Differential Response of Gray Poplar Leaves and Roots Underpins Stress Adaptation during Hypoxia^,

Abstract

The molecular and physiological responses of gray poplar (Populus × canescens) following root hypoxia were studied in roots and leaves using transcript and metabolite profiling. The results indicate that there were changes in metabolite levels in both organs, but changes in transcript abundance were restricted to the roots. In roots, starch and sucrose degradation were altered under hypoxia, and concurrently, the availability of carbohydrates was enhanced, concomitant with depletion of sucrose from leaves and elevation of sucrose in the phloem. Consistent with the above, glycolytic flux and ethanolic fermentation were stimulated in roots but not in leaves. Various messenger RNAs encoding components of biosynthetic pathways such as secondary cell wall formation (i.e. cellulose and lignin biosynthesis) and other energy-demanding processes such as transport of nutrients were significantly down-regulated in roots but not in leaves. The reduction of biosynthesis was unexpected, as shoot growth was not affected by root hypoxia, suggesting that the up-regulation of glycolysis yields sufficient energy to maintain growth. Besides carbon metabolism, nitrogen metabolism was severely affected in roots, as seen from numerous changes in the transcriptome and the metabolome related to nitrogen uptake, nitrogen assimilation, and amino acid metabolism. The coordinated physiological and molecular responses in leaves and roots, coupled with the transport of metabolites, reveal important stress adaptations to ensure survival during long periods of root hypoxia.

Higher plants are aerobic organisms and depend on the availability of O₂. A lack of O₂ in the rhizosphere affects the maintenance of numerous pathways and is therefore an important environmental stress for vascular plants (Drew, 1997). Plant adaptations to O₂ deprivation include avoidance strategies at the morphological level and physiological tolerance mechanisms (Bailey-Serres and Voesenek, 2008). One of the major cellular pathways dependent on O₂ is mitochondrial respiration. In order to maintain energy generation under conditions of decreased O₂ availability, plants switch from respiration to fermentative metabolism. Fermentation allows regeneration of NAD⁺ in the absence of respiration, thereby maintaining glycolysis and the generation of ATP under anaerobic conditions. As an initial reaction to O₂ deprivation, many plants activate lactic acid fermentation. As generation of lactic acid causes a decrease in cytosolic pH (Roberts et al., 1984), which reduces the activity of the responsible enzyme, lactate dehydrogenase (LDH; Hanson and Jacobson, 1984), lactic acid fermentation is followed by alcoholic fermentation (Davies et al., 1974; Roberts et al., 1984). The significantly lower energy yield of alcoholic fermentation, compared with mitochondrial respiration, causes an energy crisis in anaerobic tissues (Bailey-Serres and Voesenek, 2008).

A high rate of fermentation increases the demand for carbohydrates, leading to the hypothesis that carbohydrate supply becomes critical for survival under prolonged hypoxic conditions. This assumption is supported by the following: (1) improved survival of flood-sensitive species by exogenous supply of sugars (Waters et al., 1991; Perata et al., 1992; Loreti et al., 2005); (2) activation of glycolytic enzymes by hypoxia (Setter et al., 1997; Liu et al., 2005; Loreti et al., 2005); and (3) decreased flooding tolerance when the synthesis of these enzymes is inhibited (Subbaiah et al., 1994). Although carbon and energy metabolism have been identified to play crucial roles for the physiological adaptation of plants to hypoxia, it has not yet been clarified what mechanisms ensure the steady supply of carbohydrates to hypoxic tissues of tolerant plant species. Furthermore, the changes observed in response to O₂ deficit are caused and accompanied by altered gene expression patterns, such as enhanced transcription of genes required for fermentation and morphological adaptations (Bailey-Serres and Voesenek, 2008). The latter include genes involved in ethylene biosynthesis, programmed cell death, ethylene signal transduction, and cell wall lysis (Klok et al., 2002). Moreover, transcript levels of enzymes required for biosynthetic processes such as cell wall synthesis and flavonoid biosynthesis are reduced (Klok et al., 2002; Geigenberger, 2003; Branco-Price et al., 2008; Van Dongen et al., 2008).

Although the morphology and physiology of plants grown under anaerobic conditions have been studied in detail (including the targeted analysis of specific metabolites), limited work has been done to elucidate adaptations at the molecular level. While some microarray studies have been performed (Klok et al., 2002; Branco-Price et al., 2005, 2008; Liu et al., 2005; Loreti et al., 2005; Lasanthi-Kudahettige et al., 2007; Van Dongen et al., 2008), the regulation of anaerobically induced genes and the integration of metabolic responses in hypoxic organs with those of attached normoxic feeder organs have not been comprehensively studied at the molecular level. Furthermore, there have been no previous reports of the differences between trees and annual species, such as Arabidopsis (Arabidopsis thaliana), with respect to their responses to O₂ deficit. The aim of this study was to gain a comprehensive insight into plant adaptation mechanisms to root hypoxia, with a focus on carbon metabolism, using the flooding-tolerant species gray poplar (Populus × canescens). Gray poplar is a hybrid of Populus alba and Populus tremula that naturally occurs in regularly flooded areas of alluvial forests along several European rivers (Fossati et al., 2004; Lexer et al., 2005; van Loo et al., 2008). Its high flooding tolerance has also been demonstrated in a series of physiological studies (Kreuzwieser et al., 1999, 2002). In this work, a global analysis of gene expression patterns was performed using the Affymetrix poplar genome array representing 56,055 transcripts. This approach was complemented with metabolomic analysis and physiological measurements in order to assess whether observed gene expression changes were likely to be responsible for the plant's adaptation to anaerobic conditions.

