Inducible knockdown of MONOGALACTOSYLDIACYLGLYCEROL SYNTHASE1 reveals roles of galactolipids in organelle differentiation in Arabidopsis cotyledons.

Monogalactosyldiacylglycerol (MGDG) is the major lipid constituent of thylakoid membranes and is essential for chloroplast biogenesis in plants. In Arabidopsis (Arabidopsis thaliana), MGDG is predominantly synthesized by inner envelope-localized MONOGALACTOSYLDIACYLGLYCEROL SYNTHASE1 (MGD1); its knockout causes albino seedlings. Because of the lethal phenotype of the null MGD1 mutant, functional details of MGDG synthesis at seedling development have remained elusive. In this study, we used an inducible gene-suppression system to investigate the impact of MGDG synthesis on cotyledon development. We created transgenic Arabidopsis lines that express an artificial microRNA targeting MGD1 (amiR-MGD1) under the control of a dexamethasone-inducible promoter. The induction of amiR-MGD1 resulted in up to 75% suppression of MGD1 expression, although the resulting phenotypes related to chloroplast development were diverse, even within a line. The strong MGD1 suppression by continuous dexamethasone treatment caused substantial decreases in galactolipid content in cotyledons, leading to severe defects in the formation of thylakoid membranes and impaired photosynthetic electron transport. Time-course analyses of the MGD1 suppression during seedling germination revealed that MGDG synthesis at the very early germination stage is particularly important for chloroplast biogenesis. The MGD1 suppression down-regulated genes associated with the photorespiratory pathway in peroxisomes and mitochondria as well as those responsible for photosynthesis in chloroplasts and caused high expression of genes for the glyoxylate cycle. MGD1 function may link galactolipid synthesis with the coordinated transcriptional regulation of chloroplasts and other organelles during cotyledon greening.

In dicotyledonous plants, cotyledons, which are formed during embryogenesis, initially serve as storage organs during seed germination but mainly function in photosynthesis after seedling establishment. During the developmental switch from heterotrophic to autotrophic growth in germinated seedlings, metabolic activities change greatly in cotyledon cells. Before the development of photosynthetic capacity in cotyledons, the seedlings of oilseed plants such as Arabidopsis (Arabidopsis thaliana) grow heterotrophically depending on triacylglycerol (TAG) stored within oil bodies in cotyledon cells. In this stage, peroxisomes function as the glyoxysome, which converts fatty acids bound to TAG to succinate via b-oxidation and the glyoxylate cycle, to provide carbon sources and energy for growth. After chloroplast development, plants rely on photosynthesis, which converts solar energy into chemical energy and fixes carbon dioxide into carbohydrates. Concomitant with photosynthesis, photorespiration, performed by cooperation among chloroplasts, peroxisomes, and mitochondria, is activated to recycle 2-phosphoglycolate, the product of oxygenation reaction instead of carboxylation by Rubisco (Peterhansel et al., 2010).
Chloroplast biogenesis involves the remarkable development of thylakoid membranes consisting of photosynthetic protein-pigment complexes and the membrane lipid bilayer. In chloroplasts, monogalactosyldiacylglycerol (MGDG) and digalactosyldiacylglycerol (DGDG) account for approximately 50 and 25 mol % of thylakoid membrane lipids, respectively (Block et al., 1983). In addition to providing a lipid matrix for thylakoid membranes, these galactolipids may be required for the structure and function of photosynthetic complexes (Mizusawa and Wada, 2012).
The last step of MGDG synthesis is catalyzed by MONOGALACTOSYLDIACYLGLYCEROL SYNTHASE (MGD), which transfers a Gal moiety of UDP-Gal to diacylglycerol in plastid envelopes (Kobayashi et al., 2009b). Because DGDG is synthesized by the galactosylation of MGDG (Dörmann and Benning, 2002), MGD is the key enzyme for the biosynthesis of both of these galactolipids and, therefore, the biogenesis of thylakoid membranes. Three MGD isoforms identified in Arabidopsis are MGD1, MGD2, and MGD3 (Kobayashi et al., 2009b). The contribution of MGD2 and MGD3 to galactolipid synthesis is limited; the isoforms are localized at the outer envelope membrane of chloroplasts (Awai et al., 2001) and affect galactolipid synthesis only under phosphatelimited conditions (Kobayashi et al., 2009a). By contrast, MGD1 is targeted to the chloroplast inner envelope membrane and is expressed actively in photosynthetic tissues (Awai et al., 2001). The significance of MGD1 in galactolipid biosynthesis and thylakoid membrane biogenesis was demonstrated by the study of two Arabidopsis mgd1 mutants: the knockdown mutant mgd1-1 (Jarvis et al., 2000) and the knockout mutant mgd1-2 (Kobayashi et al., 2007). The mgd1-1 mutant reduces MGDG to 58% of that in wild-type plants and has chloroplasts with fewer thylakoid membranes (Jarvis et al., 2000). These defects result in impaired thylakoid membrane energization and decreased photoprotective capacity (Aronsson et al., 2008). Similar results were recently reported in an MGD1 knockdown mutant (M18) in tobacco (Nicotiana tabacum), which further revealed an involvement of MGD1-mediated MGDG biosynthesis in photosynthetic electron transport between PSII and PSI (Wu et al., 2013). However, the mgd1-2 knockout mutant accumulated only negligible amounts of both MGDG and DGDG, thus resulting in the absence of thylakoid membranes in leaf plastids and the complete dysfunction of photosynthetic activities (Kobayashi et al., 2007).
Recently, we revealed that the loss of galactolipids in mgd1-2 caused strong down-regulation of both plastidand nucleus-encoded photosynthesis-associated genes, whereas partial complementation of the mutation by the alternative MGD2/MGD3 pathway under phosphate starvation attenuated the down-regulation of such genes (Kobayashi et al., 2013). Our findings suggest that galactolipid biosynthesis plays a crucial role in coordinating the formation of photosynthetic proteinpigment complexes with the development of thylakoid membranes during chloroplast biogenesis. However, how the genes and processes associated with photosynthesis are intertwined with the development of thylakoid membrane bilayers remains largely unknown because of the strong pleiotropic effects of the constitutive loss of galactolipids on plant growth and development. Furthermore, no detailed analysis has been reported on the effect of galactolipid biosynthesis and consequent thylakoid membrane biogenesis on the differentiation of other organelles such as peroxisomes and Figure 1. Characterization of the amiR-MGD1 seedlings. A, Schematic representation of an inducible amiR-MGD1 construction. The artificial microRNA precursor fragment designed to target MGD1 was cloned under six times repeat of the pOp promoter. The rat glucocorticoid receptor (GR) fused to the transcription activator LhG4, which consists of a high-affinity DNA-binding mutant of the lac repressor (lacI His-17 ) fused to transcriptionactivation domain II of GAL4 from Saccharomyces cerevisiae (Gal4-II), was cloned under the control of the cauliflower mosaic virus (CaMV) 35S promoter. amiR-MGD1 was designed to target 20 nucleotides across the stop codon of MGD1 mRNA. LB, Left border; RB, right border; ter, transcriptional terminator; UTR, untranslated region. B, Photograph of wild-type (WT) and amiR-MGD1 (L2, L4, L5, and L7) seedlings grown on DEX-free (2DEX) or 10 mM DEX-containing (+DEX) medium. C, Chlorophyll content per seedling in each amiR-MGD1 line grown on 2DEX or +DEX medium. The horizontal line in each box represents the median value of the distribution. The top and bottom of each box represent the upper and lower quartiles, respectively. The whiskers represent the range (n = 24). D, Expression of MGD1 in the wild type and each amiR-MGD1 line under +DEX and 2DEX conditions. E, Expression of three MGD genes in phosphatedeficient L4 seedlings grown under +DEX and 2DEX conditions. Data are fold differences from the +DEX wild type in D and from the 2DEX control in E (means 6 SE, n = 3). Plants used in B to E were all 5-d-old seedlings. mitochondria during leaf development, although a tight metabolic coordination occurs between chloroplasts and these organelles during the transition from heterotrophic to photoautotrophic growth in leaf cells (Peterhansel et al., 2010;Nunes-Nesi et al., 2011).
To gain insight into the role of galactolipid biosynthesis through MGD1 in the coordinated formation of photosynthetic complexes and thylakoid membranes, we constructed an artificial microRNA gene (amiR-MGD1) targeting Arabidopsis MGD1 under the control of the dexamethasone (DEX)-inducible promoter and introduced it into the Arabidopsis Landsberg erecta (Ler) ecotype. The resulting plants showed a wide phenotypic variation in cotyledon development with strongly reduced MGD1 expression with DEX treatment, which allowed us to analyze the function of the MGD1-mediated galactolipid biosynthesis during the initial stage of chloroplast biogenesis. Moreover, we investigated a linkage between MGD1 function and gene expression associated with peroxisomal and mitochondrial functions during early seedling development.

