Arabidopsis lipins mediate eukaryotic pathway of lipid metabolism and cope critically with phosphate starvation

Abstract

Phosphate is an essential nutrient for plant viability. It is well-established that phosphate starvation triggers membrane lipid remodeling, a process that converts significant portion of phospholipids to non-phosphorus-containing galactolipids. This remodeling is mediated by either phospholipase C (PLC) or phospholipase D (PLD) in combination with phosphatidate phosphatase (PAP). Two PLC genes, NPC4 and NPC5, and PLD genes, PLDζ1 and PLDζ2, are shown to be involved in the remodeling. However, gene knockout studies show that none of them plays decisive roles in the remodeling. Thus, although this phenomenon is widely observed among plants, the key enzyme(s) responsible for the lipid remodeling in a whole plant body is unknown; therefore, the physiological significance of this conversion process has remained to be elucidated. We herein focused on PAP as a key enzyme for this adaptation, and identified Arabidopsis lipin homologs, AtPAH1 and AtPAH2, that encode the PAPs involved in galactolipid biosynthesis. Double mutant pah1pah2 plants had decreased phosphatidic acid hydrolysis, thus affecting the eukaryotic pathway of galactolipid synthesis. Upon phosphate starvation, pah1pah2 plants were severely impaired in growth and membrane lipid remodeling. These results indicate that PAH1 and PAH2 are the PAP responsible for the eukaryotic pathway of galactolipid synthesis, and the membrane lipid remodeling mediated by these two enzymes is an essential adaptation mechanism to cope with phosphate starvation.

Phosphate (Pi) is an essential nutrient for plant viability. Due to the limited availability in many soils, however, plants often suffer from phosphate deficiency (1, 2). One of major adaptation responses to Pi shortage is the membrane lipid remodeling: upon Pi deficiency, a significant portion of membrane phospholipids is replaced by non-phosphorus galactolipids and sulfolipid presumably to use phospholipids as plant inner Pi reserve (3–6).

Galactolipids such as monogalactosyldiacylglycerol (MGDG) and digalactosyldiacylglycerol (DGDG) are unique but ubiquitous lipid classes in plants. In addition, they have decisive function in plant photosynthetic growth (7, 8). In Arabidopsis, they are synthesized from diacylglycerol (DAG) by MGDG synthases (MGD) and DGDG synthases (DGD). Although galactolipid biosynthesis occurs exclusively at plastid envelopes (9), there are two metabolic pathways for DAG supply, either prokaryotic or eukaryotic pathways localized at plastids or ER, respectively (10). During Pi starvation, galactolipid synthetic pathway is up-regulated to increase DGDG levels (11–15). However, how the substrate DAG is provided is still an issue of discussion (16). Since extraplastidic phospholipids such as phosphatidylcholine (PC) or phosphatidylethanolamine (PE) are replaced by DGDG, an ER-localized eukaryotic pathway was thought to be involved in the supply of DAG.

There are two pathways to produce DAG from PC or PE, either by phospholipase C (PLC) or phospholipase D (PLD) in combination with phosphatidate phosphatase (PAP). Since non-specific phospholipase C4 (NPC4), which assumes major PLC activity during Pi starvation, does not hydrolyze phosphatidic acid (PA) (i.e., NPC4 has no detectable PAP activity), these two pathways are considered to be independent (17). We previously reported two Pi starvation-inducible PLC, NPC4 and NPC5, and showed that NPC5 is involved in the membrane lipid remodeling (17, 18). However, the remodeling was affected only partially in the npc5 mutant, possibly because the other pathway by PLD in combination with PAP might complement this defect. Although two PLD (PLDζ1 and PLDζ2) are reported to be involved in the remodeling, the function deduced by knock out mutant analysis is restricted to roots (19, 20). This suggests that the following step by PAP may be a limiting step for the remodeling (20). Because no extraplastidic PAP involved in eukaryotic pathway of galactolipid biosynthesis is known, however, further investigation was hampered to date. Thus, isolation of a key enzyme in membrane lipid remodeling was anticipated to understand the physiological significance of the widely observed membrane lipid remodeling during Pi starvation.

