Comparative analysis between plant species of transcriptional and metabolic responses to hypoxia

Authors

  • Reena Narsai,

    1. ARC Centre of Excellence in Plant Energy Biology, MCS Building M316 University of Western Australia, 35 Stirling Highway, Crawley 6009, WA, Australia
    2. Centre of Excellence for Computational Systems Biology, University of Western Australia, 35 Stirling Highway, Crawley 6009, WA, Australia
    Search for more papers by this author
  • Marcio Rocha,

    1. Energy Metabolism Research Group, Max Planck Institute of Molecular Plant Physiology, Am Muehlenberg 1, D–14476 Potsdam-Golm, Germany
    Search for more papers by this author
  • Peter Geigenberger,

    1. Department Biologie I, Ludwig-Maximilians-Universität München, D–82152 Martinsried, Germany
    Search for more papers by this author
  • James Whelan,

    1. ARC Centre of Excellence in Plant Energy Biology, MCS Building M316 University of Western Australia, 35 Stirling Highway, Crawley 6009, WA, Australia
    Search for more papers by this author
  • Joost T. van Dongen

    1. Energy Metabolism Research Group, Max Planck Institute of Molecular Plant Physiology, Am Muehlenberg 1, D–14476 Potsdam-Golm, Germany
    Search for more papers by this author

Author for correspondence:
Joost T. van Dongen
Tel: +49 331 5678353
Email: dongen@mpimp-golm.mpg.de

Summary

  • The variation in tolerance to low oxygen is likely explained by divergent sets of molecular and metabolic responses between species.
  • We analysed the versatility of the response to low oxygen of primary metabolism by comparing nine previously published metabolome profiling studies. Data were juxtaposed with expression profiles of genes encoding enzymes involved in the metabolic pathways of rice, Arabidopsis and poplar. Furthermore, full transcript profiles were compared to determine commonalities in the expression of orthologous genes and genes that serve similar functions.
  • Activation of fermentation and the accumulation of alanine plus succinate were observed in all species, but transcriptional regulation of these metabolic pathways varied. Global analysis of orthologue expression revealed that most differentially expressed genes either had no orthologues or were not affected in the other species. Expression analysis of nearly all gene clusters with common functions varied significantly between species.
  • The resemblance of the metabolic response to hypoxia indicates that this occurs independent of the level of tolerance. However, regulation of these processes at transcriptional level varied between species. An important role is suggested for signalling and post-transcriptional regulation to be involved in the mechanisms that lead to tolerance to hypoxia.

Introduction

Plants display a remarkable variation in their ability to tolerate limitations in oxygen availability, ranging from a small decrease in the oxygen concentration (hypoxia) to total absence of oxygen (anoxia) (Bailey-Serres & Voesenek, 2008). Such changes not only result from a reduced supply of oxygen from the environment owing to periodic submergence or waterlogging (Bailey-Serres & Voesenek, 2008), but also because the oxygen uptake by plants is strongly affected by the natural diffusion resistance for oxygen (Armstrong et al., 2009). The continuous consumption of oxygen via mitochondrial respiration can lead to plant internal oxygen concentrations that are below one-fifth of the concentration of oxygen in air. In roots (Armstrong et al., 1994; Zabalza et al., 2009), seeds (Borisjuk & Rolletschek, 2009) and tubers (Geigenberger et al., 2000), steep oxygen gradients have been measured with lowest concentrations of oxygen around the phloem (van Dongen et al., 2003). Developmental fluctuations of internal oxygen concentration have been observed during seed germination and growth (van Dongen et al., 2003; Benamar et al., 2008).

The oxygen concentration of a cell has been shown to affect metabolic activity both in animal (Semenza, 2007) and plant tissues (Gupta et al., 2009), as well as in bacteria (Li et al., 2001) and yeast (Burke & Poyton, 1998). Carbohydrate and energy metabolism is affected in a variety of species when the oxygen availability declines. In plants even a small decrease of the oxygen concentration in the environment results in a decrease of the cellular energy status (ratio of ATP to ADP), whereas the redox poise (NADH/NAD ratio) is not affected until environmental oxygen concentrations drop below 5% (v : v) (Geigenberger, 2003). The accumulation of NADH and the concomitant depletion of available NAD would ultimately lead to an inhibition of glycolysis, which is prevented by NADH oxidation via fermentation. Indeed, the accumulation of the fermentation products lactate and ethanol, together with the increase of the NADH to NAD ratio and a reduction of the cellular energy charge are generally considered to be a common response to hypoxia of aerobic organisms including plants (Bailey-Serres & Voesenek, 2008).

Despite the highly conserved common metabolic reaction to low oxygen, that is, the upregulation of glycolysis and fermentation, large differences in hypoxic tolerance are observed between species. Many reasons are described that are responsible for this variation in stress resistance to low oxygen. For example, species differ in the ability to supply oxygen to hypoxic tissues from the aerated shoot to waterlogged roots via diffusion or convective gas flow through stem, rhizome and root tissues (Armstrong & Drew, 2002; Armstrong & Armstrong, 2009). Furthermore, various rice (Oryza sativa) varieties have been described to have evolved different rescue strategies to complete flooding stress. One survival syndrome is known as the ‘quiescent strategy’, whereas the second is described as ‘fast escape strategy’ (Bailey-Serres & Voesenek, 2008; Nagai et al., 2010). The quiescent strategy is a response to survive short periods (up to a few weeks only) of submergence. The inhibition of growth saves energy reserves until the water level recedes, after which the plant can restart growing immediately using the preserved energy resources. Induction of this tolerance strategy depends on the activation of the transcription factor Submergence-1A (Sub1a), which is an ethylene responsive gene that belongs to the Ethylene Response Factor (ERF) transcription factor family (Fukao et al., 2006; Xu et al., 2006). The second strategy that is adopted by varieties that are able to withstand long-term flooding is to accelerate growth via rapid internode elongation to keep their leaves above the water level. This requires the rapid allocation of carbohydrate resources to supply both energy and carbon skeletons that are used for growth. For this purpose, starch reserves are rapidly hydrolysed, which is a feature that not all species are able to achieve during anoxic stress (Guglielminetti et al., 2001; Perata & Voesenek, 2007). Interestingly, the fast escape strategy is activated by the transcription factors SNORKEL1 and 2 (SK1 and SK2) that both belong to the same ethylene responsive transcription factor family as Sub1a does (Hattori et al., 2009). It is very intriguing that these two flooding-tolerance strategies are both regulated by ethylene. Finally, a third molecular mechanism exists in rice where in some varieties seeds germinate in the absence of oxygen that appears to be linked to a CIPK15 kinase (calcineurin B-like (CBL) interacting protein kinase) signalling cascade, that integrates sugar and hypoxia signalling in rice (Lee et al., 2009).

The versatility of responses to hypoxic stress within one species provokes the question how diverse the responses are in other species. Most research on describing responses to low oxygen has investigated common adaptive responses, most notably the upregulation of glycolysis and fermentation. However, alone, these common responses do not explain the diversity of responses that occur to distinguish between adaptation and response to low oxygen. The differences as well as the similarities between species could be equally important. The aim of the present study is to compare various metabolome and transcriptome profiling studies from both monocot and dicot species that are currently available in the literature to discover both similarities as well as to investigate the divergence. The analysis confirmed the existence of a core-set of hypoxia responses both at the level of gene expression and metabolism, but striking differences were also observed and are discussed with reference to tolerance and survival in low oxygen environments.

