Mono- and digalactosyldiacylglycerol (MGDG and DGDG, respectively) constitute the bulk of membrane lipids in plant chloroplasts. Mutant analyses in Arabidopsis have shown that these galactolipids are essential for chloroplast biogenesis and photoautotrophic growth. Moreover, these non-phosphorous lipids are proposed to participate in low-phosphate (Pi) adaptations. Under Pi-limited conditions, a drastic accumulation of DGDG occurs concomitantly with a large reduction in membrane phospholipids, suggesting that plants substitute DGDG for phospholipids during Pi starvation. Previously, we reported that among the three MGDG synthase genes (MGD1, MGD2 and MGD3), the type-B MGD2 and MGD3 are upregulated in parallel with DGDG synthase genes during Pi starvation. Here, we describe the identification and characterization of T-DNA insertional mutants of Arabidopsis type-B MGD genes. Under Pi-starved conditions, the mgd3-1 mutant showed a drastic reduction in DGDG accumulation, particularly in the root, indicating that MGD3 is the main isoform responsible for DGDG biosynthesis in Pi-starved roots. Moreover, in the roots of mgd2 mgd3 plants, Pi stress-induced accumulation of DGDG was almost fully abolished, showing that type-B MGD enzymes are essential for membrane lipid remodeling in Pi-starved roots. Reductions in fresh weight, root growth and photosynthetic performance were also observed in these mutants under Pi-starved conditions. These results demonstrate that Pi stress-induced membrane lipid remodeling is important in plant growth during Pi starvation. The widespread distribution of type-B MGD genes in land plants suggests that membrane lipid remodeling mediated by type-B MGD enzymes is a potent adaptation to Pi deficiency for land plants.
Life in all organisms relies on the presence of lipid bilayer cell membranes, which separate the interior of cells from their environment. In the plastids of higher plants, the non-phosphorous galactolipids monogalactosyldiacylglycerol (MGDG) and digalactosyldiacylglycerol (DGDG) are the main constituents of the membrane lipids, whereas phospholipids are the predominant components of other subcellular membranes, such as the plasma membrane, the endoplasmic reticulum (ER) and the mitochondrial membranes (Joyard et al., 1998). In chloroplasts, MGDG and DGDG account for approximately 50 and 25 mol%, respectively, of thylakoid membrane lipids, and play essential roles in photosynthetic membrane biogenesis (Joyard et al., 1998). These lipids may also be directly involved in the photosynthetic reaction in higher plants and cyanobacteria (Hölzl and Dörmann, 2007).
In higher plants, the last step in MGDG synthesis occurs in plastid envelope membranes. This reaction is catalyzed by a galactosyltransferase or MGDG synthase (MGD), which transfers a galactosyl residue from uridine diphosphate (UDP)-galactose to the sn-3 position of sn-1,2-diacylglycerol (DAG) (Benning and Ohta, 2005). Not only is MGDG the main constituent of thylakoid membranes, it is also a substrate for DGDG synthesis; thus, MGDG synthesis is a key step for the biosynthesis of both galactolipids. In Arabidopsis, the DAG used in galactolipid synthesis comes from two different pathways: the prokaryotic pathway inside the plastid and the eukaryotic pathway from the ER (Joyard et al., 1998). Plastid-derived DAG contains 18-carbon fatty acids in the sn-1 position, and 16-carbon fatty acids in the sn-2 position, whereas ER-derived DAG contains 16- or 18-carbon fatty acids in the sn-1 position, and 18-carbon fatty acids in the sn-2 position. Under phosphate (Pi)-starved conditions, ER-derived DAG containing C16:0 in the sn-1 position is actively used for galactolipid biosynthesis (Härtel et al., 2000; Kelly et al., 2003).
