Global Analysis of the Role of Autophagy in Cellular Metabolism and Energy Homeostasis in Arabidopsis Seedlings under Carbon Starvation[OPEN]

The multifaceted phenotypes of carbon-starved, autophagy-deficient mutant plants during seedling establishment are ascribed to the global effects of autophagy on central metabolism and energy status. Germination and early seedling establishment are developmental stages in which plants face limited nutrient supply as their photosynthesis mechanism is not yet active. For this reason, the plant must mobilize the nutrient reserves provided by the mother plant in order to facilitate growth. Autophagy is a catabolic process enabling the bulk degradation of cellular constituents in the vacuole. The autophagy mechanism is conserved among eukaryotes, and homologs of many autophagy-related (ATG) genes have been found in Arabidopsis thaliana. T-DNA insertion mutants (atg mutants) of these genes display higher sensitivity to various stresses, particularly nutrient starvation. However, the direct impact of autophagy on cellular metabolism has not been well studied. In this work, we used etiolated Arabidopsis seedlings as a model system for carbon starvation. atg mutant seedlings display delayed growth in response to carbon starvation compared with wild-type seedlings. High-throughput metabolomic, lipidomic, and proteomic analyses were performed, as well as extensive flux analyses, in order to decipher the underlying causes of the phenotype. Significant differences between atg mutants and wild-type plants have been demonstrated, suggesting global effects of autophagy on central metabolism during carbon starvation as well as severe energy deprivation, resulting in a morphological phenotype.


INTRODUCTION
Macroautophagy (referred to hereafter as "autophagy") is a conserved eukaryotic mechanism, which is classically defined as the degradation of cytoplasmic constituents in the lytic organelle (vacuoles in yeast and plants and lysosomes in mammals) (Reggiori and Klionsky, 2013). The general targets of autophagy vary from long-lived proteins to protein complexes and even entire organelles (Reumann et al., 2010). Morphologically, autophagy begins with the formation of cupshaped double membranes, which expand to form autophagosomes engulfing malfunctioning or unneeded macromolecules and organelles and transport them for degradation inside the vacuole. Upon arrival of the autophagosomes to the vacuoles, their outer membranes fuse with the tonoplast, creating single-membrane vesicles inside the vacuole, termed "autophagic bodies." The autophagic bodies and their contents are then degraded inside the vacuole, providing recycled materials to build new macromolecules (Bassham, 2009). Nutrient starvation has been one of the hallmark inducers of autophagy in yeast, mammals, and plants (Reumann et al., 2010;Singh and Cuervo, 2011;Li and Vierstra, 2012). The autophagy mechanism in presumed to play a role in recycling during starvation, thus supplying the cell with nutrients during the stress period (Li and Vierstra, 2012). It has been shown to be involved in protein and lipid degradation (Onodera and Ohsumi, 2005;Singh and Cuervo, 2011), as well as to be associated with several metabolic disorders in animals (Singh and Cuervo, 2011).
The genes participating in the autophagic process (termed ATG genes) were originally discovered in yeast (Saccharomyces cerevisiae) via the isolation of autophagy-defective mutants whose cells show little or no accumulation of autophagic bodies during nutrient starvation (Liu and Bassham, 2012). Many ATG genes are conserved in evolution, and homologs to the yeast genes have been found in many organisms including mammals and plants Ryter et al., 2013). Several atg mutants have been characterized in the model plant Arabidopsis thaliana (Doelling et al., 2002;Hanaoka et al., 2002;Yoshimoto et al., 2004;Xiong et al., 2005;Harrison-Lowe and Olsen, 2008;Phillips et al., 2008). Among the common phenotypes of these mutants are early senescence in comparison to wild-type plants as well as hypersensitivity to carbon and nitrogen starvation (Bassham, 2009). Interestingly, the early senescence phenotype of atg mutants was shown to be the consequence of salicylic acid (SA) accumulation, apparently regulated by autophagy (Yoshimoto et al., 2009). Autophagy in plants has also been shown to function during abiotic stresses such as salt, drought, and oxidative stress (Xiong et al., 2007;Liu et al., 2009) as well as in response to pathogens (Liu and Bassham, 2012). In addition, selective autophagic degradation of chloroplasts has been demonstrated (Ishida et al., 2014), and evidence of peroxisome and starch degradation by autophagy has been recently provided (Farmer et al., 2013;Kim et al., 2013;Wang et al., 2013;Avin-Wittenberg and Fernie, 2014). Although hypersensitivity to nutrient starvation is a major phenotype of atg mutants, only very few recent studies have attempted to uncover the metabolic implications underlying it (Guiboileau et al., 2012;Izumi et al., 2013;Masclaux-Daubresse et al., 2014), and although we are making great strides in our understanding of the metabolic impact of autophagy in plants, many questions still remain open.
Seed germination and seedling establishment are two developmental periods in which there is no carbon assimilation. (A) atg mutant seeds and their respective wild-type controls (Col for atg5-1, atg5-3, and atg7-2; Ws for atg4a4b) were sown on 0.53 MS plates without sucrose and grown in the dark for 7 d after imbibition. Representative pictures of the etiolated seedling are shown as well as measurement of hypocotyl length performed using ImageJ (n = 15). Data are presented with SE; significant difference compared with the wild type (P < 0.05 in Student's t test) is denoted by an asterisk. (B) atg mutant seeds and their respective wild-type controls (Col for atg5-1, atg5-3, and atg7-2; Ws for atg4a4b) were sown on 0.53 MS plates containing 1% sucrose and grown in the dark for 7 d after imbibition. Representative pictures of the etiolated seedling are shown as well as measurement of hypocotyl length performed using ImageJ (n = 15). Data are presented with SE. (C) Wild-type, NahG, and atg5.NahG seeds were sown on 0.53 MS plates without sucrose and grown in the dark for 7 d after imbibition. Representative pictures of the etiolated seedling are shown as well as measurement of hypocotyl length performed using ImageJ (n = 15). Data are presented with SE; significant difference compared with the wild type (P < 0.05 in Student's t test) is denoted by an asterisk. (D) atg mutant and wild-type control seeds were sown on 0.53 MS plates without sucrose, imbibed for 48 h, and transferred to continuous light conditions. The number of germinated seedlings (defined by radical protrusion) was scored each day for 4 d, and average percent germination (n = 6) is presented with SE. Detailed results of the assay for the depicted lines as well as additional lines are presented in Supplemental Table 1. (E) atg mutant and wild-type control seeds were sown on 0.53 MS plates without sucrose, imbibed for 48 h, and transferred to continuous light conditions. The number of germinated seedlings (defined by the appearance of two green cotyledons) was scored each day for 4 d, and average percent germination (n = 6) is presented with SE. Detailed results of the assay for the depicted lines as well as additional lines are presented in Supplemental Table 1. Seeds possess storage tissues, containing reserves necessary for the establishment of the seedling (Arc et al., 2011). Seed germination is associated with the degradation and mobilization of these storage reserves (Fait et al., 2006). This process is crucial in providing fuel for growth until the growing seedling becomes photoautotrophic (Pritchard et al., 2002). Since autophagy functions as a degradation pathway facilitating nutrient mobilization, it is safe to assume it serves a function during seedling establishment. Indeed, atg mutants have been shown to display impaired seedling establishment . However, the reasons for this phenotype, and specifically the impact on energy metabolism, have not been well studied.
In this study, we investigated etiolated Arabidopsis seedlings as a model for carbon starvation during seedling establishment.
We demonstrate that under our experimental conditions, atg mutants display delayed growth in comparison to wild-type seedlings. Metabolic profiling revealed reduced levels of free amino acids in the mutants, while proteomic analysis displayed accumulation of proteins in the mutants. Metabolic flux analysis also suggests increased respiration in atg mutants in addition to decreased net protein biosynthesis compared with the wild type. Examination of the lipid composition of the mutants revealed altered lipid levels, strengthening our hypothesis of altered metabolism as well as suggesting the occurrence of starvation stress responses in the mutants. We suggest that only a comprehensive large-scale analysis including both steady state metabolite levels as well as metabolic flux will enable us to decipher such compound phenotypes as that of autophagy deficiency. atg mutants as well as their respective wild-type and SA controls (NahG and atg5.NahG) were sown on 0.53 MS plates and grown in the dark for 6 d after imbibition. Metabolic content was analyzed using GC-MS (n = 4 to 8). Detailed results of the assay for the depicted lines as well as additional lines are presented in Supplemental Table 2. (A) PCA of primary metabolite levels. (B) Log 2 values of the relative metabolic content are presented as a heat map. Significant differences compared with the wild type following Student's t test are denoted by one asterisk (P < 0.05) or two asterisks (P < 0.01). (C) Relative levels of BCAA compared with the wild type. Data are presented with SE; significant differences following Student's t test are denoted by an asterisk (P < 0.05).

