Selective mRNA translation coordinates energetic and metabolic adjustments to cellular oxygen deprivation and reoxygenation in Arabidopsis thaliana

Authors

  • Cristina Branco-Price,

    1. Center for Plant Cell Biology, University of California, Riverside, CA 92521, USA
    2. Department of Botany and Plant Sciences, University of California, Riverside, CA 92521, USA
    3. Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa, 2780-901 Oeiras, Portugal
    Search for more papers by this author
  • Kayla A. Kaiser,

    1. Center for Plant Cell Biology, University of California, Riverside, CA 92521, USA
    2. Department of Chemistry, University of California, Riverside, CA 92521, USA
    Search for more papers by this author
  • Charles J. H. Jang,

    1. Center for Plant Cell Biology, University of California, Riverside, CA 92521, USA
    2. Department of Botany and Plant Sciences, University of California, Riverside, CA 92521, USA
    Search for more papers by this author
  • Cynthia K. Larive,

    1. Center for Plant Cell Biology, University of California, Riverside, CA 92521, USA
    2. Department of Chemistry, University of California, Riverside, CA 92521, USA
    Search for more papers by this author
  • Julia Bailey-Serres

    Corresponding author
    1. Center for Plant Cell Biology, University of California, Riverside, CA 92521, USA
    2. Department of Botany and Plant Sciences, University of California, Riverside, CA 92521, USA
      (fax +951 827 4437; e-mail serres@ucr.edu).
    Search for more papers by this author

(fax +951 827 4437; e-mail serres@ucr.edu).

Summary

Cellular oxygen deprivation (hypoxia/anoxia) requires an acclimation response that enables survival during an energy crisis. To gain new insights into the processes that facilitate the endurance of transient oxygen deprivation, the dynamics of the mRNA translation state and metabolites were quantitatively monitored in Arabidopsis thaliana seedlings exposed to a short (2 h) or prolonged (9 h) period of oxygen and carbon dioxide deprivation and following 1 h of re-aeration. Hypoxia stress and reoxygenation promoted adjustments in the levels of polyribosomes (polysomes) that were highly coordinated with cellular ATP content. A quantitative comparison of steady-state and polysomal mRNA populations revealed that over half of the cellular mRNAs were restricted from polysome complexes during the stress, with little or no change in abundance. This selective repression of translation was rapidly reversed upon reoxygenation. Comparison of the adjustment in gene transcripts and metabolites demonstrated that profiling of polysomal mRNAs strongly augments the prediction of cellular processes that are altered during cellular oxygen deprivation. The selective translation of a subset of mRNAs promotes the conservation of ATP and facilitates the transition to anaerobic metabolism during low-oxygen stress.

Introduction

Oxygen is essential for aerobic metabolism in eukaryotic cells. In plants, the availability of oxygen is limited in highly metabolically active cells, internal tissues and as a consequence of soil compaction, flooding or submergence (Bailey-Serres and Voesenek, 2008; Fukao and Bailey-Serres, 2004; Geigenberger, 2003). Cellular metabolism must be rapidly and reversibly configured when oxygen levels are insufficient to maintain the production of ATP via mitochondrial oxidative phosphorylation. The rearrangement of metabolism during moderate to severe hypoxia stress (HS) involves increased catabolism of soluble carbohydrates for substrate-level production of the ATP needed to maintain critical processes such as RNA transcription, protein synthesis and the activity of plasma and vacuolar membrane proton pumps that limit acidification of the cytosol (Bailey-Serres and Voesenek, 2008; Gibbs and Greenway, 2003). The response of the plant to oxygen deprivation includes elevation of mRNAs encoding enzymes that promote sucrose breakdown and entry into glycolysis, as well as the conversion of pyruvate to the fermentation end-products lactate and ethanol (Branco-Price et al., 2005; Klok et al., 2006; Lasanthi-Kudahettige et al., 2007; Liu et al., 2005; Loreti et al., 2005). Although transcriptional induction is crucial for reconfiguration of metabolism during HS, it has long been recognized that protein synthesis is limited to translation of a subset of mRNAs during this stress (Fennoy et al., 1997; Sachs et al., 1980). To date, the contribution of transcriptional and translational regulation to the initial and prolonged adjustments in metabolism that aid survival of transient moderate to severe HS has not been addressed at a global level in either plants or animals.

Cellular conditions that deplete ATP reserves necessitate a reduction of energetically costly cellular processes, such as protein synthesis. The primary rate-limiting step in mRNA translation is the initial recruitment of the 43S pre-initiation complex. This process consumes multiple molecules of ATP in the removal of mRNA secondary structure and one molecule of GTP in the recognition of the initiation codon by the scanning 40S ribosomal subunit. Polyribosomes (polysomes) form as a result of sequential initiation of ribosomes. It follows that a reduction in the number of ribosomes per mRNA is diagnostic of a restriction in initiation of translation (Proud, 2007). Those mRNAs that are not undergoing translation are sequestered into a cytosolic mRNA ribonucleoprotein (mRNP) complex where they may be stabilized or degraded (Hoyle et al., 2007; Parker and Sheth, 2007). Since the addition of each amino acid to the nascent polypeptide requires consumption of four nucleotide triphosphates, any reduction in translation facilitates energy conservation (Proud, 2007).

It was shown previously that the majority of cellular mRNAs are translated at reduced levels in Arabidopsis thaliana seedlings exposed to prolonged oxygen deprivation, although a subset of the highly induced mRNAs escape translational repression (Branco-Price et al., 2005). Differential regulation of mRNA translation under HS was also confirmed in gene expression studies in mammalian cells (Koritzinsky et al., 2005; Koumenis et al., 2002; Thomas and Johannes, 2007). These analyses, however, did not consider the independent contribution of transcript accumulation versus translational regulation in the reconfiguration of metabolism and other cellular processes during HS and reoxygenation. Here, a quantitative evaluation of translational and metabolic adjustments in response to short term to prolonged oxygen deprivation exposes the critical importance of selective mRNA translation in limiting energy-consuming and non-vital processes via a mechanism that expedites the return to homeostasis upon reoxygenation.

