Comparative transcriptome analysis reveals significant differences in gene expression and signalling pathways between developmental and dark/starvation-induced senescence in Arabidopsis


(fax 02476 574500; e-mail


An analysis of changes in global gene expression patterns during developmental leaf senescence in Arabidopsis has identified more than 800 genes that show a reproducible increase in transcript abundance. This extensive change illustrates the dramatic alterations in cell metabolism that underpin the developmental transition from a photosynthetically active leaf to a senescing organ which functions as a source of mobilizable nutrients. Comparison of changes in gene expression patterns during natural leaf senescence with those identified, when senescence is artificially induced in leaves induced to senesce by darkness or during sucrose starvation-induced senescence in cell suspension cultures, has shown not only similarities but also considerable differences. The data suggest that alternative pathways for essential metabolic processes such as nitrogen mobilization are used in different senescent systems. Gene expression patterns in the senescent cell suspension cultures are more similar to those for dark-induced senescence and this may be a consequence of sugar starvation in both tissues. Gene expression analysis in senescing leaves of plant lines defective in signalling pathways involving salicylic acid (SA), jasmonic acid (JA) and ethylene has shown that these three pathways are all required for expression of many genes during developmental senescence. The JA/ethylene pathways also appear to operate in regulating gene expression in dark-induced and cell suspension senescence whereas the SA pathway is not involved. The importance of the SA pathway in the senescence process is illustrated by the discovery that developmental leaf senescence, but not dark-induced senescence, is delayed in plants defective in the SA pathway.


Developmental senescence in ageing leaves is a complex and highly organized process resulting in many changes in gene expression and metabolic processes. Within a senescing leaf, the individual cells are usually at many different stages of senescence with the distal parts of the leaf senescing first and the cells surrounding the veins tending to remain active longer to maximize transport from the leaf. Thus, the study of senescence has been complicated by the lack of coordinated development of the cells within an individual leaf and various methods have been used to artificially induce senescence to obtain a synchronous process. For example, dark-induced senescence has been used frequently as a useful method to induce synchronous senescence as many typical senescence symptoms such as chlorophyll degradation and loss of protein occur. Also, suspension cultures can be used to synchronize developmental events and senescence-like symptoms are induced in ageing cell suspensions. In this paper, we compare gene expression in three different types of senescence and our results indicate that, while dark-induced or suspension culture senescence have many features common to developmental senescence and could be useful model systems for certain aspects, there are also considerable differences in other key processes that take place.

Leaf ageing in plants is accompanied by a genetically programmed senescence process, in which nutrients are mobilized from the dying leaves to support active growth in the younger parts of the plant (Hörtensteiner and Feller, 2002). The metabolic changes occurring during senescence are essential to support plant growth and reproduction and the cells of the senescing leaf must remain viable until the mobilization of salvaged nutrients is complete. Senescence in plants is highly regulated and depends upon the modulated expression of many different genes (Buchanan-Wollaston et al., 2003; Gepstein et al., 2003). The identification of genes that are up- or downregulated during leaf senescence has been a focus of a considerable amount of research and in recent years the putative functions of senescence-related genes has increased our understanding of the regulation and metabolic basis of this important developmental process.

Several different groups have identified genes that show increased expression during the onset of senescence in several plant species although the majority of experiments have been conducted with Arabidopsis (reviewed in Buchanan-Wollaston et al., 2003). The current number of such genes is approximately 150 (Gepstein et al., 2003) but is certainly not a complete picture. Guo et al. (2004) have identified a collection of over 2000 expressed sequence tag clones that represent genes that are expressed in senescing leaves of Arabidopsis. However, the proportion of these that show senescence-enhanced expression has not been determined. Microarray analysis has been used to investigate the expression patterns of several thousand genes during autumnal leaf senescence in poplar, and similar types of genes have been identified (Andersson et al., 2004). Recently Lin and Wu (2004) have used microarray analysis to identify a large group of genes that show differences in transcript abundance in response to dark treatment of Arabidopsis.

Swidzinski et al. (2002) showed that Arabidopsis suspension cultures undergoing natural ageing, or after heat treatment, expressed symptoms typical of programmed cell death (PCD). They found that a number of senescence-related genes were upregulated in these treated cells indicating that there were common processes occurring. Plant PCD occurs in a wide range of specialized situations such as response to pathogen attack, waterlogging and in various aspects of development including senescence. PCD in plants exhibits many different phenotypes but is usually characterized by controlled vacuolar collapse and DNA laddering (Jones, 2001). DNA laddering has been seen very rarely in leaf senescence and vacuolar collapse may only occur in the very final stages (Lee and Chen, 2002). However, senescing cells eventually die and the mechanism of this may have processes in common with other PCD events.

In this paper, we report the use of an Affymetrix GeneChip system to identify genes that are upregulated in developmental senescence in Arabidopsis. Gene expression data from the dark-induced senescence experiments described by Lin and Wu (2004) and also data from Arabidopsis cell suspension cultures undergoing starvation-induced senescence and showing clear symptoms of PCD (Swidzinski et al., 2002) were used to compare and contrast gene expression changes that occur in developmental senescence with those taking place in these two other types of senescence. Exploitation of the global expression patterns obtained in these three separate data sets has allowed us to show clear differences between the three types of senescence.

Gene expression during senescence is under the control of a complex combination of signalling pathways, some of which have key roles in other plant processes such as in response to environmental stresses. For example, the ethylene signalling pathway has a role in modulating the rate of leaf senescence, mutants defective in this pathway show delayed senescence (Grbic and Bleecker, 1995). Also the jasmonic acid (JA) pathway is involved in senescence, treatment with JA induces premature senescence and various senescence-enhanced genes are expressed in response to JA treatment (He et al., 2002). However, mutants affected in JA signalling do not show obvious alteration in senescence phenotype. Salicylic acid (SA), a key signalling molecule in plant pathogen responses, has also been shown to be required for expression of some senescence-enhanced genes (Morris et al., 2000). The elucidation of key senescence-specific pathways is complicated by the crosstalk between these non-specific stress-induced pathways. We have extended the microarray analysis to examine the role of the ethylene, JA and SA pathways on gene expression during senescence. Comparisons made with expression of these genes in the other types of senescence or PCD has allowed us to determine the role of these pathways in dark-induced and cell suspension senescence.

Results and discussion

Genes showing increased transcript abundance in leaf senescence

The array data described in this paper are based on levels of hybridization to Affymetrix GeneChip microarrays. This determines transcript abundance and may not necessarily represent increased rates of transcription of individual genes as post-transcriptional regulation will vary between different genes. Also, transcript levels do not necessarily reflect the amount of final active protein product. However, for simplicity, increased transcript abundance is often referred to as induced gene expression.

Arabidopsis Affymetrix GeneChip arrays were probed with RNA from mature green leaves (before flowering) and from leaves in the mid senescent stage (5–15% visible chlorosis). In the green leaf sample, expression of 8538 genes was detected, while in the senescent leaf sample, expression of 9152 genes was detectable. This suggests that, of the 24 000 genes represented on the array over 30% are expressed at a detectable level in each of the leaf samples.

Genes showing increased abundance in leaf senescence were identified using array data from two independent experiments and genes that showed at least a threefold upregulation were selected. This resulted in 827 upregulated genes (Table S1). This list of genes has been generated by combination of the two different data sets, which strengthens confidence in the data as the genes selected show the same relative changes in abundance during senescence under two different growth regimes. Thus, an extensive new set of senescence-enhanced genes has been identified.

