Chloroplast Envelopes Play a Role in the Formation of Autophagy-Related Structures in Plants

Autophagy is a degradation process of cytoplasmic components that is conserved in eukaryotes. One of the hallmark features of autophagy is the formation of double-membrane structures known as autophagosomes, which enclose cytoplasmic content destined for degradation. Although the membrane source for the formation of autophagosomes remains to be determined, recent studies indicate the involvement of various organelles in autophagosome biogenesis. In this study, we examined the autophagy process in Bienertia sinuspersici: one of four terrestrial plants capable of performing C4 photosynthesis in a single cell (single-cell C4 species). We demonstrated that narrow tubules (stromule-like structures) 30–50 nm in diameter appear to extend from chloroplasts to form the membrane-bound structures (autophagosomes or autophagy-related structures) in chlorenchyma cells of B. sinuspersici during senescence and under oxidative stress. Immunoelectron microscopic analysis revealed the localization of stromal proteins to the stromule-like structures, sequestering portions of the cytoplasm in chlorenchyma cells of oxidative stress-treated leaves of B. sinuspersici and Arabidopsis thaliana. Moreover, the fluorescent marker for autophagosomes GFP-ATG8, colocalized with the autophagic vacuole maker neutral red in punctate structures in close proximity to the chloroplasts of cells under oxidative stress conditions. Together our results implicate a role for chloroplast envelopes in the autophagy process induced during senescence or under certain stress conditions in plants.


Introduction
Autophagy is a process responsible for the degradation of cytosolic and organellar materials for nutrient recycling and the removal of undesirable components in eukaryotes. Although there are several types of autophagy, two main autophagic pathways have been described in plants: microautophagy and macroautophagy [1,2]. Microautophagy involves the direct engulfing of cytoplasmic contents by the invagination of the tonoplast, followed by their release inside the vacuole. Macroautophagy is a process where a portion of cytoplasm, including organelles, is sequestered into a double-or multi-membrane structure called an autophagosome. Subsequently, the autophagosome is transported into lysosome in mammals or vacuole in yeast and plants for hydrolysis and degradation by proteases. Finally, the degraded products, such as amino acids, are reallocated and recycled [for plant autophagy, see reviews: [3][4][5][6][7]. In plants, autophagy is induced under nutrient-limiting conditions such as carbon and nitrogen starvation [8][9][10][11], during developmental events such as senescence [12][13][14], and in response to oxidative stress and pathogen attacks [15][16][17]. Autophagy observed under normal growth conditions has been suggested to serve as a housekeeping role [18,19].
Evidence connecting the relationship between autophagosome and autophagic vacuole biogenesis has been suggested in previous studies. For example, provacuoles formed tubule structures sequestering portions of cytoplasm, which eventually led to the formation of autophagic vacuoles that have been documented in the root meristematic cells of Euphorbia characias [20]. In a study using tobacco miniprotoplasts in which central vacuoles were removed from protoplasts, it was shown that cytoplasmic materials in newly formed vacuoles and in the presence of the cysteine protease inhibitor E-64d, indicating the participation of autophagy concomitant with the biogenesis of new vacuoles [21]. However, autophagy was not inhibited by macroautophagy inhibitors such as 3-Methyladenine, wortmannin, or LY294002, and thus the authors proposed that different types of autophagy could occur in miniprotoplasts. Similarly, lytic vacuoles have been observed specifically in chloroplastcontaining cells of senescing Arabidopsis leaves and termed senescence-associated vacuoles (SAV) [22,23]. Moreover, their presence in mesophyll protoplasts of the atapg7-1 Arabidopsis mutant indicated that the biogenesis of the SAV was independent of the ATG-dependent autophagic pathway.
Although autophagy has been discovered for more than 40 years, the origin of autophagosomes and the membrane source appears to involve numerous organelles. In yeast, the preautophagosomal structure containing the ATG proteins played a central role in autophagosome formation [19,[24][25][26]. Several lines of research supported that the endoplasmic reticulum (ER), or the trans-Golgi network was responsible for autophagosome generation [2,27,28]. Studies using electron 3D tomography showed interconnections between ER and premature-autophagosome membranes in mammalian cells [29,30]. Hailey et al. reported that the outer membrane of mitochondria was utilized for autophagosome biogenesis during starvation [31]. Similarly, the involvement of the plasma membrane in the formation of autophagasome has been suggested by the recruitment of Atg16L1 to the plasma membrane, autophagosomes, and clathrin-coated structures [32]. These findings implicate that various organelles may contribute to the origin of autophagosome membranes depending on cell types and situations. Although the molecular machinery involved in autophagy is conserved among eukaryotes, including mammals, yeasts, and plants [19], it is also possible that different organisms have developed novel pathways for autophagosome formation using different organelles as a source of the membrane.