Results

Oxygen availability regulates cellular polysome and adenylate status

The goal of this study was to evaluate the modulation of multiple levels of gene regulation and metabolic acclimation in response to the energy crisis imposed by oxygen deprivation (Figure S1). Arabidopsis seedlings propagated on the surface of solid medium were deprived of oxygen by replacement of ambient air with argon, a treatment that reduces the availability of oxygen and carbon dioxide. Because the plants in this system endure a progressive decrease in oxygen availability, we describe this as HS, although some cells may reach a state of anoxia. Seedlings were stressed for a brief (2 h, 2HS) or prolonged (9 h, 9HS) period, or reoxygenated following 9HS by re-aeration for 1 h (1R). Control seedlings were mock treated for 2 h (2NS) and 9 h (9NS) in air. The stress treatments were non-lethal, resulting in death of the primary root tip but promotion of lateral root development (data not shown). A quantitative analysis of ribosome complexes confirmed that 2HS and 9HS caused an approximately 50% decrease in polysomes that is accompanied by an increase in 80S monosomes and ribosomal subunits (Figure 1a), indicative of inhibition of translational initiation. This repression in translation was rapidly reversible, since polysome levels were 88% recovered within 1R (Figure 1b). As expected, HS stimulated a reversible decline in ATP that was accompanied by a slight increase in ADP (Figure 1c). Together, these data expose a strong correlation between cellular ATP and polysome contents.

Figure 1.

 Coordinate alterations in polysome and ATP levels.
Absorbance profiles of rRNA complexes obtained from seedlings treated for 2 and 9 h without stress (NS) (black line), 2 and 9 h with hypoxia stress (HS) (gray line) and 9 h HS followed by 1 h reoxygenation (red line, 1R) obtained by pelleting ribosomes of detergent-treated extracts through 1.75 m sucrose cushions followed by fractionation in 20–60% (w/v) sucrose gradients (a). Traces represent absorbance after subtraction of the baseline of a blank sucrose gradient.
(b) Polysome content quantified by integration of the polysome-containing area divided by the area of the profile containing subunits, ribosomes and polysomes.
(c) The ATP and ADP content determined from seedling extracts.
In (b) and (c), values in parentheses are the percentage relative to the non-stress control, data are mean ± SE of three independent experiments, and an asterisk indicates a significant difference from the corresponding NS control, as determined by Student’s t-test, < 0.05.

Positive and negative regulation of gene expression through selective mRNA translation

Steady-state (total) and polysomal mRNA populations were profiled following short and prolonged HS and reoxygenation to monitor the dynamics of translational regulation of individual mRNAs. The analyses were performed with a transgenic Arabidopsis line expressing a FLAG-tagged form of ribosomal protein L18 (RPL18) that allows for efficient immunopurification (IP) of ribosome complexes with mass and abundance consistent with the complete polysome population (Zanetti et al., 2005). A benefit of this approach is that polysomes are purified from crude cell extracts without contamination by other mRNP complexes, such as stalled 43S pre-initiation complexes, P-bodies or stress granules. The hybridization of three biological replicate samples of total and polysomal mRNA to Affymetrix ATH1 arrays yielded data with a strong correlation in signal intensity (R2 > 0.97,Table S1), confirming the reproducibility of the seedling responses to varying oxygen availability and the method of purification of polysomal mRNA.

To assess translational regulation under HS, total and polysomal mRNA abundance was evaluated for the ∼23 000 genes represented by probe pair sets on the array. The ∼9000 mRNAs with signal intensity above background in all 30 hybridizations (mRNAs displaying only present calls by the Affymetrix mas 5.0 algorithm) were further analyzed (Table S2). These represent over 90% of all mRNA signals detected in the hybridizations. The signal log2 ratio (SLR) data showed only a modest correlation between alteration in transcript abundance in the total and IP polysomal RNA populations following HS (R2 = 0.45 and 0.59 following 2HS and 9HS, respectively) (Figure 2a,b).

Figure 2.

 Comparison of abundance of total and polysomal mRNA following hypoxia stress (HS) and reoxygenation.
Change in total mRNAs (x-axis) versus change in immunopurified polysomal (IP) mRNA: (a) 2 h (2HS/2NS; NS = mock treatment), (b) 9 h of HS (9HS/9NS), (c) and (d) 1 h of reoxygenation following 9 h HS (1R/9HS and 1R/9NS respectively). = 3 independent experiments. Values represent log2-transformed signal log2 ratio (SLR) data of comparisons between different treatments (e.g. SLR = log2 signal Treatment A gene i:log2 signal Treatment B gene i), using genes with ‘present’ call in all 30 hybridizations (= 8863 probe pair sets). The diagonal dashed line indicates perfect correlation between samples.

The traditional profiling of total cellular RNA content exposed the downregulation of 4.2% of the genes after 2HS and 22.1% after 9HS, respectively (Table 1). In marked contrast, a significant decrease in the amount of individual mRNAs in polysomes was observed for 67.5% and 73.8% of the mRNAs at 2HS and 9HS, respectively. The repression of translation was both rapid and sustained, since 91% of the mRNAs that declined in the polysome population after 2HS remained poorly loaded after 9HS (Figure S2). The consequence of a reduction of a polysomal mRNA is a decline in the amount of gene product synthesized. The overall decrease in mRNA translation also restricted the synthesis of proteins of HS-induced mRNAs. Of the 6.0% and 20.9% of the mRNAs that increased significantly in steady-state abundance after 2HS and 9HS, respectively, only 2.0% and 4.6% were detected at significantly higher levels in polysomes (Table 1). One hour of re-aeration reversed the restriction in translation for 90% of the RNAs with reduced initiation during HS (Figures 2c,d and S2). By contrast, only 10% of the transcripts showed an increase at the steady-state level (Table 1). The translation of many stress-induced mRNAs continues upon reoxygenation, since only 20% (89) of mRNAs that were highly induced in polysomes after 9HS were significantly reduced in polysomes after 1R (Figure S2).

Table 1.   Limited overlap between total and polysomal mRNA with a significant change in abundance following hypoxia stress (HS) and reoxygenation
Genes2 h HS9 h HS9 h HS + 1R
Total mRNAOverlapPolysomal mRNATotal mRNAOverlapPolysomal mRNATotal mRNAOverlapPolysomal mRNA
  1. Percentage and number (no.) of mRNAs present above the detection limit in 30 hybridizations in all three independent biological replicates (8863 probe pair sets) and with a significant increase or decrease in abundance in total or polysomal mRNA (false discovery rate cut-off, 0.001). Overlap: mRNAs with a significant change in both total and polysomal mRNA populations. 1R, re-aeration for 1 h.