In Arabidopsis, three functional MGDG synthases (MGD1, MGD2 and MGD3) have been identified and classified into type-A (MGD1) and type-B (MGD2 and MGD3) enzymes based on their amino acid identity (Awai et al., 2001). There are several differences between type-A and type-B enzymes, with respect to their substrate specificity, subcellular localization and gene expression profiles. Type-A enzyme MGD1 is targeted to the inner envelope membranes of plastids, whereas type-B MGD2 and MGD3 are localized to the plastid outer envelope membranes (Awai et al., 2001). Of the three, MGD1 is the most abundant isoform in chloroplasts, and expression of MGD1 is widespread in photosynthetic tissues. In contrast, expression of MGD2 and MGD3 is specifically detected in non-photosynthetic tissues such as flowers and roots, and is rarely observed in green tissues (Awai et al., 2001; Kobayashi et al., 2004). An MGD1 knock-out mutant in Arabidopsis showed an absence of both galactolipids, and a disruption of the photosynthetic membranes, leading to the complete impairment of photosynthetic ability and photoautotrophic growth (Kobayashi et al., 2007). These data demonstrate that, in photosynthetic tissues, MGD1 function underlies the bulk of galactolipid biosynthesis, and that MGD2 or MGD3 cannot compensate for a loss of MGD1 function. Previous studies have, however, shown that Pi starvation strongly activates the expression of MGD2 and MGD3 in the shoot and the root (Awai et al., 2001; Kobayashi et al., 2004). This activation occurs in synchrony with the induction of the DGDG synthase genes, DGD1 and DGD2, which leads to a substantial increase in DGDG content during Pi starvation (Kelly et al., 2003). Although DGDG is thought to be localized exclusively to plastids under nutrient-sufficient conditions, it accumulates in extraplastidic membranes under Pi-starved conditions (Härtel et al., 2000), such as in the plasma membrane (Andersson et al., 2003), the tonoplasts (Andersson et al., 2005) and the mitochondrial membranes (Jouhet et al., 2004). In parallel with DGDG accumulation, phospholipid degradation occurs during Pi starvation (Essigmann et al., 1998). These observations suggest that DGDG replaces phospholipids as the major membrane component under Pi-limited growth conditions. Despite the increased demand for MGDG for DGDG synthesis during Pi starvation, no induction of MGD1 expression is observed in response to Pi starvation (Awai et al., 2001), indicating the possibility that type-B but not type-A MGD enzymes are responsible for Pi starvation-induced galactolipid accumulation.
To unravel the role of type-B MGD enzymes in plant growth, particularly in adaptation to Pi-limited conditions, we isolated loss-of-function mutants for MGD2 and MGD3. Characterization of these mutants showed that type-B MGD genes, mainly MGD3, are largely responsible for DGDG accumulation during Pi starvation, although they are not essential for galactolipid biosynthesis and plant growth under nutrient-sufficient conditions. Loss of type-B MGD functions resulted in growth impairment during Pi starvation, providing direct genetic evidence that type-B MGD enzymes are crucial for adaptation to Pi deficiency, and that replacement of phospholipids by DGDG is a potent adaptation mechanism of plants to Pi-limited growth conditions.
Isolation of T-DNA-tagged mutants of MGD2 and MGD3 in Arabidopsis
Both MGD2 and MGD3 encode functional MGDs, and have expression patterns that are different from that of MGD1, particularly in non-photosynthetic tissues, suggesting a role for these genes that is distinct from MGD1 (Awai et al., 2001; Kobayashi et al., 2004, 2006). To understand the in vivo function of type-B MGD genes, we isolated loss-of-function mutants of these genes in Arabidopsis (Figure 1a). A plant carrying a T-DNA insertion in the fifth intron of MGD2 (mgd2-1) was identified by PCR screening of the T-DNA insertional mutant population of the Kazusa DNA Research Institute. Sequencing of the T-DNA/genomic DNA junction in mgd2-1 revealed that both ends of the insertion were the left border of the T-DNA, indicating that the insertion was an inverted repeat of T-DNAs. We also obtained a T-DNA-inserted mutant of MGD3 from the SALK Institute Genomic Analysis Laboratory (mgd3-1; SALK_084186), which carries the insertion in the first exon of MGD3. A second plant with a T-DNA insertion in MGD3 (mgd3-2) was isolated from the T-DNA-tagged lines of the Kazusa DNA Research Institute. This T-DNA insertion was in the second intron of MGD3. PCR analysis and DNA gel blot analysis revealed that the nucleotide sequence between the fifth and eighth exons of MGD3 was missing in the mgd3-2 genome, indicating that the T-DNA insertion caused a large deletion of the 3′ nucleotide sequence of the MGD3 coding region (data not shown). We successfully generated a homozygous mgd2 mgd3 double mutant by crossing mgd2-1 and mgd3-1. To examine transcript levels of the three MGD genes in these mutants, semi-quantitative reverse transcription (RT)-PCR was carried out (Figure 1b). In mgd2-1, expression of MGD2 was reduced to undetectable levels, whereas transcript levels of MGD1 and MGD3 were very similar to those in wild type. Likewise, MGD3 expression was not detectable in the mgd3-1 mutant, although expression levels of both MGD1 and MGD2 were not affected. MGD3 expression was also disrupted in the mgd3-2 mutant (data not shown). In the mgd2 mgd3 double mutant, expression of both MGD2 and MGD3 was disrupted, but the MGD1 transcript level was normal. These results indicate that the T-DNA insertions in MGD2 and MGD3 resulted in the knock-out of gene expression, but did not affect the transcript levels of other MGD isoforms.