Materials and Methods

Publically available rice and Arabidopsis thaliana microarrays

The publically available Affymetrix microarray data used for the analysis in this study were downloaded from the Gene Expression Omnibus within the National Centre for Biotechnology Information database or from the MIAME ArrayExpress database (http://www.ebi.ac.uk/arrayexpress/). For Arabidopsis and rice, the data was downloaded as CEL files in the sets; GSE9719 (Total RNA – 2 h and 9 h exposure to anaerobic conditions) and E-MEXP-2267 (Total RNA – 3 h and 6 h exposure to anaerobic conditions; 27A-N and 30A-N). Note that for Arabidopsis, the anaerobic treatments were carried out on 1-wk-old seedlings (on media with 1% sucrose), for rice the anaerobic treatments were carried out on 1-d-old germinated embryos and for poplar, the plants were 3 months old. Details of the respective experiments are described in Branco-Price et al. (2008), Kreuzwieser et al. (2009) and Narsai et al. (2009) for Arabidopsis, rice and poplar respectively.

Microarray analyses

All raw intensity CEL files were imported into Avadis 4.3 (Strand Genomics, Bangalore, India) and the standard MAS5.0 (Affymetrix, Santa Clara, CA, USA) normalization was first carried out in order to determine present/absent/marginal calls for each probeset. Probesets which had present calls across two or more replicates were considered to be expressed and used for further analysis. For each experiment, the GC content based Robust Multi-array Average (GC-RMA) normalized data (control aerobic vs anaerobic treated) were used as the input set for the differential expression analysis for both time points. This analysis was carried out using the Cyber-T method, which implements a Bayesian method for determination of probesets showing significant changes in transcript abundance. The posterior probability of differential expression (PPDE) method within Cyber-T was used for false discovery rate calculation. All input criteria were set according to the Cyber-T recommendations applicable for each experimental set. For a probeset to be defined as differentially expressed under anaerobic conditions, the fold change had to be greater than or equal to −2 or 2, this fold change had to be statistically significant at < 0.01 and had to be associated with a PPDE of > 0.96 (false discovery rate; FDR) (Baldi & Long, 2001).

pageman (Usadel et al., 2006) and mapman (Thimm et al., 2004; Usadel et al., 2005) analyses were carried out using a reduced set of unique probesets representing the differentially expressed genes. For poplar, the set of FDR-corrected differentially expressed genes at (< 0.01) (Table S1 from Kreuzwieser et al., 2009) was filtered to only include genes differentially expressed by a minimum of twofold, making it comparable to the rice and Arabidopsis differentially expressed sets. For all three studies –Arabidopsis (Branco-Price et al., 2008), rice (Narsai et al., 2009) and poplar (Kreuzwieser et al., 2009) – the differential expression analysis involved a comparison at the time-points within each experiment using the GC-RMA normalized values. Therefore, in order to visualize this data in mapman/pageman, the differentially expressed gene list was generated on the criteria that a probeset had to be significantly changing (up/down) at one or more time points in the same direction (up/down) with no significant changes in the opposite direction. In this way, the maximum/minimum fold changes were visualised in mapman. The Fisher’s test for ORA (over-representation analysis) analysis was carried out in pageman in order to determine statistically significant changes in specific BINS. Given that pageman does not allow visualization of data from more species at once, the raw data was exported, matched and visualised in parallel using Partek Genomics Suite (version 6.5) (St Louis, MO, USA) for the pageman output seen in Fig. 1.

Figure 1.

Defining common and exclusive anaerobic responsive genes in rice and Arabidopsis. (a) The number of genes significantly (< 0.01, posterior probability of differential expression (PPDE) > 0.99) upregulated/downregulated in abundance in rice (R) and Arabidopsis (A) under anaerobic conditions were compared in terms of orthology and transcriptomic response. The Inparanoid database was used for determination of orthologues. The number of transcripts significantly responding is shown as follows: the number of transcripts that had no gene orthologues in the respective other species (lightest shade), transcripts with known gene orthologues (darker shade) and the number of transcripts that were both orthologous and showing a conserved response or opposite response, in rice and Arabidopsis (darkest shade). (b) The proportion of genes in the rice and Arabidopsis genomes that have orthologues in Arabidopsis and rice, respectively, is indicated (dark orange). For each subset of differentially expressed genes, the proportion of transcripts that have orthologous genes responding in a similar way (Up/Down-regulated in A & R; indicated in red/blue), have orthologous genes responding in an opposite manner (e.g. upregulated in rice and downregulated in Arabidopsis; denoted Opposite response; indicated in purple), have orthologous genes not significantly changing in abundance in the respective other species (light orange) or do not have orthologues (indicated in grey) is shown. (c) pageman analysis of the microarray data in response to anaerobic conditions. For rice and Arabidopsis, the subsets of genes that were both orthologous and showing conserved responses were analysed separately from the rest of the differentially expressed genes (i.e. those that were either not orthologous/showed an opposite response or had orthologues that were unchanging in the respective other species). For each subset, significant fold changes in transcript levels in response to switching from aerobic to anaerobic conditions were analysed using the pageman tool. Statistical analysis of over-represented functional categories was carried out using the Fisher method. Nonsignificantly changing functional categories were collapsed for display. Statistical significances are represented by a false colour heat map (up, red; down, blue) where a z-score of 1.96 represents a P-value of 0.05. The green and orange shading indicates conserved and divergent responses, respectively, based on functional classifications.

Analysis of orthologues

The InParanoid: Eukaryotic Ortholog Groups database (version 7.0) was used to analyse all orthologues between rice and Arabidopsis (Remm et al., 2001). The orthologous group file was downloaded for the whole-genome comparison of rice vs Arabidopsis. This produced information for orthologues identified by The Institute for Genomic Research (TIGR) identifiers for rice and Arabidopsis gene identifiers (AGIs) for Arabidopsis.

Phylogeny analysis of plant haemoglobin genes

All multiple sequence alignments were carried out using mafft (Katoh et al., 2005) and visualized using multiple align show (http://www.bioinformatics.org/sms/multi_align.html). The program iqpnni (Vinh le & Von Haeseler, 2004) was used to reconstruct a maximum likelihood phylogeny assuming the Whelan and Goldman model (Whelan & Goldman, 2001). Phylogenetic trees were finally visualized using the program geneious (http://www.geneious.com).

Publically available metabolome data

To compare metabolic responses to low oxygen, a collection of metabolite profiling studies was compiled from the literature that were obtained from experiments in which hypoxia or anoxia was imposed by various approaches. To allow comparison of the data between experiments, all changes of metabolite abundance were calculated relative to the respective normoxic control treatment. A relative increase of the metabolite amount with > 25% was marked red, whereas a relative decrease of > 25% was marked in blue. Smaller changes (between 25% decrease and 25% increase) were coloured in yellow.