Significant increasePercent (no.)4.0 (355)2.0 (178)0.4 (35)16.3 (1444)4.6 (405)0.4 (32)0.4 (34)7.1 (632)59.5 (5276)
Significant decreasePercent (no.)0.0 (0)4.2 (372)63.3 (5609)0.4 (32)21.7 (1926)52.1 (4620)3.5 (310)1.3 (112)1.2 (106)

These data confirm that oxygen deprivation imposes a strong, selective and sustained impairment of initiation of translation of the majority of mRNAs [63.3% (2HS) and 52.1% (9HS) of cellular mRNA content], without a concomitant decrease in transcript abundance. This rapid re-recruitment of the majority of these mRNAs to polysomes upon reoxygenation is likely to help the restoration of cellular homeostasis.

Distinction in regulation of abundance and translation of individual mRNAs

To further investigate the seedling response to oxygen deprivation, genes were identified that had similar patterns of transcript accumulation and translation in response to HS and 1R (Figure 3, Table S3a). A k-means cluster analysis (= 10) resolved three subgroups based on change in total mRNA abundance relative to the NS samples after 2HS and/or 9HS: mRNAs for clusters 1–3 increased after 2HS and/or 9HS, mRNAs for clusters 4 and 5 were reduced and mRNAs for clusters 6–10 were generally maintained. Reoxygenation altered the abundance of some mRNAs, as assessed in the 1R/9HS and 1R/9NS SLR comparisons. The cluster analysis exposed a strong dichotomy between the dynamic regulation of transcript abundance and translation. The effect of HS and reoxygenation on polysomal mRNA levels was generally independent of the change in steady-state abundance for most gene clusters. An exception was the strong concordant induction in accumulation and translation of cluster 1 mRNAs. These transcripts, which were highly induced and recruited to polysomes during HS, comprise 0.4% and 4.0% of the total cellular mRNA content under NS and HS conditions, respectively (Table S3b). Although 25.0% and 60.7% of cluster 2 mRNAs showed a significant increase at the steady-state level after 2HS and 9HS, only 18.6% and 20.1% showed a significant increase in the polysomal population (Table S3c). In the case of cluster 3 mRNAs, the HS induction was largely restricted to steady-state abundance with recruitment to ribosomes only occurring upon reoxygenation, indicating that the elevation of these mRNAs during HS may contribute to preparation of the cell for reoxygenation.

Figure 3.

 Oxygen deprivation promotes discord between transcript abundance and translation status.
K-means clustering analysis performed on four mean signal log2 ratio (SLR) comparisons of total and immunopurified polysomal (IP) mRNA levels of transcripts detected with ‘present’ call in all 30 hybridizations (8863 probe pair sets) from three independent experiments. A representative panel for each of the ten clusters is shown; colors indicated increase (yellow), decrease (blue) or no-change (black) in mRNA abundance. Columns indicate the number of genes (no. genes) and percentage of the total genes (% genes) per cluster. 2HS/2NS, change in response to 2 h HS; 9HS/9NS, change in response to 9 h HS; 1R/9HS, change upon reoxygenation; and 1R/9NS, change after reoxygenation relative to non-stress (NS).

The comparison of the mRNAs induced after 2HS and 9HS revealed distinctions between the response to short and prolonged stress. There were 155 mRNAs that were significantly induced in the polysomal fraction at both time points (Table S4). However, polysomal mRNAs induced after 2HS were enriched in transcripts encoding signaling proteins and transcription factors, whereas after 9HS they were biased towards proteins involved in the response to external stimuli and stress. The mRNAs of cluster 1 include enzymes that enhance catabolism of sucrose and promote anaerobic fermentation, as well as signaling components (transcription factors, protein kinases and receptors), enzymes involved in cell wall modification and amino acid degradation, proteins involved in regulation of redox and reactive oxygen species, and proteins of unknown biological function (Table S5).

Translational repression limits the expression of a large proportion of cellular mRNA during oxygen deprivation. The 7546 mRNAs that decreased in polysomes under HS (clusters 4–10) were further distinguished by whether the decline was accompanied by a reduction (clusters 4, 5 and 9) or little change (clusters 6, 7 and 10) in steady-state abundance, or whether an increase in abundance was offset by reduced translation (cluster 8). Messenger RNAs of clusters 4, 5 and 9 displayed concordant regulation in the total and polysomal populations (cluster 9 at 9HS only), antithetical to regulation of cluster 1 mRNAs. Notably, reduced association with ribosomes was not necessarily coupled with mRNA destabilization, as indicated by mRNAs of clusters 6, 7 and 8. For these mRNAs, translational repression under HS and re-recruitment to polysomes upon reoxygenation occurred without alteration in abundance. Together, clusters 6, 7 and 8 mRNAs comprise ∼50% of total cellular mRNA detected under both 2HS and 9HS (Table S3b). Cluster 10 mRNAs, which comprise ∼16% of the mRNA content, also showed a modest reduction at both steady-state and translational levels. The marked repression of translation of ∼50% of the cellular mRNA content was remarkably consistent with the stabilization of ATP content at ∼50% of NS levels.

Restriction of translation in hypoxic seedlings contributes to energy conservation

The mRNAs that were downregulated by restriction of translational initiation during HS (clusters 4–8) encode proteins involved in a spectrum of cellular functions (Table S5). An overall effect of the downregulation of translation of cellular mRNAs during the stress is the conservation of cellular ATP reserves. The cluster analysis further resolved the coordinate transcriptional/translational regulation of mRNAs encoding proteins with similar biological or functional characterization. For example, proteins that regulate cell wall formation, transcription, signaling, cell division as well as hormone and lipid metabolism were abundant in cluster 4 (Table S5b). The severe translational repression of these mRNAs under HS was markedly reversed after 1R.

