Regulation and execution of molecular disassembly and catabolism during senescence


Author for correspondence: John E. Thompson Tel: +1 519 8884465 Fax: +1 519 7460614 Email:

I. Introduction

Senescence occurs naturally at the end of the life span of an organ or, in the case of monocarpic senescence, the whole plant (Pic et al., 2002). It is induced by up-regulated expression of genes encoding proteins that are capable of implementing cell death. In as much as it is genetically regulated, the death of cells during senescence is classified as programmed cell death. Indeed, studies with Affymetrix GeneChips have revealed that > 1400 genes exhibit changes in expression during leaf senescence (Buchanan-Wollaston et al., 2003). One of the earliest physiological manifestations of senescence is membrane leakiness, in part the result of de-esterification of membrane lipids (Thompson et al., 1998). In fact, the breakdown of cellular ultrastructure inherent in programmed cell death is, to a large extent, driven by the catabolism of macromolecules. Not surprisingly, therefore, many of the proteins expressed in senescing tissues are hydrolytic enzymes, including proteases and lipases, that are capable of degrading macromolecules and setting in motion the dismantling of cellular ultrastructure (Guo et al., 2004). Recent evidence indicates that special vacuoles, designated senescence-associated vacuoles, enriched in digestive enzymes accumulate in senescing tissue (Otegui et al., 2005).

Programmed cell death is also an inherent feature of xylogenesis and the hypersensitive response to incompatible pathogens (Fukuda, 1996; Morel & Dangl, 1997; Buchanan-Wollaston et al., 2003). However, senescence occurs more slowly than, for example, the hypersensitive response induced during pathogenesis, which is acute and rapid (Jones, 2001). This is consistent with the fact that in senescing leaves programmed cell death is integrated with remobilization of carbon and nitrogen to developing seeds. For nutrient mobilization to be successful, the cells of the senescing tissue must remain viable, and hence it is only after this mobilization has been completed that total collapse of cell structure leading to death occurs (Matile et al., 1996). Noteworthy in this context is the growing realization that catabolites formed during senescence, which are detrimental to either the structural or physiological integrity of cells, are sequestered until late in the senescence cascade (Hanaoka et al., 2002; Kaup et al., 2002; Xiong et al., 2005).

This review addresses recent findings pertaining to the molecular disassembly of membranes in senescing tissues and how this is coordinated with molecular cascades that initiate and execute programmed cell death and nutrient mobilization.

II. Role of signaling cascades in the initiation of senescence

Senescence is often described as the terminal phase of plant development, be it of a tissue, an organ or the entire plant. However, it is important to temper this notion of the terminal nature of senescence with the realization that senescence of one part of a plant is often essential for the ongoing development of other organs and tissues. Perhaps the clearest example of this is leaf senescence and the attendant translocation of nutrients from the dying leaf tissue to developing seeds (Himelblau & Amasino, 2001; Kaup et al., 2002). Moreover, in nonmonocarpic species, the timing of senescence initiation for different tissues and organs appears to be developmentally regulated, and this has prompted use of the terminology ‘developmental senescence’ (Doorn & Woltering, 2005).

Senescence is induced by endogenous signals including age, developmental cues and plant growth regulators (Raghothama et al., 1991; Gan & Amasino, 1995; Grbic & Bleecker, 1995; Nooden & Penney, 2001; Riefler et al., 2006). However, it can also be engendered prematurely by a number of exogenous environmental stresses, including light and temperature stress, dehydration, nutrient stress and pathogen ingression (Beers & McDowell, 2001; Pic et al., 2002; Xiong et al., 2005). Of particular note in this context is the recent finding that autumnal senescence of aspen leaves appears to be triggered solely by photoperiod, with no effect of temperature (Keskitalo et al., 2005). Moreover, these endogenous and exogenous signals inducing senescence appear to be coordinated through a common signaling network (Fig. 1) designed to engage the transcriptional activity underlying senescence (Buchanan-Wollaston et al., 2003; Lim et al., 2003; Thomas et al., 2003).

Figure 1.

Schematic illustration of the relationships between signaling and the regulation and execution of senescence.

Recent analysis of gene expression in mutant Arabidopsis plant lines has indicated that signaling pathways involving ethylene, jasmonic acid and salicylic acid regulate the expression of genes required for developmental senescence (Buchanan-Wollaston et al., 2005). Ethylene signaling, in particular, appears to play a pivotal role in the timing of senescence, though not in its execution (Grbic & Bleecker, 1995). For example, etr1-1 mutants of Arabidopsis, in which ethylene signaling is defective, exhibit a delay in the onset of leaf senescence, but senescence per se of the leaves is not prohibited (Grbic & Bleecker, 1995). Moreover, in Arabidopsis, the ability of ethylene to initiate senescence is tightly linked to the developmental age of the plant or plant organ (Grbic & Bleecker, 1995), and it has been proposed that this reflects a capability within the plant for monitoring cell age as development progresses (Lim et al., 2003). There are, however, many exceptions to this, in that flower senescence in certain species, for example petal senescence in roses (Thompson, 1988), appears to be independent of ethylene, and similarly there is little evidence for the involvement of ethylene in leaf senescence in species such as temperate grasses.

Studies of age-related changes required for leaf senescence in Arabidopsis have resulted in characterization of a number of old (onset of leaf death) mutants (Jing et al., 2002). The ability of ethylene to induce leaf senescence under normal conditions is limited to a precise window of age, and some of these old mutants, for example old1, display an early senescence response to ethylene (Jing et al., 2005). Premature senescence is also induced by exogenous factors, including treatment with jasmonic acid, and in such instances there is an accompanying up-regulation of many of the genes known to be involved in developmentally induced senescence (He et al., 2002). Salicylic acid signaling, which underpins key plant responses to pathogen ingression, is also involved in senescence (Morris et al., 2000). In fact, mutational defects in salicylic acid signaling delay the onset of developmental leaf senescence in Arabidopsis (Buchanan-Wollaston et al., 2005).

