Eukaryotes such as plants, animals and yeast have all evolved ways of cellular suicide that are known as programmed cell death (PCD). In multicellular organisms such as plants and animals, the organized destruction of cells is important for forming body plans and specific organ shapes and for removing unwanted, damaged or infected cells1, 2, 3, 4. Two well-characterized model systems for the study of plant PCD are those of the HYPERSENSITIVE RESPONSE (HR), which is often observed during plant–microbe interactions, and the development of TRACHEARY ELEMENTS in the XYLEM of vascular plants5, 6 (Fig. 1). A simplified depiction of the sequence of cytological events that take place during these forms of plant PCD is contrasted with those of apoptosis, the well-studied form of animal PCD (Fig. 2). The 'clean' process of apoptosis effectively contains the contents of the dead cell for removal by other cells and avoids activating an inflammatory response in animals. In the case of HR-associated cell death and the terminal differentiation of tracheary elements, the contents of the dying cells are not engulfed by other cells. In addition, the 'corpse' of the dead cell is held in place by the cell wall, and, for mature tracheary elements, the cell wall is reinforced during the early phase of PCD and carries out the essential function of mechanical support and transport after autolysis5, 6. These key characteristics therefore distinguish plant PCD from that of classic apoptosis and indicate that specialized features and pathways have probably evolved to control and execute the death programme in plant cells (Box 1).
Figure 1 | Examples of programmed cell death in plants.
a | An example of the hypersensitive response: a tobacco leaf is infiltrated with two different strains of Pseudomonas syringae. Zones 1 and 2 are infiltrated with strain NPS3121 and zones 3 and 4 with strain NPS4000. The latter strain is isogenic to NPS3121 except for a mutation in the HRP locus, which renders it incapable of inducing the hypersensitive response in tobacco. Visible cell death of the inoculated regions can be observed in zones 1 and 2 one day post-inoculation, whereas no significant morphological changes were observed in zones 3 and 4. Enlarged views with back lighting show cleared cells with little chlorophyll remaining in zones 1 and 2, in contrast to zones 3 and 4. b | An example of developmental death that has a critical role in the housekeeping function of mechanical support and long-distance transport is the formation of the xylem in vascular plants. The top panel shows a cross-section of a tobacco leaf at the mid-rib region, with the boxed region showing the central files of tracheary elements that are critical for transport of solutes and water. These cells have already undergone programmed cell death and the remaining corpses have reinforced secondary cell walls that are highly autofluorescent (shown by arrowheads in enlarged views). Top enlarged view: transmission light microscopy image. Bottom enlarged view: epifluorescence of the same viewing field with excitation filter at 436 nm and emission filter at 480 nm.
Figure 2 | Morphological comparison between programmed cell death in plants and animal apoptosis.
a | In the hypersensitive response, chromatin condensation and DNA cleavage into 50-kb fragments were observed before the apparent disruption of the vacuole, which takes places during the late stages of cell death18. Blebbing of the vacuole and plasma membranes, and late destruction of organelles were also observed. At the final stage of this cell-death process, the plasma membrane collapses and separates from the cell wall18, 19, ending with the leakage of the dead cell's content into the apoplast Fragmented nuclear DNA is shown throughout the figure as irregular, brown masses in the nuclei that are undergoing cell death. b | During the differentiation of tracheary elements, vacuole swelling and rupture is coordinated with the thickening and restructuring of the cell wall. The final collapse of the vacuole immediately precedes nuclear DNA fragmentation, which occurs at the late stages of the cell-death process before the final autolysis of the cell. Short stubbles on differentiating tracheary elements indicate reticulated secondary cell walls. Broken areas in the cell wall of terminally differentiated tracheary elements indicate spatially localized perforations. c | Apoptosis in animal cells initiates morphologically with chromatin condensation and fragmentation. Plasma-membrane ruffling is followed by the formation of apoptotic bodies from the repackaging of the cell content and their final engulfment by neighbouring cells or macrophages.
In addition to HR and tracheary element differentiation, other models of plant PCD such as those occurring during senescence, embryogenesis and ozone treatment have also been established and systematically characterized (see Refs 3,4,7 for more comprehensive reviews of the various forms of plant PCD). Recent genetic and molecular approaches have yielded many cellular components that, when mutated, can perturb plant-cell-death control under various conditions. The recent report of cell-free systems from Zinnia elegans and Arabidopsis thaliana that can be used to study nuclear changes during PCD could provide a much-needed assay for the biochemical characterization of key players in this process8, 9.