Plastids are ubiquitous organelles specifically found in photoautotrophic organisms. Plastids have been known to form membrane extensions and protrusions devoid of thylakoid membranes, and tubule structures are named "stromules" because they are stromafilled tubules [33]. Electron microscopy revealed that stromules are double-membrane structures with both outer and inner envelopes of plastids [34][35][36]. While stromules are more abundant in plastids of non-photosynthetic tissues, they are also detected in plastids of a variety of cell types and plant species [37,38]. Moreover, the close association of stromules with other organelles such as ER, mitochondria, and the nucleus has been observed in Arabidopsis thaliana, Nicotiana tabacum, and Nicotiana benthamiana [37,[39][40][41]. Although the biological function of stromules remains elusive, the flow of proteins between stromules and other compartments in close proximity has been directly demonstrated [37,[42][43][44][45]. However, transporting materials between plastids may not be the major and only function of stromules because the majority of plastids are not connected by stromules [38,40,45]. Regardless, the dynamic interactions between stromules and these organelles implicate that they likely have multiple roles in a variety of cellular processes.
Bienertia sinuspersici represents one of four terrestrial plants previously shown to perform C 4 photosynthesis in a single cell [46,47]. The Bienertia species carry out C 4 photosynthesis without requiring the traditional Kranz anatomy. In this single-cell C 4 system, the C 4 pathway is achieved by the spatial compartmentalization of dimorphic chloroplasts, mitochondria and peroxisomes, and photosynthetic enzymes in two distinct domains of the cytoplasm: the peripheral cytoplasmic compartment (PCC) and the central cytoplasmic compartment (CCC) [47,48]. The PCC replaces the mesophyll cell, whereas the CCC performs the function of the bundle sheath cells. C 3 plants evolved an important mechanism known as the chloroplast photorelocation movement to protect plants from light stress [49]. In contrast, the dimorphic chloroplasts appeared to be permanently maintained in their respective cytoplasmic subdomains, the PCC and CCC, respectively, by an elaborate cytoskeletal network with limited mobility [48]. Therefore, chloroplasts in the single-cell C 4 system are likely to be more susceptible to oxidative damage. In this study, we investigated the role of stromules in the biogenesis of autophagy-related structures during chloroplast degradation in chlorenchyma cells of Bienertia sinuspersici and Arabidopsis thaliana while undergoing senescence or under oxidative stress. It was demonstrated that chloroplast-derived stromules appeared to serve as a source of the membranes of autophagy-related structures. In addition, the in vivo localization of autophagosomes and autophagic vacuoles was illustrated by green fluorescent protein (GFP)-ATG8 fusion and neutral red (NR), respectively. The possible role of chloroplast envelopes in autophagic vacuole formation was also discussed.

Autophagy in Stress-Induced Chlorenchyma Cells of Bienertia sinuspersici
In Arabidopsis seedlings, autophagy was induced by reactive oxygen species (ROS) [16]. In order to determine whether this would occur in mature leaves, isolated chlorenchyma cells of B. sinuspersici were treated with 20 mM H 2 O 2 for up to 2 h and observed using transmission electron microscopy (TEM) (Figure 1). The central cytoplasmic compartment (CCC) appeared to be unaffected under low-resolution images, whereas some peripheral cytoplasmic compartment (PCC) chloroplasts dilated, while others maintained their normal oblong shape ( Figure 1A). However, at a higher resolution, structures that resembled autophagosomes and autophagic vacuoles were observed in addition to chloroplasts and mitochondria in the CCC ( Figure 1B). The autophagic-like bodies were mainly oval in shape with diameters ranging from 0.4 to 2.5 µm and were occasionally found in the central vacuole in some cells ( Figure 1C,D). In autophagy-related structures, cytoplasmic components, including mitochondria, were found to be surrounded by multiple membranes in both the CCC and PCC ( Figure 1E,F). Autophagic vacuoles, which were electron transparent, were also observed in the CCC and PCC and appeared to degrade cellular components within the structures ( Figure 1G,H). However, the autophagy-related structures were rarely observed in healthy mature untreated chlorenchyma cells ( Figure 1I-K). These results indicated that autophagy was induced by H 2 O 2 in the chlorenchyma cells of mature B. sinuspersici leaves.

