Introduction

Macroautophagy (hereafter referred to as “autophagy”) is a conserved eukaryotic mechanism, which is classically defined as the degradation of cytoplasmic constituents in the lytic organelle (vacuoles in yeast and plants and lysosomes in mammals) (Xie and Klionsky 2007). The general targets of autophagy vary from long-lived proteins to protein complexes and even organelles (Reumann et al. 2010). Morphologically, autophagy begins in the formation of cup-shaped double membranes, which expand to form autophagosomes engulfing malfunctioning or unneeded macromolecules and organelles and transport them for degradation inside the vacuole. Upon arrival of the autophagosomes to the vacuoles, their outer membranes fuse with the tonoplast, creating single-membrane vesicles inside the vacuole, termed “autophagic bodies.” The autophagic bodies and their contents are then degraded inside the vacuole, providing recycled materials to build new macromolecules (Bassham 2009). The genes participating in the autophagic process (termed A u T opha G y-related or ATG genes) were originally discovered in yeast (Saccharomyces cerevisiae), using autophagy-defective mutants whose cells show little or no accumulation of autophagic bodies during nutrient starvation (Xie and Klionsky 2007). Many of the ATG genes are conserved in evolution, and homologs to the yeast genes have been found in many organisms including mammals and plants (Reumann et al. 2010; Tanida 2011). The core autophagy machinery is divided into three groups: (i) the ATG9 cycling system including ATG1, ATG2, ATG9, ATG13, ATG18, and ATG27; (ii) the PI3 kinase complex comprised of ATG6/VPS30/BECLIN1, ATG14, VPS15, and VPS34; and (iii) the ubiquitin (Ub)-like protein system comprised of ATG3, ATG4, ATG5, ATG7, ATG8, ATG10, ATG12, and ATG16 (for examples, see the following reviews: Xie and Klionsky 2007; Farré et al. 2009; Suzuki and Ohsumi 2010; Tanida 2011). This review focuses on selected aspects of plant autophagy, which have not been extensively reviewed before. A more comprehensive view of the plant autophagy machinery and its multiple functions can be obtained in the following representative reviews as examples (Thompson and Vierstra 2005; Bassham et al. 2006; Bassham 2007, 2009; Kwon and Park 2008; Hayward and Dinesh-Kumar 2010; Reumann et al. 2010; Yoshimoto et al. 2010).

The impact of autophagy on plant growth and response to stress

Autophagy studies in plants initially focused on nutrient starvation and senescence. Homologs of many of the yeast ATG genes were found in plants and studies of T-DNA knockout mutants in these genes existing in the model plant Arabidopsis thaliana (such as atg7, atg5, atg4a4b, atg9, atg10, atg12) revealed increased sensitivity to carbon and nitrogen starvation (Doelling et al. 2002; Hanaoka et al. 2002; Yoshimoto et al. 2004; Thompson et al. 2005; Phillips et al. 2008; Chung et al. 2010). In addition, under short-day conditions, the autophagy knockout Arabidopsis mutants exhibited early senescence of rosette leaves and cotyledons and also had a lower seed set than wild-type plants. Under favorable (nonstress) growth conditions, most of the Arabidopsis ATG knockout lines examined so far develop relatively normally (Doelling et al. 2002; Hanaoka et al. 2002; Yoshimoto et al. 2004; Thompson et al. 2005; Phillips et al. 2008; Chung et al. 2010). The normal development of ATG knockout plants is surprising since, in mammals, knockout of ATG genes may lead to severe developmental phenotypes such as embryo lethality (Kuma et al. 2004; Komatsu et al. 2005). One exception to this rule is the ATG6 knockout plant which displays aberrant pollen germination, and therefore homozygous knockout plants were not achieved (Fujiki et al. 2007; Qin et al. 2007). It has been argued that this phenotype stems from the possible involvement of ATG6 in cellular processes and is different than the classical autophagy process of nutrient recycling (Fujiki et al. 2007).

In the past few years, our knowledge of the functions of autophagy in plants has been greatly expanded. Autophagy in plants has been shown to occur at basal levels under favorable (nonstress) growth conditions (Sláviková et al. 2005; Inoue et al. 2006) and was also shown to be involved in the response of plants to various abiotic and biotic stresses as well as in hormonal control (Xiong et al. 2007; Slavikova et al. 2008; Liu et al. 2009; Hayward and Dinesh-Kumar 2010; Vanhee et al. 2011) . It is also important to note that degradation of the major plastid protein ribulose-1,5-bisphosphate carboxylase oxygenase (RUBISCO) as well as entire chloroplasts during senescence was shown to be performed by autophagy (Ishida et al. 2008; Wada et al. 2009).

Another fascinating aspect is the role of autophagy in plant–pathogen interactions (Cacas 2010; Hayward and Dinesh-Kumar 2010). Initially, autophagy was suggested to function as an “anti-death” mechanism, limiting the extent of programmed cell death (PCD) during the hypersensitive response (Liu et al. 2005). Later, however, autophagy was suggested to take part in an opposite “pro-death” activity, since ATG mutants seemed to exhibit lower levels of PCD in response to pathogen infection (Hofius et al. 2009). A recent study (Lenz et al. 2011) attempted to settle this contradiction by suggesting that autophagy is involved in limiting necrosis, a form of disease-associated cell death. In addition, autophagy has been shown to be involved in plant–pathogen resistance by fine-tuning the salicylic acid-mediated immune response, although the specific functions of autophagy in these processes are still a matter of debate (Love et al. 2008; Cacas and Diamond 2009).