RESULTS AND DISCUSSION

Validation of the Experimental Setup

Induction of root hypoxia caused an onset of ethanolic fermentation in roots of poplar trees, as demonstrated by increases in transcript levels of Pyruvate Decarboxylase (PDC; protein identifier 560932) and Alcohol Dehydrogenase (ADH; protein identifier 643524; Fig. 1A). This finding was supported by the observed 3-fold increase in ADH activities (Fig. 1A). Although the mean PDC transcript level was 15 times higher after 5 h of hypoxia, this transcriptional up-regulation was only associated with a 3-fold increase in PDC activity in poplar roots. Possible explanations for this difference include reduced translation efficiency (Branco-Price et al., 2008), decreased protein stability, or some other form of posttranscriptional or posttranslational regulation of enzyme activity. The observed relatively high basal activities of PDC in normoxic controls have also been discussed as a reason for such observations (Bouny and Saglio, 1996; Tadege et al., 1998). In contrast to roots, leaves did not show altered PDC and ADH expression. However, leaf ADH activity increased as a consequence of hypoxic treatment of the roots. While this observation indicates that the expression of an additional ADH isoform may be enhanced in leaves, the microarray data do not provide evidence for this. Increased ADH activity in leaves of flooded trees may be required to cope with the xylem-transported ethanol produced in the roots (Kreuzwieser et al., 2001), as suggested from considerable emission of the volatile oxidation product of ethanol, acetaldehyde. Acetaldehyde emission from leaves increased from 22 ± 9 nmol m⁻² min⁻¹ (controls) to 423 ± 158 and 282 ± 164 nmol m⁻² min⁻¹ after 24 and 168 h of hypoxia, respectively, consistent with previous studies (Kreuzwieser et al., 1999, 2001). Obviously due to the transport of ethanol from roots to leaves, the alcohol did not accumulate to any great degree in the roots, with root ethanol concentrations remaining below 4 μmol g⁻¹ fresh weight (Figs. 2 and 3). In contrast, ethanol concentrations in the xylem sap increased by a factor of 200 after 24 h of flooding.

Figure 1.

Effects of hypoxia on expression and activity of fermentative enzymes (A) and general transcript abundance (B) in poplar roots and leaves. A, Transcript levels and activities of PDC (left) and ADH (right) in leaves (top row) and roots (bottom row) of gray poplar. Transcript levels of normoxic controls (black squares) and flooded plants (red circles) are normalized to β-tubulin mRNA levels. Means ± se of the enzyme activities (bars) from four to six independent experiments are shown. Statistically significant differences at P < 0.05 were calculated by Student's t test and are indicated by asterisks. B, Number of differentially expressed genes (≥2-fold changes; q < 0.05) following root hypoxia as determined by microarray analysis. Data represent averages of three biological replicates for each time point/tissue. Applying the criteria given in “Materials and Methods” for leaves, no differentially expressed genes were observed.

Figure 2.

Effects of hypoxia on metabolite abundance in leaves and roots of poplar. Normoxic controls and hypoxically treated trees were harvested and metabolite concentrations were determined by GC-MS (roots) at 5, 24, and 168 h and by HPLC (leaves) at 168 h. Ethanol concentrations were determined enzymatically. Metabolites listed in red showed significantly lower concentrations in treated trees than in controls. Blue and green metabolites indicate higher and unchanged concentrations, respectively. Diagrams behind metabolite names indicate fold changes (log₂ converted) of results obtained by GC-MS.

Figure 3.

Carbohydrate (A), ethanol (B), and amino acid (C) concentrations in leaves and roots of poplar affected by root hypoxia. Normoxic and hypoxically grown trees were harvested and metabolite concentrations were determined. Results shown are means ± se from at least four trees per time point and treatment. Statistically significant differences between treatments per carbohydrate species at P < 0.05 were calculated with the Tukey test under ANOVA (A) or Student's t test (B and C) and are indicated by different letters/asterisks above the bars. Cit, Citrulline; f.wt., fresh weight; 3-m-his, 3-methyl-His.

Transcript Profiling

Based on these observations, microarray analysis was performed using root samples taken after 0, 5, 24, and 168 h of flooding and leaf samples taken at 0 and 168 h after treatment. To assess if the microarray data were comparable to the highly sensitive quantitative reverse transcription (qRT)-PCR approach, the results for 12 genes analyzed by both techniques were examined (Supplemental Fig. S1). The high correlation (r² = 0.83) of these results validated the microarray analysis and confirmed that the normalization method employed provided reliable results.

Exposure to root hypoxia caused altered transcript levels of over 5,000 genes in roots during the experiment (Fig. 1B; Supplemental Table S1). Interestingly, the number of genes with more than 2-fold changed (q < 0.05) transcript abundance strongly differed between roots and leaves. No genes in the leaves showed different expression levels after 1 week of root hypoxia, indicating that the response of poplar to O₂ deficiency involves adaptations mainly in the stressed tissue itself. This finding is surprising, considering that root hypoxia led to reduced abundance of mRNA coding for some low molecular weight proteins in leaves of flooding-tolerant Populus trichocarpa × Populus deltoides (Neuman and Smit, 1993). The reason for this discrepancy might be either the different species studied or the very stringent conditions applied in that study, in which only genes showing a statistically significant difference of at least 2-fold (q < 0.05) were considered. It should be mentioned that unaffected transcript abundance does not necessarily mean unchanged protein abundance, as seen from the work of Branco-Price et al. (2008) with anoxia-treated Arabidopsis. Those authors clearly demonstrated that translation efficiencies can selectively be altered independently of transcription. Changed metabolite concentrations in poplar (Figs. 2 and 3) may be a result of this phenomenon or may be caused by modified xylem or phloem transport. In root tissue, the transcript abundance of 5,250 genes was significantly altered. After 5 h of root hypoxia, only 195 genes showed altered expression, of which more than 85% were up-regulated. The expression of 147 and 144 of these genes was still altered after 24 and 168 h of flooding, respectively. Prolonged duration of flooding caused considerably more genes to show altered expression, with 3,728 and 5,066 genes differing in their expression compared with controls after 24 and 168 h of root hypoxia, respectively. In contrast to the early response to hypoxia, approximately two-thirds of these genes were found to be down-regulated. This is consistent with the gene expression response in another flooding-tolerant species, rice (Oryza sativa), in which 1,364 and 1,770 probe sets showed increased and decreased expression, respectively (Lasanthi-Kudahettige et al., 2007).

Differentially expressed genes in roots were clustered into hierarchical dendrograms, resulting in the genes being classified into 15 clusters according to their expression profiles (Fig. 4; Supplemental Table S1). About 70% of all differentially expressed genes were classified into the main clusters 4, 7, 13, and 14; the largest cluster, 14, alone contained 26% of all differentially expressed genes. Genes in clusters 4 and 7 were up-regulated in the roots during hypoxia. In contrast, members of clusters 13 and 14 exhibited down-regulated gene expression in roots in response to hypoxia; genes in cluster 14 were less strongly reduced than genes in cluster 13. Comparable to the Arabidopsis study by Liu et al. (2005), genes in the roots were either up- or down-regulated during the whole course of the experiment, and expression patterns changing from up- to down-regulation or vice versa were scarce.

Figure 4.