Contribution of Stromules in Autophagy in Chlorenchyma Cells of Bienertia sinuspersici
It has been reported that chloroplasts change their morphology under some stress conditions. Holzinger et al. described that chloroplast protrusions appeared at high temperatures (35-45 • C) while these structures were not detected at low temperatures 5-15 • C in Arabidopsis mesophyll chloroplasts [39]. To observe such changes under oxidative stress, stromal proteins such as ribulose 1,5-bisphosphate carboxylase/oxygenase (Rubisco) large subunit (RbcL) or pyruvate Pi dikinase (PPDK) were immunolabelled in chlorenchyma cells and observed using TEM ( Figure 2). In the CCC of the control cells, the dilation and elongation of chloroplast envelopes containing stromal proteins were observed (Figure 2A,B). Small vesicles with a diameter in the range of 90-240 nm containing stromal proteins were detected mostly in the PCC and the peripheral region of the CCC along the central vacuole ( Figure 2C). Only stroma-containing vehicles (SCVs) with spherical shapes were chosen and measured to prevent the inclusion of stromules. Some SCVs appeared to be releasing content to the central vacuole. On the other hand, oxidative stress-treated cells showed stromules at a higher frequency ( Figure 2D,E) and varied in size and shape. Stromules were also observed in cytoplasmic strands connecting the CCC and PCC ( Figure 2F). In previous studies, we have shown that RbcL mostly accumulated in CCC chloroplasts but was scarce in the PCC chloroplasts of mature B. sinuspersici leaves [47,48]. The densely labeled stromule by the RbcL antibody suggested that it extended from the CCC chloroplasts to the PCC through the cytoplasmic strand. Although chloroplast protrusions, stromules, and SCVs were found in chlorenchyma cells under the control conditions, they were more abundant in oxidative stress-treated cells.   and PCC ( Figure 2F). In previous studies, we have shown that RbcL mostly accumulated in CCC chloroplasts but was scarce in the PCC chloroplasts of mature B. sinuspersici leaves [47,48]. The densely labeled stromule by the RbcL antibody suggested that it extended from the CCC chloroplasts to the PCC through the cytoplasmic strand. Although chloroplast protrusions, stromules, and SCVs were found in chlorenchyma cells under the control conditions, they were more abundant in oxidative stress-treated cells. We have successfully captured images showing narrow stromules with an average diameter of 36 nm (including membranes and excluding base and tip parts of the stromules), originating from chloroplasts and extending in the cytoplasm, which appears to be sequestering portions of the cytoplasm ( Figure 3A-D). Careful observation revealed that labeling for stromal proteins was detected between membranes of autophagy-related structures in We have successfully captured images showing narrow stromules with an ave diameter of 36 nm (including membranes and excluding base and tip parts o stromules), originating from chloroplasts and extending in the cytoplasm, which app to be sequestering portions of the cytoplasm ( Figure 3A-D). Careful observation reve that labeling for stromal proteins was detected between membranes of autophagy-re structures in both the CCC and PCC of B. sinuspersici chlorenchyma cells ( Figure 3E The diameter of the stromal part of stromules was 10-20 nm, with some exceptions (Fi 4A,B). Although autophagy-related structures are generally surrounded by two app membranes, some structures are enclosed by multiple membranes ( Figure 4C,D). sociated with the CCC chloroplasts consist of narrow stromules showing specific reactivity to ant RbcL antibodies; (G) The membrane surrounding the autophagy-related structure in the proximit of the PCC reacted with anti-PPDK antibodies. White arrows indicate gold particles specificall bound to RbcL (A-F) or PPDK (G). CCC-central cytoplasmic compartment; PCC-peripheral cy toplasmic compartment; Ch-chloroplast; st-stromule; ap-autophagy-related structure; m-m tochondrion; PCh-PCC chloroplast; cw-cell wall. Scale bars = 200 nm in (A-F), 500 nm in (G). The contents of autophagy-related structures surrounded by stromules seemed t vary based on their appearance. To identify the contents, immunolabeling experiment using various antibodies were performed ( Figure 5). For example, labeling for catalas was found inside the autophagy-related structure, indicating that a peroxisome was en closed by membranes ( Figure 5A). The RbcL antibody labeled an SCV and its surroundin stromules ( Figure 5B). In addition, labeling for the cytosolic marker phosphoenolpyruvat carboxylase revealed that cytosolic content, along with another organelle, was als The contents of autophagy-related structures surrounded by stromules seemed to vary based on their appearance. To identify the contents, immunolabeling experiments using various antibodies were performed ( Figure 5). For example, labeling for catalase was found inside the autophagy-related structure, indicating that a peroxisome was enclosed by membranes ( Figure 5A). The RbcL antibody labeled an SCV and its surrounding stromules ( Figure 5B). In addition, labeling for the cytosolic marker phosphoenolpyruvate carboxylase revealed that cytosolic content, along with another organelle, was also captured by autophagy-related structures ( Figure 5C,D). Together with the structures related to autophagy containing mitochondria shown in Figure 1E, autophagic bodies induced by the oxidative stress treatment appeared to be generated in a non-selective manner. Furthermore, autophagy-related structures were often found in close proximity to either central vacuoles or autophagic vacuoles ( Figure 5). captured by autophagy-related structures ( Figure 5C,D). Together with the structures related to autophagy containing mitochondria shown in Figure 1E, autophagic bodies induced by the oxidative stress treatment appeared to be generated in a non-selective manner. Furthermore, autophagy-related structures were often found in close proximity to either central vacuoles or autophagic vacuoles ( Figure 5).

