Systemic analysis of inducible target of rapamycin mutants reveal a general metabolic switch controlling growth in Arabidopsis thaliana

Authors


(e-mails camila.caldana@bioetanol.org.br; giavalisco@mpimp-golm.mpg.de).

Summary

The target of rapamycin (TOR) pathway is a major regulator of growth in all eukaryotes, integrating energy, nutrient and stress signals into molecular decisions. By using large-scale MS-based metabolite profiling of primary, secondary and lipid compounds in combination with array-based transcript profiling, we show that the TOR protein not only regulates growth but also influences nutrient partitioning and central energy metabolism. The study was performed on plants exhibiting conditional down-regulation of AtTOR expression, revealing strong regulation of genes involved in pathways such as the cell cycle, cell-wall modifications and senescence, together with major changes in transcripts and metabolites of the primary and secondary metabolism. In agreement with these results, our morphological and metabolic analyses disclosed major metabolic changes leading to massive accumulations of storage lipids and starch. The implications of these data in the context of the general role of TOR in eukaryotic systems are discussed in parallel with the plant-specific aspects of TOR function. Finally, we propose a role for harnessing the plant TOR pathway by utilizing it as a potent metabolic switch, offering a possible route for biotechnological optimization of plant energy content and carbon partitioning for the production of bioenergy.

Introduction

Cell division, assimilation of carbon and coordinated cell growth, including the regulation of anabolic and catabolic processes, are essential steps in the development of every living organism (Leevers and McNeill, 2005; Goranov and Amon, 2010; Johnson and Lenhard, 2011). Their coordination relies on a complex signaling network that integrates external and internal inputs, such as nutrient availability or energy status, into molecular growth decisions (De Virgilio and Loewith, 2006; Goranov and Amon, 2010). The control of these signaling networks is coordinated by a number of conserved and partially organism-specific regulators. One of the best-described global regulators of growth in all eukaryotic systems is the target of rapamycin (TOR) kinase (Ma and Blenis, 2009; Soulard et al., 2009; Zoncu et al., 2011). The target of rapamycin not only integrates several external and internal signals but also controls a number of essential anabolic and catabolic processes, including transcription, translation, autophagy and central energy and carbon metabolism (Wullschleger et al., 2006; Proud, 2009, 2010; Soulard et al., 2009; Jung et al., 2010).

The target of rapamycin, which is ubiquitous in eukaryotic organisms, belongs to the class of phosphoinositol 3-kinase-related kinases (Wullschleger et al., 2006). These kinases have structural similarity to lipid kinase, but act as a protein (serine/threonine) kinase (Menand et al., 2002; Mahfouz et al., 2006; Soulard et al., 2009; Ahn et al., 2011; Ren et al., 2011). Thus the TOR regulatory network is coordinated by protein phosphorylation events, which lead to functional reorganization of the cell (Huber et al., 2009; Soulard et al., 2010; Hsu et al., 2011; Sun et al., 2011). The target of rapamycin function is executed by two distinct protein complexes (termed TORC1 and TORC2), which were initially detected in yeast and mammalian cells, but have since been found in several other organisms (Wullschleger et al., 2006; Soulard et al., 2009; Wang and Proud, 2009). The well-studied TORC1, which consists of TOR, RAPTOR (regulatory-associated protein of mTOR) and LST8 (lethal with SEC13 protein 8) is chemically inhibited by the anti-proliferative drug rapamycin (Benjamin et al., 2011), is mainly regulated by nutrient (Wang and Proud, 2009; Li and Guan, 2010) and energy availability (Gwinn et al., 2008), and targets translation (Huo et al., 2011) and other processes necessary for growth. The function of the rapamycin-insensitive complex TORC2, which consists of TOR, RICTOR (rapamycin-insensitive companion of mTOR), LST8 and SIN (stress-activated map kinase-interacting protein 1), is less well described, but it appears to be involved in the regulation of cytoskeleton organization and possibly the regulation of TORC1 function (Wullschleger et al., 2006; Zinzalla et al., 2011). Even though all of the essential TORC1 genes are present in the genome of Arabidopsis (Menand et al., 2002; Anderson and Hanson, 2005; Anderson et al., 2005; Deprost et al., 2005; Mahfouz et al., 2006; Moreau et al., 2010, 2012; Ahn et al., 2011; Ren et al., 2011), rapamycin does not seem to lead to growth inhibition in plants (Menand et al., 2002; Turck et al., 2004; Mahfouz et al., 2006; Deprost et al., 2007; Sormani et al., 2007). However, recently, application of 10 μm of rapamycin to a liquid seedling culture system revealed that this drug is active in plants – albeit at a concentration more than 100 times that used in studies of yeast or mammalian systems (Xiong and Sheen, 2012). The combination of the embryo lethality of TOR knockout plants (Menand et al., 2002; Deprost et al., 2007; Ren et al., 2011) and the lack of rapamycin sensitivity of physiologically growing plants is quite unfortunate as it has to date limited the possibility for functional genomic studies of this essential pathway in plants (Moreau et al., 2010; John et al., 2011).

