Heat treatment- and senescence-induced programmed cell death in an Arabidopsis suspension culture
In order to undertake a microarray analysis to distinguish changes in specific gene expression that are related to PCD from those that are specific to a particular PCD-inducing treatment, we investigated the effectiveness and timing of various treatments for induction of cell death in Arabidopsis suspension cultures. The two most effective and reproducible treatments were found to be during natural senescence of the ageing cultures and following a heat treatment at 55°C for 10 min. PCD in individual cells of the culture was scored by the absence of fluorescence with the vital stain FDA and also by identifying the morphological characteristics of PCD (Figure 1a,b). PCD morphology is recognised by the condensation of the cytoplasm and nucleus and the shrinkage of the plasma membrane away from the cell wall (McCabe et al., 1997). Conversely, healthy, living cells stain positively with FDA and are typically highly vacuolated (Figure 1c), while cells which have undergone a necrotic, or non-programmed form of cell death do not stain with FDA and fail to display the striking cytoplasmic condensation that occurs in PCD cells (Figure 1d).
Figure 1. Cellular morphology of Arabidopsis suspension culture cells during cell death.
Cellular morphology was examined by phase-contrast microscopy (400× magnification) in the following cells:
(a) 6-day-old cultures incubated for 10 min at 55°C and examined immediately thereafter.
(b) Untreated 13-day-old cultures.
(c) Untreated 6-day-old cultures.
(d) 6-day-old cultures incubated for 10 min at 80°C and examined immediately thereafter.
PCD morphology is evident as a condensed cytoplasm and shrinkage of the plasma membrane away from the cell wall (a and b). CC, condensed cytoplasm; CW, cell wall; PM, plasma membrane; LC, living cell; NC, necrotic cell. Bar is 10 µm for a-d.
Download figure to PowerPoint
As originally demonstrated by Callard et al. (1996), PCD is observed in Arabidopsis cultures when they are allowed to senesce in the absence of the usual subculturing at day 7. They demonstrated that many whole-plant senescence-specific genes could be used as markers of this response, indicating that ageing cell cultures are an appropriate system to study developmental senescence. Depending on the original culture conditions and overall state of the cultures, the time required for 100% of the cells to undergo PCD in the cultures used in this study ranges from 13 to 17 days (data not shown). The primary limiting factor in senescing cultures appears to be sucrose availability (data not shown). Senescing cell cultures consistently exhibit the hallmark features of PCD, including morphological alterations and detectable DNA laddering when at least two-thirds of the cells in culture have undergone PCD (Figure 2b). For microarray analysis, RNA was isolated from cells harvested from 13- to 14-day-old cultures when 40% of cells had undergone cell death as judged by vital staining and morphology (Figure 2a). At this time, in spite of the fact that total DNA extracts from the cultures will fail to exhibit DNA laddering (Figure 2b), the remaining cells are clearly induced to undergo PCD or are in the early stages of PCD.
Figure 2. Time course of senescence-induced PCD in Arabidopsis suspension culture cells.
(a) Cell cultures were examined at a fixed time each day beginning at 6 days after subculturing. The percentage of cells exhibiting PCD in the culture was scored on the basis of a condensed cytoplasmic morphology and the absence of FDA staining by phase-contrast microscopy. Samples used for microarray analysis were collected at days 13–14 when approximately 40% of the cells had exhibited PCD.
(b) DNA laddering during culture senescence. DNA was extracted at various timepoints during culture senescence as the total percentage of PCD increased in the culture. 10 µg of genomic DNA was fractionated by gel electrophoresis on a 1.5% (w/v) agarose gel and stained with ethidium bromide. The percentage of cells exhibiting PCD morphology at the time of extraction is indicated above each lane. Laddering is visibly evident in cultures in which over two-thirds of the cells have undergone PCD following senescence.
