Introduction

Eukaryotes need elaborate cellular systems to maintain homeostasis of various biomolecules. Such systems include gene expression machinery, membrane transport network, and degradation/recycling mechanisms. Autophagy is a degradation mechanism that delivers cytoplasmic materials to lytic compartments (i.e., lysosomes in metazoans and vacuoles in yeast and plant cells). In plant cells, two types of autophagy are known, macroautophagy (Fig. 1a) and microautophagy (Fig. 1b), where the cytoplasm is sequestered by the autophagosome and vacuole, respectively (Avin-Wittenberg, et al. 2018). Thus, autophagy can be considered a membrane trafficking route to vacuoles or lysosomes.

Fig. 1
figure 1

Molecules mediating membrane trafficking during autophagy in budding yeast. a During macroautophagy, core ATG proteins are recruited to the phagophore assembly site (PAS) located between the ER and vacuole. Upon receiving an autophagy induction signal, the ATG1 kinase phosphorylates ATG9 and PI3K, driving phagophore nucleation. During phagophore expansion, ATG2 transfers membrane lipids from the ER to the phagophore, while ATG8 is attached to the phagophore membrane by a lipidation reaction that is mediated by the ATG8/12 conjugation system (see Table 1). After ESCRT-III seals the opening of the phagophore, the autophagosome with double membrane forms and matures. The trans-SNARE complex, Rab7, and tethering factors like the HOPS complex mediate autophagosome-vacuole fusion. Further investigation is needed to elucidate the exact functions of COPII vesicles and actin filaments in phagophore nucleation and other steps during autophagosome formation. Genetic and cell biological data using Arabidopsis indicates that macroautophagy in plants is mediated by similar molecules (see main text). However, extensive localization of core ATG proteins at the PAS has not been tested in plant cells. b Microautophagy does not involve autophagosome formation. Rather, the vacuolar membrane is invaginated or protruded to sequester a portion of the cytoplasm. The ESCRT-III is involved in vacuolar membrane sealing, at least during selective microautophagy for the ER. The role of individual ATG proteins in microautophagy appears to depend on specific types of microautophagy

This mini-review will focus on macroautophagy (hereinafter, autophagy) because molecules responsible for macroautophagy are defined better than those for microautophagy. Autophagosome formation begins with the nucleation of an autophagosome precursor termed the phagophore (Fig. 1a). The phagophore expands and grows into a cup-shape membrane structure to sequester a portion of the cytoplasm. The phagophore seals its narrow neck by membrane scission, followed by the double-membrane autophagosome maturing and subsequently fusing with a vacuole or lysosome. Autophagic cargo enclosed by the inner autophagosomal membrane is released into the vacuolar lumen as an autophagic body, which is rapidly degraded by resident hydrolases (Fig. 1a). Yeast, metazoans, and plants harbor conserved sets of core Autophagy-related (Atg) genes (Table 1). The products of these genes are involved in multiple steps during autophagosome formation and turnover (Fig. 1a).

Table 1 Core ATG genes in budding yeast, mammals, and Arabidopsis

Genetic analysis of core atg mutants in several plant species has helped define functions of autophagy in numerous biological processes, such as nutrient recycling, responses to biotic and abiotic stress, and plant development (Shinozaki, et al. 2020; reviewed by Avin-Wittenberg, et al. 2018; Liao and Bassham 2020; Marshall and Vierstra 2018; Su, et al. 2020). Nevertheless, progress in defining plant autophagosome dynamics is relatively slow, compared to extensive autophagosome-related studies using yeast and mammalian cells. For this reason, plant models of autophagosome dynamics are largely based on yeast and mammalian models (Chung 2019; Zhuang, et al. 2018). This mini-review updates our model of plant autophagosome dynamics, focusing on recent studies using Arabidopsis and yeast.