Hierarchical clustering of changes in transcript abundances following flooding. Following microarray analysis, fold changes were determined from three biological replicates for each time point/tissue type. A, Hierarchical clustering was performed using the TMeV software package from log₂ signal ratio data. To distinguish between clusters, violet triangles are drawn to the left of clusters and cluster numbers are given. B, Color scale indicating signal log₂ ratios. The criteria given in “Materials and Methods” were applied, and only genes with greater than 2-fold changed expression (q < 0.05) are displayed.

The genes with significant (q < 0.05) and more than 2-fold changes in expression in the roots were classified into functional groups using the MapMan Bin code (Thimm et al., 2004; Supplemental Table S1) and the software Pageman (Usadel et al., 2006; Supplemental Figs. S2 and S3). More than one-third of these genes had unassigned functions. A relatively high percentage (approximately 7%) seemed to play a direct role in metabolism or to be involved in transport processes; the latter genes were mostly reduced in expression (Supplemental Fig. S2). The high proportion of genes involved in the regulation of transcription, signaling, and hormone metabolism and the numerous genes with changed expression coding for protein kinases indicate the dynamic adjustment of metabolic processes taking place in the stressed plants. Similar changes have also been reported in other studies on the transcriptome of plants stressed by O₂ deficiency (Klok et al., 2002; Liu et al., 2005; Loreti et al., 2005; Lasanthi-Kudahettige et al., 2007; Branco-Price et al., 2008; Van Dongen et al., 2008).

Effects on Carbon Metabolism

The switch from mitochondrial respiration to fermentation is likely to strongly affect energy and carbon metabolism. Indeed, some important genes involved in glycolysis were up-regulated in roots due to hypoxia (Fig. 5). The transcript abundance of key enzymes of this pathway (Plaxton, 1996), phosphofructokinase (protein identifier 695057) and pyruvate kinase (protein identifier 698714), were significantly increased due to hypoxia. As expected and verified by qRT-PCR data, transcript abundance of genes mediating alcoholic fermentation increased significantly due to root hypoxia. In good agreement with the expression profile of LDH in some herbaceous species (e.g. Arabidopsis [Klok et al., 2002] and maize [Zea mays; Christopher and Good, 1996]), the LDH transcript in poplar (protein identifier 704268) increased in abundance in the initial stages of root hypoxia (5 h of treatment) but thereafter dropped again. Consistent with this, the lactate concentration in flooded roots was six times higher than in controls after short-term flooding but thereafter decreased to control levels. This pattern was also seen in pyruvic acid concentrations that initially increased by a factor of 3 (5 h of flooding) but then dropped to control levels (Fig. 2; Supplemental Table S2). Similarly increased pyruvate and lactate abundance was recently observed in anoxia-treated Arabidopsis (Branco-Price et al., 2008). These observations support the pH-stat hypothesis (Roberts et al., 1984) and show that in poplar, induction and inhibition of LDH can occur not only at the physiological level but also at the molecular level.

Figure 5.

Expression changes of genes involved in primary carbon metabolism. After 5 h (left squares), 24 h (middle squares), and 168 h (right squares) of hypoxia, roots were harvested and transcript levels determined by microarray analysis. The log₂ values of fold changes of genes involved in carbon metabolism are displayed using the color code indicated. Protein identifiers are indicated. This figure was generated using MapMan (Thimm et al., 2004). Green arrows indicate gluconeogenesis, and orange arrows indicate the glyoxylate cycle. Details on gene expression are given in Supplemental Table S2.

In this study, the higher demand for carbohydrates required for glycolysis in hypoxia-treated plants was most probably compensated for by increased phloem transport of Suc from leaves to roots, as suggested by the higher Suc concentrations in the phloem of hypoxically treated plants (2.14 ± 0.44 μmol g⁻¹ bark fresh weight) than in controls (1.43 ± 0.32 μmol g⁻¹ bark fresh weight) and, simultaneously, significant decreases in Suc levels in leaves (Figs. 2 and 3). In the roots, altered transcript levels clearly indicated a switch in Suc degradation from cleavage by invertases (protein identifiers 711782 and 574265) to phosphorolytic degradation by Suc synthases (protein identifier 692288; Fig. 5; Supplemental Table S4). Moreover, it may be that the observed up to 10-fold decrease in pyrophosphatase transcript levels (protein identifiers 748006, 704742, 707491, 551881, and 664206) caused decreased abundance and activity of this enzyme, resulting in increased levels of pyrophosphate, the substrate for Suc synthase. Progressive induction of pyrophosphate-dependent phosphofructose kinase (protein identifiers 704902 and 550714) provides a further link between enhanced glycolysis and pyrophosphate during root hypoxia (Supplemental Table S1). The observed switch in Suc degradation, which has also been observed in several other studies at the physiological and molecular levels (Guglielminetti et al., 1995; Zeng et al., 1998, 1999; Liu et al., 2005; Loreti et al., 2005; Branco-Price et al., 2008; Van Dongen et al., 2008), is energetically advantageous, as only one pyrophosphate is used by Suc synthase compared with two ATP molecules needed by invertases. Starch biosynthesis and degradation were also found to be affected by hypoxia (Fig. 5). Starch production seemed to be slowed as transcript levels of the starch-branching enzyme (protein identifier 589574) and starch synthase (protein identifier 569276) decreased. In contrast, starch degradation was most probably altered because of an observed increase in β-amylase expression (protein identifier 679498) and decreases in α-amylase (protein identifier 577743) and α-xylosidase (protein identifier 564524) expression at the same time.

Consistent with these changes in transcript levels, some carbon metabolite concentrations in leaves and roots were altered during hypoxia (Figs. 2 and 3). As suggested from accelerated Suc and starch catabolism, Glc, Fru, and Fru-6-P concentrations increased in roots, while starch concentration showed a slight, but not significant, decrease after 1 week of hypoxia. Increased carbohydrate levels in the roots of flooded plants have been previously observed in trees and herbaceous species (Albrecht et al., 1993; Castonguay et al., 1993; Huang and Johnson, 1995) and have been explained by a reduced carbon demand due to a decline in root growth and nitrogen metabolism (Angelov et al., 1996). In contrast, the results of this study suggest that improved sugar supply to hypoxic roots occurs in order to compensate for increased carbohydrate demand under these conditions. Interestingly, reduced carbohydrate concentrations have been reported in roots of hypoxically treated flood-sensitive species (Vu and Yelenosky, 1991), highlighting the importance of sugar delivery to roots in tolerating periods of O₂ deficiency.