Here, we report Arabidopsis homologs of lipin, phosphatidate phosphohydrolase 1 and 2 (AtPAH1 and AtPAH2) which encode a type of PAP. We characterized that disruption of AtPAH1 and AtPAH2 (pah1pah2) affect the eukaryotic pathway of galactolipid biosynthesis significantly. Furthermore, pah1pah2 plants showed severe overall growth defect under Pi starvation. These results suggest that [1] AtPAH1 and AtPAH2 encode PAP involved in eukaryotic pathway of galactolipid biosynthesis and [2] the membrane lipid remodeling mediated by AtPAH1 and AtPAH2 is an essential adaptation mechanism for plants to circumvent Pi starvation.

Results

Arabidopsis Expresses Two Lipin Genes, AtPAH1 and AtPAH2, Which Encode Functional Phosphatidate Phosphatase Activities.

To identify the PAP gene involved in membrane lipid remodeling during Pi starvation, we searched for Arabidopsis homologs of lipin. Lipins are recently identified PAPs that affect lipid metabolism in yeast and animals, as evidenced by gene knockout models (21–24). Two lipin homologs were identified in Arabidopsis, which we designated as AtPAH1 (At3g09560) and AtPAH2 (At5g42870). These proteins each have a calculated molecular mass of approximately 101 kDa, which is much larger than previously characterized Arabidopsis PAPs [lipid phosphate phosphatase (LPP) family] (25–27). The subcellular localization prediction performed by TargetP did not detect transit peptides or transmembrane regions in AtPAH1 and AtPAH2. Both AtPAH1 and AtPAH2 have two domains, the amino-terminal lipin (NLIP) and carboxy-terminal lipin (CLIP) domains, which are conserved among lipins (Fig. S1) (22). The catalytic motif [or HAD-like domain (21)] was found in both isoforms, suggesting that they both may be functional PAPs. A phylogenic tree created with over 200 organisms suggested that lipin homologs are present only in eukaryotes (Fig. S2). AtPAH1 and AtPAH2 are only distantly related to previously characterized yeast or mammalian (human and mouse) lipins and are more closely related to lipins identified in algae and protists, suggesting that lipins are widely distributed among eukaryotes. Arabidopsis has both prokaryotic and eukaryotic pathways of membrane lipid biosynthesis, both of which are mediated by PAP (28). To date, no PAP involved in the eukaryotic pathway has been identified in Arabidopsis; because lipins are found only in eukaryotes, however, it is reasonable to predict that PAP function(s) may be mediated by lipins in the Arabidopsis eukaryotic pathway.

To examine whether AtPAH1 and AtPAH2 encode a functional PAP, we cloned the full-length coding sequence of each gene and introduced them into an yeast (Saccharomyces cerevisiae) Δdpp1Δlpp1Δpah1 mutant, which shows significantly decreased PAP activity and a temperature-sensitive phenotype, and therefore cannot grow at 37 °C (21). As shown in Fig. 1A, Δdpp1Δlpp1Δpah1 harboring AtPAH1 or AtPAH2 rescued the temperature-sensitive phenotype, indicating that either of these two Arabidopsis lipin homologs could complement yeast Δdpp1Δlpp1Δpah1 in vivo. Next, we measured PAP activity in the crude precipitate fraction (15,000 × g pellet) prepared from yeast Δdpp1Δlpp1Δpah1 transformed with AtPAH1 or AtPAH2 since major PAP activity attributable to AtPAH1 and AtPAH2 was detected in this fraction. We could not detect PAH1/2-derived PAP activity in the soluble fraction reproducibly. This is because defect of Mg²⁺-dependent PAP activity in Δdpp1Δlpp1Δpah1 is much more obvious in the membrane fraction than that in the cytosolic fraction (6). The PAP activity of yeast Pah1p is dependent on Mg²⁺ (1 mM for maximal activity) (21); thus, we examined whether the PAP activity of AtPAH1 and AtPAH2 also required Mg²⁺. As shown in Fig. 1B, there was a Mg²⁺-dependent increase in PAP activity (up to 1 mM Mg²⁺) in membranes of Δdpp1Δlpp1Δpah1 transformed with either AtPAH1 or AtPAH2, suggesting that AtPAH1 and AtPAH2 also encode Mg²⁺-dependent PAPs. Thus, the two isolated lipin homologs in Arabidopsis encode functional PAPs with enzymatic features similar to those of yeast Pah1p.

Fig. 1.

Identification of Arabidopsis lipins as functional phosphatidate phosphatases. (A) Complementation of temperature-sensitive phenotypes observed in the Δdpp1Δlpp1Δpah1 triple mutant by Arabidopsis PAH1 or PAH2. (B) Dependency of in vitro PAP activity on Mg²⁺. Data are means ± SD of triplicate assays with independent samples.

