The APG8/12-activating Enzyme APG7 Is Required for Proper Nutrient Recycling and Senescence in Arabidopsis thaliana *

The vacuole/lysosome serves an important recycling function during starvation and senescence in eukaryotes via a process called autophagy. Here bulk cytosolic constituents and organelles become sequestered in specialized autophagic vesicles, which then deliver their cargo to the vacuole for degradation. In yeasts, genetic screens have identified two novel post-translational modification pathways remarkably similar to ubiquitination that are required for autophagy. From searches of the Arabidopsisgenome, we have identified gene families encoding proteins related to both the APG8 and −12 polypeptide tags and orthologs for all components required for their attachment. A single APG7gene encodes the ATP-dependent activating enzyme that initiates both conjugation pathways. Phenotypic analysis of anAPG7 disruption indicates that it is not essential for normal growth and development in Arabidopsis. However, theapg7-1 mutant is hypersensitive to nutrient limiting conditions and displays premature leaf senescence. mRNAs for both APG7 and APG8 preferentially accumulate as leaves senesce, suggesting that both conjugation pathways are up-regulated during the senescence syndrome. These findings show that the APG8/12 conjugation pathways have been conserved in plants and may have important roles in autophagic recycling, especially during situations that require substantial nitrogen and carbon mobilization.

Plants, like other organisms, have developed sophisticated mechanisms for recycling intracellular constituents during periods of growth, developmental remodeling, and nutrient-limiting conditions (1,2). Especially critical is the degradation of protein, given the importance of reused amino acids to the nitrogen and carbon economy. Selective protein removal is accomplished primarily by the ubiquitin (Ub) 1 /26 S proteasome pathway (3,4). In this pathway, the covalent attachment of Ubs is used as a signal to target specific proteins for degrada-tion by the 26 S proteasome. Another major recycling system employs the vacuole, or its animal equivalent, the lysosome, as a lytic organelle (5,6). Here cytosolic proteins are delivered to the vacuole and then degraded by a wide variety of vacuolar proteases. In some situations, entire organelles can be targets (7). Unlike the Ub/26 S proteasome pathway, vacuolar proteolysis is for the most part non-selective, thus targeting proteins indiscriminately. As a consequence protein turnover by the vacuole is thought to play less of a role in cellular regulation and a more prominent role under conditions when rapid remobilization and resorption of nutrients are crucial. These conditions include nutrient deprivation and environmental stress and developmental periods that require extensive cell and organ remodeling such as senescence and programmed cell death (5,6).
The vacuole degrades cytoplasm by three unique but overlapping mechanisms, chaperone-assisted import, microautophagy, and macroautophagy (6 -9). Chaperone-assisted import is activated during starvation; it employs Hsp70-related proteins that help transport individual proteins directly into the vacuole without the use of a vesiculated intermediate (9). Microautophagy defines the sequestration of cytosol by invaginations of the tonoplast, thus forming small intravacuolar vesicles. These vesicles and their cargo are then broken down by resident vacuole hydrolases (7). In methylotrophic yeasts for example, microautophagy is important for removing peroxisomes in a process called pexophagy, as the cells switch from methanol to glucose metabolism (6,7). Macroautophagy involves the engulfment of large portions of cytoplasm by vesicles called autophagosomes (5,6,8). These vesicles appear to arise from the endoplasmic reticulum as provacuole tubules; as the tubules elongate, they form long threaded structures that eventually encircle portions of cytoplasm in a cage-like structure. Autophagosomes are then created when these threads fuse to seal completely the region in a double-membrane compartment. The outer membrane of the autophagosome fuses with the tonoplast thus releasing the internal vesicle into the vacuole where its cargo is degraded in a process similar to microautophagy.