The conservation of the autophagy machinery within the plant kingdom: comparing Arabidopsis, Chlamydomonas, and rice

The proteins composing the core autophagy machinery in plants (particularly studied in the model plant Arabidopsis) have been extensively studied and reviewed (see, for example, Bassham 2007; Kwon and Park 2008). In contrast, studies on the evolution of the autophagy machinery among photosynthetic organisms, ranging from lower species, such as Chlamydomonas reinhardtii, to major crop plants, such as cereals, are still in their infancy and have not been extensively reviewed. The morphological conservation of starvation-induced autophagy structures and the published sequence conservation of ATG genes and proteins in fully sequenced genomes of various photosynthetic organisms allow us to predict: (i) which autophagy-associated mechanisms and core signal transduction cascades might have been conserved between ancient photosynthetic organisms and different species of higher plants; and (ii) which autophagy-associated biological processes evolved plant-specific components and strategies required for the adaptation of plant physiology to various ecological needs (Reumann et al. 2010). To address this issue, we discuss in the following the comparative evolution of autophagy in the photosynthetic alga C. reinhardtii and in the monocot cereal rice (Oryza sativa) in comparison to Arabidopsis. All of these organisms have fully sequenced genomes and are also amongst the few members of the plant kingdom that already have relatively detailed experimental evidence for the existence of ATG genes (Crespo et al. 2005; Su et al. 2006; Díaz-Troya et al. 2008a; Shin et al. 2009; Pérez-Pérez et al. 2010). To address this issue, we rely on published data (Bassham et al. 2006; Su et al. 2006; Meijer et al. 2007; Díaz-Troya et al 2008a; Shin et al. 2009) and on our own basic local alignment search tool (BLAST)-generated analyses of sequence homology comparisons of the core autophagy genes of Chlamydomonas, rice, Arabidopsis, and the yeast Sacharomyces cerevisae. A schematic representation summerizing the existence of core autophagy genes in these photosynthetic model organisms is provided in Fig. 1, while Table 1 provides a more detailed description of the ATG gene terminology in the chosen organisms compared to yeast.

Fig. 1
figure 1

The core autophagy machinery genes in three different photosynthetic model organisms: Chlamydomonas, Arabidopsis, and Rice, compared to yeast. The yeast 18 core autophagy genes were divided into three functional groups according to latest reviews (Xie and Klionsky 2007; Farré et al. 2009; Suzuki and Ohsumi 2010; Tanida 2011): Atg9 cycling system (orange), PI3 kinase complex (purple) and Ub-like protein system (blue). Continues line boxes stand for single genes, dotted line boxes stand for missing genes, and filled boxes stand for gene families. Upper panel Chlamydomonas, middle panel Arabidopsis, and lower panel rice (Oryza sativa). Full gene terminology is given in Table 1

Full size image
Table 1 Core autophagy machinery orthologs in the model organisms: yeast, algae, and plantsa,b
Full size table

The alga C. reinhardtii is an excellent example of a simple single-celled model of a lower photosynthetic organism whose full genome sequence is available (Merchant et al. 2007). As can be seen in Table 1, the Chlamydomonas genome contains homologous genes for most of the core autophagy machinery genes existing both in yeast and Arabidopsis. Similarly to yeast, all predicted Chlamydomonas core ATG genes are single-copy genes, and this organism possesses no unique ATG genes that are missing in Arabidopsis. In contrast, some of the ATG genes present in Arabidopsis possess no homologs in Chlamydomonas (ATG2, ATG9, ATG13, ATG16), suggesting that either these ATG genes do not exist in Chlamydomonas, or that functional homologs of those genes exist in Chlamydomonas but possess sufficient diversification that render them unidentified in a regular BLAST search (Fig. 1). Moreover, higher plants evolved either small or large families of specific ATG genes to perform specialized functions that are apparently not present and not needed in ancient algae. Pioneering work identified and characterized the Chlamydomonas ortholog of target of rapamycin (TOR) kinase, CrTOR (Crespo et al. 2005). In yeast and mammals, rapamycin treatment is used to induce autophagy via inhibition of the TOR kinase (Cebollero and Reggiori 2009; Jung et al. 2010). A significant advantage in studying autophagy regulation in Chlamydomonas is the fact that it is sensitive to rapamycin, in contrast to higher plants that are rapamycin-resistant (Menand et al. 2002; Díaz-Troya et al. 2008b). Working with Chlamydomonas may present the unique opportunity to induce autophagy in photosynthetic organisms using rapamycin. Moreover, a single Chlamydomonas ortholog to the ATG8 gene family of higher plants, CrATG8, has been discovered (Pérez-Pérez et al. 2010). Utilization of the CrATG8 protein as an in vivo marker for various autophagy-associated processes in this algal species looks to be a promising approach for future discoveries of autophagy-associated processes in Chlamydomonas, rendering this organism an emerging new model system for autophagy research.

