The galactolipids monogalactosyldiacylglycerol (MGDG) and digalactosyldiacylglycerol (DGDG) are the predominant lipids in thylakoid membranes and indispensable for photosynthesis. Among the three isoforms that catalyze MGDG synthesis in Arabidopsis thaliana, MGD1 is responsible for most galactolipid synthesis in chloroplasts, whereas MGD2 and MGD3 are required for DGDG accumulation during phosphate (Pi) starvation. A null mutant of Arabidopsis MGD1 (mgd1-2), which lacks both galactolipids and shows a severe defect in chloroplast biogenesis under nutrient-sufficient conditions, accumulated large amounts of DGDG, with a strong induction of MGD2/3 expression, during Pi starvation. In plastids of Pi-starved mgd1-2 leaves, biogenesis of thylakoid-like internal membranes, occasionally associated with invagination of the inner envelope, was observed, together with chlorophyll accumulation. Moreover, the mutant accumulated photosynthetic membrane proteins upon Pi starvation, indicating a compensation for MGD1 deficiency by Pi stress-induced galactolipid biosynthesis. However, photosynthetic activity in the mutant was still abolished, and light-harvesting/photosystem core complexes were improperly formed, suggesting a requirement for MGDG for proper assembly of these complexes. During Pi starvation, distribution of plastid nucleoids changed concomitantly with internal membrane biogenesis in the mgd1-2 mutant. Moreover, the reduced expression of nuclear- and plastid-encoded photosynthetic genes observed in the mgd1-2 mutant under Pi-sufficient conditions was restored after Pi starvation. In contrast, Pi starvation had no such positive effects in mutants lacking chlorophyll biosynthesis. These observations demonstrate that galactolipid biosynthesis and subsequent membrane biogenesis inside the plastid strongly influence nucleoid distribution and the expression of both plastid- and nuclear-encoded photosynthetic genes, independently of photosynthesis.
During chloroplast biogenesis, plant cells coordinate several intracellular processes such as gene expression in the nucleus and plastids, protein transport into plastids, and biosynthesis of pigments and lipids. The biogenesis of thylakoid membranes is one of the most prominent processes that occur during chloroplast development. Thylakoid membrane synthesis involves complex processes for assembling photosynthetic proteins and pigments within membrane lipids, which mainly comprise the galactolipids monogalactosyldiacylglycerol (MGDG) and digalactosyldiacylglycerol (DGDG). In chloroplasts, MGDG and DGDG account for approximately 50 and 25 mol% of thylakoid membrane lipids, respectively, and provide a membrane scaffold for photosynthetic pigment-protein complexes (Hölzl and Dörmann, 2007; Mizusawa and Wada, 2011). MGDG has non-bilayer-forming characteristics due to its cone-like shape, whereas DGDG is a bilayer-forming lipid with a cylindrical shape (Lee, 2000). These characteristics may be important for the organization and stabilization of highly stacked thylakoid membranes. In addition to providing the membrane scaffold, galactolipids are also directly involved in the photosynthetic machinery. MGDG is required for thylakoid membrane energization (Aronsson et al., 2008) and energy coupling between light-harvesting chlorophyll complex II (LHCII) and photosystem II (PSII) core complexes (Zhou et al., 2009). Crystallization of cyanobacteria PSI and PSII revealed that MGDG is associated with the cores of the reaction centers, indicating that MGDG may be involved in electron transfer in the photosystem complexes (Jordan et al., 2001; Guskov et al., 2009; Kern and Guskov, 2011; Umena et al., 2011). DGDG has also been found in the crystal structure of cyanobacterial PSII (Guskov et al., 2009; Kern and Guskov, 2011; Umena et al., 2011), and plays an important role in the stability and activity of PSI and PSII complexes (Hölzl and Dörmann, 2007; Hölzl et al., 2009; Mizusawa and Wada, 2011). In addition, this lipid is involved in stabilization of LHCII trimers (Hölzl and Dörmann, 2007; Hölzl et al., 2009).
Because of the great abundance of these lipids in thylakoid membranes, biosynthesis of galactolipids is a critical process for construction of photosynthetic membrane systems and thus chloroplast biogenesis. The synthesis of MGDG is catalyzed by MGDG synthase (MGD), which transfers galactose from UDP-galactose to diacylglycerol (Kobayashi et al., 2009b). DGDG is synthesized by galactosylation of MGDG (Hölzl and Dörmann, 2007), indicating that MGDG synthesis is also required for DGDG production. In Arabidopsis, three MGDG synthases have been identified, and are classified into type A (MGD1) and type B (MGD2 and MGD3) (Kobayashi et al., 2009b). Type A MGD1 is the major isoform and is localized to the inner envelope membranes of chloroplasts, whereas type B MGD2 and MGD3 are very minor proteins that are associated with the outer envelope membranes (Awai et al., 2001). Of the three, MGD1 is responsible for the bulk of galactolipid biosynthesis in photosynthetic tissues. Deficiency of this enzyme causes severe loss of both galactolipids, which leads to the absence of thylakoid membranes and photosynthetic activities (Kobayashi et al., 2007). In contrast, loss of MGD2 and MGD3 causes no abnormal phenotype under nutrient-sufficient conditions, showing that these enzymes do not play an important role under these conditions (Kobayashi et al., 2009a). However, under phosphate (Pi)-deficient conditions, expression of the type B MGD genes is strongly activated, concomitant with up-regulation of the DGDG synthase genes DGD1 and DGD2 (Awai et al., 2001; Kelly et al., 2003; Kobayashi et al., 2004), which results in a dramatic accumulation of DGDG, but not MGDG, in plastidic and extra-plastidic membranes (Härtel et al., 2000). Indeed, the outer envelope pathway mediated by MGD2/3 acts as an alternative pathway for Pi stress-induced production of DGDG, which contributes to plant growth during Pi starvation by substituting membrane phospholipids (Kobayashi et al., 2009a).