The genes listed in Table S1 provide a valuable indication of the metabolic activities that are changing in the leaf when senescence takes place. The genes have been sorted depending on the likely role of each gene that is upregulated in senescence, according to their annotation (TIGR v. 5.0). This confers a potential function for each encoded protein but in many cases this is due to similarity to known proteins and may not reflect their true activity. The range of gene functions represented in the 827 senescence-enhanced genes is illustrated in Figure 1 and summarized in Table 1. Some of these groups are discussed in the next section.

Figure 1.

Potential functions of senescence-enhanced genes.
The 827 genes that show at least threefold increase in transcript abundance were sorted into putative functional groups according to their TIGR v. 5.0 annotation as shown in Table S1 and Table 1. These groups are illustrated by the central pie chart with extra charts used to divide the genes encoding putative regulatory proteins and genes involved in macromolecule degradation into smaller functional groups.

Table 1.  Potential functional groups of genes showing increased transcript abundance in developmental leaf senescence (from Table S1)
Gene classPutative functionNumberExamples
RegulationTranscription factors96AP2 domain, bZIP, CCAAT binding, Leu zipper, Myb, NAC domain, WRKY, zinc finger, other DNA binding
Protein:protein interaction9Armadillo, WD-40
Ubiquitination control30F-box, ubiquitin binding, RING finger
Protein kinase/phosphatase66Protein kinase, LRR domain, protein phosphatase
Signalling17RALF, RelA/SpoT, remorin
Calcium binding13C2 domain, EF hand, calmodulin binding
Hormone pathways17Auxin responsive, ethylene responsive, ABA signalling, cytokinin oxidase
Macromolecule degradation and mobilizationProtein degradation29Aspartyl and cysteine proteases, serine carboxypeptidase, vacuolar processing
Amino acid degradation and N mobilization27Glutamine synthetase, homogentisate 1,2-dioxygenase, lysine-ketoglutarate reductase, uricase
Nucleic acid degradation and phosphate mobilization14Nucleases, acid phosphatase
Lipid degradation and mobilization29Lipase, α-dioxygenase, acyl-CoA dehydrogenase
Chlorophyll degradation2ELIP protein, Rieske domain protein
Sulphur metabolism 2Allinase, ATP-sulfurylase
Carbohydrate metabolism 63Amylase, β-1,3-glucanase, polygalacturonase, exotosin, glucosyl hydrolases, invertase, pectinesterase, glucosyl transferases
Lignin synthesis 3Cinnamyl-alcohol dehydrogenase, caffeoyl-CoA 3-O-methyltransferase
Transport 74ABC transporters, amino acid permease, cation exchangers, MATE efflux, sugar and peptide transporters
ATPases 7AAA type, plasma membrane ATPase
Metal binding 10Ferritin, copper chaperone, zinc binding
Stress relatedAntioxidants11Alternative oxidase, glutaredoxin, peroxidase
Stress and detoxification17DNAJ, heat shock, glutathione S transferase
Defence related11Chitinase, isochorismate synthase, major latex protein, osmotin
Secondary metabolismAlkaloid biosynthesis9Strictosidine synthase, reticuline oxidase
Flavonoid/anthocyanin pathway19Chalcone synthase, dihydroflavonol reductase, leucoanthocyanidin dioxygenase
Autophagy 5Autophagy genes APG7, 8a, 8b, 8h, 9
Structural 4Actin binding, myosin, tropinone reductase
Unclassified enzymes unknown role 110Cytochrome p450s, copper amine oxidase, dehalogenases, short-chain dehydrogenases
Unknown genes 132Unknown proteins

Regulatory genes.  The massive shift in gene expression patterns and metabolic functions that occurs during senescence is reflected in the substantial increase in transcript abundance of putative regulatory genes. Around 100 genes encoding putative transcription factors show increased expression. These include MYB factors, AP2 domain proteins, zinc finger proteins and a number of DNA- and RNA-binding proteins (listed in Table S1). In addition, 24 members of the family of NAC domain proteins show enhanced expression. In Arabidopsis, there are around 100 of these genes, which contain a plant-specific, highly conserved N-terminal domain (Ooka et al., 2003) and certain members have been implicated in different developmental or stress response processes (e.g. Fujita et al., 2004). A similar range of transcription factors expressed during developmental senescence (Guo et al., 2004) and showing increased expression in dark-induced senescence (Lin and Wu, 2004) has been described previously.

A number of genes involved in potential ubiquitination pathways are upregulated and this indicates that protein degradation via the 26S proteosome is active during senescence. Genes expressed encode a number of F-box proteins, C3HC4-type RING finger proteins and other members of the ubiquitin ligase complex such as the ASK1 protein and E2 conjugating enzymes. The presence of a RING finger domain is a characteristic of RING-class E3 ubiquitin protein ligases that act by transferring ubiquitin from an E2 enzyme to a substrate protein that is targeted for degradation. F-box proteins are also a variable component of an ubiquitin ligase complex and different E3 ligase complexes allow the targeting of specific proteins for degradation (reviewed in Smalle and Vierstra, 2004). The upregulation of a number of these genes during senescence indicates that specific targeting of proteins for degradation may be an important mechanism for the control of senescence. The importance of such genes in the control of senescence has been illustrated by the discovery of the F-box gene ORE-9; mutation in this gene results in a delayed senescence phenotype, indicating that the target of this protein has an important role in senescence (Woo et al., 2001).

Many genes involved with protein modification, such as protein kinases and phosphatases, are upregulated during senescence indicating that kinase-signalling cascades are likely to function during senescence. Expression of several genes involved with calcium regulation is increased including genes encoding calcium and calmodulin-binding proteins. Calcium signalling may be an important component in the regulation of senescence processes. The involvement of calcium in many types of cell death has been described (Jones, 2001) and increased levels of calcium ions have been observed to correlate with senescence of parsley mesophyll cells (Huang et al., 1997).

Two RelA/SpoT genes (RSH2 and RSH3) are upregulated; these proteins have been suggested to play a part in ppGpp signalling in plant stress responses (van der Biezen et al., 2000). Also expressed are two RALF (rapid alkalinization factor) and a phytosulphokine gene both of which have been postulated as polypeptide hormones (Ryan et al., 2002). RALF proteins respond to pH changes in cells and appear to have roles in plant development or stress responses (Haruta and Constabel, 2003).

Macromolecule degradation.  Many senescence upregulated genes encode enzymes likely to be involved in degradation of macromolecules and mobilization of metabolites. Many types of protease, including cysteine, serine, aspartyl and vacuolar processing proteases are found. Also, genes encoding enzymes that degrade amino acids such as lactoylglutathione lyase which is involved in threonine degradation and proline oxidase, involved in proline degradation are upregulated. Mobilization of nitrogen from the senescing cell to other parts of the plant is likely to occur via amino acid transport in the phloem. Several genes encoding enzymes involved in glutamate/glutamine metabolism are upregulated including two glutamate decarboxylase genes, two glutamate receptor proteins and three cytosolic glutamine synthetases. It has been previously reported that much of the N transported from senescing leaves is in the form of glutamine (Finnemann and Schjoerring, 2000). Many genes encoding amino acid permeases (eight genes) and peptide transporters (seven genes) are upregulated indicating increased transport from the leaf cells.