Arabidopsis Etiolated atg Mutant Seedlings Display a Delayed Growth Phenotype
We have chosen to use etiolated Arabidopsis seedlings as a model system for carbon starvation since the only carbon source available to these seedlings is that coming from the seed, rendering the system heterotrophic. Additionally, no photosynthesis has occurred, thus eliminating the possibility of differential carbon accumulation in the atg mutants compared with the wild type. We began our investigation by assessing the morphological phenotype of atg mutants germinated under carbon starvation conditions in comparison to wild-type seedlings. Seed were sown on 0.53 Murashige and Skoog (MS) plates (in the absence of a carbon source) and grown in the dark for 7 d after imbibition. All atg mutants displayed visibly shorter hypocotyls than their respective wild-type controls (Columbia [Col] for atg5-1, atg5-3, and atg7-2 and Wassilewskija [Ws] for atg4a4b). This significant difference was also evident when hypocotyl length measurements were conducted ( Figure 1A). We were able to recover the phenotypes by supplementing the medium with an external carbon source. The size of atg mutants grown on 0.53 MS medium plates containing 1% sucrose was not different from that of their respective wild-type controls ( Figure  1B). In order to verify that the autophagy mechanism is indeed functioning under our experimental conditions, we used Arabidopsis plants expressing the fusion protein GFP-ATG8f-HA (containing ATG8f fused to green fluorescent protein and the human influenza hemagglutinin [HA] tag) in the wild-type background to determine autophagic flux as described previously (Sláviková et al., 2005). During autophagosome elongation, the C terminus of ATG8 is cleaved by ATG4, and the protein is subsequently conjugated to the lipid phosphatidylethanolamine (PE) and inserted into the growing autophagosome membrane. Upon fusion of the autophagosome with the vacuole, ATG8 decorating the inner membrane of the autophagosome is degraded by vacuolar proteases (Li and Vierstra, 2012). By using a construct in which ATG8 is tagged from both termini, it is possible to follow the rate of ATG8 cleavage by ATG4, suggesting autophagosome elongation, and the rate of ATG8 atg mutants as well as their respective wild-type and SA controls (NahG and atg5.NahG) were sown on 0.53 MS plates and grown in the dark for 6 d after imbibition. Secondary metabolite content was analyzed by LC-MS (n = 5 to 6). Detailed results of the assay for the depicted lines as well as additional lines are presented in Supplemental Table 3. (A) PCA of secondary metabolite levels. (B) Log 2 values of the relative metabolic content are presented as a heat map (n = 5 to 6). Significant differences compared with the wild type following Student's t test are denoted by one asterisk (P < 0.05) or two asterisks (P < 0.01). degradation, suggesting autophagosome fusion with the vacuole. The ratio between these two steps is used as a measure of autophagic flux. GFP-ATG8f-HA seeds were grown in the dark for 6 d after imbibition on 0.53 MS plates with or without 1% sucrose; protein extracts from the seedlings were then run on SDS-PAGE gel, and an immunoblot assay was performed using an antibody raised against GFP (Supplemental Figure 1). The samples taken from seedlings grown on sucrose contained three protein bands, corresponding to the full-length fusion protein as well as its degradation products ATG8f-GFP and free GFP. However, the samples taken from seedlings grown without sucrose contained only the free GFP band, suggesting increased degradation of the GFP-ATG8f-HA protein and, thus, an increase in autophagic flux under our experimental conditions. This observation is in agreement with a previous study showing similar results (Sláviková et al., 2005).
It has been previously demonstrated that atg mutants accumulate SA as they age, causing an early senescence phenotype (Yoshimoto et al., 2009). This phenotype could be recovered by crossing atg mutant plants with plants overexpressing the bacterial protein NahG, encoding a SA hydroxylase, thus producing stay-green atg mutants. In order to verify that the short hypocotyl phenotype was not the result of early senescence during seed development or of SA accumulation in the seed tissue, we examined the hypocotyl length of 7-d-old atg5.NahG seedlings grown on plates without sucrose. The atg5.NahG seedlings displayed significantly shorter hypocotyls than wildtype plants, demonstrating that decreasing the SA levels in the plants was not able to recover the morphological phenotype ( Figure 1C).
We next wished to ascertain whether the difference in size stems from delayed germination or reduced growth following germination. For this purpose, we compared the germination of wild-type and atg mutants, as well as the SA control lines (atg5.NahG seedlings; Figures 1D and 1E; Supplemental Table 1). Two parameters were scored: radical protrusion, indicating germination; and appearance of two green cotyledons, indicative of early seedling development (Gao et al., 2011). A delay in both parameters was observed for the atg mutant lines in comparison to the wild type, though eventually all lines reached the same germination percentage ( Figures 1D and 1E). This behavior was also observed for the SA control lines (Supplemental Table 1). These results may indicate that the difference in plants size stems, at least in part, from delayed growth.

