Multilevel genomic analysis of the response of transcripts, enzyme activities and metabolites in Arabidopsis rosettes to a progressive decrease of temperature in the non-freezing range

Authors


M. Stitt. Fax: +49 331 567 8101; e-mail: mstitt@mpimp-golm.mpg.de

ABSTRACT

This paper characterizes the transcriptional and metabolic response of a chilling-tolerant species to an increasingly large decrease of the temperature. Arabidopsis Col-0 was grown at 20 °C and transferred to 17, 14, 12, 10 or 8 °C for 6 and 78 h, before harvesting the rosette and profiling >22 000 transcripts, >20 enzyme activities and >80 metabolites. Most parameters showed a qualitatively similar response across the entire temperature range, with the amplitude increasing as the temperature decreased. Transcripts typically showed large changes after 6 h, which were often damped by 78 h. Genes were induced for sucrose, proline, raffinose, tocopherol and polyamine synthesis, phenylpropanoid and flavonoid metabolism, fermentation, non-phosphorylating mitochondrial electron transport, RNA processing, and protein synthesis, targeting and folding. Genes were repressed for carbonic anhydrases, vacuolar invertase, and ethylene and jasmonic acid signalling. While some enzyme activities and metabolites changed rapidly, most changed slowly. After 6 h, there was an accumulation of phosphorylated intermediates, a shift of partitioning towards sucrose, and a perturbation of glycine decarboxylation and nitrogen metabolism. By 78 h, there was an increase of the overall protein content and many enzyme activities, a general increase of carbohydrates, organic and amino acids, and an increase of many stress-responsive metabolites including raffinose, proline, tocopherol and polyamines. When the responses of transcripts and metabolism were compared, there was little agreement after 6 h, but considerable agreement after 78 h. Comparison with the published studies indicated that much, but not all, of the response was orchestrated by the CBF programme. Overall, our results showed that transcription and metabolism responded in a continuous manner across a wide range of temperatures. The general increase of enzyme activities and metabolites emphasized the positive and compensatory nature of this response.

INTRODUCTION

The response of plants to low temperature is an important determinant of their ecological range (Boyer 1982; Larcher 1995). Temperatures change within minutes because of fluctuations of the light intensity, hours as a result of the diel light/dark cycle, days as a result of changing weather and weeks to months as a result of seasonal changes (Larcher 1995). Temperature affects almost all metabolic and cellular processes (Larcher 1995; Atkin & Tjoelker 2003; Atkin et al. 2006). The typical Q10 for protein-dependent catalysis is between 2 and 3. Membrane-based processes are additionally affected by changes in the physical properties of lipids. Adaptation to low temperature is often divided into freezing resistance to allow survival at sub-zero temperatures, and chilling tolerance to allow survival and growth at low but non-freezing temperatures.

Species from tropical or subtropical zones typically show wilting, chlorosis, necrosis, slow growth or death at temperatures as high as 10–12 °C (Lyons 1973; Wang 1990; Van der Ploeg & Heuvelink 2005). Temperatures in this range can result in photoinhibition (Lyons 1973; Huner, Öquist & Sarhan 1998; Allen & Ort 2001). Sucrose and starch synthesis are inhibited more strongly than carbon fixation, leading to sequestration of orthophosphate (Pi) in phosphorylated intermediates and a transient Pi limitation of photosynthesis (Leegood & Furbank 1986; Sharkey et al. 1986; Stitt & Grosse 1988). As low temperatures lead to carbohydrate accumulation (Guy 1990; Larcher 1995; Strand et al. 1999; Artuso et al. 2000; Hurry et al. 2000; Jompuk et al. 2005; Van der Ploeg & Heuvelink 2005), carbon utilization must be inhibited even more strongly than photosynthesis. Chilling inhibits phloem transport (Krapp & Stitt 1995), respiration (Huner et al. 1998; Taylor, Day & Millar 2002), delays leaf development and interferes with plastid biogenesis, leading to delayed greening, chlorosis and thickening or deformation of new leaves (Sowinski et al. 2005; Van der Ploeg & Heuvelink 2005; Atkin et al. 2006). Chilling leads to oxidative stress (Kocsy, Galiba & Brunold 2001; Foyer et al. 2002; Zhou et al. 2004; Aroca et al. 2005; Mora-Herrera et al. 2005), leakage of calcium and other ions across membranes (Lyons 1973; Huner et al. 1998; Knight & Knight 2000), reduced water conductivity (Melkonian, Yu & Setter 2004; Aroca et al. 2005; Lee et al. 2005b) and wilting (Bloom et al. 2004). As many crops originate from warm climes (Larcher 1995; Lee et al. 2002b; Venema et al. 2005), chilling tolerance is a major trait for plant breeding. Genetic diversity from related chilling-tolerant species is being introduced into breeding lines, but the genes responsible have not been identified (Goodstal et al. 2005; Jompuk et al. 2005; Pimentel et al. 2005; Van der Ploeg & Heuvelink 2005; Venema et al. 2005).

The physiological and molecular basis of freezing tolerance has been intensively researched, with especially detailed studies in Arabidopsis (Xin & Browse 1998; Thomashow 1999, 2001; Stitt & Hurry 2002). Freezing tolerance is promoted by a process termed cold acclimation, which occurs at low but non-freezing temperatures (Guy 1990; Larcher 1995). Many studies of freezing tolerance have therefore investigated the metabolic and transcriptional response after decreasing the temperature to 0–4 °C. While some aspects of this response may be related to the development of freezing tolerance, others may be relevant for survival at less extreme temperatures.

The physiological and metabolic changes after transferring plants to 0–4 °C include changes in lipid composition and membrane fluidity (Somerville & Browse 1996), induction of genes encoding proteins that modify the physical characteristics of membranes (Steponkus et al. 1998), accumulation of compatible solutes like sucrose (Strand et al. 1999), raffinose and proline (Guy 1990; Cook et al. 2004), and an increase in the capacity to detoxify active oxygen species (Kocsy et al. 2001; Iba 2002; Yoshimura et al. 2004). Leaves that develop at low temperatures are thicker and have a higher protein content, even in chilling-tolerant species (Strand et al. 1999; Hurry et al. 2000; Pinedo et al. 2000; Atkin et al. 2006). This increased protein content may compensate for the slower rate of catalysis at low temperatures. For example, leaves that develop at 4 °C have higher concentrations of Calvin cycle enzymes, allowing higher rates of photosynthesis on a fresh weight (FW) or leaf area basis (Strand et al. 1999; Hurry et al. 2000; Atkin et al. 2006).

Transfer to 0–4 °C modifies the levels of hundreds of transcripts (Thomashow 2001; Fowler & Thomashow 2002; Hannah, Heyer & Hincha 2005; Lee et al. 2005a; Vogel et al. 2005). The CBF family of transcription factors plays an important role in the orchestration of this response (Thomashow 1999, 2001). CBF1, CBF2 and CBF3 belong to the AP2/ERF domain family of DNA-binding proteins (Riechmann & Meyerowitz 1998). CBF transcripts rise rapidly and transiently after transfer to low temperature. This increase is mediated by phosphorylation of the cMYC-like basic helix loop transcription factor ICE1 (Chinnusamy, Zhu & Zhu 2006; Nakashima & Yamaguchi-Shinozaki 2006), allowing it to bind to regulatory ICEr sequences in the CBF promoters (Chinnusamy et al. 2003; Zarka et al. 2003; Lee et al. 2005a; Benedict et al. 2006a). The CBF proteins recognize a cis-acting regulatory element known as the C-repeat/dehydration response element (CRT/DRE) that is present in the promoters of many cold-inducible genes (Baker, Wilhelm & Thomashow 1994; Yamaguchi-Shinozaki & Shinozaki 1994; Stockinger, Gilmour & Thomashow 1997). The CBF regulon contains most of the genes that are strongly induced after transfer to 4 °C (Vogel et al. 2005). Over-expression of CBF genes leads to constitutive expression of CBF-targeted genes (Jaglo-Ottosen et al. 1998; Liu et al. 1998; Gilmour, Fowler & Thomashow 2004), production of cryoprotective polypepetides like COR15a (Steponkus et al. 1998), increased levels of cryoprotective solutes like sucrose, proline and raffinose (Gilmour et al. 2000, 2004; Cook et al. 2004), and increased freezing tolerance in non-acclimated plants (Jaglo-Ottosen et al. 1998; Liu et al. 1998).

CBF-independent pathways also contribute to low temperature responses (Fowler & Thomashow 2002). The Arabidopsis mutants eskimo-1 (Xin & Browse 1998) and ada2 (Vlachonasios, Thomashow & Triezenberg 2003) show enhanced freezing tolerance, even though expression of CBF pathway components is unaltered. Addition of abscisic acid (Chen & Gusta 1983; Mäntylä, Lång & Palva 1995) or salicylic acid (Tasgin, Atici & Nalbantoglu 2003; Mora-Herrera et al. 2005) can improve freezing tolerance. Constitutive freezing tolerance can be achieved by over-expressing the transcription regulator ZAT12 (Vogel et al. 2005). A recent bioinformatics analysis indicates that ZAT12 and ABA interact with ICE1-CBF signalling (Benedict et al. 2006a). Studies of pho1 and pho2 mutants indicate that changes of Pi modulate some of the changes in photosynthetic and carbohydrate metabolism at low temperature (Hurry et al. 2000).

It is unclear which of the responses that are seen at 0–4 °C occur only at very low temperatures, and which also occur at less extreme temperatures. A comprehensive analysis of the response of a chilling-tolerant species to a small decrease of the temperature would provide a benchmark data set that would aid interpretation of studies of chilling-sensitive species. The CBF family is transiently induced at temperatures as high as 14 °C (Zarka et al. 2003), implying that the CBF programme may contribute to adaptation to small decreases of the temperature. A contrasting conclusion was reached in a study that investigated the response of 8000 transcripts 2 d after transferring Arabidopsis from 24 to 13 °C (Provart et al. 2003). The response was rather different to that reported in earlier experiments in which Arabidopsis seedlings were transferred to 4 °C for 1 d (Fowler & Thomashow 2002; Kreps et al. 2002). Provart et al. (2003) suggested that a 2 d treatment at 13 °C initiated responses that promoted metabolism and growth at chilling temperatures, but did not induce processes that were required for survival at sub-zero temperatures. However, as they pointed out, there could be other explanations for the discrepancies between the 13 and 0–4 °C data sets. Comparison is complicated because they have been generated using different array technologies. Comparison of single time points could be misleading, because the temporal kinetics may be temperature dependent. Comparison of data sets from different studies can also be complicated by experiment-to-experiment variation. Changes of expression in soil-grown plants have been analysed at various times after transfer to 0–4 °C using 8K (Fowler & Thomashow 2002) and 22K (Hannah et al. 2005; Vogel et al. 2005) arrays. The response of seedlings on non-supplemented and sugar-supplemented agar medium has been analysed using the 22K ATH1 arrays at various times after transfer to 0–4 °C (Lee et al. 2005a; Vogel et al. 2005). There are discrepancies between the responses of soil and nutrient medium-grown plants (Vogel et al. 2005) and even between plants grown in similar set-ups; for example, Lee et al. (2005a) reconfirmed only a third of the genes that were shortlisted as cold responsive by Vogel et al. (2005).

The following experiments investigate the short- and mid-term response of Arabidopsis to a progressive decrease of the temperature. The main aims were to investigate if there were strong discontinuities in the response, to identify the major metabolic and physiological components, and to investigate how far the response resembled that expected from the CBF programme. We show that a small and large decrease of the temperature leads to qualitatively similar responses, but the amplitude increases at lower temperatures. Inspection of the responses of the CBF regulon (Fowler & Thomashow 2002; Vogel et al. 2005) reveals that the CBF programme mediates much but not all of the response to a small decrease of the temperature. The results identify many examples where changes of transcripts correlate with changes in metabolism, and highlight changes in sucrose and raffinose metabolism, photorespiration, organic acid and nitrogen metabolism, mitochondrial electron transport and protein synthesis as potentially important components of the chilling response.

