Distinctly localized lipid phosphate phosphatases mediate endoplasmic reticulum glycerolipid metabolism in Arabidopsis

Abstract

Inter-organelle communication is an integral subcellular process in cellular homeostasis. In plants, cellular membrane lipids are synthesized in the plastids and endoplasmic reticulum (ER). However, the crosstalk between these organelles in lipid biosynthesis remains largely unknown. Here, we show that a pair of lipid phosphate phosphatases (LPPs) with differential subcellular localizations is required for ER glycerolipid metabolism in Arabidopsis (Arabidopsis thaliana). LPPα2 and LPPε1, which function as phosphatidic acid phosphatases and thus catalyze the core reaction in glycerolipid metabolism, were differentially localized at ER and chloroplast outer envelopes despite their similar tissue expression pattern. No mutant phenotype was observed in single knockout mutants; however, genetic suppression of these LPPs affected pollen growth and ER phospholipid biosynthesis in mature siliques and seeds with compromised triacylglycerol biosynthesis. Although chloroplast-localized, LPPε1 was localized close to the ER and ER-localized LPPα2. This proximal localization is functionally relevant, because overexpression of chloroplastic LPPε1 enhanced ER phospholipid and triacylglycerol biosynthesis similar to the effect of LPPα2 overexpression in mature siliques and seeds. Thus, ER glycerolipid metabolism requires a chloroplast-localized enzyme in Arabidopsis, representing the importance of inter-organelle communication in membrane lipid homeostasis.

IN A NUTSHELL

Background: The endoplasmic reticulum (ER) is the major site of glycerolipid biosynthesis for membrane and storage lipids, and phosphatidic acid phosphatase catalyzes a crucial reaction step of the glycerolipid biosynthesis pathway.

Question: What enzyme(s) are responsible for ER-localized phosphatidic acid phosphatase activity and what is their effect on lipid metabolism and plant growth in Arabidopsis?

Findings: We found that suppression of a pair of lipid phosphate phosphatases, LPPα2 and LPPε1, redundantly affected gametogenesis and lipid metabolism. LPPα2 localizes to the ER, but LPPε1 localizes at the chloroplast outer envelope. Nevertheless, both enzymes contribute to ER lipid metabolism. We concluded that ER glycerolipid metabolism requires a chloroplast-localized enzyme.

Next steps: Inter-organelle communication is integral to cellular homeostasis and we will further examine the mechanisms of inter-organelle communication in membrane lipid homeostasis.

Introduction

Inter-organelle communication is an integral subcellular process for maintaining cellular function. In membrane lipid metabolism of seed plants, the metabolic pathway for de novo assembly of glycerolipid structure, or Kennedy pathway, to synthesize phosphatidic acid (PA) resides both in plastids and endoplasmic reticulum (ER), which are catalyzed by a separate set of enzymes (Nakamura 2017). De novo synthesis of PA appears to occur separately in these organelles, but some subsequent metabolic steps involve inter-organelle trafficking (Li et al. 2016). For example, a galactolipid monogalactosyldiacylglycerol (MGDG) is synthesized exclusively in the plastid envelope, but its synthesis heavily relies on ER-derived precursory lipid compounds (Nakamura et al. 2010). In Arabidopsis (Arabidopsis thaliana), in which about a half of MGDG synthesis relies on an ER-derived precursor, a multipartite transporter complex, the trigalactosyldiacylglycerol (TGD) complex, plays a crucial role in ER-to-chloroplast lipid trafficking in galactolipid biosynthesis (Hurlock et al. 2014). Because each component fulfils an essential role in lipid trafficking and the lipid phenotype of the mutant indicates impaired ER-to-chloroplast lipid trafficking in galactolipid biosynthesis (Xu et al. 2003; Awai et al. 2006; Lu et al. 2007; Wang et al. 2012; Fan et al. 2015), this process is assumed to occur at the ER–chloroplast contact site (Fan et al. 2015). Thus, functional integrity of the TGD complex suggests the importance of an ER–chloroplast contact site in membrane lipid metabolism.

In glycerolipid biosynthesis, dephosphorylation of PA to synthesize diacylglycerol (DAG) is a pivotal reaction step because both the substrate and product are precursors for the biosynthesis of different glycerolipid classes (Nakamura 2017). PA phosphatase (PAP), an enzyme responsible for this reaction step, has two types: soluble phosphatidate phosphohydrolase (PAH) and membrane-integrated lipid phosphate phosphatase (LPP) families, with two and nine isoforms, respectively, in Arabidopsis. PAH1 and PAH2 are involved in phospholipid biosynthesis (Eastmond et al. 2010; Craddock et al. 2015) and membrane lipid remodeling under phosphate (Nakamura et al. 2009) or nitrate (Yoshitake et al. 2017) starvation.

The LPP family has two sub-families: the “eukaryotic” type (LPPα1 to α4) homologous to animal LPPs and the “prokaryotic” type (LPPβ, γ, δ, ε1 and ε2) homologous to cyanobacterial LPPs (Nakamura et al. 2007; Sato and Awai 2017). LPPα2 dephosphorylates PA in vivo and is involved in abscisic acid signaling in seed germination (Katagiri et al. 2005). Of the “prokaryotic” LPPs, LPPγ, ε1 and ε2 are all chloroplast-localized PAPs (Nakamura et al. 2007), and LPPδ functions as a long-chain base 1-phosphate phosphatase (designated sphingosine 1-phosphate phosphatase 1) (Nakagawa et al. 2012; Yanagawa et al. 2017). The function of LPPβ has not been reported (Nakamura and Ohta 2010). LPPγ knockout has a lethal effect on plant viability, so LPPγ is likely a crucial PAP isoform in chloroplasts. However, the role of the other two chloroplast-localized LPPs (LPPε1 and ε2) remains elusive because the lppε1 lppε2 double knockout mutant does not affect membrane glycerolipid contents (Nakamura et al. 2007). Thus, among the nine isoforms in the LPP family, which isoform(s) play a critical role as a primary PAP in membrane glycerolipid metabolism remains elusive. Moreover, whether plastidic and extraplastidic LPP isoforms have any interplay in glycerolipid metabolism remains to be addressed (Fig. 1A).

Figure 1.

Construction and phenotype observation of artificial microRNA-mediated LPPα2 knock-down lines in lppε1-2 background. A) Metabolic reaction catalyzed by lipid phosphate phosphatases (LPPs) and candidate isozymes in chloroplasts and endoplasmic reticulum (ER). DAG, diacylglycerol; G3P, glycerol 3-phosphate; LPP lipid phosphate phosphatase; PA, phosphatidic acid; TAG, triacylglycerol. B) Schematic representation of the Pro35S:amiLPPα2 constructions. The artificial microRNA (amiRNA) precursor fragments designed to target LPPα2 were cloned under the cauliflower mosaic virus (CaMV) 35S promoter. LB, left border; RB, right border; ter, transcriptional terminator. Two target sites of amiRNAs designed are indicated. C) RT-qPCR analysis of relative transcript level of LPPα2 in 7-day-old seedlings of wild type (WT) and independent transgenic lines harboring Pro35S:amiLPPα2-1 or Pro35S:amiLPPα2-2 in an lppε1-2 background. Data are mean ± SD from 3 biological replicates, each with a median of 3 technical replicates. The asterisks indicate statistical significance compared with WT by Student's t-test (*P < 0.05). D) Mature siliques of WT, Pro35S:amiLPPα2-1 lppε1-2 (lines #5 and #7) and Pro35S:amiLPPα2-2 lppε1-2 (lines #6 and #19). Bars, 1 mm. (E, F) Length E) and percentage of empty seed slots F) of siliques shown in D). Data are mean ± SD from 25 siliques in each genotype (n = 25). (G, H) Percentage of viable pollen by Alexander staining G) and pollen germination rate by in vitro culture H). More than 800 pollen grains were counted for each line. Data are mean ± SD. The asterisks indicate statistical significance compared with WT by Student's t-test (***P < 0.001).

To explore the potential interplay between plastidic and extraplastidic LPPs, we examined the genetic interaction between LPPα2 and LPPε1, representative LPP isoforms of “eukaryotic” and “prokaryotic” subtypes whose knockout mutants are viable (Nakamura et al. 2007). To avoid the possible lethal effect of the double knockout that is usual with impairment of the major lipid biosynthetic pathway, we created transgenic knockdown mutant lines suppressing LPPα2 in the knockout mutant background of LPPε1 and vice versa, both of which showed a male gametophyte defect. Although these LPPs showed similar tissue-specific expression patterns, LPPα2 was localized at the ER while LPPε1 was localized at the outer envelope of the chloroplast and close to ER-localized LPPα2. Lipid analysis of the knockdown mutant lines showed greater effect in siliques and seeds than the other tissues examined. In planta overexpression of LPPε1 affected ER-localized glycerolipid metabolism but not in chloroplasts, for a lipidomic phenotype similar to the overexpression of LPPα2, which resulted in an accumulation of triacylglycerols in mature siliques and seeds. We suggest that ER-localized LPPα2 and chloroplastic LPPε1 redundantly play a crucial role in ER glycerolipid metabolism.

Results

Transgenic knockdown of LPPα2 in lppε1-2 and LPPε1 in lppα2-1 revealed male gametophyte lethality

To investigate functional crosstalk between “eukaryotic-” and “prokaryotic-” type LPPs, we focused on LPPα2 and LPPε1 because only LPPα2 among the four “eukaryotic”-type LPPs affects PA dephosphorylation (Katagiri et al. 2005), whereas among the three chloroplast-localized “prokaryotic”-type LPPs, LPPε1 is expressed much higher than LPPε2, and LPPγ is an essential isoform whose knockout causes a lethal effect (Nakamura et al. 2007). To avoid a possible lethal phenotype by the double knockout of LPPα2 and LPPε1, we generated transgenic knockdown mutants suppressing LPPα2 in the lppε1-2/− background by using an artificial microRNA-mediated gene suppression system expressed under the control of the cauliflower mosaic virus 35S promoter (Pro35S:amiLPPα2 lppε1-2; Fig. 1B; Schwab et al. 2006). Two alternative artificial microRNA sequences (amiR-LPPα2-1 and amiR-LPPα2-2) were designed to target LPPα2 (Fig. 1B). After screening 50 individual transgenic Arabidopsis plants, we selected Pro35S:amiLPPα2-1 lppε1-2 lines #5, #7, and Pro35S:amiLPPα2-2 lppε1-2 lines #6, #19 as representative lines, showing significantly reduced LPPα2 expression as compared with the wild type (WT; Fig. 1C). The lines had a short silique phenotype (Fig. 1D) but no obvious morphological phenotype in vegetative organs. For further functional characterization, we used Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7 because of their stronger phenotype expression in siliques. The length of siliques in Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7 was about 6 mm shorter than that in WT siliques (Fig. 1, D and E) with 41% and 58% empty slots, respectively, containing remnants of aborted seeds (Fig. 1, D and F), which suggests reduced viability of gametophytes. Although the morphology of ovules appeared normal and indistinguishable from that of the WT, the pollen viability of Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7 was reduced to about 25% of the WT viability on Alexander staining (Fig. 1G; Alexander 1969), and the pollen germination rate was significantly lower, at less than 20%, compared to that of the WT, at 52% in vitro (Fig. 1H).

