The isolation of the Arabidopsis thaliana pah1 pah2 mutant and the generation of transgenic plants overexpressing PAH1 or PAH2 and harboring GFP in a pah1 pah2 background were previously described (Nakamura et al., 2009). Surface-sterilized seeds of WT A. thaliana (Columbia-0), the pah1 pah2 mutant, and the transgenic mutant lines were incubated at 4°C in darkness for 3 days prior to plating on Murashige and Skoog medium (Murashige and Skoog, 1962) containing 0.8% (w/v) agar supplemented with 1% (w/v) sucrose. Plants were then incubated at 22°C under continuous white light (40–50 μmol m^-2 s^-1) for all growth conditions. Arabidopsis seeds were grown on solidified Murashige and Skoog agar supplemented with 1% (w/v) sucrose for 10 days and then were grown for another 7 days on a solidified N-sufficient (4.5 mM N) or N-depleted (0 mM N) medium (Estelle and Somerville, 1987; Gaude et al., 2007) supplemented with 1% (w/v) sucrose, with KNO₃ and Ca(NO₃)₂⋅4H₂O replaced with KCl₂ and CaCl₂, respectively.

Total RNA was isolated from three independent plant samples using the SV Total RNA Isolation System (Promega). Reverse transcription was performed using the PrimeScript RT reagent kit (TaKaRa Bio), and cDNA amplification was carried out using SYBR PreMix Ex Taq (TaKaRa Bio). Signal detection and quantification were performed in duplicate using the Thermal Cycler Dice Real Time System (TaKaRa Bio). Quantitative PCR determination of MGD1, MGD2, MGD3, DGD1, DGD2, PES1, and PES2 transcripts was normalized using the Arabidopsis UBQ10 transcript level (Sun and Callis, 1997). Expression levels were obtained from at least three replicates. The gene-specific primers used were as follows: PAH1 Fw (5′ GGATAACGAGGACAGGAAGACTG 3′); PAH1 Rv (5′ AGCAGCTGCGCTAAGTCCCATAC 3′); PAH2 Fw (5′ CTCAAGCCTCAGTCACAAGACAA 3′); PAH2 Rv (5′ AAGGAAAGAGACCATCAGGAGAGA 3′); PES1 Fw (5′ CCTGTCACCGCAACCAATC 3′); PES1 Rv (5′ ATTGTTGCACCAAACCGTGCT 3′); PES2 Fw(5′ CTCTTCTCCTATACTTACCTGG 3′); PES2 Rv (5′ CCTCAATAAGCTTCACCAAGT 3′); MGD1 Fw (5′ AGGTTTCACTGCGATAAAGTGGTT 3′); MGD1 Rv (5′ AACGGCAATCCCTCCTCAC 3′); MGD2 Fw (5′ GATTCGATCACTTCCTATCATCCTC 3′); MGD2 Rv (5′ TGTGCTAAACCATTCCCCAAC 3′); MGD3 Fw (5′ TCGTGGCGGATTGGTTTAG 3′); MGD3 Rv (5′ CGTTGTTGTTGTTGGGATAGATG 3′); DGD1 Fw (5′ CTGAAGAGAGATCCCGTGGTG 3′); DGD1 Rv (5′ TCCCAAGTTCGCTTTTGTGTT 3′); DGD2 Fw(5′ TGCAGAACCTATGACGATGGA 3′); DGD2 Rv (5′ GCTCTGTAAGTTGCGATGGTTG 3′); UBQ10 Fw (5′ GGCCTTGTATAATCCCTGATGAATAAG 3′); UBQ10 Rv (5′ AAAGAGATAACAGGAACGGAAACATAGT 3′).

The 17-day-old seedlings were incubated in 1 mL dimethylformamide (Wako) at 4°C for 20 h in the dark, and then absorbance was measured at 646.8 and 663.8 nm. Chlorophyll content was calculated as follows: Chla = (13.43A_663.8 – 3.47A_646.8)/sample fresh weight, and Chlb = (22.9A_646.8 – 4.88A_663.8)/sample fresh weight (Moran and Porath, 1980).

The maximum quantum efficiency (F_v/F_m) of photosystem II was measured using a Dual-PAM system (Walz). The minimum chlorophyll fluorescence (F_o) was determined by measuring light (intensity set at 20). After the F_o determination, a saturating pulse of red light (intensity set at 15) was applied to determine the maximum chlorophyll fluorescence (F_m). F_v/F_m was calculated as (F_m – F_o)/F_m. Before the F_v/F_m measurements, samples were incubated for 10 min in the dark. Samples were in plastic plates when measured.