Three genes that have a role in ureide metabolism (uricase, xanthine dehydrogenase and allantoinase) show increased expression. This may indicate a role for allantoin or other ureides as alternative N-rich molecules that could be transported from the senescing leaf. Alternatively, or additionally, these enzymes could be involved in mobilization of N from nucleic acid degradation. Massive degradation of RNA occurs during senescence and several nuclease genes are upregulated. Nucleic acid breakdown is not only a source of nitrogen but also an important source of phosphorous. Four different acid phosphatase genes are upregulated; the role of these proteins in plants is not clear but it has been postulated that they may be involved in phosphate acquisition (Duff et al., 1994). Increased levels of these proteins in senescing leaves may indicate a role for these proteins in the storage or mobilization of phosphate released by nucleic acid degradation. Phosphate transport from the senescing leaf may be via the AtPT2 gene which is also upregulated. AtPT2 was identified in phosphate starved Arabidopsis leaves and shown to have a role in phosphate transport (Muchhal et al., 1996).

Many of the genes involved in the chlorophyll degradation pathway are expressed constitutively in the leaf, implying that the pathway is not under transcriptional control (reviewed in Hörtensteiner, 2004). Recently, the acd1 gene (accelerated cell death 1) was shown to encode the enzyme pheophorbide a oxygenase (Pruzinskáet al., 2003), a key enzyme in chlorophyll degradation. This gene showed increased expression during leaf senescence (Table S1) but enzyme activity was increased to a greater extent indicating the possibility of post-transcriptional control (Pruzinskáet al., 2003). The upregulated expression of the gene encoding early light-induced protein (ELIP) may be linked to chlorophyll degradation. The ELIP proteins bind free chlorophyll and have been postulated to have a role in protection from oxidative stress caused by the release of phototoxic free chlorophyll. This protein may have an important role in the initial stages of degradation and bind the chlorophyll as it is released from the pigment protein complexes (Binyamin et al., 2001; Hutin et al., 2003).

Transport.  Many transporter genes show increased transcript levels during senescence including a number of ABC transporters and several sugar, peptide, amino acid and cation transporters, many of which are associated with metal transport of H+ exchange. The upregulation of a range of transporters, especially those involved in sugars, amino acids and peptides, reflects the new function of a senescing leaf, as it becomes a source tissue for the mobilization of nutrients to the rest of the plant.

Antioxidant genes.  During senescence, maintaining protection against oxidative stress is important to protect the cells from premature death and delayed senescence mutants have been found to have increased tolerance to oxidative stress (Woo et al., 2004). Two genes encoding alternative oxidases (AOX) are upregulated in senescing leaves. The role of the AOX proteins is thought to limit ROS formation by reducing the activity of the electron transport chain thereby reducing oxidative stress in the mitochondria (Millenaar et al., 1998). Overexpression of AOX in tobacco cell cultures was shown to reduce the overall ROS levels (Maxwell et al., 1999). In addition, three different glutaredoxin genes are upregulated during senescence as well as two peroxidase genes (Table S1). Also, genes encoding enzymes involved in the synthesis of tocopherol such as 4-hydroxyphenylpyruvate dioxygenase and homogentisate phytylprenyltransferase are upregulated. Tocopherol, a free radical scavenger, has been previously implicated as an antioxidant in senescing leaves (Falk et al., 2003).

Autophagy.  An interesting class of genes that are upregulated in tandem during senescence is a group encoding autophagy-related proteins. Autophagy is a regulated recycling process where double-membrane-bound vesicles, known as autophagosomes, traffic cytosolic contents or organelles to the vacuole where they are broken down and reallocated to essential processes. The role of autophagy in senescence is not clear but it is likely that some of the massive degradation of cellular components could occur via this mechanism. Disruption in Arabidopsis autophagy genes has been shown to result in accelerated leaf senescence (Doelling et al., 2002; Hanaoka et al., 2002) and this may indicate that controlled degradation via autophagy is necessary to stabilize cellular viability and carry out efficient senescence.

Gene expression in other types of senescence

Dark-induced senescence.  One of the problems that is encountered when studying developmental senescence in Arabidopsis is the lack of consistency in the rates of senescence between leaves and also in different cells within the same leaf. Many research groups have exploited the use of dark-induced senescence as a convenient method to synchronize the senescence process as many of the events that occur in dark treated leaves are known to mirror those that occur in developmental senescence. However, there have been concerns expressed about the validity of this method as a direct comparison to natural senescence (e.g. Becker and Apel, 1993). Weaver et al. (1998) showed that several senescence-associated genes were differentially expressed after stress treatments including darkness. Lin and Wu (2004) recently carried out microarray analysis to characterize gene expression during dark-induced senescence in Arabidopsis leaves and they identified many metabolic pathways that were altered. The availability of global gene expression data for developmental and dark-induced senescence has allowed us to carry out a direct comparison of gene expression in natural developmental senescence with that in dark-induced senescence to clarify the different and common processes that occur in the two types of senescence.

We used the Affymetrix expression data to compare the two types of senescence. Genes showing a threefold expression change between untreated plants and 5-day dark-treated plants were identified and compared with the gene list described above. Of the 827 senescence-enhanced genes, 53% (437) were also at least threefold upregulated in the dark-treated leaves. However, perhaps surprisingly, 34% (277) of the genes were not upregulated at all in the dark (the remaining 13% were upregulated two- to threefold). The effect of dark treatment on the expression of senescence-enhanced genes can be seen in Table S1 and many differences are obvious. In addition, there were nearly 2000 genes that were at least threefold upregulated in the dark treatment, meaning that around 1500 of these were not altered in developmental senescence. Examination of the differences in gene expression between these two types of senescence offers a new insight into the events taking place.

Starvation-induced senescence in suspension culture cells

Programmed cell death is observed in heterotrophic suspension cultures that are starved of sucrose and allowed to enter a senescent phase. The hallmark features of PCD such as DNA laddering and membrane shrinkage are observed in these ageing and dying cultures (Swidzinski et al., 2002). An Affymetrix GeneChip analysis was carried out using RNA isolated from suspension culture cells that had been growing for 10 days without subculturing and gene expression patterns were compared to control cultures that had obtained the correct culturing regime with fresh sucrose (as described in Swidzinski et al., 2002). Genes that showed at least a threefold increase in transcript abundance in the starved cultures were identified and subjected to a similar analysis as that described above for the dark-induced senescence experiment. Of the 827 senescence-enhanced genes, 326 also showed at least threefold upregulation in the senescing suspension culture cells (data shown in Table S1). However, this means that around 500 of the genes upregulated in developmental senescence were not upregulated in senescing suspension cultures. Also, over 1000 genes showing at least a threefold increased expression in the senescing cultures were not upregulated in leaf senescence.

The expression of genes in dark-induced senescence was included in a combined analysis of the three types of senescence (Figure 2). The majority (65%) of the genes that were common to developmental and cell culture senescence also showed increased expression during dark-induced senescence. Thus, there is a group of 229 genes that are expressed in all three types of senescence and a group of 218 genes only upregulated in leaf senescence. Also of interest is a larger group of 308 genes that show increased expression only in dark-induced and cell culture senescence and not in leaf senescence. This may imply that there is more in common between dark- and starvation-induced senescence than with developmental leaf senescence.

Figure 2.

Venn diagram to illustrate the number of genes upregulated in the three different types of senescence.
Numbers of genes in each category are shown. All the lists represented in the developmental senescence groups can be determined from Table S1. The groups only include genes that show at least a threefold upregulation in the relevant experiments.