Although expression of photosynthetic proteins and the biosynthesis of pigments are expected to be strictly coordinated with the biogenesis of thylakoid membranes, little is known about the mechanism responsible for this coordination. Here, we characterize the Arabidopsis mutant mgd1-2, an MGD1 knockout, under Pi-deficient conditions. In the mutant, activation of the alternative galactolipid pathway by Pi starvation induced partial recovery of membrane biogenesis inside plastids. This was accompanied by various modifications in plastid development and gene expression but did not result in photosynthetic activity. Our data provide new insight into the multiple effects of galactolipid biosynthesis and membrane biogenesis on chloroplast development and photosynthesis.
Pi stress-induced membrane lipid remodeling in mgd1-2
To examine the effects of Pi starvation on galactolipid metabolism in mgd1-2, we first investigated the transcript levels of MGD2 and MGD3 by quantitative RT-PCR. Similar to their expression levels in wild-type, expression levels of MGD2 and MGD3 in mgd1-2 were substantially elevated in response to Pi starvation (Figure 1a). Lipid analysis in the mutant revealed that, whereas DGDG was present only at trace levels under Pi-sufficient conditions, the proportion of DGDG in membrane lipids was drastically increased after Pi starvation (Figure 1b). This result is in agreement with the observation that the alternative galactolipid pathway mediated by MGD2/3 is responsible for Pi stress-induced DGDG accumulation (Kobayashi et al., 2009a). Although MGDG also accumulated in the mgd1-2 mutant during Pi starvation, the proportion of MGDG remained very low, showing that the role of the alternative pathway is essentially confined to the supply of MGDG as a substrate for DGDG synthesis. In addition to galactolipids, the proportion of sulfoquinovosyldiacylglycerol (SQDG) in the mutant increased during Pi starvation, but accounted for only 2.2 mol% of the membrane lipids (Figure 1b). Pi starvation also affected the fatty acid composition of the galactolipids in mgd1-2, but those of other membrane lipids such as SQDG, phosphatidylglycerol and phosphatidylcholine were not significantly changed (Figure S1 and Appendix S1).
Pi starvation induces plastid development in mgd1-2
Loss of galactolipids in mgd1-2 resulted in an albino phenotype and photosynthetic dysfunction (Kobayashi et al., 2007). The mutant had no visible green tissues under Pi-sufficient conditions (Figure 2a, left panel) and accumulated chlorophylls to barely detectable levels (Table 1). However, under Pi-deficient conditions, pale yellow-green leaves were observed in the mutant (Figure 2a, right panel), and chlorophyll levels were markedly increased, with a chlorophyll a/b ratio similar to that in Pi-starved wild-type (Table 1). Concomitant with the chlorophyll accumulation, the amount of LHC proteins (LHCP) also increased in mgd1-2 during Pi starvation, although the level was still very low compared with wild-type (Figure 2b). In the leaves of the Pi-starved mgd1-2 mutant, chlorophyll autofluorescence was detected within the plastids (Figure 2c), but was undetectable under nutrient-sufficient conditions (Kobayashi et al., 2007). Unlike in wild-type seedlings, chlorophyll fluorescence in the Pi-starved mgd1-2 mutant was irregularly granulated. Moreover, leaf cells of the mutant contained fewer and larger plastids than in wild-type cells (Figures 2c and S2), which may reflect a lower dividing activity of the mutant plastids.
Table 1. Chlorophyll (Chl) amount in wild-type and mgd1-2 under phosphate-sufficient (+Pi) or -deficient (−Pi) conditions
Chl (nmol g−1 fresh weight)
Values are means ± SD from four independent experiments.
875.7 ± 59.8
671.2 ± 43.6
204.5 ± 16.2
3.28 ± 0.05
842.4 ± 11.7
617.0 ± 8.61
225.4 ± 6.62
2.74 ± 0.08
0.80 ± 0.46
0.47 ± 0.31
0.34 ± 0.20
1.68 ± 0.93
12.3 ± 2.08
9.07 ± 1.56
3.19 ± 0.53
2.85 ± 0.18
To examine the morphology of mgd1-2 plastids during Pi starvation, transmission electron microscopic analysis was performed. As reported previously (Kobayashi et al., 2007), only abnormal plastids were present in the leaves of the mgd1-2 mutant under Pi-sufficient conditions, with no or severely under-developed internal membrane structures (Figure 3a, left panel, and Figure S3a). However, in the Pi-starved mgd1-2 mutant, thylakoid-like laminated membrane structures were observed within the plastids (Figure 3a, right panel, and Figure S3b). Moreover, invagination of inner envelope membranes, which was observed in the plastids of the mgd1-2 mutant under Pi-sufficient conditions (Figure S4a,b) (Kobayashi et al., 2007), was actively detected in the mutant plastids under Pi-deficient conditions, particularly at the edge of the laminated internal membranes (Figure 3b and S4c-f). These data indicate that Pi stress-dependent galactolipid synthesis induces internal membrane biogenesis inside mutant plastids.