To rule out the possibility that these morphological phenotypes were caused by a functionally irrelevant second-site mutation in lppε1-2, we created transgenic lines that suppressed LPPε1 in lppα2-1. Two alternative miRNA sequences designed to specifically target LPPε1 (amiR-LPPε1-1 and amiR-LPPε1-2) (Supplemental Fig. S1A) were used for overexpression of the amiRNA in the lppα2-1 mutant background. We screened 24 transgenic lines for each construct and selected two representative lines for observation (Pro35S:amiLPPε1-1 lppα2-1 #6, #7 and Pro35S:amiLPPε1-2 lppα2-1 #20, #23) (Supplemental Fig. S1B). We confirmed the morphological phenotypes in siliques (Supplemental Fig. S1C), including silique length (Supplemental Fig. S1D) and percentage of aborted seeds (Supplemental Fig. S1E) as well as percentage of viable pollen (Supplemental Fig. S1F), which were similar to that of Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7.

Although the lppα2-1 or lppε1-2 single knockout mutant did not show these morphological phenotypes, the double heterozygous mutant plants lppα2-1/+ lppε1-2/+ produced siliques with empty seed slots similar to the suppression lines shown in Fig. 1D and Supplemental Fig. S1C (Supplemental Fig. S2, A–C). We failed to isolate a double homozygous mutant nor heterozygous mutants of one LPP isogene in the homozygous mutant background of the others (i.e. lppα2-1/− lppε1-2/+ and lppα2-1/+ lppε1-2/−). The lppα2-1/+ lppε1-2/+ plants significantly reduced the silique length along with an increased percentage of empty seed slot, as well as reduced percentage of viable pollen in the stamen (Supplemental Fig. S2, D–F), which were similar to the morphological phenotypes observed in the knockdown mutant lines (Fig. 1; Supplemental Fig. S1). To confirm whether the morphological phenotype in siliques is still observed when maximal fertilization is ensured, we manually self-crossed double heterozygotes, which showed silique lengths similar to those shown in Supplemental Fig. S2D compared with those from back-crossing with wild-type plants (Supplemental Fig. S3). Thus, we confirmed that these morphological phenotypes were due to suppression of LPPα2 and LPPε1.

The male gametophyte lethality was associated with defective pollen growth

To further examine the male gametophytic defects, we took a closer look at the pollen architecture in stamens of Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7 using scanning electron microscopy. These lines showed a considerable number of pollen grains with aberrant surface architecture, whereas WT pollen grains showed normal morphology (Fig. 2, A and B). Thus, suppression of LPPα2 in the knockout mutant of LPPε1 may cause a male gametophyte defect. Next, we used transmission electron microscopy to analyze pollen morphology of Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7 plants at four developmental stages (polarized microspore, mitosis I, bicellular, and mature stage; Fig. 2C). At the polarized microspore stage, pollen grains of Pro35S:amiLPPα2-1 lppε1-2 #5 and #7 showed an ultrastructure similar to that of the WT with no major morphological abnormality except a little odd round shape of pollen grains in the mutants. However, at later stages, pollen grains of Pro35S:amiLPPα2-1 lppε1-2 #5 and #7 showed a loosened cytoplasm, which eventually resulted in shrunken pollen at the mature stage. WT pollen showed no abnormal morphology. Therefore, the defects by suppressing LPPα2 in lppε1-2 during male gametogenesis occurred at the stage of pollen mitosis I, which suggests that LPPα2 and LPPε1 are crucial at this stage of male gametophyte development in Arabidopsis.

Figure 2.

Male gametophytic defect in Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7. A, B) Cryo-scanning electron microscopy of stamens A) and magnified pollen grains B). White asterisks indicate aberrantly shaped pollen grains. C) Transmission electron microscopy of pollen at four different stages (polarized microspore stage; mitosis I stage; bicellular stage; mature stage). Bars in A), 100 μm; in B), 10 μm; in C), 2 μm.

To further investigate how these morphological defects are associated with reduced fertility, we performed a pollen tube growth assay in vitro. Compared with WT pollen grains, pollen grains of lppα2-1/+ lppε1-2/+ had significantly reduced germination rates (Fig. 3A). Also, Pro35S:amiLPPα2-1 lppε1-2 (lines #5 and #7) and Pro35S:amiLPPε1-1 lppα2-1 (lines #6 and #7) showed further reduced germination rate, which suggests that suppression of LPPα2 and LPPε1 may affect pollen germination. To test if the phenotype is associated with decreased reaction product of the LPPs, we performed the germination assay in the presence of sn-1,2-dioctanoylglycerol (di08:0-DAG), a cell-permeable DAG molecular species commonly used for pharmacological assays in animal, yeast and plant cells (Davis et al. 1985; Kearns et al. 1997; Henneverry et al. 2001; Peters et al. 2014). We found that the reduced germination rate of pollen grains from lppα2-1/+ lppε1-2/+ and other suppression lines was rescued to the WT level by exogenous supplementation of DAG (Fig. 3A). We also measured the length of pollen tubes germinated in vitro, which was significantly reduced in lppα2-1/+ lppε1-2/+ and the other four suppression lines compared to in the WT (Fig. 3B). In the presence of di08:0-DAG, the reduced length was rescued to the WT level. These results suggest that suppression of LPPα2 and LPPε1 affected pollen germination and elongation due to reduced DAG production.

Figure 3.

Effect of LPPα2 and LPPε1 suppression on pollen tube growth. Percentage of germinated pollen A) and elongated pollen tube length B) of the wild type (WT; Col-0), lppα2-1/+ lppε1-2/+ and transgenic suppression lines Pro35S:amiLPPα2-1 lppε1-2 (#5 and #7) and Pro35S:amiLPPε1-1 lppα2-1 (#6 and #7) with mock treatment (left graph) or dioctanoyl diacylglycerol (di08:0-DAG) treatment (right graph). More than 50 pollen grains were counted in each assay. Statistical significance was analyzed by Student's t test: *P < 0.05; ***P < 0.001.

Tissue-specific localization of LPPα2 and LPPε1

To investigate the tissue-specific expression pattern of LPPα2 and LPPε1, we first referred to the GENEVESTIGATOR database, a publicly available transcriptomic database (Supplemental Fig. S4). Among LPP isoforms with available data, LPPα2 and LPPε1 showed ubiquitous tissue transcript profiles with slightly preferred expression in sperm cell and pericarp for LPPα2 and anther abscission zone, ovule and embryo for LPPε1. Next, we created transgenic plants harboring ProLPPα2:LPPα2-GUS or ProLPPε1:LPPε1-GUS reporter constructs for histochemical GUS reporter assays. We screened 24 transgenic lines for each construct and selected two representative lines for observation (ProLPPα2:LPPα2-GUS #16 and #17 and ProLPPε1:LPPε1-GUS #13 and #14). On observing GUS-stained tissues at different developmental stages of ProLPPα2:LPPα2-GUS #16 and ProLPPε1:LPPε1-GUS #13, LPPα2 and LPPε1 showed highly similar expression patterns (Fig. 4, A–M). In germinating seedlings, LPPα2-GUS and LPPε1-GUS were highly expressed in root tips (Fig. 4A); cotyledons (Fig. 4, B–D), particularly in the vasculature (Fig. 4E); and emerging leaf primordia (Fig. 4F). In roots, they were expressed in vasculature and tips (Fig. 4G). In 14-day-old seedlings, GUS staining was clearly observed at the shoot apices, leaf veins, and roots (Fig. 4, H and I). In the inflorescence stem, staining was obvious in stems but not cauline leaves (Fig. 4J). In reproductive tissues, GUS staining was detected in young buds but not clearly in mature flowers (Fig. 4, K and L). A closer look at the young floral buds of ProLPPα2:LPPα2-GUS and ProLPPε1:LPPε1-GUS plants showed clear GUS stain in the floral reproductive organs (Supplemental Fig. S5, A, B, E and F). Transverse sections of GUS-stained anthers at stage 7 showed GUS stain in the developing pollen grain (Supplemental Fig. S5, C, D, G and H). No clear GUS staining was observed in developing siliques (Fig. 4M). Similar GUS staining patterns were confirmed in other transgenic plant lines of ProLPPα2:LPPα2-GUS (line #17) and ProLPPε1:LPPε1-GUS (line #14) (Supplemental Fig. S6). To observe the expression pattern during seed development precisely, we produced transgenic Arabidopsis plants that harbor LPPα2 or LPPε1 fused with Venus fluorescence protein (ProLPPα2:LPPα2-Ven or ProLPPε1:LPPε1-Ven). The fluorescent signal patterns were highly similar between LPPα2-Ven and LPPε1-Ven: at the globular to heart stages, the signal was mainly observed in peripheral endosperm, which was then observed in cellular endosperm as well as embryos at later stages of seed development (Fig. 4N). Thus, LPPα2 and LPPε1 showed highly similar tissue-specific expression patterns, which suggests possible functional overlap.

Figure 4.

Tissue-specific localization of LPPα2 and LPPε1. A–M) Histochemical GUS staining of LPPα2-GUS and LPPε1-GUS in transgenic Arabidopsis plants harboring ProLPPα2:LPPα2-GUS or ProLPPε1:LPPε1-GUS in: seedlings at A) 1 d after sowing, B) 2 d after sowing, and C) 3 d after sowing; D) 4-d-old seedlings; E) 5-d-old seedlings; F) cotyledons and true leaf primordia of 5-d-old seedlings; G) the primary root tip of 5-d-old seedlings; H) 14-d-old seedlings; I) true leaf of 14-d-old seedlings; J) an inflorescence stem; K) an inflorescence with flowers; L) flowers at different developmental stages; M) developing siliques. N) Localization of LPPα2-Ven and LPPε1-Ven in developing seeds of transgenic Arabidopsis plants harboring ProLPPα2:LPPα2-Ven or ProLPPε1:LPPε1-Ven using confocal laser scanning microscopy. Bars in A–D), 500 μm; in (F, G), 200 μm; in (E, H, J–M), 2 mm; in I), 1 mm, in N), 50 μm.

Subcellular localization of LPPα2 and LPPε1

LPPε1 localized in chloroplasts on subcellular fractionation analysis with specific antibodies against LPPε1 (Nakamura et al. 2007). However, the subcellular localization of LPPα2 has not been reported. Thus, we used transgenic plants that harbored LPPα2 fused with Venus fluorescence protein (ProLPPα2:LPPα2-Ven). We confirmed that this construct is functional in vivo, because the previously reported low germination rate of lppα2-1 seeds after abscisic acid treatment (Katagiri et al. 2005) was complemented by transducing ProLPPα2:LPPα2-Ven (Supplemental Fig. S7). The Venus signal of ProLPPα2:LPPα2-Ven lppα2-1 line #16 was observed only in the roots, where it was in vascular cells (Fig. 5A) and clearly overlapped with ER marker dye (ER-Tracker). The similar ER-localized Ven signal was consistently observed in another transgenic line (ProLPPα2:LPPα2-Ven lppα2-1 line #34) (Supplemental Fig. S8A). To confirm the ER localization of LPPα2-Ven, we transiently co-expressed LPPα2-Ven with an ER marker (ER-rk) (Nelson et al. 2007) in Arabidopsis protoplasts (Fig. 5B) and Nicotiana benthamiana leaf epidermal cells (Fig. 5C). In both cases, the Venus fluorescent signal clearly overlapped with the ER-rk signal, which confirmed the ER localization in roots (Fig. 5A). Thus, LPPα2 may be localized at the ER.

Figure 5.

Subcellular localization of LPPα2-Ven. A) LPPα2-Ven signals in a root tip of 7-day-old seedlings harboring ProLPPα2:LPPα2-Ven (line #16) merged with ER-Tracker Red staining. Blue dashed lines indicate root shape, and areas indicated by white dashed squares are magnified in lower images. B) LPPα2-Ven with ER marker (ER-rk) in Arabidopsis leaf protoplast. C) LPPα2-Ven with ER-rk in N. benthamiana leaf epidermal cells. Bars in A), 20 μm; B, C), 10 μm.