This article focuses on integrating information on the orchestration of cell death in plant cells, with emphasis on new results obtained from analyses of gene functions using mutants or transgene expression. This effort is especially timely given that the A. thaliana model system is now readily amenable to both forward- and reverse-genetic approaches, and that cell-biological markers for various subcellular compartments are well established for plants. So, the functional relevance of a given gene and the cytological events during cell death in plants can now be characterized with unprecedented ease.
Cytology of controlled cell death in plants
The cytological events that accompany cell-death activation have been documented comprehensively for several plant-PCD model systems. In PCD that takes place during the transdifferentiation of MESOPHYLL cells into mature xylem cells in Z. elegans cell cultures6, vacuoles accumulate degradative enzymes and swell, and the cell wall is remodelled into a highly reticulated form. Nuclear and organelle DNA fragmentation closely follows the collapse of the vacuole during the autolytic phase at the end of this process10, 11. Chromatin condensation and other typical apoptotic morphologies are not seen (Fig. 2). This type of vacuole-directed-death process has also been documented during other plant developmental processes such as senescence and AERENCHYMA formation in the root12.
In a series of recent studies, the cytology of embryogenesis-related cell deaths were characterized using in vitro cell cultures of fertilized ovules from Norway spruce (Picea abies). Chromatin fragmentation into clear nucleosomal fragments was observed without obvious chromatin condensation or nuclear BLEBBING, whereas vacuolation of the cytoplasm occurred earlier. Vacuole rupture occurs very late in the autolysis of the dying cells when most cellular compartments are already dismantled. Two separate waves of cell-death events were defined for the transition from pro-embryogenic mass to a somatic embryo13, with the latter wave mirroring those events observed during the developmental cell death that results in the selection of a single surviving embryo within the polyembryonic GYMNOSPERM ovule14, 15. Rapid cytoskeletal changes are observed in the early phase of embryogenesis and these could be important in orchestrating the relocation of essential cellular components that are involved in this developmental cell-death process16.
More-classic apoptosis-like features were observed for the HR, with nucleosomal laddering and apparent cellular fragmentation akin to apoptotic bodies reported in some cases of HR cell death17 but not in others5, 18, 19 (Fig. 2). Rapid reorganization of the cytoskeleton of plant cells has been observed at the infection site by fungal pathogens, and pharmacological studies have indicated that this might be important in disease resistance and HR cell death during an incompatible interaction20. However, a recent study indicates that the dramatic changes in the cytoskeleton might be necessary but they are insufficient to account for HR cell-death activation21. New imaging tools22 should provide a more detailed and dynamic description of plant-cell-death morphologies under various scenarios.
Signalling for plant cell death
The ubiquitous nature of PCD highlights the idea that this might be a default pathway for a cell in the absence of the proper survival signal23. In plants, this idea was supported by the finding that PCD of carrot cells cultured at low density could be reversed by putative intercellular factor(s) that are present in 'conditioned media' that was obtained from cultures at higher densities24. So, numerous signals are probably constantly integrated by the cell to decide whether to activate the cell-death programme or not. In addition, the control of the boundaries and kinetics of the cell-death event would be crucial for the optimal deployment of this process. For example, in senescence, PCD affects entire organs, occurs over a relatively long period and is linked with the recycling of fixed nitrogen. By contrast, for HR cell death, a rapid but restricted zone of cell death is used to effectively contain pathogens at their site of entry, and efficient recycling of the contents of the dead cell might be secondary.
The isolation and characterization of LESION-MIMIC mutants showed that cell death can be activated cell-autonomously and that survival signals are required to restrict lesion size25. A relatively large set of genes that, when mutated, result in the lesion-mimic phenotype has now been isolated, predominantly from A. thaliana26 (Table 1). In addition, some transgenes that perturb the physiology of the cell can also induce lesion-mimic phenotypes27. Although these findings are aiding our appreciation of cell-death control in plants, separating primary signals for a cell-death pathway from those that might affect the physiology of the cell, and thereby potentiate the observed death phenomenon as a secondary consequence, is not trivial. This is especially problematic when core regulators for cell-death pathways are not defined for plants as yet, and so a mechanistic description of how a signal ultimately connects to a particular cell-death pathway would be difficult at best.