Comparative analysis of genes expressed in three types of senescence

The gene expression data were analysed to identify groups of genes that showed different expression patterns in the three types of senescence. There is likely to be more noise in the dark-induced and cell suspension senescence experiments compared with the leaf senescence data as this comes from a combination of two independent experiments. Therefore, the numbers of genes differentially expressed only in dark senescence or cell suspension senescence is probably higher than if more replicates had been analysed. However, in spite of this, a study of general trends can be carried out and obvious differences between the three types of senescence can be identified.

Differential expression in the different types of senescence

Hormone signalling pathways.  Analysis of genes involved in different hormone signalling pathways allows us to assess their relative importance in different types of senescence. Abscisic acid (ABA) has been implicated in the regulation of stress-induced senescence (Yang et al., 2003) and expression of several senescence-enhanced genes is induced on treatment with ABA (Weaver et al., 1998). The gene encoding the key enzyme in ABA biosynthesis, 9-cis-epoxycarotenoid dioxygenase (NCED), and two aldehyde oxidase genes AAO1 and AAO3, which are involved in ABA biosynthesis (Seo et al., 2000), show increased transcript abundance, indicating that ABA levels probably increase in senescing leaves. Also, two protein phosphatase genes, implicated in early events of ABA signalling, ABI1 and ABA2 (Leung et al., 1997), are upregulated (Table 2). Investigation of the expression patterns of these genes in dark-induced and cell suspension senescence shows that most of these genes are also upregulated (Table 2) which indicates that the ABA signalling pathway is active in all three types of senescence.

Table 2.  Expression changes in genes involved in ABA and cytokinin biosynthesis and signalling in different types of senescence
LocusAnnotationLeaf (sen/MG)Dark (dark/con)Cell suspension (sen/con)
  1. Yellow boxes indicate over threefold increase in expression compared with the control (green leaf or untreated sample), blue boxes indicate over threefold downregulation, grey indicate undetectable expression in that experiment.

  2. aSenescence-enhanced in ATGE experiment only.

inline image

Cytokinin levels are reduced in senescing leaves and this is thought to be one of the key signals for the initiation of senescence (Gan and Amasino, 1995). A number of cytokinin inducible genes such as the type A response regulators, ARR4, ARR6, ARR7 and ARR9, that have key roles in cytokinin signalling and two genes involved in cytokinin synthesis, an isopentenyl phosphotransferase and a cytokinin synthase gene, show reduced transcript abundance in senescence (Table 2). Parallel to that is an increase in a cytokinin oxidase transcript that may be involved in cytokinin degradation.

However, analysis of expression patterns of these genes does indicate a difference between the three types of senescence (Table 2). Many of the cytokinin signalling genes that are downregulated in developmental senescence are not expressed at all in suspension culture cells (senescent or control) and this indicates that loss of cytokinin is unlikely to be a signal for the senescence/cell death processes that are occurring in these cells. Cytokinin signalling appears to be affected in dark-induced senescence as reduced expression of cytokinin synthase and one of the cytokinin-related response regulators is evident. Also, increased expression of cytokinin oxidase occurs in dark-induced senescence. Interestingly, increased expression of an ARR1-like type B response regulator gene (represented by two probes on the array) is seen in both dark-induced and cell suspension senescence. Type B ARRs act as transcriptional activators of the type A ARR genes but their N-terminal domains may negatively regulate this activation in the absence of cytokinin (reviewed in Hutchinson and Kieber, 2002).

To help in the investigation of the putative functions in the different groups of genes, we made use of the mapman programme recently described by Thimm et al. (2004). For this programme, each Arabidopsis gene of known or predicted function has been assigned to a particular functional compartment (bin). Entering the array data for a series of genes allows the display of the relative changes in expression of different genes on to diagrams of metabolic pathways.

Nitrogen mobilization. mapman data for metabolic pathways for the genes expressed only in leaf senescence showed few easily identifiable pathways (Figure S1a). However, it does show that genes involved in ammonia assimilation, such as three cytosolic glutamine synthetase genes, are not expressed in dark-induced or suspension culture senescence (Table 3). These enzymes are thought to have a key role in N mobilization from senescing leaves and this may imply that N mobilization does not occur in the dark (or an alternative pathway may operate, see below). Two genes encoding glutamate decarboxylase are also only upregulated in leaf senescence. Glutamate decarboxylase is involved in the conversion of glutamate to GABA (4-aminobutyrate). The role of GABA in the plant is unclear but it may have a role as a signalling molecule to coordinate the C:N balance in changing nutrient environments as is occurring during senescence (reviewed in Bouche and Fromm, 2004). It is thought that the C:N balance in plants could be controlled via a family of glutamate receptors (ATGLRs) (Kang and Turano, 2003) and two members of this family show increased expression in senescence (Table S1). Metabolism of GABA is via the GABA shunt, which involves three enzymes including the cytosolic glutamate decarboxylase and mitochondrial enzymes GABA transaminase and succinic semialdehyde dehydrogenase. This pathway has a potential role in reducing oxidative stress in the mitochondria (Bouche et al., 2003) and it is possible that this could also have an important protective role in senescence.

Table 3.  Metabolic genes that show differential expression in the three types of senescence
LocusAnnotationLeaf (sen/MG)Dark (dark/con)Cell suspension (sen/con)
  1. Colour representation as in Table 2.

inline image

In contrast, two genes encoding glutamate dehydrogenase, and the genes encoding aspartate amino transferase and asparagine synthase are upregulated in dark-induced and suspension culture senescence but not in developmental senescence (Figure S1b; Table 3). Lin and Wu (2004) proposed a novel pathway involving these genes for the generation of asparagine for export from dark-treated leaves. It may be that glutamine is the N metabolite mobilized in developmental senescence, while asparagine is primarily mobilized in dark-induced senescence. Lin and Wu (2004) showed an increase in asparagine levels in the dark-treated Arabidopsis leaves used for this analysis. An increase in asparagine levels in parallel with a decrease in glutamine levels has been shown in dark-treated Arabidopsis plants (Lam et al., 1995) and in maize seedlings (Brouquisse et al., 1998).

However, PPDK expression is induced in all three types of senescence and as genes encoding glyoxylate cycle enzymes are not induced (see below), the role of this gene in developmental senescence may be to provide precursors for glutamine synthesis via PEP carboxylase and/or PEP carboxykinase and the TCA cycle.

Lipid catabolism.  Many lipid catabolism genes appear to be upregulated in dark-induced senescence but many of these do not show a parallel increase in leaf senescence (Figure S1b). Lin and Wu (2004) used the lipid catabolism genes defined by Graham and Eastmond (2002) to examine their expression in dark-induced senescence and concluded that many genes involved in β-oxidation of lipids were upregulated. We have now compared the expression of these genes in all three types of senescence and large differences are evident (Table 3). Interestingly there are many more lipid catabolism genes upregulated in dark and cell suspension senescence than in leaf senescence. The raw data for these experiments show that these genes are expressed at a considerably higher level in dark-induced senescence. Similarly, sucrose degradation genes are upregulated in dark-induced senescence and to a lesser extent in cell suspension senescence (Table 3). The parallel regulation of these genes in dark-induced and cell suspension senescence is likely to be related to carbohydrate starvation. Similar effects of sucrose starvation have been seen in other studies on Arabidopsis plants (Thimm et al., 2004) and sucrose-starved suspension cultures (Contento et al., 2004). Loss of photosynthetic fixed carbon, when plants are placed in the dark, results in rapid depletion of the sugar levels in the leaves (Brouquisse et al., 1998; Tcherkez et al., 2003). Sugar levels in the medium supplying the senescent suspension culture cells fall to below 20% of the original level (JS, unpublished data). Carbohydrate starvation has been shown to result in significant induction of β-oxidation activity (Graham and Eastmond, 2002) and this is reflected by the observed upregulation of many genes in dark-induced and cell suspension senescence. In developmental leaf senescence, endogenous sugar levels tend to increase which would explain why there is no extensive induction of β-oxidation genes. However, there is some suggestion that senescence in petals could be induced by sugar starvation as external application of sugar can delay senescence (van Doorn, 2004).