The robustly regulated mRNAs of clusters 5–7 encode many components of the synthesis machinery for cytosolic protein, including ribosomal proteins (RPs), initiation, elongation and release factors (most highly represented in cluster 7). An evaluation of the 353 genes known to encode proteins associated with cytosolic or organellar protein synthesis confirmed that HS strongly inhibits the production of nearly all proteins involved in translation or translational regulation (Figure 4a). These mRNAs account for 9.2% and 9.3% of the total mRNA at 2NS and 9NS, respectively (Table S3b). A strong and sustained repression of translation of cytosolic RP mRNAs occurred within 2HS, with an average three-fold decline in polysome association (SLR = −1.5 ± 0.46). Figure 4(a) also reveals that curtailment of RP synthesis was ensured by reduced steady-state abundance of the mRNA encoding the RPs that were not strongly translationally repressed. The coordinate repression of synthesis of cytosolic and organellar RP was observed in Arabidopsis following mild dehydration stress (Kawaguchi and Bailey-Serres, 2005; Kawaguchi et al., 2004) and sucrose starvation (Nicolaïet al., 2006), indicating that ribosome biogenesis is routinely limited under conditions that require conservation of energy reserves.

Figure 4.

 Dynamics of accumulation and translation of mRNAs encoding proteins involved in translation and primary metabolism.
(a) Heat map display of mRNAs encoding components of cytosolic and organellar ribosomes and translation factors. Genes organized by ribosome complex or function and then sorted by signal log2 ratio (SLR) for immunopurified polysomal (IP) mRNA at 2 h hypoxia stress (2HS).
(b) Heat map display of mRNAs encoding enzymes involved in central carbon and nitrogen metabolism. Genes organized by pathway. Columns represent mean SLR comparisons of total and IP mRNA levels of transcripts of selected genes (all detected with ‘present’ call in all 30 hybridizations). Colors indicated increase (yellow), decrease (blue) or no-change (black) in mRNA abundance. Table S6 contains the gene lists used.

Another example of mRNAs that were similarly regulated by HS are those that encode proteins and enzymes involved in cell wall degradation, modification and synthesis (Figure S5b). In the case of cell wall biosynthesis, most of the mRNAs declined at both the steady-state and polysomal mRNA levels, indicating multiple levels of control. The mRNAs encoding proteins involved in cellulose biosynthesis were poorly translated during HS, possibly restricting the highly energy consuming processes of glucan chain polymerization. On the other hand, transcripts of three genes that encode xyloglucan endotransglycosylase/hydrolases were strongly induced in both total and polysomal mRNA populations during HS. These enzymes promote cell wall extensibility and play a role in the promotion of cell elongation during flooding stress (Bailey-Serres and Voesenek, 2008). Transcripts encoding enzymes and proteins involved in photosynthetic processes were modestly or strongly repressed at the level of translation during the stress (enriched in clusters 6 and 10). Similarly, many mRNAs encoding enzymes involved in transcription and signal transduction were strongly negatively regulated by translational repression (Figure S5c,d). These data emphasize that translational repression is a key mode of regulation of production of proteins involved in a number of cellular processes during low oxygen stress.

Selective mRNA translation mirrors metabolic adjustments to low oxygen stress and reoxygenation

The dynamic modulation of accumulation of polysomal and total mRNA in response to oxygen deprivation predicts that the enhanced catabolism of sucrose and glycolysis leads to production of the anaerobic fermentation end-products lactate, ethanol, alanine and γ-aminobutyric acid (GABA) (Figure 4b), consistent with earlier studies of metabolic adjustments to low oxygen stress (Bailey-Serres and Voesenek, 2008). By contrast, the modest to pronounced regulation of total and polysomal mRNAs encoding enzymes involved in the pentose phosphate pathway and tricarboxylic acid (TCA) cycle (Figure 4b) suggests that these pathways may have reduced importance during the stress. To directly investigate the extent to which the reconfiguration of metabolism during HS and reoxygenation is driven by dynamics in steady-state mRNA accumulation or translation, metabolite levels were evaluated by 1H nuclear magnetic resonance (NMR) spectroscopy or enzymatic assay (Tables 2 and S7). The five tissue samples used for metabolic analysis included the three used in the RNA profiling study.

Table 2.   Fold-change in metabolites in response to hypoxia stress (HS) and reoxygenation
MetaboliteShift2HS/2NS9HS/9NS1R/9NS1R/9HS
  1. Metabolite quantification by 1H NMRa or enzymaticb assay for three or more independent biological replicate samples for 2NS, 9NS, 2HS, 9HS and 1R. Absolute values for metabolites measured by enzymatic methods are reported in Table S7. Shift is the 1H chemical shift of the resonance used for metabolite quantification. Significant changes in metabolite levels were evaluated by the Student’s t-test (*< 0.05; **< 0.001). Lactate was detected by 1H NMR in the 2H, 9H and 1R samples, but the fold change could not be quantified because of the low level and overlap with lipid resonances in the NS samples. For lactate, + and +++ denote moderate and strong increases, respectively. NA, not applicable. Seedlings were stressed as follows: 2HS, 2 h HS; 9HS, 9 h, 9HS; 1R, reoxygenated following 9HS by re-aeration for 1 h; 2NS, 9NS control seedlings mock treated for 2 h or 9 h in air.

ADPbNA1.7*1.8*1.20.7
ATPbNA0.6*0.5*0.81.7*
Alaninea1.423.3*6.5**6.8**1.0
Asparaginea2.791.00.7*0.7**1.1
Cholinea3.130.80.8*0.81.1
EthanolbNA1.5**1.31.10.8
Fructosea4.020.81.11.21.0
GABAa2.961.01.3**1.2**0.9
Glucosea5.150.81.21.21.0
Glutamatea2.101.00.7*0.7**0.9
Glutaminea2.060.90.7**0.7**1.0
Lactatea1.13+ ++++
Ornithinea1.750.81.01.21.2
PyruvatebNA4.5**5.3**4.7*0.9
Succinatea2.564.2**11.4**7.4**0.7*
Sucrosea5.341.91.11.11.0
Valinea0.991.11.31.6**1.2

Global and local approaches were used to interpret 1H NMR spectra of seedling metabolites. First, the global response was characterized using the unsupervised multivariate method principal components analysis (PCA). The plot of the scores of the first versus the second principal component revealed that the metabolic profiles of the control samples (2NS and 9NS) were well segregated along the first component axis from the 9HS and 1R samples, whereas the 2HS samples fell between the other samples in this analysis (Figure S3a). The resonances of alanine, sucrose, fructose and glutamine were found as primarily responsible for the variance within the dataset (Figure S3b). Second, the 1H NMR spectra and enzymatic assay data were used to assess changes in metabolites associated with carbohydrate consumption, fermentation and nitrogen metabolism during low oxygen stress.