An alternative to the use of rapamycin or constitutive knockdown lines is the use of inducible knockdown lines leading to conditional inhibition of TOR gene expression (Deprost et al., 2007; Liu and Bassham, 2010; Ahn et al., 2011; Ren et al., 2011). Such an approach not only introduces the possibility of repressing TOR at a specific developmental stage, but also provides the means to analyze the repressive effects of TOR in a time-resolved manner. Accordingly we developed several independent, artificial microRNA (amiR) lines of TOR. These lines, which showed strong growth-related phenotypes if the TOR gene was repressed early in development, were subjected to a thorough, large-scale transcriptional and metabolic profiling, including metabolic profiles of primary, secondary and lipid metabolites. This range of metabolic information in combination with whole-genome transcript profiling allows integration and comparison of several related and unrelated molecular responses on a global scale. The whole experiment was performed using material harvested at two early time points (3 and 6 days) after induction of TOR repression by the amiR construct. The results obtained by this systemic profiling approach demonstrate that massive changes of the central energy and carbon metabolism are accompanied by a strong alteration in the more distant lipid and secondary metabolism. Many of these metabolic changes are in agreement with transcriptional changes, but several observations were not anticipated by the transcript data. Taken together, these data demonstrate that TOR functions as an essential sensor and regulator of growth and development in plants.

Results

Conditional down-regulation of the TOR transcript results in suppressed plant growth

Given the embryo lethality of TOR knockout plants (Menand et al., 2002; Mahfouz et al., 2006; Deprost et al., 2007; Ren et al., 2011) and the insensitivity of physiologically grown Arabidopsis plants to rapamycin (Menand et al., 2002; Mahfouz et al., 2006; Sormani et al., 2007), we generated inducible artificial microRNA (amiR) lines (Ossowski et al., 2008) for conditional down-regulation of the AtTOR gene. The amiR system is not only highly efficient but also provides high specific down-regulation of the targeted genes at the transcriptional and translational level (Ossowski et al., 2008). The unique amiR-tor target site within the single-copy AtTOR gene was obtained using an online microRNA designer tool (Ossowski et al., 2008), and targets a central region of the AtTOR gene between 3503 and 3523 bp of the mature transcript (Figure 1a).

Figure 1.

Inducible amiR-tor lines. (a) Target position of amiR in the cDNA sequence of the AtTOR gene. (b) Twenty-day-old amiR-tor seedlings. Seedlings were germinated and grown for the first 6 days in half-strength MS medium supplemented with 1% sucrose, before transferring the seedlings for an additional 14 days to half-strength MS medium supplemented with 1% sucrose and either ethanol (−) or 20 μm EST (+). (c) AtTOR expression in amiR-tor and empty vector (EV) plants. Seedlings were grown for the first 14 days in half-strength MS medium supplemented with 1% sucrose, before transferring the seedlings for an additional 6 days to half-strength MS medium supplemented with 1% sucrose and either ethanol (mock, −) or 20 μm EST (induced, +). Samples were taken 3 days after transferring them to the induction plates. Relative expression levels of TOR were calculated as follow: first, the ΔCT was obtained by subtracting CT of UBQ10 (reference gene) from the Ct of TOR (gene of interest). A second calculation was performed to evaluate the effect of EST treatment on the repression of TOR levels, using the equation inline image. Data are means of three biological replicates ± SD. Significant differences between the EST- and mock-treated plants, using a paired t test, are indicated by asterisks (< 0.05).

Several independent estradiol (EST)-inducible amiR-tor lines were generated, of which three independent transformants (amiR-tor9, amiR-tor17 and amiR-tor20), differing in the extent of their transcript repression and their growth inhibition phenotype, were selected for further analysis (Figure 1b,c and Figures S1–S3). Initially we screened for the growth inhibitory effect of TOR repression in plants grown on half-strength MS agar plates supplemented with 1% sucrose (Figure 1b). These plants were germinated and grown for 6 days regularly, before transferring them either to estradiol-containing amiR-tor inducing, or ethanol-containing control plates. The plants were grown for up to 10 additional days under induction or control conditions, and analyzed for phenotypic alterations. As shown in Figures S1 and S2, EST-induced repression of AtTOR leads, as shown previously (Deprost et al., 2007), to severe growth arrest, which is clearly visible from day 3 after induction of TOR repression. To quantify these growth differences, we validated several parameters such as the leaf size, mesophyll cell size, root length and fresh weight of these plants (Figures S1 and S2). The plants are not only producing less biomass, but they also have significantly smaller leaves and roots, which are explained by significantly smaller cells (Figures S1 and S2). In addition to the severe growth arrest, we also observed a strong induction of chlorosis and cell death in amiR-tor lines 9 and 17 at day 10 of TOR repression, but line 20 appears to be unaffected (Figure S1). Taken together, we reproduced most of the previously described phenotypes of TOR repression plants, indicating that our lines are fully comparable with previous described inducible RNAi TOR plants (Deprost et al., 2007). However, the severity of the observed morphological phenotypes (Figures S1–S3) and the amplitude of the molecular phenotypes (Figures 3–5) are not always directly correlated with the degree of TOR transcript repression (Figure 1c), which may indicate that the amount of active TOR protein may also be affected by post-transcriptional regulation (Ossowski et al., 2008).

Transcript profiling of amiR-tor lines

Due to the strong phenotype after TOR down-regulation by the artificial microRNAs and the low amount of plant material obtained from these plants (Figure S1), we decided to change the growth conditions used for morphological phenotyping of the amiR-tor lines when performing molecular phenotyping (transcript and metabolic) of these plants. We noticed that growing the plants to a larger size before inhibiting TOR expression leads to a less severe phenotype. Therefore, we decided to grow the plants for 14 instead of 6 days on medium without EST, before transferring them to induction medium (EST/control). As shown in Figure S3, except for line 9, no significant growth phenotype was apparent for these lines up to day 6 after TOR repression. Leaving the plants for longer (up to 20 days) on the TOR repression-inducing plates leads to a significant reduction in growth and biomass (Figure 1b). Interestingly, although severe chlorosis or cell-death phenotypes were not observed for the plants pre-cultured for 14 days, a strong, up to fourfold, repression of expression of the AtTOR gene was measured by quantitative RT-PCR (Figure 1c).