Phytohormones as PCD-signal regulators
Salicylic acid and nitric oxide. Different phytohormones are involved in modulating HR cell death under various conditions. Salicylic acid is a well-known mediator of SYSTEMIC ACQUIRED RESISTANCE (SAR) in plants28. Aside from this key role, the involvement of salicylic acid in cell death has been tested by crossing A. thaliana plants that are engineered to suppress salicylic-acid accumulation with various lesion-mimic mutants. Interestingly, although the spontaneous cell-death phenotype could be abolished in some of these mutants, others showed only a quantitative reduction or no effect at all when salicylic acid is removed (reviewed in Ref. 26). So, salicylic acid might promote cell death in some signalling pathways but have only a minor role in others. Studies of the lesion-mimic mutants agd2 (Ref. 29) and acd6 (Ref. 30) also showed a possible involvement of the salicylic-acid-dependent pathway(s) on HR-cell-death suppression as well as cell growth. One current view of salicylic-acid functions is that, at high concentrations, such as those that might be produced by plant cells at the site of pathogen entry, salicylic acid probably functions as a cell-death promoter in collaboration with other signals. As salicylic acid, and possibly other systemic signals, is transported outward from these initial sites of synthesis, salicylic acid at low concentrations becomes a survival signal instead and helps to establish the border of spreading lesions28. Added to this complexity is the likely existence of feedback controls through the action of reactive oxygen species (ROS) on salicylic-acid synthesis and the salicylic-acid-dependent generation of ROS31.
Nitric oxide (NO) cooperates with salicylic acid to induce HR cell death and activate defence, which is analogous to its role in animal systems32. Increased NO production is sufficient to induce cell death in an A. thaliana cell culture33. It has been proposed that the balance between intracellular NO and hydrogen peroxide (H2O2), but not superoxide, concentrations is the key determinant for the HR cell-death response34. The recent identification of two plant nitric-oxide-synthase genes revealed that they share little, if any, structural similarities with their animal counterparts35, 36. This indicates that there are new modes of NO synthesis and that possible convergent evolution occurred in plant and animal systems to generate similar signalling molecules using structurally distinct enzymes. Mutants of these genes should provide useful tools to facilitate our understanding of the involvement of NO in cell-death processes from HR to senescence in plants.
Jasmonic acid and ethylene. In addition to salicylic acid and NO, two other phytohormones, jasmonic acid and ethylene, regulate cell death under stress conditions and during development. Jasmonic acid, a phytohormone that is produced after wounding, negatively regulates cell death in A. thaliana under oxidative stress by ozone treatment37, but it might work as a positive factor to promote cell death that is induced by a fungal toxin fumonisin B1 (Ref. 38). Ethylene is involved in promoting senescence39 as well as other forms of developmental cell death, such as the formation of aerenchyma in hypoxic roots40 and ENDOSPERM cell death in cereals41. Ethylene also promotes cell-death activation by ROS as well as lesion formation in the lesion-mimic mutant acd5 (Ref. 42) and in cell-death induction by the toxin fumonisin B1 (Ref. 38).
Brassinosteroid, gibberellic acid and abscisic acid. Three phytohormones that are involved in plant growth and development are also known to control cell-death signalling in a context-dependent manner. In the tracheary-element differentiation system of Z. elegans, brassinosteroid is produced by Z. elegans cells to initiate the final stage of tracheary-element differentiation, which involves the autolytic death programme6, 43. Two other plant hormones have opposing effects on PCD in the ALEURONE cell layers of developing barley endosperm. Gibberellic acid promotes cell death in this context in cooperation with ROS, whereas abscisic acid counteracts this death-promoting effect of gibberellic acid. The mechanism of this interplay between the two hormones could relate to their opposing effects on the expression of ROS-scavenging enzymes44. The role of abscisic acid in repressing cell death has also been reported for developing barley anthers45, so this type of hormonal potentiation of cell death might not be restricted to aleurone cells.
Membrane channels and lipid metabolism
Rapid ionic fluxes are among the earliest documented responses to HR that are induced by avirulent pathogens. Plasma-membrane-localized receptors, as well as enzymes that are involved in ROS metabolism, are also important in the perception and propagation of signals for cell growth and death in both animals and plants. It is therefore no surprise that several plant-cell-death signalling proteins associate with membrane channels or have lipid-associated functions.