III. Senescence transcriptome

That the transcriptome for senescing Arabidopsis leaves contains 2491 unique genes (Guo et al., 2004) implies that senescence is an active process requiring gene expression. Proteins encoded by senescence-associated genes (SAGs) include proteases, lipases, nucleases, carbohydrate and nitrogen-metabolizing enzymes and stress-responsive proteins (Gepstein et al., 2003; Lin & Wu, 2004). In addition, upwards of 100 different transcription factors from at least 20 different gene families are expressed during developmental leaf senescence (Guo et al., 2004; Buchanan-Wollaston et al., 2005). Among these are members of the MYB, AP2, NAC and WRKY transcription factor families. Several members of the WRKY transcription factor family are thought to be involved in senescence (Eulgem et al., 2000; Robatzek & Somssich, 2001, 2002). WRKY53, for example, regulates other WRKY transcription factors as well as stress genes and SAGs (Hinderhofer & Zentgraf, 2001; Miao et al., 2004). During leaf senescence, as many as 20 different NAC transcription factor family members exhibit increased expression (John et al., 1997; Andersson et al., 2004; Lin & Wu, 2004; Buchanan-Wollaston et al., 2005), and a T-DNA insertional knockout of AtNAP, an NAC transcription factor, has been shown to delay senescence (Guo & Gan, 2006). Transcription factors also act as negative regulators of leaf senescence. For instance, in studies with rice, RNAi-mediated suppression of OsDOS, a zinc finger family transcription factor, was shown to accelerate leaf senescence, resulting in hyperactivity of the jasmonate pathway (Kong et al., 2006). These findings were interpreted as indicating that OsDOS negatively regulates senescence by integrating developmental cues to the jasmonate pathway (Kong et al., 2006). DNA microarray analyses of aspen have indicated that large changes in transcriptional activity also underlie the onset of autumnal leaf senescence. Increased transcriptional activity is apparent before the onset of visible symptoms of aspen leaf senescence, and as senescence progresses there is a clear shift in gene expression away from transcripts required for photosynthesis to those necessary for mitochondrial respiration, fatty acid oxidation and nutrient mobilization (Andersson et al., 2004). It is thus apparent that developmental senescence is regulated by a complex molecular signaling network in which transcription factors play a central role.

IV. Execution of senescence

1. Post-transcriptional regulation

Although it seems clear that signaling and transcriptional networks regulate where and when senescence is initiated, there is growing evidence for a still finer level of control over the actual execution of senescence, which entails multiple cascades of molecular activation and inactivation. Moreover, for the most part, this finer level of control appears to be regulated post-transcriptionally (Thomas et al., 2003).

Recent evidence suggests that post-transcriptionally activated eukaryotic translation initiation factor 5A (eIF5A) plays a role in the post-transcriptional regulation of senescence (Wang et al., 2001; Thompson et al., 2004; Duguay et al., 2006). However, it apparently does so not by participating in the initiation of protein synthesis, which is the conventional role of translation initiation factors, but rather by selective recruitment of mRNAs from the nucleus (Kang & Hershey, 1994; Hanauske-Abel et al., 1995; Xu & Chen, 2001; Parker & Gerner, 2002; Xu et al., 2004). Plant and animal eIF5A are both post-translationally modified by two enzymes, deoxyhypusine synthase (DHS) and deoxyhypusine hydroxylase (DHH), which results in the conversion of a conserved lysine residue to the unusual amino acid, hypusine. The hypusine-modified eIF5A is thought to be the active form of the protein (Caraglia et al., 1999; Ober & Hartmann, 1999; Kruse et al., 2000; Wang et al., 2001).

Eukaryotic translation initiation factor 5A and DHS have been cloned and partially characterized for several plant species, including Senecio vernalis (Ober et al., 2003), rice (Mehta et al., 1991, 1994; Chou et al., 2004), tomato (Wang et al., 2001, 2005), Arabidopsis (Wang et al., 2001, 2003), tobacco (Chamot & Kuhlemeier, 1992; Ober & Hartmann, 1999), maize (Dresselhaus et al., 1999), and rubber tree (Chow et al., 2003). The first indication of a possible role for post-translationally activated plant eIF5A in programmed cell death came from the finding that eIF5A and DHS transcripts are up-regulated in senescing tissues (Wang et al., 2001). Studies with Arabidopsis have indicated that only one of three isoforms of eIF5A, eIF5A-1 (GenBank accession number AAG53646), is expressed in senescing leaf tissue (Thompson et al., 2004). In keeping with this, GUS reporter gene analyses have indicated that the promoter for AteIF5A-1 is active in senescing tissue (Duguay et al., 2006). Moreover, the promoter for AtDHS, which encodes one of the enzymes mediating post-translational modification of eIF5A, is also active in senescing tissue (Duguay et al., 2006), indicating that it is indeed the hypusinated form of eIF5A that is active in senescence. This contention is supported by the finding that transgenic Arabidopsis plants exhibiting suppressed DHS (Wang et al., 2003; Wang et al., 2005; Duguay et al., 2006) or eIF5A-1 (Fig. 2) exhibit delayed leaf senescence. Suppression of DHS has also been shown to delay postharvest senescence of cut carnation flowers (Fig. 3) and tomato fruit (Wang et al., 2005). Thus, hypusination of eIF5A appears to be an important element of the post-transcriptional regulation of senescence.