Rice is a staple monocot crop plant, and the availability of its full-genome sequence renders it an additional important model plant for studying various biological processes including those associated with autophagy. Nevertheless, compared to the relatively significant identification and functional analysis of the Arabidopsis ATG orthologs, elucidation of the core autophagy genes and their functions in rice has, in general, not been extensively studied. As can be deduced from Table 1, based on DNA and protein sequence homology, rice possesses, as expected, all the genetic potential for fully active core autophagy machinery. In addition, expansion of the core ATG genes into gene families occured in rice quite extensively compared to Arabidopsis (Fig. 1). The operation of autophagy in rice suspension culture cells under sucrose starvation has been demonstrated some time ago (Chen et al. 1994). Yet, so far, only four ATG genes have been identified and characterized in rice. Su et al. (2006) had cloned two of the rice ATG4 and ATG8 orthologs based on sequence homology with Arabidopsis ATG4b and ATG8a genes. The OsATG4 protease was also shown to interact with the OsATG8 protein in a yeast-two-hybrid analysis, resulting in the cleavage of the C-terminal tail of OsATG8, implying a functional conservation of the autophagy process in dicotyledonous and monocotyledous plant species (Su et al. 2006). Similarly, Shin and coworkers had cloned two rice orthologs for the Arabidopsis ATG10 genes based on their DNA sequence homology. The two rice ATG10 proteins are 75% identical to each other, but possess only 39–43% protein sequence homology to the single Arabidopsis ATG10 (Shin et al. 2009). Moreover, the C terminal domain of the rice ATG10, especially the cysteine residue that is necessary for binding to ATG12, is conserved in the Arabidopsis and rice homologs (Shintani et al. 1999; Shin et al. 2009).

A rice mutant lacking the ATG10b protein (the more abundantly expressed of the two rice ATG10 proteins) was more sensitive than the wild-type plant to salt and oxidative stresses, had decreased number of autophagosomes and showed accumulation of oxidized proteins (Shin et al. 2009). Similar phenotypes were also demonstrated in autophagy-defective Arabidopsis ATG18a RNA interface (RNAi) plants (Xiong et al. 2007). Arabidopsis atg10 plants were additionally hypersensitive to nitrogen and carbon starvation, initiated senescence more quickly than the wild type and failed to accumulate autophagic bodies inside the vacuole (Phillips et al. 2008), all phenotypes that were not examined in the rice atg10 mutant. Even though different ATG8 proteins were used in these experiments, the fact that both ATG10 and ATG18 are central to the operation of autophagy supports a functional conservation of autophagy in dicotyledonous and monocotyledonous plant species. Interestingly, a functional rice ATG10a homolog does not compensate for the phenotype of the absence of ATG10b, suggesting that these two genes are not functionally redundant in rice (Shin et al. 2009). It is important to note that the autophagy machinery has also been demonstrated in maize, another staple crop plant which was not discussed in this review (Chung et al. 2009).

Playing hide and seek — the missing links in the plant core autophgay machinery

In yeast, 18 out of 35 ATG genes comprise the core autophagic machinery (Meijer et al. 2007; Xie and Klionsky 2007). The Arabidopsis genome has homologs to almost all of the yeast ATG genes, some of which are present in more than one isoform, particularly ATG8 and ATG18 (see discussion below; Table 1 and Fig. 1). Seven of the yeast core ATG genes (ATG1, ATG2, ATG3, ATG13, ATG16, VPS15, VPS34) have only predicted Arabidopsis orthologs that were found on the basis of sequence similarity but were never functionally studied (Hanaoka et al. 2002; Meijer et al. 2007). Interestingly, two well studied and characterized yeast genes of the core autophagy machinery, namely, ATG14 and ATG27, have no known or predicted orthologs in Arabidopsis plants.

The yeast ATG14 is a hydrophilic protein with a coiled-coil motif in its N-terminus. It was shown to be essential for proper autophagy in S. cerevisiae and was also shown to function as a part of an autophagy-specific phosphoinositide 3 (PI3) kinase complex consisting of VPS34, VPS15, ATG6, and ATG14 (Kametaka et al. 1998; Kihara et al. 2001). Basic BLAST analysis could not clearly identify an ATG14 gene homolog in mammals and plants. Yet, a functional mammalian ortholog for yeast ATG14 was independently identified in human and mouse by two specifically dedicated approaches. First, bioinformatic analysis comparing the yeast and Candida glabrata ATG14 protein sequence to the protein databases of mouse and human: the analysis revealed a weak homolog to yeast ATG14 (13% identity and 37% similarity; Itakura et al. 2008). Second, in vivo isolation of Beclin-1 complexes from various mouse tissues and identification of their components using mass spectrometry: this yielded in a mouse ATG14 homolog (Zhong et al. 2009). Furthermore, similarly to yeast, the ATG14-Beclin1-VPS15-VPS34 complex was identified in mammals as an autophagy-specific PI3 kinase complex (Itakura et al. 2008; Sun et al. 2008; Matsunaga et al. 2009, 2010; Zhong et al. 2009). In contrast to yeast and mammalians, no ATG14 homolog has so far been identified in plants. Yet, the important principal role of ATG14 in yeast and its homologs in mammalian cells, which is expected also to operate in plant cells, calls for additional efforts to search for a putative plant protein carrying a comparable function to ATG14 from yeast and mammalian cells.