Specific Suppression of MGD1 by the amiR-MGD1 Transgene Is DEX Dependent
To control the timing of MGD1 expression during chloroplast development, we used a microRNA-mediated gene suppression technique (Schwab et al., 2006) combined with a glucocorticoid-inducible system (Craft et al., 2005;Fig. 1A). We obtained seven DEX-responsive lines from 48 individual transgenic Arabidopsis lines that harbor amiR-MGD1. We chose lines 2, 4, 5, and 7 (L2, L4, L5, and L7) for subsequent analyses. First, we examined the effect of continuous DEX treatment on the early seedling growth of amiR-MGD1 transgenic lines. The phenotypes of all amiR-MGD1 lines were indistinguishable from the wild type in the absence of DEX (Fig. 1B). In the presence of 10 mM DEX, cotyledon greening was impaired in L2, L4, and L5, although a wide variety of color phenotypes, from albino-like to wild-type-like, were observed in these lines. Meanwhile, L7 showed no remarkable difference from the wild type in visible phenotype even in the presence of 10 mM DEX. Chlorophyll measurements in 5-d-old seedlings supported these results (Fig. 1C). Although the chlorophyll content deviated considerably in each line, the median values were substantially lower in DEX-treated L2, L4, and L5 seedlings than in the DEX-untreated control of each line. However, chlorophyll content was almost the same in DEX-treated L7 seedlings and the DEX-untreated control.
To address the relationship between the amiR-MGD1 phenotypes and the suppression of MGD1 expression, we examined the expression of MGD1 in 5-d-old amiR-MGD1 seedlings by quantitative reverse transcription (qRT)-PCR (Fig. 1D). In the absence of DEX, no lines showed decreased MGD1 expression, which confirms that the amiR-MGD1 transgene does not suppress MGD1 without DEX treatment. By contrast, in the presence of 10 mM DEX, MGD1 expression decreased by 70% to 80% in L2, L4, and L5 seedlings as compared with the wild type. Even in L7 seedlings, which did not show reduced chlorophyll content with DEX treatment, MGD1 expression was decreased by half of the wild-type level with DEX treatment (Fig. 1D).
Because Arabidopsis possesses two paralogs of MGD1, namely MGD2 and MGD3 (Kobayashi et al., 2009b), we also examined whether the amiR-MGD1 transgene targets the expression of MGD2 and MGD3 in L4 seedlings. Under phosphate-sufficient normal growth conditions, the expression of MGD2 and MGD3 was too low to quantify in both the wild type and amiR-MGD1 lines. We then grew seedlings under the phosphate-deficient condition to up-regulate MGD2 and MGD3, as described previously (Awai et al., 2001;Kobayashi et al., 2004). In phosphate-deficient L4 seedlings, DEX treatment suppressed the expression of MGD1 but not MGD2 or MGD3 (Fig. 1E). Therefore, the amiR-MGD1 construct specifically targeted the MGD1 transcript for suppression and had no effect on the expression of MGD2 and MGD3. We confirmed the DEX responsiveness and target specificity of the amiR-MGD1 system.

Tight Link between Chlorophyll Accumulation and Galactolipid Biosynthesis in amiR-MGD1 Seedlings
To further analyze the phenotypes of L2 and L4, we classified DEX-treated seedlings of these lines into three groups, namely white, pale green, and green, using a chlorophyll-deficient hemA1 mutant (Kobayashi et al., 2013) as a color reference. The white seedlings had paler cotyledons than those of hemA1 mutants, whereas the cotyledon color of green seedlings was similar to that of DEX-untreated seedlings ( Fig. 2A). The remaining seedlings, with intermediate phenotypes between white and green seedlings, were defined as pale green, which were the majority in both L2 and L4 (Fig. 1B). This classification was supported by quantifying chlorophyll content in each class of DEX-treated seedlings (Fig. 2B).
We compared the expression of MGD1 in white and green seedlings grown for 5 d in the presence of DEX and in DEX-untreated controls (Fig. 2C). In both L2 and L4 lines, MGD1 expression was lower in white seedlings than in green seedlings. Together with the results in Figure 1D, our data show a link between the MGD1 suppression and impaired cotyledon greening and support our previous conclusion that galactolipids are necessary for chloroplast development (Kobayashi et al., 2007). Why a wide variation in cotyledon phenotypes is found in L2, L4, and L5 remains elusive; the amiR-MGD1 activity may differ individually even within a single line, and the small differences in the amiR-MGD1 activity may lead to the considerable disparity in chlorophyll accumulation and chloroplast development in cotyledons.
Next we analyzed galactolipid content in each color class of 5-d-old DEX-treated L4 seedlings and in the untreated control (Fig. 2D). The proportion of MGDG in total glycerolipids decreased by 25%, 77%, and 85% in green, pale-green, and white seedlings, respectively, as compared with the DEX-untreated control. The MGDG level in these seedlings was associated with their chlorophyll content (Fig. 2B), which suggests that MGDG biosynthesis strongly affects chlorophyll accumulation and thus the formation of photosynthetic machinery, presumably by varying the size and/or quality of thylakoid membrane bilayers. In green seedlings, the proportion of DGDG did not change, and thus the MGDG-DGDG ratio decreased. These findings agree with the previous reports on the Arabidopsis mgd1-1 mutant and the tobacco M18 line showing that the decrease in MGD1 activity primarily results in a loss of MGDG (Jarvis et al., 2000;Wu et al., 2013). A possible explanation for the decrease in MGDG-DGDG ratio by partial defects of the MGD1 function is that the decreased MGD1 activity results in an overbalance of DGDG biosynthesis relative to MGDG biosynthesis and consequent consumption of MGDG in the formation of DGDG, thus leading to a reduced distribution of MGDG to membrane bilayers. In the pale-green and white seedlings, the proportion of DGDG decreased together with MGDG, which reduced the proportion of total galactolipids by 80% in white seedlings. The ratio of non-bilayer-forming MGDG to bilayer-forming DGDG may affect properties of chloroplast membranes and so is tightly regulated through a yet unknown mechanism (Dörmann and Benning, 2002). The inhibition of MGD1 expression may induce the coordinated down-regulation of DGDG biosynthesis in the paler amiR-MGD1 seedlings so that the MGDG-DGDG ratio is maintained in photosynthetic membranes.
In addition to examining membrane lipids, we analyzed the levels of TAG in these seedlings. The proportion of TAG was 1.5-fold higher in white L4 seedlings than in the DEX-untreated control, while the proportion in the green and pale-green seedlings was unchanged (Fig. 2D). As in the wild type, TAG in white seedlings contained eicosenoic acid (20:1) as a major constituent (Supplemental Fig. S1), which may reflect a high retention of seed TAG in white seedlings after germination, because eicosenoic acid is specifically accumulated in TAG during seed maturation (Li et al., 2006). The strong MGD1 suppression in white seedlings may decrease the demand of diacylglycerol for galactolipid biosynthesis, thereby influencing TAG metabolism. However, unlike white seedlings, pale-green seedlings showed no change in TAG content, despite the substantial decrease in galactolipid content. Therefore, changes in other metabolic processes caused by severe defects in chloroplast biogenesis may affect TAG metabolism in white seedlings.