To assess the transcriptional changes triggered by the silencing of AtTOR, the three amiR-tor lines (lines 9, 17 and 20) and one empty vector line were subjected to whole-genome transcript profiling using Affymetrix ATH1 arrays. Samples were collected from 14-day pre-cultured plants at two time points (3 and 6 days after EST or mock treatment). The relative expression level was calculated by comparing the expression abundance of each amiR-tor line under EST- or non-induced conditions. To exclude the EST effect and identify only transcriptional changes caused by repression of AtTOR, we furthermore corrected the fold change ratios by subtracting the log2 fold changes obtained for EST- and non-induced empty vector controls (see 'Experimental Procedures').

Genes were considered up- or down-regulated when the absolute fold change was ≥ 2 in at least two lines (Table S1). Comparison of the transcriptional responses of the various lines showed that their transcriptional changes qualitatively overlapped and quantitatively reflected the differences in the severity of repression of the AtTOR gene. In our subsequent analysis, only transcripts that were significantly differentially regulated in at least two of the three independent lines (Table S1) were considered. Applying this criterion to the whole dataset led to detection of 270 (174 up- and 96 down-regulated) and 515 (359 up- and 156 down-regulated) differentially expressed genes after 3 and 6 days of AtTOR repression, respectively. Most of the transcriptional changes at 3 days after amiR-tor induction persisted in the 6-day samples, indicating that the initial repression pattern is maintained and further transcriptional changes are induced in a cascade-like structure.

Differentially expressed transcripts were categorized into functional categories using PageMan (Usadel et al., 2006), and the network representation was realised using the BiNGO tool within the framework of the Cytoscape package (Maere et al., 2005; Kohl et al., 2011). As shown in Figure 2, the functional over-representation analysis suggests that AtTOR is a key regulator of plant growth and development, as not only central functions such as the cell cycle and cell-wall modifications but also degradation processes such as senescence and autophagy show coordinated changes at the transcriptional level (Figure 2 and Figure S4). These changes were observed alongside severe alterations of transcripts coding for components of the primary metabolism, including nitrogen and carbon utilization (Figure 2 and Figure S4). Additionally, we found that expression of several genes responsible for the synthesis of the major secondary metabolites (glucosinolates and flavonoids) was changed over the course of AtTOR repression (Figure 2 and Figure S4). Finally, transcripts that changed in AtTOR inhibited lines showed significant over-representation of genes involved in chromatin structure, hormone metabolism, protein signaling, lipid transfer, transporters and stress responses (Figures S4 and S5, and Table S1).

Figure 2.

Gene ontology (GO) over-representation analysis of genes regulated in the amiR-tor lines. (a) Analysis of 359 up- and (b) 152 down-regulated genes 6 days after amiR-tor induction. The size of nodes is proportional to the number of genes annotated to that node (category). The color of the node represents the corrected P value: white nodes are not significantly over-represented; colored nodes indicate significantly over-represented genes, ranging from yellow to dark orange, representing a P values from 0.05 to 5E-7.

Primary metabolite profiling of amiR-tor lines

To confirm the molecular influence of TOR repression on the central metabolism, we performed comprehensive metabolite profiling using GC-TOF MS, which allowed for the quantitative analysis of 90 compounds of primary metabolism (Caldana et al., 2011) (Table S2).

The relative changes in the individual metabolites after EST treatment in the various lines are shown in Table S2, together with the results of statistical testing of differences between EST-treated and control plants. A change in the metabolite level was determined and validated in the same manner as described for the transcripts. As most of the significant changes in the transcript data were observed at the later time point after EST induction, we focused our metabolic analysis on samples derived from plants 6 days after EST application.

Down-regulation of the AtTOR gene leads to global changes in the primary metabolism (Figure 3 and Table S2). A significant increase in the levels of branched chain (Leu, Ile and Val), aromatic (Tyr and Trp) and other amino acids such as Lys, β-Ala, His, Pro, Thr and gamma amino butyric acid (GABA) was observed in the induced amiR-tor lines (Figure 3 and Table S2). Next to these changes almost all the tricarboxylic acid (TCA) cycle intermediates, including citrate, α-ketoglutarate, succinate, fumarate and malate, were significantly increased in the induced amiR-tor seedlings (Figure 3, and Tables S2 and S3). In addition to the strong up-regulation of amino and organic acids, the catabolism of Arg and other polyamines appears to be significantly affected. Accordingly levels of Arg, ornithine and spermidine were strongly reduced (Figure 3 and Table S2) in the amiR-tor lines, strongly suggesting altered nitrogen metabolism/usage.

Figure 3.

Disruption of AtTOR leads to severe and significant metabolic changes. The ratio of metabolite abundance is represented in log2 scale on metabolic pathways. Each box represents the median of five biological replicates of one of the three independent amiR-tor lines (left is line 9, middle is line 17 and right is line 20). Blue indicates induction and red indicates repression. The shading intensity represents the strength of alteration. Significant differences between the EST- and mock-treated plants, using a paired t test, are indicated by asterisks within the boxes (< 0.05).

Analysis of storage carbohydrates and comprehensive lipid profiling

Starch is the initial form in which plants store reduced carbohydrates (Smith and Stitt, 2007). Recently, it has been found that the abundance of this metabolite negatively correlates with growth (Sulpice et al., 2009). Consequently, we analyzed the starch content in our inducible EST lines. As expected and described previously for knockout of another TORC1 component (LST8) (Moreau et al., 2012), we observed a strong increase in starch content for all three amiR-tor lines upon EST induction (Figure 4a), indicating that the assumption that starch accumulation negatively correlates with growth also holds true for our plant lines.