Interestingly, genes involved in the glyoxylate cycle, such as malate synthase and isocitrate lyase do not show any increase in expression in any of the types of senescence. These genes have been shown to be upregulated in senescing tissues of other plants such as cucumber (reviewed in Graham and Eastmond, 2002) and are required for the conversion of lipid to sugar via gluconeogenesis. This implies that the acetyl CoA released from the β-oxidation of fatty acids is respired directly in Arabidopsis and is not used for gluconeogenesis.

Two genes involved in branched chain amino acid (leu, ile and val) catabolism, BCKDH subunits E1 and E2 are upregulated in dark-induced and cell suspension senescence but not in developmental leaf senescence (Figure S1b; Table 3) (Fujiki et al., 2001). Branched chain amino acid catabolism may be activated specifically in low carbohydrate conditions in order to provide an alternative carbohydrate source to the cell. β-Oxidation is required for complete degradation of the ketoacids.

Genes involved in α-oxidation of lipids show increased transcript abundance in all three types of senescence (Table 3) and may have a role in a response to stress (De Leon et al., 2002; Hamberg et al., 1999). Lipid 2-hydroperoxides, which are produced by α-oxidation, may be precursors of the oxylipin signalling molecules, including JA. The increased expression of these genes may be related to a stress response rather than having a fatty acid degradation role.

Trehalose metabolism.  Another pathway that shows altered expression in dark-induced and cell suspension senescence but not developmental senescence is that involved in trehalose metabolism (Figure S1; Table 3). Trehalose is synthesized from glucose by the action of trehalose-6-phosphate synthase (TPS) and trehalose-6-phosphate phosphatase (TPP). The role of trehalose in plants is not clear but the pathway is essential for development (Schluepmann et al., 2003) and there is evidence that trehalose-6-phosphate is required for regulation of sugar metabolism (Eastmond et al., 2003). Genes encoding TPS and TPP are upregulated in dark-induced senescence. The induction of this pathway has previously been reported in plants exposed to a long night period, and in sugar-starved suspension cultures by Thimm et al. (2004) and Contento et al. (2004), respectively. It is likely that the induced expression of this pathway is to try to redress the sugar imbalance caused by the dark or starvation treatments and thus this pathway is not required during developmental senescence.

Flavonoid biosynthesis.  Another group of genes that stand out as being differentially expressed in the three types of senescence are those genes involved with flavonoid biosynthesis. mapman analysis shows that approximately the same number of genes that have been classified in this group are expressed in developmental and dark-induced senescence (Figure S1a,b) but a closer examination of these indicates that only one is common to both (Table 4). It appears that a different flavonoid biosynthesis pathway may operate in the absence of light. Flavonoid pathway enzymes such as those encoded by chalcone synthase (At5g13930) and dihydroflavonoid reductase (At5g42800) were considered to be represented by single copy genes in Arabidopsis as mutants in these cause a transparent testa phenotype and flavonoids were absent in other parts of the plant (reviewed in Winkel-Shirley, 2001). However, this flavonoid biosynthesis pathway is regulated in leaves by light in a diurnal fashion (Harmer et al., 2000) and therefore may not be activated in dark-induced senescence. This may indicate specific regulation of an alternative pathway in dark-induced senescence with increased expression of potential alternatives to these genes, e.g. At1g02050 (chalcone and stilbene synthase family, similar to rice chalcone synthase), At4g35420 (dihydroflavonol reductase family) and two different flavonol synthase genes. In cell suspension senescence this pathway appears not to be induced (Table 4). This raises the question of the role of this pathway in senescence. In developmental leaf senescence, it is thought that increased light stress due to the degradation and release of chlorophyll increases the need for protective flavonoid and anthocyanin in the senescing tissues. Indeed this is why leaves turn orange and red in the autumn (Feild et al., 2001). In dark-induced senescence different stresses might be evident, macromolecule degradation results in the release of ROS and increased flavonoid production might have a protective role. However, the same stresses are present in the cell suspension senescence but no similar pathway is expressed.

Table 4.  Genes that may have a role in flavonoid biosynthesis
LocusAnnotationLeaf (raw)Leaf (sen/MG)Dark (raw)Dark (dark/con)Cell suspension (raw)Cell suspension (sen/con)
  1. Colour representation as in Table 2.

inline image

Role of signalling pathways in gene expression during senescence

The initiation and progression of plant senescence appears to involve a complex combination of signalling pathways with considerable crosstalk between other plant responses. The elucidation of the role and importance of each of these pathways will help in the clarification of the genes and signals that are senescence-specific. As described above, three well-characterized stress response pathways (those involving ethylene, JA and SA signals) have been implicated in senescence (Grbic and Bleecker, 1995; He et al., 2002; Morris et al., 2000). Levels of these signalling molecules increase during senescence and induce the expression of specific genes. We have used Affymetrix array experiments to identify alterations in senescence-related gene expression in plants that are defective in each of these pathways. This has allowed us to group the senescence-enhanced genes into classes that are dependent on one or other signalling pathways and identify the genes that are independent of these stress-related signals.

Of the 827 genes that are upregulated during senescence, 19% (165) show at least twofold reduced expression in the NahG transgenic plant that is defective in the SA signalling pathway (Gaffney et al., 1993); 12% (103 genes) show reduced expression in the coi1 mutant defective in the JA signalling pathway (Xie et al., 1998) and 9% (73 genes) show reduced expression in the ethylene signalling mutant ein2 (Alonso et al., 1999). The expression ratio for each of the senescence-enhanced genes in senescing leaves of each mutant compared with the wild type is shown in Table S1. The senescence-related expression of the remaining 575 senescence-enhanced genes is therefore independent of these signalling pathways although more replicate experiments would be needed to prove that there was no effect of the mutation. There are several genes that appear to be downregulated in two of the three mutants and a few that require all three pathways for optimum expression. For example, 26 genes are downregulated in both coi1 and ein2 backgrounds and 25 genes show lower expression in NahG and coi1. A small number of genes are upregulated in each mutant and this may indicate that the pathway that controls these genes is suppressed by the presence of the other signalling pathway. For example, several genes that require the JA and ethylene pathways for expression are upregulated in the NahG background. It has been shown previously that the presence of the SA signalling pathway represses gene expression induced by the JA/ethylene pathway system during pathogen responses (Gupta et al., 2000) and the same may be true in senescence.