Membrane-associated signals. The observation that transgenic expression of a bacterial proton channel, bacterioopsin (bO), in tobacco (Nicotiana tabacum) resulted in a lesion-mimic phenotype indicated that perturbation of the pH homeostasis in plant cells might activate HR-like cell death and the associated defence response46. Point mutants that abolish the proton-channelling activity of bO failed to induce lesion-mimic phenotypes. Both bO and the various point mutants accumulate to similar levels in transgenic tobacco plants and probably localize to the plasma-membrane fraction47. Another recent study points to an anion transporter that is involved in nitrate efflux as an essential component of ELICITOR signalling for HR cell death in tobacco48. In addition, two genes that encode members of the CYCLIC-NUCLEOTIDE-GATED CHANNEL (CNGC) family in A. thaliana have been identified as regulators of HR cell death. CNGC2, a cation channel that conducts calcium, is encoded by the DND1 (defence, no death-1) locus49. The dnd1 mutant suppresses HR cell death when challenged with certain avirulent pathogens without altering disease resistance. However, this mutant also shows micro-lesions, indicating a low level of constitutive activation of HR cell death. Recently, a lesion-mimic mutant, hlm1, which shows constitutive HR-like cell death and pathogen resistance, was shown to encode CNGC4, a channel protein that can mediate transport of K+ and Na+, and that is activated by both cyclic GMP and cyclic AMP50. These results support the view that the alteration of ionic homeostasis is a key step in the early signalling process for HR cell death.
Lipid-related signals. Intracellular signals that are derived from lipid metabolism are also being recognized as mediators of HR induction and associated cell death. Two mutants that are defective in pathogen-response signalling, pad4 and eds1, revealed genes that encode proteins that are related to triacylglycerol lipases51, 52. Interestingly, these genes are also required for the spontaneous phenotypes of some of the lesion-mimic mutants such as lsd1, cpr1 and cpr6 (reviewed in Ref. 26; Table 1). This provided genetic evidence that phospholipid signalling is involved in the induction of HR cell death. Phosphatidic acid — which can be produced by the action of phospholipase D (PLD) on the precursor phosphatidylcholine, or from phosphorylation of diacylglycerol by a kinase — increases after induction of the plant-cell-defence response53. One particular isoform of PLD in A. thaliana, PLD
, has been genetically shown to attenuate H2O2-mediated activation of cell death that might mediate many biotic and abiotic responses54. This indicates that the phosphatidic-acid production, which was mediated by PLD in plant tissues that were challenged with pathogens, might serve as a negative signal of PCD for the management of cell-death propagation during the development of the HR. PLD is apparently stimulated by oleic acid55, and deficiency in oleic acid in the ssi2/fab2 mutant (Table 1) leads to a lesion-mimic phenotype, possibly due to a decrease in phosphatidic-acid production from lower levels of PLD activity56. These results support the importance of the phosphatidic-acid-lipid-derived signal as a negative regulator of certain cell-death pathways, in particular those that are mediated through oxidative stress.
Recent suppressor analyses of the ssi2/fab2 mutant resulted in the identification of the SFD4/FAB6 gene (which encodes a plastidic 
6-desaturase) and other SFD loci, which function as modulators of the ssi2 phenotypes57. However, the differential effects of these suppressors on ssi2 plants indicate a complex interplay between lipids containing polyunsaturated fatty acid and the various signalling pathways that mediate defence and wound responses. Finally, a loss-of-function mutation in the A. thaliana gene ACD11, which encodes a sphingosine-transfer protein, also results in a lesion-mimic phenotype that is salicylic acid and light dependent, in addition to a requirement for PAD4 and EDS1 (which are involved in salicyclic-acid signalling; Ref. 58). The facilitation of sphingosine transfer between membranes might also potentiate cell-death activation and its perturbation could trigger cell death in plants.
NADPH oxidases and ROS are cell-death signals
The involvement of ROS in the activation of PCD is well known in animals, plants and yeast5, 7. The generation of an 'oxidative burst' during the early and late phases of plant–pathogen interaction has been well documented59, and more-recent observations support a role for ROS in cell-death signalling60, 61, 62. In animal systems, a membrane-associated NADPH oxidase in complex with the small G-protein RAC is responsible for ROS production63. Recently, mutations in two A. thaliana genes, rbohD and rbohF, that encode orthologues of the mammalian GP91phox NADPH oxidase catalytic subunit were analysed. Loss-of-function mutants in A. thaliana rbohD and rbohF resulted in decreased ROS production during the HR, with a corresponding decrease in cell death64. These results provided the first direct genetic evidence for ROS generation by plant NADPH oxidases and their roles in a cell-death response. Paradoxically, although A. thaliana rbohD seems to be more prominent in the generation of total detected ROS, A. thaliana rbohF is more important for the activation of HR cell death. The intracellular location of ROS synthesis might be an important factor that determines the efficacy of cell-death induction by this signal. Consistent with the role of plant orthologues of GP91phox in ROS generation and HR induction, studies with rice-derived RAC have provided evidence that this small G-protein functions in the control of HR cell-death induction and the associated resistance to fungal and bacterial pathogens65. However, a recent study that reported gene suppression of RACB in barley indicated that this small G-protein is required for successful fungal infection in a genotype-specific manner66. Perhaps, the differential accumulation of superoxide, instead of H2O2, could be an important determinant of resistance and possibly HR cell death?
