Organ-specific quality control of plant peroxisomes is mediated by autophagy

ABSTRACT Peroxisomes are essential organelles that are characterized by the possession of enzymes that produce hydrogen peroxide (H2O2) as part of their normal catalytic cycle. During the metabolic process, peroxisomal proteins are inevitably damaged by H2O2 and the integrity of the peroxisomes is impaired. Here, we show that autophagy, an intracellular process for vacuolar degradation, selectively degrades dysfunctional peroxisomes. Marked accumulation of peroxisomes was observed in the leaves but not roots of autophagy-related (ATG)-knockout Arabidopsis thaliana mutants. The peroxisomes in leaf cells contained markedly increased levels of catalase in an insoluble and inactive aggregate form. The chemically inducible complementation system in ATG5-knockout Arabidopsis provided the evidence that these accumulated peroxisomes were delivered to vacuoles for degradation by autophagy. Interestingly, autophagosomal membrane structures specifically recognized the abnormal peroxisomes at the site of the aggregates. Thus, autophagy is essential for the quality control of peroxisomes in leaves and for proper plant development under natural growth conditions.


INTRODUCTION
Peroxisomes are essential organelles characterized by the possession of enzymes that produce hydrogen peroxide (H 2 O 2 ) as part of their normal catalytic cycle (De Duve and Baudhuin, 1966). In plant leaves, two major types of peroxisomes play crucial roles in the postgerminative development of seedlings and photosynthesis (Beevers, 1979). A type of plant peroxisome, the glyoxysome, is involved in fatty acid b-oxidation to convert storage oil into carbohydrates in cotyledons, the embryonic first leaves, to support the postgerminative growth of seedlings until the commencement of photoautotrophic growth. The other prominent type of plant peroxisome, the leaf peroxisome, is unique in that it plays an important role in photorespiration (i.e. the photosynthesis-related oxidative carbon cycle). In both metabolic pathways, especially in photorespiration, H 2 O 2 is generated by oxidative reactions (Foyer and Noctor, 2003) and is scavenged by catalase, which is present in both types of peroxisomes (del Río et al., 2006). Proteins in peroxisomes are damaged by reactive oxygen species (ROS) produced during such peroxisomal oxidation processes. Plants with impaired peroxisome function exhibit a defect in seedling development and have decelerated growth phenotypes (Reumann and Weber, 2006;Fulda et al., 2004;Rylott et al., 2006), indicating that the maintenance and turnover of peroxisomes are important for plant development and growth. Despite their importance, the mechanism by which the basal quality of peroxisomes is maintained is unknown.
Macroautophagy (referred to hereafter as autophagy) is an evolutionarily conserved intracellular process in eukaryotes, which involves the degradation of cytoplasmic constituents and organelles by sequestration in double-membraned vesicles, the autophagosomes (Mizushima et al., 2011). In yeast, phytopathogens and animal cells, there is a type of autophagy that is specific for peroxisome degradation, known as 'pexophagy'. This involves the removal of excess or redundant peroxisomes under particular artificial growth conditions in which peroxisomal functions are no longer required (Sakai et al., 2006;Asakura et al., 2009;Iwata et al., 2006). However, the role of pexophagy in the quality control of peroxisomes remains obscure.

RESULTS AND DISCUSSION
We investigated the behavior of peroxisomes in ATGknockout Arabidopsis thaliana (atg mutants) expressing green fluorescent protein (GFP) fused with PTS1 (Mano et al., 2002), a peroxisomal targeting signal. The number of peroxisomes was increased in the leaves of all atg mutants examined ( Western blotting analysis using antibodies against the representative peroxisomal protein catalase further confirmed these imaging data. Although there were no substantial differences in the transcript levels of catalase genes (supplementary material Fig. S2A; Frugoli et al., 1996), the catalase protein levels were markedly increased in atg5 leaves, but not in the roots, compared to wild-type controls (Fig. 1C). These results suggest that the number of photosynthesisrelated peroxisomes is increased in atg mutants.
As it has been reported that overexpression of PEX11 genes, which encode peroxisomal membrane proteins, causes an increase in peroxisome number (Orth et al., 2007), we checked the transcript levels of the PEX11 genes in atg mutants. However, the transcription levels of PEX11 genes in atg mutants were similar to those in wild type (supplementary material Fig. S2B). We further confirmed that peroxisome numbers were increased even in NahG atg5 plant leaves (supplementary material Fig. S3), in which endogenous salicylic acid (SA) is depleted and consequently the early senescence phenotype is suppressed (Yoshimoto et al., 2009). This indicated that the increase in peroxisome number was not caused by side effects, such as highly accumulated SA or the onset of senescence in atg mutants. These results further support the view that peroxisomes in leaves undergo less degradation in atg mutants, resulting in an increased number of peroxisomes.