Although LPPα2 and LPPε1 have a redundant function in gametogenesis (Figs. 1, 2 and3; Supplemental Figs. S1, S2 and S3), their subcellular localization was distinct: LPPα2 at ER (Fig. 5; Supplemental Fig. S8A) and LPPε1 at chloroplasts (Nakamura et al. 2007). To address how the differentially localized LPPs function redundantly, we performed an extensive subcellular localization study. First, we elaborated the chloroplast localization of LPPε1 based on previously reported subcellular fractionation analysis with anti-LPPε1 antibodies (Nakamura et al. 2007). We used transgenic Arabidopsis plants that harbor LPPε1 fused to Venus fluorescence protein (ProLPPε1:LPPε1-Ven). In leaf mesophyll cells of the transgenic line #19, the Venus signal was associated with chlorophyll autofluorescence but appeared in patches (Fig. 6A). This signal pattern was confirmed with another transgenic line harboring ProLPPε1:LPPε1-Ven (line #20; Supplemental Fig. S8B). Next, we transiently expressed LPPε1-Ven in Arabidopsis protoplasts and merged the signal with chloroplast autofluorescence to confirm this signal pattern, which again showed LPPε1-Ven signal in patches at the periphery of chloroplasts (Fig. 6B). To confirm whether these patched signals colocalize with chloroplasts, we transiently expressed LPPε1-mRFP in N. benthamiana leaf epidermal cells and merged the mRFP signal with chlorophyll autofluorescence, which indicated signal overlap at the periphery of the chloroplast envelope (Fig. 6C). To further test whether LPPε1 is also localized at the ER, we performed microsome fractionation of ProLPPε1:LPPε1-Ven transgenic plants using sucrose density-gradient centrifugation and BIP1 as the ER-localized marker protein (Chen et al. 2002). ER-localized proteins are known to show Mg²⁺-dependent shift of the peak of enriched fractions (Schaller 2017). As shown in Supplemental Fig. S9, BIP1 showed an intense signal in Fractions 2 to 4 in the presence of Mg²⁺, which was shifted to Fractions 5 to 13 in the absence of Mg²⁺. Here, LPPε1-Ven protein detected by anti-GFP antibodies did not show such a Mg²⁺-dependent shift of peaks, which suggests that LPPε1 is unlikely localized at the ER. Altogether, these results suggest that LPPε1 may be enriched in the chloroplast envelope, possibly at a peripheral subdomain, which is close to the ER-rk signal in Arabidopsis protoplasts (Fig. 6D).

Figure 6.

Subcellular localization of LPPε1-Ven. A) LPPε1-Ven signals in Arabidopsis leaf epidermal cells of 7-day-old transgenic plants harboring ProLPPε1:LPPε1-Ven (line #19). Bars, 20 μm. B) LPPε1-Ven with chlorophyll autofluorescence in Arabidopsis leaf protoplasts. C) LPPε1-mRFP with chlorophyll autofluorescence in N. benthamiana leaf epidermal cells. D) LPPε1-Ven with ER-rk in Arabidopsis leaf protoplasts. Lower panels are the magnified view of white dashed squares in upper panels. Bars in A), 20 μm; B–D), 10 μm.

Prompted by this observation, we further performed a series of transient assays to investigate the subcellular localization of LPPα2 and LPPε1. To test whether LPPα2 and LPPε1 are closely co-localized, we co-expressed LPPα2-Ven and LPPε1-mRFP in Arabidopsis protoplasts. Although these signals did not overlap clearly (Fig. 7A), a merged image showed a close localization of the two signals (Fig. 7B). The similar pattern of contact was observed using an N-terminal Venus fusion protein (Ven-LPPα2) (Supplemental Fig. S10), which suggests that the close contact may not be due to the orientation of the Ven reporter. To further confirm these localization patterns, we performed the transient assay in N. benthamiana leaves. Although LPPα2-Ven and LPPε1-mRFP did not overlap (Fig. 7C), a closer look at the signals showed that LPPα2-Ven surrounded LPPε1-mRFP (Fig. 7D). Likewise, the LPPε1-Ven signal was surrounded by the ER-rk signal (Fig. 7E). We further observed the differential localization pattern of LPPα2-Ven and LPPε1-Ven in developing embryos of ProLPPα2:LPPα2-Ven and ProLPPε1:LPPε1-Ven transgenic plants. As shown in Supplemental Fig. S11, the LPPα2-Ven signal overlapped with the ER tracker stain whereas the LPPε1-Ven signal was enriched at the periphery of chloroplasts, which confirmed the localization patterns observed in the transient assays (Figs. 5 and6). We performed quantitative analysis of the proximity using, DiAna, a widely used ImageJ analysis tool for analyzing object-based 3D co-localization and distance (Supplemental Fig. S12). Using the microscope images of the transient assay in Arabidopsis protoplasts and N. benthamiana leaves, we found that about 40% to 50% of the objects were co-localized (Supplemental Fig. S12). Taken together, these observations suggest that LPPα2 and LPPε1 may be localized closely to function redundantly in ER glycerolipid metabolism and male gametogenesis.

Figure 7.

Subcellular localization of LPPα2 and LPPε1. (A, B) Global A) and magnified B) image of LPPα2-Ven and LPPε1-mRFP in Arabidopsis leaf protoplasts. (C, D) Global C) and magnified D) image of LPPα2-Ven and LPPε1-mRFP in N benthamiana leaf epidermal cells. E) Magnified image of LPPε1-Ven with ER-rk in N. benthamiana leaf epidermal cells. Bars in A–E), 10 μm.

Involvement of LPPα2 and LPPε1 in glycerolipid metabolism

To investigate the role of LPPα2 and LPPε1 in polar glycerolipid metabolism, we analyzed the content of major glycerolipid classes in floral buds, mature flowers, siliques, and rosette leaves of Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7 as compared with that in the WT. In floral buds, we first quantified the level of PA, a substrate of LPP (Pierrugues et al. 2001; Nakamura et al. 2007). To quantify minute quantities of PA accurately, we radiolabeled floral buds of the WT and Pro35S:amiLPPα2-1 lppε1-2 #5 and #7 with [³²P]phosphate and quantified the amount of [³²P]PA (Fig. 8A). The amount of radiolabeled PA was increased in floral buds of Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7 as compared with in WT floral buds (Fig. 8B), which suggests that LPPα2 and LPPε1 may function as PA phosphatases in vivo. Next, we analyzed the content of major glycerolipid classes in mol%. In floral buds, the content of PC was decreased, whereas that of DGDG was increased in Pro35S:amiLPPα2-1 lppε1-2 suppression line (Fig. 8C). Although mature flowers showed no marked change in content, the PC level was consistently decreased in siliques (Fig. 8C). Rosette leaves showed no obvious or consistent changes (Fig. 8C). Next, by analyzing the amount of lipid contents per tissue dry weight, we confirmed that the content of PC but not PE in siliques was significantly reduced in both suppression lines (Fig. 8D). Also, we noted that the content of PI but not PG was reduced in the suppression line despite the fact that both phospholipid classes are formed from PA and CDP-DAG (Fig. 8D). Next, we analyzed the fatty acid composition of each glycerolipid class. In rosette leaves, we consistently observed increased contents of 18:1 in PC in the two transgenic mutant lines compared with in the WT (Supplemental Fig. S13). In floral buds, we observed a decreased content of 18:1 in DGDG and PE in the transgenic lines (Supplemental Fig. S14). Although mature flowers showed no consistent changes except a reduction of 18:3 content in PI (Supplemental Fig. S15), siliques of the suppression lines had consistently increased 16:1 content in PG and 16:0 content in PI at the expense of 18:2 content (Supplemental Fig. S16). We also analyzed total fatty acid composition of the pollen grains; however, no significant changes were observed (Supplemental Fig. S17). Thus, mature siliques had the most obvious changes in polar glycerolipid contents.

Figure 8.

Glycerolipid profiles in Pro35S:amiLPPα2-1 lppε1-2 plants. A) Schematic illustration of radiolabeling experiment to quantify PA level. Flower buds were immersed in MS medium containing [³²P]phosphate. After radiolabeling, total lipid was extracted and separated by 2-D thin-layer chromatography (TLC), and radioactive spots were quantified by densitometry. B) Quantification of phosphatidic acid (PA) content in the floral buds following [³²P]phosphate radiolabeling. PA level is shown as % of total radiolabeled phospholipids. Data are mean ± SD from 3 biological replicates. Statistical significance was analyzed by Student's t test: *P < 0.05; **P < 0.01. C, D) Content of polar glycerolipid classes in floral buds, mature flowers, mature siliques and rosette leaves in the wild type (WT) and Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7 shown in mol % C) and absolute amount D). Data are mean ± SD from 3 biological replicates. Statistical significance was compared with WT by one-way ANOVA; different letters indicate significant differences among the genotypes at P < 0.05. MGDG, monogalactosyldiacylglycerol; DGDG, digalactosyldiacylglycerol; PA, phosphatidic acid; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PI, phosphatidylinositol.

To confirm the effect of LPPα2 and LPPε1 suppression on polar glycerolipid contents, we analyzed the lipid profiles in mature siliques of Pro35S:amiLPPε1-1 lppα2-1 lines #6 and #7 and Pro35S:amiLPPε1-2 lppα2-1 lines #20 and #23. The results showed that PC content was consistently decreased in all of these suppression lines (Supplemental Fig. S18A). Also, MGDG content was increased significantly in Pro35S:amiLPPα2-1 lppε1-2 line #7 but not line #5 (Fig. 8C). Fatty acid composition showed that some of the lines had minor changes similar to those of Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7 but the changes were not consistent among the four transgenic lines (Supplemental Fig. S18B). These results suggest that LPPα2 and LPPε1 play a significant role in the glycerolipid metabolism of mature siliques, which showed a defective growth phenotype in transgenic suppression lines due to male gametophyte defect.

Overexpression of chloroplast-localized LPPε1 affects ER glycerolipid metabolism in vivo

To test our hypothesis that LPPα2 and LPPε1 function redundantly despite their different subcellular localization, we investigated whether chloroplast-localized LPPε1 affects ER-localized glycerolipid metabolism and/or whether ER-localized LPPα2 affects chloroplastic glycerolipid metabolism in vivo. We produced transgenic lines that overexpress LPPα2 or LPPε1 in a WT background, selected two representative transgenic lines from 12 candidate lines for each construct based on their high transcript levels (Pro35S:LPPα2 #7, #9, and Pro35S:LPPε1 #1, #6; Supplemental Fig. S19), and analyzed the membrane glycerolipid contents in mature siliques because a clear and consistent change of lipid profiles was observed in the knockdown lines (Fig. 8). Regardless of LPP isoform, the overexpression lines showed a similar change in relative lipid content (mol%), including a consistent increase in PE content and decrease in PI content but not in chloroplastic lipid classes such as MGDG or DGDG (Fig. 9A). However, the absolute amount of lipid classes per dry tissue weight did not alter significantly (Fig. 9B). Although the fatty acid composition of MGDG and DGDG showed no marked change, that of PC showed a consistent increase in 18:3 content in all lines and a decrease in 18:1 in most lines (Fig. 9C), implying an involvement of delta-15 desaturase activity. For PG, 16:1 content was increased consistently in all overexpression lines. 16:1 is produced in the plastids as the major PG species by FATTY ACID DESATURASE 4 (FAD4). These results therefore imply that extraplastidic PG content devoid of t16:1 may be specifically decreased. Thus, overexpression of LPPα2 or LPPε1 enhanced the biosynthesis of ER-localized phospholipids rather than chloroplastic glycerolipid classes.