MATERIALS AND METHODS

Plant growth conditions

Arabidopis thaliana ecotype Col-0 was grown on soil in a 12 h light/12 h dark cycle, at a light intensity of 130 µmol m−2 s and at a constant temperature of 20 °C (Thimm et al. 2004). Experiments started after 4 weeks, at which time flowering had not commenced.

Experimental design and sampling

After 4 weeks ambient growth, Arabidopsis plants were simultaneously moved from 20 to 20, 17, 14, 12, 10 and 8 °C, 4 h after the beginning of the light period. A first batch of plants was harvested 6 h later. A second batch of plants was kept for further 72 h under the same range of temperatures, and then was harvested. All harvest times were therefore 10 h into the light period. Plants were harvested by transferring rosettes into liquid nitrogen under ambient irradiance. Typically, five replicate samples, each containing three rosettes, were collected. The same material was used for metabolite, enzyme activity and expression profiling; samples were analysed separately for metabolites and enzymes, and were pooled for transcript profiling. Two independent biological experiments were carried out at 6 month intervals.

RNA preparation, array hybridization, data analysis and MapMan display

RNA preparation, array hybridization and data analysis was carried out as in Thimm et al. (2004). The normalized signal intensities for all 22 746 ATH1 probe sets from our 24 arrays are compiled in Supplementary Table S1a. The raw Affymetrix signals (CEL files) were processed using log scale robust multi-array analysis (RMA) software. It was based on the quantile normalization method and had better precision than MicroArray Suite 5.0 (MAS5, Affymetrix, Santa Clara, CA, USA) and dCHIP (http://www.dchip.org/), especially for low expression values (Irizarry et al. 2003). The averaged signals for a given treatment were filtered using a LIMMA (Smyth 2005) value of P < 0.05 and were imported into the MapMan software, which converts the data values to a false colour scale and paints them out onto the diagrams (Thimm et al. 2004; Usadel et al. 2005) at http://gabi.rzpd.de/projects/MapMan/. For the LIMMA analysis, a linear model was fitted to the whole data set treating all samples harvested at 20 °C as the control (n = 4). Individual contrasts between given time points (n = 2) and the control were than extracted and P-values adjusted separately for each contrast using the Benjamini & Hochberg (1995) correction.

The overview figures in the present paper were prepared using MapMan version 1.6.3. Wilcoxon's P-values and average responses of genes in individual BINS were calculated and visualized as in Usadel et al. (2006).

Quantitative real-time RT-PCR

Total RNA was extracted from shoot using the Qiagen RNeasy plant RNA kit (Qiagen, Hilden, Germany). Genomic DNA was removed using DNAse (Ambion Europe Ltd, Huntingdon, UK), then RNA quality was checked by agarose gel electrophoresis. DNA-free RNA (2.5 µg) was used as template for cDNA synthesis (Promega, Mannheim, Germany) with oligo dT primers. The quality of cDNA synthesis was assessed by RT-PCR using the following primers: AT1G27450-F: 5′-TCCCAGA ATCGCTAAGATTGCC-3′ and AT1G27450-R: 5′-CCTT TCCCTTAAGCTCTG-3′. Primers for real RT-PCR were designed using Primer Express software (version 2.0; Applied Biosystems, Foster City, CA, USA) for amplicon lengths of between 60 and 150 bp: CBF1-F: 5′-CCGCC GTCTGTTCAATGGAATCAT-3′, CBF1-R: 5′-TCCAA AGCGACACGTCACCATCTC-3′, CBF2-F: 5′-CGGAAT CAACCTGTGCCAAGGAAA-3′, CBF2-R:, 5′-AGACC ATGAGCATCCGTCGTCATA-3′, CBF3-F: 5′-TTATAT TCCGACGCTTGCGAGC-3′, CBF3-R: 5′-CGAAACTT CTTACGACCCGCC-3′, CBF4-F: 5′-TGGTCGCTCTGC TTGTCTCAATTT-3′, CBF4-R: 5′-GTCTCAGGAATAC GAAGCCGCCAA-3′, At3g18780-F: 5′-TCCCTCAGCAC ATTCCAGCAGAT-3′ and At3g18780-R: 5′-AACGATT CCTGGACCTGCCTCATC-3′. Real-time PCR reactions were performed in a 384-well plate with an Applied Biosystems ABI Prism 7900 Sequence Detection System (Applied Biosystems), using POWER-SYBR Green (Eurogentec, Cologne, Germany) to monitor cDNA amplification. PCR reactions and data analysis were performed as described previously by Czechowski et al. (2004)

Enzyme activities

Leaf tissue was stored in liquid nitrogen, and extracts for enzyme activities were prepared and enzyme activities were measured exactly as in Gibon et al. (2004).

Metabolite measurements

Sucrose, glucose, fructose and starch were measured in the soluble and residual fractions of an ethanol–water extract (Scheible et al. 1997a,b), as described in Stitt et al. (1989). Amino acids were determined in the ethanol/water extracts by high-performance liquid chromatography (HPLC) as described in Geigenberger et al. (1996). Frozen material was used for extraction of phosphorylated metabolites with perchloric acid and assayed as in Stitt et al. (1989). GC-MS and LC-MS metabolite profiling was performed exactly as in Gibon et al. (2006) and Osuna et al. (2006). Starch, sugars and amino acids were analysed as in Gibon et al. (2002).

RESULTS AND DISCUSSION

Experimental design

The plants were grown in soil for 4 weeks at 20 °C, and then were left at 20 °C or were shifted to 17, 14, 12, 10 or 8 °C and harvested 6 and 78 h later. The plants were shifted 4 h after the beginning of the light period and were harvested 10 h into the light period. Source leaves contributed most of the tissue in the rosettes. Two separate experiments were performed at a 6 month interval.

Metabolite profiling

Over 80 metabolites were measured, using a combination of GC-TOF, LC-MS, dedicated HPLC and enzymatic tests. The results are summarized using a false colour scale in Fig. 1a, and selected examples are shown in Fig. 1b–o. Most metabolites responded to a decrease of the temperature, with many showing significant changes at 17 or 14 °C.

Figure 1.

Metabolites. Metabolite levels were measured 6 and 78 h after transfer from 20 to 17, 14, 12, 10 and 8 °C, and at 20 °C in plants harvested at the same time as the low-temperature treatments. (a) Heat map. The figure uses a false colour scale to depict the changes of metabolites. The metabolite levels were normalized on the average level at 20 °C at that time, the ratios converted to a log2 scale and expressed on a false colour scale. Metabolites that increase at low temperature are coloured blue, and metabolites that decrease are coloured red. The results are the average of two biological replicated experiments, each with five independent replicate samples. Significant changes (P < 0.05) are marked with an asterisk. (b–o) Changes of selected metabolites: (b) glucose-6-phosphate, (c) sucrose, (d), glucose, (e) fructose, (f) starch, (g) pyruvate, (h) acetyl CoA, (i) 2-oxoglutarate, (j) glutamate, (k) glutamine, (l) glycine, (m) serine, (n) total amino acids and (o) protein. The results are the mean and SE of five replicates from one experiment. They are given as absolute levels, or as the ratio compared to control samples at 20 °C. A similar result was obtained in a biological replicate (data not shown). The data are provided in Supplementary Table S1. CHO, carbohydrate; CoA, coenzyme A.

Partitioning shifts towards sucrose after transferring Arabidopsis to 4 °C (Strand et al. 1999; Hurry et al. 2000; Stitt & Hurry 2002). Our results show that this shift occurs within 6 h, and at temperatures as high as 12 °C. It involves a strong increase of sucrose (Fig. 1c), a small increase of glucose and fructose (Fig. 1d–e), and a slight decrease of starch (Fig. 1f). While the accumulation of sugars might be partly due to inhibition of phloem export, the trend to lower starch shows that partitioning has shifted to favor sucrose synthesis. This shift was accompanied by a rapid increase of UDP-Glc (Fig. 1a), Glc6P (Fig. 1b) and other glycolytic intermediates including 3-phosphoglycerate (Fig. 1a) and pyruvate (Fig. 1g), and a decrease of Pi (Fig. 1a). This resembles other species, where low temperatures inhibit end-product synthesis more strongly than carbon fixation, leading to an accumulation of phosphorylated intermediates and a transient Pi limitation of photosynthesis (see Introduction).

Accumulation of sugars should stimulate N assimilation and amino acid biosynthesis (see Morcuende et al. 1998; Fritz et al. 2006). Any restriction of phloem export by low temperature (see Introduction) should also lead to accumulation of amino acids. Unexpectedly, many organic acids and amino acids decreased after 6 h. Acetyl coenzyme A (AcCoA) decreased slightly (Fig. 1h); malate increased (Fig. 1a), 2-oxoglutarate (2-OG) (Fig. 1i) and fumarate (Fig. 1a) decreased slightly; succinate decreased markedly (Fig. 1a); Glu decreased (Fig. 1j); Gln rose twofold (Fig. 1k); Gly increased (Fig. 1l), and Ser decreased (Fig. 1m). The resulting increase of the Gly/Ser ratio was perceptible at 17 °C and was significant at 14 °C. Some minor amino acids decreased (His, Lys, Thr); some were not significantly or consistently altered (Arg, Met); some showed a small increase at intermediate temperatures that became less marked or disappeared at low temperatures (Phe, Tryp, Val, Thr), and some showed a consistent increase (Tyr, Leu) (Fig. 1a). There was a marked decrease of shikimate and a non-significant decrease of citrulline, which are intermediates in the aromatic amino acid and arginine biosynthesis pathways, respectively (Fig. 1a). These results indicate that low temperatures perturb organic acid and nitrogen metabolism at several sites, including the entry of carbon into the trichloroacetic acid (TCA) cycle, glutamate synthase, glycine decarboxylation and some amino acid biosynthesis pathways.

By 78 h, there was a general trend to accumulation of metabolites. There was a general accumulation of carbohydrates (Fig. 1a,c–f). There was an increase of AcCoA (Fig. 1h), and a general accumulation of organic acids including citrate, malate, fumarate, succinate (Fig. 1a) and 2-OG (Fig. 1i). There was a general increase of amino acids (Fig. 1n), including all the major amino acids (Gln, Glu, Asp, Ala, Asn, Gly, Ser) and most of the minor amino acids (Phe, Tryp, Tyr, Val, Arg, Leu) (Fig. 1a,j–l). This was accompanied by an increase of the leaf protein content (Fig. 1o; see also later). The increase of protein was already significant at 17 °C (P > 0.05). There were marked changes of the lipid profile (Fig. 1a) including an increase of unsaturated fatty acids like C18:2, C18:3 and C16:3, and a decrease of C16:2 and 16:1. Most of these changes were detectable at 17 or 14 °C. Unexpectedly, some highly saturated fatty acids also increased, including 16:0 and the long-chain unsaturated fatty acids C24:0 and C26:0. Raffinose and proline increased markedly (Fig. 1a), as did other metabolites implicated in resistance to abiotic stress like myoinositol, putrescine, tocopherol, GABA, DOPA, quinol and coenzyme Q (Fig. 1a). Ferulic and sinapic acids increased, indicating that phenylpropanoid metabolism has been stimulated (Fig. 1a). Isopentyl pyrophosphate, which is a precursor for terpenoid metabolism, decreased. Many metabolites increased by 14–17 °C, including raffinose, proline, putrescine and tocopherol.

Figure 3.

Transcript levels analysed by ATH1 arrays. The data are calculated from two biologically replicated experiments, in which plants were transferred from 20 to 17, 14, 12, 10 or 8 °C for 6 or 78 h. All changes are compared to 20 °C. The original data are in Supplementary Table S3. (a) Number of genes showing a significant response (PFDR < 0.01). The P-values for the individual genes are provided in Supplementary Table S3. (b) Number of genes increasing by greater than twofold or decreasing by >50% at 17, 14, 12, 10 and 8 °C after 6 h (solid line) or 78 h (dotted line) at a given temperature.