Jasmonate and ethylene pathways.  Examination of the genes that are altered in each mutant background gives an indication of which genes may be controlled by each pathway. A large proportion of the genes altered by the mutants in the JA and/or ethylene pathways encode hydrolases (Figure 3) and many are present in the carbohydrate metabolism section of Table S1. Genes that are downregulated in both coi1 and ein2 mutants are shown in Table 5 and have been divided into three groups. These genes mostly encode proteins such as polygalacturonase and pectinesterase that may have a role in cell wall degradation. Also two nuclease genes RNAse1 and BFN1 are expressed at a lower level in the mutants. Therefore, some senescence-related degradation functions (carbohydrates and nucleic acids) may depend on signalling pathways involving JA and ethylene. The presence of the potential regulator, the NAC domain gene, NAC3, in this group indicates a potential role for this gene in controlling this aspect of senescence. Expression of the alpha-DOX1 (At3g01420) is dependent on both JA and ethylene pathways and is also slightly downregulated in the NahG plants. This gene (described above as being involved in α-oxidation of fatty acids; Table 2) has been shown to have a role in oxylipin production protecting plants from oxidative stress and cell death (De Leon et al., 2002). In this paper, expression of the alpha-DOX1 gene was shown to require the presence of the SA pathway but JA and ethylene signalling was not examined. The expression of this gene, which probably leads to JA and other oxylipin biosynthesis is reduced in the absence of JA and ethylene signals and this may indicate a feedback control of this signalling pathway.

Figure 3.

Functional categories of genes dependent on different hormone signalling pathways for senescence-enhanced expression.
Each senescence-enhanced gene that was at least twofold downregulated in any of the three hormone-deficient backgrounds was assigned a putative molecular function according to their GO annotations (Gene Ontology Consortium, 2000). Pie charts show the proportions of each function that was reduced in expression in the Coi1, Ein2 mutant and NahG transgenic plants.

Table 5.  Senescence-enhanced genes showing reduced expression in Coi1 and Ein2 mutants
IdentifierAnnotationSen/MGNahGCoi1Ein2Dark/lightCell death/control
  1. Colour representation as Table 2.

inline image

Some genes that depend on JA and ethylene show increased expression in the absence of the SA pathway. Examples of such genes are a chitinase, a glucanase and an osmotin-like gene (thaumatin-like) all of which have been implicated in pathogen responses. There are also a few genes that appear to be dependent on all three signalling pathways for expression. This group includes a monooxygenase, a putative subtilase and an ABC transporter. All genes shown in Table 5 are senescence-enhanced and it appears that even within this small group there are three overlapping but distinct pathways that control their expression.

The majority of the genes that depend on JA and ethylene for expression during leaf senescence show increased abundance during both dark-induced and cell suspension senescence (Table S1, Table 5). This indicates that these pathways have an active signalling role in these other two types of senescence.

Salicylic acid pathway.  The majority of the genes that are dependent on the SA pathway encode kinases, transferases and hydrolases (Figure 3; Table S1). The SA pathway has a key role in the disease resistance response and this may be reflected in the fact that there are many putative disease resistance leucine-rich repeat genes dependent on the SA pathway for expression during senescence. Hydrolases that show dependence on the SA pathway include a class IV chitinase and also the senescence-specific protease SAG12. This result is supported by previous work in which we found that expression of both these genes was downregulated in several mutants defective in the SA pathway (Morris et al., 2000). Other hydrolases downregulated in NahG include three different AAA-type ATPases all of which are in close proximity in the genome. AAA proteins act in a variety of cellular functions, including cell cycle regulation, protein degradation, organelle biogenesis and vesicular transport; hence, the origin of their name (ATPases associated with different activities). Dependence on the SA pathway for expression may implicate these genes in pathogen responses or may indicate a senescence-specific role for the SA pathway. Similarly, several genes involved in calcium signalling and a group of putative glucosyl and fucosyl transferases do not show senescence-enhanced expression in the NahG transgenic plant.

Significantly, many of the genes that are the most reduced in expression (over fourfold) in the NahG senescent leaves do not show increased expression in dark-induced or in cell suspension senescence (Table 6; Table S1). Genoud et al. (2002) reported that the efficient activation of the SA signalling pathways during pathogen responses was dependent on light signals and both SA-induced gene expression and the hypersensitive response were strongly reduced in tissues in the dark. The same is evident in dark-induced senescence.

Table 6.  Developmental senescence-enhanced genes dependent on SA pathway (at least fourfold downregulated in NahG), not upregulated in the dark or in cell suspension senescence (less than twofold)
Locus identifierAnnotation (putative function)Sen/MGNahG/WTDark/lightCell suspension/con
  1. The value 0 indicates undetectable expression in any sample of that experiment.

At3g28580AAA-type ATPase9.780.0800.40
At3g28540AAA-type ATPase5.170.150.070
At3g28510AAA-type ATPase19.330.0200
At3g13100ABC transporter3.880.1600
At3g47480Calcium-binding EF hand4.970.071.440.26
At5g39670Calcium-binding EF hand3.180.131.620.22
At1g33960Avirulence-responsive (AIG1)3.570.0500
At2g33080Leucine-rich repeat family protein53.380.1600
At3g24954Leucine-rich repeat family protein3.430.1900
At3g24900Leucine-rich repeat family protein4.450.040.480
At4g28490Leucine-rich repeat family protein5.580.211.620
At4g13920Leucine-rich repeat family protein6.220.2300
At1g47890Leucine-rich repeat family protein12.170.2600
At1g74710Isochorismate synthase 1 (ICS1)3.300.101.330
At2g19190Receptor-like kinase,(SIRK)9.860.1800
At4g04500Protein kinase family protein8.300.0800
At4g23150Protein kinase family protein4.380.0400
At1g21240Wall-associated kinase, putative5.830.0600
At4g35580No apical meristem (NAM) family11.820.082.920
At2g26400Acireductone dioxygenase (ARD/ARD′)7.420.0200.26
At3g13620Amino acid permease family protein10.480.0500
At4g10500Oxidoreductase, 2OG-Fe(II) oxygenase5.930.042.620
At1g19250Flavin-containing monooxygenase15.380.1000
At3g07600Heavy-metal-associated domain9.920.1200
At5g24550Glycosyl hydrolase family 1 protein6.380.0700
At5g11920Glycosyl hydrolase family 32 protein4.360.2100.65
At3g53150UDP-glucosyl transferase12.840.0400
At1g35230Arabinogalactan-protein (AGP5)
At1g14080Xyloglucan fucosyltransferase (FUT6)96.000.1700
Protease genes partially downregulated in the NahG background
At5g45890SAG12, cysteine proteinase203.070.37038.79
At1g44130Nucellin protein, putative5.700.352.450.23
At1g32960Subtilase family protein9.130.2900.28

Senescence phenotype of NahG transgenic plants

As shown above, a group of genes that show enhanced expression in developmental senescence but not in dark-induced senescence are dependent on the SA pathway for expression (Table 6). Therefore, the potential importance of this group of genes in senescence could be assessed by comparing the progress of the two types of senescence in NahG plants. The third and fourth leaves were sampled at different times of development (for the developmental senescence analysis) and after 0, 2 or 4 days in the dark for the dark-induced senescence analysis. Developmental senescence is delayed considerably in the NahG transgenic plants, both by assessment of chlorophyll levels and by measurement of photosynthetic activity Fv/Fm (Figure 4). However, there is no difference in the rates of change of these two parameters in the dark-induced senescence between the wild type and the NahG transgenic plants. This is clear evidence that the SA pathway has a very important specific role to play in developmental senescence.

Figure 4.