Figure 4.

Storage carbohydrate and storage lipid accumulation in EST-induced amiR-tor lines. (a) Alteration of starch content after EST induction for 6 days in empty vector (EV) and amiR-tor seedlings. Significant differences between the EST- (E) and mock-treated (C) plants, using a paired t test, are indicated by asterisks (*< 0.05; **< 0.01). (b) Changes in triacylglyceride abundance after EST induction for 6 days in empty vector (EV) and amiR-tor seedlings. Blue indicates induction and yellow indicates repression. The shading intensity represents the strength of alteration. The scale is saturated at an amplitude [log2 (maximum/minimum)] of 1.

After starch, triacylglycerides represent a major storage form for highly reduced carbon and energy (Graham, 2008). Using a recently established lipid-profiling platform (Giavalisco et al., 2011; Hummel et al., 2011), more than 140 lipid species, covering the 12 major lipid classes (Table S3), were quantitatively analyzed. The most dramatic effect was accumulation of various species of TAGs (Figure 4b and Table S3). Interestingly, this accumulation predominantly affected TAGs containing long-chain polyunsaturated fatty acids, a phenotype that was recently also observed in cold-treated plants (Degenkolbe et al., 2012). Quite unexpectedly, and in contrast to these severe alterations in the total storage lipid composition, only a few minor changes were detectable for the other lipid classes (Table S3), indicating that the general composition of the membrane systems within the various compartments is not much altered.

Profiling of secondary metabolites

As the levels of several transcripts responsible for the regulation of secondary metabolism were severely altered, we decided to determine to what extent secondary metabolism is changed as a result of AtTOR inhibition. For this purpose, a non-targeted liquid chromatography/high resolution mass spectrometry-based analysis (Giavalisco et al., 2011) was performed. From the complex spectra obtained, the 46 most abundant secondary metabolites, associated with the classes of phenylpropanoids and glucosinolates, were analyzed and quantified in more detail (Table S4).

Six compounds of flavonoid biosynthesis were significantly increased in the repressed AtTOR plants, paralleling the observed up-regulation of flavonoid synthesis-related transcripts (Figure S4 and Table S1). In addition to up-regulation of products of sinapate ester biosynthesis, such as 1-O-sinapoyl-β-d-glucose and sinapoyl-(S)-malate (Milkowski and Strack, 2010) (Figure 5a), the level of 4-amino-4-deoxychorismate significantly increased in the amiR-tor lines (Figure 5c).

Figure 5.

Alteration of secondary metabolism pathways in silenced AtTOR plants. The ratio of metabolite abundance is represented in log2 scale on metabolic pathways. Blue indicates induction and red indicates repression. The shading intensity represents the strength of the alteration. Significant differences between the EST- and mock-treated plants, using a paired t test, are indicated by asterisks (< 0.05). (a) Pathway from the aromatic amino acids. (b) Pathway from methionine. (c) Pathway from shikimate.

In addition to changes in these abundant carbon-, hydrogen- and oxygen-containing metabolites linked to the phenylpropanoid pathways, significant changes within the pools of nitrogen- and sulfur-containing glucosinolates were detected. EST-treated amiR-tor plants tended to have higher levels of the highly abundant tryptophan-derived glucosinalate glucobrassicin (indol-3-ylmethylglucosinolate, Figure 5a) and of several methionine-derived aliphatic glucosinalates (Figure 5b), such as 3-methylsulfinylpropyl glucosinolate, 4-methylsulfinylbutyl glucosinolate, 5-methylsulfinylpentyl glucosinolate and 5-methylsulfinylpentyl glucosinolate.

Discussion

AtTOR repression affects central energy and carbon metabolism

We observed metabolic and transcriptional changes that confirm and extend previous descriptions of the effects of TOR inhibition in eukaryotes (Duvel et al., 2010; Moreau et al., 2012). In addition to the severe growth inhibition phenotype of the amiR-tor lines, a strong and coordinated regulation of several transcriptional and metabolic phenotypes was observed in this study. Many of these changes indicate that the various regulatory networks between growth and development are highly inter-connected at the gene expression and metabolite levels, and that TOR appears to be situated at their intersection. As expected, most of the biological processes down-regulated by AtTOR inhibition are associated with anabolic activities, and many of catabolic activities are globally up-regulated. The integration of the transcripts and metabolites does not always fully match, indicating transcription-independent mechanisms, but, in most cases the metabolic responses seem to follow the genetic regulatory networks.

AtTOR down-regulation leads to the accumulation of amino acids

Down-regulation of TOR has been suggested to mimic nitrogen (Matsuo et al., 2007) or carbon (Mayordomo et al., 2002) starvation-like responses, suggesting that the TOR signaling complex may act as a nutrient-sensitive switch. One of the most prominent changes observed at the metabolic level was the massive increase in the levels of several amino acids, including Leu, Ile, Val, Tyr, Trp, Lys, His, Thr, β-Ala and GABA (Figure 3 and Table S2). A similar phenotype, although not as severe, was observed recently in Arabidopsis T-DNA knockout plants of the TORC1 component LST8 (Moreau et al., 2012). The most likely explanations for accumulation of these amino acid pools are: (i) up-regulation of protein breakdown by nutrient recycling processes such as senescence or autophagy, (ii) a decrease of amino acid incorporation into newly synthesized proteins (decreased translation), (iii) increased de novo synthesis of amino acids, or (iv) combinations of all these possibilities.