Differential Metabolic Response of atg Mutants under Carbon Starvation
Autophagy has been linked to nutrient recycling under starvation conditions (Singh and Cuervo, 2011;Guiboileau et al., 2013;Izumi et al., 2013;Reggiori and Klionsky, 2013;Ryter et al., 2013). We therefore set out to investigate the metabolic content of etiolated atg mutants under carbon starvation compared with wild-type plants. We were able to identify 30 primary metabolites using gas chromatography-mass spectrometry (GC-MS) analysis (Supplemental Table 2). Principal component analysis (PCA) demonstrated a good separation between the wild-type and atg mutant lines. NahG and atg5.NahG clustered with the wild-type line and atg mutant lines, respectively. A significant decrease in free amino acid levels was observed in all atg mutants compared with the wild type ( Figure 2B). This decrease was not recovered in the SA control lines, suggesting that the phenotype is likely to be SA independent ( Figure 2B, NahG and atg5.NahG). Of special interest were the reduced levels of lysine and the branched-chain amino acids (BCAAs) isoleucine and valine in atg mutants ( Figure 2C). These amino acids have been shown to function as electron donors for the tricarboxylic acid (TCA) cycle and mitochondria electron transport chain under carbon starvation (Araújo et al., 2010), and their decreased steady state levels in the atg mutants as well as wild-type and SA controls (NahG and atg5.NahG) were sown on 0.53 MS plates and grown in the dark for 6 d after imbibition.
(A) Total protein amount was analyzed using BCA protein assay (n = 6). Significant differences compared with the wild type following Student's t test are denoted by one asterisk (P < 0.05) or two asterisks (P < 0.01). (B) Separation of proteins on SDS-PAGE gel. Bands displaying higher intensity in atg mutants compared with the wild type and NahG (marked by arrowheads) were cut from the gel for mass spectrometry identification. Mass spectrometry results are detailed in Table 1. mutants likely suggest altered mitochondrial function under carbon starvation. Additionally, reduced steady state levels of the organic acids malate, fumarate, and dehydroascorbate were observed ( Figure 2B).
We also identified secondary metabolites using liquid chromatography-mass spectrometry (LC-MS) analysis ( Figure 3; Supplemental Table 3). Surprisingly, while PCA demonstrated a good separation between the wild-type and atg mutant lines, a cross between atg5 and NahG abolished this separation (atg5. NahG, Figure 3A). This was also evident in many of the metabolites exhibiting a significant difference in at least one of the atg mutant lines compared with the wild type, where the difference did not exist in the atg5.NahG line ( Figure 3B). This observation implies a putatively considerable role for SA in determining the secondary metabolic composition of the atg mutants. However, several secondary metabolites were, by contrast, less abundant in atg mutants compared with the wild type in a presumably SAindependent manner ( Figure 3B). However, several of these are currently unknown with respect to their chemical identity. Nevertheless, they may serve as interesting subjects for future research.

Mutant Seedlings under Carbon Starvation
Following our observation of reduction of the steady state level of many free amino acids in atg mutants and given that proteins are degradation targets of the autophagy machinery (Li and Vierstra, 2012), we next decided to examine the protein content of carbonstarved atg mutants compared with the wild type. We initially measured the total protein content of carbon starved atg mutants. Interestingly, both atg5 mutant lines presented with higher protein content than wild-type plants, suggesting a possible impairment in protein degradation. The atg5.NahG plants displayed the same accumulation ( Figure 4A); however, surprisingly, atg7-2 seedlings did not accumulate higher protein content. This may point to a milder phenotype and misaccumulation of fewer proteins. Indeed, when equal amounts of protein were separated by SDS-PAGE, several protein bands displayed stronger intensity in all atg mutants compared with the wild type, suggesting the differential accumulation and degradation of specific proteins ( Figure 4B; Supplemental Figure 2). It is important to note that the same protein accumulation was observed in atg5.NahG but not in the NahG samples ( Figure 4B). The bands showing a difference in accumulation ( Figure 4B, marked by arrows; Supplemental Figure  2) were excised and subjected to in-gel digestion with trypsin, and the resulting peptides were identified using liquid chromatography coupled to tandem mass spectrometry (Table 1). We were able to identify an accumulation of cruciferin 3 in atg mutants compared with the wild type. Cruciferin is the most abundant storage protein in Arabidopsis seeds (Wan et al., 2007); thus, its accumulation in atg mutants points to a possible impairment in protein degradation, complying with the known function of the autophagy mechanism (Li and Vierstra, 2012). Two other interesting proteins accumulating in atg mutants compared with the wild type were malate synthase (MLS) and multifunctional protein 2 (MFP2). Both proteins have been shown to be involved in b-oxidation of fatty acids and seedling establishment (Rylott et al., 2006;Kim et al., 2013). Additionally, MLS has been previously shown to accumulate in atg mutant seedlings, further validating our proteomic analysis (Kim et al., 2013). Surprisingly, we identified one protein overaccumulating in wild-type seedlings rather than atg mutants. Cytosolic triosephosphate isomerase is an enzyme involved in glycolysis (Giegé et al., 2003), suggesting a possible upregulation of this pathway in wild-type seedlings compared with atg mutants. Mass spectra obtained from peptides from in-gel digestion of protein bands marked in Figure 6B were matched using Mascot and a database of Arabidopsis proteins. Individual ions scores >22 indicate identity or extensive homology (P < 0.05). Data are presented for two biological replicates (denoted as 1 and 2 under the line name). Band size is the size of the band corresponding to the bands marked in Figure 6B. The description is the annotation of the matched protein. AGI is the genome accession number. MW, calculated mass of the matched protein; MS, mascot protein score; P, number of matching peptides; nd, not detected.

Etiolated atg Mutants Display Increased Respiration in Response to Carbon Starvation
Given that the morphological and metabolic phenotype of atg mutants observed thus far suggests an energy deprivation phenotype, we were interested in directly checking the energy status of the plants. As the seedlings are etiolated, photosynthesis does not play a role in energy generation, and we therefore directly examined the respiration parameters of carbon starved atg mutants. We evaluated 14 CO 2 evolution using positionally labeled 14 C-Glc in order to assess flux through the major pathways of carbohydrate oxidation (Nunes-Nesi et al., 2005). Wild-type and atg5-1 seedlings were grown in the dark without sucrose for 7 d and then supplied with [1-14 C], [3:4-14 C], or [6-14 C]Glc (1 mM with specific activity of 1.85 MBq/mmol) for a duration of 2 h. During this time, we collected the 14 CO 2 evolved at 40-min intervals. A relatively short time frame and low glucose concentration were chosen in order not to disturb the starvation status of the plants, thereby avoiding recording the differences in recovery from starvation between the mutant and the wild type. CO 2 can be released from the C1 position by enzymes that are not associated with mitochondrial respiration, unlike CO 2 evolution from the C3:4 positions of Glc (Rees and Beevers, 1960). Thus, the ratio of CO 2 evolution from C1 to C3:4 positions of Glc provides an indication of the relative rate of the TCA cycle with respect to other processes of carbohydrate oxidation. Though the level of C1 emission was higher than the level of C3:4 emission in both lines ( Figures 5A and 5B), the level of CO 2 emission from C1 was comparable between atg5-1 and the wild type, while the level of C3:4 emission was slightly higher in atg5-1 than in the wild type. Thus, when considering the C1/C3:4 CO 2 emission ratio, the ratio for the wild type is higher than that of atg5-1 (1.85 and 1.66 after 120 min, respectively), suggesting a higher proportion of carbohydrate oxidation performed by the TCA cycle in atg5-1. CO 2 emission from the C6 position was relatively similar between both lines, suggesting that pentane synthesis was unaltered ( Figure 5C). Since some amino acids, such as BCAA and lysine, may be used as substrates for respiration (Araújo et al., 2011), and since we have observed a decrease in the steady state levels of these amino acids in atg mutants compared with the wild type ( Figure 2C), we set out to investigate lysine metabolism in atg mutants. Wild-type and atg5-1 seedlings were grown in the dark without sucrose for 7 d and then supplied with [U-13 C]Lys for a duration of 2 h. Seedling samples were collected every 40 min, washed, and immediately frozen. 13 C label distribution was calculated as described in Methods in order to determine metabolic flux through several metabolites (summarized in Table  2). Increased label distribution in atg5-1 seedlings was demonstrated for malate, a TCA cycle intermediate, further strengthening our observation from CO 2 evolution. In addition, atg5-1 showed increased label distribution in glutamate, alanine, and aspartate compared with the wild type, suggesting increased metabolic flux in the mutants.