Download figure to PowerPoint
A brief heat treatment of cell suspension cultures has previously been demonstrated to induce a PCD response in the cells as characterised by the typical biochemical and morphological markers (McCabe and Leaver, 2000; McCabe et al., 1997). In this paper, we show that heat treatment of Arabidopsis suspension cultures for 10 min at 55°C in a shaking water bath is sufficient to induce PCD in an immediate and synchronous fashion, with 100% of the cells dying within 12 h of the heat treatment (Figure 3a). Morphological features of PCD are visible in the cultures immediately following heat treatment, and DNA laddering in total culture extracts of heat-treated cells becomes detectable in the cultures when approximately two-thirds of the cells have undergone PCD (Figure 3b). For microarray analysis, RNA was isolated from cultures immediately following the 10 min heat treatment (time 0). At this time-point, as in the senescence samples, DNA laddering in the cultures was not detectable (Figure 3b) but the majority of cells were induced to undergo or were in early stages of PCD, therefore making this a suitable time for analysing changes in gene expression. Approximately 20% of the cells had actually undergone PCD at this time.
Figure 3. Time course of heat-induced PCD in Arabidopsis suspension culture cells.
6-day-old cell cultures were induced to undergo PCD by treatment for 10 min at 55°C.
(a) The percentage of cells exhibiting PCD in the culture was scored on the basis of a condensed cytoplasmic morphology and the absence of FDA staining by phase-contrast microscopy. Cell samples used for extraction of RNA for microarray analysis were collected at time 0 h, immediately following heat treatment, when just under 20% of the cells exhibited PCD.
(b) DNA laddering following heat treatment. DNA was extracted at various timepoints following a 55°C, 10 min heat treatment as the total percentage of PCD increased in the culture (indicated above each lane). Control DNA before heat treatment is in lane 7. 10 µg of genomic DNA was fractionated by gel electrophoresis on a 1.5% (w/v) agarose gel and stained with ethidium bromide. The percentage of cells exhibiting PCD morphology at the time of extraction is indicated above each lane. Laddering is visibly evident in cultures in which over two-thirds of the cells have undergone PCD following heat treatment.
Download figure to PowerPoint
Importantly, in both the senescing and heat-treated cultures, samples are amenable to molecular analysis as indicated by the integrity of total RNA extracts (Figure 4). One hour following heat treatment proved to be unsuitable for array analysis since, due to the rapidity of the response, over 40% of the cells have undergone PCD (Figure 3a) and total RNA was already significantly degraded as judged by gel electrophoresis (data not shown). In spite of the rapid response of the cultures to the heat treatment, the response can clearly be identified as programmed cell death and not disorganised necrotic cell death. Cells treated at 80°C for 10 min undergo a necrotic cell death response in which 100% of the cells are dead immediately following treatment and the typical PCD morphology is clearly absent in over 95% of the cells (Figure 1d). Additionally, since RNA extracted immediately following this heat shock is completely degraded (Figure 4) and PCD is defined by its requirement for active gene expression (Jones, 2001), only the milder treatment at 55°C allows for a PCD pathway to be invoked. It is noteworthy that the DNA laddering occurring in senescent cells is more apparent than in heat-treated cells, which may be due to the rapidity of the response in the latter cells that results in faster and perhaps less precise DNA breakdown (i.e. a mixture of oligonucleosome cleavage and general DNA breakdown).
Figure 4. RNA integrity following induction of PCD in Arabidopsis suspension culture cells. RNA was extracted from control 6-day-old cultures (lane 1), 6-day-old cultures immediately following a 55°C, 10 min heat treatment (time 0 h in Figure 3a) (lane 2), or senescing, 13–14 day-old cultures (13–14 days in Figure 2a) (lane 3) for use as a template in reverse transcription and microarray analysis. In the heat and senescence treatments, which induce PCD, RNA integrity is fully maintained during PCD induction at the timepoints analysed. Conversely, 6 day-old cultures treated at 80°C for 10 min show immediate RNA breakdown following the heat shock (lane 4), indicative of necrotic cell death.