As it has been shown previously that nutrient depletion triggers senescence and autophagy responses in plants (Gan and Amasino, 1997; Bassham, 2007), the possibility that the accumulating amino acids are derived from a starvation-like response is very plausible, particularly as it was shown recently that TOR repression in Arabidopsis as well as the green alga Chlamydomonas reinhardtii directly targets and activates autophagy (Liu and Bassham, 2010; Perez-Perez and Crespo, 2010). Thus we found a significant increase in genes related to autophagy and senescence (Figure 2 and Table S1) in the amiR-tor lines.

In addition to the autophagy-related explanation of amino acid accumulation, our data also support the assumption that the amino acids in our samples may be derived from down-regulation of genes coding for cytosolic (80S) and chloroplastic (70S) ribosomal proteins (Table S1). This repressed ribosomal protein synthesis, which, in combination with decreased synthesis of the 45S rRNA precursor, leads to decreased ribosome biogenesis and subsequently reduced translational efficiency and/or capacity. Both, repressed 45S rRNA synthesis (Ren et al., 2011) and decreased de novo protein synthesis (Ahn et al., 2011), have been shown to occur in Arabidopsis and to be dependent on TOR activity. In addition to these direct data, it was shown that TOR additionally blocks protein synthesis by decreasing the polysomal loading (Deprost et al., 2007; Ahn et al., 2011; Ren et al., 2011), which, over time, leads to a steady decrease in total soluble protein (Deprost et al., 2007).

Taken together, we conclude that there is evidence that both processes are operating, with the reduction in translation and the increase in autophagy very likely contributing to the observed accumulation of amino acids, while there is no evidence supporting the assumption that amino acids are accumulated due to increased de novo synthesis.

AtTOR affects the TCA cycle flux and the redox state of the cell

An increase in several TCA cycle intermediates, including citrate, ketoglutarate, succinate, fumarate and malate, was observed (Tables S2 and S3). The increase in TCA cycle intermediates may be partially explained as a consequence of decreased energy-consuming anabolic activities (e.g. translation, cell-wall synthesis) and increased energy- and nutrient-providing catabolic activities (e.g. autophagy). Additionally, the observed growth arrest, in combination with full nutrition (long days and 1% sucrose in the growth medium), accounts for increased loading of chemically reduced substrates into the TCA cycle (Sweetlove et al., 2010). In addition to known physiological substrates, namely acetate and pyruvate, levels of alternative substrates such as the highly accumulated branched chain and aromatic amino acids (Taylor et al., 2004; Araujo et al., 2011), and also GABA, highly increase (Table S2). As a non-proteogenic amino acid synthesized from Glu and subsequently metabolized via succinic semialdehyde dehydrogenase to succinate, GABA fuels the TCA cycle via the so-called GABA shunt (Landrieu et al., 2004).

The assumption that the TCA cycle is up-regulated is further supported by the observation that the alternative oxidases AOX1A and AOX1D, the potential alternative NAD(P)H dehydrogenase NDA2 and the ubiquinol-cytochrome c reductase complex subunit At5g25450, all of which are associated with the mitochondrial electron transport chain, display significantly higher transcript levels (Figure S4 and Table S1). These proteins usually function by disconnecting electron transport from energy consumption, by uncoupling substrate oxidation and proton transport. The alternative oxidases and the alternative NAD(P)H dehydrogenase in particular function to decrease the build-up of the proton motive force, which is the motor for ATP synthesis (Affourtit et al., 2002). Additionally it seems that an increased flow through the regular electron transport chain, leads to an overly reduced status of the cell. This assumption is further supported by the increase in transcript levels of genes related to the redox balance of the cell, such as glutathione S-transferase, glutaredoxin and thioredoxin (Figure S4 and Table S1).

Carbon/nitrogen balance is affected by AtTOR

As described above, down-regulation of AtTOR appears to mimic starvation conditions, leading to induction of senescence-like phenotypes. Under limited availability of organic nitrogen or carbon, Asn is the major compound for the nitrogen/carbon transport (Lee et al., 2008). Interestingly, based on the fact that nitrogen is not really limiting in our experiments, the level of Asn does not increase, but instead slightly decreases in the induced transgenic lines (Table S2). We did also not observe any increase in the levels of Glu and Gln (Table S2), which are the central intermediates involved in regulation of the nitrogen status in nitrogen-assimilating cells. This observation is contrary to the molecular phenotype observed in four recent studies, in which nitrate assimilation and the accumulation of Gln in TOR-deficient plants were shown to be strongly up-regulated, either mediated through the down regulation of TORC1 genes or the activation of the TOR-repressed phosphatase PP2A (Deprost et al., 2007; Ahn et al., 2011; Heidari et al., 2011; Moreau et al., 2012). In fact, we not only detected a significant decrease in Gln, but also in the levels of other nitrogen-containing intermediates (Arg, ornithine and citrulline) in induced amiR-tor plants. The catabolism of Arg appears to play an important role in nitrogen recycling, with the amines being precursors of polyamine synthesis (Takahashi and Kakehi, 2010). Silencing of AtTOR expression also resulted in slight changes in the levels of Arg, ornithine and spermidine (Figure 3), which is of particular interest as spermidine has been shown to play a significant role in plant growth (Lehmann et al., 2012) and thus may additionally function as a signal amplifier. Taken together, it is apparent that the observed nitrogen phenotype contradicts previous data from studies analysing the nitrogen balance in TOR-repressed plants, while perfectly phenocopying the described negative effect of increased CO2 on nitrogen fixation (Bloom et al., 2010). Thus, an important finding of this study is that, even though TOR repression appears to activate nitrate reduction and therefore nitrogen assimilation under normal carbon conditions (Ahn et al., 2011; Moreau et al., 2012), it leads to repression of nitrogen assimilation under increased carbon availability (e.g. high CO2 or an surplus of sucrose in the growth medium).