Enzyme activities

Enzyme activities were measured in optimized conditions at 25 °C using the robotized platform presented in Gibon et al. (2004). Figure 2 summarizes the results on an FW basis. After 6 h, there was a slight increase of sucrose phosphate synthase (SPS) activity, a strong decrease of ADP-glucose pyrophosphorylase (AGPase) activity, and smaller decreases of invertase and glucokinase activity. There was a slight increase of nitrate reductase (NR) and a significant decrease of glutamine synthetase (GS) and aspartate aminotransferase activity.

Figure 2.

Enzyme activities. Changes in enzyme activity were calculated as log2 (value/average of controls) and were displayed using the Image Annotator module from the MapMan software, and shown using a false colour scale, with blue indicating an increase of activity and red a decrease relative to that in full nutrition (see figure for scale). The results are the mean of two separate biological experiments, each with five replicate samples. The original data are available as Supplementary Table S2 in the Supplementary Material, in which activities are given as nanomole per gram fresh weight (FW) per hour. As leaf soluble protein also increased on an FW basis (see Fig. 1o; the change of protein is shown on the same scaling as the enzyme activities as the last row of the figure), many of the changes in activity reflect the change in leaf soluble protein. Significant changes (P < 0.05) are marked with an asterisk. The data are provide in Supplementary Table S2. NADP-GAPDH, NADP-glyceraldehyde-3-phosphate dehydrogenase; AGPase, ADP-glucose pyrophosphorylase; G6PDH, glucose-6P dehydrogenase; cFBPase, cytosolic fructose-1,6-bisphosphatase; PFP, PPi-phosphofructokinase; SPS, sucrose phosphate synthase; NAD-GAPDH, NAD-glyceraldehyde-3-phosphate dehydrogenase; PK, pyruvate kinase; PEPCase, phosphoenolpyruvate carboxylase; NADP-ICDH, NADP-isocitrate dehydrogenase; NR, nitrate reductase; GS, glutamine synthetase; Fd-GOGAT, ferredoxin-glutamate synthase; GDH, glutamate dehydrogenase; ferredoxin-glutamate synthase; Asp AT, aspartate: 2-oxoglutarate transaminase; Ala AT, alanine: 2-oxoglutarate aminotransferase; shikimate DH, shikimate dehydrogenase.

By 78 h, many more enzyme activities had increased. Examples include NADP-glyceraldehyde-3-phosphate dehydrogenase (NADP-GAPDH), glucose-6-phosphate dehydrogenase (G6PDH), SPS, fructokinase and glucokinase, cytosolic fructosebiphosphatase (cFBP), PPi-phosphofructokinase (PFP), phosphenolpyruvate carboxylase (PEPCase), pyruvate kinase (PK), NR and aspartate and alanine aminotransferase. In some cases, the increase resembles the increase of leaf protein (see Fig. 2). SPS, cFBP and NR activity increased more than the leaf protein content. SPS, cFBP and NR activities also increase strongly at 4 °C (Guy, Huber & Huber 1992; Strand et al. 1999; Stitt & Hurry 2002; Tucker & Ort 2002). An opposite response was found for AGPase, acid invertase, fumarase, GS, ferredoxin-glutamate synthase (Fd-GOGAT) and glutamate dehydrogenase (GDH), whose activity remained unaltered or decreased on an FW basis and decreased on a protein basis (Fig. 2).

Expression arrays

Transcript profiling was carried out using ATH1 arrays. The original data are available in Supplementary Table S3. The software package RMAExpress (Bolstad et al. 2003) was used to normalize and estimate signal intensities from the raw data files (.cel). Of the 22 746 probe sets on the ATH1 arrays, 14 608 were called present by the MAS5 software in one or more samples at 6 or 78 h in both experimental replicates. Two approaches were taken to test reproducibility between the two biological replicates.

In the first approach, LIMMA (Smyth 2005) was used to estimate P-values for each gene. A control set of four 20 °C samples was formed by grouping the 6 and 78 h samples from the two replicate experiments. Pairs of samples were available for the 6 and 78 h time points for the 17, 14, 12, 10 and 8 °C treatments. Decreasing the temperature from 20 to 17, 14, 12, 10 and 8 °C led to significant changes (PFDR < 0.01) for 2, 96, 602, 1263 and 1075 genes after 6 h, and 0, 4, 134, 698 and 1321 genes after 78 h (see Fig. 3a). The PFDR values for each gene are provided in Supplementary Table S3. Table 1 shows the genes that show significant changes after 6 h at 14 °C (PFDR < 0.01). The list includes genes that are involved in the regulation of transcription, translation and protein degradation (e.g. several transcription factors, several ubiquitin-targeting proteins, eukaryotic elongation factor 6, APG8e), several protein kinases and phosphatases, several genes encoding enzymes in primary carbon metabolism, lipid metabolism and secondary metabolism, two known low-temperature-responsive genes (COR15a, LT150) and many genes that encode for proteins with a poorly understood or unknown function.

Table 1.  Genes that show significant changes after 6 h at 14 °C when compared to the 20 °C controls (PFDR ≤ 0.01 in the LIMMA test)
Affy ID (ATH1)AGI locus (TIGR 5)Log2 ratio to 20 °CFunctional category (Bin name)Annotation (TIGR5)
  1. For each gene, the MapMan functional category is indicated.

  2. SPS, sucrose phosphate synthase; PS, photosynthesis; OPP, oxidative pentose phosphate; PP, pentose phosphate; FA, fatty acid; Misc., miscellaneous; SAT, serine acyl transferase; PRPP, phosphoryl ribose pyrophosphate.

Chilling repressed
258055_atAt3G16250−0.71PS.light reaction.other electron carrier (ox/red.ferredoxinFerredoxin-related
260014_atAt1g68010−0.58PS.photorespiration.hydroxypyruvate reductaseGlycerate dehydrogenase/NADH-dependent hydroxypyruvate reductase
253966_atAt4g26520
At4g26530
−0.75PS.Calvin cyle.aldolaseFructose-bisphosphate aldolase, cytoplasmic
255016_atAt4g10120−1.07Major CHO metabolism.synthesis.sucrose.SPSSPS4, sucrose-phosphate synthase
246114_atAt5g20250−0.91Minor CHO metabolism.raffinose synthasesRaffinose synthase family protein/seed imbibition protein, putative (din10)
253966_atAt4g26520
At4g26530
−0.75Glycolysis.aldolaseFructose-bisphosphate aldolase
249694_atAt5g35790−0.68OPP.oxidative PP.G6PDGlucose-6-phosphate 1-dehydrogenase/G6PD (APG1)
252011_atAt3g52720−1.06TCA/org.transformation.carbonic anhydrasesCarbonic anhydrase family protein
264931_atAt1g60590−0.97Cell wall.degradation.pectate lyases and polygalacturonasesPolygalacturonase, putative
257880_atAt3g16910−1.00Lipid metabolism.FA synthesis and FA elongation.acyl coa ligaseAMP-dependent synthetase and ligase family protein
249774_atAt5g24150−1.13Lipid metabolism.‘exotics’ (steroids, squalene, et.)Squalene monooxygenase 1,1/squalene epoxidase 1,1 (SQP1,1)
263987_atAt2g42690−0.84Lipid metabolism.lipid degradation.lipasesLipase, putative
260014_atAt1g68010−0.58Amino acid metabolism.degradation.serine-glycine-cysteine group.serineGlycerate dehydrogenase/NADH-dependent hydroxypyruvate reductase
253874_atAt4g27450−0.77Hormone metabolism.auxin.induced-regulated-responsive-activatedExpressed protein
263668_atAt1g04350−0.58Hormone metabolism.ethylene.synthesis-degradation1-Aminocyclopropane-1-carboxylate oxidase – putative
245242_atAt1g44446−0.90Tetrapyrrole synthesisChlorophyll a oxygenase (CAO)/chlorophyll b synthase
254553_atAt4g19530−0.84Stress.bioticDisease resistance protein (TIR-NBS-LRR class), putative
252485_atAt3g46530−0.53Stress.bioticDisease resistance protein, RPP13-like (CC-NBS class), putative
255088_atAt4g09350−0.72Stress.abiotic.heatPutative protein heat shock protein dnaJ
250017_atAt5g18140−0.67Stress.abiotic.heatDNAJ heat shock N-terminal domain-containing protein
261144_s_atAt1g19660−1.11Stress.abiotic.touch/woundingFlag_XH2 – >wound-responsive family protein
262780_atAt1g13090−0.56Misc.cytochrome P450Cytochrome P450 71B28, putative (CYP71B28)
262830_atAt1g14700−0.94Misc.acid and other phosphatasesPurple acid phosphatase, putative
250255_atAt5g13730−0.58RNA.transcriptionRNA polymerase sigma subunit SigD (sigD) / sigma-like factor (SIG4)
264781_atAt1g08540−0.47RNA.transcriptionRNA polymerase sigma subunit SigB (sigB) / sigma factor 2 (SIG2)
260769_atAt1g49010−0.59RNA.regulation of transcription.MYB-related transcription factor familyMYB family transcription factor
249862_atAt5g22920−1.56RNA.regulation of transcription.unclassifiedZinc finger (C3HC4-type RING finger) family protein
266899_atAt2G34620−1.42RNA.regulation of transcription.unclassifiedMitochondrial transcription termination factor-related/mTERF-related
264507_atAt1g09415−0.94RNA.regulation of transcription.unclassifiedNPR1/NIM1-interacting protein 3 (NIMIN-3)
259038_atAt3g09210−0.52RNA.regulation of transcription.unclassifiedKOW domain-containing transcription factor family protein
252548_atAt3g45850−0.79DNA.synthesis/chromatin structureKinesin-related protein
249008_atAt5g44680−0.75DNA.repairMethyladenine glycosylase family protein
257801_atAt3g18750−0.60Protein.post-translational modificationProtein kinase family protein
260317_atAt1g63800−1.06Protein.degradation.ubiquitin.E2E2, ubiquitin-conjugating enzyme 5 (UBC5)
266604_atAt2G46030−0.79Protein.degradation.ubiquitin.E2E2, ubiquitin-conjugating enzyme 6 (UBC6)
262986_atAt1g23390−1.31Protein.degradation.ubiquitin.E3.SCF.FBOXKelch repeat-containing F-box family protein|
266106_atAt2G45170−1.10Protein.degradation.autophagyAutophagy 8e (APG8e)
248303_atAt5g53170−0.65Protein.degradation.metalloproteaseFtsH protease, putative similar to ATP-dependent metalloprotease FtsH1
257615_atAt3g26510−1.19Not assigned.no ontologyOcticosapeptide/Phox/Bem1p (PB1) domain-containing protein
262884_atAt1g64720−1.15Not assigned.no ontologyExpressed protein
252433_atAt3g47560−0.76Not assigned.no ontologyEsterase/lipase/thioesterase family protein
259500_atAt1g15740−0.74Not assigned.no ontologyLeucine-rich repeat family protein
251218_atAt3g62410−0.54Not assigned.no ontologyCP12 domain-containing protein
250394_atAt5g10910−0.52Not assigned.no ontologymraW methylase family protein
256754_atAt3g25690−0.51Not assigned.no ontology.hydroxyproline-rich proteinsHydroxyproline-rich glycoprotein family protein
259104_atAt3g02170−1.41Not assigned.unknownExpressed protein
253305_atAt4g33666
At4g33670
−1.14Not assigned.unknownExpressed protein
258468_atAt3g06070−1.08Not assigned.unknownExpressed protein
245602_atAt4g14270−0.88Not assigned.unknownExpressed protein
259017_atAt3g07310−0.84Not assigned.unknownExpressed protein
266707_atAt2G03310−0.82Not assigned.unknownExpressed protein
253849_atAt4g28080−0.80Not assigned.unknownExpressed protein
252036_atAt3g52070−0.79Not assigned.unknownExpressed protein
259207_atAt3g09050−0.72Not assigned.unknownExpressed protein
254187_atAt4g23890−0.70Not assigned.unknownExpressed protein
245321_atAt4g15545−0.65Not assigned.unknownExpressed protein
249231_atAt5g42030−0.64Not assigned.unknownExpressed protein
265457_atAt2g46550−0.62Not assigned.unknownExpressed protein
254848_atAt4g11960−0.62Not assigned.unknownExpressed protein
248329_atAt5g52780−0.61Not assigned.unknownExpressed protein
258738_atAt3g05750−0.49Not assigned.unknownExpressed protein
259240_atAt3g11590−0.42Not assigned.unknownExpressed protein
Chilling induced
264668_atAt1g097800.67Glycolysis.phosphoglycerate mutase2,3-Biphosphoglycerate-independent phosphoglycerate mutase, putative
267280_atAt2G194500.50Lipid metabolism.TAG synthesisDiacylglycerol O-acyltransferase/acyl CoA:diacylglycerol acyltransferase (DGAT)
247776_atAt5g587000.74Lipid metabolism.lipid degradation.lysophospholipasesPhosphoinositide-specific phospholipase C family protein
257194_atAt3g131100.65Amino acid metabolism.synthesis.serine-glycine-cysteine group.cysteine.SATSerine acetyl transferase 1, SAT1
264261_atAt1g092401.01Metal handling, binding, chelation and storageNicotianamine synthase, AtNAS3
250207_atAt5g139301.34Secondary metabolism, flavonoids, chalconesChalcone synthase (naringenin-chalcone synthase)
246468_atAt5g170501.42Secondary metabolism.flavonoids.dihydroflavonolsUDP glucose:flavonoid 3-o-glucosyltransferase-like protein
262440_atAt1g477100.60Stress.bioticSerpin, putative/serine protease inhibitor, putative
263497_atAt2G425400.66Stress.abiotic.coldCold-responsive protein/cold-regulated protein (cor15a)
258321_atAt3g228401.37Stress.abiotic.lightChlorophyll ab binding family protein / early light-induced protein (ELIP)
252102_atAt3g509702.39Stress.abiotic.unspecifiedDehydrin xero2 (XERO2)/low-temperature-induced protein LTI30 (LTI30)
257252_atAt3g241700.78Redox.ascorbate and glutathioneGlutathione reductase, putative
253252_atAt4g347401.00Nucleotide metabolism.synthesis.purine2-purF-PRPP amidotransferase
246468_atAt5g170501.42Misc.UDP glucosyl and glucoronyl transferasesUDP glucose:flavonoid 3-o-glucosyltransferase-like protein
247378_atAt5g631200.83RNA.processingEthylene-responsive DEAD box RNA helicase, putative (RH30)
260776_atAt1g145801.01RNA.regulation of transcription.C2H2 zinc finger familyZinc finger (C2H2 type) family protein
260209_atAt1g685500.74RNA.regulation of transcription.AP2/EREBPPutative AP2 domain transcription factor
251899_atAt3g544000.79RNA.regulation of transcription.unclassifiedNucleoid DNA-binding – like protein
247277_atAt5g644200.58DNA.repairDNA polymerase V family
257237_atAt3g148900.75DNA.repairDNA nick sensor, putative
251776_atAt3g556200.50Protein.synthesis.initiationEukaryotic translation initiation factor 6 (EIF-6)
266908_atAt2G346500.56Protein.post-translational modificationPutative protein kinase
265886_atAt2G256200.65Protein.post-translational modificationPutative protein phosphatase 2C
262440_atAt1g477100.60Protein.degradation.serine proteaseSerpin, putative/serine protease inhibitor, putative
247776_atAt5g587000.74Signalling.phosphinositidesPhosphoinositide-specific phospholipase C family protein
259426_atAt1g014700.80Development.unspecifiedLate embryogenesis abundant protein, putative/LEA protein
248326_atAt5g528200.80Development.unspecifiedWD-40 repeat family protein
249063_atAt5g441100.94Transport.ABC transporters and multidrug resistance systemsABC transporter family protein
259450_atAt1g138700.47Not assigned.no ontologyExpressed protein
254227_atAt4g236300.57Not assigned.no ontologyReticulon family protein (RTNLB1)
263334_atAt2g038200.66Not assigned.no ontologyNonsense-mediated mRNA decay NMD3 family protein
261651_atAt1g277600.98Not assigned.no ontologyInterferon-related developmental regulator family protein
257253_atAt3g241900.84Not assigned.no ontology.ABC1 family proteinABC1 family protein
257188_atAt3g131500.83Not assigned.no ontology.pentatricopeptide (PPR) repeat-containing proteinPPR repeat-containing protein
254656_atAt4g180700.42Not assigned.unknownExpressed protein
264366_atAt1g032500.51Not assigned.unknownExpressed protein
248762_atAt5g474550.68Not assigned.unknownExpressed protein
250151_atAt5g145700.72  
266532_atAt2G168900.94  