Type-B MGD genes are dispensable for plant growth under nutrient-sufficient conditions
Although the expression of both MGD2 and MGD3 was disrupted in the mgd2 mgd3 double mutant (Figure 1b), the mutant had a healthy phenotype and was indistinguishable from wild type when grown in nutrient-sufficient soil (data not shown). Promoter analysis of MGD genes in Arabidopsis has shown that expression of MGD2 and MGD3 is upregulated during pollen maturation and pollen tube growth (Kobayashi et al., 2004). However, the double mgd2 mgd3 mutations showed no influence on in vitro pollen tube growth, or on reproductive processes (data not shown), demonstrating that type-B MGD genes are not necessary for either vegetative growth or reproduction under optimal growth conditions. Expression of MGD3 is also highly activated during early seedling growth (Awai et al., 2001; Kobayashi et al., 2004). To examine the role of type-B MGD genes in early seedling growth, we compared the lipid composition of 3-day-old mgd2 mgd3 plants with that of wild-type plants (Figure 2a). There was no difference in lipid composition between the double mutant and the wild type. Furthermore, we could not detect any differences in the fatty acid composition of the galactolipids between wild-type and mutant seedlings (data not shown), indicating that type-B MGD genes are dispensable for galactolipid biosynthesis during early seedling growth.
In addition to thylakoid membranes of chloroplasts, galactolipids are also the major membrane lipid component of prolamellar bodies inside etioplasts, which develop in the cotyledons of dark-grown seedlings (Selstam and Sandelius, 1984). Although MGD1 is the most highly expressed of the three MGD isoforms during etiolated growth (Kobayashi et al., 2004), a knock-down mutant of MGD1 showed no obvious defect in etioplast biogenesis, suggesting the possibility that MGD1 is not important for plastid development in the dark (Jarvis et al., 2000). To assess whether type-B MGD genes are responsible for galactolipid biosynthesis in etiolated cotyledons, lipid analysis was carried out on dark-grown seedlings of wild-type and of mgd2 mgd3 plants (Figure 2b). The lipid composition of the double mutant was indistinguishable from that of the wild type. Furthermore, the fatty acid composition of MGDG and DGDG was identical for the double mutant and wild type (data not shown). These results demonstrate that type-B MGDs also have no central role in galactolipid biosynthesis during etioplast development.
Type-B MGD genes are required for galactolipid accumulation during Pi starvation
The expression of type-B MGD genes, and the DGDG synthase genes DGD1 and DGD2, is co-activated by Pi starvation (Awai et al., 2001; Kelly et al., 2003), suggesting that MGD2/3 and DGD1/2 cooperate in DGDG accumulation during Pi starvation. To understand the role of type-B MGD genes in Pi starvation-induced galactolipid biosynthesis, we analyzed the membrane lipid composition of type-B MGD mutants in both shoots and roots under Pi-controlled conditions. In the shoots of Pi-sufficient plants, there was no difference in lipid composition between the wild type and any of the type-B MGD mutants (Figure 3a), consistent with the results shown in Figure 2. In the roots of Pi-sufficient mgd2 mgd3, however, DGDG was reduced by 41%, as compared with wild type (Figure 3b), showing the importance of type-B MGD genes for DGDG synthesis in the roots. Moreover, under Pi-starved conditions, a reduction in DGDG content was observed not only in mgd2 mgd3, but also in mgd3-1; the levels of DGDG in the shoot was significantly reduced by 19 and 15%, respectively, compared with wild type, whereas no DGDG reduction was observed in mgd2-1 (Figure 3c). The level of MGDG also decreased by 11% in the mgd2 mgd3 double mutants, compared with the wild type in the Pi-starved shoots. In parallel with the decrease in galactolipids, the proportion of phosphatidylethanolamine (PE) and phosphatidylcholine (PC) increased by approximately twofold in the shoots of Pi-starved mgd2 mgd3, whereas the levels of phosphatidylglycerol (PG) and phosphatidylinositol (PI) were not significantly changed, compared with wild type. This suggests that type-B MGD enzymes are mainly involved in the replacement of PC and PE with DGDG in the membrane during Pi starvation. The mutation effect was much stronger in the root than in the shoot under Pi-starved conditions: the DGDG content was reduced by 62 and 79% in mgd3-1 and mgd2 mgd3, respectively (Figure 3d). In particular, the double mutant showed no Pi stress-induced accumulation of DGDG in the root. As observed in the Pi-starved shoots, the proportion of PE and PC, but not PG and PI, increased concomitantly with the DGDG reduction in the roots of mgd3-1 and mgd2 mgd3. The same results were also observed in the mgd3-2 mutant (Figure S1). On the other hand, the lipid composition of mgd2-1 was the same as that of wild type, even in the Pi-starved roots, showing that MGD2 has only a limited role in galactolipid biosynthesis. We also analyzed the accumulation of DAG and triacylglycerol in the roots of mgd2 mgd3 under Pi-controlled conditions to ascertain whether the disrupted galactolipid biosynthesis in type-B MGD mutants affects neutral lipid homeostasis by perturbing the DAG-metabolizing pathway. No difference in the accumulation of neutral glycerolipids was, however, observed between the wild type and the double mutant (data not shown), perhaps indicating that impairment of galactolipid biosynthesis in plastid outer envelope membranes affects phospholipid degradation negatively, resulting in the accumulation of PC and PE in type-B MGD mutants during Pi starvation.