Figure 2.

Photographs illustrating the delayed onset of rosette leaf senescence in transgenic Arabidopsis plants exhibiting suppressed expression of AteIF5A-1, the senescence-specific isoform of eIF5A (eukaryotic translation initiation factor 5A). The plants were photographed at age 5 wk. AS-AteIF5A-1, antisense AteIF5A-1; WT, wild-type. (eIF5A-1 also regulates programmed cell death during xylogenesis, and thus the transgenic plants are smaller than the WT plants.)

Figure 3.

Photographs illustrating the delayed onset of petal senescence in cut flowers of transgenic carnation plants exhibiting suppressed expression of deoxyhypusine synthase. The flowers were cut just before the onset of petal-inrolling and photographed 6 d after harvest. AS-DHS, antisense deoxyhypusine synthase; WT, wild-type.

Protein modification by kinases and phosphatases is another mechanism by which senescence-related processes such as nutrient remobilization are regulated. More than 50 different kinases and phosphatases are expressed during developmental senescence (Guo et al., 2004; Buchanan-Wollaston et al., 2005; van der Graaff et al., 2006). One of the larger families of protein kinases in plants is the group comprising Ca2+-dependent protein kinases (CDPKs) and CDPK-related kinases (CRKs), which are serine/threonine kinases bearing a calmodulin-like domain that allows detection of calcium signals (Cheng et al., 2002; Hrabak et al., 2003). One member of this family of kinases in Arabidopsis, AtCRK3, has been found to be up-regulated during leaf senescence and to phosphorylate AtGLN1;1, a cytosolic glutamine synthetase, and may therefore facilitate nitrogen mobilization in senescing tissues (Li et al., 2006). The mitogen-activated protein kinase (MAPK) signaling pathways, which have been extensively studied in mammalian systems, have been identified in plants as well. Sequencing of the Arabidopsis genome revealed the presence of 20 MAPKs, 10 MAPK kinases and 60 MAPK kinase kinases (Ichimura, 2002), and a stress-responsive MPK6 cascade has been found to phosphorylate specific isoforms of 1-aminocyclopropane-1-carboxylic acid (ACC) synthase, leading to enhanced protein stability and increased ethylene biosynthesis (Liu & Zhang, 2004). In an earlier study, Ouaked et al. (2003) reported that MPK6 is activated in response to the ethylene precursor, ACC, and may in turn activate ethylene target genes. However, this appears not to be the case in light of the finding that ACC is unable to activate MPK6, indicating that the real role for MPK6 is in ethylene biosynthesis and not in ethylene signaling (Ecker, 2004; Liu & Zhang, 2004). Sugar homeostasis is another process that appears to be regulated by phosphorylation during senescence. Inhibition of sugar transporter function via phosphorylation appears to play an important role in programmed cell death in both plant (Bourque et al., 2002; Norholm et al., 2006) and animal systems (Berridge et al., 1996; Ahmed & Berridge, 2000; Chi et al., 2000). Senescence is also negatively regulated through the action of kinases. For instance, the antisenescence properties of cytokinins are achieved in part through the activity of kinases. Cytokinin-mediated phosphorylation of ARR2, an Arabidopsis response regulator, leads to transactivation of cytokinin-responsive genes involved in increasing leaf longevity and suppressing senescence (Kim et al., 2006). Thus, modification of protein function by the action of kinases allows minute and timely control over the execution of senescence.

2. Disruption of functional compartmentalization

Loss of selective permeability, the property of membranes that underpins functional compartmentalization in eukaryotic cells, is an early event in the senescence cascade. Indeed, it is the selective permeability of membranes, in conjunction with their specific transporters, that enables establishment and maintenance of metabolite and ion gradients such as the H+ gradient, which are essential for cell function (Thompson, 1988). Moreover, many sugar and amino acid transporters are functionally dependent upon maintenance of ion or metabolite gradients across the membrane (Thompson et al., 1997). One of the clearest manifestations of declining membrane selective permeability in senescing tissues is the early onset of leakiness as these gradients dissipate. In senescing carnation flowers, for example, changes in permeability reflecting membrane deterioration are initiated before the climacteric rise in ethylene production (Eze et al., 1986). This early onset of membrane leakiness raises the question of how sufficient membrane structural integrity is maintained throughout the course of senescence to enable recruitment and translocation of nutrients from the dying cells. Furthermore, the execution of senescence requires energy, and thus there is also a requirement for retention of at least some capacity for ATP formation by thylakoid membranes or mitochondrial membranes into the later stages of senescence (Matile, 1992). This latter requirement appears to be met in part by the fact that, although thylakoids undergo degradative changes very early in the senescence cascade, loss of mitochondrial membrane function is delayed and does not occur until the later stages of senescence (Kolodziejek et al., 2003; Keskitalo et al., 2005).