The metabolite analysis demonstrated dynamic reconfiguration of central carbon and nitrogen metabolism in response to oxygen deprivation and reoxygenation. Sucrose, fructose and glucose each had a proton resonance that was well resolved and useful for quantification. However, the univariate analysis did not reveal significant differences in levels of these carbohydrates between treatments and the independent enzymatic quantification confirmed only minor changes (Table 2). Consistent with the prediction from the total and polysomal mRNA data (Figure 4b), there was a four- to five-fold increase in pyruvate (< 0.001). Pyruvate levels were determined enzymatically because the low abundance of this metabolite in the NS sample precluded reliable quantification by 1H NMR. Lactate levels also increased in response to the stress. The level of this metabolite was determined by 1H NMR following spectral subtraction, which was necessary because of the low abundance of lactate in the NS sample and overlap with other resonances. Independent enzymatic measurements of lactate were consistent with the NMR data, but the extremely low basal level of lactate prohibited an accurate estimate of the major increase that occurred during the first 2 h of the stress. Lactate rose a further 3.4-fold after 9HS and then dropped to the 2HS level after 1R. This indicates that lactate is produced in Arabidopsis seedlings during both short and prolonged periods of oxygen deprivation, but is actively metabolized or excreted upon reoxygenation. Alterations in seedling ethanol content were monitored by enzymatic assay, because this compound evaporates during sample preparation for NMR. This analysis confirmed a 1.3- to 1.5-fold increase in ethanol after HS, but this is likely to be an underestimate of the production of this excreted fermentation end-product.

The 1H NMR analysis also confirmed significant stress-induced increases in alanine, GABA, succinate and valine, and significant decreases in glutamine, asparagine, choline and glutamate (Table 2). Changes in these metabolites are indicative of alterations in pyruvate metabolism, due to the restriction of coupled TCA cycle and mitochondrial electron transport activity. Alanine levels climbed to three- and seven-fold after 2HS and 9HS, respectively. The increase in succinate was also progressive, with a three- and six-fold increase after short and prolonged HS, respectively. Of these metabolites, alanine, valine and GABA remained elevated after 1R, whereas succinate declined significantly. Other TCA cycle metabolites were evaluated by 1H NMR, but either their abundance and/or resolution prevented reliable quantification. Overall, this analysis confirmed distinctions in the 1H NMR metabolite profiles of Arabidopsis seedlings following short and prolonged HS, as well as varying readjustment of metabolites following a short period of reoxygenation.

Discussion

Differential mRNA translation conserves energy during low oxygen stress

Oxygen availability dynamically regulates gene expression at both the level of transcript abundance and translation in seedlings of Arabidopsis. Here, the quantitative profiling of alterations in steady-state and polysomal mRNA populations in response to transient HS revealed regulation of abundance of a minority of cellular transcripts, along with a rapid and reversible restriction of initiation of translation of the majority of mRNAs. The major finding is that differential mRNA translation is a key component of the response to oxygen deprivation as well as reoxygenation. The translational repression observed during the stress was not indiscriminate, since ribosome recruitment of the majority of the highly stress-induced mRNAs continued during the stress. The selective inhibition of translation of ∼70% of cellular mRNA during HS has three important consequences. First, the restriction of protein synthesis economizes the consumption of ATP, which is depleted to ∼50% of control values after 2HS and remains at that level for up to 9HS. Second, efficient translation is limited to a small subset of cellular mRNAs, a number of which encode enzymes that facilitate anaerobic production of ATP. We identified 106 mRNAs, accounting for about 0.4% of cellular mRNA, that increase significantly in abundance and association with polysomes during HS (Figure 3, cluster 1). The ability of this cohort of mRNAs to circumvent translational repression ensures the synthesis of the proteins they encode. Third, the reversal of mRNA sequestration upon reoxygenation facilitates rapid recovery of protein synthesis. Thus, differential mRNA translation is a crucial mechanism for the control of gene expression and energy conservation in response to transient oxygen deprivation.

Reconfiguration of central carbon and nitrogen metabolism during low oxygen stress is controlled by transcriptional and post-transcriptional mechanisms

The transcript and metabolite analyses revealed that the reconfiguration of carbon and nitrogen metabolism during HS to promote substrate-level ATP production involves both transcriptional induction and selective translation of mRNAs encoding key enzymes. The observed alterations in total and polysomal mRNA predict that HS promotes catabolism of sucrose via the more energetically favorable sucrose synthase (SUS) pathway, as opposed to the ATP-consuming invertase route (Figures 4b and S4) (Bailey-Serres and Voesenek, 2008; Geigenberger, 2003). There was no marked change in levels of soluble sugars (Table 2). The level of ATP-dependent phosphofructokinase (PFK, At4g32840) mRNA increased in both total and polysomal populations, raising the possibility that increases in this enzyme may augment glycolytic flux through promotion of the conversion of fructose-6-P to fructose-1,6-P2. By contrast, in oxygen-deprived rice this reaction is most likely promoted by the induction of the bidirectional pyrophosphate-dependent phosphofructokinase (PFP), thereby increasing the output of ATP per mol of hexose metabolized (Bailey-Serres and Voesenek, 2008; Lasanthi-Kudahettige et al., 2007).