Although both macro-and microautophagy have been observed and their morphological characteristics have been described in plants (10 -16), the molecular mechanisms underpinning these processes are not understood. One of the main barriers has been the paucity of molecular markers for autophagic vesicles and lack of mutants that impair their assembly and delivery. To date, the only known plant vacuole mutant is vacuoleless-1, which encodes an essential gene required for organelle assembly (17). However, substantial progress has been made recently with yeasts using genetic screens to identify autophagy-deficient mutants. These include mutants in Saccharomyces cerevisiae sensitive to starvation (autophagy or apg (18,19)) or defective in cytoplasm to vacuolar targeting (cvt (20)), and the isolation of mutants in the methylotrophic yeast Pichia pastoris blocked in pexophagy (glucose-stimulated autophagy or gsa (21)). Many of the apg/cvt/gsa mutants simultaneously abrogate microautophagy, macroautophagy, and pexophagy, indicating that the corresponding proteins participate in steps common to each of these processes. However, the mutants still retain functioning vacuoles indicating that they do not alter vacuolar ontogeny.
Characterizations of a subset of yeast apg/cvt/gsa mutants have identified two novel protein conjugation pathways required for autophagy (22,23). In S. cerevisiae, the tags are APG8 and Ϫ12, small proteins of 117 and 186 amino acids, respectively. Although bearing little sequence similarity to Ub, APG8 and Ϫ12 become attached to other cellular factors by an enzymatic cascade remarkably similar to ubiquitination (8,24). As shown in Fig. 1, the first step of each pathway requires the activity of APG7 (also known as CVT2 and GSA7). This E1 activates both APG8 and Ϫ12 by first directing the formation of a phosphoanhydride bond between the AMP moiety of ATP and the C-terminal glycine of the tag. Then APG7 becomes linked to the tag via a high energy thiol ester bond between the glycine moiety of the tag and a unique cysteine in APG7. APG7 subsequently donates the activated tag via transesterification to a unique cysteine within an E2. For APG8 and Ϫ12, this E2 activity is provided by APG3 and Ϫ10, respectively. The E2bound tag is then donated to appropriate targets. In yeast, APG12 becomes attached to a single target, APG5, via an isopeptide bond between the C-terminal glycine of APG12 and a single lysine (Lys-149) in APG5 (22). In contrast, APG8 is bound to phosphatidylethanolamine (PE) via an amide bond between the C-terminal glycine of APG8 and the terminal amino group in this lipid (23). By an unknown sequence of events, production of the APG12-APG5 and APG8-PE conju-gates then promotes the formation of autophagic vesicles and their subsequent delivery to the vacuole. The yeast APG8 pathway also requires an additional protease step directed by APG4, which removes an extra C-terminal amino acid (arginine) in the APG8 precursor to expose the glycine essential for conjugation ( Fig. 1).
Orthologs for several yeast APG proteins have been discovered recently in animals, suggesting that this autophagic system is present in all eukaryotes (21,25,26). As a consequence, we attempted to find similar proteins in plants through searches of available genome sequences. Here we report that potential orthologs for all components of the APG8 and Ϫ12 conjugation pathways are present in Arabidopsis thaliana (see Fig. 1). An Arabidopsis T-DNA mutant that interferes with the accumulation of APG7, the ortholog of the yeast E1 required for both APG8 and Ϫ12 conjugation, is phenotypically normal under non-stressed conditions, suggesting that these two pathways are not essential in plants. However, the mutant plants show accelerated senescence and are highly sensitive to nutrient-limiting growth conditions, suggesting that these conjugation pathways become important when carbon and nitrogen remobilization is essential. The discovery of these APG genes/ proteins may now provide important new tools to dissect the autophagic process in plants at the genetic, cytological, and biochemical levels.