Download figure to PowerPoint
Selection of genes for microarray analysis
The purpose of this study was to examine changes in expression profiles for genes which have previously been implicated in PCD-related responses, such as senescence and the hypersensitive response (HR), and also those genes which, based on our knowledge of animal PCD and the central role of mitochondria in this process, we hypothesised may also play a role during plant PCD. A large number of the genes selected were those previously identified as senescence-activated or senescence-related genes, as programmed cell death marks the endpoint of senescence and typical markers of PCD, including DNA laddering, occur during the senescence programme (Delorme et al., 2000; Gan and Amasino, 1997; Quirino et al., 2000; Yen and Yang, 1998). Numerous studies have identified genes that are up-regulated during senescence—a time when the majority of cellular transcripts are down-regulated—although no conserved senescence regulatory elements have been discovered. Moreover, because there appears to be multiple regulatory pathways functioning during senescence, the question of whether many of these up-regulated transcripts are components of senescence-specific processes or may play a more general role in plant PCD has become of central importance in plant PCD studies (Pontier et al., 1999; Quirino et al., 2000). The evidence thus far indicates that there are certainly some common elements in the pathways leading to plant cell death, but to date no consistent markers or regulatory genes involved in plant PCD have been identified.
The array used in this study included several previously identified senescence-activated genes to determine if the expression of these genes was also up-regulated during the induction of PCD using an unrelated treatment (heat). Obviously it was anticipated that expression of these genes would be induced during culture senescence but a change in the expression of the same genes during heat-induced PCD would suggest a common underlying molecular mechanism for PCD in the two situations. Senescence-related genes selected for array analysis included those encoding the senescence activated gene SEN1 (Oh et al., 1996), the Arabidopsis homologue to tomato SENU3 (Drake et al., 1996), SRG1, SRG2 and SRG3 which were originally identified in senescing cell cultures (Callard et al., 1996), an SPF1-like protein (Kim et al., 1997), cysteine proteases (Solomon et al., 1999), and several other ESTs identified only as ‘senescence-activated genes’. Additionally, senescing cucumber cotyledons are currently being used in our laboratory to identify genes up-regulated at the onset of PCD (Delorme et al., 2000; Kim et al., 1997). A number of genes have been identified in this system, including transcription factors, DNA binding proteins, a matrix metalloproteinase (Delorme et al., 2000), and several other genes of unknown function. Arabidopsis homologues of these genes were selected for the array where possible.
Genes identified as being up-regulated during the hypersensitive response (HR) were also included and, while the typical apoptotic morphology has not been consistently recognised during HR cell death, several studies have reported DNA laddering during this process and suggested that some aspects of HR cell death may be shared in other plant PCD pathways (Heath, 2000; Pontier et al., 1999; Quirino et al., 2000). HR-related genes selected for array analysis included those identified in the Arabidopsis HR-related studies of Lacomme and Roby (1999) and Sanchez et al. (2000). The former study identified several Athsr genes up-regulated during Arabidopsis HR, including genes encoding the alternative oxidase, the mitochondrial voltage-dependent anion channel (VDAC), and several unknown proteins. In the studies of Sanchez et al. (2000), two putative Arabidopsis Bax-inhibitor homologues were identified as being up-regulated during wounding and pathogen challenge.