We next examined whether autophagy preferentially degrades only peroxisomes and not other organelles. For this, western blotting analysis was performed using antibodies against representative organelle marker proteins to compare the protein levels of each organelle between wild-type and atg5 leaves. As shown in Fig. 1D,E, only peroxisomal proteins were increased in atg5 leaves. In contrast to the increase in peroxisome number, there was no increase in the number of mitochondria in atg5 leaves (supplementary material Fig. S4), strongly suggesting that leaf peroxisomes are selectively degraded by autophagy.
Next, the fine structure of peroxisomes in the leaves of atg mutants was examined by electron microscopy (EM) ( Fig. 2A). Consistent with the results of fluorescence imaging (Fig. 1A, upper-right panel), many peroxisomes were present in atg mutants. In addition, we found that many peroxisomes in atg mutants contained some electron-dense regions ( Fig. 2A, right panel; Fig. 2B; supplementary material Table S1).
As the levels of catalase protein were markedly increased in atg5 leaves compared to those of other peroxisomal proteins (Fig. 1E), we examined the intracellular states of the highly accumulated catalase by immuno-EM analysis (Fig. 2C). Gold particles conjugated with anti-catalase antibodies were evenly distributed in normal peroxisomes. In abnormal peroxisomes, however, the density of gold particles was markedly elevated in showing the protein levels of organelle marker proteins in wild-type and atg5 leaves. Anti-Arf1, -COXII, -SMT1 and -PsbA antibodies were used as markers for the Golgi apparatus, mitochondria, ER and chloroplast, respectively. Anti-GO, -HPR and -PEX14 antibodies were used as peroxisome markers. (E) Quantification of the protein levels of peroxisomal proteins in leaves. Fold change ratios of peroxisomal protein levels in atg5 leaves against those in wildtype leaves were calculated using ImageQuant TL software. Each value represents the mean 6 s.d. of at least three independent experiments. the electron-dense regions and, in such peroxisomes, the density of gold particles was reduced in normal regions, indicating that catalase was concentrated in the electron-dense regions. The electron-dense regions of atg mutant peroxisomes had the appearance of structures that have been reported as aggregates of dysfunctional peroxisome-matrix proteins in the peroxisomal Lon protease-deficient fungus (Bartoszewska et al., 2012). Based on these observations, we hypothesized that the electron-dense regions are dysfunctional protein aggregates, one of the major components of which is catalase. Indeed, there were differences in the biochemical features of catalase between wild-type and atg5 leaves; high levels of insoluble catalase were present in atg5 leaves ( Fig. 2D), and its enzyme activity was substantially lower than that of soluble catalase ( Fig. 2E), indicating that atg mutants accumulated an inactive form(s) of catalase. Furthermore, the redox status of a proportion of peroxisomes in the atg5 mutant was disturbed, showing a more oxidized state than wild-type ones (Fig. 2F). These results suggest that peroxisomes with electrondense regions in atg mutants are dysfunctional. We also found that peroxisomes in atg mutant leaves did not show the appropriate intracellular localization (Fig. 2G). Leaf peroxisomes are usually tightly attached to chloroplasts and/or mitochondria, where they cycle metabolites among these organelles during photorespiration. In wild-type leaves, almost 80% of peroxisomes were located within a distance of 0.1 mm from chloroplasts and/or mitochondria. In atg mutant leaves, however, more than 45% of peroxisomes were separated from these organelles and were often clustered together, which would also lead to a reduction of peroxisomal capability (Prestele et al., 2010). Leaves of atg mutants were indeed hypersensitive to the catalase inhibitor 3aminotriazole ( Fig. 2H), even though the levels of catalase protein were increased, and this result closely correlated with the accumulation of inactive form(s) of catalase and the mislocalization of peroxisomes in atg mutants. Consistent with our findings, atg mutants showed characteristic phenotypes similar to those observed in peroxisomal-functiondefective plants. Arabidopsis photorespiratory mutants show stunted or decelerated growth under standard atmospheric conditions but grow normally in air enriched in CO 2 , when photorespiration is suppressed (Reumann and Weber, 2006). Likewise, the decelerated growth phenotype observed in the atg5 mutant was partially suppressed under high CO 2 conditions (Fig. 2I). We also examined postgerminative seedling establishment on sugar-free medium as an index of peroxisomal ability. As Arabidopsis cotyledons catabolize fatty acids into sugars in glyoxysomes, a decline in peroxisomal activity would result in a lower percentage of seedlings developing true leaves on sugar-free medium (Fulda et al., 2004;Rylott et al., 2006). Indeed, the atg5 mutant showed a lower frequency of seedling development without supplemental sugars. Cotyledons of ,50% of atg5 seedlings did not fully expand and the production of true leaves was prevented, whereas this was not the case in wild-type seedlings (Fig. 2J). This growth defect in atg5 was rescued by exogenous application of sugars (Fig. 2K).