In plastids of the leaves of the mgd1-2 mutant under Pi-sufficient conditions, electron-translucent areas, which correspond to plastid nucleoids (Hashimoto, 1985), were observed in the central region of the stroma and were not detected in the periphery of the organelle (Figure 3a, arrowheads). In contrast, these areas were seen near the laminated internal membrane structures in the mutant plastids under Pi-deficient conditions. Such morphological differences were clear between mgd1 plastids grown under Pi-sufficient and Pi-deficient conditions (Figure S3), suggesting that internal membrane development within the mutant plastids affects the distribution and/or morphology of plastid nucleoids. Nucleoid morphology was also examined using a DNA-specific fluorochrome 4′,6-diamidino-2-phenylindole (DAPI) (Figure 4a,b). Unlike in wild-type protoplasts, nucleoids in the mutant were large and aggregated under Pi-sufficient conditions. However, under Pi-deficient conditions, the mutant nucleoids became smaller and dispersed within the plastid like those in the wild-type. These observations were quantitatively confirmed by measurement of nucleoid diameter in these plastids (Figures 4d and S5).
Photosynthetic capability in mgd1-2 is not restored by Pi starvation
To assess whether Pi stress-induced glycolipid accumulation affects photosynthetic function in mgd1-2, we attempted to determine the photosynthetic characteristics using a pulse-amplitude modulation chlorophyll fluorometer. However, the chlorophyll fluorescence in the mutant was too low to determine the activity even under Pi-deficient conditions. We therefore observed the transient kinetics of chlorophyll fluorescence (Kautsky effect) (Govindjee, 1995). In wild-type, a typical transient curve was observed under both conditions, although the oscillation cycle labeled SMT (steady state, maximum, terminal state) (Govindjee, 1995) was quickened in the Pi-starved seedlings (Figure 5a, left panel). In mgd1-2, no chlorophyll fluorescence was observed under Pi-sufficient conditions. Under Pi-deficient conditions, although faint fluorescence was detected in the mutant, the typical oscillation curve disappeared; instead, only a very slow fluorescence induction was observed (Figure 5a, right panel), demonstrating that the normal electron transfer in PSII was abolished in the mgd1-2 mutant, even when membrane biogenesis and chlorophyll accumulation occurred in the plastid during Pi starvation.
To examine the state of chlorophyll–protein complexes in the mutant, the chlorophyll fluorescence spectra of the seedlings were measured at 77 K. In the emission spectra of wild-type grown under both conditions, shoulder bands at 685 and 695 nm were observed (Figure 5b, left panel), which primarily originate from PSII CP43 and CP47, respectively (Govindjee, 1995). In addition, an emission peak at 735 nm was detected in these seedlings, which may be attributed to the PSI-LHCI complex (Govindjee, 1995). In mgd1-2 mutants grown under Pi-sufficient conditions, no obvious emission peaks were observed, suggesting that no antenna pigment–protein complexes of PSII and PSI were present (Figure 5b, right panel). However, in the mutant grown under Pi-deficient conditions, two emission shoulders were observed at approximately 685 and 700 nm, with a peak at 720 nm (Figure 5b, right panel). The emission band at 685 nm suggests that some components of chlorophyll protein complexes such as CP43 are present in the Pi-starved mgd1-2 mutant. Because aggregates of LHCII in vitro were reported to emit fluorescence at approximately 700 nm (Kirchhoff et al., 2003), the fluorescence shoulder at 700 nm observed in the Pi-starved mgd1-2 mutant may be attributed to emission from abnormally aggregated LHCII. The emission peak at 720 nm observed in the Pi-starved mgd1-2 mutant suggests the presence of PSI reaction center complexes lacking LHCI (Kuang et al., 1984). These results demonstrate that Pi stress-dependent glycolipid synthesis induces formation of chlorophyll–protein complexes in mgd1-2, although their forms are different from those in the wild-type.
We then performed immunoblot analyses of photosynthetic proteins in the mutant (Figure 5c,d). Total membrane proteins (20 μg) from mgd1-2 seedlings grown under Pi-sufficient or -deficient conditions were compared with a dilution series (0.2–20 μg) of total membrane proteins from Pi-sufficient wild-type. Coomassie blue staining of total membrane proteins from the mutant was similar to that from the wild-type except for the large reduction in bands at approximately 30 kDa, irrespective of Pi status (Figure 5c). On the other hand, all photosynthetic proteins investigated in Figure 5(d) were barely detectable in the mutant under Pi-sufficient conditions, consistent with a previous report (Kobayashi et al., 2007). However, the amount of these photosynthetic proteins increased in the mutant during Pi starvation. In particular, increases in the amount of LHCP and an extrinsic PSII protein PsbO were more evident than increases in D1 and D2 (reaction center proteins of PSII) or PsaA/B (reaction center proteins of PSI), suggesting that the galactolipid status in the mutant differentially affects core and peripheral components of the photosystems.