Nucleation of the Phagophore

A master regulator of phagophore nucleation is the protein kinase Atg1. In yeast, the Atg1 complex is critical for organizing the phagophore assembly site (PAS), where core ATG proteins are recruited during autophagosome formation (Fujioka et al. 2020) (Fig. 1a). The functions and composition of Atg1 complexes are well defined in mammals and yeast; however, the composition of plant ATG1 kinase complexes is poorly understood and is likely similar to that of its mammalian counterparts (Table 1). Most plant ATG1 homologs have catalytic and regulatory domains. For example, the Arabidopsis thaliana genome contains four genes encoding ATG1, one of which encodes the truncated ATG1 isoform ATG1t, which lacks a regulatory domain. Protein interaction data suggest that the ATG1 complex in Arabidopsis consists of the ATG1, ATG13, ATG11, and ATG101 proteins (Li et al. 2014; Suttangkakul et al. 2011). Autophagic flux is severely inhibited in atg13a atg13b double mutants plus atg11 single mutants. Interestingly, these mutants and atg1abct quadruple mutants (Huang et al. 2019) were more tolerant of fixed carbon starvation than atg7 and atg5 mutants. This phenotypic difference may be explained by an adaptive role of ATG1-independent autophagy, which is activated during prolonged carbon starvation and requires phosphatidylinositol 3-kinase (PI3K) and SnRK1 activities (Huang et al. 2019).

Plants are similar to yeast and mammals with regard to harboring conserved components of PI3K complexes I and II, which generate phosphatidylinositol 3-phosphate (PI3P) at the membrane of the phagophore and endosome, respectively (Table 1). VPS34, VPS15, and ATG6/VPS30 are found in both complexes, whereas ATG14 and VPS38 are specific components of PI3K complexes I and II, respectively. No viable homozygous knockout mutants have been reported for common components, which indicates the essential roles of PI3P production in plant development (Fujiki et al. 2007; Harrison-Lowe and Olsen 2008; Lee et al. 2008). In contrast, viable vps38 mutants showed defects in endosomal trafficking, but only a minor change in autophagic flux, suggesting that functions of distinct PI3K complexes are conserved between yeast and plants (Lee et al. 2018; Liu et al. 2018). More work on putative ATG14 homologs in plants will clarify their contribution to autophagosome biogenesis. A further discussion about PI3P and its effectors involved in plant autophagy is provided in a recent review (Chung 2019).

Vesicles containing ATG9, a multi-spanning transmembrane protein, are proposed to provide membrane precursors for phagophore formation in yeast and mammals (Karanasios et al. 2016; Mari et al. 2010; Orsi et al. 2012; Yamamoto et al. 2012). Similarly, Arabidopsis atg9 mutants showed reduced autophagy (Hanaoka et al. 2002; Shin et al. 2014), and Arabidopsis ATG9 appears to localize at the Golgi apparatus and endosomes, and transiently interact with the phagophore (Zhuang et al. 2017). A recently reported three-dimensional structure of Arabidopsis ATG9 (Lai et al. 2020) indicates the formation of a trimer complex via multiple interfaces, with the C-terminal region playing a self-interaction role.

Expansion and Sealing of the Phagophore

In yeast and mammals, Atg2 homologs form complexes with specific ATG18 homologs (Bakula et al. 2017; Suzuki et al. 2007). Yeast Atg2 and Atg18 are located at the phagophore edge, and the mammalian homolog Atg2A is located on the phagophore and lipid droplets (Suzuki et al. 2013; Velikkakath et al. 2012). Simultaneous silencing of mammalian Atg2A and Atg2B resulted in the accumulation of unclosed autophagic vesicles in the cytoplasm, which is consistent with their role in phagophore expansion. Indeed, studies using yeast and mammalian ATG2 homologs demonstrated that ATG2 tethers the ER to the phagophore via an interaction with PI3P-binding ATG18 homologs (Chowdhury et al. 2018; Gomez-Sanchez et al. 2018; Kotani et al. 2018) and possess an in vitro phospholipid transfer activity (Maeda et al. 2019; Osawa et al. 2019, 2020; Osawa and Noda 2019; Valverde et al. 2019). Thus, the ATG2-ATG18 complex contributes to autophagosome biogenesis by providing membrane lipids from the ER to the phagophore (Osawa and Noda 2019). Supporting this hypothesis (Fig. 2a), de novo synthesis of fatty acids near the ER was shown to drive phagophore expansion in yeast (Schutter et al. 2020).