In the second approach, the signals for a given probe set from the two replicate experiments were plotted against each other, and the Pearson correlation coefficient (CC) was calculated (Supplementary Fig. S1a & Supplementary Table S3). Transcripts that do not show marked changes of their transcript levels are subject to random noise, which generates symmetrical population of weak positive and weak negative CCs. Very few genes showed a CC below −0.5, indicating that the vast majority of genes with CCs above +0.5 show qualitatively replicated responses (see also Bläsing et al. 2005). This interpretation is supported by comparing the CC with the magnitude of the temperature-dependent change of the transcript level (Supplementary Fig. S1b). In total, 52 and 17% of the genes called present by the MAS5 software had CCs >0.5 and >0.8, respectively. Of 627 genes that showed a greater than twofold change in one of the treatments, 94 and 338 had CCs >0.8 and >0.5, respectively.

Global changes of transcripts are larger after 6 h than 78 h

Figure 3 investigates the temperature and time dependence of the response. As the temperature decreases, increasing numbers of genes are affected. At a given temperature, more genes show a significant change (Fig. 3a) or a twofold increase or 50% decrease of their transcript (Fig. 3b) after 6 h than after 78 h. The only exception is that the 6 h response saturates at 10 °C, whereas the 78 h response increases down to 8 °C. This is probably because the initial response is slowed down so much that it is still underestimated after 6 h at 8 °C.

K-clustering identifies subsets of transcripts with different temperature and temporal responses

The temperature and time dependence of the response of individual genes was visualized by performing K-clustering on 2173 genes that showed significant changes (PFDR < 0.01 in the LIMMA analysis) in their transcript levels in at least one treatment. A large number of clusters (20) were generated to display different responses (Fig. 4). The genes in each cluster are listed in Supplementary Table S3. Many clusters show a monotonic change across the entire temperature range. This shows that the difference between the global response to a small and large decrease of the temperature is mainly due to a change in the amplitude of the response, rather than a change in which genes are affected. There was often good qualitative agreement between the response of individual genes at 6 and 78 h. Thus, 71 genes were induced at 6 h and equally or more strongly at 78 h (clusters 1 and 5); 525 genes were induced at 6 h and strongly damped at 78 h (clusters 2, 4, 11 and 16); 287 genes were unaffected or weakly induced at 6 h and more strongly induced at 78 h (clusters 17 and 19); 355 genes were repressed at 6 h and equally or more strongly at 78 h (clusters 3, 5, 10 and 14); 617 genes were repressed at 6 h, and the response was damped at 78 h (clusters 12, 15, 18 and 20); and 243 genes were only marginally affected at 6 h and repressed at 78 h (clusters 7, 8 and 13). A few clusters showed different responses at 6 and 78 h. Clusters 8, 13, 17 and 20 responded across the entire temperature range after 6 h, but the changes at higher temperatures were reversed after 78 h. In clusters 1, 4, 14 and 16, the response saturated at 10–12 °C after 6 h but increased down to 8 °C after 78 h. This may be because the response is not completed after 6 h (see previous discussion). A small set of 30 transcripts (cluster 9) showed opposing responses, with a small decrease at 6 h and a small increase at 78 h.

Figure 4.

K-means clustering of transcripts. For the 2173 genes that showed significant changes in their transcript levels (P < 0.01 in the LIMMA analysis, Table 1), the responses in the four 20 °C treatments were averaged (control), and the responses of the two biological replicates for each temperature and time treatment were averaged, normalized on the 20 °C control and expressed as a log2 ratio. This yielded 10 experimental values for each gene, corresponding to the ratios at 17, 14, 12, 10 and 8 °C after 6 h, and the ratios at 17, 14, 12, 10 and 8 °C after 78 h. These values were subjected to K-means clustering, using a total of 20 clusters to allow the responses of individual genes to be well displayed. The response at 6 h is shown on the left-hand side, and the response at 78 h is on the right-hand side of each panel, with the temperature decreasing in each case from the left to the right. The number of genes in each cluster is given in the figure, and the red line shows the average response of the genes in the cluster. Members of the core set of the CBF regulon (Fowler & Thomashow 2002; Vogel et al. 2005) are shaded blue. A full list of the genes, their annotations, MapMan ontologies and the clusters they were assigned to is provided as Supplementary Table S3.

Responses of genes in different functional categories and responses of individual genes

The ∼22 000 genes on the ATH1 array were organized into >700 functional categories, using the hierarchical MapMan ontology (Thimm et al. 2004; Usadel et al. 2005; see also http://gabirzpd.de/projects/MapMan/). The assignments are included in Supplementary Table S3. For each gene, we calculated the ratio between the level of the transcript at a given temperature and the level at 20 °C, converted it to a log2scale and averaged the two independent biological replicates. The data were then analysed in two ways. Firstly, we searched for functional categories that showed a coordinated response to low temperature. To do this, we calculated the average change of all of the genes in each category and Wilcoxon's P-value, using the PageMan application (Usadel et al. 2006). Wilcoxon's P-value reveals whether the response of the genes in a given category is statistically different to the response of all the other genes on the array. The averages and P-values for selected functional categories are visualized on a false colour scale in Fig. 5. The colour intensity indicates the magnitude of the change of the average value or significance of the P-value, and blue and red signify categories whose transcripts levels increase or decrease (a positive or negative value was manually assigned to the P-value, depending on whether the average signal increased or decreased). The response after transfer to 17, 14, 2, 10 and 8 °C is shown from left to right in each block. The complete analysis is provided in Supplementary Table S4. Secondly, we used the MapMan application (Thimm et al. 2004; Usadel et al. 2005) to visualize the responses of individual genes. Formatted files are provided in Supplementary Table S5, and selected screen shots in Fig. 6. Genes called ‘absent’ by the Affymetrix software are shown as grey, and genes that increase and decrease by an increasingly intense blue and red coloration, respectively. A scale was selected in which a change of <log20.5 was shaded white and the response saturated at a twofold change. In addition, a threshold significance filter was applied. All genes whose response was not significant in the LIMMA analysis (P > 0.05) were shaded white. A similar set of files were generated for the 1187 genes whose transcript levels changed after transfer from 24 to 13 °C (Provart et al. 2003; Supplementary Table S6), and in CBF2 over-expressors, after transfer from 20 to 4 °C (Vogel et al. 2005; Supplementary Table S7).

Figure 5.