Nuclear- and plastid-encoded gene expression in mgd1-2
It is known that expression of photosynthesis-associated nuclear genes is regulated according to the functional state of plastids (Larkin and Ruckle, 2008). To address whether the defect in galactolipid synthesis in mgd1-2 affects expression of nuclear-encoded photosynthetic genes, we measured the transcript levels of the LHC genes LHCA4 and LHCB6 and the chlorophyll biosynthesis (CHL) genes CHLH, CHL27 and CHLP (Figure 6a), which form a large co-expression gene network with photosynthesis-related genes (Masuda and Fujita, 2008; Kobayashi et al., 2012). We also investigated the expression levels of GLK1 and GLK2, which are genes encoding transcription factors involved in chloroplast biogenesis (Waters et al., 2009). Under Pi-sufficient conditions, the transcript levels of these genes were very low in the mgd1-2 mutant compared with the wild-type, showing that galactolipid deficiency causes dramatic down-regulation of these light-harvesting genes in the mutant. Under Pi-deficient conditions, expression of LHC and CHL was considerably decreased in the wild-type, consistent with the report that photosynthetic genes are down-regulated during Pi starvation (Wu et al., 2003). Expression of GLK2, but not GLK1, was also reduced by Pi starvation. In contrast, in mgd1-2, expression of these genes increased during Pi starvation to a level similar to that in Pi-deficient wild-type plants, consistent with the marked accumulation of chlorophylls upon Pi starvation in the mutant.
To examine whether galactolipid metabolism also affects plastid-encoded gene expression, we investigated the expression levels of psaA, psbA and rpoB in mgd1-2 (Figure 6b). The psaA and psbA genes, which encode the core proteins of PSI and PSII, respectively, are transcribed by plastid-encoded RNA polymerase, whereas rpoB, which itself encodes the RNA polymerase β-chain, is transcribed by nuclear-encoded plastid RNA polymerase (De Santis-MacIossek et al., 1999). In the mgd1-2 mutant under Pi-sufficient conditions, expression of not only the photosynthetic genes psaA and psbA but also the housekeeping gene rpoB was markedly lower than that in the wild-type. However, expression of these genes in the mutant substantially increased after Pi starvation. Next we examined the expression levels of genes encoding sigma factors (SIGs) (Figure 6c), which are nuclear-encoded transcriptional initiation factors required for binding of plastid-encoded RNA polymerase to specific promoters of plastid genes. The Arabidopsis genome encodes six SIGs that are localized in plastids and activate subsets of plastidic gene promoters in a partly redundant manner (Schweer et al., 2010). Under Pi-sufficient conditions, expression of the SIG1, SIG2, SIG4 and SIG6 genes was reduced in the mgd1-2 mutant compared to the wild-type, but SIG5 expression was highly active in the mutant (Figure 6c). Like the nuclear- and plastid-encoded photosynthetic genes, expression of SIG1 and SIG4 was up-regulated in the mutant upon Pi starvation. However, we did not observe a clear correlation between the expression levels of SIG genes and those of plastid-encoded genes overall, possibly due to the redundant roles or post-transcriptional regulation of these factors (Schweer et al., 2010). Our results suggest that galactolipid synthesis and its possible consequence, i.e. membrane biogenesis inside the plastid, globally influence the expression of genes involved in chloroplast biogenesis.
Effects of Pi starvation in chlorophyll biosynthesis mutants
To assess whether Pi starvation generally promotes chloroplast biogenesis in albino mutants, we analyzed the chlh mutant (SALK_062726), which has a null mutation in the CHLH gene encoding a subunit of magnesium chelatase that is essential for chlorophyll biosynthesis (Huang and Li, 2009). We first investigated expression levels of plastid-encoded genes in the chlh mutant (Figure 7a). In contrast to mgd1-2, which showed substantial down-regulation of plastid-encoded genes under Pi-sufficient conditions (Figure 6b), expression of these genes was not markedly reduced in the Pi-sufficient chlh mutant. Moreover, Pi starvation did not increase the expression levels of these genes in this mutant. Differences between chlh and mgd1-2 were also observed in expression of nuclear-encoded genes (Figure 7b). As observed in mgd1-2, the expression of LHCA4, LHCB6 and GLK1 was strongly down-regulated in chlh under Pi-sufficient conditions. However, the reduced expression of these genes was not restored in the chlh mutant by Pi starvation, in clear contrast with the marked up-regulation of these genes in the Pi-starved mgd1-2 seedlings (Figure 6a). We also investigated the expression levels of SIG genes in chlh under Pi-controlled conditions. Under both growth conditions, expression of SIG2 and SIG4 was reduced in this mutant but the SIG5 expression was greatly increased, as observed in the mgd1-2 mutant. Again, we did not observe a clear correlation between the expression levels of the SIG genes and those of plastid-encoded genes.