Secondary metabolism

Under limited nutrients, remobilization of carbon and nitrogen may further occur through secondary metabolism pathways. We observed coordinated changes in genes and metabolites associated with the phenylpropanoid and glucosinolate pathways in all EST-treated amiR-tor lines (Figure 5, Figure S4, and Tables S1 and S4). This observation is supported by the increase in levels of the MYB75/PAP1 transcription factor in transgenic seedlings with reduced AtTOR levels. MYB75 functions as an activator of anthocyanin biosynthesis and a repressor of the phenylpropanoid branch of lignin and secondary cell-wall polysaccharide biosynthesis, suggesting a role for allocation of carbon between the two branches of the phenylpropanoid pathways (Borevitz et al., 2000). Additionally, we observed up-regulation of metabolites and genes belonging to the flavonoid biosynthetic pathway, such as F3H/SGR1, cinnamoyl CoA reductase, DFR and UF3GT, and precursors of the lignin pathway (CAD1, CAD7 and CAD8; Routaboul et al., 2006).

Similarly, the levels of glucosinolates, which are nitrogen- and sulfur-rich secondary metabolites, may be influenced by environmental factors such as abiotic stress or altered mineral nutrition and auxin signaling (Halkier and Gershenzon, 2006). Their synthesis, which is regulated by the tryptophan regulation protein MYB34/ATR1, is accomplished in five steps comprising oxidation of tryptophan (or other amino acids) to aldoximes by cytochrome P450 monooxygenases (Mikkelsen et al., 2000). Although we observed an increase in CYP79B2, CYP79B3 and CYP71B15 transcript levels, the MYB34 levels were clearly repressed in our transgenic amiR-tor lines (Table S1). Additionally, catabolism of glucosinolates may be regulated by nitrilases (NIT), which function in detoxification and nitrogen recycling, because the nitrogen of the nitrile group is recovered in other nitrogen metabolites such as asparagine or ammonia (Janowitz et al., 2009). Expression of NIT3 and NIT4 was strongly induced in the transgenic lines with reduced AtTOR expression levels.

Storage carbon accumulation is regulated by AtTOR activity

Our metabolomics study further showed that sucrose levels tend to decrease in all the amiR-tor lines (Table S2). This observation is partly supported by the amiR-tor transcript data, in which genes involved in sucrose breakdown such as those encoding alkaline/neutral invertase (A/N-INVH), sucrose synthase (SUS3) and phosphofructokinase (PFK3) display higher transcript levels (Table S1). Furthermore, we found increased expression of the glucose-6-phosphate/phosphate transporter-encoding gene GTP2. GTP2 has been shown to be transcriptionally regulated under environmental stress, sugar-induced senescence or phosphate starvation conditions, enabling fine-tuned regulation under conditions of impaired carbohydrate metabolism (Kunz et al., 1993). Indeed, disruption of TORC1 in yeast leads to carbon storage through accumulation of glycogen (Schmelzle et al., 2004). Starch, which is highly accumulated in the amiR-tor lines (Figure 4a), appears to redirect the carbon fluxes, contributing to the regulation of biomass. Similarly to the increase in starch, a strong increase in polyunsaturated, long-chain fatty acid-containing TAGs was observed (Figure 4b), providing a major sink for the accumulated carbon. It has been shown previously that the metabolic disruption occurring in senescence causes the conversion of galactolipids to TAGs, preceding conversion to sucrose (Livaja et al., 2008). Additionally, it has been reported that the transcription factor ABI4, which is activated by ABA signaling under nitrogen starvation, binds and activates the promoter of DGAT1 (diacylacetyltransferase1), which is a crucial protein for production of TAGs (Yang et al., 2011). However, neither of these genes (ABI4 and DGTA1) nor a significant breakdown of galactolipids was observed in the induced amiR-tor lines, indicating that there must be alternative routes to the observed accumulation of TAGs.

Even though the complete mechanism of triacylglyceride and starch accumulation has not been determined in our study, there is another highly relevant implication of the carbon partitioning observations: as illustrated in Figure 1, growth of the non-induced amiR-tor lines is unaltered, compared to wild-type plants, meaning that these transgenic plants produce a biomass equivalent to the wild-type plants. Having been allowed to reach full growth, these transgenic plants could then be subjected to in planta refinement of their metabolic inventory, by simply repressing the activity of the TOR gene, which would consequently lead to a severe conversion and accumulation of highly reduced carbon (Figure 6). Such an approach may allow to increase the yield and quality of extractable bioenergy for biofuel production. One of the main limitations in converting plant biomass to fuels involves saccharification, namely breakdown of biomass to monosaccharides (Carroll and Somerville, 2009). Saccharification of starch is much easier and cheaper than saccharification of lignocellulosic material such as cellulose. Similarly, increased amounts of TAGs may be used as an energy-rich product as a substitute for conventional diesel (Durrett et al., 2008).

Figure 6.

Schematic illustration of a possible biotechnological strategy for the refinement of metabolic composition for production of plant-derived bioenergy and biofuels.