EXPERIMENTAL PROCEDURES
Identification and Sequence Analysis of APG Genes-The A. thaliana genomic and expressed sequence tag (EST) data bases (genome-www. stanford.edu/Arabidopsis/) were searched by BLAST (27) for potential APG genes using the sequences of yeast and human proteins as queries. A full-length cDNA for AtAPG12a (162B16T7) and a genomic BAC fragment (T14L9) containing AtAPG7 were obtained from the Arabidopsis Biological Resource Center (Ohio State University, Columbus). The complete coding sequence of APG7 was determined from a reversetranscribed (RT) PCR clone generated with Pfu polymerase and the primers CATATGCACCATCATCATCATCACATGGCTGAGAAAGAA-ACTCCAGCA (predicted start codon underlined) and GAAGCCACCT-AGTGAATATAGTCTATGGAC using A. thaliana Col-0 RNA as the template. The forward primer was designed to add codons for an Nterminal His 6 tag. The product was first ligated into the EcoRV site of pBluescript; the NdeI-Acc65I fragment of this plasmid was isolated and cloned into the corresponding sites of pET32 (Novagen, Madison, WI) for expression of recombinant APG7 protein in Escherichia coli (see below). The complete coding sequences for AtAPG8a and AtAPG8h were determined from RT-PCR clones generated with the primers TCGCATATGGCTAAGAGTTCCTTCAAGATCTCT and GTGAGCTCA-AGCAACGGTAAGAGATCCAAAGT, and TCGCATATGAAATCGTTC-AAGGAACAATACA and GTGAGCTCAACCAAAGGTTTTCTCACTGC-TA, respectively (predicted start and stop codons underlined). DNA sequences were determined by the PCR-based dideoxy method (PerkinElmer Life Sciences). Intron/exon boundaries were identified by comparing cDNA and EST sequences with their respective genomic DNA sequences or predicted by comparing related genes. Amino acid sequence comparisons and phylogenic analyses were performed using MACBOXSHADE (Institute of Animal Health, Pirbright, UK) and ClustalW (www.ebi.ac.uk/clustalw), respectively. The GenBank TM accession numbers for the sequences described in this article are AF492761 (AtAPG7), P38862 (ScAPG7), NP_006386 (HsAPG7), AF492759 (AtAPG8a), AF492760 (AtAPG8h), NP_009475 (ScAPG8), NP_009216 (HsAPG8), AF492758 (AtAPG12a), P38316 (ScAPG12), and O94817 (HsAPG12).
Isolation and Complementation of apg7-1-The apg7-1 insertion mutant was identified by a PCR screen of a T-DNA-transformed population of A. thaliana ecotype WS generated by Dr. Kenneth Feldman (DuPont) using the 5Ј-CAATCTCTCAGAAAGATAAGATCAGCCATG or 3Ј-GAA-GCCACCTAGTGAATATAGTCTATGGAC gene-specific primers in combination with either a left border or right border T-DNA-specific primer (28). The presence of this T-DNA in subsequent generations was followed by PCR and by kanamycin resistance conferred by the T-DNA. The apg7-1 mutant was backcrossed 3 times to the wild-type WS ecotype to help eliminate secondary mutations and then the homozygous apg7-1 mutant was obtained by segregation analysis.

FIG. 1. Description of the APG conjugation pathways in yeast.
By ATP-dependent reaction cascades involving an E1 and E2s, APG12 and APG8 become covalently attached to their respective targets APG5 and phosphatidylethanolamine (PE). Prior to activation, the APG8 precursor is processed by the APG4 protease to remove the extra Cterminal residues and thus expose the penultimate glycine. Both APG12 and Ϫ8 are activated by the E1-APG7 and become bound to APG7 by a thiol ester bond. The activated tags are then donated to their respective E2s, APG10 and APG3 by transesterification, and finally attached to their targets by an amide bond. For APG5, a specific lysine (K) is involved which forms an isopeptide bond with APG12. For PE, the amide group of ethanolamine is used for attachment to APG8. ment from BAC T14L9 encompassing the APG7 gene was ligated into the KpnI site of pCAMBIA 1301 (Medical Research Council Laboratory of Molecular Biology, Cambridge, UK). The APG7C-S mutant with the Cys-558 codon substituted for that encoding Ser was generated by the QuickChange method (Stratagene, La Jolla, CA) using the primers CGAACTCTAGACCAACAAAGCACTGTTACACGCC and GGCGTG-TAACAGTGCTTTGTTGGTCTAGAGTTCG (altered nucleotides underlined) and a 1.5-kb BamHI-EcoRV fragment of APG7 cloned into pBluescript as the template. The presence of the desired mutation and the absence of secondary mutations were confirmed by DNA sequencing. Both pCAMBIA1301 APG7 constructs were introduced into Agrobacterium tumefaciens strain C81 (pMP90) and then transformed into hemizygous apg7-1 plants using the floral dip method (29). Transformed seedlings (T1) were selected on medium containing 15 mg/liter hygromycin B. T2 plants homozygous for both the apg7-1 allele and the APG7 transgene were identified by testing progeny for growth on 25 mg/liter kanamycin and 15 mg/liter hygromycin B and confirmed by PCR. T3 seeds were used for the phenotypic analyses.