Organelles and cell death in plants
As the energy status of the cell can potentiate cell death in animal and yeast systems, it is not surprising that the mitochondria and plastids in plants are also important signal generators for cell-death regulation. In addition to their abilities to generate ROS through electron-transfer intermediates that are intimately involved in functions associated with these organelles, a number of other cell-death signalling pathways also seem to depend on organelle components. Several well-characterized cell-death mediators exit the mitochondria to promote apoptotic cell death in animal systems. Cytochrome c (the release of which activates procaspase-9, Ref. 67), apoptosis-inducing factor (AIF; which functions as a CASPASE-independent activator of nuclear DNA cleavage68), and the recently described endonuclease G (Ref. 69) are some of these animal proteins that might have counterparts in plants.
In plants, although cytochrome-c translocation to the cytoplasm has been reported (reviewed in Ref. 5), its possible function as a cell-death signal — rather than it being a consequence of mitochondrial destruction — remains to be shown. Recent studies indicate that breakdown of the permeability of mitochondria might correlate with cell-death induction in some plant systems70, 71, 72, but these pharmacological studies require more molecular and genetic studies to further elucidate the mechanisms involved. A Mg2+-dependent nuclease was recently identified in the intermembrane space of the mitochondria using a cell-free nuclear system that recapitulates chromatin condensation and DNA fragmentation, and it might be biochemically similar to the recently described mitochondrial endonuclease G (Ref. 9). Increased phosphorylation of a mitochondrial chaperone, prohibitin, has also been reported in the rice lesion-mimic mutant cdr1 (Ref. 73). This correlates with the activation of NADPH oxidase in this mutant and might be part of the cellular responses that lead to the cell-death phenotype.
The plastid-localized enzymes that are involved in porphyrin metabolism, which is important for haem and chlorophyll biosynthesis, are known to be involved in spontaneous cell-death activation in plant lesion mimics. These include the lin2 and acd2 mutants in A. thaliana and several lesion-mimic mutants in Zea mays26, 74. The effects of these mutations are not well defined, although it is likely that photo-oxidative damage mediated by accumulated toxic intermediates might be a common cause. The level of a plastid homologue of the bacterial protease FtsH, which is involved in clearing damaged proteins, negatively correlates with the induction of HR cell death. This tobacco gene, DS9, seems to positively affect photosynthetic-electron flow and its overexpression leads to an apparent delay in HR-cell-death activation75. How these plastid-derived signals are all integrated together for cell-death signalling remains to be determined.
New plant-specific PCD regulators
Two plant genes that are emerging as interesting signalling mediators for HR-cell-death activation are A. thaliana LSD1 and barley MLO. Both are conserved across monocots and dicots but seem to be specific to plants.
LSD1. lsd1 was originally identified as a lesion mimic of the propagation class and it exhibits a so-called 'runaway cell death' (rcd) phenotype — that is, lsd1 plants cannot restrict cell-death progression after an initial stimulus76. It specifically responds to O2-, but not to H2O2, to activate HR cell death, and recent genetic studies show that it is probably an important negative regulator downstream of the triacylglycerol lipases EDS1 and PAD4, as mutants of these loci suppressed the lsd1 cell-death phenotype26. In addition, salicylic acid is also required for the rcd phenotype, which provides strong genetic evidence for LSD1 as a key repressor of cell-death progression during plant defence77. A highly conserved relative of LSD1, known as LOL1 (LSD1-like-1), was recently characterized in A. thaliana. Both proteins contain a set of conserved zinc-finger motifs and are putative transcription factors or scaffold proteins78. Unlike LSD1, which is predicted to negatively regulate cell death, LOL1 is thought to be a positive regulator of cell death. Suppression of LOL1 expression in an lsd1 background suppressed rcd, whereas increased expression of LOL1 accelerated HR cell death in a wild-type background. It has been suggested that LSD1 and LOL1 might represent plant-cell-death potentiators that function in an antagonistic fashion to establish a threshold for cell-death activation. Although the roles of these proteins in HR cell death is clearly established, an interesting question is whether they are specific for pathogen signalling or whether they could be involved in controlling other forms of PCD in plants. It would also be interesting to further clarify the possible in vivo association between these two proteins, their biochemical activities and target(s), as well as the role of the third member, LOL2, of this emerging family of plant-cell-death regulators.