Impaired Chloroplast Biogenesis and Cotyledon Cell Organization with MGD1 Suppression
Partial deficiency of MGDG by MGD1 knockdown leads to decreased amounts of thylakoid membranes with altered architecture (Jarvis et al., 2000;Wu et al., 2013), whereas crucial lack of both galactolipids by MGD1 knockout results in no or severely underdeveloped internal membrane structures in leaf plastids (Kobayashi et al., 2007). We examined the effect of MGD1 suppression on chloroplast development in green and white cotyledons from DEX-treated L4 seedlings by transmission electron microscopy. Consistent with previous reports (Jarvis et al., 2000;Wu et al., 2013), chloroplasts in green cotyledons, with slightly reduced MGDG content (Fig. 2D), appeared a little less mature than those from DEX-untreated controls (  et al., 2007) and more severe than in the mgd1-1 knockdown mutant (Jarvis et al., 2000), agrees with their phenotype in galactolipid content, which indicates a direct impact of the MGD1-derived MGDG synthesis on the development of thylakoid membrane networks. In addition to the abnormal chloroplasts, oil bodies were frequently observed in close contact with glyoxysomes in the cotyledon cells of the white seedlings (Supplemental Fig. S2, G and H), which is consistent with the high TAG level in the seedlings.
The MGD1 suppression also influenced the structure and organization of cotyledon cells (Fig. 3, D-F). Both epidermal and mesophyll cells in the L4 white cotyledons had irregular shapes, often with indented cell outlines, whereas those in the green cotyledons appeared similar to the DEX-untreated control. In the white cotyledons, the alignment of mesophyll cells was disordered, with irregular organization of palisade cells and spongy cells. Moreover, the size of mesophyll cells in the white cotyledons was smaller than in the green cotyledons of DEX-treated and DEX-untreated seedlings. Because seeds of all amiR-MGD1 lines were collected from parents grown in the absence of DEX, and therefore the embryonic development progressed normally in all these lines, the morphological disorder of the white cotyledons could be attributed to impaired postgerminative development. Considering that cell proliferation for cotyledon formation is nearly completed during embryogenesis and the cell expansion mainly contributes to the postgerminative cotyledon growth in the light (Tsukaya et al., 1994;Stoynova-Bakalova et al., 2004), the morphological disorder and the dwarf phenotype in the white L4 cotyledons may be due to distorted mesophyll cell expansion during postgerminative growth. The distorted organization of mesophyll cells has been observed in several mutants with defective chloroplast development (Wang et al., 2000;Wycliffe et al., 2005;Garcion et al., 2006;Sulmon et al., 2006), including the mgd1-2 mutant (Kobayashi et al., 2007), thus showing a tight developmental link between chloroplasts and host mesophyll cells. Furthermore, application of norflurazon, which inhibits carotenoid biosynthesis and impairs chloroplast development, to Arabidopsis seedlings was recently found to affect the transition of leaf cells from the proliferation stage to the expansion stage, thus resulting in decreased cell size in young true leaves (Andriankaja et al., 2012). Our data reveal that inhibition of cell expansion also occurred in cotyledons with impaired MGDG biosynthesis during postgerminative growth, which suggests that the thylakoid membrane biogenesis and consequent chloroplast development play a role in regulating cell development even in cotyledons with basic structure established during embryogenesis.