ATG27 is a second integral membrane protein of the core autophagy machinery (the first is ATG9). It is also a PI3P-binding protein. Similarly to ATG9, the ATG27 protein cycles between two different membrane compartments in yeast cells, the mitochondria and phagophore assembly site (PAS), but its additional localization to the Golgi apparatus suggests the involvement of the Golgi complex itself in the autophagy pathway, perhaps by lipid delivery from the Golgi to the forming double-membrane autophagosome vesicles (Yen et al. 2007; Ohashi and Munro 2010). When ATG27 is depleted in yeast cells, autophagy still occurs, but at a substantially reduced level (Yen et al. 2007). To our knowledge, there is no mammalian or plant ortholog to ATG27. We propose that the absence of an ATG27 ortholog in plant and mammalian genomes makes biological sense as the yeast ATG27 protein is localized to the PAS, a compartment missing in mammalians and plants. In addition, alternative membrane sources other than the Golgi are used for autophagosome assembly in mammalian and plant cells such as the ER (Hayashi-Nishino et al. 2009; Ylä-Anttila et al. 2009). ATG27 may thus be a protein that was specifically evolved in yeast cells to fulfill yeast-specific functions, or, on the contrary, the ancestor ATG27 protein was lost during mammalian and plant evolution.

The multigene families of plant autophagy — ATG8 and ATG18

Two of the proteins belonging to the core autophagy machinery, namely, ATG8 and ATG18, are encoded in higher plants by multigene families. The model plant Arabidopsis possesses nine genes encoding ATG8 isoforms and eight genes encoding ATG18 isoforms (Doelling et al. 2002; Xiong et al. 2005). Mammalian cells possess eight ATG8 genes, but in sharp contrast to higher plants, they possess only two to four genes encoding ATG18 isoforms (Polson et al. 2010; Weidberg et al. 2010). These two gene families, ATG8 and ATG18, were suggested as candidates for functional specification of autophagy in plants (Hayward et al. 2009). The reasons for the existence of these two large gene families in plants is yet unknown, but are apparently related to the multiple functions of these proteins in plants as will be discussed later on during this review.

ATG8

ATG8 is one of two proteins containing a Ub-fold in the autophagy core machinery. It is synthesized as a pro-protein that is cleaved by ATG4 to expose a glycine at the C-terminus of the protein. In a Ub-like conjugation system, the processed ATG8 protein then binds through the exposed glycine to phosphatidylethanolamine (PE) molecules on mvembranes that are programmed to differentiate into autophagosomes (Tanida 2011). ATG8-PE located on the outer membrane of the autophagosome is cleaved off during autophagosome deposition by ATG4. ATG8-PE located on the inner membrane of the autophagosome enters the vacuole with the autophagic body and is degraded (Chung et al. 2010). Yeast possesses one ATG8 gene, that has been shown to be involved in autophagosome formation and membrane expansion, possibly by mediating membrane tethering and hemifusion (Nakatogawa et al. 2007).

In contrast to yeast, mammals possess eight ATG8 isoforms, while Arabidopsis plants possess nine ATG8 isoforms (Doelling et al. 2002; Yoshimoto et al. 2004). The mammalian ATG8 proteins are principally divided into two subfamilies according to protein sequence homology, namely, LC3 subfamily (four members) and GABARAP/GATE-16 subfamily (four members) (Weidberg et al. 2010). The Arabidopsis ATG8 homologs all display high sequence similarity to the yeast ATG8 (between 73% and 90%), and the processed proteins following the C-terminal cleavage by the plant ATG4 proteins contain a glycine residue at their C-terminus (Doelling et al. 2002; Hanaoka et al. 2002). They are divided into three subfamilies according to protein sequence similarity: AtATG8a, AtATG8c, AtATG8d, and AtATG8f (four members); AtATG8b, AtATG8e, and AtATG8g (three members); and AtATG8h and AtATG8i (two members; Doelling et al. 2002). AtATG8h and AtATG8i differ from the other AtATG8s since they lack extra residues at the C-terminus following the glycine residue. The ATG8 Ub-like conjugation system, existing in yeast and mammals, also exists in plants and ATG8 cleavage by the ATG4 protease as well as its lipidation has been demonstrated in many independent studies (Doelling et al. 2002; Yoshimoto et al. 2004; Thompson et al. 2005; Fujioka et al. 2008; Chung et al. 2009, 2010).

A number of reports also implied multiple functions for ATG8 proteins in both mammalians and plants, which appear to be unrelated to the basic process of autophagosome formation and may be differentially executed by the different ATG8 isoforms. The mammalian and plant ATG8 homologs were shown to bind microtubules, implying a connection between autophagy and the cytoskeleton (Mann and Hammarback 1994; Wang et al. 1999; Ketelaar et al. 2004). In addition, the mammalian ATG8 homologs were shown to be connected to intracellular trafficking unrelated to autophagosomes (Wang et al. 1999; Sagiv et al. 2000; Kittler et al. 2001). Work previously published by our group has also demonstrated a connection between overexpression of an Arabidopsis green fluorescent protein (GFP)-ATG8f fusion protein and an altered response to plant hormones, suggesting a role for ATG8 or autophagy in regulation of hormonal signaling (Slavikova et al. 2008). Furthermore, ATG8 in yeast and mammals has been shown to mediate target recognition during selective autophagy by specific binding to protein targets (Noda et al. 2010). This target recognition is discussed further in a later section of this review.