We also analyzed the fatty acid composition of galactolipids in the wild type and in the mgd2 mgd3 mutant under Pi-controlled conditions. No difference in the fatty acid composition of MGDG and DGDG was detected between the wild type and the double mutant under Pi-sufficient conditions in either the shoots or roots (Figure 4a,b), reflecting the fact that type-B MGD enzymes do not contribute significantly to galactolipid synthesis under Pi-sufficient conditions. Under Pi-starved conditions, however, large differences in the fatty acid composition of galactolipids were observed between the wild type and the double mutant, particularly in the root. In the roots of wild-type plants, Pi starvation induced an increase in C16:0 and C18:2 fatty acids, concomitantly with a reduction in C18:3 fatty acids in both MGDG and DGDG (Figure 4c,d). The same changes in fatty acid composition of DGDG were reported for the leaves of Pi-starved Arabidopsis plants (Kelly et al., 2003), and were also observed for Pi-starved shoots in our study, although the changes were smaller than those in the roots (Figure 4b). In contrast, the fatty acid composition of galactolipids of Pi-starved mgd2 mgd3 was almost the same as that of the Pi-sufficient double mutant in both the shoots and the roots, demonstrating that the galactolipid metabolic pathway activated by Pi starvation was strongly perturbed in the double mutant. Whereas Pi starvation-induced changes in the fatty acid composition of galactolipids were also abolished in mgd3-1 and mgd3-2, those in mgd2-1 were comparable with those in the wild type (Figure S1 and data not shown), showing again that MGD3, but not MGD2, is the main enzyme for galactolipid metabolism during Pi starvation.
Importance of type-B MGD-mediated membrane lipid remodeling in adaptation to Pi starvation in Arabidopsis
Our lipid analyses demonstrated that, particularly in the root, type-B MGD genes are indispensable for DGDG accumulation during Pi starvation, which implies that these genes play an important role in adaptation to Pi starvation in Arabidopsis. To address the physiological contribution of these genes during Pi starvation, primary root growth was investigated in the seedlings of wild-type plants and type-B MGD mutants under Pi-controlled conditions (Figure 5a,b). Under Pi-sufficient conditions, no significant differences in root growth were detected among any plants (Figure 5a). Under Pi-limited conditions, however, growth inhibition was observed in the primary roots of mgd3-1 and mgd2 mgd3, whereas mgd2-1 showed only a small reduction, compared with the wild type. In 16-day-old plants, the primary root length of both mgd3-1 and the double mutant was 22% less than that of the wild type, indicating the importance of MGD3 for root growth during Pi starvation. Under Pi-sufficient conditions, the fresh weight of shoots and roots for all type-B MGD mutants was the same as that of wild-type plants (Figure 5c,d). In addition, mgd2-1 showed no significant difference from the wild type in the fresh weight of either tissue, even under Pi-starved conditions. In contrast, the fresh weight of shoots and roots of Pi-starved mgd3-1 was 25 and 33% less, respectively, than that of the wild type. The double mutant showed the same reduction as mgd3-1 in the fresh weight of shoots and roots. These growth phenotypes in type-B MGD mutants closely correlated with the DGDG content phenotype, indicating that DGDG accumulation during Pi starvation is important for plant development under Pi-starved conditions. Moreover, a pulse amplitude modulation (PAM) analysis revealed that the disruption of type-B MGD genes also affected photosynthetic activity during Pi starvation (Figure 5e). Under Pi-sufficient conditions, the effective quantum yield of photosystem II (ФPSII) was comparable between the wild type and all type-B MGD mutants, both in the cotyledons and in the rosette leaves. In Pi-starved wild-type plants, the ФPSII of rosette leaves and cotyledons was reduced by 15 and 60%, respectively, compared with that of Pi-sufficient leaves, suggesting a degradation of the photosynthetic apparatus, particularly in cotyledons, during Pi starvation. Leaves of the mgd2-1 mutant showed no significant difference compared with wild type, even under Pi-starved conditions. In the Pi-starved mgd3-1 mutant, however, the ФPSII in cotyledons was greatly reduced to a trace level, whereas the yield in rosette leaves was comparable with that of Pi-starved wild-type plants. ФPSII was also very low in the cotyledons and rosette leaves of Pi-starved mgd2 mgd3 plants. Given the result that the MGDG content was decreased in the shoots of mgd2 mgd3 plants under Pi-starved conditions (Figure 3c), MGDG reduction may affect photosynthetic activity in the rosette leaves of the double mutant.