Experimental Procedures

Generation of EST-inducible transgenic plants and growth conditions

The amiR-tor-specific sequences were selected using the Web MicroRNA designer (http://wmd2.weigelworld.org/cgi-bin/mirnatools.pl; Schwab et al., 2006), targeting the region between 3503 and 3523 bp of the AtTOR (At1 g50030) mRNA. Primers for amiRNA precursors were designed using the Web MicroRNA designer (http://wmd2.weigelworld.org/cgi-bin/mirnatools.pl; Schwab et al., 2006). Briefly, amiRNAs were generated by sequential PCR using RS300 (an miR-319a pBSK backbone) and the primers listed in Table S5. The resulting PCR product was digested using XhoI and SpeI, inserted into pCR2.1-TOPO (Invitrogen, www.invitrogen.com) and then cloned via XhoI and SpeI sites into the pER8 EST-inducible vector (Zuo et al., 2000). All constructs were confirmed by sequence analysis. Arabidopsis thaliana (Col-0) was transformed by the floral-dip procedure (Clough and Bent, 1998) using Agrobacterium tumefaciens strain GV3101. The seeds of transgenic plants were selected on hygromycin plates and verified by PCR. Experiments were performed in T2 transgenic lines. Transgenic amiR-tor seeds were surface-sterilized and germinated in MS plates containing hygromycin under continuous light. Two-week-old transgenic seedlings were transferred to MS plates with 20 μm EST to induce over-expression of amiR-tor (and, consequently, repression of AtTOR levels) under the control of an EST-inducible promoter. Identically treated wild-type Col-0 and pER8 empty vector-transformed seedlings were used as controls. All the genotypes were also transferred to MS plates without EST as an additional control. Seedlings were harvested after 0, 3 and 6 days. For phenotypic analysis, seedlings were selected in MS medium containing 1% sucrose and hygromycin. After 1 week, seedlings were transferred to plates or Mason jars with or without 20 μm EST for at least 2 weeks, and grown under 16 h light/8 h dark at day/night temperatures of 20/16°C. The intensity of the fluorescent light was 100 μmol m−2 sec−1 and the relative humidity was 60/75%. Seedlings were harvested after 0, 3 and 6 days for molecular studies.

RNA preparation and transcript expression profiling

Total RNA extraction, cDNA synthesis and quantitative RT-PCR were performed as described previously (Czechowski et al., 2004; Caldana et al., 2007). Repression of AtTOR was confirmed by quantitative RT-PCR using three primer pairs spanning different regions of the TOR gene. The relative expression level of TOR was calculated by first normalizing the TOR transcript levels relative to the reference gene (UBQ10) (ΔCT) for each transgenic line and treatment (i.e. EST and control). Next, the percentage expression level was calculated by comparing the EST-treated plants and controls in the same genotype using the equation inline image, as described by Portereiko et al. (2006).

All the primer sequences used in this study are given in Table S5. Microarray analysis was performed using Affymetrix ATH1 arrays of three independent amiR-tor lines (amiR-tor9, amiR-tor17 and amiR-tor20) and empty vector lines harvested 3 and 6 days after amiR induction by EST or controls. Total RNA was isolated from pools of ten Arabidopsis seedlings and labeling was performed using 1 μg RNA as previously described. Affymetrix ATH1 hybridizations were performed by Atlas Biolabs. Raw CEL files were analyzed using Bioconductor software (Gentleman et al., 2004) for R (http://www.atlas-biolabs.de), and the GC Robust Multi-array Average (GC-RMA) expression estimation was obtained using the gcrma package (Wu et al., 2004). Expression data obtained for EST-inducible amiR-tor and empty vector lines were submitted to the NCBI Gene Expression Omnibus (GEO) repository (http://www.ncbi.nlm.nih.gov/geo) under accession number GSE38878.

Transcript fold change was calculated as follow: first the expression levels of non-induced plants were subtracted from the levels of EST-induced plants, generating a fold change for each genotype. Second, the fold change of empty vector control was subtracted from the fold change of the amiR-tor lines. (Table S2). A transcript was considered induced or repressed when the absolute fold change was ≥2.

Significant over-representation of functional annotations was analyzed using PageMan with Benjamini–Hochberg correction (Usadel et al., 2006). Gene clusters of functional categories were visualized by heatmaps or using the BiNGO 2.3 plugin tool in Cytoscape version 2.6 with MapMan bin files (Maere et al., 2005; Usadel et al., 2005). Over-represented MapMan bins were identified using a hypergeometric test with a significance threshold of 0.05 after a Benjamini–Hochberg false discovery rate correction (Benjamini and Hochberg, 1995).

Metabolite extraction

For the metabolite extraction, five independent pools of ten seedlings were ground to a fine powder, and 50 mg of ground tissue of each independent pool was extracted in 1 ml of a pre-cooled (−15°C) mixture of Methyl-tert-butyl ether (3:1:1 v/v/v), as described previously (Giavalisco et al., 2011). MTBE buffer was added to the frozen powder and vortexed until the tissue was fully re-suspended. The sample was incubated for 30 min at 4°C on an orbital shaker, before incubating it for another 10 min in an ultra-sonication bath with ice. Then 700 μl of UPLC-grade water/methanol (3:1 v/v) were added. The tubes were then 20 800 g at room temperature for 5 min, and 700 μl of the upper MTBE (green) phase, containing the lipids, were transferred to a fresh 1.5 ml Eppendorf tube and dried down in a Speed-Vac (www.eppendorf.com). The remaining organic phase was then completely removed from the extract, and 150 μl of the lower, polar fraction were transferred to a fresh 1.5 ml Eppendorf, and used for GC analysis after drying down. Another aliquot of 500 μl of the same fraction was transferred to a fresh 1.5 ml Eppendorf tube. After drying it down, this aliquot was used for the secondary metabolite analysis. The remaining pellet in the tube, which contains starch and proteins, was either used for protein analysis (e.g. Western blots) or for starch measurements as described by Smith and Zeeman (2006).