MLO. The MLO gene family was originally identified in barley, with a loss-of-function mutant showing a spontaneous lesion-mimic phenotype of the initiation class — that is, it spontaneously activates PCD in the absence of any obvious inductive signals. MLO, as well as its related proteins, is predicted to contain multiple transmembrane helices and is located in the plasma membrane. Loss of MLO leads to enhanced disease resistance in barley, and it has also been shown that the spontaneous cell death of the mesophyll cells in mlo mutants might be the result of partly accelerated senescence79. Furthermore, MLO seems to be transcriptionally activated under a variety of biotic and abiotic stresses and this might correlate with its role as a negative potentiator of cell-death activation. A clue to the function of MLO came from recent studies showing that CALMODULIN interacts specifically with the carboxy-terminal portion of MLO proteins80. Point mutations that abolish this interaction result in a quantitative decrease in the ability of MLO to suppress resistance and this can be phenocopied by suppression of a barley calmodulin isoform by RNA interference81. These interesting observations provide evidence that MLO function could be modulated by Ca2+ concentrations and that it mediates signalling downstream of calcium channels to suppress cell death and resistance.
The elusive executioners or core regulators
Determinants of the 'point-of-no-return.' The first irreversible step in the activation of the cell-death programme would require high precision by the cell. Proteases such as metazoan caspases that function as regulatory switches can provide such a key function. They are usually expressed in a dormant form and subsequently activated, or they are sequestered in a subcellular compartment to protect the cell from misfiring the essential death trigger.
The characterization of a putative protease switch with a clearly defined set of subsequent targets and effectors in plant PCD has been relatively slow. For tracheary-element differentiation, multiple types of protease are thought to be important at different stages. Proteasome-inhibitor studies indicate that ubiquitin-mediated protein turnover is essential during the commitment phase, but not the autolytic phase, of this process6, 82. A potential role for extracellular serine proteases has also been postulated83. Early work with ROS-induced cell death has indicated the importance of serine- and cysteine-protease activities19. Subsequent studies identified cystatin, a cysteine-protease inhibitor, as a possible inhibitor of ROS- and bacteria-induced HR cell death84. On the other hand, caspase-specific peptide inhibitors specifically abolished HR cell death in tobacco leaves85. Caspase-like protease (CLP) activity was also detected in extracts that were prepared from tissues undergoing HR cell death. Notably, cell death was uncoupled from the defence-gene response by peptide inhibitors, which indicated that the loss of cell-death induction was not due to inhibition at the early step of host–pathogen signalling. Although the use of pharmacological reagents has the usual caveat of specificity and the identity of the targets is still uncertain, these early results, nevertheless, might signify that a protease or proteolytic cascade involving multiple proteases and their regulation by endogenous inhibitors could function as a critical switch in plant cell death.
Since the publication of the complete A. thaliana genome, it has become clear that canonical caspases with high structural similarities to the well-studied metazoan homologues are unlikely to be found in plants. Nevertheless, proteases with activities akin to caspases have been implicated in plant-cell-death activation using a variety of experimental conditions and model systems4, 5, 85. In addition to peptide inhibitors, biological inhibitors for metazoan caspases have also been reported to suppress plant cell death. The baculovirus proteins Op-IAP (inhibitor of apoptosis protein) and p35 have been shown to interact with caspases in vitro and to suppress apoptosis activation in animal systems86. Transgenic expression of Op-IAP in tobacco suppressed cell death that was induced by necrotrophic fungal pathogens, and expression of the p35 protein in tobacco delayed HR cell death in several different systems85, 87, 88, 89, 90. Mutants of p35 that cannot inhibit animal caspases are also unable to suppress cell death in transgenic plants88, 89, 90 — providing important support for the specificity of the phenomena that are observed as the result of CLP inhibition.
Identifying elusive CLPs. What might be the identity of plant CLPs? Iterative searches based on the structural features of caspases have identified two related families of proteases, known as PARACASPASES and METACASPASES91. These families of cysteine proteases do not contain any obvious variants of the QXCRG active-site motif that is found in all canonical caspases, but their predicted structures are likely to have strong homologies to the caspase–haemoglobinase fold. Metacaspases are found in fungi, protozoa and plants, including nine predicted metacaspase-encoding genes in the A. thaliana genome. Three of these A. thaliana genes belong to the type-I class, containing a prodomain that has interesting homologies to the LSD1 protein, whereas the other six type-II metacaspases contain a conserved insertion in the carboxy-terminal half of the predicted protein sequence.