The accumulation of abnormal peroxisomes in atg mutants prompted us to examine whether autophagy transports peroxisomes to the vacuolar lumen for degradation. For this purpose, an atg5 transgenic was generated. In this mutant, the autophagic defect can be complemented through induction of ATG5 gene expression by treatment with 17-b-estradiol, and peroxisomes can be visualized by monitoring GFP fluorescence by confocal fluorescence microscopy. The number of peroxisomes was reduced in the transgenic plants after 17-bestradiol treatment (Fig. 3A, upper-right panel), whereas control leaves still retained a large number of peroxisomes (Fig. 3A, upper-left panel). To confirm that peroxisomes were delivered into the vacuolar lumen by autophagy, transgenic leaves were incubated in the dark after 17-b-estradiol treatment as GFP fluorescence cannot be detected in the vacuolar lumen in plants that are kept in the light (Tamura et al., 2003). The number of peroxisomes was reduced in 17-b-estradiol-treated leaves, and the vacuolar lumens in these cells contained GFP (Fig. 3A, lower right panel). When plant cells are treated with the V-ATPase inhibitor concanamycin A, autophagic bodies can also be detected inside the vacuolar lumen as small randomly moving vesicles (Yoshimoto et al., 2004). In concanamycin-A-treated cells without 17-b-estradiol treatment, in which autophagy is defective, peroxisomes were detected in the cytoplasm and did not show much movement, sometimes showing rectilinear motion along the cytoskeleton (supplementary material Movie 1). After 17-b-estradiol treatment, however, highly mobile peroxisomes were observed (supplementary material Movie 2), suggesting that peroxisomes were present in autophagic bodies within the vacuole. These results indicat that peroxisomes are delivered to the vacuole by autophagy for degradation.
We then sought to determine the mechanism by which selectivity in the degradation of plant peroxisomes by autophagy is guaranteed. On EM images, some of the peroxisomes in atg mutants appeared to be fragmented and consisted only of electron-dense material ( Fig. 2A,C). This might explain the marked accumulation of catalases in atg5 leaves (Fig. 1E). A proportion of peroxisomes containing protein aggregates appeared to be segregated or 'torn off' from the whole peroxisome by unknown mechanisms (Fig. 3B), thus dysfunctional peroxisomal proteins could be selectively degraded by autophagy. In addition, the data for the frequency of peroxisomes with electron-dense regions indicated that the increased number of peroxisomes in atg mutants was mainly due to the increase of peroxisomes with electron-dense regions  Table S1). (C) Immuno-EM images of wild-type and atg-mutant leaf peroxisomes using an antibody against catalase. Black dots represent gold particles labeled with anti-catalase antibody. The white dotted square indicates the enlarged area shown in the lower right panel. Scale bars: 1 mm. (D) High levels of accumulated insoluble catalase in atg5. The pellet samples were subjected to immunoblotting analysis with anti-catalase antibody. (E) The catalase activities in the soluble and insoluble fractions from atg5 leaves. The activity is expressed in arbitrary units as the mean 6 s.d. of three different experiments. (F) Redox states within peroxisomes in leaves of wildtype and atg5. Using redox-sensitive GFP (roGFP2), redox states within peroxisomes in leaves were compared between wild-type and atg5. Ratiometric images of the fluorescence excited by 405 and 488 nm lasers are shown. The color scale for the ratio values indicates the reduced state in dark color and the oxidized state in bright color. Scale bar: 10 mm. (G) Distributions of the distances between peroxisomes and chloroplasts and/or mitochondria. For all peroxisomes on randomly taken EM images the distance from each peroxisome to the nearest chloroplast and/or mitochondrion was measured (supplementary material Table S2) and the percentages at each distance are shown. No organelles, there were no chloroplasts or mitochondria around the peroxisomes. (H) Hypersensitivity of autophagy-defective Arabidopsis leaves to catalase inhibitor. Two-week-old seedlings grown on Murashige and Skoog (MS) medium without sucrose (MS-C) were transferred to MS-C with or without 10 mM 3-aminotriazole (3-AT) for an additional 14 days. The eighth or ninth leaves are shown. Scale bar: 1 cm. (I) The phenotype of atg5 under high-CO 2 conditions. Seedlings of wild-type and atg5 were grown hydroponically and supplied with standard nutrients under ambient (360 ppm CO 2 ) or elevated CO 2 (2000 ppm CO 2 ) conditions for 3 weeks. Scale bar: 1 cm. (J,K) The frequency of seedling establishment on medium without (J) or with (K) supplemental sugars (1%). Seedling establishment was scored 7 days after germination as the ability to develop true leaves. Error bars indicate the s.d. (n560, three independent experiments).