An inhibition of respiration under conditions of reduced O₂ supply has been observed in herbaceous plants (Geigenberger, 2003). As this inhibition occurs at O₂ concentrations higher than the K_m of cytochrome oxidase, it has been proposed that under hypoxic conditions respiration is slowed in order to reduce O₂ consumption, thereby avoiding anoxia (Geigenberger, 2003). However, in this study, transcript abundance of most respiratory chain components was found to be unaffected by root hypoxia using qRT-PCR (Supplemental Fig. S4) and microarray approaches (Supplemental Table S1), suggesting (1) that any inhibition of respiration occurs at the protein or enzyme activity level, or (2) as also concluded from work on Arabidopsis (Branco-Price et al., 2008), that substrate availability for the tricarboxylic acid (TCA) cycle and respiration is limited because of increased flux in the direction of fermentation.

The expression of most genes encoding enzymes of the TCA cycle was unaffected or only slightly reduced by root hypoxia (Fig. 5), supporting the observations of Loreti et al. (2005) and Branco-Price et al. (2008) working with Arabidopsis. The gene showing the most strongly reduced expression was a NADP-dependent isocitrate dehydrogenase (IDH; protein identifier 55426), which dropped almost 15-fold after 1 week of hypoxia (Fig. 5). This isoform showed high similarity to cytosolic (At1g65930) and mitochondrial (At1g54340) IDH isoforms of Arabidopsis. In contrast to IDH, the transcript abundances of Isocitrate Lyase (ICL; protein identifier 741351) and Malate Synthase (MS; protein identifiers 57656 and 570219) increased, indicating an enhanced glyoxylate cycle under hypoxia. Increases in succinate concentrations in roots of hypoxia-treated poplar, which have also been demonstrated in anoxia-treated Arabidopsis (Branco-Price et al., 2008), are consistent with this proposal (Fig. 2; Supplemental Table S2). The glyoxylate cycle exerts a gluconeogenic function linking lipid degradation and carbohydrate production (Eastmond et al., 2000) but may also play an anaplerotic role, as seen in microorganisms (Kornberg and Krebs, 1957). It is interesting that ICL and MS activities are also enhanced in carbon-starved plants (Kim and Smith, 1994), a situation that may exist in plants experiencing O₂ deficiency (Loreti et al., 2005). In this case, an involvement of the glyoxylate cycle in gluconeogenesis is more likely than an anaplerotic role (Smith, 2002). Through the glyoxylate cycle, excess acetyl-CoA resulting from β-oxidation of fatty acids would be channeled into carbohydrate biosynthesis. Stimulated degradation of fatty acids is indeed suggested by the 5-fold up-regulation in expression of acyl-CoA synthetase (protein identifiers 709380 and 709180; Supplemental Table S1), an enzyme involved in this pathway in peroxisomes (Fulda et al., 2002). Consistent with this observation, the biosynthesis of fatty acids seemed to be repressed, as the transcript levels of S-malonyl transferase (protein identifier 722565), acetyl-CoA carboxylase (protein identifiers 684485, 559199, and 673504), 3-ketoacyl-ACP synthase (protein identifiers 554288, 714498, and 691067), and enoyl-ACP reductase (protein identifiers 576267 and 684702) dropped during hypoxia (Supplemental Table S1).

Inhibition of Energy-Consuming Processes

Numerous energy-demanding processes were repressed in hypoxically treated poplar (Supplemental Figs. S2 and S3). For example, the transcript abundance of many transporters strongly decreased (Supplemental Table S1). Moreover, the increased concentrations of the hemicellulose components Ara and Gal (Fig. 2; Supplemental Table S2) in flooded poplar roots indicated that changes in cell wall formation and/or degradation occurred. Consistently, changes in transcript levels suggested that the biosynthesis of cellulose, hemicellulose, and cell wall proteins was inhibited (Supplemental Fig. S2). Eight genes with homology to Arabidopsis cellulose synthases were found in poplar. In the roots, the expression of cellulose synthases required for secondary cell wall formation was dramatically reduced (up to 44-fold; protein identifier 555650). In contrast, the expression of only one cellulose synthase gene (protein identifier 714760) involved in primary cell wall biosynthesis declined, and the extent of decrease (up to 4.4-fold) was much less pronounced. These results may indicate the importance of the primary cell wall for cellular integrity, whereas formation of the secondary cell wall seems to be of minor importance and may be initiated after survival of the stress period. However, cellulose degradation was also affected, as the expression of most root-specific cellulase isozymes was strongly reduced due to hypoxia (e.g. protein identifier 651956; 17-fold; Supplemental Table S4).

Supplemental Table S5 summarizes the expression data of the relevant genes of lignin biosynthesis together with other genes of the phenylpropanoid metabolism that were present on the microarray. In the roots, all steps of this pathway, starting from the conversion of chorismate to prephenate (catalyzed by chorismate mutase), to the production of Phe from arogenate (by arogenate dehydratase), to its conversion to cinnamate (by Phe ammonia lyase), seemed to be down-regulated. For example, the transcript levels of root-specific Phe ammonia lyase isozymes decreased by 7.6-fold (protein identifier 696959) and 11.5-fold (protein identifier 82117). In addition, the expression of the root-specific chalcone synthase, a key enzyme of phenylpropanoid metabolism, was reduced by 8-fold (protein identifier 572875). Moreover, the subsequent modifications of cinnamate leading finally to the formation of lignin and flavonoids were repressed.

Effects on Nitrogen Metabolism

Nitrogen metabolism was studied in this work by analyzing amino acid concentrations in leaves and roots after 168 h of hypoxia (Figs. 2 and 3), the progression of metabolic changes in roots up to 168 h of hypoxia (Supplemental Table S2), and transcript levels of enzymes involved in nitrogen metabolism (Supplemental Table S6). As expected, NO₃⁻ uptake was most probably impaired by root hypoxia, as the transcript levels of some NO₃⁻ transporters (protein identifiers 551887 and 591881) decreased up to 17-fold. This is in good agreement with previous observations of reduced nitrogen uptake in poplar (Kreuzwieser et al., 2002) and other species (Drew, 1991). The situation for NH₄⁺ uptake was different, as two transporters were up-regulated. Considering slightly reduced NH₄⁺ uptake by flooded poplar (Kreuzwieser et al., 2002), up-regulated transporters could be involved in root internal transport, such as NH₄⁺ loading into the xylem. Consistent with the assumption of reduced nitrogen uptake into poplar roots, the gene expression of enzymes involved in nitrogen assimilation seemed to be down-regulated. In roots, NH₄⁺ assimilation was transcriptionally affected, as Gln synthetase isozymes were strongly down-regulated, a finding consistent with considerably reduced concentrations of Gln (Fig. 2; Supplemental Table S2). The subsequent NH₄⁺ assimilation proceeds via GOGAT (Lea et al., 1990). In plants, an NADH-dependent and a ferredoxin (Fd)-dependent GOGAT isoform exist (Suzuki and Gadal, 1984). In accordance with results of anaerobically germinating rice (Mattana et al., 1996), Fd-GOGAT expression (protein identifier 695596) was increased whereas expression of NADH-GOGAT (protein identifier 569758) was reduced (Supplemental Table S4). Notably, the opposite was the case in Arabidopsis (Liu et al., 2005). Fd-GOGAT seems to be essential for reassimilation of NH₄⁺ arising from the catabolism of nitrogen-containing compounds such as proteins (Aurisano et al., 1995; Mattana et al., 1996), which is supported by decreased protein levels in flooded poplar roots (Kreuzwieser et al., 2002). Another interesting effect was the 6-fold increase in Glu decarboxylase transcripts after 5 h of hypoxia (Supplemental Table S6), which was observed together with reduced levels of Glu but increased levels of γ-aminobutyrate (GABA), succinate, and Ala (Fig. 2; Supplemental Table S2), indicating onset of the GABA shunt (Fig. 2).