Figure 9.

Lipid profiles in mature siliques of Pro35S:LPPα2 and Pro35S:LPPε1. A, B) Polar glycerolipid content shown in mol% A) and absolute amount B). C) Fatty acid composition of glycerolipids presented in A, B). Data are mean ± SD from 3 biological replicates. 16:1 in PG was in trans configuration. Statistical significance was compared with the wild type (WT) by one-way ANOVA; different letters indicate significant differences among the genotypes at P < 0.05. SQDG, sulfoquinovosyldiacylglycerol. See Fig. 8 legend for abbreviations for lipid classes.

Because the primary metabolic fate of ER-localized phospholipids is triacylglycerol (TAG) in developing seeds, we measured TAG content in mature dry seeds of these overexpression lines. To confirm the effect of 35S promoter activity in mature fresh seeds, we analyzed transcript levels of the target genes for the overexpression lines as well as amiR suppression lines. We found that all the overexpression lines used for the lipid analysis had more than a 100-fold increase in the transcript levels, while suppression lines showed significantly lower transcript levels than that of the WT (Supplemental Fig. S20). These data indicate that 35S promoter activity in these transgenic lines was effective in seeds. The result of lipid analysis showed that the TAG level in Pro35S:LPPα2 #7 and #9 was significantly increased as compared with that in the WT (Fig. 10A), which indicates that ER-localized LPPα2 contributes to TAG production. Of note, Pro35S:LPPε1 #1 and #6 seeds showed a similar increase in TAG content despite the chloroplast localization (Fig. 10A; Nakamura et al. 2007). The fatty acid composition of seed TAG accumulated in these four overexpression lines showed a slight decrease in 18:1 content (Fig. 10B).

Figure 10.

Triacylglycerol (TAG) analysis in dry mature seeds and pollen tubes. A, B) Absolute amount per dry weight A) and fatty acid composition B) in Pro35S:LPPα2 and Pro35S:LPPε1. C, D) Absolute amount per dry weight C) and fatty acid composition D) in Pro35S:amiLPPα2-1 lppε1-2 and Pro35S:amiLPPε1-1 lppα2-1. For A), B), C), and D), data are mean ± SD from 7 biological replicates. 50 seeds were used for each replicate. (E, F) Nile red staining in the pollen tube of wild type (WT), lppα2-1/+ lppε1-2/+ and transgenic suppression (Pro35S:amiLPPα2-1 lppε1-2 #5 and Pro35S:amiLPPε1-1 lppα2-1 #6) and overexpression (Pro35S:LPPα2 #7 and Pro35S:LPPε1 #1) lines shown in images E) and by quantification of signal intensities F). Bars, 100 μm. For A), B), C), D), and F), statistical significance was compared with WT by one-way ANOVA; different letters indicate significant differences among the genotypes at P < 0.05. G) TAG content in pollen grains of wild type (WT), lppα2-1/+ lppε1-2/+, Pro35S:amiLPPα2-1 lppε1-2 (#5, #7) and Pro35S:amiLPPε1-1 lppα2-1 (#6, #7). Data are mean ± SD from 4 biological replicates. Pollen grains from 500 flowers were used for each replicate. The asterisks indicate statistical significance compared with WT by Student's t-test (*P < 0.05; **P < 0.01). See Fig. 8 legend for abbreviations for lipid classes.

To test if increased TAG content in the overexpression lines is due to “extra” PA phosphatase activity that is not necessarily dependent on LPPα2 and LPPε1 isoforms, we produced transgenic lines that overexpress two other characterized LPP isoforms, LPPγ and LPPε2 (Nakamura et al. 2007) in a WT background and selected two representative transgenic lines from 11 candidate lines for each construct based on their high transcript levels (Pro35S:LPPγ #2, #6, and Pro35S:LPPε2 #1, #6; Supplemental Fig. S21, A and B). Although these overexpression lines showed about a 50 to 200-fold increase in transcript level (Supplemental Fig. S21C), TAG content was not significantly increased in the seeds (Supplemental Fig. S21D). This result suggests that increased TAG content may not be caused by an increase in activity of any PA phosphatase but specifically by LPPα2 and LPPε1 isozymes.

To examine whether the mutant lines conversely decreased TAG content, we analyzed TAG contents in the mature seeds of Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7 and Pro35S:amiLPPε1-1 lppα2-1 lines #6 and #7 compared with that in the WT. Although TAG content was not significantly decreased in these suppression lines (Fig. 10C), 18:1 content was significantly increased in contrast to the decrease in overexpression lines (Fig. 10D). These results suggest that TAG biosynthesis in the suppression lines was affected but compensated by an alternative TAG biosynthetic pathway. Since TAG is important for seed germination, we tested the seed germination rate of the suppression lines. Compared to the WT, the suppression lines as well as double heterozygous mutant (lppα2-1/+ lppε1-2/+) showed a consistent decrease in germination rate (Supplemental Fig. S22), suggesting an importance of LPPα2 and LPPε1 in seed viability. To further address how the role of LPPs for TAG biosynthesis is associated with the pollen defect observed in the suppression lines (Figs. 2 and3), we focused on TAG biosynthesis in the pollen tube since recent reports show that TAG metabolism is important for pollen tube growth (Müller and Ischebeck 2018; Bose et al. 2021). We used Nile red staining for detecting lipid droplets, which mainly consist of TAG. Compared to the WT, elongated pollen tubes of the suppression lines and double heterozygous mutant gave weaker staining signal while those of overexpressors showed more intense signals (Fig. 10E). Quantification of signal intensity from the images showed a significant decrease in suppression lines and a double heterozygous line but an increase in overexpressors, suggesting that TAG content is decreased by suppressing LPPs (Fig. 10F). To verify if TAG content was indeed decreased in the suppression lines and double heterozygous mutant, we quantified TAG content in pollen grains. Compared with the WT, pollen grains from the suppression lines and double heterozygous mutant all showed significantly reduced TAG content (Fig. 10G). These results suggest that the male gametophytic lethal phenotype in plants with suppressed LPPs may be due to impaired TAG biosynthesis that affects pollen tube growth.

Taken together, these results suggest that chloroplast-localized LPPε1 and ER-localized LPPα2 may function in ER glycerolipid metabolism but not chloroplast glycerolipid metabolism.

LPPε1 was localized at the outer envelope of chloroplasts

Since chloroplastic LPPε1 localizes closely to the ER-localized LPPα2 (Figs. 5, 6, and 7) and functions in ER glycerolipid metabolism (Figs. 8, 9, and 10), we investigated the sub-organelle localization of LPPε1 within the chloroplast. We isolated intact chloroplasts from the rosette leaves of ProLPPε1:LPPε1-Ven transgenic plants and detected LPPε1-Ven using a western blotting assay following protease treatment. Treatment with thermolysin digests proteins at the cytoplasmic leaflet of outer envelope (Cline et al. 1984), whereas trypsin digests proteins at the outer envelope and outer leaflet of the inner envelope (Cline et al. 1981, 1984; Marshall et al. 1990; Jackson et al. 1998). As positive controls, we used TRANSLOCON AT THE OUTER ENVELOPE MEMBRANE OF CHLOROPLASTS 159 (Toc159), which is localized at the cytosolic leaflet of outer envelope and thus is digested by thermolysin; TRANSLOCON AT THE INNER ENVELOPE MEMBRANE OF CHLOROPLASTS 22 (Tic22) at the inner envelope, which is digested by trypsin but not thermolysin; and CHLOROPLASTIC FORM OF GLUTAMINE SYNTHASE (GS2) at the stroma, which is resistant to both proteases (Chen et al. 2018). As shown in Fig. 11A, these marker proteins showed the expected digesting pattern based on the protein band detected by the specific antibodies. Here, the LPPε1-Ven signal detected by anti-GFP antibody was diminished to a trace level after thermolysin treatment, which suggests that LPPε1 may be localized at the cytosolic side of the chloroplast outer envelope.

Figure 11.

A proposed model of LPP function in Arabidopsis glycerolipid metabolism. A) Chloroplast sub-organelle localization of LPPε1-Ven. Intact chloroplasts isolated from rosette leaves of transgenic plant harboring ProLPPε1:LPPε1-Ven were treated with thermolysin and trypsin. An equivalent amount of protein from each treatment was analyzed by SDS-PAGE and immunoblotting with antibodies against TRANSLOCON AT THE OUTER ENVELOPE MEMBRANE OF CHLOROPLASTS 159 (Toc159), TRANSLOCON AT THE INNER ENVELOPE MEMBRANE OF CHLOROPLASTS 22 (Tic22), CHLOROPLASTIC FORM OF GLUTAMINE SYNTHASE (GS2), and GFP. Data are representative from 3 biological replicates with similar results. B) ER-localized LPPα2 and chloroplast-localized LPPε1 redundantly function in ER glycerolipid metabolism (pathways outside of chloroplast, in magenta arrows). LPPε1 does not affect chloroplast-localized glycerolipid metabolism (pathways within chloroplast, in green arrows), whereas LPPγ may play a major role. C) Subcellular compartmentalization of three classes of PA phosphatase (PAP) activities with distinct metabolic fate and tissue specificity. The ER-localized PAP is mediated by ER-localized LPPα2 and chloroplast-localized LPPε1 for the biosynthesis of ER phospholipids (PLs) and TAG but not chloroplastic glycerolipids (GLs), which play a role in flowers and siliques but not leaves. Cytosolic PAP is mediated by phosphatidate phosphohydrolase 1 (PAH1) and PAH2 for PL, TAG and possibly GL biosynthesis in leaves, flowers and siliques. Chloroplastic PAP is mediated by LPPγ, which may be crucial for leaf GL biosynthesis.

Since LPPε1 may face the cytosolic side of the outer envelope, we examined whether LPPε1 interacts with LPPα2 using a bimolecular fluorescence complementation (BiFC) assay including GTP BINDING PROTEIN BETA1 (AGB1) and GTP BINDING PROTEIN GAMMA-SUBUNIT1 (AGG1) as a positive control for their known interaction (Supplemental Fig. S23; Mason and Botella 2000; Tsugama et al. 2013). For co-expression of AGG1-nYFP and AGB1-cYFP, 132 cells indicated YFP fluorescence out of the 163 cells observed, that is, 80.9% of the cells showed an interaction. Similarly, co-expression of AGB1-nYFP and AGG1-cYFP resulted in 77.6% of the cells showing YFP fluorescence (66 of 85 cells observed), which confirmed the known interaction of AGG1 and AGB1 as a positive control of this experiment. However, co-expression of neither LPPα2-nYFP and LPPε1-cYFP nor LPPε1-nYFP and LPPα2-cYFP gave any YFP signal after observing more than 4,000 cells. These results suggest that LPPε1 and LPPα2 may not bind to each other.