Isolation and Characterization of pah1pah2 Mutant.

To investigate the in vivo function of AtPAH1 and AtPAH2 in membrane lipid metabolism, we isolated T-DNA-tagged mutants of PAH1 and PAH2, pah1 (SALK_042850) and pah2 (SALK_047457), respectively. In both mutants, positions of T-DNA insertions were determined (Fig. S3A). Reverse transcription (RT) PCR revealed that homozygous pah1 and pah2 did not express their respective full-length mRNAs (Fig. S3B). We then produced the pah1pah2 double mutant by crossing the pah1 and pah2 homozygous lines (Fig. S3B). To investigate the involvement of PAH1 and PAH2 in PA dephosphorylation, we initially measured soluble PAP activity (29) of leaf extract from the pah1pah2. As shown in Fig. 2A, PAP activity in the supernatant fraction of pah1pah2 leaves was decreased by approximately 40% as compared to that in wild-type leaves, suggesting that significant portion of PAP activity in the supernatant fraction is attributable to PAH1 and PAH2. We further confirmed defect in PAP activity in vivo by radiolabeling detached rosette leaves of pah1pah2 with [³²P]-phosphate. The result showed that the relative amount of radiolabeled PA increased to 1.61- ± 0.32-fold (n = 3) in pah1pah2 as compared to the wild-type (Fig. 2B). These results suggest that PA metabolism is impaired in pah1pah2 in vivo due to the loss of PAP activity encoded by PAH1 and PAH2.

Fig. 2.

Characterization of pah1pah2. (A) Relative decrease in PAP activity in the supernatant fraction of leaf crude extract from pah1pah2. Data shown are means ± SD of triplicate measurements. Gray bar, wild-type; black bar, pah1pah2. (B) Relative increase in labeled PA after [³²P]-phosphate labeling of detached pah1pah2 rosette leaves. Data shown are means ± SD of triplicate measurements. Gray bar, wild-type; black bar, pah1pah2. (C) Lipid composition of pah1pah2. Total lipid was extracted from the aerial part of 20-day-old plants and analyzed. Data are means ± SD of triplicate analyses with independently extracted samples. Inset graph shows PA levels. Data are means ± SD of quadruplicate analyses with independently extracted samples. Gray bars, wild type; black bars, pah1pah2. (D) Subcellular localization of PAH1-GFP and PAH2-GFP. Rosette leaves of 35S::PAH1-GFP, pah1pah2 or 35S::PAH2-GFP, pah1pah2 transgenic plants were fractionated into membrane and soluble fraction as described in the Materials and Methods section. Proteins (50 μg each) were run on SDS/PAGE, blotted onto a nitrocellulose membrane and PAH1-GFP/PAH2-GFP detected by monoclonal anti-GFP antibody. Controls used are RuBisCO large subunit (RbcL) for plastid storoma and NADPH-dependent thioredoxin reductase A (NTA) for cytosol (27). Lane 1, crude; lane 2, crude of wild-type; lane 3, pellet; lane 4, supernatant.

Next, lipid analysis was conducted with 20-day-old vegetative shoots of wild-type and pah1pah2 plants. As compared with wild-type plants, the galactolipids MGDG and DGDG decreased by 16% and 30%, respectively whereas the composition of PC and PE increased by 47% and 20%, respectively, in pah1pah2 (Fig. 2C). In addition, there was a 26% increase in PA levels in pah1pah2 plants as compared to wild-type plants. Thus, PAH1 and PAH2 may supply DAG as a substrate of galactolipid synthesis, and PA hydrolyzed by PAH1 and PAH2 may be derived from PC and PE. The composition of phosphatidylglycerol (PG) and sulfoquinovosyldiacylglycerol (SQDG) were not affected by pah1pah2, suggesting that accumulated PA was not funneled into PG, and DAG produced by PAH1 and PAH2 was not used for SQDG biosynthesis. We further analyzed the 16:3 and 18:3 fatty acid composition of MGDG because 16:3 fatty acids of MGDG are derived exclusively from the prokaryotic pathway (Fig. S4) (30). The mole percent of 16:3 was 30.7 ± 2.0% in wild-type and 34.4 ± 0.6% in pah1pah2, whereas that of 18:3 was 62.0 ± 1.9% and 56.0 ± 1.9%, respectively. Although this increase in the mole percent of 16:3 and concomitant decrease in the mole percent of 18:3 was significant (P < 0.05), the contribution of PAH1 and PAH2 to the eukaryotic portion of MGDG is rather limited compared with TGD1, a component involved in ER to chloroplast lipid transfer, whose mutation causes much clearer changes in the fatty acid composition (31).