The reason for the presence of such large ATG8 families in mammalians and plants is not entirely clear, but may imply that the different ATG8 isforms possess multiple nonparallel functions in the various biological processes associated with Autophagy. Interestingly, the different mammalian ATG8 subfamilies have been suggested to play nonparallel roles in the different stages of autophagosome formation, namely, the LC3 subfamily being involved in phagophore membrane elongation, while members of the GABARAP/GATE-16 subfamily take part in autophagosome maturation (Weidberg et al. 2010). In Arabidopsis, previous studies conducted by our group as well as by others have shown differential expression of diverse AtATG8 homologs in various plant organs and plant tissues (Yoshimoto et al. 2004; Sláviková et al. 2005; Thompson et al. 2005). This nonuniform expression pattern may suggest specific roles for each AtATG8 homolog. Interestingly, although expression of all the Arabidopsis ATG8 homologs is induced by nitrogen starvation, each homolog shows a different pattern of induction (Yoshimoto et al. 2004). This may suggest discrete roles in autophagosome formation for the various AtATG8 proteins, as previously seen for mammalian ATG8 homologs (Weidberg et al. 2010). To test the expression patterns of the different Arabidopsis AtATG8 isoforms in response to various stresses, hormonal and nutritional cues, we used the public microarray database located in the Nottingham Arabidopsis Stock Centre (http://affymetrix.arabidopsis.info/AffyWatch.html). The gene expression of the different AtATG8 isoforms is differentially induced in response to the various cues, implying the specific involvement of autophagy, or the different AtATG8 isoforms, in the various conditions. In addition, only four of the nine ATG8 homologs displayed a considerable change in expression over the cues examined (Fig. 2A). The lack of change in expression in five other ATG8 homologs can be explained either by a lack of response of these specific homologs to the treatments examined in the database or by the relatively limited expression of these homologs in specific tissues or developmental stages. It is important to note that the homologs that showed marked differences in expression belong to all three AtATG8 subfamilies, so the ability to respond to stresses at the level of gene expression is not limited to one subfamily. It can also be seen in Fig. 2 that the expression pattern varies between the homologs in both regulation type (induction or repression) and also in the actual level of induction of the genes. These results may imply differential functions of the various ATG8 proteins in plants, similarly to what has been observed in mammals.

Fig. 2
figure 2

Variable expression pattern of AtATG8 and AtATG18 family members under various treatments using a database of publically available microarray data (Nottingham Arabidopsis Stock Centre, http://affymetrix.arabidopsis.info/AffyWatch.html). We analyzed the expression levels of the AtATG8 (A) and AtATG18 (B) homologs. The treatments displayed are ones in which at least one of genes tested demonstrated a difference in expression. The AtATG8 homologs displayed are the ones who showed a massive change in expression. AtATG18b does not have a probset and therefore was not included in the analysis

Full size image

ATG18

Similarly to ATG8, yeast possesses only one ATG18 encoding gene. The yeast ATG18 is a WD-40 repeat containing protein, which specifically recognizes PI3P (Tooze and Yoshimori 2010). Yeast contains two more WD repeat proteins, ATG21 and Ygr223c, that together, with ATG18, make up a family of three PI-binding proteins. WD repeat protein-interacting phosphoinositides 1 and 2 (WIPI1 and WIPI2, respectively) are both mammalian orthologs proteins of the yeast ATG18 (Polson et al. 2010). Both WIPI proteins were shown to be involved in the initiation of autophagy in mammalian cells (Burman and Ktistakis 2010; Polson et al. 2010).

Like ATG8, ATG18 has undergone diversification in plants. In Arabidopsis, there is an eight member family of ATG18 genes with multiple splice variants named ATG18 a-to-h (Xiong et al. 2005; Bassham et al. 2006). Those genes were divided to three major subgroups: AtATG18a, AtATG18c, AtATG18d, and AtATG18e are surprisingly more similar to the yeast Ygr223c gene; AtATG18b is the sole gene most similar to yeast ATG18 itself; and AtATG18f, AtATG18g, and AtATG18h are more divergent, forming a less conserved subgroup with longer peptide sequences (Xiong et al. 2005). Also, plant ATG18 genes are characterized by a great variety of expression patterns upon exposure to stress conditions (Hayward et al. 2009). Very few studies have tried to elucidate the functions of the different ATG18 isoforms in Arabidopsis. Xiong and coworkers (2005) showed significant differences in the expression of AtATG18 genes in response to various treatments and in different plant organs. Apparently, AtATG18a, which is closer in sequence homology to the yeast Ygr223c than to yeast ATG18, was the only isoform that showed increased transcription in response to both starvation and senescence and was also detected in all plant organs (Xiong et al. 2005). Later studies showed that AtATG18a is upregulated in response to oxidative, salt and drought stresses as well (Xiong et al. 2007; Liu et al. 2009). The rest of the AtATG18 genes were involved either in starvation or in senescence or were partially expressed in the different plant organs. The AtATG18e isoform was not detected in any of the plant organs or in the stress conditions tested. The changes in expression of the various AtATG18 genes in response to various treatments and in different plant organs did not follow the subgroup division made according to sequence homology, suggesting that in the ATG18 gene family, sequence similarity does not necessarily correspond to protein functions. The Bassham group generated unique AtATG18a RNAi plants with reduced expression levels of AtATG18a gene (Xiong et al. 2005). Surprisingly, in spite of the high sequence similarity with other members of the family, the reduction in AtATG18a transcript level was gene-specific and did not affect the transcript level of closer or farther homologs (Xiong et al. 2005). In addition, despite the extensive number of AtATG18 isoforms, AtATG18a RNAi plants displayed a similar phenotype to classical ATG knockout plants (Xiong et al. 2005), as well as higher sensitivity to oxidative stress conditions and higher production of oxidized proteins and reactive oxygen species (ROS) then wild-type plants (Xiong et al. 2007). AtATG18a RNAi plants exhibited also higher sensitivity to salt and osmotic/drought conditions than wild-type plants (Liu et al. 2009). At the cellular level, The AtATG18a RNAi plants had no detectable autophagosomes, during nitrogen and carbon starvation (Xiong et al. 2005) oxidative stress (Xiong et al. 2007) and salt and osmotic stresses (Liu et al. 2009). These results indicate that AtATG18a is not functionally redundant with other members of the ATG18 family and is critical for autophagosome formation under different starvation and stress conditions in Arabidopsis. It will be interesting to test whether other AtATG18 isoforms also possess isoform-specific functions. The AtATG18f, AtATG18g, and AtATG18h genes, belonging to a divergent subgroup in the ATG18 family, were all upregulated during nutrient starvation but had some differences in expression under various abiotic stresses (Rizhsky et al. 2004).