3. A role for senescence-associated lipases

The lipid bilayer serves as the structural framework of membranes and is also the basis for their selective permeability. That is, lipid bilayers are inherently impermeable to polar metabolites and ions, and are rendered selectively permeable by large protein complexes that serve as specialized transporters (Thompson et al., 1998). The onset of membrane leakiness during the early stages of senescence arises because of changes in the composition and molecular organization of lipid bilayers that disrupt their impermeable nature. Paramount among these is extensive de-esterification of membrane phospholipids and galactolipids. Indeed, declining lipid phosphate concentrations reflecting phospholipid catabolism are often evident well before morphological manifestations of tissue aging (Paliyath & Thompson, 1990) and have been demonstrated for a variety of senescing tissues, including leaves, cotyledons, flower petals and ripening fruit (Thompson et al., 1998). This in turn leads to phase separations within the lipid bilayers of senescing membranes, because the de-esterified fatty acids, once formed, come together by moving laterally through the plane of the membrane to form free fatty acid-enriched domains. The resultant mixture of lipid phases causes the bilayer to become leaky because of packing imperfections at the phase boundaries, and leads to loss of ion and metabolite gradients that are essential for normal cell function (Barber & Thompson, 1980, 1983; Thompson, 1988; Yamane et al., 1993).

It can be inferred from these observations that lipases capable of de-esterifying membrane lipids play a key role in initiating the decline in membrane structural integrity that leads to loss of intracellular compartmentalization in senescing tissues. This inference is supported by the fact that senescence-associated lipases exhibiting lipolytic acyl hydrolase activity have been identified (Hong et al., 2000; He & Gan, 2002). Indeed, one such lipase cloned from senescing carnation petals, which are climacteric, is inducible by treatment with ethylene at concentrations that invoke early flower senescence (Hong et al., 2000). Lipolytic acyl hydrolases de-esterify fatty acids at both the sn-1 and sn-2 positions of phospholipids, thereby enabling the rapid de-esterification of membrane fatty acids that is a characteristic feature of both natural senescence and postharvest senescence (Thompson et al., 1998; Page et al., 2001). In studies with transgenic plants, overexpression of lipolytic acyl hydrolase has been shown to induce precocious leaf senescence, and antisense suppression of its activity to delay leaf senescence (He & Gan, 2002). These observations are consistent with the contention that lipases capable of de-esterifying membrane fatty acids play a proactive role in initiating senescence. De-esterification of galactolipids, the other major membrane lipid in senescing leaves, appears to be in part mediated directly by galactolipases (Kim et al., 2001; Matos et al., 2001). Further to this, a lipase gene (accession number AAM14092), which was identified in the transcriptome for Arabidopsis leaf senescence (Guo et al., 2004), has been shown to be localized in the chloroplast (C.A. Taylor & J.E.Thompson, unpublished), which suggests it may de-esterify galactolipid-rich thylakoid membranes during senescence. However, it is also apparent that galactolipids in senescing leaves are de-esterified indirectly through the combined actions of galactosidase, which releases diacylglycerol, and a lipolytic acyl hydrolase capable of de-esterifying diacylglycerol (Lee et al., 2004).

That catabolism of membrane lipids is an inherent feature of senescence is further supported by recent analyses of senescence-associated changes in gene expression (Gepstein et al., 2003; Guo et al., 2004). For example, of the approximately 6200 expressed sequence tags corresponding to 2491 unique genes in the transcriptome for senescing Arabidopsis rosettes, 11 of the unique genes are lipases/acyl hydrolases, six are phospholipases, two are lipoxygenases, and nine are β-oxidation enzymes (Guo et al., 2004). These proteins are all involved in membrane lipid catabolism.

4. Asynchronous destabilization of membrane structure

It is clear that de-esterification of membrane lipids will lead to loss of membrane structural integrity. Phospholipids and, in the case of thylakoids, galactolipids are important structural molecules in lipid bilayers. However, free fatty acids, one of the immediate products of lipid de-esterification, act like detergents when released into the bilayer and destabilize membrane structure. Moreover, lipase-mediated de-esterification of fatty acids in lipid bilayers is autocatalytic. This arises because free fatty acids formed by the action of lipases perturb bilayer structure, and perturbed bilayers are more effective substrates for de-esterifying lipases than their unperturbed counterparts (Goormaghtigh et al., 1981). Indeed, in senescing membranes, bilayer perturbation is not only attributable to the free fatty acids themselves, but also to lipid phase separations as well as the formation of nonbilayer lipid configurations such as inverted micelles within the bilayer (Thompson, 1988).

That fatty acid de-esterification occurs early in the senescence cascade and is autocatalytic posits an interesting dilemma in the context of nutrient mobilization, because translocation of nutrients out of the dying tissue is dependent upon maintenance of membrane functional integrity. This predicament appears to be addressed in part by the fact that different types of membranes undergo structural deterioration at different stages of senescence. For example, in senescing carnation flowers, lipid phase separations, which render bilayers leaky, are evident in the membranes of the endoplasmic reticulum before the climacteric rise in ethylene production and petal – inrolling. Yet, lipid phase separations in the plasmalemma are not discernible until a later stage of petal senescence (Paliyath & Thompson, 1990). Thus, the selectively permeable nature of the plasmalemma is retained during the period when there is a need for nutrient translocation out of senescing cells. Freeze-fracture electron microscopy has indicated that, even within a senescing membrane, there is strict regulation of lipid de-esterification and ensuing phase separations. During the early stages of endoplasmic reticulum senescence, only small domains of gel phase lipid are evident in the plane of the membrane bilayer, and these are separated by large intervening regions of normal liquid-crystalline phase lipid which have retained structural integrity (Paliyath & Thompson, 1990), indicating that bilayer destabilization does not occur synchronously over the entire membrane. Moreover, it is well established that as domains of gel phase lipid form, proteins are laterally displaced into adjacent liquid crystalline domains and presumably remain functional (Thompson et al., 1998). Thus, even though membrane senescence in a particular organelle, such as the endoplasmic reticulum, may be initiated early in the cascade, loss of structural integrity is a gradual process, and regions of the membrane continue to serve as a scaffold for functional proteins into the later stages of senescence.