To understand the functions of PAH1 and PAH2, it is important to determine subcellular localization of PAH1 and PAH2. We therefore produced transgenic pah1pah2 plants that harbor either 35S::PAH1-GFP or 35S::PAH2-GFP transgenes. The transgenic plants (35S::PAH1-GFP, pah1pah2 and 35S::PAH2-GFP, pah1pah2) recovered phenotype observed in pah1pah2 mutant (see later for pah1pah2 phenotype), indicating that these transgenes were functional in vivo. Although GFP fluorescence was not observed by microscopy, Western blotting with anti-GFP antibody revealed that PAH1-GFP and PAH2-GFP were both expressed and detected in the soluble fraction (Fig. 2D). This suggests that PAH1 and PAH2 are mainly soluble proteins that are accessible to the lipid substrates located at different membranes.

The ER-Localized Eukaryotic Pathway of Membrane Lipid Metabolism Is Compromised in pah1pah2.

To examine PAP utilization in pah1pah2, an acetate labeling pulse-chase experiment was used (31). [¹⁴C]acetate is incorporated into fatty acids, which are subsequently acylated to a glycerol backbone by an acyltransferase to form glycerolipids. Because fatty acids are synthesized in plastids, glycerolipids synthesized in the prokaryotic pathway are labeled faster than those synthesized in the eukaryotic pathway in the pulse-chase experiment. In wild-type plants, the radiolabeled MGDG reached a maximum level of incorporation at 1 h after pulsing with [¹⁴C]acetate (Fig. 3A). This MGDG was then gradually disappeared, whereas radioactivity in PC, a signature lipid class of eukaryotic pathway, was increased transiently. In accordance with Xu et al. (31), radiolabeled PC levels in wild-type decreased with a concomitant increase in radiolabeled MGDG within 10 h of chase, indicating that MGDG was now produced by the eukaryotic pathway. However, in pah1pah2, radiolabeled MGDG showed a continual decrease even after 10 h of chase, and radiolabeled PC increased over the same time course (Fig. 3B). This result indicated that eukaryotic MGDG biosynthesis may be affected in pah1pah2, as was observed in tgd1-1 (31).

Fig. 3.

Impairment of eukaryotic pathway in pah1pah2 by in vivo pulse-chase labeling. (A and B) [¹⁴C]acetate labeling of fatty acids associated with individual lipids in (A) wild-type and (B) pah1pah2 plants were analyzed. Solid arrows indicate the plastid pathway component of MGDG labeling; the broken arrows represent the ER pathway component. Closed diamonds, MGDG; gray squares, PC; open triangles, PE; closed triangles, DGDG; closed circles, PG. (C and D) [¹⁴C]glycerol labeling of glycerol backbone in (C) wild-type and (D) pah1pah2 were analyzed. Closed diamonds, MGDG; gray squares, PC; open triangles, PE+PG; closed triangles, DGDG; closed circles, PI. Experiments were repeated three times, each with similar results. Shown is a representative result.

To further confirm disruption in eukaryotic MGDG biosynthesis, we performed pulse-chase labeling with [¹⁴C]glycerol (32). After the pulse of [¹⁴C]glycerol to excised leaves of wild-type and pah1pah2, radioactivity was predominantly incorporated into PC (32). At this time point, there was no significant difference in the percentage of radiolabeled PC and MGDG between wild-type and pah1pah2. Then, radioactive glycerol was totally washed off and chase was continued for 12 h. As shown in Fig. 3 C and D, the prominent difference between wild-type and pah1pah2 was observed till 3 h (2.5 h after chasing): whereas radioactive PC remains stable and MGDG showed gradual increase in wild-type, the simultaneous increase in radioactive PC and drop in radioactive MGDG was observed in pah1pah2, supporting an idea that conversion of PC to MGDG is affected in pah1pah2. Thus, it is suggested that PAH1 and PAH2 are involved in the ER-localized eukaryotic pathway of membrane lipid metabolism.

pah1pah2 Is Defective in Membrane Lipid Remodeling to Cope with Pi Starvation.