Contribution of Stromules in Autophagy in Arabidopsis Mesophyll Cells
Autophagic bodies surrounded by chloroplast-derived stromules were observed in stress-induced chlorenchyma cells of mature B. sinuspersici. To determine whether this process is conserved in other plant species, Arabidopsis was used for EGFP expression and immunoelectron microscopic analyses. First, Arabidopsis protoplasts were transfected with RbcS-EGFP plasmid DNA by the polyethylenglycol (PEG)-mediated method [50] and treated with H 2 O 2 to observe chloroplast morphology. The RbcS-EGFP signal was observed mostly in the chloroplasts of non-treated protoplasts, whereas it was also found in vesicles attached to or away from chlorophyll autofluorescent signals in H 2 O 2 -treated Plants 2023, 12, 443 9 of 21 protoplasts ( Figure 6). Next, Arabidopsis leaves were treated with H 2 O 2 , immunolabelled with an anti-RbcL antibody, and observed using TEM (Figure 7). Cytoplasmic components, including SCVs, were observed in the central vacuole of stressed cells ( Figure 7A,B). Specific reactions to RbcL were also found in stromules encasing cytoplasmic regions. Some autophagic-related structures surrounded by stromules appeared to release their contents into the central vacuole ( Figure 7C), while others appeared to have degraded cellular components inside of them ( Figure 7D-F). The size of autophagic bodies, excluding stromules ranged between 0.4 and 1 µm in diameter. However, stromules sequestering a larger area of the cytoplasm were occasionally observed ( Figure 7G). The chloroplast invagination of cytoplasmic materials was also observed in stress-treated Arabidopsis leaves ( Figure 7H-J). The engulfed contents appeared to be degraded in autophagic vacuoles formed in chloroplasts. These results indicate that stromules and chloroplasts contributed to autophagy in the oxidative stress-induced mesophyll cells of Arabidopsis, similar to that observed in B. sinuspersici.
stress-induced chlorenchyma cells of mature B. sinuspersici. To determine whether this process is conserved in other plant species, Arabidopsis was used for EGFP expression and immunoelectron microscopic analyses. First, Arabidopsis protoplasts were transfected with RbcS-EGFP plasmid DNA by the polyethylenglycol (PEG)-mediated method [50] and treated with H2O2 to observe chloroplast morphology. The RbcS-EGFP signal was observed mostly in the chloroplasts of non-treated protoplasts, whereas it was also found in vesicles attached to or away from chlorophyll autofluorescent signals in H2O2-treated protoplasts ( Figure 6). Next, Arabidopsis leaves were treated with H2O2, immunolabelled with an anti-RbcL antibody, and observed using TEM (Figure 7). Cytoplasmic components, including SCVs, were observed in the central vacuole of stressed cells ( Figure 7A,B). Specific reactions to RbcL were also found in stromules encasing cytoplasmic regions. Some autophagic-related structures surrounded by stromules appeared to release their contents into the central vacuole ( Figure 7C), while others appeared to have degraded cellular components inside of them ( Figure 7D-F). The size of autophagic bodies, excluding stromules ranged between 0.4 and 1 μm in diameter. However, stromules sequestering a larger area of the cytoplasm were occasionally observed ( Figure 7G). The chloroplast invagination of cytoplasmic materials was also observed in stress-treated Arabidopsis leaves ( Figure 7H-J). The engulfed contents appeared to be degraded in autophagic vacuoles formed in chloroplasts. These results indicate that stromules and chloroplasts contributed to autophagy in the oxidative stress-induced mesophyll cells of Arabidopsis, similar to that observed in B. sinuspersici.