Requirement of Galactolipids for Photosynthetic Electron Transport
A partial loss of MGDG (approximately 50% of wildtype levels) results in decreased intersystem electron transport between PSII and PSI (Wu et al., 2013) and increased conductivity of thylakoid membranes (Aronsson et al., 2008), whereas the absence of MGDG (approximately 5% of wild-type levels) causes severe disorder in photosystem complexes and complete deficiency of photosynthetic electron transport (Kobayashi et al., 2013). Because the amiR-MGD1 L4 seedlings showed wide variation in MGDG content, from 15% to 75% of wild-type levels with DEX treatment (Fig. 2D), we used this line and the similar L2 line to investigate the relationship between MGDG biosynthesis and the functionality of the photosynthetic electron transport chain. The maximum quantum efficiency of PSII in the dark (F v /F m ), which represents the intrinsic photosynthetic efficiency of PSII, and the actual PSII efficiency (F II ) under light (36 mmol photons m 22 s 21 ) were determined in individual cotyledons by using a chlorophyll fluorescence imaging system based on pulseamplitude modulation techniques. Maximal chlorophyll fluorescence after dark adaptation (F m ) was used as a measure of chlorophyll accumulation in cotyledons to indicate the severity of amiR-MGD1-mediated suppression of galactolipid biosynthesis. Unlike in the wild type, both F v /F m and F II were decreased in L2 and L4, being correlated with a decrease in F m values (Fig. 4, A and B). These result suggests that the galactolipid biosynthesis through MGD1 is closely associated with the efficiency of light utilization in PSII. Very similar scatter diagrams between L2 and L4 indicate that these lines are mostly equivalent.
To elucidate the influence of MGD1 deficiency on photosynthetic electron transport in cotyledons in more detail, several photosynthetic parameters were characterized in three color classes of 5-d-old DEX-treated L2 seedlings. F v /F m was most decreased in white seedlings, followed by pale-green and green seedlings (Fig. 4C). A dgd1 mutant showed a slight decrease in F v /F m (approximately 0.74) as compared with the wild type (approximately 0.81), although the mutant contains only approximately 10% of the wild-type level of DGDG (Hölzl et al., 2009), which indicates that the loss of DGDG has only a minor impact on the maximal PSII efficiency. Thus, in the white amiR-MGD1 cotyledons, the deficiency in MGDG may be primarily responsible for the substantial decrease in maximal PSII efficiency.
Next, we analyzed quantum yields of PSII under low photosynthetically active radiation (45 mmol photons m 22 s 21 ). Whereas F II in wild-type and green seedlings increased within 2 min after actinic illumination started, that in white and pale-green seedlings remained very low during the measurement (Fig. 4D). A similar pattern was observed in the coefficient of photochemical quenching (qP; Fig. 4E). Because qP represents the redox state of primary electron-accepting plastoquinone of PSII (Q A ) and thus the openness of PSII, the low qP in white and pale-green seedlings suggests acceptor-side limitation in PSII resulting from defective electron transfer downstream of PSII. MGDG deficiency decreased levels of the cytochrome b 6 f complex and blocked the intersystem electron transport in the tobacco MGD1 mutant (Wu et al., 2013). Therefore, the PSII acceptorside limitation in the amiR-MGD1 cotyledons may be attributed at least in part to the inhibition of intersystem electron transport around the cytochrome b 6 f complex. In addition, the maximum quantum efficiency of open PSII (F v 9/F m 9) was constantly lower for the white and pale-green seedlings than for the wild-type and green seedlings (Fig. 4F). The low F v 9/F m 9 indicates dysfunctional photochemical reactions in open PSII and is in line with the low F v /F m in these seedlings (Fig. 4C). In the white and pale-green seedlings, defective electron transfer both within and downstream of PSII could cause decreased light utilization for photosynthesis represented in the low F II .
We also evaluated the quantum yield of light-induced nonphotochemical quenching (F NPQ ) and the quantum yield of non-light-induced nonphotochemical quenching (F NO ), which represent the proportion of regulated and nonregulated dissipation of light energy in PSII, respectively (Kramer et al., 2004). In both white and pale-green seedlings, the low F II was inversely related to the high F NO (Fig. 4G), whereas F NPQ stayed at low levels ( Fig. 4H), which indicates that severe galactolipid deficiency causes nonregulated dissipation of excess light energy from PSII even under low light. The high value of F NO shows the limited photoprotective capacity in DEX-treated amiR-MGD1 cotyledons, consistent with the observation of accelerated photodamage in the tobacco MGDG-deficient mutant under high-light conditions (Wu et al., 2013). Although pale-green seedlings retained a certain amount of DGDG compared with the strong reduction in MGDG, the seedlings showed severe photosynthetic defects similar to white seedlings. Thus, deficiency of MGDG may be the major cause of photosynthetic dysfunction in these seedlings.
To dissect the functionality of PSII in the white cotyledons of amiR-MGD1 lines, we analyzed the transient kinetics of chlorophyll fluorescence in a logarithmic timing series (Fig. 4I). Wild-type cotyledons showed a typical polyphasic fluorescence increase called the origin-inflection-intermediary peak-peak (OJIP) transient (Govindjee, 1995). The O-J phase and the J-I phase reflect the photochemical process in PSII and the reduction in the plastoquinone pool, respectively, whereas the I-P phase is related to the process in PSI. White cotyledons of L2 and L4 showed a very fast increase in chlorophyll fluorescence during the photochemical phase (the O-J phase), which indicates impaired electron transfer from excited chlorophylls downstream within the PSII. In the presence of 3-(3,4-dichlorophenyl)-1,1dimethylurea (DCMU), a rapid reduction of total Q A occurs by illumination due to the inhibition of electron transfer from Q A to the secondary electron-accepting plastoquinone of PSII, thus resulting in a fast increase in chlorophyll fluorescence without inflection, as observed in wild-type cotyledons (Fig. 4J). The fluorescence increase in the presence of DCMU was much faster for L2 and L4 than for the wild type, which suggests decreased electron-accepting capacity of Q A in these transgenic lines. The limitation in electron-accepting capacity within PSII decreases the utilization of light energy for photosynthesis and thus increases excess energy, which would be dissipated in nonregulated forms of heat and fluorescence in DEX-treated amiR-MGD1 cotyledons, as indicated by the high F NO value (Fig. 4G). Crystallographic studies in cyanobacteria have revealed that many galactolipid molecules reside in the PSII complex and that some are present in the vicinity of the reaction center (Guskov et al., 2009;Umena et al., 2011). Therefore, deficient galactolipids in the PSII reaction center may be involved in the dysfunction of the PSII photochemical reaction in DEX-treated amiR-MGD1 cotyledons, as was proposed for the mgd1-2 mutant (Kobayashi et al., 2013).

Role of MGD1 in the Formation of Photosystems
Previously, we revealed by chlorophyll fluorescence analysis at 77 K that the lack of MGDG in the Arabidopsis mgd1-2 knockout mutant caused severely disordered formation of both PSII and PSI complexes (Kobayashi et al., 2013). Meanwhile, knockdown of MGD1 in Arabidopsis and tobacco had only a slight impact on chlorophyll fluorescence spectra at 77 K (Aronsson et al., 2008;Wu et al., 2013). To evaluate the importance of galactolipids in the formation of photosystem complexes, we determined chlorophyll fluorescence spectra at 77 K in white cotyledons of L2 and L4, whose phenotypes are intermediate between the mgd1-2 knockout mutant (Kobayashi et al., 2007) and the knockdown mutants (Aronsson et al., 2008;Wu et al., 2013). In the wild type, three major emission peaks were detected at 682, 690, and 729 nm (Fig. 5A). The peak at 729 nm is attributed to the PSI associated with lightharvesting complex I (LHCI), whereas the peaks at 682 and 690 nm primarily originate from CP43 and CP47 in PSII, respectively (Govindjee, 1995). In white cotyledons of L2 and L4, peak wavelengths did not differ from the wild type, except for a very slight shift of the peak at 732 nm, which suggests that the global formation of photosystem complexes is not largely perturbed in the L2 and L4 white cotyledons. When fluorescence intensities were normalized to the emission maximum at 682 nm among these samples, emission peaks at 729 nm were lower for both L2 and L4 than for the wild type. However, the fluorescence intensity from PSII can vary by chlorophyll concentration in samples because of self-absorbance of the PSII emission by the PSI complex. Therefore, next we compared 77 K chlorophyll fluorescence spectra in membrane fractions from each color class of DEX-treated L2 seedlings at the same chlorophyll concentration (0.8 mg mL 21 ; Fig. 5B). Fluorescence from PSI largely decreased in white and pale-green seedlings compared with the wild type (Fig. 5B), which suggests a decrease in PSI proteins or reduced energy transfer to the PSI reaction center from antenna complexes. Even in green seedlings, which had reduced MGDG but a wild-type level of DGDG, the emission from PSI decreased, which suggests a specific requirement of MGDG for the formation of the PSI-LHC complexes.
Moreover, emission peaks from PSII and PSI were both blue shifted in L2 seedlings in correlation with the extent of the MGD1 suppression. This result suggests that LHCs in membrane fractions from the L2 seedlings are largely dissociated from core complexes in both photosystems. Considering that emission peaks were not blue shifted in intact white seedlings (Fig. 5A), LHCs may interact weakly with both photosystems in vivo but be dissociated from core complexes during membrane fractionation. Indeed, an in vitro analysis revealed that MGDG intensifies the physical interactions between LHCII and PSII core complexes and increases their energy coupling (Zhou et al., 2009). Thus, galactolipids, and particularly MGDG, may be essential for strong interactions between photosystem core complexes and LHCs.
To address whether the MGD1 suppression affects the abundance of membrane photosynthetic proteins, we performed immunoblot analysis in DEX-treated L2 palegreen seedlings. Total membrane proteins (20 mg) from pale-green seedlings were analyzed together with a dilution series (1, 5, and 20 mg) of membrane proteins from the wild type. Although both PSI (PsaA, PsaB, LHCA1, and LHCA3) and PSII (D1, D2, LHCB1, and LHCB3) proteins largely decreased in pale-green seedlings compared with the wild type, the balance of protein abundance between PSI and PSII was maintained. Thus, the reduced PSI fluorescence at 77 K (Fig. 5B) is not due to a decrease in PSI protein levels but presumably due to reduced energy transfer to the PSI reaction center from LHCI or from the PSII antenna system, both of which may be dissociated from reaction centers in the membrane fraction from pale-green seedlings (Fig. 5B).
The immunoblot analysis also revealed that core proteins (PsaA/B, D1, and D2) were more decreased than antenna proteins (LHCA1, LHCA3, LHCB1, and LHCB3) in pale-green seedlings. Similar results were observed in the phosphate-deficient mgd1-2 mutant (Kobayashi et al., 2013). Thus, MGDG may be required for the accumulation or maintenance of photosystem core complexes.