As well as investigating genes that are associated with particular types of PCD such as senescence and HR, we assayed the expression of genes associated with metabolic events that are thought to generate signals that form part of the molecular mechanism of PCD in plants. These include genes associated with antioxidant metabolism such as superoxide dismutases, glutathione and ascorbate peroxidases, and genes involved in Ca2+ signalling, including a number of calmodulin and calcium-dependent protein kinases. Oxidative stress can trigger PCD in plant cells and, in some cases, it has been shown that reactive oxygen or nitrogen species serve as signals to induce changes in gene expression associated with PCD (Clarke et al., 2000; Desikan et al., 1998, 2000; Dorey et al., 1999; Levine et al., 1994; Rao et al., 2000). Additionally, the production of ROS may also occur as a consequence of the activation of a death pathway and the suppression of antioxidant enzyme activity may actually be necessary to allow progression of PCD (Fath et al., 2001; Mittler et al., 1999). Interestingly, in a screen to identify plant suppressors of Bax-induced cell death in yeast, a glutathione-S-transferase/peroxidase was identified, again linking the production of ROS to cell death (Kampranis et al., 2000). Similarly, changes in intracellular Ca2+ have also been demonstrated to be a common event during plant PCD (Levine et al., 1996; Sanders et al., 1999).
Finally, we investigated the role of genes encoding mitochondrial proteins. In mammalian systems mitochondria have been shown to play a central role in the execution of the death pathway, but to date the evidence for a similar involvement of mitochondria in plant PCD is largely circumstantial. We arrayed a number of markers of mitochondrial function in an attempt to address the issue of involvement of mitochondria in plant PCD. These included VDAC, alternative oxidase, ANT1 and ANT2, MSD1, HSP60, and two subunits of the ATP Synthase complex, the mitochondrially encoded Atp1 and nuclear-encoded Atp2. In addition to these genes of interest, several enzymes and 16 Arabidopsis‘housekeeping’ genes (Reymond et al., 2000) were arrayed as markers of general cellular activity, together with a range of negative controls.
Microarray hybridisation experiments and expression analysis
The custom-designed microarrays used in this analysis were screened with cDNA synthesised from RNA samples derived from each of the PCD-inducing treatments, senescence and heat. Total RNA was isolated from senescing cells or from heat-treated cells as described earlier. cDNA was synthesised in a reverse transcription reaction during which the fluorescent dyes, Cy3 or Cy5, were incorporated as α-dCTP precursors into the control or test (PCD) samples, respectively. For each treatment, arrays were replicated three times, including one reverse labelling reaction in which control samples were Cy5 labelled and test samples Cy3 labelled. Ratios of < 1 were transformed to – 1/ratio and the results from the 3 arrays were then averaged.
In the final data analysis, only genes for which the expression varied more than 1.5-fold when averaged across 3 array results for each treatment were taken to represent significant changes in gene expression. This threshold ratio is lower than that used in some previous array studies (Reymond et al., 2000; Schenk et al., 2000), but has been demonstrated to represent significant changes in differential expression in other microarray analyses (Desikan et al., 2001; Kawasaki et al., 2001; Perez-Amador et al., 2001). Moreover, during PCD the majority of cellular transcripts decrease in abundance, and total RNA yield may decrease up to 10-fold (Buchanan-Wollaston, 1994; Quirino et al., 2000). Equal amounts of total RNA were used for reverse transcription for both control and PCD cells but, due to the decreased RNA yields per cell in PCD-treated cultures, a gene for which the steady-state expression level remains unchanged on a cellular basis during PCD will appear to increase up to 10-fold in copy number relative to other transcripts. It is important to note that such a gene may not actually be up-regulated during PCD, but it may not be degraded at the same rate as other transcripts (Buchanan-Wollaston, 1994) and could be indicative of an important function of that gene product during PCD. While PCD is a particular case in which widespread mRNA degradation occurs, the expression profiles inferred from microrarray data may always be influenced by the possibility that total mRNA yields per cell are not equal across control and treated cells, and researchers working with such systems should account for this possibility in their interpretation of array results.