Two weeks after germination, plants growing in a short-day photoperiod were transferred to soil and grown until seed maturation. Seeds collected from individual plants were weighed and counted. One-weekold plants growing in liquid medium under continuous light were washed two times with modified Murashige and Skoog medium lacking nitrogen and sucrose (nitrogen/carbon (N/C)-free medium: Murashige and Skoog micronutrient salts (Invitrogen), 3 mM CaCl 2 , 1.5 mM MgSO 4 , 1.25 mM KH 2 PO 4 , 5 mM KCl, 2 mM MES (pH 5.7)) and then grown in this N/C-free medium. One-week-old plants grown on Gamborg B5 agar were transferred to N/C-free agar medium for further growth and subsequently transferred back to Gamborg B5 agar. For the detached leaf assay, first and second true leaves were detached with a sharp blade from 2-week-old plants grown under a long-day photoperiod on Gamborg B5 medium and placed in the dark with their abaxial sides down on 3MM paper wetted with 3 mM MES (pH 5.7) with or without 1 M kinetin (Sigma). Leaves were collected at the indicated times, blotted dry, frozen in liquid nitrogen, and used for chlorophyll, protein, and RNA isolation.
Chlorophyll Extraction and Measurement-Chlorophyll was extracted from frozen leaves into 1 ml of 96% EtOH by overnight incubation at 4°C and quantified spectrophotometrically (30). Values given were per leaf and were expressed as a percentage of the maximum chlorophyll level for each set of treatments.
Protein Isolation and Immunoblot Analysis-Total soluble protein was isolated from detached leaves or soil-grown plants by homogenization in 50 mM Tris (pH 8.0), 1 mM Na 4 EDTA, 1% Triton X-100, 100 mM KCl, and 30 mM Na 2 S 2 O 5 , and the extracts were clarified by low speed centrifugation (31). Proteins were subjected to SDS-PAGE and either stained with silver or transferred to Hybond-C Extra membranes (Amersham Biosciences) for immunoblot analysis using alkaline phosphatase-labeled goat anti-rabbit immunoglobulins (Kirkegaard & Perry Laboratories, Gaithersburg, MD) for detection. Sample sizes were adjusted to reflect a per leaf basis. For production of anti-APG7 antibodies, pET32His6APG7 was introduced into E. coli BL21 Codon Plus (Stratagene) for expression of recombinant APG7. Following a 3-h induction of log phase cells with 1 mM isopropyl-␤-D-thiogalactoside, inclusion bodies were collected, solubilized in urea, and injected into rabbits (Polyclonal Antibody Service, Madison, WI). Antibodies against the large subunit of spinach Rubisco and the CH-42 subunit of pea magnesium-protoporphyrin chelatase (32) were provided by Drs. Archie Portis and Tanya Falbel, respectively. The anti-PBA1 and RPT2 antibodies were as described (33).