Involvement of Galactolipid Biosynthesis in Coordinated Regulation among Genes Associated with Photosynthesis, Photorespiration, and the Glyoxylate Cycle during Cotyledon Development
We recently reported in the mgd1-2 mutant that galactolipid biosynthesis and subsequent membrane biogenesis inside the plastid strongly influence the expression of both plastid-and nucleus-encoded photosynthetic genes independent of photosynthesis (Kobayashi et al., 2013). To investigate the effect of the MGD1 suppression on photosynthetic gene expression during cotyledon development, we examined the expression of psaA and psbA, which are plastid-encoded genes for PsaA and D1 of the PSI and PSII core complexes, respectively. Consistent with the substantial decrease in PsaA and D1 proteins (Fig. 5C), both psaA and psbA were down-regulated in white cotyledons and green cotyledons of DEX-treated L4 seedlings (Fig. 6A). In developing plastids, a large DNA-protein complex named nucleoids exists in close contact with envelope and thylakoid membranes (Sato, 2001). Considering that plastid transcription is regulated by the activity of the RNA polymerase complexes and by the structural organization of plastid DNA (Sekine et al., 2002), the galactolipid biosynthesis in the plastid envelope and subsequent thylakoid membrane biogenesis may affect transcriptional activities in the nucleoid. In fact, we previously showed that galactolipid biosynthesis modifies the morphology of nucleoids in leaf plastids, which suggests an association between galactolipid biosynthesis and nucleoid activity (Kobayashi et al., 2013).
We then examined the expression of LHCB6 and CHLH, which are nuclear genes encoding LHCII subunit 6 and the H subunit of magnesium chelatase involved in chlorophyll biosynthesis, respectively. Coordinately with plastid-encoded genes (psaA and psbA), nucleus-encoded photosynthetic genes (LHCB6 and CHLH) were downregulated in both white and green cotyledons of DEXtreated L4 seedlings (Fig. 6A). Consistent with their galactolipid phenotypes, the suppression of both plastidand nucleus-encoded genes was stronger in white than in green cotyledons, which suggests that MGDG A and B, The 77 K chlorophyll fluorescence spectra in wild-type (WT) and white seedlings of L2 and L4 in the presence of DEX (A) and in the membrane fraction (0.8 mg mL 21 chlorophyll) from the wild type and each color class of DEX-treated L2 seedlings (B). C, Immunoblot analysis of membrane photosynthetic proteins from wild-type and L2 pale-green seedlings in the presence of DEX. A 20-mg aliquot of total membrane protein from the L2 seedlings was compared with a dilution series (20, 5, and 1 mg) of membrane proteins from the wild type.
biosynthesis is tightly linked to photosynthetic gene expression during chloroplast biogenesis. Inhibition of chloroplast protein translation results in down-regulated nuclear photosynthetic gene expression during early seedling growth, which suggests an involvement of plastid translation in plastid signaling that down-regulates nuclear photosynthetic gene expression in response to chloroplast dysfunction (Nott et al., 2006). Furthermore, plastid gene expression mediated by nucleus-encoded s factors plays a role in plastid signaling (Woodson et al., 2013). Therefore, one possibility is that the decreased expression of plastid-encoded genes in DEX-treated amiR-MGD1 seedlings triggers the down-regulation of photosynthetic genes in the nucleus through plastid signaling pathways.
During cotyledon greening, peroxisomes and mitochondria transform their functions in concert with chloroplast development (Hayashi and Nishimura, 2006). To examine whether the MGD1 suppression affects organelle differentiation early during seedling development, we analyzed the expression of genes associated with peroxisomal and mitochondrial functions in 5-d-old L4 seedlings (Fig. 6, B and C). GOX1 and HPR1, which encode glycolate oxidase and hydroxypyruvate reductase of the photorespiratory pathway, respectively, were used as leaf peroxisome markers. In addition, GLDP1 and SHM1, which encode the P protein of the Gly decarboxylase complex and Ser hydroxymethyltransferase, respectively, were used as markers for the functionality of photorespiration in mitochondria. All of these photorespiratory genes are reported to be specifically expressed in photosynthetic tissues (Kamada et al., 2003;Voll et al., 2006;Engel et al., 2007;Timm et al., 2008). Meanwhile, ICL and MLS, which encode isocitrate lyase and malate synthase of the glyoxylate cycle, respectively, were used as markers for glyoxysomes. ICL and MLS are transiently expressed during the very earliest postgerminative growth stages, when peroxisomes actively operate the glyoxylate cycle as glyoxysomes (Eastmond et al., 2000;Cornah et al., 2004). Our qRT-PCR analysis revealed that the photorespiratory genes GOX1, HPR1, GLDP1, and SHM1 were down-regulated concomitantly with the photosynthetic genes by DEX treatment, particularly in white seedlings (Fig. 6B). The photorespiratory metabolism is associated with photosynthetic functionalities (Peterhansel et al., 2010); therefore, co-down-regulation of the photorespiratory genes with photosynthesisassociated genes is reasonable.
Because many photorespiratory genes are downregulated together with photosynthesis-associated nuclear genes in norflurazon-treated seedlings through plastid signaling pathways (Strand et al., 2003), the MGD1 suppression may down-regulate these genes coordinately with nuclear photosynthetic genes by activating plastid signaling. By contrast, the expression of the glyoxysomal genes ICL and MLS remained at higher levels in both DEX-treated green and white seedlings than in the untreated control (Fig. 6C). The low expression of photorespiratory genes and the high expression of glyoxylate cycle-associated genes suggest that the differentiation of peroxisomes and mitochondria to leafspecific types during cotyledon development is inhibited by amiR-MGD1-mediated suppression of galactolipid biosynthesis. Considering that many glyoxysomes were observed close to oil bodies in 5-d-old white L4 cotyledon cells (Supplemental Fig. S2, G and H), which is typically observed during the early seed germination stage in the case of the wild type (Hayashi et al., 2001), the white plants may keep the b-oxidation and the glyoxylate cycle activities high to maintain heterotrophic growth in response to disrupted chloroplast biogenesis. Of note, we observed high expression of glyoxysomal ICL and MLS genes in green DEX-treated seedlings (Fig. 6C), although they did not show extra TAG possession in cotyledons (Fig. 2D), which implies that the glyoxysomal gene expression is not simply associated with TAG levels but rather appears to be linked to lipid metabolism in plastids. Furthermore, the high expression of ICL and MLS without strong down-regulation of GOX1 and HPR1 in green seedlings indicates that the expression of glyoxylate cycle genes can be regulated independently of photorespiratory genes during peroxisome differentiation in cotyledons.