We next investigated whether the DGDG reduction in the mgd2 mgd3 mutant affected free Pi pools during Pi starvation (Figure 5f). The level of free Pi was equally decreased in the wild type and the double mutant during Pi deficiency, suggesting that the membrane lipid remodeling during Pi starvation did not affect the size of free-Pi pools within the seedlings. In cultured Arabidopsis cells, DGDG accounts for 18 mol% of mitochondrial membrane lipids under Pi-deficient conditions, implying that DGDG contributes to the membrane organization in mitochondria as a substitute for phospholipids during Pi starvation (Jouhet et al., 2004). To assess whether the disruption of membrane lipid remodeling that is induced during Pi starvation affects respiration activity in the roots of mgd2 mgd3 plants, we measured oxygen consumption in the roots of wild-type and mgd2 mgd3 plants under Pi-limited conditions (data not shown). In the wild type, there was no difference in the oxygen consumption rate between Pi-sufficient and Pi-starved roots, which is consistent with the observation that the respiration activity of isolated mitochondria is not affected by Pi starvation (Jouhet et al., 2004). The roots of the mgd2 mgd3 plants also showed the same level of oxygen consumption as did those of the wild type under both conditions, suggesting that the DGDG reduction in mgd2 mgd3 did not affect respiration activity of mitochondria in the root.
Studies on MGD1 mutants in Arabidopsis showed that type-A MGD1 is the major isoform of MGD responsible for the bulk of galactolipid biosynthesis in photosynthetic tissues (Jarvis et al., 2000; Kobayashi et al., 2007). By contrast, it was anticipated that type-B MGD2 and MGD3 are involved in an alternative galactolipid pathway activated by specific growth conditions, such as Pi deficiency (Awai et al., 2001; Benning and Ohta, 2005; Kobayashi et al., 2004). Moreover, a recent study by Xu et al. (2008) suggested that the upregulation of type-B MGD genes is important for the activation of the alternative pathway in response to reactive oxygen species. Although these data imply unique roles of type-B MGD enzymes differing from type-A MGD1, there was no genetic evidence to demonstrate the physiological roles of type-B MGD genes. Here, we characterized loss-of-function mutants for type-B MGD isoforms in Arabidopsis, and provided direct evidence that type-B MGDs are indispensable for the alternative galactolipid pathway that plays an important role in efficient growth during Pi starvation.
Our data show that Pi stress-induced DGDG accumulation was severely disrupted in the roots of the type-B MGD mutants mgd3-1 and mgd2 mgd3 (Figure 3). These results demonstrate that type-B MGD isoforms, particularly MGD3, play a crucial role in membrane lipid remodeling during Pi starvation. Concomitant with the large reduction in DGDG content, developmental defects were also observed in mgd3-1 and mgd2 mgd3 during Pi starvation (Figure 5). These type-B MGD mutants, which are deficient only in Pi stress-induced DGDG biosynthesis, provide direct evidence that membrane lipid remodeling is a potent adaptation system for Pi deficiency in Arabidopsis. In contrast, no significant defects were observed in the mgd2-1 single mutant, suggesting that MGD2 has only a limited role in galactolipid biosynthesis. However, because the mgd2 mgd3 double mutant showed more severe defects in galactolipid biosynthesis than the mgd3-1 single mutant during Pi starvation, MGD2 must contribute to galactolipid accumulation by some degree. Because the T-DNA insertion in mgd2-1 is in the fifth intron of MGD2, we cannot rule out the possibility that truncated N-terminal fragments of MGD2 are expressed and translated in mgd2-1. Bottéet al. (2005), however, have shown that the putative galactose recognition sites located in the C-terminal domain of plant MGDG synthases are necessary for their galactosyltransferase activity. We showed by RT-PCR analysis that expression of the C-terminal region of MGD2, including the galactose recognition sites, was abolished in mgd2-1 (Figure 1b), suggesting that mgd2-1 is unable to express functional MGD2 because of the T-DNA insertion.