Decreased Flux to Net Protein Synthesis in Carbon-Starved atg Mutants
As the seedlings were grown under carbon starvation condition, we hypothesized these conditions might have an effect on other major pathways of carbon metabolism. We thus performed [U-14 C] Glc labeling of wild-type and atg5-1 seedlings. The samples were incubated in the presence of 1 mM [U-14 C]Glc for 80 min, extracted, and fractioned into organic acids, amino acids, starch, protein, cell wall, phosphoesters, and sucrose in order to analyze flux alterations between wild-type and atg5-1 plants (Table 3). This time point was chosen as an intermediate time point derived from our CO 2 evolution experiment, again to assure examination of the response to carbon starvation rather than recovery from it. The atg5-1 seedlings incorporated and metabolized significantly higher amounts of label compared with wild-type seedlings, possibly suggesting the mutants are more energy deprived, in accordance with the observed morphological phenotype. Lower CO 2 evolution was also observed for the atg5-1 seedlings. Label redistribution in the protein fraction was also altered in atg5-1 seedlings, with significantly lower labeling of the protein pool in atg5-1 plants. In order to estimate absolute fluxes in atg5-1 seedlings compared with the wild type, we calculated the specific activities of the hexose phosphate pool and used these to calculate the fluxes to starch, sucrose, cell wall, respiration, and protein using the assumptions documented by Geigenberger et al. (2000). Surprisingly, no difference in respiration was observed between the wild type and atg5-1, in contrast to our observation from CO 2 evolution and [U-13 C]Lys feeding experiments. However, this contradiction is likely due to the time frame of the experiments. The major changes in respiration were observed after 2 h in both experiments. It is possible that after 80 min of incubation, the label was not distributed sufficiently for the difference to be evident. On the other hand, as also seen from the label redistribution, less flux in funneled to protein in atg5-1 mutants, suggesting lower net protein synthesis in this line. Taken together with the higher level of total proteins observed in atg mutants ( Figure 4A), we postulate that the higher level of proteins stems from decreased degradation rather than increased synthesis.

Altered Lipid Composition of atg Mutants under Carbon Starvation
Oil, in the form of triacylglycerol (TAG), is a major seed storage reserve in many plant species, including Arabidopsis (Graham, 2008). Since etiolated seedlings rely on storage reserves for their supply of energy, we set out to profile the lipid composition of our carbon starved mutants. A total of 159 individual lipids were identified, belonging to 12 different lipid classes. PCA revealed a distinct separation between wild-type and atg mutant lines, demonstrating a clear difference between the lipid compositions of the individual lines ( Figure 6A). The NahG line was also clearly separated from the wild type; however, the lipid composition of atg5.NahG could not be differentiated from that of atg5-3. Interestingly, not all atg mutants clustered together in the PCA, possibly pointing to a gradient in their phenotypes. In addition, an examination of the lipid composition of Ws and atg4a4b revealed an ecotypic difference between the wild-type lines, increasing the level of complexity of this analysis (Supplemental Data Set 1). We first focused on the TAG composition of the different lines, as it is the main storage lipid in the seed. A significant accumulation of most TAG species was observed for both atg5-1 and atg5-3 ( Figure 6B). While a mild accumulation was also observed for atg7-2, in this instance, it was not statistically significant. During seed germination and early seedling establishment, TAGs are hydrolyzed into fatty acids (FAs) and glycerol. FAs are then degraded in the glyoxysome by b-oxidation (Graham, 2008). An accumulation of FAs similar to that of TAGs was observed for atg5-1 and atg5-3, suggesting a possible impairment of b-oxidation. However, this accumulation was not seen for atg7-2 ( Figure 6C). Interestingly, an increase in lysophosphatidylcholine and lysophosphatidylethanolamine (LysoPE) was also observed for all atg mutants analyzed ( Figure  6D). These lysolipids, which are editing products of the membrane lipids PC and PE, have previously been shown to accumulate under freezing stress in Arabidopsis (Welti et al., 2002). These data are in keeping with the apparent general retarded lipid degradation of the mutants.
Two lipid species have been documented as being involved in autophagosome formation. Phosphatidylinositol (PI) is phosphorylated during autophagosome vesicle nucleation by PI3-kinase and is used as an anchor for the PI3K complex (Li and Vierstra, 2012). In addition, PE is conjugated to ATG8 during autophagosome elongation, thus enabling its incorporation into the growing autophagosome membrane (Liu and Bassham, 2012). We detected an increase in the PI levels of all atg mutant lines compared with the wild type, possibly suggesting an attempt to compensate for the lack of autophagy ( Figure 7A). By contrast, the levels of large PE molecules (PE42), containing very-long-chain fatty acids, significantly decreased in all atg mutants examined, corresponding to the increase in LysoPE levels ( Figure 7B).