Requirement of Galactolipids for the Initiation of Chloroplast Biogenesis
To elucidate the role of MGD1 expression and galactolipid biosynthesis in the initial stage of chloroplast development during seed germination, we suppressed MGD1 expression early in chloroplast biogenesis by adding 10 mM DEX after a lapse of a certain period after seeding. Application of DEX within 36 h after seeding strongly suppressed chlorophyll accumulation in 5-d-old seedlings, whereas DEX application 48 h after seeding decreased the inhibition in chlorophyll accumulation (Fig. 7A). In 5-d-old seedlings with DEX added 72 h after seeding, chlorophyll content was almost equal to that in DEX-untreated controls, although MGD1 expression was suppressed to nearly minimal levels of that of amiR-MGD1 constructs (Fig. 7B). Even when the seedlings were treated with DEX for 5 d from 72 h after seeding, chlorophyll content did not differ substantially from that in untreated controls (Supplemental Fig. S3A). Consistent with the chlorophyll levels, the expression of both nucleus-and plastid-encoded genes was not impaired in 5-and 8-d-old L4 seedlings treated with DEX 72 h after seeding ( Fig. 7C; Supplemental Fig. S3B), which suggests that galactolipid biosynthesis and presumably thylakoid membrane biogenesis at the initiation of chloroplast biogenesis are sufficient to ensure the induction of photosynthesis-associated genes. The expression of genes associated with peroxisomal and mitochondrial functions was also unchanged by the DEX treatment after initial seedling growth (Fig. 7C). Therefore, once MGDG is properly synthesized by MGD1 at the beginning of chloroplast biogenesis, the development of chloroplasts and other organelles in cotyledons proceeds without severe retardation even if MGD1 expression is inhibited afterward. MGD1 expression may be crucial particularly at the initiation of chloroplast biogenesis in cotyledons but less required as seedlings grow.

Effect of MGD1 Suppression on the Development of True Leaves
The continuous induction of amiR-MGD1 also affected the development of true leaves in L2 and L4 seedlings, Figure 7. Involvement of galactolipids in initial organelle development. A, Inhibition of chlorophyll accumulation in 5-d-old L2 and L4 seedlings by DEX treatment at different times from 0 to 72 h after seeding. Chlorophyll content in 5-d-old seedlings grown continuously in the absence of DEX is shown as an untreated control (2DEX). The horizontal line in each box represents the median value of the distribution. The top and bottom of each box represent the upper and lower quartiles, respectively. The whiskers represent the range (n = 24). B, MGD1 expression in 5-d-old L2 and L4 seedlings with DEX added 72 h after seeding (72 h) or grown continuously in the absence of DEX (2DEX). C, Expression of genes associated with photosynthesis, photorespiration, and the glyoxylate cycle in 5-d-old L4 seedlings with DEX added 72 h after seeding or grown continuously in the absence of DEX. In B and C, data are fold differences from the 2DEX control (means 6 SE, n = 3). Figure 8. Characteristics of amiR-MGD1 true leaves. A, Eight-day-old seedlings of the wild type (WT), L2, and L4 grown continuously on DEXcontaining medium. B, Chlorophyll content in true leaves of 14-d-old L4 seedlings grown on DEX-free (2DEX) or DEX-containing (+DEX) medium. FW, Fresh weight. C, Changes in F v /F m in true leaves (left) or cotyledons (right) of DEX-treated wild-type, L2, and L4 seedlings from day 7 to day 10 after seeding. Data are means 6 SE (n = 10). D, MGD1 expression in true leaves of 10-d-old L4 seedlings grown on 2DEX or +DEX medium. E, Proportion of galactolipids in total glycerolipids in true leaves of 14-d-old L4 seedlings grown on 2DEX or +DEX medium. F, Relative radioactivity in MGDG per total radioactivity in lipid fractions extracted from [ 14 C]acetatelabeled leaves of L4 seedlings grown on 2DEX or +DEX medium. True leaves from 10-d-old seedlings or cotyledons from 5-d-old seedlings were used. G, Expression of genes associated with photosynthesis, photorespiration, and the glyoxylate cycle in true leaves of 10-d-old L4 seedlings grown on 2DEX or +DEX medium. *Not detected. In D and G, data are fold differences from the DEX-untreated control. In B and D to G, data are means 6 SE from three or four independent experiments. All +DEX amiR-MGD1 seedlings used in these experiments had white cotyledons. although the effect was milder than in cotyledons. In the L2 and L4 seedlings with white cotyledons, first and second true leaves appeared smaller and paler than those in the wild type in the presence of DEX (Fig. 8A). Consistent with color, chlorophyll content in L4 true leaves was decreased with DEX treatment (Fig. 8B). When the PSII activity was monitored by measuring F v /F m in the presence of DEX, we consistently detected lower values in the true leaves of both L2 and L4 white seedlings than in the wild type (Fig. 8C). However, in these lines, the PSII impairment in the true leaves was considerably weaker than in the cotyledons, which shows a diminished impact of amiR-MGD1 in true leaves.
To evaluate the suppression effect of the amiR-MGD1 transgene on the MGD1 expression in true leaves, we grew L4 seedlings with white cotyledons on DEX-containing medium for 10 d and compared MGD1 expression in the true leaves with that in the DEX-untreated control. DEX treatment suppressed the MGD1 expression in true leaves (Fig. 8D) to the same level as in the white cotyledons of 5-d-old seedlings (Fig. 2C), so the weak effect of the DEX treatment in the true leaves was not due to the release of MGD1 suppression. To investigate whether the reduced MGD1 expression was sufficient for galactolipid biosynthesis in true leaves, we analyzed galactolipid content in true leaves in the white L4 class seedlings grown for 14 d in the presence of DEX (Fig. 8E). The proportion of MGDG in total membrane glycerolipids was decreased by 25% in DEX-treated true leaves as compared with the DEX-untreated control, whereas the proportion of DGDG was unchanged. Consistent with the lipid data, the MGDG synthesis rate estimated by [ 14 C]acetate incorporation was partially decreased in true leaves by DEX treatment, whereas that in cotyledons was very low compared with the DEX-untreated control (Fig. 8F). These data indicate that MGDG biosynthesis was also decreased in true leaves by amiR-MGD1, although the suppression effect was weaker than that in the white cotyledons. The reason why the MGDG synthesis activity largely differs from the MGD1 transcriptional level in the true leaves remains unclear. MGD2 and MGD3 expression appeared decreased even in true leaves of amiR-MGD1, and their contribution to leaf MGDG synthesis could be negligible (Supplemental Fig. S4; Kobayashi et al., 2009aKobayashi et al., , 2013. The MGD1 enzyme is reduced and activated in vitro by chloroplast thioredoxins, which could be coupled with photosynthetic activity in vivo (Yamaryo et al., 2006). Thus, the difference in photosynthetic activity (Fig. 8C) may explain the discrepancy in MGDG synthesis rate between true leaves and cotyledons in DEX-treated amiR-MGD1. In addition, the MGDG-synthesizing activity is regulated by anionic membrane lipids (Dubots et al., 2010) and also could be influenced by substrate supply. These factors may also affect the MGDG synthetic activity differently between true leaves and cotyledons. Processes of chloroplast differentiation differ considerably between cotyledons and true leaves (Pogson and Albrecht, 2011); in cotyledons, plastids partially developed during embryogenesis differentiate simultaneously into chloroplasts on light-induced germination (Mansfield and Briarty, 1996), whereas in true leaves, proplastids in the shoot apical meristem gradually differentiate into mature chloroplasts during leaf biogenesis (Charuvi et al., 2012). Thus, the requirement of MGD1 gene expression for sufficient galactolipid accumulation may vary between these two organs.
To address whether the reduced MGDG biosynthesis modifies gene expression in true leaves as observed in cotyledons (Fig. 6), we examined the expression of genes associated with organelle functions in true leaves of the 10-d-old white L4 seedlings (Fig. 8G). The expression of genes associated with photosynthesis and photorespiration was not decreased with DEX treatment, and psbA expression even appeared to be increased in true leaves. In addition, ICL expression was undetectable in true leaves regardless of DEX treatment, which implies no prominent expression of glyoxysomal genes in response to MGD1 suppression in true leaves. Our data indicate that MGD1 suppression differently affects gene expression between cotyledons and true leaves. Impaired MGDG biosynthesis in the mgd1-2 mutant strongly down-regulates both nucleus-and plastid-encoded photosynthetic genes in true leaves, which indicates a requirement of galactolipid biosynthesis in photosynthetic gene expression in true leaves (Kobayashi et al., 2013). Therefore, the 75% reduction in the MGD1 expression would be sufficient to down-regulate photosynthetic genes in cotyledons but not in true leaves. Distinct regulation of chloroplast differentiation between cotyledons and true leaves is evident from the identification of several mutants with cotyledon-specific defects in chloroplast biogenesis (Pogson and Albrecht, 2011). Most of the cotyledon-specific mutants show impaired transcription and translation of plastid-encoded genes (Shirano et al., 2000;Yamamoto et al., 2000;Ishizaki et al., 2005;Albrecht et al., 2006;Ruppel and Hangarter, 2007;Chi et al., 2010;Woodson et al., 2013) and homeostasis of plastid proteins (Shimada et al., 2007;Albrecht et al., 2008;Tanz et al., 2012), which suggests that cotyledons possess a particular set of components essential for photosynthetic gene expression during early seedling growth, in which galactolipid biosynthesis by MGD1 may be involved.