Our arrays demonstrate a high correlation coefficient between replicates (Table 1); even the lowest R2 value obtained, 0.876, is still well within the range of significance (P < 0.05). Statistically, to detect differential expression of genes included in the microarrays at greater than or equal to 1.5-fold, the log-2 expression ratio of 1.5 (0.58) should fall at least two standard errors away from zero (no change in log-2 expression) when considering the 3 replicate arrays. It can therefore be determined that the log-2 standard deviation for each gene, averaged across 3 replicate arrays, must be less than or equal to 0.50. In fact, in 172 ratios considered in the final analysis, 93% of the means calculated had a sd ≤ 0.50. Taken together with the correlation coefficients, this data indicates that our results give excellent reproducibility and, using 3 replicate arrays for both heat and senescence, we conclude that differential expression can be detected at ± 1.5-fold.
Table 1. Correlation coefficients of ratios for each microarray replicate from control Arabidopsis cultures versus heat- or senescence-induced cultures undergoing PCD
|Replicate||C-Cy5 H-Cy3 (1)||H-Cy5, C-Cy3||C-Cy5, S-Cy3 (1)||S-Cy5, C-Cy3|
|C-Cy5 S-Cy3 (2)||–||–||0.949||0.969|
To further support our threshold of 1.5-fold, competitive quantitative RT–PCR was carried out on 4 transcripts. The method of RT–PCR employed involved the reverse transcription of RNA samples derived from either control or PCD-induced cells using unique oligo(dT) primers for each. Both primers had identical PCR primer binding sites and oligo(dT) domains, but one primer contained an extra linker of 40 nucleotides. In this manner, each cDNA in one sample was synthesised with an additional 40 bp at its 3′ end. This allowed both populations of cDNA from control, and either heat- or senescence-induced PCD samples, to be mixed together and a single RT–PCR reaction to be carried out. The cDNA from each sample therefore competed for the amplification reagents, and the PCR product derived from each sample was distinguished by a 40-bp difference. The quantity of each band was then measured, thus giving a more accurate indication of the relative transcript abundance in each RNA sample and avoiding comparisons across different PCR reactions.
For our RT–PCR analysis, we deliberately analysed transcripts which represented ratios between ± 1.5 and 2.0 in one or both of the treatments. To normalise the RT–PCR data, each reaction was carried out in parallel with an RT–PCR reaction using primers for β-tubulin. Ideally, one would wish to normalise RT–PCR against a gene whose expression level is known to remain constant (1.0 ratio) but, because we have been unable to identify such a gene during PCD (see below), normalisation against the β-tubulin transcript was undertaken for each gene. In a similar manner, the microarray ratio for each gene analysed was normalised against the microarray ratio obtained for β-tubulin. This allowed a direct comparison between the β-tubulin-normalised RT–PCR ratio and the β-tubulin-normalised microarray ratio for each transcript investigated. In this manner, we tested whether the ratio obtained by microarray analysis could be replicated by an independent method and, especially, whether the change in expression for genes whose ratios were between ± 1.5–2.0 could be confirmed by RT–PCR analysis. Figure 5 indicates a good correlation between microarray ratios and those calculated by RT–PCR against a β-tubulin normalisation, confirming the quality of our array data and justifying the choice of a 1.5-fold threshold. It is important to stress, however, that the microarray expression ratios should not be interpreted as exact numbers for fold-induction or -repression of a gene; previous studies have failed to demonstrate a direct linear relationship between transcript abundance and fluorescence intensity (Ruan et al., 1998). Nevertheless, the data is sufficiently reproducible to indicate general changes in gene expression patterns during PCD.
Figure 5. Competitive RT–PCR analysis of selected gene transcripts.
(a) Competitive RT–PCR was undertaken on 4 transcripts to confirm the changes in steady state levels obtained by microarray analysis. In each lane the higher band (PCD) represents the RT–PCR product from the PCD sample (heat or senescence), while the lower band (control) represents the RT–PCR product from the control sample. The intensity of each band was quantified by Multi-Analyst (Bio-Rad) and used to calculate an expression ratio, heat : control or senescence : control. The ratios obtained were normalised to a β-tubulin competitive RT–PCR reaction run in parallel. RT–PCR reactions were carried out using primers for β-tubulin, Athsr5, KNAT2, SRG3 and Athsr2/VDAC.