To examine whether PAH1 and PAH2 are involved in membrane lipid remodeling during Pi starvation, we observed phenotypes of 20-day-old wild-type and pah1pah2 plants that had suffered 10 days of Pi starvation. The growth of wild-type vegetative shoots in Pi-starved conditions was retarded as compared to those grown in Pi-replete conditions (compare Fig. 4 A and C), as reported (3). However, pah1pah2 mutant shoots exhibited a more severe growth retardation phenotype compared with wild-type plants under Pi-starved conditions (compare Fig. 4 C and D). We compared the fresh weight of Pi-starved wild-type and pah1pah2 vegetative shoots. Indeed, the fresh weight of Pi-starved shoots was significantly lower in pah1pah2 (wild-type: 148.3 ± 11.4 mg, pah1pah2: 93.2 ± 10.3 mg; n = 84), whereas there was no significant difference in weight under normal growth conditions (wild-type: 319.6 ± 25.0 mg, pah1pah2: 331.0 ± 32.9 mg; n = 72), indicating that this severe phenotype was not due to the generally small size of pah1pah2 plants. This result suggested that disruption of PAH1 and PAH2 affects vegetative shoot growth during Pi starvation. Root elongation of pah1pah2 was nearly halved in Pi-starved conditions as compared with Pi-replete conditions (Fig. 4E). To clarify if these phenotypes were due to changes in membrane lipid composition, we analyzed membrane lipid composition of 20-day-old vegetative shoots and roots after 10 days Pi starvation (Fig. 4F and Figs. S4 and S5). In the vegetative shoots of Pi-starved pah1pah2, a decrease in galactolipids and an increase in phospholipids were evident, as well as an increase in PA levels. (Fig. 4F). It should be noted that not only DGDG but also MGDG were decreased significantly in the double mutant. This phenotype is largely different from any other mutants deficient in Pi starvation-induced membrane lipid remodeling reported so far (15, 18–20). We also analyzed lipid contents of Pi-starved pah1pah2 roots because Pi-starvation inducible DGDG accumulation is more evident in roots (4). As shown in Fig. 4G, the decrease in DGDG and increase in PC were also observed in the roots of pah1pah2. Fatty acid composition of root galactolipids (Fig. S4) showed that 16:0 composition of DGDG was greatly reduced in pah1pah2 roots under normal growth condition, suggesting that PAH1 and PAH2 may provide a major 16:0 source of DGDG. However, the reduction of 16:0 was not observed under Pi-deprived conditions. These results suggest that this 16:0 source is different from that used to increase DGDG level during Pi starvation. In Arabidopsis, a galactolipid MGDG is synthesized by both inner and outer envelope–localized pathways, mediated by MGD1 and MGD2/3, respectively (10). These pathways are rather compartmentalized: whereas MGD1 synthesizes the bulk of MGDG (7), MGD2 and MGD3 are conditional enzymes that synthesize MGDG used particularly for DGDG synthesis at phosphate starvation in roots (15). In Pi-starved roots, indeed, MGD2/3 account for the net increase in DGDG since it is abolished in mgd2mgd3 roots (15). Because pah1pah2 decreases MGDG contents, it is suggested that DAG produced by PAH1 and PAH2 serves as a substrate mainly for MGD1, rather than MGD2/3 in vegetative shoots. However, in roots, PAH1/2-derived DAG can be used by MGD2/3 to some extent because apparent decrease in DGDG was still observed in pah1pah2 roots (Fig. S5). Since MGD1 is localized at the inner envelope of plastids, how DAG is transported to the inner envelope of plastids remains to be clarified. Thus, PAH1 and PAH2 play a pivotal role in membrane lipid remodeling during Pi starvation, and this remodeling is crucial for plants to cope with Pi starvation.

Fig. 4.

Phenotype of pah1pah2 under phosphate starvation. (A–D) Ten-day-old seedlings were transferred to solid media with or without phosphate for an additional 10 days. Phenotypes of 20-day-old wild-type and pah1pah2 were compared under phosphate-replete (A and B) or phosphate-starved (C and D) conditions. (A and C), wild-type; (B and D), pah1pah2. Scale bars, 1 cm. (E) Root growth in phosphate-starved pah1pah2. Seeds were germinated on vertically maintained solid media plates with or without phosphate (n = 8). (F) Lipid composition of pah1pah2 vegetative shoots under phosphate starvation. The 20-day-old plants that suffered 10-day phosphate starvation were used for lipid analysis. Data are means ± SD of triplicate experiments with independently extracted samples. Inset graph shows PA levels in vegetative shoots. Data are means ± SD of quadruplicate analyses with independently extracted samples. Gray bars, wild-type; black bars, pah1pah2.