Because CHLH (also referred to as GENOMES UNCOUPLED5) is one of the causal genes for genome-uncoupled phenotypes, and mutations in CHLH influence plastid signaling pathways, we also analyzed another chlorophyll-deficient mutant, hema1 (SALK_053036), which has a T-DNA insertion in the HEMA1 gene (Woodson et al., 2011) and lacks full-length HEMA1 transcripts (Figure S6a–c). HEMA1 encodes the major isoform of glutamyl-tRNA reductase, catalyzing the biosynthesis of 5-aminolevulinic acid (the first committed step of tetrapyrrole biosynthesis) (Tanaka et al., 2011). Deficiency in HEMA1 resulted in a large reduction in chlorophyll content and a pale-green phenotype (Figure S6d,e), showing a crucial role for this gene in chloroplast biogenesis. As shown in Figure 7(c), the expression patterns of LHCB6, CHL27, psaA and psbA in hema1 were quite similar to those in chlh but substantially different from those in mgd1-2. Moreover, in contrast to mgd1-2, hema1 showed a decrease in chlorophyll content upon Pi starvation (Figure 7d), confirming the clear differences between mgd1-2 and chlorophyll biosynthetic mutants with respect to responses to Pi starvation.
We then analyzed nucleoid morphology in plastids of chlh leaf protoplasts by staining with DAPI (Figure 8a). Consistent with its albino phenotype, no chlorophyll fluorescence was observed in chlh plastids regardless of Pi supply. Within the plastids, small DAPI-stained particles were detected under Pi-sufficient conditions, and the pattern remained unchanged under Pi-deficient conditions. The nucleoid diameter in plastids was even increased in the chlh mutant upon Pi starvation (Figures 8b and S5). These data indicate that dysfunction of chlorophyll biosynthesis in the chlh mutant has less impact on the nucleoid morphology of plastids, in contrast to the dysfunction of galactolipid biosynthesis in mgd1-2.
In this study, we analyzed the MGD1 knockout mutant in Arabidopsis, which is deficient in the major galactolipid biosynthetic pathway located in the inner envelope of plastids. As shown previously (Jarvis et al., 2000; Kobayashi et al., 2007; Aronsson et al., 2008), deficiency in MGD1 causes a loss of galactolipids and results in photosynthetic dysfunction. In contrast, deficiency in both type B isoforms MGD2 and MGD3 located in the outer envelope causes no phenotypic changes under nutrient-sufficient conditions, demonstrating a minor role for these synthases in galactolipid biosynthesis (Kobayashi et al., 2009a). However, under Pi-deficient conditions, the alternative galactolipid pathway was activated, resulting in substantial accumulation of DGDG in extra-plastidic membranes as well as in plastids, which may substitute for phospholipids to counteract the Pi deficiency (Kobayashi et al., 2009b). Our study directly demonstrates that Pi stress-induced DGDG biosynthesis also contributes to photosynthetic membrane biogenesis because the mgd1-2 mutant showed partial membrane formation inside plastids under Pi-deficient conditions. Thus, in addition to its role in membrane lipid remodeling to counteract Pi deficiency, the alternative pathway mediated by MGD2/3 may also function to supply DGDG to photosynthetic membranes. Because MGDG and SQDG levels were slightly increased in the Pi-starved mgd1-2 mutant, these lipids would also contribute to plastidic membrane formation in the mutant. The partial complementation of plastidic membrane formation without functional photosynthetic electron transport allowed us to analyze the function of galactolipids in chloroplast biogenesis.
Involvement of galactolipids in the formation of photosynthetic complexes
In the Pi-starved mgd1-2 mutant, photosynthetic electron transport was severely impaired despite partial restoration of internal membrane biogenesis and chlorophyll accumulation inside plastids, suggesting that the galactolipids that accumulate in response to Pi starvation are insufficient to enable photosynthetic reactions. Mutant analyses in Arabidopsis and cyanobacteria have shown that deficiency of DGDG does not cause complete dysfunction of photosynthetic activities but instead results in partial defects (Kelly et al., 2003; Awai et al., 2007; Sakurai et al., 2007; Hölzl et al., 2009). Moreover, deficiency of SQDG causes no obvious photosynthetic disruption in Arabidopsis (Yu et al., 2002). Because the amount of MGDG was very low in the mgd1-2 mutant even under Pi-deficient conditions, but the proportion of DGDG was greatly increased, the photosynthetic dysfunction in the Pi-starved mutant may be mainly attributed to the deficiency of MGDG in the photosynthetic membranes. In the crystal structure of cyanobacterial PSI, one MGDG molecule is located in proximity to the core of the reaction center (Jordan et al., 2001). Moreover, in the structure of cyanobacterial PSII, 11 MGDG molecules are present in total, and three MGDG molecules are present in the vicinity of QA binding pockets (Guskov et al., 2009). Also, two MGDG molecules are located near the plastoquinone exchange cavity, which may be important for the fast exchange of reduced plastoquinol with fresh plastoquinone (Guskov et al., 2009). Although some variations of the lipid composition in PSII have been reported (Sakurai et al., 2006; Umena et al., 2011), these data indicate the possibility that the very low levels of MGDG in the mgd1-2 mutant result in a deficiency of MGDG molecules in the core of the reaction centers, resulting in impairment of photosynthetic electron transport. It is interesting to note that the abundance of the reaction center proteins D1, D2 and PsaA/B was lower than that of LHCP and PsbO in the Pi-starved mgd1-2 mutant compared to wild-type (Figure 5d), although expression of psaA encoding PsaA and psbA encoding D1 was strongly up-regulated in the mutant during Pi starvation (Figure 6b). The data support our hypothesis that MGDG is involved in the formation or maintenance of reaction centers in photosystems.