GC-TOF MS data analysis

Prior to GC-TOF MS analysis, the samples were derivatized. A variation of the two-stage technique used by Roessner et al. (2001) was employed as described previously (Krueger et al., 2011). First the carbonyl moieties were protected via methoximation in a 90 min reaction at 30°C using 5 μl of 40 mg/ml methoxyamine hydrochloride (www.sigmaaldrich.com) in pyridine (www.merck.com), followed by derivatization of acidic protons via a 30 min reaction at 37°C with addition of 45 μl MSTFA (N-methyl-N-trimethylsilyltrifluoroacetamide) (www.mn-net.com). Forty microliter of a mixture of retention time standards (fatty acid methyl esters), containing 3.7% w/v heptanoic acid, 3.7% w/v nonanoic acid, 3.7% w/v undecanoic acid, 3.7% w/v tridecanoic acid, 3.7% w/v pentadecanoic acid, 7.4% w/v nonadecanoic acid, 7.4% w/v tricosanoic acid, 22.2% w/v heptacosanoic acid and 55.5% w/v hentriacontanoic acid dissolved in tetrahydrofuran at 10 mg/ml total concentration was added prior to trimethylsilylation. One microliter of the derivatized sample was injected onto the column, and analysis was commenced in non-split mode. The analysis was performed under the following temperature program: 5 min isothermal heating at 70°C, followed by a 5°C min−1 increase in oven temperature to 350°C and a final 5 min heating at 330°C. The samples were run in an Agilent 7683 series autosampler (www.agilent.com) coupled to an Agilent 6890 gas chromatograph coupled to a Leco Pegasus 2 time-of-flight mass spectrometer (www.leco.com). Chromatograms were exported from Leco ChromaTOF software (version 3.25) to R software (www.r-project.org). Peak detection, retention time alignment and library matching were performed using the Target Search R package (Cuadros-Inostroza et al., 2009). Metabolites were quantified by the peak intensity of a compound-specific mass.

Lipid analysis

For lipid analysis, which was performed as described previously (Giavalisco et al., 2011; Hummel et al., 2011), the dried pellets were re-suspended in 250 μl of a mixture of acetonitrile/isopropanol (7:3 v/v), thoroughly vortexed and centrifuged. Then 100 μl aliquots were transferred into glass vials and 2 μl were injected and separated on an Acquity UPLC system (www.waters.com) using a reversed-phase C8 column. The UPLC solvents (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 μl/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 re-equilibrated for 4 min (22 min total run time). The samples were either measured in positive or negative ion mode. The mass spectra were acquired using an Exactive high resolution mass spectrometer (www.thermofisher.com). The spectra were recorded using alternating full-scan and all-ion 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 msec with a target value of 1 × E6 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 UPLC gradients. Obtained raw chromatograms were further processed using Excalibur software version 2.10 (Thermo Fisher) or Refiner MS® software version 6.0 (GeneData, www.genedata.com). Peaks from raw chromatograms were first determined, then aligned by their parent masses, chemical noise is subtracted, and a final alignment file of all chromatograms is the output, which contains information about m/z ratio, retention times and retention time deviations for each annotated peak. The peaks were assigned using the chemical formula database GoBioSpace (gmd.mpimp-golm.mpg.de) to compounds with sum formulas, according to retention time and m/z ratios, and were finally manually corrected (Hummel et al., 2011). The obtained data matrix was then further processed for normalization and statistical analysis.

Secondary metabolite analysis

For the secondary metabolite analysis, the dried pellets from the polar phase were fully resuspended in 200 μl UPLC-grade water, centrifuged and separated on the Acquity UPLC system using an HSS T3 C18 reversed-phase column (length 100 mm, internal diameter 2.1 mm, particle size 1.8 μm; Waters) (Giavalisco et al., 2011). The mobile phases were 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B). Two microliter samples were loaded per injection and separated in the following gradient: 1 min at 99% A, a 13 min linear gradient from 99% A to 65% A, a 14.5 min linear gradient from 65% A to 30% A, a 15.5 min linear gradient from 30% A to 1% A, hold at 1% A until 17 min, a 17.5 linear gradient from 1% A to 99% A, and re-equilibration of the column for 2.5 min (20 min total run time). The samples were measured in positive or negative ion mode and processed as described above for the lipid measurements.

Statistical analyses

All statistical analyses were performed using R version 2.12.2. (www.r-project.org) Metabolite data obtained from GC-TOF MS analysis were normalized by fresh weight, followed by sample total ion count and global outlier replacement as described previously (Giavalisco et al., 2011; Huege et al., 2011; Hummel et al., 2011). Lipophilic and polar fractions analyzed by liquid chromatography high resoloution mass spectrometry were normalized in a same manner, except using internal standards (phosphatidylcholine 34:0 and ampicillin, respectively) instead of total ion content. The significance of differences were tested for each metabolite by a paired t test (< 0.05), comparing the same genotype under EST-treated conditions and control conditions. All raw data values are given in Table S6.

Acknowledgements

We acknowledge Änne Eckardt, Gudrun Wolter and Antje Bolze for excellent technical assistance, Dirk Steinhauser and Salma Balazadeh for helpful discussions, and Josef Bergstein for photographic work.

Ancillary