DISCUSSION
Hypersensitivity to carbon starvation is one of the most wellcharacterized phenotypes of plant autophagy mutants (Doelling et al., 2002;Hanaoka et al., 2002;Thompson et al., 2005;Xiong et al., 2005). We therefore selected a carbon starvation system to study the metabolic implication of autophagy in plants. Thus far, few studies have tackled this issue, using either nitrogen starvation (Guiboileau et al., 2013;Masclaux-Daubresse et al., 2014) or carbon starvation in the form of darkness combined with starch accumulation defects (Izumi et al., 2013) as their experimental systems. In this research, we chose etiolated Arabidopsis seedlings grown on medium without supplementary sucrose as a carbon starvation model system. The rationale behind this was that it allowed us to utilize the advantages of a "closed system" as well as providing insights concerning the function of autophagy during .53*10 25 6 9.53*10 25 0 6 0 1.26*10 24 6 1.09*10 24 0 6 0 4.30*10 24 6 2.14*10 24 2.25*10 25 6 1.60*10 25 8.36*10 24 6 1.65*10 24 4.17*10 24 6 1.14*10 24 Seven-day-old etiolated seedlings grown without exogenous sucrose were incubated in the presence of 10 mM MES-KOH (pH 6.5) containing 5mM [U-13 C]-lysine and collected at different time points. 13 C sum accumulation (nmol/g fresh weight) is presented as the average of four biological replicates, and variation is given as 6 SE. Significant differences following Student's t test (P < 0.05) compared to the wild type are denoted in bold. GABA, g-aminobutyric acid.
early seedling establishment, in which nutrient acquisition through reserve mobilization plays an important role. Indeed, we were able to see a clear growth phenotype for atg mutants compared with the wild type, notably a phenotype that was recovered by the addition of exogenous sucrose ( Figures 1A and 1B). This phenotype is in keeping with the previous observation that atg5 seedlings were impaired in their ability to generate true leaves in a medium without sucrose . Taken together with the results of our germination assay ( Figures 1D and 1E), we postulate that the growth impairment phenotype of atg mutants becomes more severe as seedling growth proceeds. The contribution of nutrients from the mother tissue has previously been shown to play an important role in seed content and also to have an impact on future seedling establishment (Andriotis et al., 2012). Since atg mutants additionally possess an early senescence phenotype that affects seed set (Yoshimoto et al., 2004;Thompson et al., 2005), we used a previously characterized stay-green atg mutant to control for maternal tissue effects (Yoshimoto et al., 2009). The fact we could not recover the phenotype ( Figure 1C), taken alongside our documentation of increased autophagic flux under our experimental conditions (Supplemental Figure 1), suggests the phenotype is indeed likely to stem from lack of function of the autophagy mechanism during germination and seedling establishment. During our analysis of primary metabolites, we observed a reduction in free amino acids in atg mutants compared with wild-type seedlings ( Figure 2B). This result is in accordance with previous results from mice and yeast (Onodera and Ohsumi, 2005;Ezaki et al., 2011). In addition, autophagy was implicated in amino acid release in Arabidopsis during nighttime starvation (Izumi et al., 2013), further corroborating our observation. Surprisingly, a recent study examining autophagy under nitrogen starvation detected, by contrast, an amino acid accumulation in atg mutants compared with the wild type (Masclaux-Daubresse et al., 2014). A possible explanation for this discrepancy may stem from the age of the studied plants. While Masclaux-Daubresse et al. (2014) analyzed plants ranging from 30 to 75 d after sowing, we examined 6-d-old seedlings. Not only are these two developmental systems morphologically distinct, but older plants are also autotrophic, in contrast to etiolated seedlings. Thus, the considerable differences in experimental systems might be responsible to the phenotypic inconsistency. The observation of context-dependent autophagy results is not without precedence since previous studies investigating the role of autophagy in the hypersensitive response demonstrated that shorter incubation times (Hofius et al., 2009) yielded opposite results compared with longer incubation times (Liu et al., 2005). Considering this previous example implies that the opposite metabolic phenotypes observed here may additionally arise from the longer exposure to nitrogen starvation in the previous study compared with carbon starvation in our study. It is important to note that the time frame of the nutrient restriction in both instances is most likely the most physiologically relevant since carbon and nitrogen starvation operate on different timescales.
Given that the levels of free amino acids were reduced in atg mutants, we had anticipated seeing an increase in total protein amount in the mutants. Indeed, such an increase was observed in atg5 mutants, but not in atg7-2 ( Figure 4A). It has been previously shown that, although all Arabidopsis atg mutants lack autophagosomes, some display a more severe phenotype than others for unknown reasons (Yoshimoto et al., 2009). It is thus not unexpected to observe a milder phenotype for atg7-2 Seven-day-old etiolated seedlings grown without exogenous sucrose were incubated in the presence of 10 mM MES-KOH (pH 6.5) containing 1 mM [U-14 C]Glc (specific activity of 1.85 MBq/mmol). Each sample was extracted with boiling ethanol, and the amount of radioactivity in each metabolic fraction was determined as described in Methods. Values are means 6 SE (n = 3 to 4). Significant differences following Student's t test (P < 0.05) compared to the wild type are denoted in bold. FW, fresh weight. compared with atg5 in our experimental system. This milder atg7-2 phenotype was also observed in the lipid analysis of the mutants (Figures 6A to 6C). On the contrary, accumulation of selected proteins was observed in all atg mutants ( Figure 4B; Supplemental Figure 2). Proteomic analysis of the differential bands (Table 1) revealed the accumulation of several interesting proteins, among which was cruciferin, an Arabidopsis storage protein (Wan et al., 2007). The higher abundance of this protein in atg mutants, together with the increase in total protein amounts observed in atg5 mutants, corresponds with the known role of autophagy in protein degradation (Li and Vierstra, 2012). Also of note is the fact that although the total protein size of cruciferin is ;60 kD, we identified the protein in the 30-kD region of the gel, suggesting that partial degradation of the protein occurs in the mutants. Although Arabidopsis contains storage proteins, its main storage compound is oil, rather than proteins (Kim et al., 2013). It would therefore be interesting in future studies to examine whether seedling establishment is more severely impaired in autophagy mutants of species in which proteins are a more abundant storage compound, such as legumes (Gallardo et al., 2008). Since autophagy is intimately connected to nutrient recycling and the maintenance of cellular energy status (Singh and Cuervo, atg mutants as well as their respective wild-type and SA controls (NahG and atg5.NahG) were sown on 0.53 MS plates and grown in the dark for 6 d after imbibition. Lipid content was analyzed by UPLC-MS (n = 4 to 6). Detailed results of the assay for the depicted lines as well as additional lines are presented in Supplemental Data Set 1. (A) PCA of lipid levels. (B) Heat map of log 2 values of TAG relative levels in comparison to the wild type. Significant differences following Student's t test are denoted by one asterisk (P < 0.05) or two asterisks (P < 0.01). (C) Heat map of log 2 values of FA relative levels in comparison to the wild type. Significant differences following Student's t test are denoted by one asterisk (P < 0.05) or two asterisks (P < 0.01). (D) Heat map of log 2 values of lysophosphatidylcholine (LysoPC) and LysoPE relative levels in comparison to the wild type. Significant differences following Student's t test are denoted by one asterisk (P < 0.05) or two asterisks (P < 0.01). 