In contrast to the Pi-starved conditions, nutrient-sufficient conditions produced no visible defective phenotypes in type-B MGD mutants, including the double mutant (data not shown). Even in lipid analyses, these mutants showed no significant differences from the wild type under Pi-sufficient conditions, with the exception of 20-day-old roots of mgd2 mgd3, in which the DGDG content was slightly decreased (Figures 2–4). Moreover, quantitative analyses of primary root length (Figure 5a), fresh weight (Figure 5c) and photosynthetic activity (Figure 5e) revealed that mutations in type-B MGD genes caused no developmental alteration under Pi-sufficient conditions. These results clearly show that type-B MGD genes have no essential role under Pi-sufficient conditions, even during early seedling growth and pollen development, when the specific expression of type-B MGD genes has been detected (Kobayashi et al., 2004). Because expression of MGD1 was also detected during early seedling growth and pollen development (Kobayashi et al., 2004), it is likely that MGD1 activity is sufficient for the development of these tissues under normal nutrient conditions.
In the roots of mgd2 mgd3 plants, Pi stress-induced DGDG accumulation was almost fully abolished, demonstrating that type-B MGD enzymes are responsible for the majority of DGDG biosynthesis in Pi-starved roots, and that type-A MGD1 cannot compensate for this role (Figure 3d). In contrast to the crucial defect in lipid metabolism in the root, shoots of the double mutant still showed an increase in DGDG content upon Pi starvation, although the accumulation rate was reduced compared with the wild type (Figure 3c). In Arabidopsis, two biosynthetic pathways for DGDG have been proposed, namely the MGD1–DGD1 pathway and the MGD2/3–DGD2 pathway (Benning and Ohta, 2005). The MGD1–DGD1 pathway is suggested to be responsible for the majority of galactolipid biosynthesis, whereas the MGD2/3–DGD2 pathway is important under conditions of Pi stress. In addition to DGD2, however, DGD1 is also induced by Pi starvation, and contributes to Pi starvation-induced DGDG accumulation, at least in the shoot (Kelly et al., 2003). Thus, in mgd2 mgd3, the MGD1–DGD1 pathway may contribute to DGDG accumulation in the shoot during Pi starvation, even though MGD1 expression is not induced by Pi starvation. On the other hand, Pi stress-induced DGDG accumulation in the root is predominantly achieved through DGD2 (Härtel et al., 2000). Combining this information with our data (Figure 3d), we conclude that the alternative MGD2/3–DGD2 pathway is exclusively involved in Pi stress-induced DGDG accumulation.
Analysis of the fatty acid composition also indicated a strong relationship between MGD2/3 and DGD2 (Figure 4). Mutant studies of Arabidopsis DGDG synthases have shown that DGD2 is responsible for the synthesis of DGDG molecular species that are rich in C16:0 and C18:2 fatty acids and poor in C18:3 fatty acids, which are Pi starvation-induced molecular species, whereas DGD1 is involved in the synthesis of DGDG molecular species that are rich in C18:3 fatty acids, and poor in C16:0 and C18:2 fatty acids (Kelly et al., 2003). Our results in Figure 4 showed that type-B MGD isoforms are responsible for the synthesis of MGDG molecular species that are rich in C16:0 and C18:2 fatty acids, and poor in C18:3, fatty acids during Pi starvation. These data strongly suggest that Pi stress-induced changes in DGDG molecular species require a coupled reaction of MGD2/3 and DGD2, all of which are localized to the outer envelope membrane of plastids in Arabidopsis (Benning and Ohta, 2005).
Along with the large reduction in DGDG content, primary root growth upon Pi starvation was also diminished in the mgd3-1 mutant and in the mgd2 mgd3 double mutant. Under these conditions, plants increase the total surface area of roots by developing lateral roots and root hairs to maximize Pi absorption from the soil, resulting in an increase in the root-to-shoot growth ratio (Ticconi and Abel, 2004). Membrane lipid remodeling upon Pi starvation, which involves an increase in DGDG and a decrease in phospholipids, would contribute to this massive root development by decreasing the demand for Pi in the membranes, and reprioritizing internal Pi use. Lipid analyses in mgd3-1 and mgd2 mgd3 showed that the defect in membrane lipid remodeling during Pi starvation was much more severe in the root than in the shoot (Figure 3c,d), but that the fresh weights of the shoot and root were equally reduced (Figure 5d). It is known that under Pi-starved conditions, Pi in older leaves is translocated to young leaves and growing roots, indicating that Pi demand is carefully controlled throughout the entire plant (Schachtman et al., 1998). Pi redistribution from the shoot to the root may explain why developmental impairment in fresh weight occurred equally in the shoots and in the roots of these mutants, despite the fact that the defect in lipid metabolism was more severe in the roots. Photosynthetic performance in the cotyledons was also reduced by Pi starvation (Figure 5e). Because Pi starvation induces a redistribution of Pi from old leaves to young leaves and growing roots (Schachtman et al., 1998), it is likely that the senescence of cotyledons is accelerated under Pi-starved conditions, so as to recycle Pi for other important tissues such as developing roots. In mgd3-1 and mgd2 mgd3, the photosynthetic performance in cotyledons was affected by Pi starvation more severely than in the wild type (Figure 5e). In these mutants, Pi starvation-induced degradation may be accelerated in the cotyledons because of the defect in the membrane lipid remodeling. Although the membrane lipid remodeling during Pi starvation was severely impaired in mgd2 mgd3, the size of free-Pi pools within the mutant remained the same as those within the wild type under Pi-starved conditions (Figure 5f). Under severe Pi-deficient conditions, plants should have no surplus Pi pools within the cells; Pi molecules recycled from membrane phospholipids may be used immediately for other principal phosphorus-containing compounds, such as nucleic acids and phosphoproteins.