Comparison of senescence parameters in wild-type (Col) and NahG plants during age-dependent senescence (a, b) and dark-induced senescence (c, d).
Age-dependent senescence symptoms in the Col and NahG transgenic line. Chlorophyll content (a) and photochemical efficiency (Fv/Fm) of PSII (b) was examined every 4 days from 20 to 44 DAE.
Dark-induced senescence symptoms in the Col and nahG transgenic line. Chlorophyll content (c) and photochemical efficiency (Fv/Fm) of PSII (d) was examined every 2 days during dark incubation.
Fv/Fm, maximum quantum yield of PSII electron transport (maximum variable fluorescence/maximum yield of fluorescence). Error bars indicate SD. DAE, days after emergence.

Some of the genes that depend on the SA pathway for full expression in developmental senescence but are not expressed in dark-induced senescence are shown in Table 6. Within this list there are likely to be some key genes that are essential for the normal progression of senescence and functional analysis of these is an important next step to define the importance of each gene. The list includes several leucine-rich repeat genes similar to known disease resistance genes and many of these are known to depend on the SA pathway for expression (Durner et al., 1997). Also there are a number of other kinase genes and two transcription factors that may have a regulatory role. The senescence-specific SAG12 gene appears on this list, this gene has been shown previously to be partly dependent on the SA pathway for expression (Morris et al., 2000) but there is still some gene expression in the NahG plants so it is unlikely that the loss of this protein has a strong influence on the delayed senescence phenotype.


Three different types of senescence have been studied in this paper and differences in gene expression have been characterized. The similarities and differences that have been observed are summarized by the model in Figure 5. In all cases, the senescence results in dismantling of cellular constituents and degradation of macromolecules including proteins, lipids and nucleic acids. Thus it is likely that genes that we have found to be expressed in all three types of senescence are involved in these common processes. In addition, to extend viability in the senescing cells there are a number of stress pathways including those dependent on JA and ethylene that are expressed in all three senescence processes.

Figure 5.

Model illustrating the similar pathways and alternative pathways that operate in the three types of senescence.
Genes that are expressed in the different types of senescence are illustrated in coloured boxes, green for developmental senescence only, blue for the starvation-induced senescence that is caused by both dark treatment and in the ageing suspension cultures and orange for the genes that are expressed in all three types of senescence.
Developmental senescence only is controlled at least partially by reduction in cytokinin levels and results in the differential expression of genes for flavonoid biosynthesis and glutamine metabolism. In addition, the SA pathway is only active in developmental senescence resulting in the expression of a number of putative pathogen-related genes and kinases.
Starvation-induced senescence, which probably results from sugar depletion, induces the expression of genes involved in asparagine metabolism, fatty acid degradation and a putative alternative flavonoid pathway. ABA response genes are expressed in all three types of senescence implicating this hormone in senescence regulation. However, an absolute requirement for this signal for any genes in senescence has not been determined, hence the dotted line. Also, certain genes that depend on JA and ethylene for expression are expressed in all three types of senescence, including a number of hydrolases and carbohydrate metabolism enzymes. A number of other groups of genes including those required for fatty acid α-oxidation are expressed in all three types of senescence. Some but not all of these are dependent on JA and/or ethylene signalling.

There are also considerable differences in processes that take place in the three types of senescence. In developmental senescence, photosynthesis continues, although at a reduced rate, and this presumably provides some energy for the process to take place. However, leaves undergoing senescence in the light have to contend with oxidative stress resulting from the degradation products of chlorophyll and other macromolecules. Genes related to flavonoid synthesis show increased expression in developmental senescence only although an alternative pathway of unknown function may be involved in dark-induced senescence. In dark-induced and cell suspension senescence the main signal for the process could be a rapid reduction in sugar levels, leading to a more significant switch to lipid degradation to supply an energy source. The differences in the C:N balance or the source of carbon skeletons in the light grown tissues compared with the other two treatments may be the reason why these cells appear to synthesis glutamine instead of asparagine for export. In the senescing suspension culture cells there are also clear indications of PCD occurring and this is likely to require a set of genes not expressed in the senescing leaves, at least not at the time they were harvested.

Signalling pathways that are induced during senescence have a downstream effect on transcription and genes that depend on the three stress response pathways, involving SA, JA and ethylene, for senescence-enhanced expression, have been identified. The SA pathway is not expressed in dark-induced or cell suspension senescence while the JA and ethylene signals are clearly active in all three types of senescence. Developmental senescence is delayed in plants defective in SA signalling but dark-induced senescence progresses normally in these plants. This shows that one or more of the genes that depend on the SA pathway for expression during senescence has an important role in the developmental senescence process.

Experimental procedures

Plant material

Developmental senescence and mutant analysis.  Arabidopsis plants (wild types Col-0, Col-5, transgenic NahG, and mutants ein2 and coi1) were grown at 22°C under 12 h day/12 h night conditions. The two wild-type green leaf samples Col-0 and Col-5 were harvested at Boyes stage 3.9 (Boyes et al., 2001). Three to four fully expanded leaves were harvested from approximately 15 individual plants. For the senescent samples, wild-type Col-0 and Col-5 and mutants in genes in the ethylene pathway (ein2) and the jasmonate pathway (coi1) and the NahG transgenic plant which is defective in the SA pathway were grown until the mid flowering stage (Boyes stage 6.0–6.9; Boyes et al., 2001). Three to four partially senescent leaves, showing 0–10% yellowing were harvested from approximately 15 plants at this stage. The leaves from the NahG plants were harvested when they were at the same senescent stage as the wild type and other mutants. Because of the delayed senescence in the NahG plants, the harvested leaves were likely to be older than the wild-type leaves. RNA was isolated by the method of Logemann et al. (1987) and purified using RNAeasy columns (Qiagen, Valencia, CA, USA). Two independent biological replicates were used for each mutant. For the AtGenExpress experiments leaf 6 or 8 were used from 17-day-old plants for the green leaf samples and leaves 6–8 were used from 35-day-old plants for the senescent leaf sample.

Dark-induced senescence.  The methodology and data used in this paper were originally described in Lin and Wu (2004). Arabidopsis thaliana ecotype Col-0 was grown at 22°C under short-day conditions (8-h L/16-h D) with 100 μmol m−2 sec−1 white light. For dark treatments, Arabidopsis plants at stage 1.10 (10 true leaves; approximately 40–50 days after germination) were placed in the same growth chamber (22°C) without light (light switched off at dusk of the day prior to treatment). The plant materials for control (0 day), 1 day and 5 days were harvested at the days indicated. At each time point, 12 plants were harvested for RNA isolation and subsequent microarray hybridization.

Arabidopsis total RNA samples were isolated as described previously (Chang et al., 1993). The population of mRNA was then isolated from total RNA with use of the Oligotex mRNA kit (Qiagen). Affymetrix ATH1 Genome Array hybridization and data acquisition were performed as described previously (Lin and Wu, 2004).

Suspension cultures.  Suspension culture cells were grown, harvested and RNA isolated as described in Swidzinski et al. (2002).