The asynchronous deterioration of the energy-producing membranes, chloroplast thylakoids and the cisternae of mitochondria, during senescence is a further manifestation of regulated membrane degradation in senescing tissues. A decline in leaf chlorophyll reflecting molecular degradation within thylakoids is widely accepted as being one of the earliest indices of the onset of leaf senescence (Matile et al., 1996). Yet, it is not until much later in the senescence cascade that the inner mitochondrial membranes show signs of deterioration (Kolodziejek et al., 2003). Thus chloroplastic energy production is lost during the early stages of senescence, but mitochondrial energy production is preserved until much later. Energy production by chloroplasts is very efficient, and since the ongoing execution of senescence requires energy, one might wonder why chloroplast energy production is forfeited so early in senescing leaves. Some insight into this comes from the realization that thylakoid lipids are the most abundant carbon source in leaves, and that capture and mobilization of this carbon are likely to be most efficient if they occur during the earlier stages of senescence, while much of the cellular metabolic capability is still functional. Moreover, some of the fatty acids de-esterified from galactolipids are catabolized by β-oxidation (Buchanan-Wollaston, 1997; Charlton et al., 2005), and thus contribute to mitochondrial ATP formation as senescence progresses.

5. Release of membrane lipid catabolites

Remobilization of membrane fatty acid carbon during senescence is an elaborate process that entails β-oxidation and/or gluconeogenesis. Yet for this to be possible, the de-esterified fatty acids have to be removed from the bilayer, and this does not readily occur. Indeed, because of their amphipathic nature, individual free fatty acids partition out of membrane bilayers only sparingly. This can lead to an increase in the free fatty acid : esterified fatty acid ratio of senescing membranes (Thompson et al., 1997). However, the accumulation of free fatty acids within senescing membranes is limited and they appear to be released from the bilayer in a modulated fashion as they are formed, thus minimizing their disruptive detergent-like effects on bilayer structure. This arises because of the fact that free fatty acids move laterally through the plane of the bilayer to form discrete gel-phase domains (Paliyath & Thompson, 1990; Thompson et al., 1998). These domains maintain only a weak association with bulk membrane lipid because of packing imperfections at the phase boundaries and, accordingly, tend to be voided from the membrane surface into adjacent aqueous compartments. Evidence supporting this comes in part from studies demonstrating that cytosolic particles enriched in free fatty acids of membranous origin are abundant in senescing tissues (McKegney et al., 1995).

Comparative chemical analyses of the lipid composition of senescing membranes have indicated that free fatty acids are also assimilated into metabolites that can be more easily accommodated within membrane bilayers. There is, for example, a striking increase in concentrations of steryl/wax esters and of triacylglycerol in senescing membranes (Thompson et al., 1997; Kaup et al., 2002). Both of these molecular species have less of a disruptive effect on bilayer structure than free fatty acids. Indeed, the formation of triacylglycerol in membranes is increasingly being regarded as a means of temporarily sequestering free fatty acids in a structurally inert form (Kaup et al., 2002). Moreover, the finding that lipid particles isolated from the cytosol of senescing tissue are enriched not only in free fatty acids, but also in triacylglycerol and steryl/wax esters suggests that free fatty acids as well as their metabolites are released from senescing membranes as voided particles (McKegney et al., 1995).

6. Catabolism of membrane proteins

Upwards of 75% of the total cellular nitrogen in leaves is localized in chloroplasts (Hoertensteiner & Feller, 2002). Much of this is embedded in stromal ribulose bisphosphate carboxylase (Rubisco), but a sizable proportion is also tied up in thylakoid proteins, in particular the apoprotein, LHCP II (Matile, 1992). Recovery and remobilization of this nitrogen during the breakdown of chloroplasts are of vital importance if the plant is to derive full benefit from the senescence of its leaves. To this end, of the 2491 unique genes identified as components of the transcriptome for Arabidopsis leaf senescence, 116 are predicted to be involved in some aspect of protein degradation. Of these, 41 encode proteases, and the remaining 75 appear to encode elements of the ubiquitin-proteasome pathway (Guo et al., 2004). In fact, cysteine proteases, in particular SAG12, are the most abundant components of this transcriptome (Guo et al., 2004). Chloroplast proteolysis is initiated during the early stages of senescence when much of the cellular metabolic capability is still intact (Matile et al., 1996; Masclaux et al., 2000), and this presumably helps to ensure successful remobilization of the extensive nitrogen localized in this organelle.

Recovery of the nitrogen equivalents within thylakoid proteins is a challenging process because the most abundant apoproteins are tightly associated with chlorophyll. In light of this, it is perhaps not surprising to find that proteolysis of LHCPs in senescing leaves is tightly linked to chlorophyll degradation (Tsuchiya et al., 1999). Among the most compelling lines of evidence supporting this argument is the finding that genetic intercession in the pathway for chlorophyll degradation causes a corresponding disruption of LHCP II proteolysis during leaf senescence (Ougham et al., 2005). Further evidence for the synchronization of LHCP II degradation and chlorophyll catabolism in senescing leaves has come from studies of stay-green mutants (Bachmann et al., 1994). These mutants contain genetic lesions in the chlorophyll degradation pathway and retain much of their chlorophyll as senescence progresses. Since chlorophyll degradation and LHCP II proteolysis are codependent, the stay-green mutants also exhibit dramatically delayed proteolysis of LHCP II, yet the degradation of Rubisco is unimpeded. Thus, the effect of disrupted chlorophyll degradation on proteolysis appears to be localized to those proteins that are structurally associated with chlorophyll, an association that may preclude access of proteases to such proteins. The codependence of LHCP II proteolysis and chlorophyll degradation is consistent with the observation that membrane proteins must undergo a conformational change in order to be recognized by proteases. Recently, a Zn2+-dependent metalloprotease, FtsH6, was shown to be responsible for the senescence-associated degradation of LHCII (Zelisko et al., 2005). Moreover, apparently it is the nonphosphorylated form of LHCPII that gets degraded in senescing leaves (Yang et al., 1998). Ultimately, there is complete degradation of thylakoid proteins in senescing leaves (Kolodziejek et al., 2003), and this may in part be facilitated by changes in the molecular organization of thylakoids as senescence progresses. For example, relocation of light harvesting complex II from the stacked granal region to the stromal lamellar region housing Photosystem I complexes has been observed for senescing Cucumis cotyledons, and it has been proposed that this is related to remobilization of thylakoid molecular components (Prakash et al., 2001).