Because Pi easily forms insoluble compounds with aluminum and iron in soils, Pi concentration in the soil rarely exceeds 10 μm, and thus plants tend to suffer from phosphorus deficiency in nature (Schachtman et al., 1998). Gene expression analyses in Arabidopsis have demonstrated that all Pi starvation-inducible genes involved in glycolipid biosynthesis (MGD2, MGD3, DGD1, DGD2, SQD1 and SQD2) are fully activated when concentrations of Pi are below 10 μm (Awai et al., 2001; Essigmann et al., 1998; Kelly et al., 2003; Yu et al., 2002). These data suggest that membrane lipid remodeling is a common occurrence in wild plants in natural habitats. In fact, Pi starvation-induced membrane lipid remodeling has also been observed in Avena sativa (Andersson et al., 2003), Glycine max (Gaude et al., 2004), Phaseolus vulgaris (Russo et al., 2007) and Acer pseudoplatanus suspension-cultured cells (Jouhet et al., 2004). Recently, we found that replacement of phospholipids by DGDG is also used by Sesamum indicum to adapt to Pi deficiency (T. Watanabe, M. Shimojima, R. Koizumi, S. Masuda and H. Ohta, unpublished data). These studies are indicative of a widespread use of the lipid remodeling system upon Pi starvation in plants. In silico analysis using the GenBank Expressed Sequence Tags database revealed that type-B MGDG synthases are also found in monocotyledons (Awai et al., 2001). A BLAST search of the latest data from the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov), and a subsequent phylogenetic analysis using a 122-amino acid sequence from MGD proteins according to Awai et al. (2001), showed that type-B MGD family members are widely present in angiosperms, and are strictly conserved (Figure S2; Table S1), which implies a universal importance for type-B enzymes in angiosperms. MGD orthologs identified in the moss Physcomitrella patens are classified as type-A MGD enzymes, and are distant from type-B MGD isoforms (Bottéet al., 2005; Figure S2). We also found that the fern Selaginella moellendorffii possesses an ortholog of type-A MGD. Because type-B MGD has thus far been identified only in angiosperms, it is possible that, during the evolutionary process, seed plants have acquired a special requirement for MGDG biosynthesis that is catalyzed by type-B MGD isoforms, presumably to adapt to Pi-limited growth conditions in the soil. Like type-B MGD enzymes, DGD2 isoforms, which are distinguishable from DGD1 at their N termini, are also present in many plant species (Gaude et al., 2004; Kelly and Dörmann, 2002). These data suggest that the alternative galactolipid pathway of type-B MGD–DGD2 may have been developed and conserved widely in seed plants to improve their survival and fecundity under Pi-limited growth conditions.
Plant material and growth conditions
Surface-sterilized seeds of Arabidopsis thaliana wild type (Columbia-0) and type-B MGD mutants were incubated at 4°C in darkness for 3 days prior to seeding on medium. The incubation temperature after seeding was 23°C for all growth conditions. For the lipid analysis, plants were germinated in liquid MS medium containing 1% (w/v) sucrose under continuous light for 3 days (Figure 2a), or in darkness for 7 days (Figure 2b). For the analyses shown in Figures 1b, 3, 4 and 5f, plants were grown on solidified MS medium with 0.8% (w/v) agar for 10 days, and were then grown on solidified Pi-replete (1.0 mm) or Pi-depleted (0 mm) medium, prepared as described by Härtel et al. (2000), for another 10 days. For the growth analysis (Figure 5a–e), plants were seeded on Pi-replete or Pi-depleted medium after low-temperature treatment, and were then grown on vertically placed medium for 16 days under continuous light.