RESULTS
Detection of the APG Genes in Arabidopsis-By reiterative BLAST searches, we attempted to find counterparts to many of the yeast APG/CVT/GSA genes (6,24) in the near-complete genome data base of Arabidopsis (www.arabidopsis.org/ BLAST). Remarkably, potential orthologs for almost all loci were discovered including APG1-10, Ϫ12-13 (Figs. 2-4 and data not shown). For clarity, we have adopted the same nomenclature developed by Ohsumi (24) for these APG loci. Unlike yeast where each protein is encoded by a single gene (24), many of the Arabidopsis proteins are encoded by two or more loci, suggesting a greater complexity of the APG system in plants. However, not all loci were detected; notably absent were obvious counterparts to APG14, Ϫ16, and Ϫ17, suggesting that they have substantially diverged during evolution. Multiple cDNAs for many of the APG genes were also detected in various EST collections from Arabidopsis, rice, soybeans, Medicago, tomato, potato, corn, barley, wheat, sorghum, and Chlamydomonas to name a few, implying that this autophagic system may be active in all plants.
Importantly, a collection of Arabidopsis genes predicted to encode all components of the APG8 and Ϫ12 conjugation pathways was identified (8,24). These include loci for APG8 and Ϫ12, the E1-APG7, the E2s-APG3 and Ϫ10, the target APG5, and the protease APG4 (see Fig. 1). In all cases, the positionally important amino acids necessary for enzymatic activity and/or function are present. Nine loci encode a family of APG8 proteins with 69 -96% amino acid sequence similarity to each other and 73-90% similarity to yeast APG8 (Fig. 2B). Intron/ exon arrangements are also similar among the Arabidopsis gene family, indicating that this family arose from a common progenitor ( Fig. 2A). Like yeast and human APG8 (23,26), seven of the Arabidopsis APG8 proteins (APG8a-g) contain additional residues beyond the predicted mature C terminus that presumably must be removed post-translationally by APG4 to expose the glycine necessary for conjugation. Phylogenic analysis of the APG8 family revealed that APG8a-g cluster into one clade of 84 -96% amino acid sequence similarity that also includes yeast APG8 (Fig. 2C). APG8h and -i grouped into a separate clade; they are only 69 -75% similar to members of the other clade and lack extra residues that protect the C-terminal glycine. ESTs are available for seven of the nine APG8 genes, the exceptions being APG8f and APG8i, indicating that most are expressed.
Two Arabidopsis genes were identified that encode 94-and 96-amino acid proteins related to yeast APG12 (Fig. 3A). A full-length EST was identified in the data bases for only APG12a, which was used to verify the correct coding region. No ESTs were available for APG12b; its gene structure was pre-dicted by alignment with APG12a and its yeast and animal orthologs. The predicted amino acid sequences for APG12a and -b are 95% similar to each other and 36 -45% similar to their yeast and human orthologs (22,36). Both Arabidopsis isoforms end in the C-terminal glycine essential for conjugation, indi-cating that post-translational processing is not needed to generate active APG12 proteins. The human and yeast orthologs of Arabidopsis APG12a and -b are substantially larger, due to the presence of long hydrophilic N-terminal extensions of 45 and 90 residues, respectively (22,36). The function of these extensions is not yet known.
The yeast APG8 and Ϫ12 conjugation pathways both require an E1 activity directed by APG7 (21,22,26,37) (Fig. 1). In Arabidopsis, we detected a single APG7 gene. It contains 10 introns and 11 exons and encodes a 697-amino acid protein with 56 and 62% similarity to yeast and human APG7, respectively (Figs. 4A and 5A). Consistent with its predicted enzymatic activity, the Arabidopsis APG7 protein contains the positionally conserved cysteine (Cys-558) expected to form the thiol ester linkage with APG8 and Ϫ12 (21,22), and the consensus nucleotide-binding site ( 364 GXGXXG 369 ) possibly required for either binding ATP or the activated AMP tag intermediate (Fig. 4A). By using antibodies directed against recombinant APG7, we could detect a protein of the expected size (76 kDa) in crude extracts from various Arabidopsis tissues of both etiolated and green seedlings (Fig. 4B). This distribution suggests that APG7 and the corresponding APG8 and Ϫ12 conjugation pathways are active in all cell/tissue types.