Chilling-induced changes in transcripts, organized at the level of functional categories. The original analysis used >700 functional categories defined in the MapMan hierarchical ontology (Thimm et al. 2004; Usadel et al. 2005; see also http://gabirzpd.de/projects/MapMan/). For each gene called present, the ratio between the transcript level at a given temperature and at 20 °C was calculated and converted to a log2 scale. The right-hand pair of columns display the average change for all genes in a functional category on a false colour scale (increasing blue and red indicate an increasingly large increase and decrease, respectively), using a linear scale between +4 and −4. The left-hand pair of columns show the P-values obtained when Wilcoxon's test was applied to identify functional categories in which the genes in the category showed a statistically different response to the remaining genes on the array. The P-values are assigned a positive or negative value, depending on whether the average signal in the category increased or decreased, and are visualized on a false colour scale (increasing blue and red indicate an increasingly significant increase and decrease, respectively). To set the scale, the P-values were converted to their respective z-scores using the inverse normal cumulative distribution function, treating the P-value as two-tailed. These values can be obtained in Excel by using the NORMSINV (2 × P-value) formula. At the settings used for this display, the pale, mid and deep red shadings represent P-values <1.2 × 10−2, 6 × 10−7 and 6 × 10−14, respectively. In each column, the response after transfer to 17, 14, 12, 10 and 8 °C is shown from left to right. This figure displays examples that were selected manually, because they showed a strong response to temperature. A full analysis with average changes and P-values for each BIN and sub-BIN is provided in Supplementary Table S4. AGPase, ADP-glucose pyrophosphorylase; PK, pyruvate kinase; PEPCase, phosphoenolpyruvate carboxylase; LHC, light harvesting complex; TPS, trehalose phosphate synthase; TPP, trehalose phosphate phosphatase; LDH, lactate dehydrogenase; OASTL, O-acetylserine thiolyase; PAL, phenylalanine lyase.

Figure 6.

Responses of transcripts for individual genes. The results are displayed using the MapMan software (Thimm et al. 2004; Usadel et al. 2005). A full description of the bins and layout can be obtained from Thimm et al. (2004) and at http://gabi.rzpd.de/projects/MapMan/. The results are the mean of two biological replicates. Genes that are called ‘absent’ by Affymetrix software are shown as grey; genes that do not change by more than a threshold significance of P < 0.05 in the LIMMA analysis are shown as white, and genes that increase and decrease are shown by an increasingly intense blue and red coloration, respectively. A scale was selected in which a value of 0.5 and 2.0 on a log2 scale gave a faint and full saturation, respectively. (a) Selected features from the central metabolism. (b) mRNA and protein synthesis. Files to allow these and other responses to be visualized in MapMan are provided in the supplementary material (Supplementary Table S5).

Changes of transcripts of genes involved in central metabolism

After 6 h, there was a general trend to repression of genes that are involved in photosynthesis (Figs 5 & 6a), including the light reactions and, to a lesser extent, photorespiration and the Calvin cycle. This was perceptible at 17 °C (Fig. 5). Some changes were attenuated after 3 d, while others became more marked, for example, the repression of the light reactions. One exception was LHCB2.4 (At3g27690), which was induced (see also Provart et al. 2003) (Fig. 6a). Several members of the carbonic anhydrase gene family (At3g52720, At1g58180, At5g14740) were repressed. This might be related to the lower rates of carbon fixation and increased solubility of carbon dioxide at low temperatures, prompting the hypothesis that the expression of carbonic anhydrase might respond to changes in the internal carbon dioxide concentrations.

Several genes were induced that were required for sucrose synthesis (Fig. 5). This included the envelope transporter TPT and specific members of the SPS (SPS3B; see also Provart et al. 2003) and SPP (SPP4) families, but not cFBP (cytosolic FBPase) (Fig. 6a). Transfer of Arabidopsis to 4 °C for 10 d also led to increased levels of transcripts that encode enzymes in the pathway of sucrose synthesis, but not of transcripts that encode Calvin cycle enzymes (Strand et al. 1999). There was a coordinated repression of several genes involved in starch and sucrose breakdown, including several members of the invertase family (At1g12240, At1g62660, At3g06500) (see also Provart et al. 2003) (Figs 5 & 6a).

There was no general trend for genes that encode enzymes in glycolysis or the TCA cycle (Figs 5 & 6a). However, transcripts levels increased for individual genes that encode PFP (At5g56630), PEPCase (At2g42600) and PK (At5g56350, At5g08570, At2g36580), as well as selected enzymes in the TCA cycle (Figs 5 & 6a). One specific member of the sucrose synthase family was induced (SUS5.2;At5g20830) (Fig. 6a). There was a strong induction of genes involved in fermentation including pyruvate decarboxylase (At5g54960, At4g33070) and ADH1 (At1g77120) (Figs 5 & 7a; see also Provart et al. 2003; Peters & Frenkel 2004; Vogel et al. 2005). There was a weak induction of selected genes that encode components of the mitochondrial electron transport chain, in particular At1g07180 encoding NDA1 (a non-phosphorylating NADH dehydrogenase) and At3g22370, which encodes alternative oxidase 1a. These responses were detectable at 17 °C, and appeared within 6 h. After 78 h, the induction of alternative oxidase and genes involved in fermentation was weakened or reversed; the induction of NADH dehydrogenase was strengthened, and there was weak induction of cytochrome c oxidase (At4g10040) and genes for ATPase components (Figs 5 & 6a). Although small, these changes were confirmed at two or more temperatures. Provart et al. (2003) reported that transfer to 13 °C induced cytochrome oxidase and mitochondrial ATPase components. Alternative oxidase and subunits of cytochrome oxidase and ATP synthase were induced by low temperatures in tomato fruits (Holtzapffel et al. 2002, 2003).

Figure 7.

Evaluation of the relation between changes of transcripts, enzyme activities and metabolites. (a) Frequency plot of the changes of transcript levels (red), enzyme activities (blue) and metabolite levels (green) after transfer from 20 to 17, 14, 12 or 10 °C for 6 h (dotted line) or 78 h (solid line). The frequency plots were computed using a band width of 0.1. (b) Box plot of the changes of transcript levels (red), enzyme activities (blue) and metabolite levels (green) after transfer from 20 to 17, 14, 12 or 10 °C. The 6 and 78 h data sets are shown on the left and right side, respectively. The plot shows the median, quartiles (box) and the respective quartile 1.5 times the interquartile range (horizontal line). (c) Frequency plot of the agreement between changes of transcript levels and changes of enzyme activities 6 and 78 h after transfer to a lower temperature. For each enzyme, scatter plots were made of the activity against the level of each transcript from the gene family for that enzyme (with six data points corresponding to 20, 17, 14, 12, 10 and 8 °C), and the correlation coefficient (CC) (R2) was calculated. The plot shows the number of gene–enzyme pairs that have a CC in a given range 6 h (open column) and 78 h (closed column) after transfer. The full analysis is available in the supplementary material (Supplementary Fig. S5). (d) Plots showing selected scatter plots for relation between enzyme activity and the level of a major transcript 6 h (open symbols) and 78 h (closed symbols). The six data points correspond to the values at 20 °C and the values at 17, 14, 12, 10 and 8 °C. Enzyme activities were normalized on protein basis. R values are provided for the comparison after 6 h(italics) and 78 h. PK, pyruvate kinase; GDH, glutamate dehydrogenase.

In nitrogen metabolism, the most striking response was the sustained induction of genes for proline synthesis (At3g55610, At2g39800) at temperatures below 14 °C (Figs 5 & 6a). Other changes included induction of NIA1 and NII at temperatures below 12 °C (Fig. 5), a decrease of transcripts for the GOGAT pathway (e.g. At3g53180) after 78 h (Fig. 5), induction of genes for cysteine synthesis (At5g28030, At3g13110; see also Table 1) and repression of genes that encode enzymes for branched-chain amino acid degradation (e.g. At3g06850, At3g45300, At1g03090). Induction of NIA at low temperature was reported by Tucker & Ort (2002). There was a weak coordinated induction of genes for nucleotide synthesis and salvage (Fig. 5).

A small decrease of the temperature induced galactinol synthase (At1g09350, At2g47180) and raffinose synthase (At5g40390) by (Fig. 6a). There was also a small increase of the transcripts for two of three members of the myoinositol phosphate synthase family (Fig. 6a), which are required to provide myoinositol for galactinol synthesis (Fig. 5) (Loewus & Loewus 1983). Transcripts decreased strongly for an α-galactosidase family member (At5g20250), which might remove galactose during raffinose degradation (Pennycooke, Jones & Stushnoff 2003).

Several genes for polyamine synthesis were induced rapidly in response to small changes of the temperature, including ADC1 (At2g16500) and ADC2 (At4g34710) and spermine synthase (At5g19530) (Fig. 5). There was a strong and sustained induction of several genes for phenylpropanid and flavonoid synthesis at 14–17 °C (Figs 5 & 6a), confirming Provart et al. (2003). Genes for tocopherol synthesis were progressively induced as the temperature decreased (Fig. 5). Most genes for lipid biosynthesis were unaffected or repressed (Fig. 5).

Low temperatures induced many genes that contribute to cell wall synthesis (Fig. 5), including genes encoding enzymes that are required to synthesize NDP-sugar precursors, and genes encoding cell wall proteins including an extensin (At1g76930), and HRGP (At4g14900) and RGP (At5g15650, At3g02230) proteins. Low temperatures repressed several AGP proteins (At4g16980, At2g46330, At1g03870, At2g46330) (Figs 5 & 6a). Small decreases of the temperature repressed several members of the PIP (At4g23400, At1g01620, At2g45960) and TIP (At3g16240, At2g36830) gene families (Fig. 5). These showed a larger response after 78 h than after 6 h (Fig. 5). Transfer of Arabidopsis (Jang et al. 2004) or rice (Sakurai et al. 2005) to 4 °C repressed several members of the PIP family. As discussed in the Introduction, low temperatures decrease conductivity to water. These results prompt the hypothesis that small decreases of the temperature lead to a transcriptional changes that reduce conductivity to water, promote cell wall synthesis and inhibit cell expansion.

Global comparison of the temporal kinetics of transcripts, enzymes and metabolites

The next sections provide a combined evaluation of the changes of transcripts, enzyme activity and metabolites. We start by comparing the kinetics and direction of the global responses of transcript levels, enzyme activities and metabolite levels.

Figure 7a,b summarizes the responses as frequency distributions and box plots. The 8 °C treatment is omitted because the response may be attenuated at 6 h (see earlier discussion). These plots show that while the response of transcripts is damped between 6 and 78 h (see also Fig. 3), the response of enzyme activities and metabolites increases between 6 and 78 h. They also reveal a strong bias to higher enzyme activities and metabolite levels at low temperatures. This comparison cannot be made for transcripts, because their signals are normalized on the total signal on the array, which reflects the total amount of poly-A RNA. The higher enzyme activities often mirror the higher protein content (Fig. 2). The bias to higher levels of metabolites reflects the general accumulation of carbohydrates, the recovery of nitrogen metabolism and the general increase of the levels of amino acids, and the accumulation of metabolites that are involved in cryoprotection and stress responses after 78 h (see Fig. 1). Overall, Fig. 7a,b allows two general conclusions. Firstly, transcripts respond rapidly, whereas many of the changes of enzyme activities and metabolites are delayed. Secondly, there is a positive adjustment of metabolism to lower temperatures, which involves activation of many sectors of metabolism.

Comparison of the responses of transcripts and the enzymes that they encode

We next evaluated whether changes of transcript levels lead to a change in the level of the encoded protein. To do this, we focused on the genes that encode the enzymes measured in Fig. 2. We have previously argued that the enzyme activities measured by this platform are usually indicative of the level of the respective protein (Gibon et al. 2004, 2006).