Fig. 2
figure 2

ATG2 homologs supply membrane lipids during phagophore expansion by direct lipid transfer. a A model for membrane lipid transfer by the ATG2-ATG18 complex. The arrow indicates the directional flow of phospholipids. The asterisks represent phosphate groups on PI3P. b Conserved domains in human, budding yeast, fission yeast, and Arabidopsis ATG2 homologs. Two N-terminal domains, especially Chorein_N, of yeast and mammalian Atg2 are important for ER localization and phospholipid transfer activity. The ATG2_CAD domain of human ATG2A was proposed to interact with WIPI4/ATG18 (Chowdhury et al. 2018). The VPS13_C domain (also called the C-terminal localization region) of yeast Atg2 contains an amphipathic helix that is responsible for localization at the PAS (Kotani et al. 2018). Domains were identified using HHpred (https://toolkit.tuebingen.mpg.de/tools/hhpred) (Zimmermann et al. 2018)

ATG2 and ATG18 are largely conserved among eukaryotes, including plants (Table 1). Arabidopsis ATG2 is encoded by a single-copy gene (Inoue et al. 2006), whereas there are eight Arabidopsis genes encoding proteins with ATG18-like WD repeats (Xiong et al. 2005). Arabidopsis ATG2 possesses several domains conserved in yeast and mammalian homologs (Fig. 2b), an indication that these homologs likely function as lipid transfer proteins. Both atg2 and atg18a mutants exhibit accelerated leaf senescence or hyper-sensitivity to starvation, similar to atg5 or atg7 (Inoue et al. 2006; Xiong et al. 2005). However, autophagic vesicles accumulate only in atg2 and atg18a (Kang et al. 2018), suggesting that Arabidopsis ATG2 and ATG18A are also involved in phagophore expansion.

The autophagic membrane in yeast contains relatively few membrane proteins (Baba et al. 1995). ATG8 is a ubiquitin-fold protein that is lipidated to an autophagic membrane by the sequential action of the ATG8 conjugation system (Table 1). First, a glycine residue near the C-terminus of ATG8 is exposed by the ATG8-specific protease ATG4. The E1-like enzyme ATG7 activates ATG8, which is transferred to the E2-like enzyme ATG3 and finally lipidated to phosphatidylethanolamine on the phagophore membrane. ATG12 is another ubiquitin-fold protein that is conjugated to ATG5 via ATG7 and the E2-like enzyme ATG10. The ATG12-ATG5 conjugate then interacts with ATG16, forming a hexameric complex that enhances ATG8 lipidation. The ATG8/12 conjugation pathway is highly conserved across various eukaryotes including plants (Fujioka et al. 2008). Mutations in the conjugation system result in impaired nutrient recycling and accelerated senescence (Chung et al. 2010; Doelling et al. 2002; Phillips et al. 2008; Thompson et al. 2005; Yoshimoto et al. 2004).

ATG8 and ATG5 are also important markers for monitoring autophagy in Arabidopsis. Cytoplasmic puncta of ATG8 fluorescent fusions have been interpreted as either phagophores or autophagosomes, whereas autophagic bodies can be observed in the vacuolar lumen if transgenic plants expressing the fusions are treated with concanamycin A, an inhibitor of vacuolar proton pumps (Thompson et al. 2005; Yoshimoto et al. 2004). In contrast, the cytoplasmic puncta of ATG5 fluorescent fusions represent the phagophore, but not a mature autophagosome (Le Bars et al. 2014), because neither ATG12 nor ATG5 fusions are targeted to the vacuole (Chung et al. 2010; Le Bars et al. 2014).

Sequestration of cytoplasmic constituents including organelles is accomplished when the phagophore is sealed (Fig. 1a). Recent studies implicated the Endosomal Sorting Complex Required for Transport (ESCRT) in phagophore sealing. ESCRT proteins are involved in inverse membrane involution (also called reverse-topology membrane scission), which takes place during various membrane remodeling events, such as intraluminal vesicle formation in the multivesicular endosome, plasma membrane repair, nuclear envelope repair, virus budding, cytokinesis, and autophagy (Vietri et al. 2020). Central to ESCRT functions are the ESCRT-III subcomplex and its accessory proteins, including the ATPase Vps4. Compartment-specific targeting factors, other ESCRT subcomplexes, and their accessory factors recruit ESCRT-III to a membrane remodeling site, where ESCRT-III and VPS4 eventually drive reverse-topology membrane scission. Genetic null alleles, or the silencing of genes encoding ESCRT components, have been found to accumulate autophagic vesicles (Filimonenko et al. 2007; Lee et al. 2007; Rusten et al. 2007). Recent studies using mammalian cells and yeast demonstrated that unclosed phagophores accumulated in ESCRT-III mutants under autophagy-inducing conditions, indicating that ESCRT has a pivotal role in phagophore closure (Takahashi et al. 2018, 2019; Zhen et al. 2019; Zhou et al. 2019). ESCRT-III is also involved in vacuole membrane scission during selective microautophagy for the ER (Loi et al. 2019; Schafer et al. 2020). How ESCRT-III is recruited to the site of phagophore sealing remains unclear, but a possible regulator of this process is Vps21, a Rab5 GTPase homolog in yeast. Vps21 is important for phagophore sealing (Zhou et al. 2017) and acts upstream of ESCRT-III (Zhou et al. 2019).