Metabolite profiling revealed that flooding induced rapid changes in the levels of a wide variety of amino acids in the roots. As seen also in Arabidopsis (Branco-Price et al., 2008), most of the more strongly accumulating amino acids were closely derived from either pyruvate (e.g. Ala, Val, Leu) or intermediates of glycolysis (Gly, Ser, Tyr), while all of the decreasing amino acids (Glu, Gln, Asp, Asn) were derived from TCA cycle intermediates. This observation strongly supports the hypothesis that the block in aerobic respiration caused by hypoxia led to a decrease in flux into the TCA cycle (due to an accumulation of NADH and depletion of NAD⁺ in the mitochondrial matrix), resulting in a redirection of glycolytic carbon into glycolytic intermediate-derived amino acids and a decrease of flux out of the mitochondria into TCA cycle intermediate-derived amino acids.

Interestingly, changes in amino acids were frequently associated with dynamic changes in the levels of transcripts encoding enzymes involved in their metabolism. In most cases, transcriptional responses appeared to act as a mechanism of counteracting flooding-induced changes in amino acid levels. For example, rapid increases in Gly, Tyr, and Ser were followed by transcriptional up-regulation of enzymes involved in their biosynthesis and down-regulation of enzymes involved in their degradation (Supplemental Table S2; Supplemental Figure S6). Conversely, a significant decrease in Glu at 24 h was accompanied by transcriptional up-regulation of enzymes involved in its biosynthesis and down-regulation of enzymes involved in its degradation. Interestingly, in the case of Lys, which was significantly increased after 24 h of flooding, this pattern of negative feedback was replaced with an apparent positive feedback mechanism, with the transcript encoding dihydrodipicolinate synthase 2 (protein identifier 730895), the enzyme responsible for the first committed step in Lys biosynthesis, being significantly up-regulated by more than 4-fold after 5 h of flooding, before Lys levels had significantly increased. Moreover, the accumulation of Lys was followed by a down-regulation of two enzymes involved in its degradation. Some metabolite-transcript interactions appeared to be more complex. For example, a 7-fold up-regulation of the transcript encoding a reversible mitochondrial Asp aminotransferase (protein identifier 362675) at the 5-h time point was followed by a 6-fold decrease in Asp at the 24-h time point. Interestingly, by the time the decrease in Asp was at its most severe (24 h), transcript levels for the mitochondrial Asp aminotransferase had already decreased to the point that they were no longer statistically significant. At this time, the transcript encoding another Asp aminotransferase (protein identifier 71675) had become significantly down-regulated and transcripts encoding two enzymes of Asp biosynthesis, Asp semialdehyde dehydrogenase (protein identifier 564677) and Asp kinase/homoserine dehydrogenase (protein identifier 92760), had already become significantly up-regulated by approximately 4-fold. Consistent with the notion that these transcriptional responses act to maintain homeostatic levels of Asp by reducing its catabolism and enhancing its biosynthesis, Asp levels were no longer significantly decreased by the 168-h time point. Although it is not clear exactly how these transcriptional responses were mediated, the observations above are consistent with the notion that transcriptional responses of genes involved in amino acid metabolism were largely driven by changes in amino acid levels.

Surprisingly, the most dramatic increase in abundance of a metabolite (57-fold) was observed for uric acid in hypoxically treated poplar roots (Fig. 2; Supplemental Table S2). This nitrogen-containing compound is a well-known transport form of symbiotically fixed N₂ in the xylem sap of legumes (Reinbothe and Mothes, 1962) but is also a product of purine degradation (Woo et al., 1980). The latter is most likely the source of uric acid in this study, as stresses other than hypoxia that affect the nutrient balance of plants such as sulfur deficiency (Nikiforova et al., 2006) also cause a higher abundance of this nitrogen-rich compound. The elevation of AMP (Fig. 2; Supplemental Table S2) may suggest the availability of purines for degradation in hypoxic roots. It has to be tested in further studies if nitrogen is relocated from roots to the shoot (like uric acid or other nitrogen-containing compounds), as suggested by decreasing protein concentrations in roots but increasing concentrations in leaves of flooded poplar (Kreuzwieser et al., 2002).

Growth and Physiology of the Trees

Repressed nutrient uptake/transport and the inhibition of biosynthetic processes are in accordance with the assumption that plants restrict O₂ consumption in order to avoid the occurrence of complete anoxia (Geigenberger, 2003). Although (1) the roots of the hypoxia-treated trees switched from mitochondrial respiration to alcoholic fermentation with an assumed impaired energy metabolism (Fig. 1) and (2) biosyntheses were inhibited (Supplemental Tables S4 and S5), the shoot growth of poplar trees did not show any significant limitation, at least during 2 weeks of hypoxic treatment (Fig. 6A). Given the metabolic analysis performed in this study, which indicates that the availability of both sugars and amino acids was not considerably reduced (Fig. 2; Supplemental Table S2), a reduction in both root and shoot growth would not necessarily be expected. Moreover, CO₂ assimilation rates, which drop in anoxia-sensitive species as a response to this stress (Pezeshki et al., 1996; Nunez-Elisea et al., 1999), were nearly unchanged in poplar (Fig. 6B) and a steady supply of carbohydrates was therefore ensured. In addition, the leaves of poplar were supplied with reduced carbon in the form of ethanol via the transpiration stream, as suggested from high ethanol concentrations in the xylem sap and ongoing transpiration even after 1 week of root hypoxia (Fig. 6B). However, as the growth of belowground plant parts was not determined, it cannot be excluded that root growth was limited by reduced energy supply.