The following experimental treatments are described briefly. More detailed information, as well as the raw data files can be found in the original publications. Two low-oxygen experiments were available for rice (Narsai et al., 2009). Germinating seeds in liquid medium were aerated with air or nitrogen to obtain either control or anaerobic conditions. In one experiment (Table 1, ‘continuous anoxic germination’) germinating seeds were kept under anoxia continuously and samples were taken after 1, 3, 12, 24 and 48 h. In a second experiment seeds were kept in normoxic medium for 24 h before a switch to anoxia. Samples from anoxic seedlings were taken after 1, 3 and 6 h. Metabolites were analysed by GC/MS. For Arabidopsis, three different datasets were obtained. In the first set of experiments, 7-d-old seedlings on vertical solid 1× Murashige and Skoog medium (MS) phytagel plates supplemented with 1% sucrose were kept in 99.995% argon gas for 2 or 9 h before sampling (Branco-Price et al., 2008). Metabolites were analysed by nuclear magnetic resonance. Arabidopsis seedlings were grown on vertical agar plates supplemented with 0.5× MS and 1% sucrose that were placed in containers with premixed gas with different oxygen concentrations (van Dongen et al., 2009). The third dataset with results from Arabidopsis originates from Gibon et al. (2002) in which siliques and seeds were harvested from 12- to 14-wk-old plants that were pretreated for 2 h with air containing different oxygen concentrations, as indicated earlier in the text and in Table 1. Metabolites were determined by enzymatic cycling essays. Data from potato (Solanum tuberosum) tubers were obtained from an experiment in which small tuber discs were incubated in liquid medium through which premixed gasses with different oxygen concentrations was blown (Geigenberger et al., 2000). Samples were taken after 2 h. Sugars, sugar-phosphates and organic acids were determined via various enzymatic analysis methods. The levels of amino acid were obtained from the same biological material as described in Geigenberger et al. (2000) and determined by high-pressure liquid chromatography (HPLC) analysis as described by van Dongen et al. (2003). Metabolite profiling of waterlogged poplar (Populus × canescens) roots from 20 to 25 cm tall plants growing in pots was performed by HPLC pulsed-amperometric detection and GC-MS. Amino acids were detected by HPLC (Kreuzwieser et al., 2009). Metabolites from waterlogged roots and nodules of Lotus japonicus were determined by GC-MS (Rocha et al., 2010a). The unicellular alga Chlamydomonas reinhardtii was exposed to anaerobic conditions by impairing photosystem II via sulphur depletion from the liquid growth medium. Metabolite profiling was performed using nuclear magnetic resonance analysis (Matthew et al., 2009).

Table 1.   Overview of metabolic changes induced by various hypoxia, anoxia or flooding treatments
Plant speciesOryza sativumArabidopsis thaliana Solanum tuberosumRicinus communisPopulus×canescensLotus japonicusChlamydomonas reinhardtii
ReferenceNarsai et al. 2009Branco-Price et al. 2008van Dongen et al. 2009Gibon et al. 2002Geigenberger et al. 2000van Dongen et al. 2003Kreuzwieser et al. 2009Rocha et al. 2010Matthew et al. 2009
TissueSeedlingsSeedlingsRootsSeedsSilique wallTuber discsPhloemRootsRootsNodulesLiquid cell culture
TreatmentShift to anoxiaContinuous anoxic GerminationAnoxia1% O24% O28% O21% O24% O28% O21% O24% O28% O20% O21% O24% O28% O212% O20% O21% O24% O28% O212% O20% O21% O24% O28% O212% O20.5% O23% O26% O2WaterloggingWaterloggingAnoxia induced by sulfur depletion
Time1 h3 h6 h1 h3 h12 h24 h48 h2 h9 h0.5 h0.5 h0.5 h2 h2 h2 h48 h48 h48 h2 h2 h2 h2 h2 h2 h2 h2 h2 h2 h2 h2 h2 h2 h2 h3 h3 h3 h5 h24 h168 h5 d5 d24 h48 h72 h96 h120 h
  1. The values that are given in the table indicate the ratio of metabolite levels from the low oxygen treatment compared with a normoxic control. Ratios that are > 1.25 are marked in red. Ratios below 0.75 are marked in blue. Metabolites that hardly changed (0.75 < ratio< 1.25) are coloured yellow. A brief description of the experimental treatment is given in the Materials and Methods section. For the raw data as well as a detailed description of the treatments, please refer to the original publications indicated in each column.