To determine N concentration, ∼0.1 g fresh weight of shoots from 10- or 17-day-old seedlings was washed and oven dried at 80°C for a minimum of 3 days and then ground into powder. The N concentration was determined using a CN analyzer (Vario ELIII; Elementar Analysensysteme GmbH) according to Hachiya and Noguchi (2008).

Leaf segments were fixed with 2% (w/v) paraformaldehyde and 2.5% (w/v) glutaraldehyde in 0.067 M phosphate buffer (pH 7.4) for 2 h at room temperature and then for 16 h at 4°C. Samples were then washed six times in the phosphate buffer for 10 min each at room temperature. They were post-fixed with 2% (w/v) osmium tetroxide in 0.067 M phosphate buffer (pH 7.4) for 2 h at room temperature. The fixed samples were dehydrated in a graded ethanol series and embedded in an epoxy resin mixture (Quetol 651 mixture; Nisshin EM). Ultrathin 70-nm sections were cut with a diamond knife on a Leica EM-UC7 ultramicrotome and were transferred to copper grids. The sections were stained with EM stainer (Nisshin EM) for 1 h, followed by Reynolds’s lead citrate for 9 min at room temperature. The specimens were observed with a JOEL JEM-1400 Plus transmission electron microscope at an accelerating voltage of 80 kV.

Total lipids were extracted from tissues as described (Bligh and Dyer, 1959). Polar membrane lipids were separated by two-dimensional thin-layer chromatography (Kobayashi et al., 2007). Neutral glycerolipids were separated by one-dimensional thin-layer chromatography using the solvent system of hexane/diethyl ether/acetic acid (160:40:4, v/v/v). Lipids on silica gel plates were visualized with 0.01% (w/v) primuline in 80% (v/v) acetone under UV light. Lipids isolated from silica gel plates were methylated, and fatty acid methyl esters were quantified by gas chromatography using pentadecanoic acid as an internal standard (Kobayashi et al., 2006).

In vivo labeling experiments with [¹⁴C]acetate were conducted according to Nakamura et al. (2009), except that [¹⁴C]acetate was diluted to 0.05 mCi mL^-1 with N-depleted medium (Estelle and Somerville, 1987; Gaude et al., 2007). [¹⁴C]acetate was spread with a micropipette in 3-μL droplets over the surface of the third and fourth leaves cut off from 15-day-old seedlings, and the leaves were placed on 3MM paper immersed in N-depleted medium and incubated for 2 h at 22°C. After the incubation, the leaves were washed twice in non-radioactive N-depleted medium to remove exogenous radioactivity and were placed on 3MM paper immersed in N-depleted medium. After the incubation, the samples were harvested and were frozen in liquid nitrogen. Lipids were extracted and separated by two-dimensional TLC as described above. Radioactive spots were analyzed by autoradiography (FLA7000, Fuji Film) and Imaging Plate (Fuji Film).

We previously produced Arabidopsis transgenic plants overexpressing PAH1 or PAH2 in the pah1 pah2 background (Nakamura et al., 2009). Expression levels of PAH1 and PAH2 were analyzed for several lines for each transgenic combination, and PAH1OE 20-1 and PAH2OE 22-1 were selected and used as PAH1OE and PAH2OE, respectively, for further analyses (Supplementary Figure S1). Expression of PAH1 in PAH1OE and PAH2 in PAH2OE was 11-fold and 5-fold higher than that in WT, respectively, under normal growth conditions (+N), and these levels were 16-fold and 3-fold higher than that in WT, respectively, under N starvation (–N) (Figures 1A,B). When we grew plants under N-depleted conditions, seedlings of the Arabidopsis double-knockout mutant pah1 pah2 were significantly smaller than WT seedlings, whereas PAH1OE and PAH2OE seedlings were comparable to or larger than WT seedlings (Figures 1C,D). Under N-sufficient conditions, the chlorophyll content in WT, pah1 pah2, PAH1OE, and PAH2OE did not differ significantly (Figure 1E). Under N-depleted conditions, however, the chlorophyll content in pah1 pah2 was significantly lower than that in WT, whereas the content in PAH1OE and PAH2OE was similar to or greater than that in WT (Figure 1E). As for photosynthetic activity parameters under N-depleted conditions, both PAH1OE and PAH2OE showed significantly higher maximum quantum yield (F_v/F_m) as compared with those of WT and pah1 pah2 (Figure 1F). The differences in sensitivity to N starvation among WT, pah1 pah2, PAH1OE, and PAH2OE could be due to the enhanced efficiency of N uptake during the first 10 days of growth under N-sufficient conditions. Therefore, we assessed total N concentration in seedlings of WT, pah1 pah2, PAH1OE, and PAH2OE; the N concentrations across these plants were comparable under both N-sufficient and N-depleted conditions (Figure 1G), suggesting that the growth defect in pah1 pah2 during N starvation is not due to the lower N content in the seedlings as compared with the other plants.