Recently, the single yeast type-I metacaspase, Yca1, has been shown to mediate oxidative stress and ageing-related cell death. A yca1 strain shows dramatically longer life in culture in addition to high tolerance to H2O2 (Ref. 92). The wild-type Yca1 protein — but not one with a point mutation at the cysteine residue predicted to be the active site — was shown to be proteolytically activated upon oxidative stress, which correlated with the appearance of new CLP activities. These results therefore establish the first genetic and biochemical evidence that metacaspases could be candidates for the role of executioner protease in plants, fungi and protozoa. Indeed, expression of a Trypanosoma brucei metacaspase, but not a variant that was mutated at its predicted active-site cysteine, caused cell death in yeast93. Recent findings showed the increased expression of a type-II metacaspase during necrotroph-induced, but not chemical-induced, cell death of tomato (Lycopersicon esculentum) leaves, perhaps linking this class of proteases to the induction of plant cell death94. Comprehensive analysis of the metacaspase gene family, using reverse-genetic and molecular approaches, should allow us to determine their biochemical characteristics and in vivo relevance to plant cell-death control in the near future.
BI-1-related cell-death regulators. A large class of conserved core regulators that are related to the BCL-2 protein are important in potentiating death-inducing signals in animals7, 95. Although plant and yeast genomes lack any obvious orthologues of BCL-2-related proteins at the primary sequence level, expression of metazoan pro-apoptotic BAX activated the cell-death programme, whereas expression of the pro-survival BCL-2 or BCL-XL provided protection in these organisms4, 5, 7. An induced-lethality suppression screen was carried out in yeast to identify new mammalian proteins that could counter the pro-death activity of BAX96. Two proteins — BAX inhibitor-1 (BI-1) and bifunctional apoptosis regulator (BAR) — were identified, and BI-1 homologues were found as conserved genes in plants. Plant BI-1 homologues, as well as a number of proteins that are involved in ROS homeostasis such as iron-superoxide dismutase, were also found using this yeast lethality-suppression screen with A. thaliana and tomato cDNA expression libraries97. In A. thaliana, the expression of BI-1 was induced during plant–pathogen interaction, and overexpression of A. thaliana BI-1 — as well as the rice orthologue of BI-1 in animal cells, transgenic tobacco and A. thaliana — has been shown to suppress BAX-induced cell death97, 98, 99, 100. Suppression of tobacco BI-1 expression correlated with increased AUTOPHAGY, nucleosomal DNA fragmentation and cell death after carbon starvation and hypo-osmotic shock101. Another study found that decreased BI-1 expression correlated with chemical-induced resistance of barley to powdery mildew fungal infection. Furthermore, overexpression of barley BI-1 at a single-cell level induced hyper-susceptibility and could reverse the fungal resistance that is conferred by the loss of MLO, a negative regulator of resistance and HR-like cell death102. BI-1 could therefore function as a cellular survival factor that promotes fungal infection, and its downregulation would correlate with heightened fungal resistance, as is the case for mlo mutants of barley79, 81. Consistent with this conclusion, transgenic rice that overexpressed rice BI-1 showed improved cell survival upon challenge with a fungal elicitor from Magnaporthe grisea103.
Although A. thaliana BI-1 was reported to localize to the endoplasmic reticulum of plant cells99, the mode of action for this protein remains enigmatic in animal, plant and yeast systems. All known BI-1-related proteins contain 5–7 predicted transmembrane 
-helices in their deduced sequence, which are conserved in the two other A. thaliana BI-1-related genes, BI-2 and BI-3. Database searches using the amino-terminal region of A. thaliana BI-2 uncovered 13 genes, which are known as A. thaliana BI-2-related (ABR) genes5. No obvious homologues to ABRs were found in animal or yeast genomes, which indicates that these genes might have evolved specifically in plants. This new family of proteins has been speculated to function in an analogous fashion to Bcl-2-related proteins — perhaps pro-survival and pro-death members could regulate cell death under different contexts. Consistent with this is the presence of six predicted transmembrane helices in all of the ABR members, similar to BI-1. Interestingly, expression of A. thaliana BI-1 in a human fibrosarcoma HT1080 cell line induced apoptosis that could be repressed by co-expression of the caspase inhibitor XIAP104. This finding showed that A. thaliana BI-1, and perhaps its related proteins, could function as either pro-survival or pro-death factors in a context-dependent manner. Further detailed studies on the activity and function of BI-1 and its related proteins in plants should help establish whether they are true components that modulate cell-death triggers of PCD.