Discussion

PAH1 and PAH2 Act as PAPs in the Eukaryotic Pathway of Membrane Lipid Metabolism.

The lipid composition of pah1pah2 was similar to that of tgd1-1 (33). In addition, the membrane lipid profiles of pah1pah2 and tgd1-1 by acetate-labeling (31) and glycerol-labeling (32) analyses suggest that eukaryotic pathway of membrane lipid metabolism may be affected in pah1pah2. A knockout of Type B MGDG synthases (mgd2mgd3), which localize to the plastid outer envelope, did not affect levels of MGDG under normal growth conditions (15). By contrast, the MGDG content decreased significantly in pah1pah2. Since type B MGDG synthases do not contribute to the bulk of MGDG synthesis (15), we hypothesize that PAH1/2-derived DAG can serve as a substrate for the inner envelope–localized galactolipid biosynthetic pathway mediated by MGD1. However, the 16:3/18:3 ratio of MGDG differed between pah1pah2 and tgd1-1. In tgd1-1, the increase in the 16:3/18:3 ratio was probably due to a blockage in the eukaryotic pathway because 16:3 is a signature fatty acid moiety of prokaryotic pathway-derived MGDG. In pah1pah2, however, the 16:3/18:3 ratio did not change as large as that in tgd1-1, suggesting that PAH1 and PAH2 may contribute to the eukaryotic pathway differently from TGD1. A major difference between the pah1pah2 and tgd1-1 mutants is the metabolite suggested to reach the inner envelope: in pah1pah2 it is DAG, whereas in tgd1-1 it is PA. Further investigation is required to understand how PAH1/2-derived DAG is transported to the outer leaflet of the inner envelope, where MGD1 is localized (33).

PAH1 and PAH2 Act as Central Mediators of Membrane Lipid Remodeling during Pi Starvation.

It has been shown that eukaryotic pathway is activated during Pi starvation-induced phospholipid-to-galactolipid conversion in membranes. The PLC NPC5 is involved in phospholipid hydrolysis for DAG production (18). However, because npc5-1 exclusively affects the outer envelope-localized galactolipid biosynthetic pathway, the main galactolipid biosynthesis mediated by MGD1 is unaffected (18). For pldζ2 and pldζ1pldζ2, which partially block the PC to PA conversion in roots, the phenotype is subtle in leaves, suggesting that PLDζ1 and PLDζ2 are not the primary mediators of membrane lipid remodeling (19, 20). Because pah1pah2 affects the eukaryotic pathway even under normal growth conditions, we hypothesized that PAH1 and PAH2 are also involved in membrane lipid remodeling during Pi starvation. As expected, galactolipid biosynthesis in pah1pah2 was further impaired upon Pi starvation. Overall pah1pah2 plant growth was significantly affected during Pi starvation, indicating that membrane lipid remodeling is an indispensable adaptation mechanism to cope with Pi starvation. Based on these results, we propose a revised model for membrane lipid remodeling under Pi starvation (Fig. 5). Upon Pi starvation, extraplastidic phospholipids are first hydrolyzed either by PLC (NPC4,5) (17, 18) or PLD (PLDζ1, ζ2) (19, 20). The PA produced by PLDζ1 and ζ2 is further dephosphorylated by PAP activity (19, 20) catalyzed by PAH1 and PAH2. However, the severer phenotype associated with pah1pah2 compared to pldζ1pldζ2 suggests that there are some more PLD isoforms involved in this pathway. The DAG produced by PAH1 and PAH2 is transferred to the inner envelope of chloroplasts to be galactosylated by MGD1 (7, 15), and eventually converted to DGDG by DGD1 (13). On the other hand, DAG produced by NPC4 and NPC5 are provided exclusively to the outer envelope of chloroplasts to be converted to DGDG by MGD2/3 and DGD2 (18). However, some portion of DAG used by the outer envelope-localized galactolipid biosynthetic genes may be derived from PAH1/2-mediated DAG production as evidenced by the decrease in root DGDG of pah1pah2 (Fig. S5). Thus, these pathways effectively remodel the membranes from phospholipids to galactolipids to circumvent Pi starvation.

Fig. 5.

A proposed model for membrane lipid remodeling during phosphate starvation. The pathway in red arrows shows substrate supply to the MGD1-mediated DGDG production, whereas that in blue arrows indicates pathway to the MGD2/3-mediated DGDG production.