ESCRT machineries are largely conserved in eukaryotes, including plants. Several molecular mechanisms were proposed to explain the crosstalk between autophagy and endosomal trafficking, especially ESCRT-mediated endosomal sorting, in plant cells (Chung 2019; Cui et al. 2018; Kalinowska and Isono 2018; Zhuang et al. 2015). These reviews provide additional information; however, considering the aforementioned findings from yeast and mammalian studies, examining whether plant-specific PI3P effectors such as FREE1/FYVE1 (Gao et al. 2015) and CFS1/FYVE2 (Sutipatanasomboon et al. 2017) function as ESCRT-targeting factors during phagophore sealing will be interesting.

Maturation and Transport of the Autophagosome: Any Roles for Cytoskeletons?

After phagophore sealing, the autophagosome undergoes maturation, by which the autophagosome releases most ATG proteins and becomes capable of fusing with the lysosome/vacuole (Reggiori and Ungermann 2017). Although autophagosome maturation in yeast and mammals requires both PI3P turnover and ATG8, it is not known whether the same is true for plant autophagosomes (Chung 2019).

Depending on cell types, autophagosomes may rely on cytoskeletons and motor proteins to find their fusion partners. Mammalian cells have a large number of small lysosomes and during starvation, their intracellular positions can change from the cortical to perinuclear regions, where they are fused with autophagosomes that randomly form at the cell peripheries and are transported after maturation (Jahreiss et al. 2008; Korolchuk et al. 2011). Because FYCO1 binds to LC3, PI3P, and Rab7, it has been proposed to connect the autophagosome to a microtubule, as FYCO1 depletion has been found to result in the perinuclear clustering of autophagosomes (Pankiv et al. 2010). In yeast, microtubules and actin filaments are not required for bulk autophagy. Normal autophagy was observed in tubulin mutants and in cells treated with microtubule-depolymerizing drugs (Kirisako et al. 1999). Similar results were obtained from studies on actin filaments (Reggiori et al. 2005). This is not surprising since the PAS in yeast is adjacent to the vacuole (Fig. 1a), which is the fusion partner of the autophagosome. However, yeast cytoskeletons play more obvious roles in selective autophagy. For example, Atg11 acts as a scaffold for Atg9 and other core ATG proteins and interacts with autophagic cargo during the initiation of selective autophagy. Atg9 delivery to the PAS requires Atg11 and actin filaments (Reggiori et al. 2005), supporting a role for actin filaments in cargo sequestration into the autophagic membrane and/or transport to the PAS.

In mature plant cells containing a large lytic vacuole, the autophagosome may meet the vacuole without traveling a great distance. With regard to this possibility, downregulation of tobacco tubulin gene expression was found to have little effect on basal autophagy (Wang et al. 2015). Actin filaments in tobacco are also dispensable for basal and induced autophagy because autophagosome abundance and autophagic flux were not affected by a short-term disruption of actin filaments (Zheng et al. 2019). Prolonged exposure to anti-microfilament drugs or the genetic inhibition of actin gene expression has led to the activation of autophagy that targets ER components for vacuolar degradation (Zheng et al. 2019).

Although autophagy in plant cells does not strictly require microtubules and actin filaments, both types of cytoskeletons may contribute to autophagosome formation. The number of autophagic vesicles is reduced by tubulin gene silencing and by anti-microtubule drugs (Wang et al. 2015), suggesting a role of microtubules in autophagosome biogenesis. If microtubules were mainly involved in the transport of the autophagosome, autophagosomes would have accumulated. Because tobacco tubulin interacts with Atg6, it was proposed that this interaction enables recruitment of an ATG6-containing complex to the site of phagophore formation (Wang et al. 2015). A separate research group found that Arabidopsis NAP1, a component of the actin nucleation-promoting factor complex, was detected as puncta near the ER approximately 9 min after mechanical stress. Interestingly, YFP-ATG8 signal was detected on the NAP1-GFP puncta 20 min after the stress (Wang et al. 2016). Arabidopsis Pan1 homologs are another type of actin nucleation regulators that localize autophagic vesicles and are transported to the vacuole in an ATG7-dependent manner (Wang et al. 2019). Fewer autophagosomes were observed in the nap1 (Wang et al. 2016) and Pan1 RNAi lines (Wang et al. 2019), underscoring the role of branched actin filaments in autophagosome formation. Whether actin filaments have additional roles in plant autophagosome dynamics, and how actin filaments interact and cooperate with core ATG machinery, will be important questions for future research.