Discussion

We revealed that ER-localized LPPα2 and chloroplast-localized LPPε1 redundantly function in ER-localized phospholipid metabolism (Figs. 1, 2, 3, 8, 9 and 10). From the morphological and lipid phenotypes of the transgenic lines, they primarily function in reproductive organs such as pollen and siliques rather than in vegetative organs (Figs. 1, 2, 3 and 8). The, LPP suppression lines affected pollen function, which was rescued by supplementing the reaction product DAG (Fig. 3). DAG is a substrate for the biosynthesis of TAG, an important lipid class for pollen tube function (Müller and Ischebeck 2018; Bose et al. 2021), whose accumulation was compromised (Figs. 10, E, F and G) and thus led to male gametophyte lethality as observed in short siliques containing empty seed slots (Figs. 1 and 2; Supplemental Figs. S1–S3). Also, the effect of LPPα2 and LPPε1 on TAG accumulation and viability was observed in mature seeds (Fig. 10; Supplemental Fig. S22), which suggests that these LPP are involved in ER-localized TAG biosynthesis in these different reproductive organs. We propose a new class of PA phosphatases whose function to maintain the ER glycerolipid homeostasis may be achieved by a pair of differentially localized isoforms for functional importance in inter-organelle communication (Fig. 11).

The morphological phenotype of the suppression lines indicated that LPPα2 and LPPε1 are required for the maturation (Fig. 2) and tube growth (Fig. 3) of pollen grains, which affected fertility and thus produced short siliques with empty seed slots (Fig. 1; Supplemental Figs. S1, S2 and S3). Since defective pollen tube growth was rescued by supplementing DAG, the reaction product of LPPs (Fig. 3), the morphological phenotype may be caused by reduced DAG production due to the suppression of LPPα2 and LPPε1. DAG is a substrate for the biosynthesis of TAG, which is important for pollen tube growth (Müller and Ischebeck 2018; Bose et al. 2021). Since reduced TAG content was observed in the pollen grains and tube of the LPP suppression lines (Fig. 10, E–G), LPPα2 and LPPε1 may produce DAG for TAG biosynthesis, which is required for pollen function. This function is in contrast to our previously reported role of diacylglycerol kinase 2 (DGK2) and DGK4, which catalyze the reaction opposite to LPPs in that DGK converts DAG back to PA (Angkawijaya et al. 2020). The suppression lines of DGK2 and DGK4 exhibited a pollen growth defect similar to the LPP suppression lines presented in this manuscript. However, the effect was due to compromised production of PA, which is known for PA signaling in pollen tube growth (Angkawijaya et al. 2020). Thus, although these two suppression lines show similar pollen growth phenotypes, their roles may be distinct: while DGK2 and DGK4 produce PA for signaling, LPPα2 and LPPε1 produce DAG for TAG production. Both lipid classes may be important in pollen growth through different lipid-mediated mechanisms.

Consistent with pollen growth, the role of LPPα2 and LPPε1 in TAG production was observed also in mature seeds. Overexpression of LPPα2 or LPPε1 increased seed TAG content with decreased 18:1 composition (Fig. 10, A and B), possibly because of enhanced ER-localized phospholipid biosynthesis. Suppression of LPPα2 and LPPε1 conversely increased 18:1 composition (Fig. 10D) representing the opposite effect on TAG biosynthesis despite the fact that the total TAG level was unaffected (Fig. 10C), which suggests that these LPPs may be in favor of producing TAG enriched with polyunsaturated fatty acid species. Apart from TAG, LPPα2 and LPPε1 also affected polar glycerolipid content in mature siliques, while the primary metabolic fate of glycerolipids is TAG because of developing seeds. The lipid analysis of suppression lines showed that siliques have the most prominent change among the four organs examined (Fig. 8, C and D). The consistent decrease in PC content, both in terms of mol% (Fig. 8C) and the absolute amount (Fig. 8D), but not in the other polar glycerolipid classes suggests that LPPα2 and LPPε1 may be mainly involved in PC biosynthesis. The overexpression lines did not alter PC content but consistently increased PE and decreased PI content in mol% (Fig. 9A). Decreased PI content indicates that the PA pool for LPPα2 and LPPε1 is shared with the CDP-DAG synthase activity-mediated pathway that produces CDP-DAG as the common precursor of PI and PG. Of note, PG content did not show a similar decrease but significantly increased 16:1 composition in trans configuration (t16:1) (Fig. 9C). Since t16:1 is a signature acyl group of plastid-synthesized PG (Browse et al. 1985), the significantly increased t16:1 content suggests that ER-localized PG biosynthesis may be compromised specifically while total PG content was unaffected (Fig. 9A). Neither suppression nor overexpression lines altered plastidic lipid classes such as MGDG and DGDG significantly (Figs. 8 and 9). Also, we found no vegetative growth phenotype that may be associated with the photosynthetic defect, which can be caused by decreased MGDG content (Kobayashi et al. 2007). ER-localized PAP is crucial for glycerolipid homeostasis because both the substrate PA and product DAG are the common precursors for the biosynthesis of different glycerolipid classes (Nakamura 2017). PA is a precursor for PG and PI, whereas DAG is converted to the major phospholipids PC and PE and to the storage lipid TAG (Nakamura et al. 2010; Nakamura 2017). Considering that the effect of these LPPs on polar glycerolipid profiles was less obvious than their effect on TAG content, a more detailed metabolic flux analysis such as a pulse-chase labeling experiment would verify the possible role of these LPPs as discussed above. Also, whether the regulation of fatty acid desaturase activity is affected by these LPPs remains to be investigated. Thus, lipid phenotypes of the suppression and overexpression lines suggest that LPPα2 and LPPε1 may function in ER-localized glycerolipid metabolism (Fig. 11B), which is required for pollen function and involved in seed oil content through TAG biosynthesis.

PAP is a key enzyme in the highly conserved glycerolipid metabolism pathway from bacteria to plants and animals (Nakamura and Ohta 2010). Biochemical studies in plants indicate both soluble and insoluble PA activities. ER-localized PAP activity is assumed to be essential, whereas plastid-localized PAP activity was known only for “16:3” plant species, including Arabidopsis, which are so called because of the presence of the 16:3 fatty acid moiety in MGDG owing to chloroplastic PAP activity (Heinz and Roughan 1983). The 11 Arabidopsis PAP isoforms hampered our identification of the critical PAP isozymes for the respective pathways and our investigation of their roles (Nakamura et al. 2009; Nakamura and Ohta 2010). In leaves, chloroplastic PAP (mainly LPPγ) plays an important role in glycerolipid metabolism (Nakamura et al. 2007) whereas cytoplasmic PAP (PAH1 and PAH2) is involved in phospholipid and TAG biosynthesis (Nakamura et al. 2009; Eastmond et al. 2010; Fan et al. 2014). However, PAP(s) responsible for ER-localized glycerolipid biosynthesis in non-photosynthetic organs remained open for investigation. Here, our present data suggest that LPPα2 and LPPε1 may be the PAPs for an ER-localized pathway and that these three pathways likely contribute differentially in different organs. The knockout mutant of cytosolic PAP, pah1 pah2, affects leaves, flowers, and siliques (Nakamura et al. 2009, 2014; Eastmond et al. 2010), indicating that cytosolic PAP activity plays a major role in both vegetative and reproductive organs. In contrast, PAP for ER glycerolipid metabolism, catalyzed by LPPα2 and LPPε1, plays a crucial role in reproductive growth but not vegetative growth. For chloroplast-localized PAP, the lethal phenotype of the LPPγ mutant implies a crucial role during reproduction and photosynthetic growth during vegetative growth, however, there is a lack of direct evidence (Nakamura et al. 2007). We propose that functional PAP in Arabidopsis is of three classes; ER (LPPα2 and LPPε1), cytosolic (PAH1 and PAH2), and plastidic (LPPγ), which contribute differently to both vegetative and reproductive growth (Fig. 11C).

We do not know why chloroplast-localized LPP commits to ER glycerolipid metabolism. Several mechanisms are known for the commitment of ER lipid metabolism to chloroplastic lipids, such as the TGD complex channeling ER-derived glycerolipid back to chloroplasts (Hurlock et al. 2014). However, apart from the export mechanism of neosynthesized fatty acid (Li et al. 2015), evidence for chloroplast commitment to ER glycerolipid metabolism is lacking. The membrane contact sites between the ER and chloroplasts have been discussed for possible bidirectional lipid trafficking between the ER and the chloroplasts (Andersson et al. 2007; Mehrshahi et al. 2013). Our data indicated that LPPα2 is ER-localized (Fig. 5) while LPPε1 is at the outer envelope of chloroplasts (Figs. 6 and 11A; Supplemental Fig. S8). Although LPPα2 and LPPε1 unlikely bind to form a functional complex because our BiFC assay did not show any interaction (Supplemental Fig. S23), the functional redundancy of LPPα2 and LPPε1 is supported by three factors: (i) the genetic suppression of both LPPα2 and LPPε1 but not either one caused a morphological phenotype in reproductive organs (Figs. 1, 2 and 3, Supplemental Figs. S1, S2 and S3) and lipid changes (Figs. 8 and 10C–G, Supplemental Figs. S13, S14, S15, S16, S17 and S18), (ii) the tissue expression patterns of LPPα2 and LPPε1 were highly similar to each other (Fig. 4; Supplemental Figs. S5 and S6), and (iii) the overexpression of either LPPα2 or LPPε1 caused the same lipidomic change (Figs. 9, 10, A and B). Although the biological significance of functional overlap between LPPγ and LPPε1 is open for future investigation, our data show no clear commitment of LPPε1 to chloroplastic glycerolipid metabolism. A significant difference between 16:3 plants and 18:3 plants is the presence or absence of chloroplastic PAP activity. Our lipid analysis of the overexpressors showed no significant increase in C16 fatty acid content, and the result of our sub-organelle localization assay suggests that LPPε1 may be localized at the cytosolic leaflet of the outer envelope (Fig. 11A). Thus, LPPε1 does not likely take chloroplast-localized PA as a substrate, which is produced at the inner envelope (Block et al. 1983). In chloroplast galactolipid biosynthesis, outer and inner envelope-localized pathways are functionally separated (Benning and Ohta 2005). While the inner envelope-localized pathway plays a critical role in Arabidopsis leaves (Awai et al. 2001; Kobayashi et al. 2007), the outer envelope-localized pathway is used for galactolipid biosynthesis in chloroplasts of pea (Pisum sativum var Laxton's Progress No. 9) (Cline and Keegstra 1983), an 18:3 plant species whose chloroplastic PAP activity is estimated to be too low for efficient galactolipid production (Heinz and Roughan 1983). Also, the outer envelope may be important in galactolipid synthesis in reproductive organs of Petunia hybrida (Nakamura et al. 2003). Because PAP activity for ER glycerolipid metabolism is present both in 18:3 and 16:3 plants, and paralogs of “prokaryotic” LPPs (LPPγ, LPPε1 and LPPε2) are widely found in these plant species, they may have a common role in catalyzing ER-localized PAP activity, but some are used to catalyze chloroplastic PAP activity for 16:3 plants. Although the evolutionary context of diversification between 18:3 plants and 16:3 plants is open for discussion, chloroplastic enzymes functionally isolated from chloroplastic glycerolipid metabolism may have acquired a function to contribute to ER glycerolipid metabolism. Comparative studies of the PAP systems in different model plants may further unravel these outstanding questions.

Materials and Methods

Plant growth conditions

Arabidopsis (Arabidopsis thaliana ecotype Columbia-0) plants were used in this study. Plants were grown at 22°C under a continuous light condition with a light intensity of 150 μmol m⁻² s⁻¹. Seeds were surface-sterilized with 70% ethanol and stratified at 4°C in the dark for 2−4 d before seeding. Seedlings were cultured on half-strength Murashige and Skoog (MS) medium (adjusted to pH 5.6 with KOH; Murashige and Skoog 1962). Adult plants were grown in soil by transplanting seedlings from MS agar plates.