In this report, we identified and characterized Mg²⁺-dependent PAP, AtPAH1 and AtPAH2 in Arabidopsis. The double knock out mutant pah1pah2 showed decreased galactolipid composition by affecting eukaryotic pathway. The severe growth phenotype of Pi-starved pah1pah2 suggests that these PAPs have pivotal role in adaptation to Pi starvation by mediating membrane lipid remodeling. However, following questions may need to be elucidated in future. We do not know whether PAH1 and PAH2 use PA derived exclusively from membrane phospholipids (e.g., PC) or both from membrane phospholipids and de novo synthesis. At least, under Pi deprivation, significant accumulation of phospholipids in the pah1pah2 mutant was observed, suggesting that conversion of phospholipids to galactolipids is largely dependent on PAH1 and PAH2 in this condition. However, the result of pulse-chase experiment showed that eukaryotic pathway was not completely blocked in pah1pah2 mutant, which suggest the existence of additional PAP serving for eukaryotic pathway. Indeed, Yeast Δpah1 mutant still remains a certain level of soluble PAP activity (21). Since there is no more homolog of PAH in Arabidopsis as well as in Yeast S. serevisiae (21), there could be unknown form(s) of PAP or some of the known membrane-bound PAP such as LPPα1 and LPPα2 (25) might account for the remaining PAP activity in pah1pah2 mutant. Alternatively, recently unraveled TGD transporter complex involved in the trafficking of extraplastidic PA into the inner envelope of chloroplasts may also play roles in the lipid remodeling upon Pi starvation (31, 33–36). Future investigation will give clues to answer these and other question derived from current study.

Materials and Methods

Plant Materials.

Arabidopsis thaliana (Columbia-0) was used in this study, and plants were grown under continuous light (5 μmol/m²s) at 23 °C. T-DNA knockout lines for PAH1 (SALK_042850) and PAH2 (SALK_047457) were obtained from the Arabidopsis Biological Resource Center (Columbus, OH). The Pi starvation conditions were as described (17).

Complementation of Yeast Δdpp1Δlpp1Δpah1 by AtPAH1 and AtPAH2 and Enzyme Activity Assay.

Full-length coding sequences of AtPAH1 and AtPAH2 were amplified using specific primers and cloned into the pDO105 vector (37) at NotI/MluI sites for AtPAH1 and NotI/PstI sites for AtPAH2. The primers used were: PAH1Fw (GCGGCCGCATGAGTTTGGTTGGAAGAGTTGGGAG) and PAH1Rv (ACGCGTTCATTCAACCTCTTCTATTGGCAGTTTCC), and PAH2Fw (GCGGCCGCATGAATGCCGTCGGTAGGATCGG) and PAH2Rv (CTGCAGTCACATAAGCGATGGAGGAGGCAG). Vector constructs were transformed into yeast Δdpp1Δlpp1Δpah1 (21), and transformants were screened on solid SD medium lacking histidine, tryptophan, leucine, and uracil. Each transformant was collected at exponential phase and disrupted by vortexing with glass beads. Total membrane proteins of each sample were collected by centrifugation (15,000 × g, 10 min) and suspended in assay buffer (50 mM Tris-HCl, pH 7.0, 50 mM NaCl, and 5% glycerol). The enzyme activity assay was carried out as described (27). Briefly, the membrane suspension was mixed with substrate solution (45 nCi L-3-PA, 1,2-di[1-¹⁴C]palmitoyl, (Perkin Elmer) and 18 nmol L-3-PA, 1,2-dipalmitoyl dispersed in 50 mM Tris-HCl, pH 7.0, containing 0.1% (wt/vol) Triton X-100 and various concentration of MgCl₂) and incubated at 25 °C for 1 h. The reaction products were extracted, developed on 1-D TLC and radioactive DAG spots quantified by Image Plate (Fuji Photofilm) and Image Analyzer (Storm, Amersham Biosciences). The same protocol was used to measure PAP activity of crude extract of pah1pah2 leaves.

Lipid Analysis.