Sugars
Sucrose1.011.111.18     1.901.10                        0.500.831.00   1.030.95     
Fructose1.141.131.061.050.660.060.260.820.801.100.981.351.101.011.061.081.011.410.98          1.713.232.431.520.93   3.012.602.430.262.43     
Glucose1.161.010.910.990.930.050.252.010.801.201.031.351.020.930.950.941.061.430.99                     0.213.110.551.141.020.760.57
Trehalose0.840.530.48       0.891.221.000.900.910.881.161.670.94                  0.392.270.440.380.97     
Glc1P          0.821.351.041.151.131.061.571.820.860.570.380.691.001.080.730.510.640.610.631.04 0.790.751.131.001.301.30          
Glc6P0.770.760.971.011.311.111.021.69  0.891.450.940.970.960.881.092.010.930.470.470.811.281.900.600.700.700.700.700.900.960.790.911.030.501.001.25     2.642.253.572.542.54
Fru6P0.760.781.061.091.481.041.402.26  1.472.231.361.121.561.150.790.711.070.500.400.901.201.300.880.770.770.660.770.901.020.810.760.920.500.871.000.372.693.141.224.77     
Gly3P1.171.291.730.860.952.244.798.62           2.502.082.502.502.502.001.001.001.001.00        2.032.450.40       
Gly2P   1.041.321.480.960.95                             2.042.390.41       
Maltose0.820.680.801.371.082.543.012.08                             2.032.200.480.93      
Galactose   0.770.960.080.080.36                             2.273.395.86  0.531.161.080.790.58
Xylose0.260.370.060.661.210.970.340.26  0.841.310.920.971.020.921.403.190.87                  2.102.640.471.12      
Ribose0.790.530.490.791.261.500.940.81  0.921.280.980.980.991.000.921.210.93                  2.662.140.36       
Cellobiose0.730.670.851.291.130.920.660.68                             0.462.030.48       
Organic acids
3PGA                   0.750.330.331.001.330.460.640.780.710.710.981.661.901.741.310.400.600.60          
PEP                             0.631.721.241.411.71             
Pyruvate0.790.660.611.000.911.300.322.284.505.30                   0.771.181.310.900.840.661.331.668.343.052.310.70      
Citrate0.870.750.771.001.220.970.340.42  1.061.311.201.161.221.300.911.321.00          0.961.030.970.960.93      1.010.81     
Isocitrate   1.031.191.140.250.19                     1.060.780.981.210.88   3.272.770.46       
Fumarate0.971.011.600.931.522.462.634.60  1.321.581.050.750.851.011.612.441.09                  2.582.270.480.611.190.560.340.340.220.15
Lactate3.003.823.850.992.632.901.8511.30++++         2.331.661.460.761.201.660.660.660.760.7317.148.212.992.141.933.001.251.2574.031.751.752.222.70     
Malate0.130.100.350.740.751.000.701.17                     0.590.780.780.840.910.300.380.583.712.410.390.160.580.410.250.250.130.09
Succinate1.352.948.700.550.817.4916.4758.254.2011.401.411.570.881.761.231.851.581.551.24               2.402.001.002.6020.6830.273.9310.130.310.570.660.460.49
2-Oxoglutarate0.650.340.651.600.810.630.130.19                     0.821.111.191.610.83   2.502.993.781.091.49     
Glycerate0.010.090.971.231.380.770.460.40  0.961.371.071.011.041.064.054.361.01                  2.572.200.46  0.050.030.010.010.00
Acetate                2.501.671.03                            
Amino acids
Alanine2.271.041.001.521.412.011.1016.433.306.500.392.620.670.481.151.071.282.852.591.752.502.502.501.8510.001.601.301.161.005.205.093.081.551.7430.0020.0010.0041.0734.306.238.307.344.003.723.162.201.28
Glutamate0.090.110.190.760.761.460.861.941.000.700.551.321.010.761.100.641.123.560.910.170.420.861.000.910.120.720.720.800.802.192.531.911.501.350.500.630.752.200.283.031.204.230.931.470.801.000.60
Glycine1.061.001.330.941.075.012.1610.34  2.501.291.270.130.230.242.223.084.94          1.342.111.861.131.220.830.851.005.213.258.94 2.900.760.730.840.810.67
Phenylalanine0.980.751.560.680.561.071.165.79  0.721.260.940.910.960.911.141.781.76          1.801.281.081.301.122.402.001.202.550.412.380.773.240.581.371.411.320.68
Proline1.050.860.930.890.941.441.525.84  0.661.391.050.710.770.921.822.351.64                     0.035.44     
Serine0.180.180.030.910.780.440.741.52  0.351.060.590.590.930.611.332.251.82          1.201.721.491.161.100.650.570.575.242.224.410.552.430.160.160.230.120.06
Threonine1.170.820.960.960.841.011.032.31  0.531.380.870.610.990.681.292.371.23          1.011.110.870.870.952.001.001.004.060.433.810.48 1.131.411.240.930.56
Valine1.040.891.040.850.751.221.145.011.101.300.611.450.950.730.830.801.061.771.06          1.191.150.880.901.043.001.001.006.020.330.312.012.541.191.502.782.561.33
GABA1.541.851.921.021.955.202.064.831.001.300.891.650.920.820.890.800.932.580.838.006.002.401.601.405.001.002.001.601.401.440.910.911.081.0910.0010.0010.0017.633.587.627.589.25     
Aspartate0.690.590.680.300.061.070.750.33           0.180.180.450.540.810.170.450.570.660.62     1.501.001.000.330.010.210.060.661.242.171.281.410.85
Glutamine0.070.010.790.580.270.531.771.140.900.70                   1.931.310.950.871.520.420.570.852.850.090.080.08      
Tyrosine0.900.991.730.920.961.151.607.62                     0.660.570.570.690.822.221.161.059.714.297.784.385.082.904.484.763.671.86
Leucine1.070.910.970.860.961.021.105.85                     1.391.331.111.041.0220.0010.008.003.582.434.41  1.201.202.221.841.02
Isoleucine1.020.780.870.750.800.780.872.76  0.781.451.030.861.040.961.162.041.12          1.111.140.900.920.968.005.005.00 0.472.39  0.851.552.502.181.15
Tryptophan0.220.050.121.400.931.251.061.90                     1.061.250.770.730.892.211.161.052.682.063.76 1.53     
Asparagiine0.150.010.640.860.921.070.751.291.000.700.241.150.66                2.351.501.261.101.4915.0010.008.004.110.040.15 0.18     
Homoserine0.801.001.100.861.092.331.8015.20                             3.810.410.50       
lysine0.820.861.150.970.881.110.983.96  0.861.471.121.241.251.021.352.031.67          1.991.031.301.201.292.331.331.204.203.072.87  0.762.574.764.052.38
Histidine                             1.200.930.890.921.081.251.401.00          
Cysteine                                     2.932.330.47  0.200.280.190.100.06
Methionine0.890.120.970.550.310.771.113.08                     1.181.080.950.930.974.003.503.00          
Others
Putrescine1.040.931.031.060.760.790.231.06                                0.581.690.030.010.010.030.03
Urea   0.590.941.243.961.62  0.491.390.720.650.940.781.111.791.01                     0.770.54     
Allantoin0.981.081.590.731.021.871.584.28                             2.500.462.35       
Galactinol1.070.971.401.251.130.420.130.33                                0.19      
Glycerol          0.571.050.750.981.100.990.941.391.40                     1.290.94     
Mannitol0.090.261.000.540.661.091.080.40                                0.76      
Myo-Inositiol1.040.941.241.041.191.581.221.65  0.971.351.030.980.971.001.161.630.91                     0.902.95     
Sorbitol0.410.650.331.510.681.470.690.99                              3.583.320.20      
Phosphoric acid                   0.580.660.730.761.160.920.920.920.791.05           0.640.45     
AcCoA                   0.440.330.570.621.000.750.750.750.500.750.430.580.540.791.04             
Ethanol        1.501.30                        4.001.201.20 3.250.9328 1015182020

Results

Transcriptomic responses under anaerobic conditions for rice and Arabidopsis

To examine the transcriptomic response to anaerobic conditions, differential expression analysis (after false discovery rate correction) was carried out on publically available microarray data following exposure to anaerobic conditions for 2 h and 9 h for Arabidopsis (Table S5) and 3 h and 6 h for rice (Table S6). This design was based on comparing the responses of a plant that can endure – and even grow – under conditions with limited or no oxygen compared with a plant that does not respond well to such conditions and would eventually succumb if left for extended periods without oxygen (Pierik et al., 2005; Colmer & Voesenek, 2009). The fact that the two different species at different developmental stages are also compared implies that some of the differences may be species specific and/or developmental specific. However, it is the response of each species and/or tissue that is being assessed, not the basal transcriptome or metabolome. Thus, while similarities are likely to be relatively robust across several species, they alone cannot explain the differences in survival to low oxygen conditions as there is ‘great genetic diversity to survive flooding’ (Bailey-Serres & Voesenek, 2008). Rather, the differences in response between species and tissues, in combination with changes observed among a wide variety of species, have the potential to elucidate the physiological and morphological changes that occur to allow plants to endure and even grow under low oxygen conditions. For Arabidopsis, the microarray data were derived from the experiments carried out in the study by Branco-Price et al. (2008), which involved complete anaerobic treatment of 7-d-old seedlings. In order to carry out the most parallel examination possible, for rice, microarray data were derived from the experiments carried out by Narsai et al. (2009), which involved anaerobic treatment of rice embryos after 24 h of germination under aerobic conditions (see the Materials and Methods section). Analysis of these in parallel enabled the identification of similarities and differences in the transcriptomic response to anaerobiosis between rice and Arabidopsis (Fig. 1a).