Each enzyme activity was plotted against the level of the encoding transcript, resulting in a scatter plot with six data points (corresponding to the values at 20, 17, 14, 12, 10 and 8 °C). This plot was used to calculate a CC. For enzymes that are encoded by gene families, a separate plot was made for each member of the gene family. The full analysis is available in the supplementary material (Supplementary Fig. S5). The results are summarized in Fig. 7c. There is little agreement between enzyme activities and transcripts after 6 h. There is much better agreement after 78 h, when the enzyme–transcript pairs fall into two approximately equal groups, one with a negligible or negative CC, and one with a CC >0.25. It should be noted that good agreement cannot be expected for all transcript–enzyme pairs because many transcripts represent minor members of the gene families, which will not contribute significantly to total enzyme activity. About 25% of the enzyme–transcript pairs had a CC >0.5 after 78 h (Table 2). Of these, only four showed a CC >0.5 after 6 h, and in all 14 cases, the CC increased between 6 and 78 h. Four examples with little or no correlation after 6 h and a strong correlation after 78 h (fumarase versus At5g50950, PK versus At5g08750, GDH versus At5g18170, acid invertase versus At1g12240) are shown in Fig. 7d. Other examples where there was reasonable agreement after 78 h included AGPase and two major family members (At5g19220 and At5g48300), PFP and two major family members (At1g12000 and At1g20950), hexokinase and two major family members, PEPCase and all three family members, and aspartate aminotransferase (Supplementary Fig. S5). In some cases, including cytosolic FBPase, SPS, Fd-GOGAT and shikimate dehydrogenase, there is a marked discrepancy between the response of activity and the transcript for every major family member (see Supplementary Fig. S5). The level of these enzymes is presumably regulated by translation or protein degradation.

Table 2.  Relation between changes of the levels of transcripts and enzyme activities
NameNumber of genes in familyAffymetrix codeAGI
Code
Expression level (% of total family)R (6 h)R (78 h)
  1. For each enzyme, activity was calculated on a protein basis, and plotted against the level of the transcript to generate a scatter plot with six data points (corresponding to the values at 20, 17, 14, 12, 10 and 8 °C) that was used to calculate a correlation coefficient (CC) (R). This was done separately for the 6 and 78 h data sets. For enzymes that are encoded by gene families, plots were made and CCs calculated for each family member. The number of genes in the family, and relative importance of the family member is indicated in the table. The relative importance was estimated by calculating the total signal strength on the ATH1 array for all members of the family at 20 °C, and then expressing the signal of the individual family member as a percent of this total. The full analysis is available in Supplementary Fig. S5.

  2. Fd-GOGAT, ferredoxin-glutamate synthase; NAD-GAPDH, NAD-glyceraldehyde-3-phosphate dehydrogenase; NADP-ICDH, NADP-isocitrate dehydrogenase; GDH, glutamate dehydrogenase.

Acid invertase6260969_atAt1g12240430.540.96
265118_atAt1g62660180.610.73
NAD-GAPDH3262939_s_atAt1g7953020.530.80
Pyruvate kinase8250526_atAt5g08570200.290.82
260653_atAt1g3244070.280.79
252300_atAt3g4916000.630.69
NADP-ICDH2261920_atAt1g6593095−0.270.72
Fumarase1248461_s_atAt5g50950100−0.450.77
Glutamine synthetase7256524_atAt1g66200100.280.85
251973_atAt3g5318020.510.73
258160_atAt3g1782030.170.64
249710_atAt5g35630680.270.56
Fd-GOGAT2266365_atAt2g4122050.020.53
GDH3250032_atAt5g18170550.340.98

Earlier studies of diurnal cycles (Gibon et al. 2004) and carbon, nitrogen and phosphate depletion and re-addition (Osuna et al. 2006; Scheible et al. 2007) showed that there is usually a delay between responses of transcripts and enzymes in the central metabolism. Our results extend this conclusion to a further biological perturbation. The slow response of enzymes could explain why many metabolites also show a delayed response (Table 1 & Fig. 7a,b). It is tempting to speculate that the damping of transcripts after 78 h may be partly due to this gradual adjustment of metabolism to the lower temperature. There are probably also further compensatory adjustments that are not detected by our profiling platforms. For example, it would be interesting to know the rate at which COR15a and LT150 protein increase; both are known to be involved in cold acclimation, and their transcripts rise significantly and transiently at 14 °C (Table 1).

Comparison of the responses of transcripts and responses in central carbon and nitrogen metabolism

The next sections compare the responses of specific metabolites (Fig. 1), enzyme activities (Fig. 2) and transcripts (Figs. 5–6) to identify examples where changes in transcription contribute to the metabolic response to low temperature. We start by considering a series of perturbations that occur in the central metabolism after 6 h, and are partly or completely reversed by 78 h.

A small decrease of the temperature leads to a rapid switch from sucrose to starch synthesis (Fig. 1). This is associated with a rapid increase of cFBP and SPS activity (Fig. 2) and a decrease of AGPase activity, which is the key enzyme for the regulation of starch synthesis (Fig. 2). These changes are already seen by 14 °C, emphasizing that carbohydrate partitioning responds sensitively to small changes of temperature. The increase of SPS activity is associated with rapid changes in the SPS3B transcript level (Fig. 6a). Indirect evidence indicates that post-translational regulation of SPS also contributes to the stimulation of SPS activity (Strand et al. 1999; Hurry et al. 2000; Henschel & Stitt, unpublished data). The increase of cFBP activity cannot be explained on the basis of changes in cFBP transcripts, which decrease at low temperatures. AGPase activity also changes independently of AGPB and AGPS transcripts. Interestingly, low temperatures lead to a decrease of transcripts for vacuolar invertase (Fig. 5) and total invertase activity (Fig. 2). This may restrict recycling of sucrose to reducing sugars. Invertase is often induced at low temperatures, and this increase is attenuated in chilling-tolerant genotypes (Zrenner, Schüler & Sonnewald 1996; Artuso et al. 2000).

Low temperature should decrease flux through photorespiration; in addition to slowing catalysis, lower temperatures increase the affinity of ribulose 1·5-bisphosphate carboxylase/oxygenase (Rubisco) for CO2 relative to O2, and increase the solubility of CO2. One of the most marked changes after 6 h at low temperatures is an increase of Gly and a decrease of Ser (Fig. 1). This reveals that glycine decarboxylation is rapidly and preferentially inhibited by a small decrease of the temperature. Possible reasons will be discussed further. It is striking that the transcripts for SHMT show a tendency to increase (Fig. 5), whereas transcripts for most other genes for the light reactions, the Calvin cycle and photorespiration are unaffected or decrease at low temperatures. Goulas et al. (2006) reported a marked increase of SHMT protein at 4 °C. These results prompt the hypothesis that an inhibition of glycine decarboxylation leads to compensatory adjustments in the expression of genes that encode components of glycine decarboxylase.

Decreased temperatures lead to a short-term disturbance of N metabolism. This is unlikely to be caused by inhibition of nitrate assimilation. NIA transcript levels (Fig. 6a) and NR activity (Fig. 2) increase after 6 h at low temperature, and there is an increase of Gln (Fig. 1), which is the immediate product of nitrate and ammonium assimilation. The perturbations include a significant decrease of GS activity (Fig. 2), a decrease of Glu and a decrease of many minor amino acids (Fig. 1). After 3 d, most amino acids including Glu rise to higher levels than at 20 °C, revealing that the imbalance has been corrected. This readjustment occurs even though GS and Fd-GOGAT transcripts decrease (Fig. 5), GS and Fd-GOGAT activity remain unaltered or fall slightly on an FW basis and fall markedly on a protein basis (Fig. 2), and there are no large changes of the levels of most of the transcripts that encode enzymes for amino acid biosynthesis (Figs 5 & 6a). This indicates that post-transcriptional or post-translational mechanisms are primarily responsible for the recovery of N metabolism.

Low temperature leads to a short-term decrease of some organic acids, including 2-OG (Fig. 1). One possible reason for the decrease of 2-OG might be that pyruvate dehydrogenase is inhibited. The decrease of 2-OG after 6 h was accompanied by an increase of pyruvate and a small but significant decrease of AcCoA, and the recovery of 2-OG after 78 h was accompanied by a significant increase of AcCoA and other organic acids (Fig. 1). Increased activities of glycolytic enzymes like PFP, PEPCase and PK (Fig. 2) may also contribute to the gradual recovery of 2-OG. This increase is preceded by a small increase in the levels of transcripts of members of the respective gene families (see Fig. 5 & Supplementary Table S5). These changes of the 2-OG level provide an explanation for the transient inhibition of nitrogen metabolism after lowering the temperature. 2-OG is the carbon acceptor in the reaction catalysed by GOGAT. A meta-analysis has shown that the levels of 2-OG and Glu are usually rather constant and, when they do change, this often occurs in parallel (Stitt & Krapp 1999; Müller, Krapp & Stitt 2001; Matt et al. 2002; Stitt et al. 2002; Fritz et al. 2006). Thus, the initial decrease of 2-OG might be responsible for the accumulation of Gln and the decrease of Glu after 6 h, and the recovery of 2-OG after 78 h may contribute to the recovery of nitrogen metabolism.

Several aspects of our results indicate that low temperature modifies the cellular redox state. These include the transient induction of an alternative oxidase, the sustained induction of mitochondrial NAD dehydrogenases, the delayed induction of components of cytochrome oxidase and the F1-ATPase, and the strong induction of genes involved in fermentation (Figs 5 & 6a). Low temperatures may lead to excess production of NADPH by photosynthetic electron transport, and inhibit NADH oxidation in the mitochondria. Oxidation of NADH by mitochondrial electron transport is impaired by inappropriate lipid saturation (Caiveau et al. 2001). Covey-Crump et al. (2007) have shown that a small decrease of the temperature leads to a restriction of electron transport and increased reduction of the ubiquinone pool, and suggested that electron transport via alternative oxidase may help to counteract this rise. Evidence that fermentation contributes to low-temperature responses is provided by the finding that ADH double-null maize mutants are impaired in survival at 2 °C (Peters & Frenkel 2004). Low-temperature responses are also severely impaired in the Arabidopsis fro1 mutant, which is mutated in a homolog of the 18 kD Fe-S subunit of mitochondrial complex I NADH dehydrogenase (Lee et al. 2002a). The inhibition of glycine decarboxylation and 2-OG synthesis may be a consequence of oxidative stress. Taylor et al. (2002) reported that cytotoxic products of lipid peroxidation lead to loss of the lipoic acid cofactor from enzyme complexes like glycine decarboxylase and pyruvate dehydrogenase.

Comparison of transcripts and metabolites involved in cryoprotection and stress responses

Low temperatures lead to accumulation of cryoprotectants and other metabolites that are involved in stress responses. In many cases, this is preceded by a rapid increase of transcripts that encode enzymes that are required for their synthesis. There was an increase of raffinose after 3 d, which was detectable at 17 °C and became increasingly marked down to 10 °C (Fig. 1). This was preceded by an increase of transcripts for members of the small gene families for galactinol and raffinose synthases. These were among the most strongly chilling-induced genes, with changes that were significant at 17 °C at the level of the gene families (Fig. 5) and at 14 °C at the level of single genes (Fig. 6a). Previous studies in maize seedlings failed to find a relation between galactinol synthase transcript levels and raffinose accumulation (Zhao et al. 2004a,b). The accumulation of proline (Fig. 1) was preceded by a marked increase of the At3g55610/P5CSB transcript (Fig. 6a). The increase of tocopherol and polyamines (Fig. 1) was preceded by the induction of genes for their synthesis (Fig. 5). The broad and strong increase of genes for phenylpropanoid and flavonoid metabolism (Figs 5 & 6a) was accompanied by an increase of ferulate and sinapic acid (Fig. 1), which are intermediates in phenylpropanoid synthesis.