Plant materials

Mutant seeds of lppα2-1 (SALK_124384) and lppε1-2 (SALK_010330C) were obtained from the Nottingham Arabidopsis Stock Centre (NASC, Nottingham, UK). Homozygous mutants were identified by PCR-based genotyping with T-DNA–specific primers and gene-specific primers, as illustrated in Supplemental Fig. S2A. The primers were CV040/CV041 and YN749/CV041 for lppα2-1 and CV032/YN749 and CV032/CV033 for lppε1-2. These T-DNA mutants had no detectable full-length transcript based on RT-PCR analysis (Supplemental Fig. S2B). The oligonucleotide primers used are shown in Supplemental Table S1. The wild type (WT; Col-0 ecotype background) was used as a control. A list of T-DNA mutants and other transgenic suppression/overexpression lines used for the phenotype analysis in this work is provided in Supplemental Table S2.

Construction of plasmid vector and transgenic lines

For ProLPPα2:LPPα2 and ProLPPε1:LPPε1, a 2,048-bp genomic sequence of LPPα2 and 1,210-bp genomic sequence of LPPε1 were amplified by PCR with the primers CV141/CV142 and CV143/CV144, respectively. The amplified fragments were cloned into the pENTR/D_TOPO plasmid vector (Invitrogen, Thermo Fisher Scientific, Waltham, MA) to obtain pVN037 (pENTR_ProLPPα2:LPPα2) and pVN041 (pENTR_ProLPPε1:LPPε1).

For ProLPPα2:LPPα2-GUS and ProLPPε1:LPPε1-GUS, the SfoI site was added before the stop codon of pVN037 and pVN041 by PCR-based site-directed mutagenesis (Sawano and Miyawaki 2000) with the primer CV127 for LPPα2 and CV140 for LPPε1 to obtain pVN052 and pVN053, respectively. The GUS cassette was inserted into the SfoI site of these plasmids to obtain pVN038 (pENTR_ProLPPα2:LPPα2-GUS) and pVN042 (pENTR_ProLPPε1:LPPε1-GUS).

For ProLPPα2:LPPα2-Ven and ProLPPε1:LPPε1-Ven, the triple (3x) repeat of the Venus fluorescent reporter construct was created by inserting the triple Venus cassette into the SfoI site of pVN052 and pVN053 to obtain pVN039 (pENTR_ProLPPα2:LPPα2-Ven) and pVN043 (pENTR_ProLPPε1:LPPε1-Ven).

For Pro35S:LPPα2 and Pro35S:LPPε1, an 873-bp open reading frame (ORF) of LPPα2 and 840-bp ORF of LPPε1 were amplified with the primers CV118/CV119 and CV130/CV131 and cloned into the Xhol and XbaI sites of pYN2047 (Lin et al. 2015) to obtain pVN036 (pENTR_Pro35S:LPPα2) and pVN040 (pENTR_Pro35S:LPPε1), respectively.

For Pro35S:LPPγ and Pro35S:LPPε2, an 681-bp open reading frame (ORF) of LPPγ and 861-bp ORF of LPPε2 were amplified with the primers CV349/CV350 and CV346/CV347 and cloned into the Sall and XbaI sites of pYN2047 (Lin et al. 2015) to obtain pVN232 (pENTR_Pro35S:LPPγ) and pVN231 (pENTR_Pro35S:LPPε2), respectively.

The plasmids pVN037, pVN041, pVN038, pVN042, pVN039, pVN043, pVN036, pVN040, pNV231, and pNV232 were recombined into the pBGW destination vector by using LR Clonase (Invitrogen, Thermo Fisher Scientific) (Karimi et al. 2005) to obtain pVN045, pVN049, pVN046, pVN050, pVN047, pVN051, pVN044, pVN048, pVN235 and pVN236, respectively. These plant binary vectors were transduced into lppα2-1/+ lppε1-2/+ or WT plants via Agrobacterium tumefaciens-mediated gene transformation. A total of 24 Basta-resistant T₁ plants were isolated and genotyped for harvesting T₂ seeds individually. To distinguish the transgenes from corresponding endogenous genes, a specific set of primers was designed for ProLPPα2:LPPα2 (CV123/KK097), ProLPPε1:LPPε1 (CV136/KK097), ProLPPα2:LPPα2-GUS (CV123/KK098), ProLPPε1:LPPε1-GUS (CV136/KK098), ProLPPα2:LPPα2-Ven (CV123/KK104), ProLPPε1:LPPε1-Ven (CV136/KK104), Pro35S:LPPα2 (CV123/CH072), Pro35S:LPPε1(CV136/CH072), Pro35S:LPPγ (CV351/CH072), Pro35S:LPPε2 (CV348/CH072), endogenous LPPα2 (YN749/CV041), and endogenous LPPε1 (CV032/YN749). Transgenic plant lines used for the observation were ProLPPα2:LPPα2-GUS (lines #16 and #17), ProLPPε1:LPPε1-GUS (lines #13 and #14), ProLPPα2:LPPα2-Ven (lines #16 and #34), ProLPPε1:LPPε1-Ven (lines #19 and #20), Pro35S:LPPα2 WT (lines #7 and #9), Pro35S:LPPε1 WT (lines #1 and #6), Pro35S:LPPγ1 WT (lines #2 and #6) and Pro35S:LPPε2 WT (lines #1 and #6).

Pro35S:amiLPPα2-1 lppε1-2 and Pro35S:amiLPPα2-2 lppε1-2. The artificial microRNA (amiRNA) sequences targeting LPPα2 were designed using WMD3 software (http://wmd3.weigelworld.org/cgi-bin/webapp.cgi). To suppress LPPα2, two amiRNA sequences were designed: amiR-LPPα2-1 (TCAGATAACAAAATGTGGCCC) and amiR-LPPα2-2 (TCTACACTCTATGTGTGGCGC). The fragment for the hairpin structure containing miRNA and miRNA* was cloned by PCR with the plasmid vector pRS300 and the primers CV091, CV092, CV093 and CV094 for the amiLPPα2-1 construct and the primers CV095, CV096, CV097 and CV098 for the amiLPPα2-2 construct. The amiRNA precursor fragments were then cloned into the Xhol and XbaI sites of pYN2047 (Lin et al. 2015) to obtain pVN024 and pVN025. The plasmids were then recombined into the pBGW destination vector using LR Clonase (Karimi et al. 2005) to obtain pVN028 and pVN029, which were transduced into lppε1-2/− plants by A. tumefaciens-mediated transformation. In total, 24 Basta-resistant T₁ transgenic plants were isolated for each construct. Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7 and Pro35S:amiLPPα2-2 lppε1-2 lines #6 and #19 were selected for analyses.

Pro35S:amiLPPε1-1 lppα2-1 and Pro35S: amiLPPε1-2 lppα2-1. The artificial microRNA (amiRNA) sequences targeting LPPε1 were designed using WMD3 software (http://wmd3.weigelworld.org/cgi-bin/webapp.cgi). To suppress LPPε1, two amiRNA sequences were designed: amiR-LPPε1-1 (TCAGTTAAGTACAACCTACGC) and amiR-LPPε1-2 (TCTATAGCTCGAAAGTCGCGA). The fragment for the hairpin structure containing miRNA and miRNA* was cloned by PCR with the plasmid vector pRS300 and the primers CV099, CV100, CV101 and CV102 for the amiLPPε1-1 construct and the primers CV103, CV104, CV105 and CV106 for the amiLPPε1-2 construct. The amiRNA precursor fragments were then cloned into the Xhol and XbaI sites of pYN2047 (Lin et al. 2015) to obtain pVN026 and pVN027. The plasmids were then recombined into the pBGW destination vector using LR Clonase (Karimi et al. 2005) to obtain pVN030 and pVN031, which were transduced into lppα2-1/− plants by A. tumefaciens-mediated transformation. In total, 24 Basta-resistant T₁ transgenic plants were isolated for each construct. Pro35S:amiLPPε1-1 lppα2-1 lines #6 and #7, and Pro35S:amiLPPε1-1 lppα2-1 lines #20 and #23 were selected for analyses.

Pro35S:LPPα2-Ven and Pro35S:LPPε1-Ven. The Venus fluorescent reporter construct was created by inserting the triple Venus cassette C-terminally into the SfoI site of pVN100 and pVN103. The obtained entry vector pVN101 (pENTR_Pro35S:LPPα2-Ven) and pVN104 (pENTR_Pro35S:LPPε1-Ven) were recombined into the pBGW destination vector using LR Clonase (Invitrogen, Thermo Fisher Scientific). The resulting plasmids pVN102 and pVN105 were used for transient expression in Arabidopsis protoplasts or Nicotiana benthamiana leaves by particle bombardment.

Pro35S:LPPα2-mRFP and Pro35S:LPPε1-mRFP. The 870-bp ORF of LPPα2 and 837-bp ORF of LPPε1 were amplified with the primers CV200/CV201 and CV203/CV204, respectively. The amplified fragments were cloned into the pENTR/D_TOPO plasmid vector (Invitrogen, Thermo Fisher Scientific) to obtain pVN108 (pENTR_LPPα2) and pVN109 (pENTR_LPPε1). These plasmids were recombined into the pGWB654 destination vector (Nakagawa et al. 2007) using LR Clonase (Invitrogen, Thermo Fisher Scientific) to obtain pVN111 (pGWB_Pro35S:LPPα2-mRFP) and pVN112 (pGWB_Pro35S:LPPε1-mRFP). The resulting plasmids pVN111 and pVN112 were used for transient expression in Arabidopsis protoplasts or N benthamiana leaves by particle bombardment.

Pro35S:Ven-LPPα2 and Pro35S:Ven-LPPε1. The Venus fluorescent reporter construct was created by inserting the triple Venus cassette into the SfoI site at the N-terminal of pVN135 and pVN136. The obtained entry vectors pVN137 (pENTR_Pro35S:Ven-LPPα2) and pVN138 (pENTR_Pro35S:Ven-LPPε1) were recombined into the pBGW destination vector using LR Clonase (Invitrogen, Thermo Fisher Scientific). The resulting plasmids pVN139 and pVN140 were used for transient expression in Arabidopsis protoplasts or N benthamiana leaves by particle bombardment.

ProUBQ10:nYFP-LPPα2, ProUBQ10:cYFP-LPPα2, ProUBQ10:nYFP-LPPε1, and ProUBQ10:cYFP-LPPε1. Plasmids of pVN108 and pVN109 were recombined into the pUBC-nYFP and pUBC-cYFP destination vectors (Grefen et al. 2010) using LR Clonase (Invitrogen, Thermo Fisher Scientific) to obtain pVN118 (ProUBQ10:nYFP-LPPα2), pVN119 (ProUBQ10:cYFP-LPPα2), pVN120 (ProUBQ10:nYFP-LPPε1) and pVN121 (ProUBQ10:cYFP-LPPε1).