As a significant decrease in chlorophyll content was observed only in pah1 pah2 grown under N starvation, we visualized the intracellular structure of plant leaves with electron microscopy and compared the structures among WT, pah1 pah2, and PAH1OE under N-sufficient and N-depleted conditions (Figure 2 and Supplementary Figure S2). Under N-sufficient conditions, chloroplast membrane structures were similar among WT, pah1 pah2 and PAH1OE (Figures 2A–C). Under N-depleted conditions, starch accumulation in chloroplasts was observed across all of the plants (Figures 2D–F and Supplementary Figure S2). The belt shaped-structures observed in starch granules have no biological phenotype but the technical reason of cutting the sample in the resin for microscopic analysis. As for the thylakoid membrane structures in the enlarged images, however, we could often observe grana-lamellae stacking structures in the thylakoid membrane of WT and PAH1OE but not in pah1 pah2 under N-depleted conditions (Figures 2G–I and Supplementary Figure S2). Thus, it was clearly shown that the absence of PAH enhances breakdown of the chloroplast membrane structure, especially under N starvation and thus decreases chlorophyll content under N-depleted conditions as compared with WT (Figure 1E). We also observed the chloroplast membrane structure of PAH1OE. The membrane structure of PAH1OE was either comparable with WT or showed a slightly increased number of grana-lamellae stacking structures under N-depleted conditions (Figures 2C,F,I and Supplementary Figure S2). Although we could not assess these data statistically because the number of grana-lamellae stacking structures in pah1 pah2 was markedly lower than that in WT and PAH1OE, the stacking repeat distance in thylakoids of pah1 pah2 was wider than those in WT and PAH1OE (Supplementary Figure S3). Thus, our results showed that the absence of PAH affects the structure of photosynthetic membranes in chloroplasts, especially under N starvation.

Nitrogen starvation leads to a decrease in MGDG and a concomitant increase in DGDG in WT Arabidopsis (Gaude et al., 2007). In pah1 pah2, the mole percent of MGDG and DGDG in the total glycerolipids are smaller than in WT, even under N-sufficient conditions (Nakamura et al., 2009), which was also observed in our results (Figure 3). However, in pah1 pah2, the mole percent of MGDG decreased notably whereas that of DGDG only slightly increased during N starvation as compared with those in WT, PAH1OE, and PAH2OE (Figure 3). In PAH1OE and PAH2OE, the membrane lipid compositions were similar to one another, and changes in their mole percent during N starvation did not differ substantially from WT, except that the mole percent of MGDG in PAH1OE was slightly greater than that in WT under N-depleted conditions (Figures 3A,C,D). Given that PAH1 and PAH2 similarly complemented the pah1 pah2 phenotype with respect to membrane glycerolipid composition (Figures 3A,C,D), PAH1 and PAH2 may have a redundant role in maintaining membrane lipid homeostasis of the ER and chloroplasts under both N-sufficient and N-depleted conditions. Moreover, these results clearly show that PAH1 and PAH2 play an important role in maintaining the mole percent of chloroplast membrane lipids to the extraplastidial membrane lipids, namely phospholipids, especially under N-depleted conditions. Given that the fatty acid compositions of MGDG and DGDG were similar among WT, pah1 pah2, PAH1OE and PAH2OE under both N conditions (Supplementary Figure S4), it was suggested that the prokaryotic pathway and the eukaryotic pathway equally contributed to the galactolipid synthesis. We also analyzed the TAG and DAG contents (Figures 3E–G). Under N-sufficient conditions, the amounts of TAG were comparable among plants (Figure 3E). However, under N-depleted conditions, the amount of TAG in pah1 pah2 was lower and that in PAH1OE was slightly higher than that in WT (Figure 3F), suggesting that PAH1 is involved in the TAG accumulation in leaves under N starvation. PAH2OE showed a similar result to PAH1OE, but it was not significant. Given that the fatty acid compositions of TAG were similar among WT, pah1 pah2, PAH1OE and PAH2OE under both N conditions (Supplementary Figures S5A–D), it was suggested that PAH1 (and PAH2) might be only slightly involved in TAG synthesis under N starvation. The DAG contents also showed similar profile with the TAG contents (Figure 3G). Compared with WT, the lower amount of DAG in pah1 pah2 under N-sufficient conditions and the higher amount of DAG in PAH1OE under N-depleted conditions were observed (Figure 3G). However, the fatty acid compositions of DAG were varied in pah1 pah2, PAH1OE and PAH2OE (Supplementary Figures S5E–H), suggesting that the lack or overexpression of PAH might destabilize the whole flow of DAG synthesis.