Isolation of the T-DNA insertion mutants
mgd2-1 and mgd3-2 plants were screened from the T-DNA insertional mutant population of the Kazusa DNA Research Institute (Kisarazu, Japan) by PCR and DNA gel blot analysis, as described by Masuda et al. (2003). mgd3-1 (SALK_ 084186) was obtained from the Salk Institute Genomic Analysis Laboratory (http://www.salk.edu). The gene-specific and T-DNA-specific primers that were used to confirm the mutations are as follows: MGD2 forward primer, 5′-aacaggaagcctattgggcaaatgg-3′; MGD2 reverse primer, 5′-cgctagcaagaggacctcgttgctc-3′; MGD3 forward primer, 5′-atcaccaaccagccacatag-3′; MGD3 reverse primer, 5′-gtacggtggcaagtgtttag-3′; T-DNA right border (forward) primer for mgd2-1, 5′-ttcccttaattctccgctcatgatc-3′; T-DNA left border (forward) primer for mgd3-1, 5′-accatcaaacaggattttcgcctgctgggg-3′; T-DNA left border (reverse) primer for mgd3-2, 5′-ataacgctgcggacatctac-3′. The positions of the T-DNA insertion in mgd2-1, mgd3-1 and mgd3-2 were determined by sequencing the PCR fragment covering each T-DNA junction. mgd3-1 and mgd3-2 were selected on medium containing 50 μg ml−1 kanamycin; mgd2-1 was selected with 50 μg ml−1 hygromycin. Because both mgd2-1 and mgd3-1 are in the Columbia background, whereas mgd3-2 is in the Wassilewskija background, we crossed mgd2-1 and mgd3-1 to generate an mgd2 mgd3 double mutant. The homozygous mgd2 mgd3 double mutant was identified in the F2 progeny of mgd2-1 and mgd3-1 by genomic PCR analysis using the primers described above.
Total RNA was extracted from Pi-starved seedlings of wild type and type-B MGD mutants using the RNeasy Plant Mini kit (Qiagen, http://www.qiagen.com), following to the manufacturer’s instructions. RT was performed using an RT-PCR kit, RNA PCR kit v3.0 (Takara Bio, http://www.takara-bio.com) and an oligo(dT)12 primer. The PCRs for the three MGD genes and ACTIN8 (an internal standard) were performed as described in Awai et al. (2001) and Kobayashi et al. (2007). The PCR-amplified samples were electrophoresed on 1.2% (w/v) agarose gels, and were detected with 0.1 μg ml−1 ethidium bromide.
Total lipid extraction was carried out according to the method described by Bligh and Dyer (1959). The polar membrane lipids were separated by two-dimensional thin-layer chromatography as described by Kobayashi et al. (2007). Lipids isolated from silica gel plates were methylated, and fatty acid methyl esters were quantified by gas chromatography using myristic acid as an internal standard, according to the method described by Kobayashi et al. (2006).
Measurement of primary root length, fresh weight and effective quantum yield
Root growth of each plant was captured on a digital camera (COOLPIX2000; Nikon, http://www.nikon.com) from day 8 to day 16 after seeding. The primary root length was digitally analyzed using image analyzing software (Scion Image; Scion, http://www.scioncorp.com). After root length analysis on day 16, the same samples were used for PAM analysis, and subsequently for fresh weight measurement. PAM analysis was performed by mini pam (Walz, http://www.walz.com), and effective quantum yield was determined according to the protocol supplied. The fresh weights of the shoots and roots of each plant were measured separately. In the case of Pi-starved samples, eight seedlings were used for one measurement of fresh weight, and four seedlings were used for one measurement in Pi-sufficient samples.
Determination of Pi content
Inorganic Pi was extracted separately from shoots and roots, and the content was measured using a phosphomolybdate colorimetric assay according to the method described by Chiou et al. (2006). Tissue samples (0.1 g fresh weight) were homogenized with 250 μl extraction buffer (10 mm Tris, 1.0 mm EDTA, 100 mm NaCl, and 1.0 mmβ-mercaptoethanol, pH 8.0). After centrifugation (25 000 g) for 10 min, 100 μl of the supernatant was mixed with 900 μl of 1% glacial acetic acid, and was then incubated at 42°C for 30 min. After centrifugation (25 000 g) for 5 min, 300 μl of the supernatant was mixed with 700 μl of assay solution [0.35% (w/v) NH4MoO4, 0.86 N H2SO4 and 1.4% (w/v) ascorbic acid], and was then incubated at 42°C for 30 min. The Pi concentration was measured at A820.
Oxygen consumption in the roots
For respiration activity measurement in roots (data not shown), intact roots (approximately 0.4 g fresh weight) were prepared from plants grown on MS media for 10 days, followed by growth on Pi-controlled media for another 10 days. Respiration activity was measured as a decrease of O2 concentration using a Clark-type O2 electrode (Leaflab 1; Hansatech, http://www.hansatech-instruments.com) according to the manufacturer’s instructions.
This work was supported by a Grant-in-Aid for Scientific Research on Priority Areas (No. 18056007 and No. 20053005) from the Ministry of Education, Sports, Science and Culture in Japan. KK was supported by a research fellowship of the Japan Society for the Promotion of Science for young scientists.