Figure 6.

Effects of root hypoxia on shoot growth (A) and gas exchange (B) of gray poplar. A, Two weeks prior to and after starting the hypoxic treatment, shoot lengths were determined. Means ± se of five to 10 trees per treatment are given. B, Fully mature leaves were chosen and rates of net CO₂ assimilation (A), transpiration (E), and stomatal conductance for water vapor [g(H₂O)] were determined. Means ± se of eight leaves per plant and four plants per treatment are given. Statistically significant differences between treatments and controls at P < 0.05 are indicated by asterisks above the bars. White symbols and bars, Normoxic controls; gray symbols and bars, hypoxically treated plants.

Possible Flooding Adaptations

One aim of this work was to identify mechanisms mediating flooding tolerance by comparing the highly flooding-tolerant poplar with flooding-sensitive Arabidopsis. As expected, this study revealed numerous pathways connected to carbon and nitrogen metabolism that are altered by hypoxia. These pathways were affected in a similar way in Arabidopsis, although experimental conditions differed from this poplar study (Klok et al., 2002, root cultures exposed to 5% O₂ for up to 20 h; Liu et al., 2005, whole seedlings exposed to 3% O₂ for up to 24 h and use of whole plants for array analysis; Loreti et al., 2005, whole seedlings anoxically exposed for up to 72 h and use of whole plants for array analysis; Branco-Price et al., 2008, whole seedlings exposed to an argon environment free of O₂ and CO₂ for up to 9 h and use of whole seedlings; Van Dongen et al., 2008, whole seedlings exposed to varying O₂ concentrations for up to 48 h and whole seedlings used for microarrays). Adaptations seen in all experiments included increased glycolytic flux, modified Suc degradation, and a reduction in energy-consuming processes such as transport processes and biosyntheses.

However, analysis of the hypoxic response in poplar provided insights not apparent in studies performed using Arabidopsis that cannot be explained by the different experimental conditions alone. A large difference between Arabidopsis and poplar (and also rice; Lasanthi-Kudahettige et al., 2007) is the number of genes differentially expressed during hypoxia. In stress-exposed poplar tissue, over 5,000 genes showed significantly altered expression, whereas under similar conditions, the transcript abundance of only approximately 150 genes was changed in Arabidopsis (Klok et al., 2002; Liu et al., 2005). Differences in gene expression in Arabidopsis studies can partially be explained by the different O₂ concentrations applied (Van Dongen et al., 2008). At lower O₂ abundance, more genes were differentially expressed. This explains that in the study by Branco-Price et al. (2008), approximately 1,500 genes were differentially expressed in Arabidopsis. However, this was still less than one-third of the changes observed in hypoxia-treated poplar. A high percentage of altered genes in both species were transcription factors (TFs; 456 genes in poplar), protein kinases (more than 200 genes), or were involved in signaling (115 genes) and hormone metabolism (137 genes). In poplar, over 2,500 TFs have been identified and classified into 64 families (Zhu et al., 2007), of which 456 showed altered transcript abundances in this study (Supplemental Table S1). Although Arabidopsis possesses nearly the same number of TFs (estimated, 1,922), only approximately 50 of them were differentially expressed in response to hypoxia (Liu et al., 2005). TFs with modified expression levels spanned all major TF families, such as AP2 domain (17 genes), basic helix-loop-helix (28 genes), zinc finger (43 genes), myb (50 genes), Leu zipper (21 genes), and WRKY (20 genes) family proteins. However, in contrast to Arabidopsis, members of the NAC domain (26 genes) and bZIP (15 genes) families also showed changes in expression in poplar. Among the most dramatically up-regulated TFs were members of the myb, AP2/EREB, and Leu zipper families. Similarly, the strongest down-regulation was observed in TFs belonging to the myb, Leu zipper, and WRKY families (Supplemental Table S1). Some members of these families have been suggested to be involved in transcriptional regulation of wood formation (Demura and Fukuda, 2007), suggesting that the observed down-regulation of genes involved in this pathway is the result of a complex regulatory network involving regulation of TFs at the level of transcription. The observed changes in transcript abundance of genes involved in cytokinin signaling may also indicate down-regulated wood formation (Demura and Fukuda, 2007). The observed differential expression of numerous TFs and components of plant hormone signaling in poplar suggests a very complex transcriptional regulatory network leading to stress adaptation of poplar in response to hypoxia.

Another difference between Arabidopsis and poplar was the different temporal pattern of transcriptome changes. In the Arabidopsis study by Liu et al. (2005), changes in transcript abundance peaked in number and magnitude at 6 h after hypoxic treatment, whereas in the study by Van Dongen et al. (2008), a steady state was reached around 2 h after starting the stress; in poplar, differential gene expression increased steadily until 7 d of treatment. Such a pattern was evident for a broad range of genes, ranging from TFs and other regulatory elements to genes of major metabolic pathways. ADH transcript abundance in Arabidopsis, for example, peaked 2 to 4 h after hypoxia (Liu et al., 2005), whereas in poplar it was consistently up-regulated during the whole flooding period. Similar patterns were observed for genes involved in glycolysis. It is suggested that many of the changes in gene expression observed in flooded poplar are related to an improvement of carbohydrate status and to the maintenance of high carbohydrate concentrations to allow for the formation of energy equivalents. Furthermore, changes in transcript abundance in poplar occurred in an organ-specific manner, revealing that the majority of changes at the transcript level were restricted to the site of hypoxia; thus, no general systemic signal appears to be generated to trigger gene expression changes throughout the plant. However, analysis of various metabolites in both leaves and roots (Figs. 2 and 3) revealed that the aerial parts of the plant facilitate specific alterations in primary metabolism that provide hypoxic roots with sources of energy, resulting in flooding tolerance.

Interestingly, the expression of two alternative oxidase (AOX) isoforms (protein identifiers 727508 and 365930) was considerably increased in poplar roots during hypoxia. In contrast, in Arabidopsis exposed to hypoxia, AOX expression is reduced (Liu et al., 2005) or only transiently enhanced (Klok et al., 2002). AOX induction in poplar may simply be due to a restriction of electron transport via the cytochrome electron transport chain (Vanlerberghe and Mclntosh, 1994; Wagner and Moore, 1997). As the K_m of AOX for O₂ is even higher than that of cytochrome oxidase (Millar et al., 1994; Ribas-Carbo et al., 1994), it is unlikely that it plays any role during hypoxia, but it may act as a preoxidant defense mechanism on the return to aerobic conditions (Purvis and Shewfelt, 1993; Purvis et al., 1995; Millar et al., 2001; Sweetlove and Foyer, 2004). Although reactive oxygen species (ROS) have often been implicated in the induction of AOX, signal transduction pathways inducing AOX independent of ROS have been described in other plants (Clifton et al., 2006; Amirsadeghi et al., 2007), and it is possible that ROS are produced under hypoxia (Rhoads et al., 2006).

