle for the E. coli C43 (DE3) cells; Dr Andreas Wodarz for access to and Dr Michael Krahn for expert assistance with confocal microscopy; Susanne Mesters for expert plant culture; and Dr Ivo Feussner for helpful discussion. We acknowledge funding through an Emmy-Noether grant from the German Research Foundation (DFG; to I. H.).
cyan fluorescent protein
enhanced yellow fluorescent protein
main beam splitter