Interestingly, despite the much smaller size of Arabidopsis genome and genes represented on the microarray genechip compared with rice (22 000 vs 54 000 respectively), it can be seen that there was a greater number of differentially expressed genes (DEGs) in Arabidopsis (5314 genes) compared with rice (4667 genes). A study examined the core response to hypoxia across four kingdoms and identified sets of hypoxia induced/reduced genes in both rice and Arabidopsis based on several studies for each species (Mustroph et al., 2010). In order to examine these genes in the context of this study, genes that were found to be induced/reduced under hypoxic conditions in three or more samples were matched to the differentially expressed genes in this study (last three columns; Tables S1, S2). It was seen that of the 4667 rice and 5314 Arabidopsis DEGs, 3656 rice genes and 549 Arabidopsis genes were also shown to be induced/reduced under hypoxia in more than three comparisons analysed by Mustroph et al., 2010. Note that many more comparisons were involved in the identification of the rice induced/reduced genes compared with Arabidopsis, therefore a cut-off of three comparisons for Arabidopsis resulted in the smaller number seen.

Irrespective of the number of DEGs, examination of orthologues revealed that over 77% of Arabidopsis genes and over 55% of the rice genes in each differentially expressed subset have orthologues in rice and Arabidopsis, respectively, which is significantly (< 0.01) larger than the proportion of orthologues in the genomes (61% and 37%, respectively) (Fig. 1b; Tables S3, S4). In order to test if the magnitude of change displayed a greater degree of orthology, genes changing by more than fourfold (highly induced/reduced) were found to display 63% orthology. This was also examined for rice and it was seen that of all differentially expressed rice genes, 53% had Arabidopsis orthologues. The genes induced/reduced by more than fourfold were also examined for orthology and it was seen that 50% had orthologues. Therefore, for both Arabidopsis and rice, it can be seen that there is no correlation with orthology and magnitude of induction/reduction. However, the responses of these orthologues appear not to be conserved, with < 15% of all genes showing both orthology and similarity in response (Fig. 1a,b; Tables S3, S4). For example, the 386 common rice and Arabidopsis genes downregulated under anaerobic conditions, represents 14% of all the rice genes downregulated following anoxic stress, while these make up 13% of all the Arabidopsis genes downregulated following anoxic stress (Common anaerobic Arabidopsis (A) and rice (R); Fig. 1a,b). While examining similarities in the transcriptomic responses between orthologous genes in rice and Arabidopsis, it was observed that some orthologues were regulated in an opposite manner between both species. To determine the extent of this contrasting response, the entire set of rice transcripts significantly upregulated in the anaerobic responsive sets were overlapped with the corresponding Arabidopsis transcripts that were significantly downregulated and vice versa (Opposite anaerobic A & R; Fig. 1a,b). Only a small percentage (4–7%) of transcripts that had orthologues was found to be regulated in an opposite manner, indicating that most genes in rice or Arabidopsis either had orthologues that were not significantly changing in abundance or did not have defined orthologues.

Given that the definition of an orthologue can be argued based of the extent of sequence similarity or other means, it was important to consider the overall response. Thus, a comparison of the response was carried out based on function using the pageman tool (Usadel et al., 2006), to classify the functions of the proteins encoded by the genes whose transcripts were altered under anoxia. The first four columns of the pageman output (Fig. 1c) represents the output using the common anaerobic A & R sets of genes. As expected, it can be observed that most of these genes were over-represented in the same functional groups across both species. For example, it was observed that the common genes upregulated under anaerobic conditions were largely enriched in photosynthesis, major CHO metabolism and protein degradation functions (Up, A & R; Fig. 1c). Similarly, the common genes downregulated under anaerobic conditions were largely enriched in cell wall and secondary metabolism functions (Down, A & R; Fig. 1c). Collectively, these results indicated that there are common functions affected in the anaerobic response, regardless of whether a species is tolerant or intolerant to anaerobic conditions.

To examine the overall response, independent of the orthologous genes showing common response, all genes that had orthologues that were oppositely regulated, unchanging in response or did not have defined orthologues (different to Arabidopsis; diff to A, different to rice; diff to R) were also analysed for over/under-represented functional categories (last four columns; Fig. 1c). In this way, it was seen that there were several examples of conservation in the over-/under-representation of functional groups indicating that although these genes may not fit the criteria for orthology, the overall response for these functional groups is conserved, for example for genes upregulated under anaerobic conditions, it can be seen that there is a conserved under-representation of secondary metabolism functions and over-representation of protein degradation functions (Fig. 1c, green shading). By contrast, it was seen that there was over-representation of signalling functions in the subset of genes upregulated under anaerobic conditions in rice, while this functional group is under-represented under anaerobic conditions in Arabidopsis. Notably, these corroborated observations by Mustroph et al. (2010), who also found that genes involved in regulatory functions displayed diverse responses between species. Examples of oppositely regulated genes included four zinc-finger domain containing transcription factors (LOC_Os01g15300.1, LOC_Os02g05692.1, LOC_Os01g57650.1, LOC_Os03g41390.1) and several kinases (e.g. LOC_Os01g74200.1, LOC_Os04g41160.1) that were upregulated in rice, while the Arabidopsis orthologues were downregulated (Tables S3, S4). Similarly, the downregulation of genes associated with signalling was over-represented in Arabidopsis and under-represented in rice (Fig. 1c, orange shading, Tables S3, S4). Other notable differences were in protein synthesis, protein post-translational modification and lipid metabolism. Rice was over-represented in several subsets of genes encoding functions associated with RNA transcription and regulation of transcription (Fig. 1c, orange shading, Table S3, S4).

Variations in metabolic responses under hypoxic conditions between different species

To expand the comparison carried out at a transcript level between rice and Arabidopsis, the responses of plants to various treatments in which the oxygen availability was reduced were compared between species at a metabolite level. The data for this comparison were obtained from investigations published by various research groups using different species, tissues and oxygen conditions (Table 1). These data were compared with each other in order to define general metabolic responses as well as to obtain insight in hypoxia induced metabolic changes that only occur under very specific conditions or in a few plant species only. Table 1 provides an overview of the changes in metabolite levels compared with the corresponding normoxic control treatments. The Table focuses on metabolites from primary carbon and nitrogen metabolism for which results were available from at least two independent experiments.

No metabolite behaved identically in all the different hypoxia treatments that were investigated, but similar changes were commonly observed for organic acids and amino acids, whereas the levels of sugars and sugar-phosphates did not change or behaved rather variably. Aspartate decreased in nearly all hypoxia-treated plant samples; the concentration of aspartate only increased in phloem sap collected from nonphotosynthetically active seedlings of Ricinus communis in an environment with only 0.5% (v : v) oxygen. Some increase was also observed for aspartate in Chlamydomonas. In both cases the extent of the increase was small, only 50% in R. communis and in Chlamydomonas it only exceeded twofold at a single time-point. Alanine levels increased during low-oxygen treatments, although through time the accumulation of alanine diminished in some experiments (rice seedlings after switching to anoxia, poplar and Chlamydomonas). In Arabidopsis seedlings that were kept at hypoxia for < 2 h, alanine concentrations were affected, although in these experiments changes were not consistent among the treatments. The amounts of lysine and tyrosine increased in many experiments – lysine concentrations increased in all experiments that measured it except upon switching rice seedlings from aerobic to anaerobic conditions. However, the increase in lysine may take some time to occur and may possibly be missed in short-term experiments as an increase in lysine concentration was also observed in rice seeds when germinated under anaerobic conditions for 48 h, and in Arabidopsis the increase was most strongly observed in roots after 48 h under hypoxia. For tyrosine an increase in concentration was observed in all experiments that measured it except potato and Ricinus at 6% oxygen. Amounts of lactate generally increased among the different experiments, and only in Arabidopsis silique wall tissue did lactate concentrations decrease at oxygen concentrations of 1% (v : v) and above. Data on the production of ethanol were rather limited within the datasets used for the current analysis, as ethanol is not generally detected by the metabolite profiling approaches that were used in the experiments that are analysed in the current survey. Nevertheless, some investigations mentioned changes in ethanol levels as detected by enzymatic analysis techniques. In all those cases, ethanol accumulated during hypoxia. Clear changes in metabolite levels were observed for γ-aminobutyric acid (GABA) and succinate. Both these metabolites increased in the majority of the treatments that were listed in the present survey. Only in Chlamydomonas did the concentration of succinate decrease.