Plant Materials and Growth Conditions
All plants used in this study were in the Ler ecotype of Arabidopsis (Arabidopsis thaliana). Seeds were surface sterilized and then cold treated at 4°C for 4 d in the dark before seeding. Plants were grown on Murashige and Skoog (MS) medium (adjusted to pH 5.7 with KOH) containing 1% (w/v) Suc solidified with 0.8% (w/v) agar except for the experiments in Figures 1C and 7, A and B, in which the liquid form of the medium was used. All plants were grown at 23°C under continuous white light (30 mmol photons m 22 s 21 ) in a growth chamber. For DEX treatment, DEX (Wako; http://www.wako-chem. co.jp/english/) was added to a final concentration of 10 mM in the medium from a 50 mM stock in dimethyl sulfoxide. Unless stated otherwise, DEX was added to plants from the start of the cold treatment at the seed stage.
For the analyses in Figures 1C and 7, A and B, seeds cold treated in deionized water in the dark were seeded in liquid MS medium and grown for 3 d in the chamber with gentle rotation (90 rpm). Only germinated seedlings were grown on solidified MS medium for another 2 d to synchronize the germination time in the experiments. For the experiment in Figure 7, A and B, DEX was added to the medium after a lapse of each time from the start of the light growth.
For the phosphate-deficient condition in Figure 1E, the concentration of KH 2 PO 4 was reduced to 10 mM in the MS medium. Seedlings for the DEXuntreated control were grown on phosphate-deficient medium without DEX for 5 d from germination. For the DEX treatment, seedlings were first grown on phosphate-deficient medium without DEX for 3 d and then on phosphate-deficient medium containing 10 mM DEX for another 2 d.

Construction of Transgenic Lines
amiR-MGD1 was designed (miR-sense, 59-TAGATTATTAGGCAGTG-CAAC-39) according to WMD2 software and instructions at the Web site (http://wmd2.weigelworld.org/cgi-bin/mirnatools.pl.). The assembled artificial microRNA precursor fragment was cloned into pENTR entry vector (Invitrogen; http://www.lifetechnologies.com/), which as pOp6 was inserted at the NotI site. We cloned the 35S::GRLhG4 fragment into the AscI site of the same pENTR vector and selected the orientation of two promoters outgoing with each other (Fig. 1A). Then, the 35S::GRLhG4/pOp6::amiR-MGD1 fragment in pENTR was transferred to the pBGW destination vector by LR recombination. Wild-type plants (Ler ecotype) were transformed with the resulting plasmid via transfection with Agrobacterium tumefaciens strain GV3101.
Forty-eight transgenic plant lines were isolated with 0.1% (w/v) Basta spray at the T1 generation. At the T2 generation, Basta-resistant seedlings germinated on soil were sprayed with 1 mM DEX solution (pH 7) once daily until true leaves appeared. Seven lines showing color change of true leaves with DEX treatment were isolated as candidates of functional amiR-MGD1 transgenic lines. Seeds harvested individually from these seven lines showed approximately 75% Basta resistance, indicating a single-copy transgene insertion. We further harvested eight independent T3 seeds for each line (56 lines in all) and selected those showing 100% Basta resistance to be homozygous in the transgene. In this study, lines 2, 4, 5, and 7 were used for detailed analyses.

qRT-PCR Analysis
Total RNA was extracted by using the RNeasy Plant Mini kit (Qiagen; http://www.qiagen.com/). Genomic DNA digestion and reverse transcription involved the PrimeScript RT reagent kit with gDNA Eraser (TaKaRa Bio; http://www.takara-bio.com/). Complementary DNA amplification involved the Thunderbird PreMix kit (Toyobo; http://www. toyobo-global.com/) and 200 nM gene-specific primers (Supplemental Table S1). Thermal cycling consisted of an initial denaturation step at 95°C for 10 s followed by 40 cycles of 5 s at 95°C and 30 s at 60°C. Signal detection and quantification were performed in duplicate by use of 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). Three independent biological experiments were performed for each sample.