(supplementary material Table S1), suggesting that there is selective degradation of abnormal peroxisomes in wild-type leaves. This selectivity was further confirmed by the results of the EM analyses in this study. Examination of the fine structures of peroxisomes in atg mutants revealed an electron-lucent membrane structure, which was located adjacent to the endoplasmic reticulum (ER), occasionally present beside peroxisomes in atg2 leaves (Fig. 3C, black arrow). Surprisingly, these membrane structures were always tightly attached to the sites of the electron-dense regions (supplementary material Table S3). In a yeast atg2 mutant, many Atg proteins, including Atg8, accumulated at high levels in pre-autophagosomal structures (Suzuki et al., 2007), suggesting that an early autophagic structure called the 'isolation membrane' forms in the atg2 mutant but this mutant cannot complete autophagosome formation. In addition, it has been reported that the ER is one of the sources of isolation membranes in mammalian cells in culture (Hayashi-Nishino et al., 2009). These observations suggest that the electron-lucent membrane structures could be the seeds of the isolation membranes. In support of this suggestion, immuno-EM analysis using antibodies against Arabidopsis ATG8a, which is an autophagosomal membrane marker protein (Yoshimoto et al., 2004;Contento et al., 2005;Thompson et al., 2005), showed that ATG8 was localized on the electron-lucent membrane structures (Fig. 3D, left and middle panel, white arrows). Furthermore, we found that ATG8 was specifically localized on the surface of the electron-dense peroxisome (Fig. 3D, right and middle panel, white arrowheads). This finding is consistent with a previously proposed model of the selective autophagy pathway, in which ATG8 first recognizes a target by direct interaction with an ATG8-family-interacting motif and is then conjugated to phosphatidylethanolamine to recruit and expand the isolation membrane in the proximity of the target (Kuma et al., The behavior of peroxisomes in conditionally complementable atg5 transgenic leaves. Leaves from the atg5 transgenic harboring both the Arabidopsis ATG5 cDNA driven by an estradiolinducible promoter and the gene for GFP-PTS1 were incubated with 17-b-estradiol to artificially regulate autophagic ability under light (upper panels) and dark (lower panels) conditions. The green dot signal indicates peroxisomes. The red signal in the upper panels indicates chlorophyll autofluorescence. The red signal in the lower panels indicates vacuolar membranes stained by FM 4-64. Scale bars: 25 mm, upper panels; 10 mm, lower panels. (B-D) Immuno-EM (B,D) and conventional EM (C) pictures of atg2 leaves. Samples for (B,D) were prepared by high-pressure freezing and freeze substitution to observe intracellular fine structures in a state as close as possible to that existing in vivo. Electron-lucent membrane structures and gold particles labeled with anti-ATG8a antibodies are indicated by the black arrow (C) and white arrows or arrowheads (D), respectively. Ch, chloroplast; IM, seed of isolation membrane; Mt, mitochondrion; Per, peroxisome; Per-ED, electron-dense region in peroxisome; Vac, vacuole. Scale bars: 500 nm.
2007; Yamaguchi et al., 2010). Taken together, these observations indicate that dysfunctional peroxisomes must be enwrapped in autophagosomes following the recognition of the sites of electrondense regions by the machinery responsible for autophagosome formation.