The chlorophyll fluorescence at 77 K was also suggestive of a role for galactolipids in the photosynthetic machinery. The irregular emission peaks at 700 and 720 nm in the Pi-starved mgd1-2 mutant, which may be due to aggregation of LHCII and dissociation of LHCI from PSI, respectively, suggest that the formation of LHC–PS complexes is much perturbed in the mutant. Because lack of DGDG does not result in these dramatic changes in the 77 K chlorophyll fluorescence spectra in Arabidopsis (Hölzl et al., 2009) and Synechocystis (Sakurai et al., 2007), deficiency in MGDG is likely to be the major cause of the severe disruption of LHC–PS complexes. Thus, MGDG is indispensable for organization of these complexes. In fact, in vitro assembly analysis has revealed that addition of MGDG to aggregating LHCII induces the formation of stacked lamellar structures, suggesting that MGDG is important for the ordered arrangement of LHCII in thylakoid membranes (Simidjiev et al., 2000). MGDG may also be involved in the energy coupling between LHCII and PSII core complexes by increasing their ability to interact (Zhou et al., 2009). Because MGDG is the only non-bilayer lipid in thylakoid membranes, its unique characteristics may be crucial for the proper assembly of LHC–PS complexes.
Involvement of galactolipids in membrane biogenesis and nucleoid remodeling inside plastids
In plants, galactolipids are formed in the envelope membranes of plastids and transported inside the plastid to construct thylakoid membrane networks (Kobayashi et al., 2009b), but their trafficking pathway remains unclear. We found that invagination of inner envelope membranes, which may indicate an initial stage of thylakoid membrane biogenesis (Vothknecht and Westhoff, 2001; Kobayashi et al., 2007), actively occurred in mgd1-2 plastids under Pi-deficient conditions (Figures 3b and S4). Moreover, inner envelope invagination was occasionally associated with the laminated internal membranes in the mutant plastids, implying a link between envelope invagination and thylakoid membrane biogenesis, as observed during early chloroplast biogenesis (Vothknecht and Westhoff, 2001). Because invagination of the inner envelope was also reported in plastids of Arabidopsis seedlings treated with a galvestine-1, a specific inhibitor of MGDG synthase (Botté et al., 2011), MGDG biosynthesis may be required for the appropriate development of thylakoid membranes through invagination of inner envelope membranes.
During Pi starvation, nucleoid morphology in the mgd1-2 plastid changed concomitantly with internal membrane biogenesis (Figures 3a, 4, S3 and S5). In higher plants, nucleoids are subject to dynamic changes during the development of plastids; whereas nucleoids are found at the center of the proplastid, those in developing plastids and mature chloroplasts are localized close to the envelope membranes and thylakoid membranes, respectively (Hashimoto, 1985; Sato, 2001). Furthermore, the plastid proteins PLASTID ENVELOPE DNA-BINDING (PEND) and MAR BINDING FILAMENT-LIKE PROTEIN1 (MFP1) may anchor nucleoids to envelope membranes in developing chloroplasts (Sato et al., 1993) and thylakoid membranes of mature chloroplasts (Jeong et al., 2003), respectively. Recently, the YLMG1-1 protein, which co-localizes with nucleoids in punctate structures on the thylakoid membranes, was identified in Arabidopsis as a factor required for proper distribution of plastid nucleoids (Kabeya et al., 2010). These data suggest that the morphology and distribution of plastid nucleoids are closely associated with the membrane organization within the plastid. It is likely that membrane biogenesis inside plastids affects the distribution and morphology of nucleoids because, in mgd1-2, nucleoids were found at the central region of the plastid under Pi-sufficient conditions but were localized near the internal membrane structures during Pi starvation (Figure 3a and S3). In fact, similar to our observations, changes in nucleoid morphology were observed in parallel with the disruption of thylakoid biogenesis in white sectors of variegated leaves (Sakamoto et al., 2009). However, the distribution and morphology of plastid nucleoids appeared normal in chlh leaf cells (Figure 8), suggesting that chlorophyll biosynthesis itself has less impact on nucleoid morphology.