There are also reports of decreased amounts of plasmalemma proteins in senescing tissues, including petunia petals (Borochov et al., 1994), daylily tepals (Lay-Yee et al., 1992) and iris tepals (Celikel & Doorn, 1995). Moreover, one of the triggers of proteolysis in cellular membranes of senescing tissues appears to be changes in the molecular organization of the lipid bilayer. For example, the formation of gel phase lipid engenders lateral displacement of proteins and free sterols out of the forming gel phase domains into adjacent liquid-crystalline lipid (Paliyath & Thompson, 1990; Ismail et al., 1999). Sterols associate with phospholipids, reducing the rotational motion of their fatty acid side chains (Shinitzky & Inbar, 1976). Accordingly, as gel phase domains form in senescing membranes, there is a corresponding decrease in the fluidity of residual liquid-crystalline lipid, the lipid that solvates membrane proteins, including those displaced from gel phase domains (Thompson et al., 1987; Duxbury et al., 1991). This decrease in bulk lipid fluidity in turn induces conformational changes in membrane proteins that presumably render them more prone to proteolysis. For example, an increase in sterol : phospholipid ratio has been correlated with altered protein conformation and receptor function (Kirby & Green, 1980). Moreover, spin-labeling studies of microsomal membranes isolated from senescing tissue have shown that the decrease in bilayer fluidity accompanying senescence is of sufficient magnitude to alter the conformation of membrane proteins (Duxbury et al., 1991).

Ubiquitination, which directs proteins to the 26S proteasome, is one of the more common means of targeting proteins for degradation (Hershko & Ciechanover, 1998; Callis & Vierstra, 2000), and there is growing evidence that the ubiquitin-proteasome pathway may be involved in the initiation of senescence. For example, Woo et al. (2001) have reported that the ORE9 F-box protein, a component of the SCF complex in the ubiquitin pathway, positively regulates leaf senescence in Arabidopsis. They propose that it does so by facilitating ubiquitin-mediated degradation of a transcriptional repressor of senescence-associated genes (Woo et al., 2001). Furthermore, components of the ubiquitin pathway, including pSEN3, an Arabidopsis cDNA clone encoding ubiquitin, At1g14400 (ubiquitin carrier protein) and At1g53750 (26S proteosome ATPase subunit), are up-regulated in senescing leaves (Park et al., 1998; Gepstein et al., 2003). Thus, while it is apparent that ubiquitin-mediated proteolysis may play a role in regulating the onset of senescence, other evidence suggests that it is not involved in the massive degradation of proteins incurred during the actual execution of senescence. For example, Bahrami & Gray (1999) reported that a gene from tobacco with high sequence similarity to human and yeast α subunit proteasome genes is highly expressed in young tissue, but not in senescing tissue. Moreover, disruption of proteasome function in plants by virus-induced silencing of proteasome subunits has been shown to induce programmed cell death analogous to the HR cell death induced by pathogens (Kim et al., 2003).

7. Molecular disassembly of membranes

It is clear from ultrastructural studies that, as senescence progresses, cellular membranes and organelles completely disappear, although they do so gradually and in temporally different patterns (Kolodziejek et al., 2003). This poses a logistical problem in that membrane lipids and proteins are amphipathic in nature and not soluble in aqueous media. Indeed, phospholipids and some galactolipids spontaneously form bilayers in aqueous media, and fatty acids form micelles. It is apparent, however, that membranes are not disassembled molecule by molecule. Rather, it would appear that catabolism of the macromolecular constituents of membranes, in particular membrane lipids, results in the formation of bilayer-destabilizing molecules, for example free fatty acids, which phase-separate into domains that are voided from the membrane surface.

Perhaps the best documented evidence for this is the formation of plastoglobuli accompanying the dismantling of thylakoids in senescing chloroplasts (Matile, 1992; Kaup et al., 2002). Analysis of the molecular composition of plastoglobuli has indicated that they do in fact contain thylakoid lipids and their catabolites, including plastoquinone, α-tocopherol, triacylglycerol, carotenoids and free fatty acids (Steinmuller & Tevini, 1985; Kaup et al., 2002). Other stromal particles with a higher buoyant density than plastoglobuli, but still containing galactolipids and their catabolites, including thylakoid-specific free fatty acids, have also been isolated from senescing leaf tissue (Ghosh et al., 1994; Smith et al., 2000). The distinguishing feature of these additional plastoglobuli-like stromal particles is that they not only contain thylakoid lipids and their catabolites, but also thylakoid proteins and their catabolites (Ghosh et al., 1994; Smith et al., 2000). They also contain the plastoglobular protein, PAP, which is present in small amounts on thylakoid membranes as well (Smith et al., 2000) and is thought to maintain the structural integrity of plastoglobuli in a manner analogous to that for oleosin associated with oil bodies (Rey et al., 2000). Indeed, constitutive overexpression of PAP results in an enhanced abundance of plastoglobuli in the stroma of chloroplasts, an observation that has prompted the view that PAP may be involved in the voiding of plastoglobuli from the thylakoid membrane surface (Rey et al., 2000). Further evidence supporting the contention that plastoglobuli are derived from thylakoids has been obtained by high-pressure freezing/freezing-substitution, electron tomography, immunoelectron tomography and freeze-etch electron microscopy (Austin et al., 2006). These findings indicate that plastoglobuli form on thylakoid membranes and are circumscribed by a half-unit membrane that is continuous with the thylakoid outer leaflet. The authors also contend that plastoglobuli are not free-floating in the stroma, but, rather, remain physically attached to each other and to the thylakoid membrane through a continuous lipid monolayer (Austin et al., 2006).