Transcriptional stimulation of RNA synthesis, protein synthesis and protein targeting

Provart et al. (2003) reported that transfer of Arabidopsis to 13 °C for 2 d leads to an increase of many transcripts that encode components of the translational machinery. Our results reveal that low temperatures lead to a coordinated induction of large numbers of genes that are involved in RNA as well as protein synthesis (Figs 5 & 6b). This affects many functional categories including RNA processing (RNA ribonucleases and RNA helicases), amino acid activation, ribosomal proteins, translation initiation and elongation, BRIX domain-containing proteins (required for ribosome assembly; Bogengruber et al. 2003), and pentatricopeptide (PPR) repeat-containing proteins (implicated in the binding of organelle RNA; Lurin et al. 2004). Genes in these categories are induced within 6 h, and in response to a small decrease of the temperature. The induction is sustained though to 78 h, by which time it becomes more marked at lower temperatures. Many genes assigned to protein synthesis are included in clusters 17 and 19 of Fig. 4, where there is little change after 6 h and a marked change after 78 h. Although the changes of the individual transcripts are not large except at 8–10 °C after 78 h, they affect a large proportion of the genes in a particular category (see Fig. 6b), resulting in highly significant Wilcoxon P-values (Fig. 5).

Genes for the cytosolic, mitochondrial and plastid translation apparatus respond differently. While there was a general induction of genes that encode components of the cytosolic and mitochondrial ribosomes, genes assigned to the plastid ribosomes were very weakly repressed after 6 h and were unaffected after 78 h (Figs 5 & 6b). There was a parallel trend to increased expression of genes involved in protein import into the mitochondria but not into the chloroplast (Fig. 5 & Supplementary Table S4; see also Provart et al. 2003).

Functional significance of changes in the leaf protein content

It is known that the leaf protein content increases when plants are grown at low temperatures (see Introduction). However, it is unclear whether this represents a direct response to temperature or an indirect effect because leaf expansion is impaired. For example, leaf protein increases two- to threefold in Arabidopsis leaves that develop at 4 °C, whereas it increases only slightly and non-significantly in existing leaves (Strand et al. 1999; Stitt & Hurry 2002). The finding that a small temperature leads within 3 d to a significant increase of the rosette protein concentration (Figs 1 & 2) shows that this is a direct response to low temperatures.

Several factors could contribute to this increase of the protein content. The coordinated induction of genes for RNA and protein synthesis (Figs 5 & 6b) could lead to higher rates of protein synthesis. Low temperatures repress genes involved in autophagy (Table 1 & Fig. 5), indicating that there may be decreased proteolysis (see also Pinedo et al. 2000). The accumulation of carbohydrates and the recovery of nitrogen metabolism after 3 d at low temperature (Fig. 1) may also contribute, because nitrogen (Scheible et al. 2004) and carbon (Osuna et al. 2007) both induce genes for protein synthesis and repress genes for autophagy.

Abiotic stress usually represses genes that encode components of the translation apparatus. This pattern is seen in carbon, nitrogen and phosphate deficiency (Scheible et al. 2004, 2007; Thimm et al. 2004; Osuna et al. 2006), and water and salt stress (AtGenExpress, http://web.uni-frankfurt.de/fb15/botanik/mcb/AFGN/). The induction of protein synthesis represents an unusual, and potentially important, component of the response to decreased low temperatures. A higher protein content will allow higher concentrations of a wide range of enzymes (Fig. 2) and, presumably, other proteins. This may partly compensate for the lower rate of catalysis at low temperature, and may contribute to the restoration of metabolic activity and growth.

The response of photosynthesis illustrates how an increased protein concentration promotes metabolism at low temperatures, and illustrates the complexity of these interactions. A decrease of the temperature does not alter light absorption, but will decrease the utilization of the absorbed energy because of the lower rates of catalysis. This is partly compensated for by the increased concentration of protein and increased activities of Calvin cycle enzymes, both on an FW and on a leaf area basis (Strand et al. 1999; Stitt & Hurry 2002; Atkin et al. 2006). The present study included one Calvin cycle enzyme (NADP-GAPDH). Its activity increased significantly within 78 h (Fig. 2) in parallel with the increase of leaf protein (Fig. 2). This increase correlated with an increase of the level of transcripts that encode NADP-GAPDH (Fig. 7d). At a given temperature, an increase of the leaf protein content will increase the rate of CO2 fixation, which in turn will require an increased rate of entry of CO2 into the leaf. The latter can be driven by higher stomatal conductance and/or a lower CO2 concentration in the leaf. It can be predicted that increasing the leaf protein content above an optimum range will lead to a lower water use efficiency and/or a limitation of photosynthesis by low internal CO2. However, the optimum leaf protein may shift upwards when the temperature decreases. Low temperatures decrease the rate of catalysis by Rubisco, increase the selectivity factor for CO2 and increase the solubility of CO2 relative to O2, but have relatively little effect on the rate constants for H2O or CO2 diffusion. The repression of transcripts for carbonic anhydrases (Figs 5 & 6a) is consistent with the hypothesis that internal CO2 concentrations increase at low temperature. It is tempting to hypothesize that the leaf protein concentration, or, more specifically, Rubisco concentration, respond to changes to the internal CO2 concentration. The increased levels of Calvin cycle enzymes in leaves that develop at 4 °C are due to a general increase of protein, except for Rubisco whose activity rises on a leaf protein basis (Strand et al. 1999). A recent proteomics study has confirmed that the level of several Calvin cycle proteins remains unaltered on a protein basis after 10 and 40 d at 4 °C, whereas Rubisco protein increases (Goulas et al. 2006). In this context, it is also noteworthy that a meta-analysis of free air carbon dioxide enhancement (FACE) responses revealed that elevated CO2 concentrations lead to a larger increase of Rubisco Vmax activity than of the capacity for ribulose-1,5-bisphosphate regeneration (Long et al. 2006).

Signalling

Lee et al. (2005a) presented a detailed analysis of the changes of transcripts of genes that are involved in signalling after transfer of Arabidopsis to 4 °C. Our study shows that most of these genes respond to small changes of the temperature within 6 h, and many of the changes are partly reverted by 78 h (Supplementary Table S7 & Supplementary Fig. S3). This includes transcripts for members of several transcription regulator families (Table 1 & Supplementary Table S3), several classes of receptor kinases, (Supplementary Table S3), components of the ubiquitin-dependent protein degradation pathway (Table 1) and genes encoding phosphoinositide-specific phospholipase C (Table 1). Small decreases of the temperature led within 6 h to increased levels of transcripts for genes involved in ABA signalling and genes that are known to be ABA responsive (Fig. 5). This response was attenuated after 78 h. Low temperature repressed genes involved in jasmonate (see also Provart et al. 2003) and ethylene metabolism and signalling (Fig. 5). This was visible by 6 h, and more marked at 78 h. They also transiently induced genes involved in salicylic acid metabolism and signalling (Fig. 5).

Comparison with array data in the public domain for responses to low temperature

Vogel et al. (2005) identified >900 genes that respond 1 and 24 h after transferring soil-grown plants to 4 °C. Figure 8a shows how these genes respond after 6 h in our study. Many show a detectable response at 17 °C, with the amplitude increasing across the entire temperature range down to 8 °C. A similar agreement was found with a smaller core set of 514 genes (Vogel et al. 2005), which respond consistently to low-temperature treatments in soil- and agar-grown plants (data not shown). This strong agreement between two data sets from different laboratories reveals that there is a high degree of continuity in the transcriptional response across a wide range of temperatures.

Figure 8.

Comparison with published ATH1 data sets for the response after transfer to 4 °C, and the response to constitutive over-expression of CBF2. The data are compared in scatter plots, with the x-axis showing the change of expression after 24 h at 4 °C (data set of Vogel et al. 2005) and the y-axis showing the response in our experiment 6 or 78 h after transfer to 17, 14, 12, 10 or 8 °C (as indicated in the figure panel). Genes that were classified by Vogel et al. (2005) as cold induced and cold repressed are coloured blue and red, respectively. (a) Temperature dependence of 938 cold-responsive genes (gene list from Vogel et al. 2005) for soil-grown plants. The plot shows the response of these genes after 6 h at 17, 14, 12, 10 and 8 °C in our experiments. (b) Temperature dependence of a core set of 93 genes whose expression was consistently changed by constitutive over-expression of CBF2 and significantly changed in response to low temperature (gene list from Vogel et al. 2005). The plot shows the response of these genes after 6 or 78 h at 17, 14, 12, 10 and 8 °C in our experiments.

This conclusion differs from that of Provart et al. (2003), who found marked differences between the response when they transferred Arabidopsis from 24 to 13 °C for 2 d and published data sets for the response after transfer to 4 °C for 1 d. When we compared our results with their more restricted 8K analysis, we found good qualitative agreement, but large differences in the amplitude of the responses (see Supplementary Fig. S2). Quantitative comparison of the two data sets is complicated by difficulties in normalizing data sets from different array technologies, and the relatively small numbers of genes that are detected by arrays (not shown). It might be noted that we used the same array technology as Vogel et al. (2005), and focused our comparative analyses on a subset of gene that respond in a robust manner in different studies to changes of the temperature, and are therefore likely to support robust conclusions from cross-experiment comparisons. Temperature-dependent changes in the temporal kinetics (see e.g. Fig. 3) also complicate any comparison of responses at single time points and different temperatures.

Comparison of the response to chilling and to CBF over-expression

Transfer of Arabidopsis to lower temperatures leads to a transient increase of CBF transcripts (see Introduction). Fowler & Thomashow (2002) and Vogel et al. (2005) applied stringent filtering to identify a core set of 93 genes whose expression is changed at 20 °C in response to constitutive over-expression of CBF2. This set, termed the CBF regulon, includes the majority of the genes that are strongly induced at 4 °C (Vogel et al. 2005). Figure 8b investigates the response of the CBF regulon to a progressive decrease of the temperature. The most responsive genes show a detectable change at 17 °C, which becomes stronger and affects a larger number of the genes at lower temperatures. The agreement is stronger after 6 h than 78 h, when many of the initial responses are damped. Members of the CBF regulon (Vogel et al. 2005) are distinguished in Fig. 4 by blue shading. They were strongly represented in clusters that show a strong response after 6 h, which is maintained or slightly damped after 78 h (e.g. clusters 1, 4, 6, 16). Few members of the CBF regulon were found in clusters that were repressed, reflecting the low proportion of repressed genes in the CBF core set (Vogel et al. 2005). No genes from the CBF regulon were present in clusters that show a slow response (little or no change after 6 h and a larger change after 78 h). These results show the CBF regulon plays a major role in the initial transcriptional response of Arabidopsis to moderate and chilling temperatures.

It is known that low temperature leads to a transient increase of CBF transcripts. This increase is detectable at 14 °C and becomes larger at lower temperatures (Zarka et al. 2003). It reversed after 4–6 h at 4 °C (Vogel et al. 2005) and even more rapidly at higher temperatures (Zarka et al. 2003). In our ATH1 data set, CBF3 transcripts were unaffected at 17–14 °C, slightly induced at 12 °C and more strongly induced at 10 and 8 °C after 6 h, while the signals for CBF1 and CBF2 were below the detection limit (see Supplementary Table S3). The 6 h time point was selected to detect a wide range of downstream changes, and may be too late to detect the transient increase of the CBF transcripts. To provide direct evidence that CBF genes are transiently induced in our conditions in response to small decreases of the temperature, we performed a further experiment in which plants were transferred from 20 °C to 17, 14, 12, 10 or 8 °C for 1 or 4 h, and analysed using quantitative real time RT-PCR (Czechowski et al. 2004). CBF1, CBF2 and CBF3 transcripts showed a small increase after transfer to 17 °C. The increase became larger as the temperature was decreased down to 12 °C (Table 3).

Table 3.  Expression of CBF gene family members after chilling treatment for 1 and 4 h
Temperature20 °C17 °C14 °C12 °C10 °C8 °C
  1. Transcript levels were determined by qRT-PCR and are expressed as ΔCT values, compared to the house-keeping control gene At3g18780. A lower ΔCT is equivalent to an increase of the corresponding transcript. The primer efficiencies and R2 values were 1.87 ± 0.15 and 0.998, 1.82 ± 0.13 and 0.996, 1.63 ± 0.19 and 0.993, 1.81 ± 0.15 and 0.998, and 1.76 ± 0.12 and 0.996 for the primer pairs for CBF1, CBF2, CBF3 and CBF4 and the control primer set At3g18780. The results are the mean and SE of three replicate samples.