Altered composition of galactolipids modifies the expression of plastid-encoded genes
In the mgd1-2 mutant, expression of the plastid-encoded genes psaA, psbA and rpoB was repressed under Pi-sufficient conditions, whereas these genes were highly up-regulated during Pi starvation (Figure 6b). Because nucleoid morphology is associated with global transcriptional activity of plastid genes (Sekine et al., 2002), it is possible that, in mgd1-2, the change in nucleoid morphology during Pi starvation influences the transcriptional activity of plastid genes. This hypothesis is consistent with the fact that the albino chlh mutant, which displayed normal nucleoid distribution in leaf plastids (Figure 8), did not show a marked reduction in the expression of plastid-encoded genes (Figure 7a). Moreover, a recent study revealed that the nuclear-encoded plastid RNA polymerase RPOTmp, which is targeted to both mitochondria and plastids, is tightly associated with thylakoid membranes (Azevedo et al., 2008), suggesting that establishment of thylakoid membranes directly affects the activity of plastid RNA polymerases. Of note, the up-regulation of these plastid-encoded genes is independent of photosynthetic activities, because mgd1-2 showed no functional photosynthetic electron transport even under Pi-deficient conditions (Figure 5a). Because SIG1 and SIG4 showed expression patterns similar to those of plastid-encoded genes in mgd1-2 (Figure 6c), we cannot rule out the possibility that transcriptional changes in these SIG genes influenced the expression of plastid-encoded RNA polymerase-dependent plastid genes.
Altered composition of galactolipids modifies the expression of nuclear-encoded genes
In addition to plastid-encoded genes, nuclear-encoded LHC and CHL genes were strongly down-regulated in mgd1-2, together with their regulatory transcription factor GLK genes under Pi-sufficient conditions (Figure 6a). Down-regulation of photosynthesis-associated nuclear genes was also reported in mutants defective in plastid protein import (Kakizaki et al., 2009). These authors proposed that, when protein import into plastids is blocked, expression of GLK1 is repressed through GENOMES UNCOUPLED1 (GUN1) mediated plastid-to-nucleus retrograde signaling. Consequently, the reduced GLK1 expression results in down-regulation of its targets, namely nuclear photosynthetic genes. As the nuclear genes analyzed here are the targets of plastid signaling (Moulin et al., 2008; Waters et al., 2009), similar mechanisms may function in down-regulation of such genes in mgd1-2. However, the reduced expression of CHL and LHC genes in mgd1-2 was restored to a similar level as wild-type under Pi-deficient conditions. Because Pi starvation did not induce expression of these genes in chlorophyll biosynthesis mutants (Figure 7b,c), it is likely that galactolipid biosynthesis activated by Pi starvation partially compensates for the loss of MGD1 and recovers the reduced expression of nuclear photosynthetic genes. In this context, it is noteworthy that the reduced expression of GLK genes in mgd1-2 was also restored upon Pi starvation. Because the expression of GLK genes directly influences the transcript levels of target genes in Arabidopsis (Waters et al., 2009), the up- and down-regulation of nuclear photosynthetic genes in mgd1-2 in response to Pi conditions may be explained by the changes in expression level of GLK genes. These data are consistent with the hypothesis that GLK genes operate downstream of plastid signaling to synchronize expression of nuclear photosynthetic genes with chloroplast development (Kakizaki et al., 2009; Waters et al., 2009). In the wild-type, GLK2 expression was reduced after Pi starvation, accompanied by down-regulation of LHC and CHL (Figure 6a). Because GLK1 expression was not significantly reduced in the wild-type during Pi starvation, GLK2 may be responsible for regulation of the light-harvesting genes during Pi starvation in the wild-type.
In summary, Pi stress-dependent activation of the alternative galactolipid pathway in mgd1-2 led to membrane biogenesis and accumulation of photosynthetic complexes inside the plastid, accompanied by a change in the nucleoid morphology and up-regulation of photosynthetic genes encoded in both the plastid and nucleus. The concerted regulation between galactolipid biosynthesis and photosynthetic gene expression may be important for synchronizing the formation of photosynthetic complexes with the development of thylakoid membranes during chloroplast biogenesis. Despite the chlorophyll accumulation associated with membrane biogenesis inside the plastid, photosynthetic electron transport was still dysfunctional in the Pi-starved mutant, which may reflect a special requirement for MGDG in construction of the photosynthetic machinery in addition to membrane building.
Plant materials and growth conditions
For all experiments, Arabidopsis thaliana wild-type (Columbia) and the mutants mgd1-2 (Kobayashi et al., 2007), chlh (Huang and Li, 2009) and hema1 (Woodson et al., 2011) were grown in liquid medium containing 1% w/v sucrose with gentle rotation at 23°C under continuous white light (60 μmol photons m−2 sec−1). The mgd1-2 seedlings were grown in the liquid MS medium for 21 days, and then grown in liquid Pi-sufficient (1.0 mm) medium or Pi-deficient (0 mm) medium (Härtel et al., 2000) for another 7 days. For comparative analyses, wild-type and the chlorophyll biosynthesis mutants (chlh and hema1) were grown in liquid MS medium for 5 and 10 days, respectively, and then transferred to Pi-controlled medium for another 7 days, at which time they were at a similar developmental stage as 28-day-old mgd1-2 seedlings.