In order to get some clues concerning specialized functions of different ATG18 isoforms, we analyzed the gene expression of ATG18 genes from Arabidopsis under various treatments using data taken from Nottingham Arabidopsis Stock Center in a similar way to the analysis performed on the ATG8 gene family (see above). The genes were divided to three subgroups according to (Xiong et al. 2005), but due to a missing probeset of AtATG18b, only two subgroups of genes were analyzed (Fig. 2B). Our expression analysis together with previous works suggest that different AtATG18 isoforms have specialized in plants to participate in different processes that involve autophagy activation, some of them probably plant-specific processes, which are triggered by different stimuli. Those processes can be different in time and space, because diverse AtATG18 gene isoforms were expressed differently in various plant tissues (Xiong et al. 2005). Interestingly, the AtATG18e isoform was not detected at all by real-time polymerase chain reaction (RT-PCR) in any of the organs/treatments checked (Xiong et al. 2005), but public microarray data showed very low but noticeable changes in the level of expression under certain conditions. AtATG18e is the closest homolog to AtATG18a, the most responsive AtATG18 homolog. It is therefore possible that AtATG18e either serves a very specific function or is expressed in discrete tissues or developmental stages. AtATG18a is the only well-studied isoform in plants whose absence shows clear phenotype and, as far as the experimental data shows, is the closest plant ortholog to the yeast ATG18 and the mammalian WIPI2 protein. Elucidating the physiological effects of the absence of the rest of the AtATG18 gene family on the plant phenotype is a challenging issue for future research.

In contrast to the case of AtATG8 proteins, so far, no GFP-tagged AtATG18 proteins have been employed to elucidate the intracellular localization of these proteins and to search for other plant proteins that might be colocalized with them. Such studies may aid in better understanding of the compound cellular and physiological roles of ATG18 in plants.

The regulatory role of ATG8 in the selective degradation of specific cellular components —ATG8-interacting motif (AIM)

The operation of ATG8 in the selective degradation of various protein substrates requires specific interaction of ATG8 with these proteins, and the way ATG8 executes this function has only recently been addressed. Noda et al. (2008) identified by structural analysis a common tetra-peptide motif “WXXL” (X is any amino acid) in two totally different proteins, the yeast ATG19 and the mammalian sequestosome 1 (p62/SQSTM1), previously shown to interact with ATG8 and LC3, respectively (Shintani et al. 2002; Pankiv et al. 2007;). This original ATG8-interacting motif (AIM) was located within the longer (22 amino acids) LC3-interacting region (LIR) identified in the p62/SQSTM1 protein (Pankiv 2007). Later, three additional proteins, namely, the yeast ATG32 and the mammalian neighbor of BRCA1 (NBR1) and Nix, were also identified as ATG8-binding proteins in different studies. Those proteins also contained a motif similar to AIM with some minor alterations (Kirkin et al. 2009a; Okamoto et al. 2009; Schwarten et al. 2009). This observation triggered a research to identify additional AIM-containing proteins (Behrends et al. 2010; Pankiv et al. 2010; Johansen and Lamark 2011). AIMs are evolutionary conserved among the large family of ATG8-binding proteins in various species (Noda et al. 2010). The initial longer definition of the AIM was X−1X−2X−3W0X1X2LX3 (Noda et al. 2008). Later studies have shown relative “flexibility and demands” needed for functional AIMs (Noda et al. 2010; Johansen and Lamark 2011). In light of the fast progress in the field of AIM study, one can speculate that more flexibility in the AIM will be added in parallel to the discovery of new ATG8-binding proteins. This quite extensive flexibility of the AIM makes the use of this predictive approach quite challenging in the identification of novel ATG8-interacting proteins. Thus, in the discovery of novel ATG8-interacting proteins, biochemical or genetic methods generally precede the identification of an AIM in the target protein sequence (Behrends et al. 2010; Vanhee et al. 2011). The interaction between autophagic receptors and ATG8-family proteins through AIMs contributes, at least partially, to the selection of specific cargoes, possibly by linking the cargoes to autophagic membranes and/or to their forming machineries (Noda et al. 2010). In some cases, aggregation/polymerization is also required in addition to the AIM-dependent binding (Johansen and Lamark 2011). Functional AIMs have been found in various cellular proteins such as calreticulin and clathrin heavy chain (Mohrlüder et al. 2007a, b) as well as ATG8 modifying enzymes belonging to the core autophagy machinery (Yamada et al. 2007; Satoo et al. 2009; Yamaguchi et al. 2010), suggesting that ATG8 binding is involved in many biological processes.