(b) Expression ratios obtained by microarray analysis for each of the selected transcripts was normalised against the microarray expression value for β-tubulin. Ratios obtained by microarray analysis are compared to those obtained by competitive RT–PCR analysis.
Download figure to PowerPoint
Transcript analysis during heat- and senescence-induced PCD
Analysis of the microarray data revealed significant changes in the steady state levels of the transcripts of approximately 50 of the selected genes in at least one of the treatments (Figure 6). Of these, approximately 25% were significantly repressed, while the remainder, many of which are known senescence- or stress-related genes, were induced. It is noteworthy that over half of the genes on the array were differentially expressed between control and PCD-inducing treatments, thus exceeding the percentage of total array genes found to be differentially expressed in previous studies (Aharoni et al., 2000; Schenk et al., 2000). However, the genes in this study were selected specifically for the reason that they may be involved in the execution of PCD and therefore differential steady-state expression levels of a high proportion of genes was anticipated.
Figure 6. Evaluation of microarray data.
Expression ratios of genes following senescence- (a) or heat-induced (b) PCD determined by microarray analysis. The expression ratios (PCD/Control) are transformed on a logarithmic (base 2) scale. The top series represents genes up-regulated during PCD, and the bottom series genes down-regulated during PCD. In each analysis, the bars indicate the cut-off for genes for which the change in expression was not considered to be significant. Genes exceeding these thresholds were considered for further analysis and are highlighted in Table 2.
Download figure to PowerPoint
Table 2 highlights the genes with significant changes in expression profiles grouped by putative function. The first group contains so-called ‘housekeeping genes’, often used in studies as control genes whose expression profiles remain constant across treatments. It is evident that, in this study at least, no such ‘housekeeping gene’ exists, and even β-tubulin expression decreased almost 2-fold in senescing cultures. This was further supported by Northern Blot Analysis of the β-tubulin transcript abundance (data not shown). This underscores the need for caution when using a particular gene as a marker of consistent expression levels, especially when using such ‘housekeeping genes’ as normalization factors in an array.
Table 2. Expression profiles of genes following PCD-inducing treatments in Arabidopsis thaliana suspension cell cultures
The second group of genes contains those with known mitochondrial-associated functions. The key event of release of cytochrome c and other apoptotic proteins from the mitochondrial intermembrane space may not necessarily be transcriptionally regulated since it is thought to involve either a complexing of existing proteins in the inner and outer mitochondrial membranes to form a permeability transition pore, and/or the recruitment of pro-apoptotic proteins such as Bax to the outer mitochondrial membrane. However, several genes encoding mitochondrial membrane and matrix proteins are differentially expressed during mammalian cell death. During lymphoma cell death, for example, uncoupling proteins and VDAC are transcriptionally up-regulated (Voehringer et al., 2000), while a microarray analysis of neuronal cell death demonstrated that transcripts encoding mitochondrial proteins including the adenine nucleotide transporter, cytochrome b-560, cytochrome c oxidase, and several other oxidoreductases are modulated during the cell death programme (Chiang et al., 2001). In this study, we found no evidence that genes encoding mitochondrial proteins are generally up-regulated during plant PCD. Interestingly, however, one of the two genes encoding the adenine nucleotide transporter (ANT) exhibited a marked decrease in expression in both treatments. A decrease in ANT protein synthesis will ultimately reduce export of mitochondrial ATP and inhibit ATP synthesis. In neuronal cell death ANT expression was also decreased during the early phase of cell death (Chiang et al., 2001) and, in an Arabidopsis microarray analysis of defense responses, Schenk et al. (2000) demonstrated the repression of a gene encoding ANT following treatment with salicylic acid, methyl jasmonate, and the inoculation of an incompatible fungal pathogen. Collectively, these results suggest that a temporary disruption of the cell's energy supply may be a general feature of early cell death in both plant and animal cells, and may be one factor that triggers the later phases of the molecular machinery of cell death. Indeed, in animal cells it has been demonstrated that a disruption of ATP/ADP exchange between the mitochondrion and the cytosol is one of the earliest events following induction of apoptosis (Vander Heiden et al., 1999). Moreover, this decrease in ATP/ADP exchange, which may be due to a defect in the function of inner membrane ANT, outer membrane VDAC, or both, has been proposed to be one of the chief initiators of downstream apoptotic events, including mitochondrial swelling, cytochrome c release, and the opening of the permeability transition pore. Therefore, it is tempting to suggest that, in plant cells, a similar decrease in ATP/ADP exchange may be one of the triggers of cell death, and may be related to a down-regulation of ANT gene expression as demonstrated by our array data.