Quantitative RT-PCR analysis
Total RNA was extracted using an RNeasy plant mini kit (Qiagen, http://www.qiagen.com/). Reverse transcription was performed using an RNA PCR kit version 3.0 (TaKaRa Bio, http://www.takara-bio.com/). cDNA amplification was performed using a SYBR PreMix Ex Taq kit (TaKaRa Bio) and 100 nm gene-specific primers (Table S1). Thermal cycling consisted of an initial denaturation step at 95°C for 10 sec, followed by 40 cycles of 5 sec at 95°C and 30 sec at 62°C. Signal detection and quantification were performed in duplicate using a MiniOpticon (Bio-Rad, http://www.bio-rad.com/). The relative abundance of all transcripts amplified was normalized to the constitutive expression level of ACTIN8 (Pfaffl, 2001).
Total lipids were extracted from seedlings that had been crushed into powder in liquid nitrogen, and were separated using two-dimensional TLC (Kobayashi et al., 2007). Lipids isolated from silica gel plates were methylated, and fatty acid methyl esters were quantified by GLC using myristic acid as an internal standard (Kobayashi et al., 2006).
Seedlings crushed into powder in liquid nitrogen were suspended in 20 mm Tris/HCl buffer (pH 7.5). For Figure 2(b), total protein was denatured and solubilized in SDS-PAGE sample buffer by boiling. For Figure 5(c,d), the membrane protein fraction was precipitated by centrifugation (25 000 g) for 10 min, briefly washed twice with the Tris/HCl buffer and then solubilized in SDS-PAGE sample buffer by incubation at 4°C for 3 h. Protein content was quantified using an RC DC protein assay kit (Bio-Rad). Detection of proteins was performed as described previously (Kobayashi et al., 2007).
Seedlings crushed into powder in liquid nitrogen were homogenized in 80% acetone, and debris was removed by centrifugation at 10 000 g for 5 min. The absorbance of the supernatant at 720, 663, 647 and 645 nm was measured using an Ultrospec 2100 pro spectrophotometer (GE Healthcare, http://www.gehealthcare.com/). The chlorophyll (a and b) concentration of the samples was determined as described previously (Melis et al., 1987).
To detect chlorophyll autofluorescence (Figure 2c), leaf samples were examined using an LSM 510 confocal laser scanning microscope (Zeiss, http://www.zeiss.com/). Transmission electron microscopic analyses were performed as described by Kobayashi et al. (2007) and Masuda et al. (2009). For the analyses in Figures 4 and 8, leaves were cut up using a razor and digested into protoplasts in medium comprising 0.5% w/v Macerozyme R-10 (Yakult, http://www.yakult.co.jp/ypi/en/index.html), 1.0% w/v Cellulase RS (Yakult), 550 mm d-mannitol and 5 mm MES/KOH, pH 5.8, at 23°C for 30 min. Nucleoids were observed using DAPI as described previously (Sekine et al., 2002). Nucleoid size was determined by measuring the diameter of DAPI fluorescence signals using imagej software (http://rsb.info.nih.gov/ij/).
Chlorophyll fluorescence measurements
Chlorophyll fluorescence transients were measured using a FluorCam 700MF fluorescence imaging system (Photon System Instruments, http://www.psi.cz/). Seedlings on agar plates were placed in the imaging system and exposed to actinic light for 20 sec following dark adaptation for 5 min. Measurement parameters were as follows: electronic shutter = 7, sensitivity = 50, irradiance = 100 for mgd1-2, and electronic shutter = 7, sensitivity = 15, irradiance = 30 for wild-type. Values are the means of six seedlings for mgd1-2 and 12 seedlings for wild-type. The fluorescence intensity was normalized to the initial value at the start of actinic light. Low-temperature fluorescence emission spectra of chlorophyll proteins were obtained directly from the intact seedlings in liquid nitrogen. The spectra were recorded using a custom-made apparatus (Sonoike and Terashima, 1994).
Characterization of the hema1 mutant
The hema1 mutant (SALK_053036) was obtained from the Salk Institute Genomic Analysis Laboratory (http://signal.salk.edu/). Genomic DNA was extracted from 21-day-old seedlings using a DNeasy plant mini kit (Qiagen). The genotype analysis in Figure S6(b) was performed by 45 cycles of PCR using HEMA1-specific primers and a T-DNA left border primer (Table S1). For the RT-PCR analysis in Figure S6(c), cDNA fragments prepared using RNA PCR kit version 3.0 (TaKaRa) were amplified by 45 cycles of PCR using primers specific to HEMA1 and HEMA2 (Table S1). The chlorophyll contents in Figure S6(d) were determined in 7-day-old seedlings as described earlier.
We thank Nobuyoshi Mochizuki (Department of Botany, Graduate School of Science, Kyoto University, Japan) for supply of the chlh and hema1 mutants, and Mayumi Wakazaki (RIKEN, Tokyo, Japan) for performing electron microscopy. This work was supported by three Grants-in-Aid for Scientific Research on Priority Areas (numbers 18056007, 20053005 and 22370016) from the Ministry of Education, Sports, Science and Culture in Japan. K.K. was supported by research fellowships from the Japan Society for the Promotion of Science for Young Scientists and by a RIKEN post-doctoral fellowship.