In poplar, the GABA shunt was stimulated, which was in contrast to Arabidopsis, in which it was either observed to be only transiently up-regulated (Klok et al., 2002; Liu et al., 2005) or not up-regulated at all (Loreti et al., 2005). The GABA shunt is considered a metabolic adaptation to O₂ limitation, as it is a proton-consuming process stabilizing cytoplasmic pH (Crawford et al., 1994). It further represents an anaplerotic sequence for the TCA cycle, bypassing the IDH reaction. As GABA stimulates ethylene production in sunflower (Helianthus annuus), it might play a role in signaling (Kathiresan et al., 1997). Thus, changes in GABA concentrations in roots of hypoxically treated poplar may be a first step in inducing morphological adaptations, such as the formation of hypertrophied lenticels and aerenchyma. In poplar, such adaptations usually occur 2 to 3 weeks after flooding treatment and therefore were not observed in this study.

CONCLUSION

Comparison of the response of the relatively flood-sensitive plant Arabidopsis with that of the highly flood-tolerant tree poplar revealed important differences. The metabolite and transcript patterns reveal the poplar's capability to maintain carbon and energy metabolism during periods of O₂ deficiency, which may be an important adaptation that allows tolerance to anoxia (Drew et al., 1997). In poplar, this is mediated by the stimulation of glycolysis and a steady supply of sugars for this pathway, on the one hand, and the repression of energy-demanding processes, on the other hand. From the large changes in transcript abundance observed, we conclude that not a single pathway but the sum of numerous well-orchestrated processes in carbon and nitrogen metabolism and morphological adaptation is responsible for the high flood tolerance of poplar.

MATERIALS AND METHODS

Plant Material

Experiments were performed with 3-month-old hybrid poplar plants (Populus × canescens). The plants were micropropagated and cultivated as described by Leplé et al. (1992). Four-week-old stem explants were transferred from sterile agar cultures into plastic pots (12 × 10 × 10 cm) containing a mixture of perlite, sand, and potting soil (2:1:1). Each tree was supplied with 200 mL of a solution containing 3 g L⁻¹ complete fertilizer (Hakaphos Blau; Bayer) every 2 weeks and was watered daily with tap water. Trees were grown under long-day conditions (16 h of light) at day and night temperatures of 22 ± 4°C and relative humidity of 72% ± 10%.

Experimental Setup

At the beginning of the experiment, the shoots of the trees had a size of 20 to 25 cm. To induce root hypoxia, 52 trees were transferred into large plastic containers. The containers were filled with tap water until the water level exceeded the soil surface by 2 to 3 cm. Oxygen concentrations in the soil of flooded trees were analyzed representatively in four individuals using O₂ microsensors (micro-optodes; Microx TX2; PreSens). Under normoxic conditions, O₂ concentrations amounted to about 7.5 μL L⁻¹ (equal to approximately 80% air saturation), whereas under flooding conditions, values decreased to 0.1 μL L⁻¹ (1.3% air saturation) within 2 h; they further dropped to 0.05 μL L⁻¹ (0.6% air saturation) within another 1 h and then remained constant at this level for the rest of the flooding period. Besides flooded trees, 52 trees were used as nonflooded, normoxic controls. After 1, 3, 5, 7, 24, and 168 h of treatment, four trees of each treatment (flooded and controls) were harvested; tissues were immediately frozen in liquid nitrogen and stored at −80°C until further analysis. Enzyme activities were determined immediately using fresh material.

RNA Isolation and cDNA Synthesis

For total RNA extraction, the Plant RNeasy kit (Qiagen) was used according to the manufacturer's instructions with some minor modifications. Aliquots of 90 mg of frozen plant powder were added to 900 μL of Qiagen lysis (RLT) buffer supplemented with 1% β-mercaptoethanol, 1% polyvinylpolypyrrolidone, and 2% (w/v) polyethylene glycol 10,000 and then incubated for 10 min at 58°C. The samples were further processed according to the manufacturer's protocol. For microarray and qRT-PCR analyses, total RNA was treated with DNase during (RNase-free DNase kit; Qiagen) and after (DNA-free kit; Ambion) the extraction procedure. For cDNA synthesis, 1 μg of total RNA was reverse transcribed using oligo(dT) primers and SuperScript II reverse transcriptase (Invitrogen) in a total volume of 20 μL.

qRT-PCR

For measurements of transcript levels of the genes of interest, specific oligonucleotide primer sets were designed and tested for cross-hybridization and efficiency (Supplemental Table S7). qRT-PCR was performed using the iQ5 instrument with iQ SYBR Green Supermix (Bio-Rad), using 25-μL reaction volumes with 0.5 to 0.9 μm of each primer (Supplemental Table S5) and 2.5 μL of template, under conditions optimized to minimize primer-dimer formation and to maximize amplification efficiency (“hot start”: 30 s, 95°C; 40 cycles: 15 s at 95°C and 30 s at 55°C). Melting curve analysis was performed after PCR amplification to ensure that products were specific. For internal normalization, transcript levels of poplar β-tubulin (PcTUB; EMBL accession no. AY353093) were determined.