Our results show that autophagy maintains the quality of peroxisomes in an organ-specific manner, which is probably dependent on peroxisomal functions, by the removal of dysfunctional peroxisomes to the vacuole for degradation. Several lines of evidence suggest that Arabidopsis has some selective mechanism by which autophagy specifically recognizes dysfunctional peroxisomes. The observation that autophagosomal structures precisely recognize the sites of electron-dense regions, which are most likely aggregates of dysfunctional proteins, is reminiscent of the selective autophagic degradation of damaged mitochondria in mammals (Youle and Narendra, 2011). Thus, the molecular mechanism underlying the selectivity might involve protein-protein interactions between cargo receptor(s) and adaptor(s) and autophagosomal membrane protein(s) . In plants, the peroxisomal membrane can also be heavily damaged by ROS produced by peroxisomal oxidative reactions during the life cycle. Therefore, damaged membranes might be recognized by peroxisomal-membrane-localized receptor and adaptor protein(s). In mammalian cells, it has been proposed that the recognition of ubiquitinylated peroxisomal membrane proteins by p62 (SQSTM1) and/or its homolog NBR1 (next to BRCA1 gene 1) confers selectivity for peroxisome degradation by autophagy (Kim et al., 2008;Deosaran et al., 2013). Although plant NBR1, which possesses hybrid properties of mammalian NBR1 and p62, has been identified (Svenning et al., 2011;Zientara-Rytter et al., 2011), it is still not known whether it has a role in pexophagy in plants. Because our western blot analysis using isolated leaf peroxisomes does not show a substantial increase in ubiquitinylated proteins in atg2 peroxisomes (data not shown), and immuno-EM analysis using anti-ubiquitin antibodies does not show ubiquitinylation of peroxisomal membrane proteins near the electron-dense regions (data not shown), the mechanism of selective peroxisome degradation in plants might be different from that in mammals. In addition, pexophagy must have plant-specific physiological significance. Through selective elimination of dysfunctional peroxisomes during peroxisomal oxidation processes, autophagy appears to play crucial roles in the early development of oilseed plants and the maintenance of photosynthetic function. Only plants, which are sessile and photosynthetic organisms, would have evolved this type of pexophagy as a unique strategy to survive conditions of severe environmental stress.

Catalase activity measurements
Leaf samples were taken from approximately the same position of the respective leaf and were homogenized on ice in a buffer consisting of 100 mM Tris-HCl pH 8.0, 20% glycerol and 30 mM DTT. The homogenates were centrifuged at 17,750 g for 15 min to generate supernatants and pellets. The resulting supernatant samples and pellet samples that were rinsed twice with protein extraction buffer were subjected to measurement of catalase activities spectrophotometrically, by monitoring the changes of H 2 O 2 at OD 240 at room temperature as described previously (Durner and Klessig, 1996). The catalase activities in soluble and insoluble fractions were normalized to the protein levels of catalase, which were determined by immunoblotting analysis using anti-catalase antibodies.

Comparison of redox state within peroxisomes
To compare the redox state within peroxisomes between wild type and atg5, we generated transgenic wild-type and atg5 plants expressing redox-sensitive GFP (roGFP2) fused with PTS1 under the control of the cauliflower mosaic virus (CaMV) 35S promoter. The leaves were observed by confocal microscopy. roGFP2 was exited with 405 and 488 nm lasers and the fluorescence was measured with a band pass filter of 505-530 nm. Then ratiometric images of 405:488 were generated as described previously (Meyer et al., 2007).

Construction of the estradiol-inducible complementation system
The ATG5 coding region (nucleotides 1-1014) was amplified by PCR from full-length cDNA using the primers 59-TCCCCCGGGAT-GGCGAAGGAAGCGG-39 and 59-GCTCTAGATCACCTTTGAGGA-GCTTTC-39. The resulting PCR product was inserted into the EcoRV site of the pBluescript II vector and the resulting plasmid was designated as pHH52. To construct a plasmid for generating stable transformants expressing ATG5 cDNA driven by the estradiol-inducible promoter, pHH52 was digested with ApaI and SpeI and was ligated into the same sites of pER8 (Zuo et al., 2006). The construct was verified by sequencing and was introduced into atg5-1 Arabidopsis plants by the floral-dip method of in planta Agrobacterium-tumefaciens-mediated transformation (Clough and Bent, 1998). Chemical treatments were carried out both by injections using a syringe without needle and by submerging tissue in liquid medium containing 10 mM 17-b-estradiol (Sigma, St Louis, MO).

RT-PCR
RT-PCR for catalase genes and peroxisome division and proliferationrelated genes, PEX11a-11e, was performed as described previously (Yoshimoto et al., 2004). cDNA derived from 0.2 mg of total RNA was used as the template for PCR. The gene-specific primers used are described in supplementary material Table S4.