In summary, this extensive survey of metabolic responses to hypoxia showed that the well-known fermentative pathways that lead to the production of lactate or ethanol are common among plant species and are induced as a general response by many different treatments that lead to the reduction of the oxygen availability. Furthermore, accumulation of alanine, GABA and succinate and a decrease in the amount of aspartate were also commonly observed among different hypoxic treatments and species. Lysine and tyrosine increased in many experiments. Sugar and sugar phosphate levels behaved rather inconsistently, and changes seemed to be more specific for individual species, experiments or time-points.

Pathway visualization for genes involved in glycolysis, tricarboxylic acid cycle (TCA) cycle and amino acid metabolism during hypoxia in rice, Arabidopsis and poplar

The hypoxia-specific metabolic changes described earlier, were further characterized by comparing changes in transcript abundance of the genes that are involved in the metabolic pathways of primary carbon and nitrogen metabolism (Fig. 2). For this, gene expression information was used from transcriptome analyses on rice (Narsai et al., 2009; Tables S2, S6) and Arabidopsis (Branco-Price et al., 2008; Tables S1, S5), which were also discussed in the previous section. Furthermore, information on transcript abundance from waterlogged poplar plants (Kreuzwieser et al., 2009; Table S7) was incorporated.

Figure 2.

Parallel display of transcripts and metabolites for starch-sucrose metabolism, glycolysis, the tricarboxylic acid cycle (TCA) cycle, γ-aminobutyric acid (GABA) shunt, mitochondrial respiratory chain and amino acid metabolism for Arabidopsis, rice and poplar. Significant fold changes in transcripts and metabolites were displayed on a custom pathway picture using the mapman tool (the Arabidopsis Information Resource or TAIR9 annotation for Arabidopsis). For some Arabidopsis genes, the TAIR7 annotation file was also used as this allowed these genes to be annotated more comparably to rice and poplar, the expression of these additional genes, for example, enolase, are indicated in black boxes. Significant changes in metabolite levels are also indicated based on the colour of the text for the metabolite name. Enzymatic conversions between metabolites are indicated by arrows and enzyme names. Changes in transcripts encoding these enzymes are indicated in the boxes next to the enzyme names and correspond to the changes in response to switching from aerobic to anaerobic conditions. For Arabidopsis, rice and poplar, only the maximum/minimum fold change is shown in response to switching from aerobic to anaerobic conditions for Arabidopsis (2 h and 9 h), rice (3 h and 6 h) and poplar (5 h, 24 h and 168 h). In most cases, the above enzymes are encoded by small gene families. However, in some cases the enzymes are not distinguished and a more overall annotation of the relevant transcripts is shown in italics (e.g. starch degradation). Note that this also applies for transcripts encoding components of the mitochondrial electron transport chain (CI–CV) where only the transcript levels for different nuclear-encoded subunits are presented. For transcripts, changes are represented by shading where the colour saturates at a fold change (FC) value of 4.

In general, changes in gene expression appeared rather uniform between species. Genes encoding for enzymes involved in sugar and sugar-phosphate metabolism, including glycolysis, were generally upregulated during hypoxia in all three species, with only an exception for invertase. These genes were down-regulated upon the various hypoxic stress treatments. Also most pyrophosphatase genes were down-regulated. Transcript levels of genes encoding TCA-cycle enzymes (including pyruvate dehydrogenase) as well as genes encoding enzymes of amino acid metabolism were generally down-regulated. An exception to this rule was the expression of genes encoding alanine aminotransferase and the fermentation enzymes pyruvate decarboxylase and lactate dehydrogenase in Arabidopsis: transcript of these genes went up during hypoxia. Rather contrasting expression profiles were obtained between the three species for genes encoding components of the mitochondrial electron transport chain. In Arabidopsis, genes were mainly upregulated, whereas in poplar or rice these genes were downregulated.

A recent study (Christianson et al., 2010a,b) compared gene expression responses to hypoxia in cotton, Arabidopsis and poplar in order to gain an insight into the core oxygen responsive genes in dicots. The 122 genes defined as core oxygen responsive in this study were isolated from the Arabidopsis and rice datasets analysed in this study and examined (Table S8). All the Arabidopsis DEGs from Table S1 and most of the rice orthologues for these genes from Table S2 that matched these core genes, showed a comparable response, corroborating our findings and those by Branco-Price et al., 2008;. Furthermore, given that these responses were consistent despite differences in the exact growing conditions and tissue type, this indicates that these genes likely form part of an overall transcriptomic response to hypoxia. Notably, however, it was seen that five rice genes showed an opposite response from the core dicot response (in yellow; Table S8). These genes encoded an HSP23, a phosphofructokinase encoding gene, 2 genes encoding hydrolases and a gene annotated as having a role in DNA-damage repair (Table S8).

Comparison between metabolites and transcripts

To determine common regulation patterns between metabolites and transcripts, changes in their respective levels were compared (Fig. 2). Metabolic changes correlated well with the transcriptional regulation of genes in Arabidopsis that encode enzymes of which the given metabolites are either substrate or product: ethanol accumulation correlated with the upregulation of genes encoding pyruvate decarboxylase and alcohol dehydrogenase; lactate accumulation correlated with the increase of lactate dehydrogenase expression; expression of alanine aminotransferase genes was upregulated and so was the level of alanine; and the decrease in the level of aspartate correlated with the downregulation of the expression of most genes encoding for enzymes involved in both aspartate synthesis and breakdown. In contrast to these correlations between gene expression and metabolite levels in Arabidopsis, no such clear correlation could be observed in poplar or rice. In these species, the expression level of many of the genes mentioned earlier showed no change in expression during the hypoxic treatment. The expression levels of some genes encoding alcohol dehydrogenase were even downregulated, which is opposite to what was observed in Arabidopsis. A similar difference between species was observed for the expression of genes encoding homoserine synthetase that were downregulated in Arabidopsis but upregulated in rice and poplar. A comparison between the three species is complicated by the different number of genes encoding various enzymatic steps in the three species examined, their basal levels of expression and the fact that in some cases genes encoded isoforms for some enzymes that respond in the opposite manner. In light of this, downregulation of pyrophosphatases, serine synthesis, aspartate aminotransferase, pyruvate dehydrogenase, glutamine synthase and various enzymes of the TCA cycle, and the upregulation of pyruvate decarboxylase and transcripts of genes involved in sugar and sugar-phosphate metabolism appear to represent common responses in all three species examined.