The data for both the total and polysomal mRNAs predict that HS promotes the production of the fermentation end-products lactate and ethanol [pyruvate decarboxylase (PDC), lactate dehydrogenase (LDH), alcohol dehydrogenase (ADH)] (Figure 4b, Table 2). This was verified by metabolite analyses of the same tissue samples used for the microarray hybridizations. Pyruvate content increased over four- and five-fold following 2HS and 9HS, respectively, and then declined slightly within 1R. This rise coincided with a decline in pyruvate dehydrogenase (PDH) mRNAs in polysomes, as well as an increase in mRNA encoding a PDH kinase, known to inhibit PDH activity (Marillia et al., 2003). Pyruvate was routed to regenerate NAD+ necessary to sustain glycolysis, rather than to the TCA cycle, as evidenced by the rise in the fermentation end-products lactate and ethanol within 2 h after the imposition of HS. The more rapid decline in lactate than ethanol after 1R was consistent with the stronger reduction in LDH than ADH mRNA, in both the total and polysomal mRNA populations (Figure 4b). Since LDH catalyzes a reversible reaction, the increase in NAD+ upon reoxygenation would promote lactate recycling to pyruvate, which could then enter the TCA cycle and alleviate acidification of the cytosol due to lactate accumulation. These data suggest that lactate is produced continuously in the cells of Arabidopsis during HS. It will be important to resolve the spatial and temporal dynamics of lactate production and to further consider the proposed efflux of this toxic metabolite by the HS-induced nodulin intrinsic protein (NIP2;1, At2g34390; Table S2) (Choi and Roberts, 2007).

In addition to the fermentation end-products lactate and pyruvate, HS promoted significant increases in alanine and GABA (Table 2). Elevation of these metabolites has been reported in other studies of the responses of plants to oxygen deprivation (Bailey-Serres and Voesenek, 2008; Bouny and Saglio, 1996; Dixon et al., 2006; van Dongen et al., 2003; Gibon et al., 2002; Miyashita et al., 2007; Ricoult et al., 2006). The change in levels of alanine, sucrose, fructose and glutamine were primarily responsible for the segregation of NS and HS samples by PCA. The 3.3- and 6.5-fold increases in alanine after 2HS and 9HS, respectively, coincided with four-fold elevation of alanine aminotransferase (AlaAT, At1g17290, At1g72330) mRNAs in polysomes under HS (Figure 4b). Alanine aminotransferase catalyzes the interconversion of pyruvate and glutamate to 2-oxyglutarate and alanine. Coupled with this reaction is the conversion of aspartate, by aspartate aminotransferase (AspAT), to the TCA intermediate oxaloacetate. A greater than three-fold increase in one of the three AspAT mRNAs (At5g19550, cluster 1) occurred within 2HS in both mRNA populations, consistent with the decline in glutamate (Table 2). Alanine production is also coupled with that of GABA via glutamine synthase-NAD(P)H-glutamate synthase (GOGAT) and glutamate decarboxylase (GDC). In agreement, HS promoted a significant decline in glutamate and glutamine and an increase in GABA (Table 2). The rise in GABA coincided with a modest elevation of glutamate dehydrogenase (GDH2, At5g07440) mRNA in polysomes (Figure 4b).

γ-Aminobutyric acid is metabolized to succinic semialdehyde (SSA) by GABA transaminase (GABA-T), in a reaction coupled with conversion of 2-oxyglutarate to glutamate (Breitkreuz et al., 2003). Succinic semialdehyde is subsequently metabolized in two distinct ways: SSA is converted to succinate by succinic semialdehyde dehydrogenase (SSADH) in a NAD+-consuming reaction or to γ-hydroxybutyrate (GHB) by GHB dehydrogenase (GHBDH), in a fermentation reaction that would be favored under HS. The mRNAs encoding the enzymes that promote these two routes of GABA metabolism are differentially regulated at the level of transcript accumulation and recruitment to polysomes during HS and reoxygenation. GABA-T (At3g22200) mRNA was strongly reduced under HS, whereas levels of GHBDH mRNA (At3g25530) were stable. Notably, SSADH mRNA (At1g79440) was induced three-fold at the steady-state level after 9HS but this cluster 3 transcript did not increase in polysomes until reoxygenation, at which time the availability of SSADH would be beneficial.

The activity of the TCA cycle is likely to be dampened during HS as a result of re-routing of pyruvate. Here, we found that mRNAs encoding TCA cycle enzymes were translationally repressed during HS but returned to polysomes upon reoxygenation (Figure 4b). The general decline in translation of TCA cycle enzyme transcripts under HS is consistent with decreased dependence on the TCA cycle. However, succinate levels significantly increased in response to short and prolonged HS and declined after 1R (Table 2). An elevation of succinate was reported in response to oxygen deprivation in other higher plants and green algae (Good and Muench, 1993; Vanlerberghe et al., 1990). On the one hand, succinate is likely to accumulate as a consequence of reduced mitochondrial succinate dehydrogenase activity due to inhibition of electron transport at cytochrome c oxidase. However, succinate production may be beneficial for several reasons. Depending upon NAD+ availability, the continuation of oxidative TCA conversion of 2-oxyglutarate to succinate would include substrate-level ATP production by succinyl CoA synthase. A reductive mode of the TCA pathway could operate to metabolize oxaloacetate to malate and ultimately to succinate, as observed in anaerobic yeast (Camarasa et al., 2003). Such reverse TCA metabolism would regenerate NAD+ that can be consumed in substrate-level ATP production via glycolysis or oxidative TCA activity. The AlaAT- and AspAT-mediated production of alanine under HS in Arabidopsis includes conversion of aspartate to oxaloacetate and glutamate to 2-oxyglutarate, providing the intermediates for reductive or oxidative TCA metabolism. Moreover, it has also been proposed that succinate may be utilized at low levels during severe oxygen deprivation in the production of ATP via mitochondrial electron transport when nitrite is available as an electron acceptor (Stoimenova et al., 2007).

Our analysis determined that 1 h of reoxygenation was sufficient to promote dramatic alterations in transcript populations as well as minor metabolic adjustments in seedlings. Levels of ethanol, lactate, GABA and succinate declined, whereas alanine levels rose slightly within the short recovery period. The change in lactate and alanine was concomitant with a reduction of LDH and maintenance of AlaAT mRNAs in polysomes (Figure 4b). Miyashita et al. (2007) reported that AlaAT1 contributes to the recycling of alanine back to glutamate and pyruvate during the recovery of Arabidopsis seedlings from oxygen deprivation. Interestingly, a number of mRNAs increased in abundance under HS but were poorly translated until the seedlings were released from the stress (Figure 3, cluster 3), indicating that plant cells prepare for reoxygenation. As mentioned, SSADH mRNAs increased three-fold during HS but were not recruited to polysomes until reoxygenation. Succinic semialdehyde dehydrogenase is responsible for oxidative metabolism of GABA to succinate. Under HS, SSADH activity would be limited by availability of NAD+; however, upon reoxygenation this enzyme would recycle GABA to succinate. Arabidopsis SSADH loss-of-function mutants are less capable of limiting the accumulation of reactive oxygen species under high light and heat stress (Bouche et al., 2003). Thus, the increase in SSADH mRNA may also contribute to the amelioration of oxidative stress produced during reoxygenation.