Lipid Analysis
Total lipids were extracted from seedlings crushed into powder in liquid nitrogen and were separated by thin-layer chromatography with a solvent system of acetone:toluene:water (136:45:12, v/v/v) as described (Kobayashi et al., 2006). Lipids were visualized with 0.01% (w/v) primuline in 80% (v/v) acetone under UV light. MGDG, DGDG, TAG, and a mixture of other glycerolipids consisting of phosphatidylcholine, phosphatidylethanolamine, phosphatidylinositol, phosphatidylglycerol, and sulfoquinovosyldiacylglycerol were isolated from silica gel plates. Fatty acids in each lipid fraction were methyl esterified by incubation in 1 M HCl in methanol at 85°C for 2 h and quantified by gas chromatography (GC-17A; Shimadzu; http://www.shimadzu.com/) with myristic acid as an internal standard.

Chlorophyll Determination
For chlorophyll determination in Figures 2B and 8B, seedlings crushed in liquid nitrogen were homogenized in 80% (v/v) acetone, and debris was removed by centrifugation at 15,000 rpm for 5 min. The chlorophyll content of samples was determined spectrophotometrically by measuring the absorbance of the supernatant at 720, 663, and 645 nm with an Ultrospec 2100 pro spectrophotometer (GE Healthcare) according to a previous report (Melis et al., 1987). For the analyses in Figures 1C and 7A, chlorophyll was extracted from single intact seedlings by incubating each seedling in 1 mL of 80% (v/v) acetone at 4°C for 3 d. Chlorophyll content in single seedlings was determined by measuring fluorescence emission at 666 nm under 440-nm excitation with an FP-6200 spectrofluorometer (JASCO; http://www.jascoinc.com/) with a chlorophyll sample of known concentration used as a standard.  Oxborough and Baker (1997):

Photosynthetic Chlorophyll Fluorescence Analysis
. From these fluorescence yields, photosynthetic parameters were calculated as follows (van Kooten and Snel, 1990;Maxwell and Johnson, 2000): F v /F m = (F m 2 F o )/F m , F v 9/F m 9 = (F m 9 2 F o 9)/F m 9, F II = (F m 9 2 F)/F m 9, qP = (F m 9 2 F)/(F m 9 2 F o 9). The actual F II can be transformed into a product of qP and F v 9/F m 9: F II = (F m 9 2 F)/F m 9 = qP 3 F v 9/F m 9. F NPQ and F NO were determined according to the method of Kramer et al. (2004). Measurement parameters for IMAGING-PAM were as follows: measuring light intensity = 2, measuring light frequency = 4, damping = 4, gain = 1, saturation pulse intensity = 10. For quantification, averaged fluorescence values in a circular area were collected from a cotyledon or true leaf for each seedling using the software ImagingWin (Walz). For the analysis with JUNIOR-PAM, the automated induction program provided by the WinControl-3 software (Walz) was used with default settings for measuring light and saturation pulse.
For chlorophyll fluorescence induction experiments (Fig. 4, I and J), 5-d-old seedlings were dark incubated for 5 min before experiments. Eight seedlings were used for an experiment in a batch. When required, the seedlings were infiltrated with 40 mM DCMU and 150 mM sorbitol by depression before dark incubation. Chlorophyll fluorescence transients were measured in a logarithmic time series between 30 ms and 10 s after the onset of strong actinic light (1,650 mmol photons m 22 s 21 ) with a light-emitting diode pump-probe spectrometer (JTS-10; BioLogic; http://www.bio-logic.info/).
Fluorescence emission spectra of chlorophyll proteins at 77 K were obtained directly from plant tissues (Fig. 5A) or from membrane fractions (Fig. 5B) in liquid nitrogen by using an RF-5300PC spectrofluorometer (Shimadzu) under 435-nm excitation. To prepare membrane fractions, seedlings were pulverized in liquid nitrogen and homogenized in a cold buffer (0.33 M sorbitol, 5 mM MgCl 2 , 5 mM EDTA, and 50 mM HEPES-KOH, pH 7.6). The homogenate was filtered through a single layer of Miracloth (Merck Millipore; http://www. merckmillipore.com/) with gentle hand pressure. After centrifugation at 5,000g for 10 min at 4°C, the supernatant was discarded and the pellet was resuspended in a cold buffer to obtain 0.8 mg mL 21 chlorophyll-containing membrane fractions.

Microscopy Analysis
Arabidopsis seedlings were fixed in 2.5% (v/v) glutaraldehyde and 4% (v/v) paraformaldehyde in 0.1 M sodium phosphate buffer, pH 7, at room temperature for 4 h. After rinses with phosphate buffer (three times for 20 min each), samples were postfixed in 1% (w/v) OsO 4 in the same buffer for 4 h at room temperature and rinsed with phosphate buffer (three times for 20 min each). Samples were dehydrated in an ethanol series and propylene oxide, embedded in Spurr's resin, and sectioned with the use of an ultramicrotome (Reichert Ultracut S or EM UC6; Leica; http://www.leica-microsystems.com/). For light microscopy, sections (0.8 mm) were stained with 1% (w/v) toluidine blue with sodium borate and analyzed with a light microscope (BX60; Olympus; http://www.olympusglobal.com/en/). For electron microscopy observation, ultrathin sections (70-90 nm) were stained with 5% (w/v) uranyl acetate in 50% (v/v) methanol and 0.4% (w/v) lead citrate in 0.1 N NaOH. Sections were observed by transmission electron microscopy (CM 100; Philips; http://www.fei.com/) at 80 kV, and images were obtained by use of a Gatan Orius CCD camera (http://www.gatan. com/).

Immunoblot Analysis
The membrane protein fraction was prepared as described previously (Kobayashi et al., 2013) from 5-d-old wild-type and L2 pale-green seedlings treated with 10 mM DEX. Twenty micrograms of total membrane protein from an L2 sample was electrophoresed together with a dilution series (0.2-5 mg) of total membrane protein from the wild type and electrotransferred onto nitrocellulose membranes (Amersham Protran Premium 0.2 NC; GE Healthcare) as described (Kobayashi et al., 2007). Protein bands that reacted with primary antibodies were secondarily labeled with goat anti-rabbit IgG secondary antibody conjugated with horseradish peroxidase (Thermo Scientific; http://www.thermoscientific.com/) and detected using a chemiluminescence reagent (Pierce Western Blotting Substrate Plus; Thermo Scientific) and an imager (ImageQuant LAS 4000 mini; GE Healthcare). Antibodies against PsaA/PsaB were kindly provided by Ryouichi Tanaka (Hokkaido University) and those against D1 and D2 were kindly provided by Masahiko Ikeuchi (University of Tokyo). Antibodies against LHCA1, LHCA3, LHCB1, and LHCB3 were from AgriSera (http://www.agrisera.com/).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Fatty acid composition of TAG.
Supplemental Figure S3. Effect of 5-d DEX treatment from 72 h after seeding in the amiR-MGD1 L4 line.
Supplemental Figure S4. Expression of MGD2 and MGD3 in true leaves of 10-d-old amiR-MGD1 L4 seedlings grown on DEX-free or DEX-containing medium.