Regulation of transcripts encoding nonsymbiotic haemoglobins

A special class of hypoxia responsive proteins that has always obtained special attention because of their oxygen binding properties are the haemoglobins. In plants, this protein family can be separated into four subclasses: the symbiotic haemoglobins, which are only expressed in nodules of legumes and some other species, and three groups of nonsymbiotic haemoglobins; class-1, class-2 and class-3 (Smagghe et al., 2009). Comparison of transcript abundance of haemoglobin orthologues in rice and Arabidopsis revealed induction of the expression of all haemoglobin genes in Arabidopsis, whereas in rice the opposite was observed (Fig. 3). In Arabidopsis, the class-1 haemoglobin was most rapidly and most strongly expressed, whereas changes were slow and less extreme for rice haemoglobin expression and Arabidopsis genes encoding class-2 and class-3 haemoglobins. In poplar, when expression was detected it decreased in a similar manner as was observed in rice (Fig. 3).

Figure 3.

Overview of the response of nonsymbiotic haemoglobins to anaerobiosis. The nonsymbiotic haemoglobin genes from rice and Arabidopsis were clustered according to their phylogenetic similarity with the soybean leghaemoglobin gene as a reference. For Arabidopsis the time points are 2 h and 9 h after transfer from normoxia to anaerobic conditions, for rice is the time-points are 3 h and 6 h after transfer from normoxia to anaerobic conditions. Numbers in red indicate an increase in transcript abundance, numbers in blue indicate a decrease in transcript abundance and 0 in black indicates expression too low to detect on microarrays or no change. Poplar annotation for nonsymbiotic haemoglobins are indicated, but only expression of one gene (POPTR_0006s26070) was detected. The classes of nonsymbiotic haemoglobins are indicated after Garrocho-Villegas et al. (2007) and Smagghe et al. (2009).

Discussion

The diversity of mechanisms employed by plants to withstand anoxic conditions is mirrored in the diversity of the responses observed between rice and Arabidopsis at a transcriptome level. Some of the differences observed likely result from the 140 million yr of evolution that separate these plants (Paterson et al., 2010), the different experimental conditions used and different tissues examined (Branco-Price et al., 2008; Narsai et al., 2009). However, it is worth noting that despite these differences, there is a common response in primary carbon metabolism indicating that the different experimental conditions are imposing a similar stress on the plants examined. This common response is even evident outside the plant kingdom, further emphasizing that a common stress is being applied and sensed by the cells (Mustroph et al., 2010). Thus examination of the differences in response to a common stress will elucidate the diversity of responses between species and tissues and may help to explain the differences in ability to withstand anoxic conditions. However, the well-described alterations in carbohydrate metabolism that appear to be a common response, although likely contribute to tolerance – as confirmed in a variety of studies by under-expression or over-expression of various genes encoding these components (Christianson et al., 2010a,b) – do not explain adaptation to, or survival in anoxic conditions. Rather it is the differences that need to be dissected to identify factors or processes that contribute to survival. This has been elegantly demonstrated in a number of independent studies in rice where the differences between varieties has led to the identification of three genes (Cipk15, Snorkel and Sub1a) linked to tolerance to anoxia (Xu et al., 2006; Hattori et al., 2009; Lee et al., 2009). Thus even within the same species it is the difference in response that has been informative in explaining the ability to survive.

Species variation of the transcriptional response to low oxygen

The global analysis of transcriptional changes induced by low oxygen identified the functional grouping of transport, protein synthesis, targeting and post-translational modification as well as signalling and regulation of transcription (different families of transcription factors) as differences between Arabidopsis and rice (Fig. 1). It is important to remember that a variety of regulatory processes from signalling, to transcription and translation all likely combine to achieve tolerance. While the over-expression of a variety of transcription factors in Arabidopsis confer greater tolerance to anoxia (Banti et al., 2010; Hinz et al., 2010; Licausi et al., 2010), it is clear from global and specific studies that post-transcriptional regulation occurs. Selective translation of mRNAs under anoxia has been shown to occur in Arabidopsis (Branco-Price et al., 2008), and with the transcription factor HsfA2, where transcript is increased under anoxia but not protein (Banti et al., 2010). However, pretreatment with heat, where this transcript is increased and translated, leads to anoxia tolerance (Banti et al., 2010). The different responses revealed in the present study between rice and Arabidopsis of genes encoding proteins involved in protein synthesis and modification (Fig. 1) may be the basis of such differences between expression at a transcript level and at a protein level, and warrants further investigation in both models.

Comparison of transcriptional changes of genes involved in primary metabolism induced in various plant species by a reduction of the oxygen availability showed a clear separation between the transcript abundances (Fig. 2): most genes that encode enzymes that are involved in sugar metabolism and glycolysis were generally upregulated, whereas transcripts of genes involved in amino acid synthesis and the TCA cycle were downregulated. The separation of transcriptional regulation of glycolysis and TCA cycle occurs at the site of pyruvate dehydrogenase (Fig. 2). This observation can be explained by the generally occurring anoxic response of aerobic organism to induce the so-called Pasteur effect (Summers et al., 2000), which is the activation of the glycolytic flux and fermentation to increase the amount of ATP that is released by pyruvate kinase and phosphoglycerate kinase.

The expression profile of invertase genes contrasted with the other transcriptional changes of the glycolytic pathway, as most of the other genes are upregulated (Fig. 2). Apparently, sucrose degradation via sucrose synthase is favoured above the reaction catalysed by invertase because the invertase pathway is energetically less efficient compared with sucrose synthase (Bologa et al., 2003). Sucrose degradation via the sucrose synthase pathway requires pyrophosphate (PPi) as an alternative energy donor. The downregulation of genes encoding for pyrophosphatase in Arabidopsis, rice and poplar (Fig. 2) may explain why PPi is maintained at high concentrations in hypoxic tissues, in contrast to the progressive decrease of the energy status of the adenine nucleotide system (Geigenberger, 2003). The expression of enolase is also downregulated in rice and poplar. This is probably because a decrease in PEP will support the downregulation of the shikimate pathway (see Fig. 2).

Glycolysis and the TCA cycle are commonly linked via AlaAT during hypoxia

Accumulation of alanine during hypoxia was confirmed for nearly all conditions and species that were analysed. For L. japonicus, a metabolic pathway was revealed to explain the role of alanine production during waterlogging (Rocha et al., 2010a). According to the metabolic model depicted in Fig. 4, pyruvate reacts together with glutamate to form alanine and 2-oxoglutarate. This reaction is catalysed by the enzyme alanine aminotransferase, which has been shown to be strongly upregulated upon waterlogging (Rocha et al., 2010b). The glutamate that is used as substrate is provided by aspartate aminotransferase, which explains the generally observed decline of aspartate upon hypoxic stress. Subsequently, 2-oxoglutarate is used as substrate of the TCA cycle enzyme 2-oxoglutarate dehydrogenase and reacts further to succinate and ATP using NAD as cosubstrate. The NAD that is required for this reaction is provided by the TCA-cycle enzyme malate dehydrogenase, which under hypoxic conditions catalyses the reaction from oxaloacetate to malate in the reversed direction to what is normal in a normoxic TCA cycle reaction pathway. Subsequently, malate can be metabolized into either succinate via fumarase and succinate dehydrogenase, as has been shown for the green alga Selenastrum minutum (Vanlerberghe et al., 1989) or into pyruvate via NAD-dependent malic enzyme (Tronconi et al., 2008). Most likely, these two pathways act in parallel during hypoxic conditions, although their individual contribution might differ among species and conditions.