Microarray analysis

All the array experiments were carried out using the ATH1 Arabidopsis GeneChip microarray ( containing 22 810 probe sets. Two independent biological replicates were performed for all samples except the dark control sample which was carried out on one slide only. The array experiments for the developmental senescence and the mutant analysis (VBW) and the suspension culture cell death analysis (JS and CL) were carried out at Nottingham Arabidopsis Stock Centre (NASC) using Affymetrix recommended protocols for cDNA synthesis, array hybridization, chip scanning and data accumulation using the Affymetrix Microarray Suite version 5.0. The ATGenExpress data were from experiments carried out by the European consortium coordinated by Lutz Nover (Frankfurt), Thomas Altmann (Potsdam) and Detlef Weigel (Tübingen) ( These data were from triplicate slides for each time point. The raw data files for the NASC experiments and the AtGenExpress data are publicly available and can be downloaded from NASC (

Raw data from each experiment were scaled to an average of 100. Data were normalized in GeneSpring (Silicon Genetics, Redwood City, CA, USA) using the default one colour normalization procedure (each measurement was divided by the 50th percentile of all measurements in that sample, each gene was divided by the median of its measurements in all samples). GeneSpring was used to filter out unwanted genes and to identify up- and downregulated genes for further analysis. In each experiment, genes that were flagged absent in all samples and any genes that had a raw value of <50 on all slides were filtered out and not analysed further.

Gene lists to identify genes up- or downregulated in each experiment were generated as follows.

Developmental senescence.  The four slides at NASC are: Buchanan-Wollaston_A-1-bwoll-C0G, Buchanan-Wollaston_A-2-bwoll-C5G_SLD, Buchanan-Wollaston_A-3-bwoll-C0S_SLD, and Buchanan-Wollaston_A-4-bwoll-C5S_SLD, and the nine slides from ATGE are: ATGE_14 _A, ATGE_14_B, ATGE_14_C. ATGE_15_ A, ATGE_15_B, ATGE_15_C, ATGE_25_ A, ATGE_25_B, and ATGE_25_C. To determine genes showing up- or downregulation in senescence the two experiments were analysed separately. For the NASC experiment, the COG and C5G were used in GeneSpring as replicates for green leaf and the COS and C5S data as replicates for the senescent leaf. This helped to ensure that any gene expression differences obtained were general in these two variants of the Columbia accession. Normalized data for each gene were used to generate the senescent/green ratio which was used to identify those genes showing at least a twofold change in expression. For the AtGenExpress data two experiments were carried out in GeneSpring with either the slide 14 data (leaf 6) or the slide 15 data (leaf 8) compared with the senescent leaf data (slide 25). The normalized senescent/green ratios for these were averaged and any genes showing a twofold increase in expression was generated. Genes showing a twofold increase in expression were listed from the NASC and from the ATGE experiment and then genes common to these two lists were identified using a Venn diagram analysis in GeneSpring. The average ratio for the two experiments were calculated for each gene and 827 genes showing an average threefold increase in the two experiments were selected. This gene list is shown in Table S1.

Mutant analysis.  Two slides were generated for senescing leaves from each mutant. NahG (Buchanan-Wollaston_A-5-bwoll-NG1_S, Buchanan-Wollaston_A-6-bwoll-NG2_SLD), ein2 (Buchanan-Wollaston_A-7-bwoll-Ei1_SLD, Buchanan-Wollaston_A-8-bwoll-Ei2_SLD), and coiI1 (Buchanan-Wollaston_A-9-bwoll-Co1_SLD, Buchanan-Wollaston_A-10-bwoll-Co2_SLD).

For the mutant analysis, genes were filtered as described above and genes showing at least a twofold difference in expression in senescent leaves of the mutant compared with the wild type were selected. Venn diagram analysis in GeneSpring was then used to identify the senescence-enhanced genes that were altered in expression in the different mutant backgrounds. The ratio data for the senescence-enhanced genes are shown in Table S1.

Dark-induced senescence.  Affymetrix GeneChip ATH1 slides were used as described in Lin and Wu (2004). Two slides for the dark treatment and one slide for the control were analysed in GeneSpring. Genes showing an absent or very low signal were filtered out as described above. Two comparisons, each using one of the dark treatment slides and the same control slide were used and ratio data were averaged to find genes showing an expression change between the treatments. These data were used in Tables 2–6 and Table S1 to calculate dark-related expression changes for genes of interest.

Cell death.  Two replicate slides were analysed for each treatment, control and senescent (cell death) (NASC data: Swidzinski Control ATH1 Replicate 1, Swidzinski Control ATH1 Replicate 2, Swidzinski Senescence ATH1 Replicate 1, Swidzinski Senescence ATH1 Replicate 2). Genes that were flagged absent in all samples and any genes that had a raw value of <50 on all slides were filtered out of the analysis. Normalized data were used to calculate a senescent/control ratio for each gene and genes showing a threefold upregulation in the senescing cells were identified.

Measurement of senescence parameters in NahG plants.  Plants were grown in an environmentally controlled growth room at 22°C with a 16-h light/8-h dark cycle with moderate light intensity (150 μm m−2 sec−1). Chlorophyll content and photochemical efficiency were examined at several developmental ages of leaves in planta (for age-dependent experiment) or at the given times after incubating detached leaves in darkness (for dark-induced leaf senescence experiment), using the third and fourth leaves of wild type (Col) and nahG transgenic plants. Chlorophyll was extracted from individual leaves by heating the leaves in 95% ethanol at 80°C. Chlorophyll concentration per fresh weight of leaf was calculated as described by Lichtenthaler (1987). The photochemical efficiency of photosystem II (PSII) was deduced from the characteristics of chlorophyll fluorescence (Oh et al., 1996) using a portable plant efficiency analyzer (Hansatech Instruments, Norfolk, UK). The ratio of maximum variable fluorescence to maximum yield of fluorescence, which corresponds to the potential quantum yield of the photochemical reactions of PSII, was used as the measure of the photochemical efficiency of PSII (Oh et al., 1996).


We thank the Nottingham Arabidopsis Stock Centre (NASC) for carrying out the Affymetrix array experiments and the BBSRC for providing funds to subsidize the cost of these experiments under the Investigating Gene Function Initiative. We acknowledge the use of microarray data produced by the AtGenExpress project, which is coordinated by Lutz Nover (Frankfurt), Thomas Altmann (Potsdam) and Detlef Weigel (Tübingen), and supported by funds from the DFG and the Max Planck Society. The leaf data were generated by Jan Lohmann and Markus Schmid (MPI Tübingen). We thank Prof. John Turner from University of East Anglia for the gift of Coi1 mutant seed. VBW, EH, TP, EB and CJL thank the BBSRC for their financial support. We thank Prof. Brian Thomas for his comments on the manuscript.

Supplementary Material

The following material is available from

Figure S1.mapman diagrams showing differential expression patterns of metabolic pathway genes in developmental and dark-induced senescence. Ratio data were converted to a log base 2 scale and imported into mapman. Blue boxes represent individual genes that are upregulated in senescence. Intensity of the colour indicates the relative level expression.
(a) Genes that are upregulated in senescence and not in dark-induced senescence.
(b) Genes that are upregulated in dark-induced senescence and not in developmental senescence.
The genes that are discussed and compared in the paper are ringed in red.

Table S1 Genes showing at least threefold upregulation during leaf senescence. Genes flagged as present with a raw value over 50 in all senescent experiments. Raw values were scaled to average 100 for each slide. Green – NASC experiment green leaf; sen – NASC experiment senescent leaf; L6 and L8 – ATGE experiment green leaf; SEN – ATGE experiment senescent leaf; Ratio – senescent/green; DARK – ratio of DARK 5 days/control; CELL DEATH – ratio of starved cell suspension culture/control. Genes have been assigned to groups based only on their TIGR v. 5.0 annotation. This grouping does not indicate confirmed function for most of the genes. Data for NahG, Coi1 and Ein2 indicate the ratio of expression in senescing leaves of mutant/senescing wild type