Cell-death regulators from plant pathogens. The genomes of viral pathogens provide a good hunting ground for potential regulators that could repress cell-death switches in their host because of their obligate BIOTROPHIC lifestyle. The baculovirus is an excellent example in which two main classes of potent cell-death regulator — IAPs and p35 — were discovered, and cellular-IAP orthologues were subsequently identified in animals as important regulators of cell death86. It is therefore interesting that the study of plant viruses has yet to uncover cell-death repressors that would shed light on the central control machinery of the host. By contrast, recent studies of TYPE-III EFFECTORS from Gram-negative bacterial pathogens have provided the first candidates that could enrich our understanding of cell-death control in plants (Box 2).
The vacuole as a source of executioners. Alan Jones recently hypothesized12 that the loading of distinct sets of hydrolytic enzymes into the vacuole, as induced by endogenous or extracellular cell-death-triggering signals, could define the various morphotypes of plant PCD. In this scenario, calcium flux was envisioned to have a critical role in mediating the trigger of the eventual vacuole collapse by the direct or indirect action of a common executioner downstream from the various death signals. It is interesting to compare the mechanism of plant PCD and that of autophagic cell death in animal systems. Developmental PCD in animal systems is now known to occur commonly through non-apoptotic pathways95. One recently well-characterized system is that of hormone-induced PCD of the salivary gland in Drosophila melanogaster where typical cellular morphologies of autophagic cell death, instead of apoptosis, were observed. Nevertheless, regulatory genes that were related to the core executioners, such as a D. melanogaster caspase, were found to mediate this type of cell death105. So, overlapping sets of regulators could be used to activate diverse cell-death morphotypes with the choice of cell-death modes being defined by the particular cellular context as well as the identity of the trigger. The finding that Beclin-1, a mammalian orthologue of a yeast protein (Apg6) that is involved in autophagy, binds BCL-2 and inhibits virus-induced apoptosis as well as tumorigenesis106 also indicates a possible connection between autophagy, the vacuole/lysosome and apoptosis. It would be interesting to determine whether the autophagy protein Beclin-1/Apg6 could be involved in modulating apoptosis via interaction with BCL2-related proteins, or conversely, whether BCL2-related proteins might also function to modulate autophagic cell death.
In the analysis of cell-death morphologies during somatic embryogenesis in gymnosperms, Filonova et al.13 made the interesting observation that they resemble a combination of apoptotic and autophagic modes of animal PCD. It is therefore intriguing to raise the question as to whether key components that are involved in autophagy, which is normally used for recycling of cellular components via the vacuole/lysosome during nutrient starvation, could be involved in plant PCD (Box 3). Further studies of cell-death-related processes with mutants in the autophagy pathway107, 108, as well as mutants in vacuole biogenesis such as vacuoleless-1 (Ref. 109), should shed light on the possible connections between these phenomena that are related to the controlled turnover of the content within plant cells.
Conclusions
The distinct purposes of controlled cell death, coupled with the diverse cell types in which cell death must be carried out, probably contribute to the diverse PCD morphologies that are observed in plants. During senescence, which is a slower process that coordinates cell death at the organ and tissue level, efficient recycling of nutrients from the dying cells and tissues is the primary purpose, whereas a more rapid cell death is required during the HR to efficiently prevent viral pathogens from systemically infecting the host plant5, 6, 12, 88. A specialized cell-death pathway might therefore have evolved in plants for each distinct purpose and different sets of executioners might be activated. Alternatively, varied contributions from a relatively small set of distinct cell-death pathways could provide the means for generating the spectrum of cell-death morphotypes that have been reported. The latter 'combinatorial' model is consistent with the observation that when apoptotic cell death is inhibited during digit formation in Xenopus laevis, a slower necrotic cell death is observed110. The partial redundancy that is inherent to this type of combinatorial model might also explain the paucity of 'plant cell executioners' that have been genetically identified so far, compared with the many components of various signalling pathways that function to modulate the death decision25, 26. This is in contrast to C. elegans and D. melanogaster, in which a set of core regulators for various death morphotypes has been described genetically2, 7, 95.
In conclusion, elucidation of the pathway for controlling vacuole rupture, the identification of the different CLPs that have been suggested to be important for plant-cell-death pathways, and the functional characterization of BI-1-related genes should help to shed light on these possible models. To this end, the completed A. thaliana genome coupled with a large collection of sequenced insertions are invaluable resources that should help mine the genome for components in the cell-death pathways of plant cells. Efficient, inducible gene-suppressing systems111 should further enhance our ability to test for gene functions under different developmental conditions. Ultimately, the knowledge that is gained should be relevant to applications such as those related to plant-disease resistance and wood production. Moreover, comparisons with animal cell-death pathways also shed light on the evolution of this process in eukaryotes.