Fusion of the Autophagosome

Fusion of a mature autophagosome with a lysosome or vacuole is mediated by soluble N-ethylmaleimide-sensitive-factor attachment protein receptors (SNAREs), the small GTPase RAB7, RAB7 effectors, and ATG8 proteins (Kriegenburg et al. 2018). The SNARE proteins contain one or two coiled-coil SNARE domains. A specific combination of SNARE proteins forms a trans-SNARE complex, which provides physical strength for effective membrane fusion. Each SNARE protein is classified as Qa-, Qb-, Qc-, or R-SNARE by the presence of either glutamine (Q) or arginine (R) in the SNARE domain. Syntaxin17 (STX17), an autophagosomal Qa-SNARE in mammals (Itakura et al. 2012), is recruited to a closed autophagosome by its interaction with ATG8 and assembled with the Qbc-SNARE SNAP29 and the R-SNARE VAMP8 to form a trans-SNARE complex (Jiang et al. 2014; Kumar et al. 2018; Tsuboyama et al. 2016). ATG8 regulates the number and acidity of lysosomes by interacting with the Qa-SNARE Stx16, and this regulation is critical to starvation-induced bulk autophagy and selective autophagy (Gu et al. 2019). The R-SNARE Ykt6 is also involved in autophagosome fusion in yeast and mammalian cells (Bas et al. 2018; Matsui et al. 2018). The yeast Rab7 homolog Ypt7 is involved in autophagosome-vacuole fusion and vacuolar biogenesis (Kirisako et al. 1999). Autophagosome-lysosome fusion also requires Rab7 and its activation in flies (Fujita et al. 2017; Hegedus et al. 2016), although mammalian Rab7 appears to be dispensable in this process (Kuchitsu et al. 2018). A GTP-bound form of Rab7 sequentially recruits Rab7 effectors (e.g., the homotypic fusion and vacuole protein sorting, or HOPS), which are involved in autophagosome fusion.

An autophagosomal SNARE or Rab protein is expected to show at least partial co-localization with autophagosomal markers, and its mutation is expected to cause a decrease in autophagic flux and accumulation of autophagosomes in the cytoplasm. A plant autophagosomal SNARE is yet to be discovered, although a mutation in the Arabidopsis Qb-SNARE gene VTI12 resulted in hypersensitivity to starvation (Surpin et al. 2003). Among the eight RAB7 isoforms (RabG1, RabG2, and RabG3a to RabG3f) found in Arabidopsis (Vernoud et al. 2003), RabG3b was shown to co-localize with ATG8a at the double-layered membrane, where their co-localization was enhanced by the invasion of a pathogen (Kwon et al. 2013). Notably, overexpression of a constitutively active form of RabG3b, but not the dominant-negative form, induced autophagy and accumulation of autophagic vesicles. These data suggest that RabG3b positively regulates autophagosome biogenesis and raises the possibility that other RAB7 isoforms are redundantly involved in autophagosome fusion. Indeed, a recent study indicated that RabG3f has a role in autophagy (Rodriguez-Furlan et al. 2019). Immunoprecipitation data suggest that ATG8e interacts with a constitutively active form of RabG3f fused to GFP. In addition, when the small synthetic molecule Endosidin17 abolished the interaction of RABG3f (and likely other RAB7 isoforms) with the retromer subunit VPS35, HOPS assembly was affected, and the vacuole appeared to separate into multiple compartments. Endosidin17 also resulted in the cytoplasmic accumulation of autophagic vesicles decorated with GFP-ATG8e (Rodriguez-Furlan et al. 2019). Because RABG3f plays a critical role in endosomal trafficking and vacuole biogenesis (Cui et al. 2014; Ebine et al. 2014; Singh et al. 2014), these results indicate that the endosome-vacuole and autophagosome-vacuole fusions share a common machinery in plants. Alternatively, the defective biogenesis of a vacuole may indirectly lead to the inhibition of autophagosome-vacuole fusion.