The steady state levels of the transcripts encoding uncoupling proteins 1 and 2 were increased during heat-induced PCD but not during senescence. While it has previously been shown in animals that there is specific up-regulation of only one member of the UCP family following induction of apoptosis (Voehringer et al., 2000), our data do not suggest a general role for either UCP1 or UCP2 in plant PCD. It is possible that these proteins play a role in limiting reactive oxygen species formation following heat treatment (Kowaltowski et al., 1998). The differential regulation of mitochondrially encoded Atp-1 and nuclear-encoded Atp-2, both encoding subunits of the mitochondrial ATP synthase, indicates that transcripts of the former may be protected or maintained at constant steady-state levels during PCD due to their localisation within the mitochondria, since only Atp-2 transcript levels decreased during PCD. Indeed, proliferation of mitochondria has actually been observed prior to apoptotic death in human cells (Mancini et al., 1997). It is of interest to note that expression of VDAC (originally identified as Athsr2) was shown to be up-regulated during the HR of Arabidopsis cultures inoculated with pathogenic Xanthomonas (Lacomme and Roby, 1999) and yet did not change significantly in either of our PCD-inducing treatments. VDAC may therefore play a specific role during the HR that is not universally connected with PCD.
Oxidative stress clearly plays an important role in plant PCD and, undoubtedly, many genes induced during PCD in our study may be part of an oxidative stress response. However, whether the production of reactive oxygen species (ROS) in PCD is a consequence of PCD-inducing stimuli or is a necessary precursor to allow PCD to proceed remains under investigation (Fath et al., 2001; Mittler et al., 1999). Several antioxidant-related genes were found to be up-regulated during induction of PCD by both heat and senescence in our system. These genes included the superoxide dismutases, CSD1 and CSD3. The cytosolic CSD1 is up-regulated to a greater degree by heat and senescence treatments than the peroxisomal CSD3, which is in accordance with its role as a general stress-response enzyme (Kliebenstein et al., 1998). Cytosolic glutathione peroxidase and the cytosolic isoform of monodehydroascorbate reductase, an enzyme involved in the ascorbate-glutathione cycle that detoxifies reactive oxygen species (Noctor and Foyer, 1998), were induced in both heat and senescence treatments. These genes may therefore be potential markers for general oxidative stress and/or PCD.
However, several antioxidant genes examined were differentially regulated in heat and senescence treatments, suggesting that the type of oxidative stress generated by different treatments may require specific metabolic accommodation strategies. For example, expression of MSD1 encoding the mitochondrial superoxide dismutase seems to be modulated during heat-induced oxidative stress, but to a much lesser extent during senescence. This is supported by previous studies that have demonstrated the differential regulation of the seven Arabidopsis SODs during various environmental stresses, including high-light treatment, ozone fumigation, and UV-B exposure (Kliebenstein et al., 1998).