ProUBQ10:LPPα2-nYFP, ProUBQ10:LPPα2-cYFP, ProUBQ10:LPPε1-nYFP, and ProUBQ10:LPPε1-cYFP. The full-length ORFs of LPPα2 and LPPε1 including the stop codons were amplified with the primers CV118/CV119 and CV130/CV131, respectively. The amplified fragments were cloned into the pENTR/D_TOPO plasmid vector (Invitrogen, Thermo Fisher Scientific) to obtain pVN122 (pENTR_LPPα2) and pVN123 (pENTR_LPPε1). These plasmids were recombined into the pUBN-nYFP and pUBN-cYFP destination vectors (Grefen et al. 2010) using LR Clonase (Invitrogen, Thermo Fisher Scientific) to obtain pVN124 (ProUBQ10:LPPα2-nYFP), pVN125 (ProUBQ10:LPPα2-cYFP), pVN126 (ProUBQ10:LPPε1-nYFP) and pVN127 (ProUBQ10:LPPε1-cYFP). The resulting plasmids pVN118, pVN119, pVN120, pVN121, pVN124, pVN125, pVN126 and pVN127 were used for bimolecular fluorescence complementation (BiFC) assays. For positive controls, we used GTP BINDING PROTEIN BETA 1 (AGB1; At4g34460, 1,131 bp) and GTP BINDING PROTEIN GAMMA-SUBUNIT 1 (AGG1; At3g63420, 294 bp; Tsugama et al. 2013). The protein coding sequences were amplified with primers YC290/YC359 for AGB1 and YC292/YC366 for AGG1, cloned into the pENTR/D_TOPO plasmid vector (pYC70 and pYC69, respectively), and recombined to the same destination vectors as above to obtain pYC71 (ProUBQ10:AGG1-nYFP), pYC72 (ProUBQ10:AGG1-cYFP), pYC73 (ProUBQ10:AGB1-nYFP) and pYC74 (ProUBQ10:AGB1-cYFP).

RNA extraction and reverse transcription quantitative PCR (RT-qPCR)

For 7-day-old seedlings, total RNA was extracted using TRI reagent with DNase treatment as described (Lin et al. 2015). For mature seeds, a pre-treatment step was performed before RNA extraction (Meng and Feldman 2010). Briefly, 100 mg of fresh mature seeds were frozen in liquid N₂ and grinded using a pre-chilled mortar and pestle with liquid N₂ to a fine powder. One-milliliter of extraction buffer [100 mM Tris-HCl (pH 9.5), 150 mM NaCl, 1% (w/v) Sarkosyl (Sigma, L5125) and 5 mM DTT (added just before use)] was immediately added to the mortar and the suspension was transferred to a microtube. The homogenate was vortexed for 5 min and centrifuged at 11,000 × g for 5 min at room temperature. The supernatant was transferred to a new microtube containing 500 μl of chloroform and suspended. Then, 500 μl of water-saturated acidic phenol (pH 4.3, Sigma, P4682) was added, mixed by vortexing for 2 min and centrifuged at 11,000 × g for 15 min at room temperature. The upper phase of solution was transferred to a new microtube, and 90 μl of 3 M sodium acetate and 600 μl of 2-propanol were added and mixed by inverting the microtube, then centrifuged at 11,000 × g for 10 min at room temperature. The pellet containing RNA was washed with 1 ml of 75% (v/v) ethanol and used to further extract RNA using the RNeasy Plant Mini Kit (Qiagen, Hilden, Germany). cDNA was synthesized using a SuperScript III first-strand synthesis kit (Invitrogen, Thermo Fisher Scientific). qPCR was conducted with the ABI7500 Real-Time PCR System (Applied Biosystems). Relative transcript levels were determined using the comparative threshold cycle method with ACTIN2 (ACT2) as an internal control. Data are mean and standard deviations from three independent biological replicates with three technical replicates for each biological replicate. Throughout the study, biological replicates were taken from the samples harvested at different times from separate plants grown separately under the same condition. The oligonucleotide primer sets for qPCR were LPPα2 (YN102/YN103), LPPε1 (YN114/YN115), and ACT2 (KK129/KK130) (Nakamura et al. 2014).

GUS staining

Histochemical GUS staining was performed following an immediate immersion of freshly harvested tissue in ice-cold 90% (v/v) acetone for 15 min then with GUS staining solution overnight as described (Lin et al. 2015). For colored tissues, pigments were removed by immersing the samples in ethanol: acetic acid (6:1 in volume). Images were acquired with a stereo microscope (Stemi 2000-C, Zeiss, Jena, Germany) equipped with a digital camera (D700, Nikon, Tokyo). Sectioning of anthers was performed as described (Kanehara et al. 2015). For sample fixation, the sequential dehydration involved 85%, 95% and 100% ethanol for 2 h each, followed by 100% ethanol overnight. Then, dehydrated samples were infiltrated using a sequential series of ethanol: LR white resin (London Resin, London, UK) mixtures [4 h each at 3:1, 2:1 and 1:1 (v:v), and overnight at 1:2, then 4 h each at 1:3 and 1:5]. Samples were then transferred to pure resin twice, with a 1-day incubation each time. Embedding and polymerization were performed at 60°C for 1 d, then the samples were cut into 2-μm sections using an ultramicrotome (EM UC7, Leica, Wetzlar, Germany) and observed under an upright microscope (Axio Imager Z1, Zeiss, Oberkochen, Germany) equipped with an AxioCam ERc5s camera (Zeiss, Oberkochen, Germany).

Confocal laser scanning microscopy

Fluorescence of 3xVenus and mRFP reporters was observed under a confocal microscope (LSM 510 Meta, Zeiss, Jena, Germany) equipped with Plan-Apochromat 20x/0.8-NA and Plan-Apochromat 10x/0.45-NA. For staining the ER, seedlings were immersed in 10 μg/ml ER-Tracker Red dye (E34250, Thermo Fisher Scientific, Waltham, MA) for 10 min before confocal microscopy observation. Images were captured using LSM 510 v3.2 (Zeiss, Jena, Germany) with filters for Venus (514 nm laser, 520–555 nm band-pass), ER-Tracker Red dye (543 nm laser, 560 nm long pass), chlorophyll autofluorescence (488 nm laser, 650 nm long-pass), and mRFP (561 nm laser, 590–630 band-pass). For observation of Ven fluorescence in the embryos of ProLPPα2:LPPα2-Ven and ProLPPε1:LPPε1-Ven transgenic plants at different developmental stages, embryo samples were cleared using clearance solution (chloral hydrate: glycerol: water, 8:2:1 by volume) prior to confocal microscope observation. Nile red staining was conducted as described using 2 μg mL⁻¹ Nile Red (Sigma-Aldrich) in 50 mM PIPES buffer (pH 7.0) for staining and observed by a confocal microscope (LSM 510 Meta; Zeiss; Cai et al. 2015; Bose et al. 2021). The fluorescence intensities of Nile Red staining were quantified using ImageJ (https://imagej.net/plugins/distance-analysis); intensities were averages of 10 biologically independent images in each genotype. Quantitative analysis of object-based co-localization was performed using the ImageJ plugin DiAna (Gilles et al. 2017). First, the objects were segmented using the 3D spot segmentation tool with the default values and setting. Then, the distance of the centroids of the closest neighbor was measured by DiAna. The objects with a center–center distance shorter than the theoretical resolution were considered to be colocalizing. Co-localization analysis was performed based on more than 15 images (n = 60–100).

Alexander staining

Alexander staining for the pollen viability test was conducted by mounting stamens in a drop of Alexander staining buffer (Alexander 1969) and observing the color of pollen grains using an upright microscope (Axio Imager Z1, Zeiss, Jena, Germany).

In vitro germination and tube elongation of pollen

In vitro germination and tube elongation of pollen were conducted as described using pollen grains from 20 mature flowers (He et al. 2018; Angkwijaya et al. 2020). 1,2-dioctanoyl-sn-glycerol (08:0-DAG) (#800800, Avanti) was used at 50 μM. Images were taken randomly from different regions in the solidified medium using an Axio Imager A2 microscope (Zeiss) with 10 × objective lens (FC Plan-Neofluar 10x/0.3 M27, Zeiss). A pollen grain was considered germinated if the tube length exceeded the diameter of the pollen. More than 50 pollen grains were counted in each genotype/treatment.

Transmission electron microscopy analysis

Anthers at different developmental stages from the wild-type and Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7 were observed as described (Ngo et al. 2018).

Cryo-scanning electron microscopy (cryo-SEM)

Cryo-SEM was used as described (Angkawijaya et al. 2020). Briefly, fresh flower samples were frozen in liquid nitrogen slush and transferred to a preparation chamber at −160°C for 5 min. Sublimation was performed at −85°C for 15 min. Then samples were coated with platinum (Pt) at −130°C, transferred to an SEM chamber and observed at −160°C using cryo-SEM (FEI Quanta 200 SEM; Thermo Fisher Scientific, Waltham, MA) with the Cryo-SEM Preparation System (PP2000TR; Quorum Technologies, Laughton, UK) at 20 kV.

Lipid analysis

Polar glycerolipids were extracted from 14-d-old whole seedlings, rosette leaves, floral buds, mature flowers, mature siliques, dried seeds, and pollen grains and analyzed as described (Angkawijaya et al. 2017). Briefly, total lipids were extracted from the frozen samples according to the Bligh and Dyer method (Bligh and Dyer 1959) and each glycerolipid class was separated by silica-gel thin-layer chromatography (TLC). The following solvent systems were used; two-dimensional separation with chloroform/methanol/aqueous ammonia (120:80:8, by vol) for the first dimension and chloroform/methanol/acetic acid/water (170:20:15:3, by vol) for the second dimension to analyze polar glycerolipid contents (Nakamura et al. 2003); and one-dimensional separation with hexane/diethyl ether/glacial acetic acid (40:10:1, by vol) for TAG analysis (Hung et al. 2013). Lipid spots on the TLC plate were visualized using primuline spraying, scraped off, and immersed in HCl-MeOH for 2 h at 85°C including pentadecanoic acid (1 mM) as an internal control. The resulting fatty acid methyl esters were analyzed with gas chromatography coupled with a flame ionization detector (GC-2010, Shimadzu, Japan) equipped with a ULBON HR-SS-10 column (Shinwa Chemical Industries, Japan).

Radiolabeling assay

To determine the PA level, freshly harvested floral buds of the WT and Pro35S:amiLPPα2-1 lppε1-2 lines #5 and #7 were immersed in ½ MS medium containing 30 μCi KH₂³²PO₄ (PerkinElmer, Waltham, MA, USA). After incubating for 3 h under light, total lipids were extracted as described (Bligh and Dyer 1959; Lin et al. 2019), spotted on a thin layer chromatography plate and developed with the solvent system of chloroform:methanol:aqueous ammonia (120:80:8 by vol) for the first dimension and chloroform:methanol:acetic acid:water (170:20:15:3 by vol) for the second dimension. Radioactive spots were visualized using Imaging Plate (Fuji Film, Tokyo) with BAS-2500 (GE Healthcare, Chicago, IL, USA). The intensities were quantified with an image analyzer (Typhoon FLA 7000; GE Healthcare Life Sciences) and analyzed with Image J. Data are mean ± SD from three biologically independent experiments.

Transient gene expression assay by particle bombardment

To perform transient gene expression assay in N benthamiana leaves, 5 μg plasmid DNA suspended with 1.25 mg tungsten particles in water was mixed with 50 μl of 2.5 M CaCl₂ and 20 μl of 0.1 M spermidine, precipitated, and suspended in ethanol. Leaves (3–5 cm in length) from plants were placed with their abaxial side upward on MS agarose medium in a Petri dish. Bombardment was performed with the PDS-1000/He Biolistic Particle Delivery System (BioRad, Hercules, CA, USA) with helium gas pressure 200 psi.

Protoplast transformation

Protoplasts were isolated from mesophyll cells of well-expanded leaves of 3 to 5-wk-old Arabidopsis as described (Wu et al. 2009). Protoplast transfection was performed as reported (Yoo et al. 2007) with 20 μg plasmid DNA (stored at 4°C) and approximately 4 × 10⁴ protoplasts in 200 μl MMg solution (0.4 M mannitol, 15 mM MgCl₂, 4 mM MES, pH 5.7) at room temperature. Plasmids were transformed into protoplasts by adding 220 μl fresh 40% PEG solution, then incubated at room temperature for 6 min. The transformed protoplasts were then washed with W5 buffer (154 mM NaCl, 125 mM CaCl₂, 5 mM KCl, 5 mM glucose and 2 mM MES at pH 5.7) and incubated in W5 buffer for 16 h under white light (75 μmol m⁻² s⁻¹, GC-102H; Firstek, Taiwan) at 22°C. The transfection efficiency was checked by confocal microscopy the next day.