 1 h
CBF16.9 ± 0.36.4 ± 0.36.4 ± 0.53.8 ± 0.84.6 ± 1.14.2 ± 0.4
CBF29.5 ± 0.78.5 ± 0.27.5 ± 0.75.9 ± 0.66.3 ± 0.65.0 ± 0.1
CBF34.4 ± 0.44.2 ± 0.84.9 ± 0.11.7 ± 0.83.3 ± 1.02.4 ± 0.2
CBF411.8 ± 0.611.3 ± 1.812.8 ± 0.911.2 ± 1.510.8 ± 2.310.9 ± 1.7
 4 h
CBF16.8 ± 0.66.0 ± 0.55.0 ± 0.14.2 ± 0.93.8 ± 0.33.4 ± 1.1
CBF211.4 ± 0.711.0 ± 0.210.8 ± 0.79.6 ± 0.49.1 ± 0.78.3 ± 1.3
CBF35.2 ± 0.64.4 ± 0.83.6 ± 0.03.8 ± 0.93.0 ± 1.62.2 ± 1.7
CBF411.9 ± 1.812.5 ± 0.612.2 ± 0.811.3 ± 0.811.6 ± 1.010.8 ± 1.9

Comparison of functional areas affected by chilling and CBF over-expression

Figure 9 compares the response of different functional categories to chilling and to CBF over-expression. The data set comparing expression profiles in CBF over-expressors and wild-type plants at 20 °C was downloaded from Vogel et al. (2005), imported into the MapMan functional categories, Wilcoxon's test performed to calculate P-values for each functional category in the MapMan ontology, and the resulting P-values converted to a z-scale. These were plotted against the corresponding values for the chilling response at 8 °C in our experiments.

Figure 9.

Comparison of the response of genes in different functional categories to low temperatures and constitutive CBF over-expression. The data set for the response to CBF2 over-expression was downloaded from Vogel et al. (2005). The data set for the response to 78 h at 8 °C is from Supplementary Table 3. Both data sets were imported into the MapMan functional categories. For each category, Wilcoxon's test was performed to estimate a P-value that the response of the genes assigned to that category was statistically different from the response of all the other genes on the array (see Fig. 6, where the results of this analysis are shown in a false colour code for the 78 h at 8 °C data set). The P-values were converted to a log10 scale to compress the range and then were plotted against each other. Categories that show coordinated changes in response to CBF over-expression are displaced along the x-axis, and categories that show coordinated changes in response to 78 h at 8 °C are displaced along the y-axis. Some of the categories relating to RNA and protein synthesis are labelled. (a) Full scale. (b) Expanded scale, with the most significant categories lying outside the range. PPR, pentatricopeptide; AA, amino acids; JA, jasmonic acid; MIP, myoinositol phosphate.

Many functional areas show similar responses to chilling and to over-expression of CBF (Table 4). Both repress genes involved in auxin, ethylene, brassinosteroid and jasmonate metabolism and signalling, several families of receptor kinases, vacuolar invertases, PIPs and TIPs, and several members of the AGP cell wall protein family (see Supplementary Table S7). Both treatments induce sets of genes for starch and raffinose metabolism, mitochondrial metabolite transporters, cytochrome oxidase and F1-ATPase, and flavonoid metabolism. Examples of single genes that are induced in both treatments include genes for fermentation (ADH1, PDC), a carbonic anhydrase (At3g51720), SUS5.2, anthocyanin synthesis, several phosphoinositide PLCs and many transcription factors, especially members of the AP2/EREBP family. Many of the metabolic responses to a small decrease of the temperatures like the accumulation of sugars, raffinose and proline (see Fig. 1) are also seen in response to over-expression of CBF (Cook et al. 2004).

Table 4.  Functional categories showing a statistically validated similar response to CBF over-expression to and chilling in wild-type Col-0
BinCodeBin namez-score
CBF
over-expression
z-score
78 h at 8 °C
Direction
  1. The data set for the response to CBF2 over-expression was downloaded from Vogel et al. (2005). The data set for the response to 78 h at 8 °C is from Supplementary Table S3. Both data sets were imported into the MapMan in order to assign functional categories. For each category, Wilcoxon's test was performed to estimate a P-value that the response of the genes assigned to that category was statistically different from the response of all the other genes on the array. The direction of the change was calculated from the average response of all the genes in the category; an increase and decrease is indicated as up and down, respectively. The table summarizes functional categories that show statistically significant changes in both treatments. CHO, carbohydrate.

2.2.1.3.3Major CHO metabolism.degradation.sucrose.invertases.vacuolar−2.20383−2.43457Down
7.1.2OPP.oxidative PP.6-phosphogluconolactonase−2.03382−2.00644Down
11.1Lipid metabolism.FA synthesis and FA elongation−2.9509−2.284Down
11.8Lipid metabolism.‘exotics’ (steroids, squalene, etc.)−2.44204−2.49935Down
12.2.2N metabolism.ammonia metabolism.glutamine synthase−2.05415−1.99198Down
26.16Misc.myrosinases-lectin-jacalin−3.00662−4.17873Down
26.8Misc.,nitrilases−2.90428−2.26393Down
34.19.1Transport.major intrinsic proteins.PIP−2.60598−3.19493Down
34.19.2Transport.major intrinsic proteins.TIP−3.51793−2.98393Down
17.2Hormone metabolism.auxin−3.58648−2.43134Down
17.3Hormone metabolism.brassinosteroid−2.44001−2.40736Down
17.5Hormone metabolism.ethylene−2.49192−3.21466Down
17.7.1Hormone metabolism.jasmonate.synthesis-degradation−3.64648−3.91214Down
17.7.3Hormone metabolism.jasmonate.induced-regulated-responsive-activated−2.46469−3.175Down
29.4.1.51Protein.post-translational modification.kinase.receptor-like cytoplasmatic kinase I−2.01065−2.34049Down
30.2.11Signalling.receptor kinases.leucine rich repeat XI−3.92837−2.74098Down
30.2.16Signalling.receptor kinases.Catharanthus roseus-like RLK1−3.39476−3.69583Down
30.2.17Signalling.receptor kinases.DUF 26−2.01553−3.34974Down
30.2.24Signalling.receptor kinases.S-locus glycoprotein-like−2.25013−2.67526Down
30.2.25Signalling, receptor kinases.wall-associated kinase−2.80285−3.19216Down
30.2.9Signalling.receptor kinases.leucine rich repeat VIII-2−3.85206−2.12474Down
30.3Signalling.calcium−1.96913−5.30132Down
31.1Cell.organization−4.11243−4.16253Down
35.1.21Not assigned.no ontology, epsin N-terminal homology (ENTH) domain protein−2.16604−2.77198Down
2.1.2Major CHO metabolism.synthesis.starch2.8426632.215404Up
3.1.1.1Minor CHO metabolism.raffinose family.galactinol synthases.known2.4293672.431883Up
34.9Transport.metabolite transporters at the mitochondrial membrane3.9300723.711267Up
9.7Mitochondrial electron transport/ATP synthesis.cytochrome c oxidase2.6037131.987229Up
9.9Mitochondrial electron transport/ATP synthesis.F1-ATPase2.961012.053094Up
16.8Secondary metabolism.flavonoids2.6128892.488395Up
26.21Misc.protease inhibitor/seed storage/lipid transfer protein (LTP) family protein2.1029283.36106Up
17.1Hormone metabolism.abscisic acid2.954212.802754Up
27.1.19RNA.processing.ribonucleases2.3185093.307151Up
35.1.5Not assigned.no ontology.pentatricopeptide (PPR) repeat-containing protein2.6209226.453819Up

However, there are also some clear differences. The most striking is that the coordinated induction of genes in categories related to RNA synthesis, amino acid activation and protein synthesis is absent from the CBF response (Fig. 9a,b). The experiments to define a core gene set of the CBF response were carried out with nutrient medium-grown plants (Vogel et al. 2005). To check that the response is similar in soil- and agar-grown plants, we carried out a similar analysis with the data set from Vogel et al. (2005) that compares soil-grown plants at 20 °C and after 24 h at 4 °C. The coordinated induction of genes in categories related to protein synthesis was visible within 1 d after transferring agar-grown plants to low temperature (Supplementary Fig. S4). Other components of the response to chilling that are absent from the published response to CBF over-expression include the induction of genes related to DNA packaging, and the repression of genes involved in photosynthesis and biotic stress (Fig. 9B & Supplementary Table S7). At the single gene level, there were differences in the responses of individual genes that were involved in sucrose and starch synthesis (e.g. SP3b, SPP4, TPT, GPT2), mitochondrial electron transport, ATP synthesis and mitochondrial electron transport, and tocopherol synthesis (Supplementary Table S7).

Our results add to the evidence (Zarka et al. 2003; Vogel et al. 2005) that the CBF orchestrates a broad response to moderate changes of temperature. It may operate as a rheostat, allowing plants to respond and adjust to a wide range of temperatures. The CBF response is functionally conserved in other freezing-tolerant species, like poplar (Benedict et al. 2006b). While parts of the CBF pathway are conserved in chilling-sensitive species like maize and tomato, their function appears to be compromised (Jaglo et al. 2001; Qin et al. 2004; Zhang et al. 2004). Tomato contains three CBF homologs, but only one (LtCBF1) is induced by low temperatures (Zhang et al. 2004). Whereas heterologous everexpression of LtCBF1 in Arabidopsis leads to induction of the CBF regulon and increased freezing tolerance, over-expression of AtCBF3 or LtCBF1 in tomato (Zhang et al. 2004) or AtCBF1 in rice (Lee et al. 2004) does not lead to widespread changes in gene expression or improved performance at low temperatures. It will obviously be interesting to investigate the transcriptional and metabolic response to chilling in sensitive species, in order to learn what aspects are shared with chilling-tolerant species like Arabidopsis, which aspects differ and whether any of the discrepancies affect parts ofthe response that are likely to be regulated by the CBF programme.

CONCLUDING REMARKS

The results presented in this paper reveal that a small decrease of the ambient temperature leads to marked changes of gene expression and metabolism, which bear a striking resemblance to the response at 4 °C except that the amplitude is smaller. Whereas the changes of transcript levels are larger at 6 h than at 78 h, many of the changes in metabolism increase between 6 and 78 h. This indicates that many of the transcriptional changes are an early response to perturbations in metabolism and cellular physiology, which are subsequently reversed. It will be interesting to investigate if there is a more complex response in chilling-sensitive species. The results show that small decreases of the temperature trigger the synthesis of cryoprotectants and other stress metabolites, and lead to a transcriptional up-regulation of protein synthesis and an increase of the protein content. Further, they reveal that the CBF programme makes an important contribution to the orchestration of the changes of expression and metabolism to small changes of the temperature. This conclusion is of considerable interest in the light of the emerging evidence that the biological activity of the CBF family is functionally attenuated in chilling-sensitive plants.

STATEMENT

Upon request, all novel materials described in this publication will be made available in a timely manner for non-commercial research purposes, subject to the requisite permission from any third-party owners of all or parts of the material. Obtaining any permissions will be the responsibility of the requestor.

ACCESSION NUMBERS

The raw data of the transcript profiles described under Materials and Methods were deposited in the public GEO repository (http://www.ncbi.nlm.nih.gov/geo) under the accession code GSE10522.

ACKNOWLEDGMENTS

We are grateful to Florian Wagner and his team at RZPD Berlin (German Resource Center for Genome Research, Berlin) for providing expert Affymetrix array service, including all steps from total RNA to data acquisition. This research was supported by the Max Plank Society and by the German Ministry for Research and Technology in the GABI (0312277A) and the GABI-COOL (0313111D) programmes.

Ancillary