Cytosolic lipid-protein particles that in some respects appear to be analogs of plastoglobuli have also been identified. These particles are obtained by ultrafiltration or flotation centrifugation of purified cytosol and contain phospholipids, confirming that they are of membrane origin (Yao et al., 1991; Hudak & Thompson, 1996). It has been proposed that, as for oil bodies (Murphy, 1993), the phospholipid exists as a monolayer circumscribing the surface of cytosolic particles (Yao et al., 1991). The distinguishing characteristic of these particles is that, like plastoglobuli, they are enriched in free fatty acids, steryl/wax esters and triacylglycerol, all of which are formed in senescing membranes (McKegney et al., 1995; Hudak & Thompson, 1996), as well as peroxidized lipids (Yao et al., 1991, 1993). Moreover, they also contain catabolites of the plasma membrane-specific H+-ATPase (Hudak & Thompson, 1996). In light of these findings, it is at least plausible that, as in the case of plastoglobuli, cytosolic lipid-protein particles are formed by voiding of destabilizing domains of lipid and protein catabolites from the surfaces of senescing cellular membranes. This possibility is supported in part by the finding that lipid-protein particles analogous to those isolated from purified cytosol can be formed in vitro from isolated microsomal membranes following activation of phospholipid catabolism (Yao et al., 1991; Hudak & Thompson, 1996).

8. Recruitment of nutrients

The protein and lipid constituents of membranes are rich sources of nitrogen and carbon, respectively, that have to be recovered if the nutritional benefit of senescence is to be fully realized. Membrane lipids are also a rich source of phosphorus, and it is clear from studies of overwintering trees that the transfer of both nitrogen and phosphorus from senescing leaves to perennial tissues is critical to plant fitness (Hoch et al., 2003). If it is the case that molecular disassembly of senescing membranes is achieved by progressive voiding of particles from the bilayer, the lipid and protein components of these particles must be further metabolized into amino acids, phloem-mobile sucrose and nonlipid phosphate. Thus the questions: what happens to these particles; and how are the nitrogen, phosphorus and carbon equivalents of their macromolecular constituents mobilized?

There is some evidence that autophagy plays a role in nutrient recycling. Autophagy entails encapsulation of portions of the cytoplasm, including organelles, within autophagosomes which then fuse with the tonoplast, releasing their captured cargo into the vacuole where it is degraded (Marty, 1999; Kim & Klionsky, 2000; Klionsky & Emr, 2000). Autophagic degradation appears to be invoked when there is a need for rapid mobilization of nutrients as in the case of senescence (Doelling et al., 2002). The role of autophagy in senescence is thought to be of particular importance as a means of degrading cytosolic proteins (Dunn, 1994; Marty, 1999; Kim & Klionsky, 2000). However, it is reasonable to assume that cytosolic lipid-protein particles derived from senescing membranes would be swept up into autophagosomes as they form. Evidence supporting the involvement of autophagy in senescence includes the finding that analogs of components of the yeast APG8 and APG12 yeast conjugation pathways involved in autophagy have been identified in Arabidopsis, and one of these, APG7, appears to be required for leaf senescence (Doelling et al., 2002). Furthermore, AtATG18a, one of eight members of the AtATG18 family in Arabidopsis exhibiting sequence homology with the yeast autophagy gene, ATG18, has been shown to be up-regulated in senescing leaves, and its suppression engenders premature leaf senescence (Xiong et al., 2005).

That loss of chlorophyll is accelerated, rather than delayed, in Arabidopsis plants with impaired autophagic function suggests that, at least in senescing tissue, chloroplasts are not normally degraded by autophagy (Doelling et al., 2002). There is, however, one report of chlorophyll-containing plastoglobuli being exuded through the chloroplast envelope into the cytoplasm in senescing leaves (Guiamet et al., 1999). Moreover, chlorophyllase genes encoding proteins deduced to be cytosolic have been identified in Arabidopsis (Tsuchiya et al., 1999) and citrus (Jakob-Wilk et al., 1999). The implied existence of extraplastidial forms of chlorophyllase supports the notion advanced by Hoertensteiner & Feller (2002) that catabolites generated inside chloroplasts and associated with plastoglobuli may be further metabolized in the cytosol after their release across the chloroplast envelope. That such metabolism might be achieved by autophagy is supported by immunocytochemistry indicating that small spherical particles containing Rubisco and/or its degradation products are detectable in the cytosol and also in the vacuole of senescing leaves (Chiba et al., 2003). Moreover, the abundance of these particles increases with advancing senescence (Chiba et al., 2003).