Immunoblotting of ER-enriched fractions isolated from microsomal membranes

To obtain microsomal membranes, 14-day-old Arabidopsis plants cultured on ½ MS medium (Murashige and Skoog 1962) were homogenized with homogenization buffer (50 mM Tris-HCl [pH 8.2], 20% glycerol, 1 mM dithiothreitol [DTT], 2 mM ethylenediaminetetraacetic acid [EDTA], 1 mM phenylmethanesulfonylfluoride [PMSF], protease inhibitors with/without 5 mM MgCl₂) and ultra-centrifuged at 100,000 × g for 45 min as described (Schaller 2017). The precipitates containing microsomal membranes were resuspended in a suspension buffer (25 mM Tris-HCl [pH 7.5], 10% sucrose [w/v], 1 mM DTT, 2 mM EDTA, 1 mM PMSF, protease inhibitors with/without 5 mM MgCl₂) and then layered onto a 20% to 50% (w/w) sucrose gradient in 10 mM Tris-HCl (pH 7.5), 1 mM DTT, 2 mM EDTA, and 0.1 mM PMSF with/without 5 mM MgCl₂ for sucrose density gradient centrifugation at 100,000 × g for 16 h using a swinging bucket rotor and 1-ml fractions were collected. Then, 15 μl of the membrane fraction was denatured at 65°C for 3 min in SDS-containing sample buffer, size-fractionated by SDS/PAGE, transferred to PVDF membranes (Immobilon-P; Merck Millipore, Darmstadt, Germany) for western blotting and analyzed using a chemiluminescence detection system (SuperSignal West Pico, Thermo Fisher Scientific). The primary antibodies used were rabbit anti-GFP-HRP (#A-11122, Invitrogen, dilution at 1:2500) and mouse anti-BiP1 (#ADI-SPA-818F, Enzo Life Sciences, dilution at 1:3000) for ER-localized marker protein as a control. Goat anti- rabbit or anti- mouse IgG-HRP (#ab6721 and #ab97023, Abcam, dilution at 1:10,000) were used as secondary antibodies.

Bimolecular fluorescence complementation (BiFC) assay

The following pairs of plasmids were separately introduced into Arabidopsis protoplasts by transient assay or into N. benthamiana leaves by particle bombardment: pVN124/pVN127, pVN125/pVN126, pVN118/pVN121, and pVN119/pVN120. Plasmid combinations of pUBC-nYFP-AGG1/pUBC-cYFP-AGB1 and pUBC-cYFP-AGG1/pUBC-nYFP-AGB1 containing AGG1 and AGB1 integrated in the C terminus of the complete fluorescent protein were used as a positive control. After incubating for 16–20 h, the fluorescence in the transformed leaf protoplasts or N. benthamiana leaves were verified using a confocal microscope as described above.

Sub-organelle localization study of LPPε1

Intact chloroplasts were isolated from 18-day-old Arabidopsis transgenic plants harboring ProLPPε1:LPPε1-Ven (line #20) plants grown on soil as described (Chiu and Li 2008). Briefly, plant leaf material was incubated in Arabidopsis grinding buffer (AGR; 0.33 M sorbitol, 50 mM HEPES, 2 mM EDTA, 0.5% bovine serum albumin [BSA], pH 8) at 4°C and homogenized using a domestic blender (Waring, USA). Chloroplasts were pelleted by centrifugation at 3,000 × g for 3 min at 4°C and resuspended in AGR buffer. Intact chloroplasts were recovered by discontinuous Percoll (GE healthcare) gradient centrifugation and isolated intact chloroplasts were suspended in import buffer (0.33 M sorbitol, 50 mM HEPES-KOH, pH 8.0) with a concentration of 1 mg chlorophyll/ml. For protease treatment using trypsin and thermolysin (Jackson et al. 1998), aliquots of 125 μg chlorophyll were incubated in digestion buffer (0.33 M sorbitol, 50 mM HEPES-KOH, pH 8.0, 2 mM DTT, including 1 mM CaCl₂ for thermolysin treatment) containing 200 μg mL⁻¹ thermolysin or trypsin at 8°C for 30 min (thermolysin) or at 22°C for 1 h (trypsin). Reactions were stopped by adding EDTA or trypsin inhibitor to the final concentration of 10 mM, and samples were mixed with 20 μl of 6 × SDS-PAGE sample buffer and heated at 65°C for 10 min. Total proteins from protease-treated chloroplasts were analyzed by SDS-PAGE and transferred to PVDF membrane (Immobilon-P; Merck Millipore, Darmstadt, Germany). Antibodies against Arabidopsis TRANSLOCON AT THE OUTER ENVELOPE MEMBRANE OF CHLOROPLASTS 159 (Toc159, dilution at 1:1,000) (Tu et al. 2004), Arabidopsis TRANSLOCON AT THE INNER ENVELOPE MEMBRANE OF CHLOROPLASTS 22 (Tic22, dilution at 1:1,000) (Chou et al. 2003), CHLOROPLASTIC FORM OF GLUTAMINE SYNTHASE (GS2, dilution at 1:5,000; AS08 296, Agrisera) and GFP (dilution at 1: 2,000; A-11122 Invitrogen) were used for immunoblotting. Goat anti-rabbit or anti-mouse IgG-HRP (#ab6721 and #ab97023, Abcam, dilution at 1:10,000) were used as secondary antibodies.

Statistical analysis

Statistical data by Student's t-test and one-way ANOVA are provided in Supplemental Data Set S1.

Accession numbers

LPPα2 (At1g15080), LPPε1 (At3g50920).

Acknowledgments

We thank Hsou-Min Li (Institute of Molecular biology, Academia Sinica) for technical instruction with sub-organelle localization study and sharing Toc159 and Tic22 antibodies used as controls for the experiment; Kazue Kanehara (Institute of Plant and Microbial Biology, Academia Sinica) and her lab members Yueh Cho and Chao-Yuan Yu for technical assistance and plasmid construction with the transient gene expression assays and Kazue Kanehara for critical reading of the manuscript; Wann-Neng Jane, Mei-Jane Fan and Ji-Ying Huang in the Plant Cell Biology Core Lab and Live Cell Imaging Core Lab (Institute of Plant and Microbial Biology, Academia Sinica) for technical support with microscopy.

Author Contributions

V.N. performed experiments, analyzed data, and wrote the manuscript. Y.N conceived research, supervised V.N.'s experiments and data analysis, and wrote the manuscript. Both authors commented on the manuscript and approved the contents.

Supplemental Data

The following materials are available in the online version of this article.

Supplemental Figure S1. Construction and screening of Pro35S:amiLPPε1 lppα2-1 plants. (Supports Fig. 1)

Supplemental Figure S2. Isolation and characterization of T-DNA tagged mutants lppα2-1, lppε1-2 and lppα2-1/+ lppε1-2/+. (Supports Fig. 1)

Supplemental Figure S3. Length of mature siliques in the F1 progeny of ♂ lppα2-1/+ lppε1-2/+×♀ wild type (WT); ♂ WT × ♀ lppα2-1/+ lppε1-2/+; and ♂ lppα2-1/+ lppε1-2/+×♀ lppα2-1/+ lppε1-2/+. (Supports Fig. 1)

Supplemental Figure S4. Heatmap of tissue-specific expression pattern of LPP genes. Data were analyzed with GENEVESTIGATOR. (Supports Fig. 4)

Supplemental Figure S5. Histochemical GUS staining of LPPα2-GUS and LPPε1-GUS in transgenic Arabidopsis plants harboring ProLPPα2:LPPα2-GUS(A-D) or ProLPPε1:LPPε1-GUS (E-H) in floral buds and anther. (Supports Fig. 4)

Supplemental Figure S6. Tissue-specific localization of LPPα2-GUS and LPPε1-GUS by histochemical GUS staining. (Supports Fig. 4)

Supplemental Figure S7. Seed germination rate of the wild type (WT), lppα2-1 and ProLPPα2:LPPα2-Ven lppα2-1 under Mock or abscisic acid (1 μM) treatment. (Supports Fig. 5)

Supplemental Figure S8. Subcellular localization of LPPα2-Ven and LPPε1-Ven. (Supports Fig. 5)

Supplemental Figure S9. Fractionation of ER-enriched microsomal membrane fractions by sucrose density-gradient centrifugation. (Supports Fig. 6)

Supplemental Figure S10. Ven-LPPα2 with LPPε1-mRFP transiently expressed in Arabidopsis leaf protoplast cells. (Supports Fig. 7)

Supplemental Figure S11. Subcellular localization of LPPα2-Ven and LPPε1-Ven in mature embryos. (Supports Figs. 5, 6 and 7)

Supplemental Figure S12. Quantitative co-localization analysis of LPPα2 and LPPε1. (Supports Fig. 7)

Supplemental Figure S13. Fatty acid composition of major glycerolipid classes in rosette leaves. (Supports Fig. 8)

Supplemental Figure S14. Fatty acid composition of major glycerolipid classes in floral buds. (Supports Fig. 8)

Supplemental Figure S15. Fatty acid composition of major glycerolipid classes analyzed in mature flowers. (Supports Fig. 8)

Supplemental Figure S16. Fatty acid composition of major glycerolipid classes analyzed in mature siliques (Supports Fig. 8)

Supplemental Figure S17. Total fatty acid composition in pollen grain. (Supports Fig. 8)

Supplemental Figure S18. Lipid profiles in mature siliques of Pro35S:amiLPPε1 lppα2-1 plant lines. (Supports Fig. 8)

Supplemental Figure S19. Transcript levels of LPPα2 and LPPε1 in 12 independent transgenic plant lines of Pro35S:LPPα2 and Pro35S:LPPε1 in the wild-type (WT) background. (Supports Fig. 9)

Supplemental Figure S20. Relative transcript levels of LPPα2 and LPPε1 in mature seed of the wild type (WT), transgenic suppression lines, and overexpression lines. (Supports Fig. 10)

Supplemental Figure S21. Transcript levels and triacylglycerol (TAG) analysis of LPPγ and LPPε2 in transgenic plant Pro35S:LPPγ wild type (WT) and Pro35S:LPPε2 WT. (Supports Fig. 10)

Supplemental Figure S22. Seed germination rate of the wild type (WT), lppα2-1/+ lppε1-2/+ and 4 independent transgenic suppression lines. (Supports Fig. 10)

Supplemental Figure S23. Bimolecular fluorescence complementation (BiFC) assay of the interaction between LPPα2 and LPPε1 in Arabidopsis leaf protoplasts. (Supports Fig. 11)

Supplemental Table S1. Oligonucleotide sequences used in this study

Supplemental Table S2. List of mutant and transgenic lines used to characterize the function of LPPα2 and LPPε1

Supplemental Data Set S1. Statistical data.

Funding

The research was supported by Career Development Award provided by Academia Sinica (Grant ID: AS-CDA-107-L02) and International Collaboration Program with Strategic Research Partner provided by RIKEN to Y.N. The authors declare no conflict of interest.

Author notes

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plcell/) is: Yuki Nakamura ([email protected]).

Conflict of interest statement. None declared.