Perspectives

Compared to a previous model (Chung 2019), our new model of autophagosome dynamics (Fig. 1a) reflects a newly assigned role of Atg2 (Fig. 2a). The PAS in yeast is found between the ER and vacuole, and this may also be the case in plants. However, plant cells are larger than yeast and have a distinct trafficking system. Autophagosomes in plant cells likely form at multiple sites, like the case in mammalian cells. The ER-plasma membrane contact site was recently proposed as a site of autophagosome formation in Arabidopsis (Wang et al. 2019), although the vacuole may also be adjacent to the ER-plasma membrane contact site.

It is not clear whether the cytoskeleton mediates autophagosome transport. Because cytoskeletons, especially branched actin filaments, are involved in autophagosome biogenesis, future efforts should focus on specific interference of cytoskeleton motor proteins. Arabidopsis mutants accumulating mature autophagosomes (with little or no effect on other trafficking routes) would be useful for elucidating the final steps in autophagosome dynamics, but such mutants are currently unavailable. Thus, characterization of the molecular machinery responsible for autophagosome fusion is of high priority.

This mini-review did not describe the regulators of core ATG proteins, but it is worth mentioning that both transcriptional (Wang et al. 2020; Yang et al. 2020) and post-translational (Bao et al. 2020; Chen et al. 2017; Qi et al. 2020; Son et al. 2018; Soto-Burgos and Bassham 2017; Van Leene et al. 2019) regulation mechanisms are being revealed in Arabidopsis (Fig. 3). Future research will elucidate molecular mechanisms underlying the interplay between autophagy and signaling network involving plant hormones and stress responses (Jung et al. 2020; Rodriguez et al. 2020). We speculate that these multiple regulatory circuits differentially connect to diverse types of selective and non-selective autophagy in plant cells (Fig. 3). Additional research is needed on how the autophagosome forms during selective autophagy of membraneless compartments (Yoon and Chung 2019) and various organelles including the ER (Hu et al. 2020; Zhang et al. 2020). Additionally, more complexity in autophagic routes, including ATG1-independent autophagy (Huang et al. 2019), ATG7-independent targeting of ATG8 to the vacuole (Ishii et al. 2019; Jia et al. 2019), and ATG7-dependent or -independent microautophagy (Chanoca et al. 2015; Goto-Yamada et al. 2019; Nakamura et al. 2018), is becoming evident from recent studies using Arabidopsis and budding yeast. The complex nature of macroautophagic and microautophagic routes in plant cells is undoubtedly challenging. However, newly developed chemical (Dauphinee et al. 2019) and genetic (Norizuki et al. 2019) tools for plant autophagy research will help describe such complex autophagic routes.

Fig. 3
figure 3

Models for autophagy regulators recently defined in Arabidopsis. a Under non-inducing conditions, autophagy is attenuated at a basal level by several mechanisms. The Target Of Rapamycin Complex 1 (TORC1) negatively regulates autophagy likely by phosphorylating ATG13. The RING-type E3 ligases SINAT1 and 2 ubiquitylate ATG6 and ATG13, leading to their proteolysis by the proteasomes. The transcriptional regulator HY5, together with the histone deacetylase HDA9, suppresses ATG8e and ATG5 transcription. COST1 (Constitutively stressed 1) also negatively regulates autophagy through an unknown mechanism. b Upon starvation, autophagy is induced transcriptionally and post-translationally. The protein kinase complex SnRK1 promotes autophagy, probably by down-regulating TORC1 and/or by phosphorylating ATG1. A RING-finger-truncated protein SINAT6 accumulates during starvation and may compete with SINAT1/2, leading to stabilization of ATG6 and ATG13. HY5 is ubiquitylated and degraded, which releases HDA9 from the ATG8e and ATG5 promoters. The transcription factor TGA9 activates transcription of various core ATG genes, whose protein products are mostly consumed in the vacuole and need to be re-synthesized for sustained autophagy. Apart from starvation, other stresses can either inhibit or transiently induce autophagy. It is largely unclear whether selective or non-selective autophagy is responsible for each type of stress-induced autophagy. Proteasome stress and heat shock lead to accumulation of transcripts encoding the aggrephagy receptor NBR1, which is also consumed in the vacuole. During drought stress, COST1 is ubiquitylated and degraded via both proteasomal and autophagic routes (degradation pathways indicated by dashed arrows). Molecules and processes with a yellow background are in their active states. Ub, polyubiquitin chain