2011), we paid particular attention to the respiration of atg5-1 compared with the wild type by evaluating the TCA cycle flux as a consequence of CO 2 evolution ( Figure 5). This experiment was further prompted by the decrease in free amino acids, especially in BCAA and lysine ( Figure 2C). These amino acids were shown to support respiration during carbon starvation both by supplying electrons directly to the mitochondrial electron chain or by feeding breakdown intermediates into the TCA cycle (Araújo et al., 2011). Our results suggest that greater carbon flux is being funneled into the TCA cycle in atg5-1 than wild-type etiolated seedlings. This is at first glance counterintuitive given that the plants are carbon starved and, at least to some extent, protein degradation is inhibited in the atg mutant. Thus, to confirm that this is indeed the case, we measured CO 2 evolution following feeding starved seedlings with positionally labeled glucose. Our conclusion is further strengthened by our proteomic analysis and by the [U-13 C]Lys feeding experiment (Tables 1 and 2, respectively). The proteomic analysis did not detect triosephosphate isomerase, a glycolysis-related protein (Giegé et al., 2003), in atg mutant seedlings, but was able to detect the protein in wild-type plants, suggesting a potential shift between glycolysis and the TCA cycle in atg mutants. Additionally, following supply of labeled lysine, increased redistribution of label to malate and glutamate was observed in atg5-1 compared with the wild type. Malate is a TCA cycle intermediate (Araújo et al., 2011), and glutamate is produced during lysine breakdown and utilization in the electron transport chain (Araújo et al., 2010). Taken together, these results suggest increased flux toward respiration in the mutants. Although we ascertained the concentration of fed glucose was very low and kept the feeding time to a minimum in order to avoid recovery from starvation, it is possible that since atg5-1 was experiencing enhanced starvation compared with the wild type, it was utilizing the supplied glucose faster in the TCA cycle, resulting in higher CO 2 emission. Nevertheless, the combined results indicate a differential energy homeostasis in atg5-1 compared with the wild type and implicate the upregulation of amino acid degradation as being a likely mechanism underlying this. Supporting this claim are the observations of greater label uptake and label metabolism in atg5-1 seedlings during the [U-14 C]Glc fractionation experiment and decreased flux to net protein synthesis in the mutants compared with the wild type (Table 3).
Comparison of lipid contents in atg mutants and the wild type both provides important insights and raises several interesting questions. The main storage compounds in Arabidopsis seeds are TAGs; these are subsequently converted to FAs, which are later degraded by b-oxidation to supply energy for the germinating seed and during seedling establishment (Graham, 2008). This process takes place in the peroxisome (Hu et al., 2012); thus, it is tempting to attribute the accumulation of TAGs and FAs in atg5-1 and atg5-3 to an impairment in b-oxidation ( Figures 6B and 6C). This is further compounded by the observation that atg5-1 mutants demonstrated impairment in seedling establishment, interpreted as a decline in peroxisomal activity . However, this accumulation was not observed for the atg7-2 and atg4a4b lines ( Figure 6C; Supplemental Data Set 1) and a pronounced difference in TAG accumulation was observed between the wild-type samples of the two different ecotypes (Col and Ws; Supplemental Data Set 1), suggesting peroxisomal deficiency may not be a sufficient explanation for the phenotype. Moreover, our proteomic analysis (Table 1) revealed an accumulation of two b-oxidation enzymes in atg mutants, MLS and MFP2 (Rylott et al., 2006;Kim et al., 2013), rendering the assumption of impaired b-oxidation less plausible. Further strengthening this claim is a study specifically examining the role of autophagy in peroxisome maintenance during seedling growth. The authors, in addition to observing MLS accumulation in atg mutants, directly examined peroxisome function in seedlings and did not find a difference compared with the wild type (Kim et al., 2013). As not all atg mutants possess the exact same phenotype (Harrison-Lowe and Olsen, 2008;Yoshimoto et al., 2009), we postulate that the lipid phenotype we present may be an ATG5specific phenotype. By contrast, increased lipid editing was a consistent phenotype throughout the atg mutants ( Figure 6D). An increased lysolipid level has been shown to be the result of the stress response (Welti et al., 2002;Gao et al., 2010). Taken together with the observed energy deprivation phenotype, we atg mutants as well as their respective wild-type and SA controls (NahG and atg5.NahG) were sown on 0.53 MS plates and grown in the dark for 6 d after imbibition. Lipid content was analyzed by UPLC-MS (n = 4 to 6). Detailed results of the assay for the depicted lines as well as additional lines are presented in Supplemental Data Set 1. (A) Heat map of log 2 values of PI relative levels in comparison to the wild type. Significant differences following Student's t test are denoted by one asterisk (P < 0.05) or two asterisks (P < 0.01). (B) Heat map of log 2 values of PE relative levels in comparison to the wild type. Significant differences following Student's t test are denoted by one asterisk (P < 0.05) or two asterisks (P < 0.01).
can conclude an increased stress response in the atg mutants compared with the wild type. Another notable phenotype is the different levels of autophagy-related lipids observed in our mutants. Phosphatidylinositol 3-phosphate serves as a scaffold for the assembly of the autophagic apparatus (Singh and Cuervo, 2011); thus, an increase in PI levels ( Figure 7A) might suggest an attempt to compensate for the lack of autophagy. PE, on the other hand, is conjugated to ATG8 to facilitate its incorporation into the growing autophagosome (Mizushima et al., 2011). A decrease in several PE species ( Figure 7B) may indicate the lack of ability to generate autophagosomes. These data thus suggest possible feedback mechanisms operating between the autophagy mechanism and the lipids regulating it. Dissecting these mechanisms will be an important area for future research into the interplay between autophagic process and cellular metabolism.
The use of stay-green plants has proven important throughout our analysis as a control differentiating between the direct impact of autophagy plant metabolism and its indirect effects caused by SA accumulation (Yoshimoto et al., 2009). This control has been previously used in the study of older plants (Guiboileau et al., 2012); however, we suggest this might also be important when studying seed and seedling phenotypes, as seed content and, thus, germination ability could be affected by the mother plant (Andriotis et al., 2012). The use of SA control lines was of significant importance when analyzing secondary metabolites in the atg mutants (Figure 3), as some of the metabolites changing in atg mutants seem to be SA dependent. A recently published study examining the metabolic response of Arabidopsis atg mutants under nitrogen starvation (Masclaux-Daubresse et al., 2014) used these stay-green lines to control for primary metabolite changes, and the authors indeed noticed SA effects on primary metabolism. Unfortunately, this control was not applied for secondary metabolites, which we discovered to be very important, thus necessitating a second evaluation of the secondary metabolism results.
In conclusion, energy deprivation is responsible for the delayed growth phenotype of atg etiolated seedlings grown under carbon starvation, highlighting the importance of autophagy in the cellular energy economy. As mentioned above, hypersensitivity to carbon and starvation is a hallmark of atg mutants (Bassham, 2009), and the products of protein degradation have been shown to assist in energy supply during carbon starvation (Araújo et al., 2010). In light of this, and taking into account the results obtained in previous studies (Guiboileau et al., 2013;Izumi et al., 2013;Michaeli et al., 2014), it seems safe to assume that the autophagy mechanism contributes to global metabolic changes not only during seedling establishment but also during other stages of plant life. However, when our results are compared with those of the study of Guiboileau et al. (2013), as well as other recent works (Izumi et al., 2013;Masclaux-Daubresse et al., 2014), it becomes evident that the role of the autophagy mechanism alters as the plant ages and environmental conditions change, thus rendering the underlying mechanism(s) more difficult to elucidate. In addition, comprehensive analyses are clearly required in order to understand the many facets of this complex metabolic phenotype. Given the difficulties in interpreting changes in steady state metabolite levels in such a complex system, we demonstrate that augmenting such approaches with label redistribution studies is able to greatly enhance our understanding of resource mobilization and energy regulation, and we thus strongly recommend its adoption for future studies of autophagy. In applying such an approach here, we were able to clearly demonstrate the importance of autophagy in both nutrient mobilization and energy homeostasis of growing seedlings. However, future studies are required to elucidate the exact role of autophagy both in other plant tissues and other environmental circumstances.