Regulation of translation during low-oxygen stress

Although transcriptional reprogramming is crucial to the response to environmental perturbations in eukaryotes, translational regulation contributes to survival by limiting consumption of ATP and directing the synthesis of specific proteins. In plants, in addition to HS, oxidative stress, extremes in temperature, water, light and sucrose availability influence the translation of individual mRNAs (Bailey-Serres, 1999; Branco-Price et al., 2005; Kawaguchi and Bailey-Serres, 2002; Kawaguchi et al., 2004; Nicolaïet al., 2006). Differential mRNA translation in plants is likely to involve mRNA sequence determinants in untranslated regions (UTRs). We reported previously that mRNAs that are efficiently recruited to polysomes during mild dehydration and prolonged low-oxygen stress in Arabidopsis have below average 5′-UTR length, G+C content and limited potential for secondary structure formation (Branco-Price et al., 2005; Kawaguchi and Bailey-Serres, 2005). In animals, oxygen starvation also promotes differential mRNA translation (Koritzinsky et al., 2005;Thomas and Johannes, 2007). This involves a global repression of translation that is driven by phosphorylation of eukaryotic initiation factor (eIF) 2α and eukaryotic elongation factor (eEF) 2 (Koumenis and Wouters, 2006; Koumenis et al., 2002; Liu et al., 2006). Unlike animals and yeast, the phosphorylation of eIF2α is not recognized as a global regulator of translation in plants (Kamauchi et al., 2005; Le et al., 1998). At the level of translational elongation, the phosphorylation of eEF2 during HS in mammals is triggered by the cellular reduction in ATP and increase in AMP (Browne et al., 2004; Patel et al., 2002). The scenario involves phosphorylation of eEF2 kinase by AMP-activated protein kinase (AMPK), which in turn phosphorylates eEF2, resulting in reduced ribosome translocation. A similar mechanism may exist in plants since the phosphorylation of eEF2 inhibits translational elongation in a cell-free wheat germ system (Smailov et al., 1993). Plants lack AMPK but possess a Snf-related protein kinase family of which Arabidopsis KIN10/11 have been reported to regulate transcription in a manner dependent upon carbon and energy status (Baena-Gonzalez et al., 2007); however, these kinases are not known to contribute to translational regulation.

Hypoxia stress also triggers translational regulation in mammalian cells via the nutrient and stress-sensing target of rapamycin (TOR) (Liu et al., 2006). The single essential TOR ortholog of Arabidopsis is highly expressed in actively proliferating cells (Menand et al., 2002; Sormani et al., 2007). However, the embryo lethality of TOR mutants and insensitivity of plants to rapamycin has made it difficult to use genetic or chemical approaches to dissect the function of TOR (Mahfouz et al., 2006; Menand et al., 2002; Sormani et al., 2007). Plants possess the TOR regulatory protein (RAPTOR) and p70 S6 kinase (S6K) orthologs that form a functional complex that controls the phosphorylation status of ribosomal protein S6 (RPS6) (Arsham and Neufeld, 2006; Mahfouz et al., 2006; Turck et al., 1998, 2004). In mammals, the activation of AMPK during HS also negatively regulates mTOR and thereby eIF4E activity (Arsham and Neufeld, 2006; Koritzinsky et al., 2006). The consequence is that only mRNAs with limited dependence on the cap-binding complex can be efficiently translated during HS. Plants possess two cytosolic cap-binding proteins, eIF4E and eIFiso4E, that differ in abundance and preference for binding mRNAs with structured 5′-UTRs (Kawaguchi and Bailey-Serres, 2002; Monzingo et al., 2007). Structural analysis of wheat eIF4E identified two novel cysteines that form an intermolecular sulfhydryl group that may allow the cellular redox status to modulate the mRNA cap-binding activity of both eIF4E and eIFiso4E (Monzingo et al., 2007).

Another consequence of the reduction in TOR activity during HS in mammalian cells is reduced activity of S6K (Liu et al., 2006). Inactive S6K competitively inhibits the joining of eIF3 to the pre-initiation complex, contributing to the global reduction in protein synthesis (Holz et al., 2005). The activation of S6K is strongly correlated with cell size and translation of RP and other mRNAs with a 5′ terminal oligopyrimidine (5′-TOP) sequence. It was thought that S6K activation regulated these processes through the phosphorylation of serine residues at the carboxy terminus of RPS6. However, transgenic mice that cannot phosphorylate RPS6 due to the absence of S6K or mutations in RPS6 provide strong evidence that this event is not required for translation of 5′-TOP mRNAs (Proud, 2007). In plants, HS, heat and cold stress reduce the carboxy terminal phosphorylation of RPS6 (Turck et al., 1998; Williams et al., 2003). Arabidopsis encodes two functional orthologs of S6K, S6K1 (At3g08730) and S6K2 (At3g08720). Intriguingly, we show here that HS elicits a four- to six-fold increase in the abundance of Arabidopsis S6K2 mRNA, but this cluster 3 transcript is not efficiently recruited to polysomes until reoxygenation, at which time levels of S6K1 and S6K2 mRNAs increased 2.6 to 7-fold, respectively, in polysomes (Figure 4a, Table S6). It can be speculated that the elevation of these transcripts prior to reoxygenation along with their restricted translation is beneficial to the recovery from the stress. Steady-state levels of S6K1 and/or S6K2 mRNA were reported to increase significantly under salt, heat and osmotic stress (Mizoguchi et al., 1995), but the translational status of these mRNAs was not considered. Mahfouz et al. (2006) reported that overexpression of S6K1 increases the sensitivity of Arabidopsis seedlings to osmotic stress, a condition in which S6K activity is normally reduced.