The expression profiles of several HR-related genes were also examined in this study. Interestingly, three of these genes were up-regulated in both heat and senescence, indicating that these processes may share common mechanisms and that identifying a gene as ‘HR-regulated’ does not exclude a role for its product in other types of PCD. Of the Athsr genes originally identified by Lacomme and Roby (1999), Athsr5 and Athsr6 are up-regulated in non-HR responses. The two putative Arabidopsis Bax-inhibitor genes, AtBI-1 and AtBI-2 (Sanchez et al., 2000), were screened using unique probes. The array analysis indicates that only AtBI-2 expression is consistently up-regulated during both heat- and senescence-induced PCD, thus implicating this gene in a more general stress- or PCD-related response.
All but one senescence-related gene on the array exhibited marked increases in expression in senescing cell cultures, suggesting that the PCD in the cell cultures was the endpoint of a senescence response and that cell cultures are a suitable system for studying senescence. Expression of the one gene which did not increase during senescence, SRG3, was previously shown to be up-regulated just prior to the onset of senescence (Callard et al., 1996), and therefore it is not surprising that it is down-regulated in our senescence sample. It is intriguing, however, that it is up-regulated following the induction of heat-induced PCD, suggesting that this gene may play a role in the early stages, or at the onset of plant cell death. One senescence-activated gene encoding a RING-H2 finger protein (SAG EST N37587) is also up-regulated in heat-treated cells, and it was found that cysteine proteases involved in senescence, including an Arabidopsis SENU3 homologue (Drake et al., 1996), are also up-regulated during heat-induced PCD and may be more general markers of plant PCD.
Several regulatory genes were up-regulated during both heat and senescence-induced PCD, indicating a role for these regulatory elements during PCD. These include the transcription factors EREBP-4 and STZ/Zat10 (Lippuner et al., 1996), as well as the Arabidopsis knotted2-like homeobox gene (KNAT2). In contrast, other members of the KNAT gene family, KNAT1, KNAT3, and KNAT4, were not similarly induced. While no downstream targets of KNAT2 have been identified (Lincoln et al., 1994; Ori et al., 1999), the results obtained here suggest that KNAT2 may play a role in plant PCD.
In this paper, we report on changes in the expression profiles of over 90 genes that have been reported to be involved in programmed cell death and/or senescence, or for which we hypothesised there may be a role in these processes. Very few genes have been identified as markers for plant PCD that function irrespective of the inducing stimulus. While many HR-related genes, senescence-related genes, or stress-related genes have been identified, it is unclear whether changes in expression of these genes are unique to a particular developmental or physiological event, or whether their role in PCD extends to a more universal mechanism underlying all forms of plant PCD. The aim of this study was to determine if genes with direct or indirect roles in plant PCD exhibited similar expression profiles when compared across two independent, PCD-inducing treatments, and whether any other novel, molecular markers of plant PCD could be identified. Since many genes selected for the array were originally identified as being up-regulated during HR or wounding responses in plants, we investigated whether these genes were similarly up-regulated in two other PCD-inducing treatments. Additionally, by identifying any putative cell death-related genes which were up-regulated at the transcriptional level following PCD induction, we were able to more clearly suggest a role for their gene products in plant PCD. Indeed, transcription is known to be necessary for the execution of PCD, and pro- and anti-apoptotic proteins have been shown to be regulated at the transcriptional level in animal cell systems undergoing PCD (Chiang et al., 2001).
Our results indicate that some overlap in differential gene expression between heat, HR, and senescence-induced PCD may exist and that, in particular, certain oxidative stress-related genes and cysteine proteases may serve as useful markers for plant PCD. Our results support the need for future investigations to utilise more than one system of PCD induction before generally ascribing a role for a gene product in plant cell death. As with all microarray studies, it must be noted that, in spite of the fact that many putative cell death genes do not appear to be transcriptionally regulated during PCD, this does not preclude a role for their gene products in this process. In the future, the use of multiple PCD-inducing systems, combined with large-scale investigations such as whole-genome microarrays and cellular and subcellular proteome analyses, will be necessary to both identify global regulators of plant cell death and to elucidate the crosstalk between pathways involved in plant PCD.