Plant Material and Growth Conditions
Arabidopsis thaliana ecotype Col was used in this study, apart from atg4a4b, which is of the Ws ecotype and was therefore compared with the appropriate wild type control. The lines used in this study are as follows: ag4a4b (Yoshimoto et al., 2004), atg5-1 (SAIL_129B079) (Yoshimoto et al., 2009), atg5-3 (SALK_020601) (Guiboileau et al., 2013), atg7-2 (GK-655B06) (Hofius et al., 2009), GFP-ATG8f-HA , NahG overexpression (NahG) (Yoshimoto et al., 2009), and atg5.NagG (Yoshimoto et al., 2009). Seeds were surface sterilized with bleach and sown on 0.5 MS agar plates without additional sucrose (unless specified otherwise, in which case 1% sucrose was added). The plates were covered with foil and kept at 4°C for a period of 48 h and then transferred, while still covered, to a growth chamber at 22°C. Samples were collected after 6 to 7 d.

Hypocotyl Length Measurement
Arabidopsis seedlings were grown as described above. Seven-day-old etiolated seedlings were transferred to a plastic tray and a picture of the seedlings was taken (15 seedlings per line). Image analysis was performed by ImageJ (Schneider et al., 2012). The experiment was repeated twice.

Seed Germination Assay
Arabidopsis seeds were sown as described above (six plates per line, ;100 seeds per plate). After imbibition, the foil cover was removed from the plates and they were put in a growth chamber with continuous light and 22°C. The number of germinated seeds was counted every day for 4 d and % germination was calculated. Germination was scored using two parameters: radical protrusion and the appearance of two green cotyledons.

Extraction, Derivatization, and Analysis of Arabidopsis Seedling Primary Metabolites Using GC-MS
Metabolite extraction for GC-MS was performed by a method modified from that described by Roessner-Tunali et al. (2003). Six-day-old etiolated Arabidopsis seedlings (;50 mg) were collected and immediately frozen in liquid nitrogen prior to storage at 280°C. The samples were then homogenized using a ball mill precooled with liquid nitrogen and extracted in 1400 mL of methanol, and 60 mL of internal standard (0.2 mg ribitol mL 21 water) was subsequently added as a quantification standard. The extraction, derivatization, standard addition, and sample injection were exactly as described previously (Lisec et al., 2006). The GC-MS system comprised a CTC CombiPAL autosampler, an Agilent 6890N gas chromatograph, and a LECO Pegasus III TOF-MS running in EI+ mode. Metabolites were identified in comparison to database entries of authentic standards (Kopka et al., 2005). Chromatograms and mass spectra were evaluated using Chroma TOF 1.0 (Leco) and TagFinder 4.0 software (Luedemann et al., 2008). Data presentation and experimental details were provided as supplemental data in a manner consistent with recent metabolite reporting recommendations (Araújo et al., 2011) (Supplemental Data Set 2).

Extraction, Derivatization, and Analysis of Arabidopsis Seedling Secondary Metabolites Using LC-MS
Six-day-old etiolated Arabidopsis seedlings (;50 mg) collected and immediately frozen in liquid nitrogen prior to storage at 280°C. Profiling of secondary metabolite by liquid chromatography-electrospray ionization/ mass spectrometry was performed in negative and positive ion detection mode as described (Araújo et al., 2010). All data were processed using Xcalibur 2.1 software (Thermo Fisher Scientific). The resulting data matrix was normalized using an internal standard (Isovitexin; CAS 29702-25-8) in extraction buffer (5 mg/mL). Metabolites were identified and annotated based on prior knowledge (Tohge et al., 2007;Watanabe et al., 2013) and coelution profile of Arabidopsis leaf extracts characterized with purified compounds from plant extracts (Nakabayashi et al., 2009). Data are reported in a manner compliant with the standards suggested by Fernie et al. (2011) (Supplemental Data Set 2).

Protein Extraction, Quantification, and Identification by Mass Spectrometry
Six-day-old etiolated Arabidopsis seedlings were collected and immediately frozen in liquid nitrogen prior to storage at 280°C. For immunoblotting, 20 mg of plant material was ground in liquid nitrogen and total proteins were extracted in 100 mL protein extraction buffer (Gruis et al., 2002). Samples were immediately incubated for 5 min at 100°C after addition of the buffer, vortexed, and centrifuged for 10 min at 20,800g. Supernatants were collected and 100 mL of SDS-PAGE samples buffer was added (Laemmli, 1970). Proteins were separated on a 10% SDS-PAGE and electrotransferred onto polyvinylidene difluoride membrane. Equal loading was validated using Coomassie Brilliant Blue staining of the membrane. For immunodetection, mouse anti-GFP antibody (Clontech; 1:20,000 dilution) was used in combination with horseradish peroxidase-conjugated rabbit anti mouse antibody (Sigma; A9044; 1:10,000 dilution).
For protein quantification and analysis, ;50 mg plant material was ground in liquid nitrogen and suspended in protein extraction buffer (50 mM Tris, pH 7.5, 20 mM NaCl, 10% [v/v] glycerol, 0.1% SDS, and Complete EDTA free protease inhibitor cocktail [Roche]) and mixed well. Samples were centrifuged for 15 min at 16,000g, 4°C, and the supernatant was collected. Protein concentration was determined using a BCA Protein Assay Kit (Thermo). Equal protein amounts were separated on a 10% SDS-PAGE, followed by protein staining using a colloidal Coomassie Brilliant Blue dye. Bands of interest were excised from the gel and subjected to tryptic digest. The resulting peptides were resuspended in 5% acetonitrile and 0.1% trifluoroacetic acid, desalted using a C18 Ziptip (Millipore), and eluted in 50% acetonitrile and 0.1% trifluoroacetic acid. Peptides were analyzed via liquid chromatography coupled to tandem mass spectrometry (Proxeon EASY nLC coupled to a Thermo LTQ-XL Hybrid Ion Trap-Orbitrap mass spectrometer). For protein identification, the Mascot search engine (Matrix Sciences) was used to match the data generated from MS/MS spectra against the AGI proteins database (www. arabidopsis.org). Searches were done allowing one missed cleavage, methionine oxidation as a variable modification, peptide tolerance was set to 10 ppm, MS/MS tolerance to 0.8 D, and peptide charge was either +2 or +3. Standard scoring was used.