Localization of Autophagic Vacuoles
Neutral red (NR), which stains acidic compartments, is a common dye used to visualize lysosomes in mammals. NR has also been applied to plant tissues for the observation of autophagic vacuoles in Arabidopsis and tobacco [18,21,22,51]. To understand how autophagic vacuole is formed, chlorenchyma cells isolated from healthy, naturally senescing, and stress-induced B. sinuspersici leaves were stained with NR and observed under a confocal microscope (see Figure 8A). In the control cells, NR mainly accumulated around chloroplasts ( Figure 8A left panel). Interestingly, the accumulation of NR in autophagy-related structures associated with PCC chloroplasts was observed in some control cells, suggesting that acidic autophagic bodies formed in the PCC under the normal growth condition (Arrows in Figure 8A left panel). Naturally senescing and stress-induced cells showed the localized accumulation of NR, often having more than one punctate structure associated with each chloroplast ( Figure 8A middle and right panels). The preferential accumulation of NR along chloroplasts in all types of cells suggested a potential relationship between the formation of autophagic vacuole and chloroplasts. The electron micrograph of the CCC in the H 2 O 2 -treated cell further supported this idea, showing autophagic vacuole formation on chloroplast envelopes and autophagosome membranes possibly derived from chloroplasts ( Figure 8B). In addition, similar patterns in the autophagic vacuole formation between naturally senescing and H 2 O 2 -treated chlorenchyma cells indicated that cell death was induced by oxidative stress. Furthermore, partially degraded cytoplasmic components, including the mitochondria and stromules in autophagic vacuoles, suggested that these newly formed vacuoles contained hydrolase activity ( Figure 8C).

Localization of Autophagic Vacuoles
Neutral red (NR), which stains acidic compartments, is a common dye used to visualize lysosomes in mammals. NR has also been applied to plant tissues for the observation of autophagic vacuoles in Arabidopsis and tobacco [18,21,22,51]. To understand how autophagic vacuole is formed, chlorenchyma cells isolated from healthy, naturally senescing, and stress-induced B. sinuspersici leaves were stained with NR and observed under a confocal microscope (see Figure 8A). In the control cells, NR mainly accumulated around chloroplasts ( Figure 8A left panel). Interestingly, the accumulation of NR in autophagyrelated structures associated with PCC chloroplasts was observed in some control cells, suggesting that acidic autophagic bodies formed in the PCC under the normal growth condition (Arrows in Figure 8A left panel). Naturally senescing and stress-induced cells showed the localized accumulation of NR, often having more than one punctate structure associated with each chloroplast ( Figure 8A middle and right panels). The preferential accumulation of NR along chloroplasts in all types of cells suggested a potential relationship between the formation of autophagic vacuole and chloroplasts. The electron micrograph of the CCC in the H2O2-treated cell further supported this idea, showing autophagic vacuole formation on chloroplast envelopes and autophagosome membranes possibly derived from chloroplasts ( Figure 8B). In addition, similar patterns in the autophagic vacuole formation between naturally senescing and H2O2-treated chlorenchyma cells indicated that cell death was induced by oxidative stress. Furthermore, partially degraded cytoplasmic components, including the mitochondria and stromules in autophagic vacuoles, suggested that these newly formed vacuoles contained hydrolase activity ( Figure 8C).

Autophagy-Related Structures and Autophagic Vacuole Formation in Arabidopsis Mesophyll Protoplasts
Autophagy-related structures and autophagic vacuoles were often observed in close proximity. To investigate the relationship between these two autophagic organelles under oxidative stress conditions, the localization of an autophagosome marker protein ATG8 and autophagic vacuole marker NR were analyzed. In the control non-stressed protoplasts, such as EGFP-ATG8, were expressed in the cytosol and nucleus, whereas they accumulated in the vesicles of H 2 O 2 -treated protoplasts ( Figure 9A). NR fluorescence was detected around chloroplasts in the control protoplasts as punctate structures in H 2 O 2 -treated protoplasts ( Figure 9B). When EGFP-ATG8 expressing protoplasts were stained with NR to compare the localization of the fluorescent signals, some signals were independent of each other, although the other signals in the cytosol and punctate structures were overlapped in H 2 O 2 -treated protoplasts ( Figure 9C). The distinct signals of EGFP and NR indicated that the formation of autophagosomes and autophagic vacuoles was derived from the independent path, whereas the overlapped signals in the punctate structures suggested the conversion from autophagosomes to autophagic vacuoles. Moreover, in mesophyll cells of stress-treated leaves, electron micrographs of autophagosomes and autophagic vacuoles with different morphologies and contents were detected as further supporting the potential autophagosome/autophagic vacuole transformation ( Figure 10A-D). Similarly, in mesophyll cells of senescing Arabidopsis leaves, stromule-bounded autophagic structures were also frequently observed protruding from the edge of chloroplasts ( Figure 10E,F).