It is by no means certain, however, that there is not also extensive catabolism of plastoglobular components within senescing chloroplasts. For example, Rubisco, a stromal protein, is readily hydrolyzed within isolated chloroplasts, implying that there are stromal proteases (Thoenen & Feller, 1998; Rudella et al., 2006). It is thus conceivable that the protein elements of stromal plastogobuli-like particles (Smith et al., 2000) are degraded, partially or completely, by these proteases. In addition, a TAG lipase (accession number AAY78709) has recently been cloned from Arabidopsis, and confocal microscopy has indicated that its cognate protein is localized in chloroplasts in association with plastoglobular lipid (Padham et al., 2007).

9. Membrane fatty acid catabolism

Membrane fatty acids are metabolized to acetyl coenzyme A (CoA) by β-oxidation within glyoxysomes, forming acetyl CoA, which can either be used for energy production to support senescence or, in most plants but apparently not all (Charlton et al., 2005), converted to oxaloacetate, leading ultimately to gluconeogenesis and the formation of phloem-mobile sucrose (DeBellis et al., 1990; Froman et al., 2000; Page et al., 2001; Cornah & Smith, 2002). To this end, peroxisomes are converted to glyoxysomes in senescing tissues (DeBellis et al., 1990).

Key to the onset of this mobilization of carbon is the translocation of de-esterified membrane fatty acids to glyoxysomes. There is some accumulation of free fatty acids in senescing membranes, but the most dramatic change in membrane lipid composition with advancing senescence is an increased abundance of steryl/wax esters and triacylglycerol (McKegney et al., 1995; Kaup et al., 2002) (Fig. 4). The synthesis of steryl/wax esters and triacylglycerol requires fatty acids, and the enzymes required for their formation are membrane-associated (Garcia & Mudd, 1978; Lu et al., 2003). There is, for example, a progressive accumulation of triacylglycerol coincident with the dismantling of galactolipids during foliar senescence that correlates temporally with up-regulation of chloroplast membrane-associated DGAT1, the enzyme that mediates the terminal step in triacylglycerol synthesis (Kaup et al., 2002). Moreover, this triacylglycerol is enriched in hexadecatrienoic acid, which is uniquely associated with thylakoid galactolipids (Kaup et al., 2002).

Figure 4.

Formation of triacylglycerol and steryl/wax esters in membranes of senescing Arabidopsis rosette leaves. The fatty acid (FA) equivalents of triacylglycerol (TG) and steryl/wax esters (SWE) in microsomal membrane preparations from rosettes are expressed as a percentage of the corresponding total microsomal membrane fatty acid complement. Closed bars, young leaves; open bars, senescing leaves.

Recognizing that free fatty acids, if allowed to accumulate unchecked in senescing membranes, would lead to unregulated premature disassembly of bilayer structure because of their detergent-like action (Thomas, 1982), it seems reasonable to propose that the formation of steryl/wax esters and triacylglycerol enables de-esterified fatty acids released in senescing membranes to be temporarily sequestered in molecules that, to a greater degree than free fatty acids, can be accommodated within the bilayer. The finding that both cytosolic lipid-protein particles and plastoglobuli are enriched in steryl/wax esters and triacylglycerol (Martin & Wilson, 1984; Ghosh et al., 1994; McKegney et al., 1995; Kaup et al., 2002) suggests that these molecules in turn phase-separate, forming discrete domains that are subsequently voided. Thereafter, although there is no evidence to date for this eventuality, it is conceivable that these particles associate with glyoxysomes where the fatty acids of steryl/wax esters and triacylglycerol are again de-esterified, enabling their metabolism by β-oxidation and the glyoxylate cycle. Recent evidence suggesting that plastoglobuli of senescing chloroplasts are exuded through the envelope into the cytoplasm (Guiamet et al., 1999; Paramonova et al., 2004) lends further credence to this notion as a plausible explanation for how the fatty acid equivalents of plastoglobuli gain access to glyoxysomes.

V. Conclusions and outlook

A plethora of novel and interesting discoveries over the last decade has greatly enhanced our knowledge and understanding of senescence. Yet, many unanswered questions remain. It is apparent that senescence is an evolutionarily conserved process and is required for the fitness and fecundity of plants. From a practical perspective, massive annual losses in yield attributable to postharvest senescence as well as pathogen- and stress-induced premature senescence continue to threaten the commercial viability of many crops. Despite the incentive to reduce these losses through a better understanding of senescence, our knowledge of the molecular cascades regulating this terminal phase of development is far from complete. Indeed, a number of SAGs identified through transcriptional analysis have yet to be characterized (Buchanan-Wollaston et al., 2005). Moreover, many of the recent new discoveries pertaining to senescence have been obtained from experiments with Arabidopsis, and certain aspects of senescence in this species do not mirror their counterparts in typical crop species. Accordingly, generalization of data for Arabidopsis to senescence of other species is not always appropriate. Molecular expression analyses and the growing sophistication of the senescence transcriptome established through microarray analysis have provided significant insights into the nature and role of changes in gene expression during senescence, but these technologies alone will not elucidate the complexities and interactivities of the molecular cascades underlying senescence. Indeed, as our knowledge of senescence grows, it is becoming increasingly clear that a combination of physiological, biochemical, genetic and molecular approaches will be required to fully elucidate the exquisite regulation of both its initiation and execution.


The authors gratefully acknowledge funding support from the Natural Sciences and Engineering Research Council of Canada.



  • Summary 201

  • I. Introduction 202
  • II. Role of signaling cascades in the initiation of senescence 202
  • III. Senescence transcriptome 203
  • IV. Execution of senescence 203
  • V. Conclusions and outlook 210
  • Acknowledgements 210

  • References 211