Microarray Analysis

Microarray analysis of changes in transcript abundance in poplar leaves and roots was performed using a set of 30 Affymetrix GeneChip Poplar Genome Arrays. Twenty-one and nine arrays were used to analyze root and leaf samples, respectively. The root samples from flooded trees taken 5, 24, and 168 h after starting the flooding treatment (nine arrays) and the root control samples collected before and 5, 24, and 168 h after starting the flooding treatment (12 arrays) were analyzed by the microarray technique. For leaves, the controls collected before flooding and 168 h after starting the treatment (six arrays) and the leaves of the flooded poplar at this time (three arrays) were used for analysis. The quality of the isolated total RNA was tested using an Agilent Bioanalyzer (Agilent Technologies) and spectrophotometric analysis (NanoDrop ND-1000; NanoDrop Technologies) to determine the A₂₆₀-A₂₈₀ ratio. Preparation of labeled copy RNA from 3.5 μg of root total RNA and 5 μg of leaf total RNA, target hybridization, as well as washing, staining, and scanning were performed exactly as described in the Affymetrix GeneChip Expression Analysis Technical Manual, using an Affymetrix GeneChip Hybridization Oven 640, an Affymetrix Fluidics Station 450, and a GeneChip Scanner 3500 7G at the appropriate steps. Data quality was assessed using GCOS1.4 (Affymetrix) by checking raw array images for artifacts, background levels, and percentages of absent and present calls for all arrays within an experiment as well as 3′-5′ ratios of control genes. Normalization was performed using the raw data (CEL files) by applying the RMA (for robust multiarray average) algorithm using RMAExpress (version 0.4.1 release; Bolstad et al., 2003). Transcripts called as “absent” and genes with an average raw signal value below 50 in the untreated sample were filtered out. Correct normalization was checked, and differentially expressed genes were determined by applying two-tailed Student's t tests (P < 0.05). In this procedure, each hypoxically treated set of root samples was tested against all four control sets (each collected in triplicate before and 5, 24, and 168 h after starting the flooding treatment). The same was applied for leaf samples, in which leaves from flooded trees were tested against the controls taken before treatment and after 168 h of flooding treatment. False discovery rate analysis was then performed using QVALUE software (Storey and Tibshirani, 2003). For estimating q values, π₀ (an estimate of the total proportion of true null hypotheses) was determined based on the distribution of 61,413 P values using the smoothing method. A stringent approach was used to select genes differentially expressed in poplar roots and leaves. Genes were considered to be differentially expressed only when (1) the transcript abundance in hypoxia-treated plants was significantly different against all four controls as determined by Student's t test, (2) the calculated q value against the direct control was ≤0.05, and (3) the change in expression was at least 2-fold compared with the control. The Institute for Genomic Research Multiexperiment viewer version 4.0.1 was used to produce hierarchical dendrograms of differentially expressed genes revealing natural groupings in microarray data. Dendrograms displaying the “natural” groupings within the data are displayed with different color intensities based on sorted fold change data, and distances (4.00) between tree branches were determined using a Euclidean function (Eisen et al., 1998; Saeed et al., 2003). The software Pageman (Usadel et al., 2006) was used for the generation of overview graphs displaying differentially expressed genes in functional classes using Wilcoxon statistical testing. Data from this study are available from Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo) and are listed under the following accession numbers: GPL4359 (for the array platform) and GSE13109 (for the experimental series), where a link to the expression data is given.

Metabolite Determination

Concentrations of soluble carbohydrates were determined by HPLC-pulsed-amperometric detection (Heizmann et al., 2001). Ethanol concentrations were determined enzymatically as described previously (Kreuzwieser et al., 2001). Starch was analyzed colorimetrically using a commercial test kit (Boehringer). For extraction, 50 mg of powdered (under liquid N₂) plant material was added to 1 mL of dimethyl sulfoxide/25% (w/v) HCl (80:20, v/v). After 30 min of incubation at 60°C, samples were centrifuged (5 min, 12,000g) and 200 μL of the supernatant was added to 1.2 mL of ice-cold 0.2 m citrate buffer (pH 10.6). The resulting solution was then used for analyses. Amino acid concentrations were determined using the HPLC method described by Kreuzwieser et al. (2002).

In parallel with these techniques, concentrations of polar low molecular weight metabolites present in poplar roots were analyzed by gas chromatography-mass spectrometry (GC-MS). Metabolites were extracted and derivatized using a method modified from that of Roessner-Tunali et al. (2003). Derivatized samples were analyzed on an Agilent GC/MSD system (Agilent Technologies). Raw GC-MS data files in the proprietary ChemStation (.D) format were exported to generic NetCDF/AIA (.CDF) format with ChemStation GC/MSD Data Analysis Software. The NetCDF files produced were then processed using in-house MetabolomeExpress (version 0.9) software to carry out all peak detection, quantification, library matching, normalization, statistical analysis, and data visualization. Detailed methodology is presented in Supplemental Data Set S1.

Enzyme Activities

ADH and PDC activities were determined according to the protocol of Owen et al. (2004). Aliquots of 100 mg of fresh leaf or root material were homogenized in 1.5 mL of ice-cold extraction buffer (0.1 m MES buffer, pH 6.5, 15% [v/v] glycerol, 0.2% [v/v] Tween 20, 2 mm dithiothreitol, 1 mm phenylmethylsulfonyl fluoride, and 10% polyvinylpolypyrrolidone). For determination of PDC activity, 30 μL of protein extract was added to 250 μL of assay buffer (0.1 m MES buffer, pH 6.5, 80 μm NADH, 10 μm thiamine pyrophosphate, 5 mm MgCl₂, 14 units mL⁻¹ ADH, and 5 mm pyruvate) and NADH consumption was analyzed.

Gas Exchange

Net CO₂ assimilation, transpiration, stomatal conductance, and acetaldehyde exchange between leaves and the atmosphere were determined using the portable gas-exchange measuring system GFS-3000 (Walz) as described by Kreuzwieser et al. (1999). Standard conditions (1,000 μmol m⁻² s⁻¹ photosynthetic photon flux density, 25°C leaf temperature) were applied. From each plant, eight fully mature leaves were analyzed.

Statistical Analysis

Data obtained were subjected either to Student's t test (Microsoft Excel 2002) or to analyses of variance and multiple range tests (Tukey) by ANOVA (SPSS 14.0 for Windows; SPSS Inc.).

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Figure S1. Real-time PCR validation of array data.

• Supplemental Figure S2. Overview of pathways with significantly changed transcript abundance in hypoxia-treated poplar roots.

• Supplemental Figure S3. Detailed view of pathways involved in cell wall biosynthesis (graphical representation extracted from Supplemental Fig. S2).

• Supplemental Figure S4. Transcript levels of genes involved in the TCA cycle and mitochondrial respiration.

• Supplemental Table S1. Genes differentially expressed in roots and leaves of hypoxically treated poplar.

• Supplemental Table S2. Changes in metabolite concentrations.

• Supplemental Table S3. Differentially expressed genes involved in primary energy metabolism in poplar roots.

• Supplemental Table S4. Genes involved in cellulose biosynthesis or degradation.

• Supplemental Table S5. Genes involved in phenylpropanoid metabolism.

• Supplemental Table S6. Differentially expressed genes involved in nitrogen metabolism.

• Supplemental Table S7. Oligonucleotide primer sets used for qRT-PCR.

• Supplemental Data Set S1. Materials and methods for GC-MS analysis.

• Supplemental Data Set S2. The MIAME checklist.