Discussion
Accumulating evidence on the dynamics of stromules and participation of autophagy in the degradation of chloroplastic proteins led us to hypothesize on the plant-specific

Discussion
Accumulating evidence on the dynamics of stromules and participation of autophagy in the degradation of chloroplastic proteins led us to hypothesize on the plant-specific mechanism in autophagy-related structure biogenesis. Previous studies suggested that the degradation of chloroplastic proteins occurred through autophagic organelles [14,22,[52][53][54][55][56]. Although these autophagic organelles have been observed with isolation membranes, the identity of the membrane origin remained undetermined in plants. Here, we investigated the localization of stromal proteins in leaf mesophyll cells under oxidative stress. Our data showed the involvement of stromules in autophagic-related structure formation. In addition, the localization analysis of autophagic vacuoles suggested the role of chloroplast envelopes in autophagic vacuole formation.
Under oxidative stress, numerous stromules and SCVs were induced in the chlorenchyma cells of B. sinuspersici ( Figure 2D-F). In this study, the size of the stromules was 0.1-0.3 µm in diameter, which is somewhat narrower than typical stromules with a diameter ranging from 0.3 to 0.8 µm as determined in other studies [57,58]. In addition, stromules with an average diameter of 36 nm were occasionally observed ( Figure 3A-D). We observed the labeling for stromal proteins in a 10-20 nm wide gap between membranes with a white appearance indicating a stromal portion of stromules. Thin stromules with a diameter of less than 100 nm have also been reported in tomatoes, Arabidopsis, and soybeans [36,59,60]. Our immuno-EM analysis revealed thin stromules surrounding portions of cytoplasmic components and forming autophagy-related structures, which are often associated with chloroplasts in both B. sinuspersici and Arabidopsis (Figures 3E-G and 7). Ishida et al. found that a strong GFP-ATG8 signal in chloroplast protrusions co-localized with stromal DsRed signals in Arabidopsis leaves, which supports our observation [54]. Moreover, similar observations on the enclosure of intercellular components by stromules have been made in previous studies, including those performed more than 40 years ago, in tobacco, tomato, Deschampsia antarctica, and green algae suggesting that the participation of stromules in autophagy is conserved in photosynthetic organisms [61][62][63][64].
The stromule-bounded autophagy-related structure is multilamellar and composed of a central body part containing cytoplasmic components and a stromal part between the membranes of a stromule. Chiba et al. demonstrated that Rubisco-containing bodies (RCBs) with double membranes were further surrounded by other membranes, consistent with our data showing that SCVs were confined by stromules [( Figure 5B) 53]. Furthermore, autophagy-related structures with more than two apparent membranes were often found in B. sinuspersici indicating that multilamellar autophagic bodies might be a common characteristic in photosynthetic cells ( Figures 1E, 4C,D and 5A). Autophagy is generally defined as a non-selective process of degrading cellular materials. However, selective autophagy degrading certain organelles has been studied and named based on the corresponding organelle subjected to degradation: for instance, pexophagy for peroxisomes [65,66], mitophagy for mitochondria [67], and chlorophagy for chloroplasts [56,68,69]. These authors demonstrated the digestion of chloroplastic proteins via RCBs in Arabidopsis during avirulent AvrRps4 infection as well as in starchless mutants while being suppressed in a carbon-rich condition and in the starch-excess mutants. Whole chloroplast degradation via autophagy has also been observed in photodamaged and senescence-induced Arabidopsis leaves [14,56,70]. Although the degradation pathways for peroxisomes and mitochondria have not been understood in plants, the results from this study demonstrate that autophagosomes formed under oxidative stress contained various intercellular components, including peroxisomes, mitochondria, SCVs, and cytosol ( Figures 1E and 5). These data suggested that oxidative stress-induced autophagy is responsible for the non-selective degradation of cytoplasmic components and also indicated that the degradation of peroxisomes and mitochondria occurred, at least partially, via the autophagosome-mediated pathway in mesophyll cells. In addition, SCVs were also found in the non-stress-treated cells of B. sinuspersici ( Figure 2C). This is in agreement with the findings provided by Prins et al. describing that Rubisco-containing vesicles were observed in mature tobacco leaves under stress and optimal conditions as well as in young leaves [55]. Regardless, whether this pathway is specifically regulated by nutrient conditions remains undetermined in this study; the chloroplastic protein degradation through the autophagic pathway might regularly occur in a normal growth condition.