Lipid analyses were conducted as described (38). Briefly, total lipid was extracted using the Bligh and Dyer method (39). Each lipid extract was separated by 2-D silica gel TLC with solvent system described in (38), and relevant lipid spots were scraped off the plate. The recovered lipids were then subjected to hydrolysis and methylation to obtain fatty acid methyl esters for analysis by gas chromatography. For PA quantification, an Agilent 1100 high-performance liquid chromatography system coupled with an Applied Biosystems 4000 Triple Quadrupole/Ion Trap mass spectrometer (Applied Biosystems) was used. Lipid extract resuspended in chloroform:methanol (1:1, vol/vol) was combined with internal standards including 1,2-dimyristoyl-sn-Glycero-3-Phosphate (di14:0-PA). Samples were introduced into the mass spectrometer by loop injections with chloroform:methanol (1:1) as a mobile phase for positive ESI mode and chloroform:methanol:200 mM piperidine (1:1:0.1) as a mobile phase for negative ESI mode, both at a flow of 250 μL min⁻¹ (40). Based on product ion and precursor ion analysis of head groups, two comprehensive sets of multiple reaction monitoring transitions were set up for quantitative analysis of PAs and other phospholipids (41). The PAs having different combinations of acyl moieties were quantified in comparison with the single internal standard of PA.

Construction of Transgenic Plants, Fractionation, and Western Blotting.

To create 35S::PAH1-GFP and 35S::PAH2-GFP plasmid constructs, full length coding sequences of AtPAH1 and AtPAH2 were amplified with pDO105-PAH1 or pDO105-PAH2 as templates and following primer sets: PSOYN728 (CACCCTCGAGAACAATGAGTTTGGTTGGAAGAGTTGGG) and PSOYN006 (CCGGCGCCTTCAACCTCTTCTATTGGCAGTTTCC) for PAH1, and PSOYN719 (CACCCTCGAGAACAATGAATGCCGTCGGTAGGATCG) and PSOYN008 (CCGGCGCCCATAAGCGATGGAGGAGGCAG) for PAH2. The amplicons were cloned into pENTR_D_TOPO vector (InvitrogenXhoI and SfoI to clone into the XhoI and SfoI sites of pGreen-35S::GFP (42). The constructs were introduced into pah1pah2 mutant via Agrobacterium-mediated transformation.

For the Western blotting, 3-week-old rosette leaves of 35S::PAH1-GFP, pah1pah2 or 35S::PAH2-GFP, pah1pah2 were shredded in the buffer (50 mM Tris-HCl and 330 mM sucrose, pH 7.0) to avoid organelle burst. The obtained crude extract was first centrifuged at 3,000 × g to remove tissue debris, then supernatant centrifuged at 125,000 × g to separate it into membranes and soluble fractions. Fifty micrograms of proteins were run on SDS/PAGE, blotted onto nitrocellulose membranes, incubated with anti-GFP monoclonal antibody as a primary antibody and HRP-conjugated anti-mouse IgG antibody as a secondary antibody. Bands were detected by chemiluminescence substrates (SuperSignal West Pico Maximum Sensitivity Substrate, Pierce) and films (Amersham Hyperfilm ECL high performance chemiluminescence film, GE Healthcare). Antibodies against RuBisCO large subunit (RbcL) and NADPH-dependent thioredoxin reductase A (NTA) (27) were used.

In Vivo Pulse-Chase Labeling Experiments.

The in vivo pulse chase labeling experiment with [¹⁴C]acetate was conducted according to Xu et al. (31) except that leaves were detached before labeling. On labeling, droplets of [¹⁴C]-acetate were applied on the surface of leaves and subsequently washed off for chase. The cut surface of petioles was kept away from [¹⁴C]acetate droplets. For labeling with [³²P]-phosphate, detached rosette leaves were soaked on MS medium containing [³²P]-phosphate (185 kBq/mL) for 3 h under light (33). [¹⁴C]-glycerol pulse-chase was carried out as reported (32). Total lipid was extracted and subjected to 2-D silica gel TLC with the solvent system of chloroform:methanol:7 N ammonia (120:80:8, vol/vol) for the first dimension and chloroform:methanol:acetic acid:water (120:30:15:3, vol/vol) for the second dimension. Radioactive spots were identified by autoradiography (Storm, Amersham Biosciences) and Imaging Plate (Fuji Film). The specific activities of each radiochemical used are: [¹⁴C]acetate (2.1 GBq/mmol, ARC), [³²P]phosphate (314 TBq/mmol, PerkinElmer) and [¹⁴C]glycerol (4.62 GBq/mmol, PerkinElmer).

Measurement of Root Length.

Root length was measured according to Kobayashi et al. (15). Briefly, root growth was captured on a digital camera, and primary root length was measured by image analyzing software (Scion Image, Scion).