Despite the extensive identification of AIM-containing ATG8-binding proteins in mammals and yeast, there are so far only two recent publications on AIM-containing ATG8-binding proteins in plants (Svenning et al. 2011; Vanhee et al. 2011). The first has identified an Arabidopsis AtNBR1 homolog as a selective autophagy substrate (Svenning et al. 2011). The AtNBR1 was also capable of binding Ub (see the section discussing autophagy and the Ub system). The second publication has shown that the Arabidopsis tryptophan-rich sensory protein (TSPO), a sensory protein localized to the cell plasma membrane, possesses an AIM and is degraded through selective autophagy and not through the proteasome pathway (Vanhee et al. 2011). Interestingly, TSPO binds heme, which is a biologically important metabolite, but also a potentially toxic compound that can generate ROS. Thus, under physiological conditions (such as abiotic stresses) in which heme may have toxic effects, TSPO binds to heme and targets it for selective degradation in the vacuoles by using the ATG8-mediated selective autophagy process (Hofmann 2011). Our laboratory (unpublished results) has recently identified a number of other Arabidopsis plant-specific proteins that interact with ATG8 and possess AIMs, implying that plant ATG8 proteins specifically interact with multiple other plant proteins, containing AIMs, which apparently possess multiple biological functions, some of which appear to be plant-specific.

Is there a connection between autophagy and the Ub machinery in plants?

The Ub protein is a central part of the proteasomal degradation machinery. The covalent attachment of a Ub tag to a substrate protein generally leads to its degradation by the proteasome (Clague and Urbé 2010). Yet, Ub has also been shown to participate in various other cellular processes such as endocytosis, signal transduction, and DNA repair (Kirkin et al. 2009b). In the past few years, emerging evidence from mammalian cells has demonstrated an additional role for Ub in targeting proteins for degradation by selective autophagy (reviewed in Kirkin et al. 2009b; Clague and Urbé 2010; Lamark and Johansen 2010). The connection between Ub and autophagy is facilitated by so-called “adapter” proteins able to bind Ub but also capable of binding proteins in the autophagy machinery (Johansen and Lamark 2011). For example, the protein p62/SQSTM1 and its interaction partner NBR1 were shown to be involved in the degradation of protein aggregates by binding to both poly-Ub chains and LC3. This binding of LC3 occurs through an AIM located in these proteins (Bjørkøy et al. 2005; Pankiv et al. 2007; Kirkin et al. 2009a).

Even though a connection between the ubiqutination and the autophagy machineries has not yet been reported in plants, the conserved pattern of operation of both these systems in plants implies that an analogous connection between autophagy and Ub may also exist in plants. To get some clues about this possibility, we conducted a coexpression analysis of the ATG8 gene family and the ATG18 gene family of Arabidopsis, using the ATTED-II database (http://atted.jp). This analysis showed that genes of the plant autophagy machinery are coexpressed with genes belonging to the plant Ub conjugation system, mainly E3 genes (data not shown). This data may suggest the plant autophagy and ubiquitination machineries that interact with each other. It should indeed prove fascinating to elucidate the ubiquitinated targets of selective autophagy in plants and also the Ub-binding proteins that connect the targets to the autophagy machinery. While NBR1 homologs are found throughout the eukaryotic kingdom including plants (see above), p62/SQSTM1 homologs are confined to the metazoans (Svenning et al. 2011). Interestingly, the Arabidopsis AtNBR1 contains similar functional domains of the mammalian p62/SQSTM1 and hence, AtNBR1 was suggested to be a functional hybrid of the mammalian autophagic adapters p62/SQSTM1 and NBR1 (Svenning et al. 2011). Deciphering the interplay between Ub and autophagy in plants may shed light on the likely existing multiple protein targets of selective autophagy in plants.

Autophagy in seed development and germination

Although the autophagy machinery is relatively conserved between plants, yeast, and mammals, plants, being sessile organisms that are highly sensitive to changing environments, may be expected to possess many unique features that result in dedicated specializations of the autophagy machinery. For instance, to ensure survival, developing seeds of a many dicotyledonuous and monocotyledonous plant species efficiently synthesize massive amounts of storage proteins and deposit them as protein bodies inside protein storage vacuoles (PSVs), Then upon early germination, these storage proteins are degraded to enable the accumulation of sufficient energy and amino acids to synthesize new plant organs and to commence photosynthesis. During early stages of seed development, the seed storage proteins are transported to the PSVs via the Golgi (Robinson et al. 1997; Robinson and Hinz 1999; Hillmer et al. 2001; Winter and Stoger 2011). Yet, later, during seed development, when massive amounts of storage proteins are synthesized, they are transported directly from the ER to the PSVs by an intracellular process that resembles autophagy in its microscopic appearance (reviewed by Galili et al. 1993; Robinson et al. 1998; Herman and Larkins 1999; Chrispeels and Herman 2000; Bassham 2002). This process is also associated with enhanced expression of genes encoding ATG proteins of the core autophagy machinery. Our previous seed gene expression data (Angelovici et al. 2009) showed a coordinated upregulation of many of the core Arabidopsis ATG genes during seed maturation and desiccation (Fig. 3). The highest expression level is observed in dry seeds.