In mammals, autophagosome fuses with a lysosome to form an autolysosome where the degradation of sequestered contents takes place [71]. Vacuoles are analogous to animal lysosomes with lytic activity in yeast and plants. In plants, most of the volume of the mature cell is occupied by a large central vacuole which has multiple functions, including the maintenance of turgor pressure and storage of metabolites as well as the digestion of cellular constituents [72]. In addition to a central vacuole, the formation of autophagic vacuoles has been observed in various plant tissues and cells [20][21][22]64,[73][74][75]. In B. sinuspersici chlorenchyma cells, the CCC is composed of massive cytosol and numerous organelles packed together and forming a ball-like structure surrounded by a central vacuole [47,48,76]. After the formation of an autophagy-related structure, it appeared to be transported into the central vacuole only, where the distance of two organelles was close enough, such as in the PCC and the peripheral region of CCC ( Figure 5A,D). On the other hand, an autophagic body formed deep inside of the CCC seemed to be engulfed by or transformed into autophagic vacuoles, although it is necessary to note that these vacuoles were also observed in the PCC ( Figures 1H and 4C,D). Findings from the present study suggested the participation of two autophagic pathways: one involved in the transportation of cellular materials into the lumen of the central vacuole, while another utilized autophagic vacuoles synthesized de novo. However, it is still possible that autophagic vacuoles could also participate in the first pathway. After the formation of an autophagy-related structure near the central vacuole, autophagic vacuoles generated in close proximity to the autophagosomes could fuse with the central vacuole making autophagosomes protrude into the vacuolar lumen to facilitate a subsequent release ( Figure 5A,B,D). In the second pathway, there seem to be sequential events indicating the transformation of an autophagic body into an autophagic vacuole: first, stromules sequester cytoplasmic materials generating an autophagy-related structure ( Figure 10A,E,F); next, autophagic body membranes start deteriorating ( Figure 10B); third, the cellular materials are digested in the lumen of the autophagic vacuole ( Figure 10C); finally, all contents including internal membranes are degraded leaving an outermost boundary membrane ( Figure 10D).
The co-localization study of EGFP-ATG8 and NR signals in Arabidopsis protoplasts also suggests two types of autophagic organelles that are partially overlapped in punctate structures under oxidative stress (Figure 7). In agreement with this observation, numerous studies have demonstrated the requirement of ATG8a for autophagosome formation in Arabidopsis during nutrient stress and senescence [77][78][79]. Interestingly, the upregulation of autophagy-related genes (ATG8 and ATG12), which is reported to be involved in distinct pathways, in the formation of autophagosome was observed in drought-stressed Caragana korshinskii leaves: a leguminous species that thrives in arid and semi-arid habitats [80]. Moreover, the high level of NR accumulation in autophagy-related structures associated with PCC chloroplasts in non-stressed B. sinuspersici cells indicated that the single-cell C 4 species might adopt a similar drought-tolerant mechanism to survive in extreme saline and arid environment ( Figure 8A left panel). Similar events in the autophagic vacuole formation have also been observed in non-photosynthetic cells, including Arabidopsis suspensioncultured cells [74], root meristematic cells in Euphobia [20], and tobacco [81]. Zheng and Stahelin have described that the membrane extension originated from protein storage vacuoles enclosed in an area of cytoplasm forming multilamellar autophagosomes in the inner cortex and vascular cylinder cells [81]. The luminal space of the extensive membrane differentiated into pre-lytic vacuoles by 'reinflation', which expanded and engulfed an entire autophagosome, and eventually, the whole domain was transformed into a lytic vacuole. The similarities in various cell types for different conditions suggest the plasticity of the autophagosome and autophagic vacuole biogenesis and supplied membranes from different organelles.
The generation of autophagic vacuoles has been indicated to be independent of, at least, early steps of macroautophagy [21,22]. Autophagic vacuoles were electron-translucent and occasionally contained partially degraded cellular materials ( Figures 1H, 2E, 7I, 8C and 10C,D). In vivo localization analysis showed that the majority of NR were accumulated along with chloroplasts regardless of the cell conditions ( Figure 8A). In agreement with this, the generation of autophagic vacuole appeared to initiate along with chloroplast envelopes, including those on stromules ( Figures 1E,G, 2D and 8B). These observations implicated the involvement of chloroplast envelopes in the autophagic vacuole formation. Moreover, the contribution of chloroplast envelopes without an extension in the autophagic vacuole formation indicated that autophagosome formation is not essential for the generation of autophagic vacuoles. This is further supported by chloroplast invaginations of cellular components followed by the on-site generation of autophagic vacuoles without forming autophagosomes ( Figure 7H-J). While our findings have established the close proximity between stromules and autophagic bodies, future experiments will examine how they are directly involved in the autophagy pathway.