In plants grown under N-depleted conditions, the synthesis of fatty acid phytyl esters catalyzed by PES1 and PES2 is accelerated in chloroplasts to avoid the accumulation of free fatty acids and phytols derived from the degradation of thylakoid membrane lipids and chlorophyll, respectively (Lippold et al., 2012). Thus, we analyzed the expression of PES1 and PES2 under N-sufficient and N-depleted conditions (Figures 4A,B). In WT, the expression of PES1 and PES2 increased markedly under N-depleted conditions (Lippold et al., 2012) (Figures 4A,B). In pah1 pah2, the expression of PES1 and PES2 was significantly higher than that in WT under N-depleted conditions (Figures 4A,B), suggesting that the breakdown of MGDG and chlorophyll was enhanced in pah1 pah2, especially during N starvation, which is consistent with the results in Figure 1. In contrast, in PAH1OE and PAH2OE, the expression of PES1 and PES2 was comparable with that in WT under both N-sufficient and N-depleted conditions (Figures 4A,B).

To clarify the effect of the absence or the overexpression of PAH on galactolipid synthesis, we analyzed the expression of MGDG synthesis genes (MGD1, MGD2, and MGD3, Awai et al., 2001) and DGDG synthesis genes (DGD1 and DGD2, Dörmann et al., 1999; Kelly and Dörmann, 2002) and compared their expression levels among WT, pah1 pah2, PAH1OE, and PAH2OE under both N conditions (Figures 4C–G). In both WT and pah1 pah2, the expression of MGD1 and MGD2 was significantly decreased during N starvation, whereas it remained unchanged in PAH1OE and PAH2OE during N starvation (Figures 4C,D). In WT and pah1 pah2, MGD3 expression remained unchanged or slightly increased during N starvation, whereas its expression in PAH1OE and PAH2OE under N starvation was significantly higher than that under normal conditions (Figure 4E). Thus, although the three genes for MGDG synthesis showed mostly higher expression in PAH1OE and PAH2OE than in WT and pah1 pah2 during N starvation, these expression levels were not remarkably different between WT and pah1 pah2 under any N conditions. On the other hand, although the expression of DGD1 in PAH1OE and PAH2OE under N-depleted conditions seemed higher than that in WT and pah1 pah2 (Figure 4F), most of the expression was similar among plants under any conditions (Figures 4F,G). These results indicated that the absence of PAH has no effect on the expression of MGDG synthases under any N conditions, but the overexpression of PAH increases their expression levels, especially during N starvation.

In the case of [¹⁴C]acetate labeling of Arabidopsis WT seedlings under normal growth conditions, the ¹⁴C label is rapidly incorporated into PC and MGDG (Browse et al., 1986; Xu et al., 2003; Nakamura et al., 2009). Under N-depleted conditions, as well as under N-sufficient conditions, PC was the major ¹⁴C-labeled lipid. However, the second most prevalent labeled lipid was PG in WT (Figure 5A). In pah1 pah2, the labeling percent of PC in the total labeled glycerolipids was markedly higher than that of WT (Figure 5B), which was plausibly caused by the enhanced PC synthesis in pah1 pah2, even under N starvation, or by the repression of PC breakdown and lipid trafficking from the ER to chloroplasts. Accordingly, in pah1 pah2, the relative amounts for labeled chloroplast lipids such as MGDG and PG in the total labeled glycerolipids were lower than in WT (Figure 5B). However, it is noteworthy that the percent of labeled phosphatidylinositol (PI) in pah1 pah2 was about 2-fold higher than that in WT at the beginning and end of the time course (Figures 5A,B), which is consistent with the membrane glycerolipid compositions under N starvation in pah1 pah2 (Figure 3B). In PAH1OE and PAH2OE, the profile of the labeled lipids, except PC and PG, were comparable with that in WT throughout the time course (Figures 5A,C,D). In PAH1OE and PAH2OE, the percent of the labeled PC was decreased by ∼8% during the initial 2 h, whereas the decrease was <5% in WT (Figures 5A,C,D). Moreover, the percent of the labeled PG increased by ∼5% over the time course, whereas that in WT and pah1 pah2 remained unchanged (Figure 5). However, the percent of PC and PG in the membrane glycerolipids were not significantly different among WT, PAH1OE, and PAH2OE (Figures 3A,C,D). Thus, the exchange of the labeled fatty acids between PC and PG might occur in PAH1OE and PAH2OE under N starvation, but the amount is too small to affect the membrane glycerolipid composition.