Fig. 3
figure 3

Expression of depicted Arabidopsis ATG genes during seed development and early germination. The data was taken from microarray analysis performed by our group (Angelovici et al. 2009). DAF Days after flowering, DRY dry seeds, imbibed 3 days imbibition at 4°C in the dark, 1D germ 1 day under long-day conditions after imbibition

Full size image

During early germination, plant seeds synthesize proteases and efficiently target them from the ER to the vacuole to mobilize the storage proteins that were accumulated during seed development (Toyooka et al. 2000). Interestingly, some of these vacuolar proteases that degrade storage proteins inside the vacuole posses a C-terminal K/HDEL signal sequence that generally functions in the retention of ER resident proteins within the ER. Moreover, removal of the K/HDEL signal causes the transport of these proteases to the plasma membrane (Toyooka et al. 2000, 2001), implying that this signal possesses a novel plant-specific regulatory role in the massive transport of the proteases directly from the ER to the vacuole, bypassing the Golgi apparatus. In addition, upon the transport of these newly synthesized proteases to the vacuole, their C-terminus appears to be removed by a yet uncharacterized process (Okamoto et al. 2001). This unique mechanism of the transport of K/HDEL-containing proteins from the ER directly to the lytic vacuole has so far not been demonstrated in mammalian cells. The mechanism mediating this transport is still not clear. A simple likely explanation is that the PSVs develop directly out of the ER, although other routes cannot be rulled out. It will also be interesting to test whether the autophagy machinery is involved in this transport process.

Recently, the delivery of different prolamin storage proteins to the storage vacuole of aleurone cells in developing maize seeds was suggested to be associated with an atypical autophagic process, that delivers ER material directly to PSVs independently of ATG8 lipidation (Reyes et al. 2011). The authors identified autophagosome-like structures involved in transport to PSVs in maize aleurone cells. However, these structures were neither surrounded by a double-membrane nor decorated by ATG8, preventing their identification as classical autophagosomes. The authors suggested that these structures might be a specialized case of ATG8-independent selective autophagy used to deliver storage protein aggregates from ER to the storage vacuole, a hypothesis that is supported by the recent identification of an analogous ATG8-independent subcellular route in mammalian cells (Nishida et al. 2009). Additional research is needed to clearly link the above findings to the autophagy process and to elucidate whether it is a unique delivery mechanism of storage proteins in developing and germinating seeds of various plant species. The generation of atg mutants in cereal seeds in the future will be an important advance in solving these challenging issues (Chung et al. 2009).

Future prospects

In this review, we attempted to summarize the recent knowledge regarding plant autophagy while addressing issues that have not been extensively reviewed previously and that we believe will take center stage in plant autophagy research in the coming years. In order to promote the plant autophagy field, we believe that the following approaches should be taken. First, the marker collection available for cellular studies of plant autophagy should be widened. The current markers are solely based on different ATG8 isoforms tagged with GFP (Yoshimoto et al. 2004; Sláviková et al. 2005; Thompson et al. 2005; Xiong et al. 2007), while in yeast and mammalian cells, many more ATG markers are used (see, for example, Suzuki et al. 2001; Krick et al. 2008; Gao et al. 2010), enabling a better understanding of the cellular processes involving autophagy. Second, the pool of available autophagy mutants should be widened, preferably generating mutants involved in autophagy initiation, which, according to recent opinion (Hayward and Dinesh-Kumar 2010), are predicted to display a more severe developmental phenotype that the current atg mutants. Third, the autophagy machinery in species other than Arabidopsis should be investigated. Emerging data from the last years regarding Chlamydomonas autopgahy genes (Díaz-Troya et al. 2008b; Pérez-Pérez et al. 2010) render it a novel model organism for studying the cellular autophagy mechanism in photosynthetic organisms. In addition, studying autophagy in crop plants such as rice may pave the way for the utilization of autophagy for agricultural purposes. Fourth, the specific functions of the different family members in the multiple-member gene families (ATG8, ATG18) should be elucidated. This can be achieved by either testing for specific treatments or analyzing specific plant organs and tissues. Fifth, plant-specific autophagy-associated proteins using the core autophagy proteins as baits should be searched. This may help elucidate plant-specific targets of selective autophagy.Sixth, autophagy in plant-specific organs and organelles must be investigated. Recent findings regarding sequestration of chloroplasts for degradation (Ishida et al. 2008; Wada et al. 2009) demonstrate an exciting plant-unique feature of autophagy. Finally, autophagy during seed development and germination still requires further studies. The great advances made in plant autophagy research in the last years could be utilized to better investigate these issues.