Since the discovery of B. sinuspersici as one of four terrestrial single-cell C 4 species, a number of studies have focused on how this novel photosynthetic system develops [47,48,76,[82][83][84]. However, as shown here, B. sinuspersici also serves as a suitable model species for the autophagy study in mesophyll cells. Unlike mature photosynthetic cells in typical plant species, in which most of the cell space is occupied by a large central vacuole and the cytoplasm is limited to the peripheral, B. sinuspersici chlorenchyma cells contain a large cytoplasmic area in the CCC ( Figure 1A,B) [47,48]. This is advantageous, especially when ultra-thin sections are made for the TEM analysis. In addition, the compartmentation of organelles can be exploited for the study of the selective degradation of particular organelles. For instance, mitochondria are exclusively located in the CCC, whereas peroxisomes are found in both compartments ( Figures 1B and 5A,D). We have also established a transient gene expression protocol in B. sinuspersici chlorenchyma protoplasts with comparable efficiency to those previously reported for Arabidopsis [50,51]. Although most genes remain unidentified, this protocol would make it feasible to analyze the localization of autophagy-related proteins such as ATGs.
Here, we have provided data supporting the role of stromules in the generation of autophagy-related structures by sequestering cellular components, including mitochondria, peroxisomes, and cytosol of photosynthetic cells. Recently, a close association between stromules and ER has been established in Catharanthus roseus [85]. The authors documented the subcellular localization of enzymes involved in the synthesis of monoterpene indole alkaloids (MIA) in stromules and postulated a potential role for these structures in the exchange of biosynthetic intermediates to the ER where the subsequent reactions of the pathway occur [85]. Together with our data, it is possible to speculate that ER might participate in the autophagosome formation by guiding membrane extensions from various organelles to sequester cytoplasmic compartments and by supplying lipids to the extensions. Whether plastids in non-photosynthetic cells are involved in autophagosome biogenesis needs to be determined. Nevertheless, plants appear to have established a novel strategy to generate autophagic organelles by supplying chloroplast envelopes as a source of membranes. We believe that these findings offer a new perspective on autophagy in plants.

Plant Materials and Growth Conditions
Bienertia sinuspersici was grown as previously described by Lung et al. [50]. Three-to four-month-old plants were used for all experiments except senescing leaves, which were obtained from plants grown for more than ten months. Arabidopsis thaliana ecotype Columbia was grown in soil (1:1 Sunshine mix of LG3 and LC1) in chambers at 120 µmol m −2 s −1 irradiance at 21 • C with a photoperiod of 16/8 h, light and dark, respectively.

Electron Microscopy
Chlorenchyma cells of B. sinuspersici leaves were isolated, as described previously [50]. Isolated Bienertia chlorenchyma cells and Arabidopsis leaves, which were sectioned into small pieces, were transferred into a cell stabilizing solution containing 25 mM HEPES-KOH (pH 6.5), 5 mM KCl, 1 mM CaCl 2, and 1 M mannitol or 0.3 M mannitol, respectively. Samples were fixed, embedded in LR white resin (Polysciences Inc., Warrington, PA, USA), and sectioned as described previously [86]. Sections were stained with uranyl acetate and lead citrate.
Arabidopsis protoplast transfection was performed according to Yoo et al. [51].

Confocal Laser Scanning Microscopy
Isolated chlorenchyma cells of B. sinuspersici, wildtype, and EGFP-ATG8 transfected Arabidopsis protoplasts were stained with 0.01 or 0.005% (w/v) neutral red, respectively, for 30 min. Samples were washed and fixed in a fixative solution containing 0.3 M mannitol, 50 mM PIPES (pH 7.2), 1.25% (v/v) glutaraldehyde, and 2% (v/v) paraformaldehyde for 10 min. Images were taken using an Olympus FV1000 confocal laser scanning microscope (Olympus, Breiningsville, PA, USA). An EGFP signal was excited at 488 nm, and the emission was detected at 512 nm. Neutral red was excited at 577 nm, and the emission was detected at 592 nm. The excitation wavelength for chlorophyll autofluorescence was 649 nm, and the emission was 666 nm.

Hydrogen Peroxide Treatment
For an oxidative stress treatment, B. sinuspersici mature leaves and isolated chlorenchyma cells were incubated in cell-stabilizing solutions containing 20 mM hydrogen peroxide (H 2 O 2 ) for 2 h at room temperature. For the treatment of Arabidopsis, 10 or 5 mM H 2 O 2 was used for leaves or protoplasts, respectively. A different set of samples was also incubated in a cell stabilizing solution without H 2 O 2 as the negative control.