Measurement of Respiratory Parameters
Arabidopsis seeds were surface sterilized and grown in 10 mL liquid medium in 100-mL conical flasks (20 mg seeds per nine flasks). Following imbibition, the liquid cultures were grown in the dark at 20°C for 7 d. Estimations of the TCA cycle flux on the basis of 14 CO 2 evolution were performed following incubation of etiolated seedlings in 5 mL 10 mM MES-KOH, pH 6.5, containing 1.85 MBq/mmol of [1-14 C]-, [3:4-14 C]-, or [6-14 C]Glc (Hartmann Analytic). Each flask was then sealed with Parafilm and placed on an orbital shaker at 100 rpm. Released 14 CO 2 was collected in 0.5 mL of 10% (w/v) KOH in a vial suspended in the flask (Harrison and Kruger, 2008). Incubation was performed under green light. 14 CO 2 evolved was then quantified by liquid scintillation counting. The results were interpreted following Rees and Beevers (1960).

Analysis of [U-13 C]-Lys-Labeled Samples
Arabidopsis seeds were surface sterilized and grown in 10 mL liquid medium in 100-mL conical flasks (20 mg seeds per nine flasks). Following imbibition, the liquid cultures were grown in the dark at 20°C for 7 d. Seedlings from four flasks were transferred under green light into one flask containing 5 mL 10 mM MES-KOH, pH 6.5, and incubated for 1 h while shaking. [U-13 C]-Lys (Cambridge Isotope Laboratories) was added to a final concentration of 5 mM, and samples were collected every 40 min for 2 h. The seedlings from each flask were divided to four groups and used as biological replicates. Each sample was placed in a sieve, washed several times with double-distilled water, and immediately frozen in liquid nitrogen prior to storage at 280°C. Samples were subsequently extracted in 100% methanol, the fractional enrichment of metabolite pools was determined exactly as described previously (Roessner-Tunali et al., 2004;Tieman et al., 2006), and label redistribution was expressed as per (Studart-Guimarães et al., 2007).

Incubation of Arabidopsis Seedlings with [U-14 C]Glc
Arabidopsis seeds were surface sterilized and grown in 10 mL liquid medium in 100-mL conical flasks (20 mg seeds per nine flasks). Following imbibition, the liquid cultures were grown in the dark at 20°C for 7 d. Etiolated seedlings were incubated under green light in 5 mL 10 mM MES-KOH, pH 6.5, containing 1.85 MBq/mmol [U-14 C]Glc (Hartmann Analytic) to a final concentration of 1 mM. Samples were then incubated for 80 min, placed in a sieve, washed several times in double-distilled water, frozen in liquid nitrogen, and stored at 280°C until further analysis. All incubations were performed in sealed flasks under green light and shaken at 100 rpm. The evolved 14 CO 2 was collected in 0.5 mL of 10% (w/v) KOH.

Fractionation of 14 C-Labeled Tissue Extracts
Extraction and fractionation were performed as previously described (Szecowka et al., 2012). Frozen seedling samples were extracted with 80% (v/v) ethanol at 80°C (1 mL per sample) and reextracted in two subsequent steps with 50% (v/v) ethanol (1 mL per sample for each step), and the combined supernatants were dried under an air stream at 35°C and resuspended in 1 mL of water (Fernie et al., 2001). The soluble fraction was subsequently separated into neutral, anionic, and basic fractions by ion-exchange chromatography; the neutral fraction (2.5 mL) was freeze-dried, resuspended in 100 mL water, and further analyzed by enzymatic digestion followed by a second ion-exchange chromatography step (Carrari et al., 2006). To measure phosphate esters, samples (250 mL) of the soluble fraction were incubated in 50 mL of 10 mm MES-KOH, pH 6.0, with or without 1 unit of potato acid phosphatase (grade II; Boehringer Mannheim) for 3 h at 37°C, boiled for 2 min, and analyzed by ion-exchange chromatography (Fernie et al., 2001) The insoluble material left after ethanol extraction was homogenized, resuspended in 1 mL of water, and counted for starch (Fernie et al., 2001). Fluxes were calculated as described following the assumptions detailed by Geigenberger et al. (1997Geigenberger et al. ( , 2000.

Extraction, Derivatization, and Analysis of Arabidopsis Seedling Lipids Using UPLC-MS
Six-day-old etiolated Arabidopsis seedlings (;20 mg) were collected and immediately frozen in liquid nitrogen prior to storage at 280°C. Lipid analysis was performed as described previously Hummel et al., 2011). The dried pellets were resuspended in 250 mL of a mixture of acetonitrile/isopropanol (7:3 [v/v]), thoroughly vortexed, and centrifuged. Then, 100-mL aliquots were transferred into glass vials and 2 mL was injected and separated on an Acquity UPLC system (Waters) using a reversed-phase C8 column. The ultraperformance liquid chromatography solvents were as follows: A = water with 1% of a solution of 1 M ammonium acetate and 0.1% acetic acid; B = 70% acetonitrile/30% isopropanol with 1% of a solution of 1 M ammonium acetate and 0.1% acetic acid. The gradient separation was performed at a flow rate of 400 mL/min as follows: 1 min at 45% A, a 3-min linear gradient from 45% A to 35% A, an 8-min linear gradient from 25% A to 11% A, and a 3-min linear gradient from 11% A to 1% A. After washing the column for 3 min with 1% A, the buffer was returned to 45% A and the column was reequilibrated for 4 min (22 min total run time). The samples were measured in either positive or negative ion mode. The mass spectra were acquired using an Exactive high-resolution mass spectrometer (Thermo Fisher). The spectra were recorded using alternating full-scan and allion fragmentation scan mode, covering a mass range from 100 to 1500 m/z. The resolution was set to 10,000, with 10 scans per second, restricting the Orbitrap loading time to a maximum of 100 ms with a target value of 1 million ions. The capillary voltage was set to 3 kV with a sheath gas flow of 60 and an auxiliary gas flow of 35 (values are arbitrary units). The capillary temperature was set to 150°C, and the drying gas in the heated electrospray source was set to 350°C. The skimmer voltage was held at 25 V, and the tube lens was set to a value of 130 V. The spectra were recorded from min 1 to 20 of the ultraperformance liquid chromatography gradients. Obtained raw chromatograms were further processed using Excalibur software version 2.10 (Thermo Fisher). Peaks from raw chromatograms were first determined and aligned by their parent masses, chemical noise was subtracted, and a final alignment file of all chromatograms, which contains information about m/z ratio, retention times, and retention time deviations for each annotated peak, was created. Assignment of peaks, normalization, and quantification were performed by Progenesis QI (version 1.0; Nonlinear Dynamics).

Statistical Analysis
Statistical differences between groups were analyzed by Student's t tests on the raw data. Results were determined to be statistically different at a probability level of P < 0.05. Relative metabolite levels were obtained as the ratio between the lines and the mean value of the respective wild type. The lipid data, processed by Progenesis QI, were exported as a matrix to Umetrics SIMCA version 13.0. The data were scaled and mean-centered according to the vendor. PCA was performed using the Multibase Excel add-in (Numerical Dynamics).