Figure 4.

Schematic overview of the metabolic pathway that is active during hypoxia, linking glycolysis via the enzyme alanine aminotransferase (AlaAT) with the tricarboxylic acid cycle (TCA) reactions, arranged in a noncyclic mode. The visualization of the metabolic changes (red indicates increase; blue indicates decrease) is based upon the data depicted in Table 1. Abbreviations used: ADH, alcohol dehydrogenase; AlaAT, alanine aminotransferase; AspAT, aspartate aminotransferase; FM, fumarase; LDH, lactate dehydrogenase; MDH, malate dehydrogenase; ME, NAD-dependent malic enzyme; OGDH, 2-oxoglutarate dehydrogenase; PDC, pyruvate decarboxylase; SCS, succinyl-CoA synthetase; SDH, succinate dehydrogenase.

A noncyclic mode of operation of the TCA cycle as described in the preceding text occurs commonly and depends on the metabolic demands of the cell (Sweetlove et al., 2010). During anoxia, when ATP production via oxidative phosphorylation is impaired, the reorganization of the TCA cycle optimizes the minimal production of ATP. The additional ATP production from the TCA reactions doubles the total ATP gained per hexose unit in the absence of molecular oxygen from two ATP units (from glycolysis only) to four.

The present comparison of metabolic profiles reveals that the detour from pyruvate via alanine and 2-oxoglutarate to succinate, as induced by hypoxia, is not just a specific adaptation to hypoxia in L. japonicus, for which this pathway was originally described (Rocha et al., 2010a) – it generally occurs in other plant species as well. Therefore, this metabolic pathway can be considered as a true hypoxic response with similar importance for hypoxic stress resistance as the fermentative pathways. In the hypoxia tolerant species rice and poplar, the initial accumulation of alanine declined again with time. In poplar this might be explained by an efficient export of alanine from the hypoxic tissues (roots) to other plant organs that are well aerated (shoot). This kind of metabolite export from hypoxic tissues could hold at least as long as adequate import of metabolic substrate is also sustained. Such metabolic differences between species may reflect a specific adaptation strategy to improve tolerance to anoxia.

Not only alanine accumulates during hypoxia: the level of GABA also increases in nearly all experiments that were evaluated here. The production of GABA from glutamate is catalysed by the enzyme glutamate decarboxylase (GAD) and uses protons as cosubstrate. Therefore, this reaction could reduce cellular acidification during hypoxia. Indeed, activation of GAD by low pH was shown to be involved in the regulation of GAD (Carroll et al., 1994; Crawford et al., 1994), which also explains why the transcript level of GAD is not consistently increasing in the different experiments (Fig. 2). Other amino acids that are consistently upregulated among species during hypoxia are tyrosine and lysine, although it is still unclear what reason this might have.

While changes in various metabolic processes outlined above are required to survive and/or grow under hypoxic conditions these changes represent an endpoint of a variety of signal transduction processes. As outlined earlier, differences to flooding tolerance in rice, either as a result of the quiescent or fast escape strategy are based on regulatory factors (Bailey-Serres & Voesenek, 2008; Nagai et al., 2010). Similarly, the role of CIPK15 acts to integrate sugar and oxygen signalling (Lee et al., 2009). In Fig. 1, based on functional categorization in pageman bins it was evident that there were differences in the category of signalling between Arabidopsis and rice, with rice over-represented in genes that were upregulated in this functional group and Arabidopsis under-represented in genes in this group.

Haemoglobin expression and probably nitric oxide (NO) signalling varies strongly among species

One signalling pathway that has been implicated in a variety of stresses in plants Wendehenne et al., 2004), including exposure to anoxia (Igamberdiev & Hill, 2009; Igamberdiev et al., 2010), involves NO. The production of NO increases during hypoxia because of the accumulation of nitrite at low pH, which leads to increased nitrite reduction by the enzyme nitrate reductase (Ferrari & Varner, 1971; Rockel et al., 2002). Furthermore, it has been proposed that under anaerobic conditions nitrite is converted to NO by cytochrome oxidase and cytochrome bc1 complexes and so plays a role in ATP synthesis under anaerobic conditions (Igamberdiev & Hill, 2009; Igamberdiev et al., 2010). The rate of production of NO under anoxia is not clear, likely because of scavenging by a variety of mechanisms (Gupta & Kaiser, 2010). Although the function of the various nonsymbiotic haemoglobins has not yet been revealed unequivocally, it has been reported that class-1 haemoglobin can play a role in scavenging NO and thus play an important role in regulating NO signalling cascades that can interact with a variety of hormone signalling pathways (Igamberdiev & Hill, 2009). However, examination of the transcript abundance of various haemoglobin genes in Arabidopsis, rice and poplar reveals that while transcripts abundance for haemoglobin genes, irrespective of class were increased in Arabidopsis, transcript abundance decreased in both rice and poplar (Fig. 3). As the last two are more resistant to anoxia this raises the possibility that NO production and scavenging differs between various species. Consistent with this trend is that in Arabidopsis transcript abundance for AOX is induced while in rice it is decreased (Fig. 2), and remained unchanged in poplar. As AOX has been shown to be induced by NO in Arabidopsis (Hunag et al., 2005), this suggests that NO signalling is active in Arabidopsis under anoxia but not in rice.

Future perspectives

Although the last few years have seen several major advances in understanding the molecular mechanism that results in the response to anoxia, a number of question are still outstanding. Very little is known about the sensing mechanisms for low oxygen in plants. Nitric oxide is suggested to be involved in this, but the differences as well as the overlap in response to NO and anoxia in both rice and Arabidopsis need to be elucidated.

Clearly post-transcriptional regulation has emerged as a major site of regulation in response to anoxia. However, it remains to be elucidated whether the selective translation of mRNAs that occurs in Arabidopsis also occurs in other species, particularly in rice. The elucidation in Arabidopsis that cell-specific translatomes occur upon exposure to even brief periods of hypoxia reveal that cell and organ specific responses occur within a single species (Mustroph et al., 2009). It is yet not clear if the selective translation of mRNAs actually leads to sensitivity or resistance to anoxia. Furthermore, while genome-wide studies at a transcript and protein level have been carried out in a variety of studies, data on proteome changes in response to anoxia would provide information on transcriptional and post-transcriptional responses.

Finally, while several transcription factors that confer tolerance to anoxia have been identified, the regulatory network that changes the transcript abundance of > 5000 genes in a species-specific manner needs to be resolved. This goes beyond the identification of single transcription factors and cognate binding sites, but requires further analyses to be carried out with multiple factors. Some steps towards this understanding have been achieved by prediction (Narsai et al., 2009), and experimental approaches (Licausi et al., 2011) but sustained efforts will be required to achieve an understanding that allows rational approaches to be undertaken to transfer such traits to desired crops.

Acknowledgements

This work was supported by the Deutsche Forschungsgemeinschaft (grants to PG and JTvD) and by an Australian Research Council Centre of Excellence Grant CEO561495 and the Western Australian State Government Centres of Excellence scheme (JW and RN).

Ancillary