The data presented here support the hypothesis that endurance of the energy crisis caused by HS is augmented by the selective regulation of translational initiation. This could be the direct consequence of limited availability of ATP or mediated by altered activity of proteins involved in initiation of translation, ranging from the cap-binding proteins to S6K. Messenger RNAs with a 5′-UTR or low G+C content and limited secondary structure may be favored for translation during HS, due to a lower requirement for ATP in the initiation phase. However, since these attributes are not necessary or sufficient for efficient translation during HS (Branco-Price et al., 2005), it is likely that additional factors enable the efficient translation of a small cohort of mRNAs during HS. Remarkably, the majority of the translationally impaired mRNAs are stabilized in non-polysomal complexes and rapidly recruited to polysomes upon reoxygenation. Thus, conservation of cellular energy is further enhanced by minimization of the need for de novo transcription upon reoxygenation. Future studies should clarify the interplay between energy status and the sequestration versus recruitment of individual mRNAs for translation during low-oxygen stress.

Experimental procedures

Detailed experimental procedures are provided in the on-line Supporting Information.

Plant material and stress treatments

Seeds from single-copy transgene 35S:His6FLAG-RPL18B (Col-0) (Zanetti et al., 2005) were surface sterilized, stratified at 4°C for 3 days, plated onto 10-cm diameter Petri dishes with solid 1× MS media additionally containing 0.4% Phytagel (Sigma, http://www.sigmaaldrich.com/), 1% (w/v) sucrose (approximately 100 seeds per plate) and positioned vertically in a growth chamber (Percival Scientific, Inc., http://www.percival-scientific.com/; model CU36L5C8) under long-day conditions (16 h at 50 μm sec−1 m−2 light: 8 h dark) at 23°C. Seven-day-old seedlings were subjected to 2 or 9 h of NS or HS treatment, or 9 h HS followed by return to ambient air for 1 h (reoxygenation). Treatments were performed in dim light and commenced at the end of the 16-h light cycle under 6.5 μm sec−1 m−2 light, in open (NS) or sealed (HS) chambers. For HS, 99.995% argon gas was sparged through water and into the chamber while ambient air was pushed out by positive pressure (Branco-Price et al., 2005), requiring approximately 1.5 h for the air to be displaced from the Petri dishes. This treatment deprives seedlings of oxygen and carbon dioxide, limiting both aerobic respiration and photosynthesis. Seedlings were frozen under liquid nitrogen within 3 min of release from treatment and stored at −80°C until use.

Total and polysomal RNA isolation

Ribosomes were immunoprecipitated using 5 ml of packed tissue, extracted in 10 ml of Polysome Extraction Buffer and 13 ml of the clarified crude extract was incubated with 300 μL of EZview® Red ANTI-FLAG® Affinity Gel (Sigma) as described by Zanetti et al. (2005). Total RNA was extracted from 1 ml of the clarified crude extract. Both the immunoprecipitation eluate and the total RNA pellet were further purified by use of the RNeasy Plant Mini Kit (Qiagen, http://www.qiagen.com/).

Microarray hybridization and data analyses

Total and polysomal RNA samples derived from three independent biological replicate experiments were used for hybridization to Arabidopsis ATH1 Genome Arrays (GeneChip System®; Affymetrix) (see Appendix S1). Data were processed using Robust Multichip Average (RMA) and Affymetrix mas 5.0. The signal data obtained from the RMA analyses of the polysomal RNA were further adjusted based on the relative polysome content measured in three replicate experiments (2 h NS = 0.75, 9 h NS = 0.75; 2 h HS = 0.43, 9 h HS = 0.36; 1R = 0.67). Differential expression analyses (comparisons between treatments) were evaluated for significance by Linear Models for Microarray Data (Limma). False discovery rate (FDR) was generated using P-value distributions (Smyth et al., 2004). Normalized SLR data of mRNAs detected with ‘present’ calls in all 30 hybridizations (= 8863) were used as the dataset for all the subsequent cluster and gene category analyses.

Enzymatic quantification of metabolites

Assays of ATP, ADP, glucose, fructose, sucrose, lactate, pyruvate and ethanol were by standard enzymatic procedures using metabolites stabilized by perchloric acid extraction of tissue (Stitt et al., 1989) (see Appendix S1). All evaluations were performed on three or more biological replicate experiments.

NMR metabolite analyses

Frozen pulverized whole seedling tissue was homogenized using a micropestle in 700 μl of extraction buffer [1:1 (v/v) acetonitrile-d3:deuterium oxide containing 50 mm sodium acetate-d3, 50 m acetic acid-d4, 1 mm 3-trimethylsilylpropionic acid-d4 sodium salt (TMSP), as a chemical shift reference (0.000 p.p.m.), pH 4.73 with deuterium chloride]. The pH of the resulting extract was 4.54 ± 0.08. The extract was clarified by centrifugation at 2300 g and 600 μl was transferred to a 5-mm NMR tube (Wilmad, http://www.wilmad-labglass.com/). The NMR spectra were acquired for bioreplicate samples with selective saturation of the solvent resonance using a Bruker Advance spectrometer operating at 600.06 MHz. Spectra were integrated for the purpose of metabolomics using equidistant integral regions of 0.02 p.p.m. width over the range of 0.50–9.00 p.p.m., excluding the regions containing the resonances of solvent constituents: acetate and acetonitrile (1.96–2.04 p.p.m.), and residual water (4.28–4. 72 p.p.m.). For metabolite profiling a second set of integration regions were manually defined for the quantification of well-resolved peaks. In both cases integral regions were normalized to the total integrated area of each spectrum. Spectral assignments were made by comparison with spectra of metabolite standards (Sigma). Fold changes were calculated by the use of standards, literature cross-referencing and two-dimensional 1H NMR methods (data not show) (Cui et al., 2008; Fan, 1996; Govindaraju et al., 2000). Quantitative fold changes were calculated for metabolites having at least one well-resolved resonance in the spectrum. This was accomplished by manually dividing the NMR spectra into regions of variable width, so that each region contained a single resonance of a metabolite. Spectral editing was carried out to estimate fold changes for metabolites found in regions where resonance overlap precluded direct integration. Global changes in metabolites were evaluated by the use of PCA.

Acknowledgements

This work was supported by National Science Foundation grants (nos IBN-0420152 and CHE-0616811) to JBS and CKL, respectively; National Science Foundation Integrative Graduate Education Research and Training Program fellowships (DGE-0504249) to KAH and CJHJ; and a Portuguese Foundation for Science and Technology graduate fellowship (SRFH/BD/9165/2002) to CB-P.

Ancillary