Isolation of the apg7-1 Mutant-To help determine whether the APG8/12 conjugation pathways function in plants, we searched the available Arabidopsis T-DNA mutant populations for insertions affecting APG7 given its predicted central and essential role in both pathways (see Fig. 1). One useful allele was apg7-1 which contains a T-DNA sequence within the gene (Fig. 5A). Sequence analysis of the T-DNA-APG7 junctions indicated that the element inserted into the 10th and final intron of the gene and created a 20-bp deletion. A left border sequence was present at both ends of the T-DNA, and a right border sequence was present at the 3Ј end internal to the left border. Kanamycin resistance conferred by the NPTII coding region associated with the T-DNA segregated in a 3:1 pattern, demonstrating that a single insert was present. Despite the presence of the T-DNA, we could still detect by RT-PCR an APG7 transcript in homozygous apg7-1 plants (Fig. 5B). The apg7-1 mRNA could be amplified by a primer pair upstream of the insertion site but not by a primer pair in which one was downstream of the insertion site indicating that a truncated APG mRNA was expressed. Despite its apparent synthesis we could not detect the transcript by RNA gel blot analysis, sug-gesting that the truncated form is unstable (Fig. 5C). Because the T-DNA was predicted to affect only a short section of the coding sequence and was downstream of the codons for the active-site cysteine and the putative nucleotide-binding domain, it was possible that an active protein could still be expressed. However, we could not detect by immunoblot analysis a shorter protein of the expected size (70 kDa) in homozygous apg7-1 seedlings (Fig. 5D). Taken together, it appears that the apg7-1 allele severely impairs accumulation of the APG7 protein.
The apg7-1 Mutant Senesces Early and Is Hypersensitive to Nutrient-limiting Conditions-Despite the absence of detectable APG7 protein, homozygous apg7-1 seedlings underwent normal embryogenesis, germination, and seedling development ( Fig. 6A and data not shown). They also grew at the same rate  Total RNA isolated from wild-type (WT), apg7-1, and apg7-1 complemented with either APG7 or the APG7 active-site mutant (C-S) was subjected to RT-PCR using either the 1 ϩ 2 primer pair or the 3 ϩ 4 primer pair. As a control, genomic DNA from WT was included. Failure to detect the product of the 1 ϩ 2 primer pair in apg7-1 showed that the last exon was absent in mutant APG7 mRNA. C, RNA gel blot analysis of WT, apg7-1, and apg7-1 complemented with either APG7 or the APG7 active-site mutant C-S to detect transcripts of APG7. Ten g of total RNA isolated from young seedlings was separated by denaturing gel electrophoresis, transferred to a membrane, and hybridized with coding region probes for APG7 and CDC2a. D, immunoblot analysis with anti-APG7 antibodies of WT, apg7-1, and apg7-1 complemented with either APG7 or the APG7 active-site mutant (C-S). Total soluble protein was isolated from young seedlings and separated by SDS-PAGE. Recombinant His 6tagged APG7 is included on the right. as wild-type and flowered at the same time under both shortand long-day photoperiods when we used standard nutrientrich media and soil conditions. Collectively, the data indicated that the APG7 protein, and by inference the APG8/12 conjugation pathways, are not essential for Arabidopsis growth and development.
However following bolting, the apg7-1 plants showed accelerated senescence, characterized by premature chlorosis of the mature rosette leaves (Fig. 6B). This phenomenon could be seen for apg7-1 plants grown under either short-or long-day photoperiods on well fertilized soil but was dramatically enhanced when the plants were grown under short days on soil deficient or depleted in nitrogen. These early senescent plants also displayed reduced fecundity. The mutant plants developed fewer inflorescence branches, and a lower proportion of the siliques expanded and became filled with seeds when grown on nutrient-depleted soils under a short-day photoperiod (Fig. 6,  C and D).
Young apg7-1 seedlings were also hypersensitive to nutrientlimiting growth conditions. When grown under continuous light in liquid medium free of nitrogen and carbon, the apg7-1 plants slowly lost chlorophyll and ceased growing, whereas the WT plants remained green and continued to grow, albeit at a reduced rate (Fig. 7A). When young seedlings were grown on N/C-free agar medium, accelerated chlorosis was also evident. Whereas wild-type seedlings retained some chlorophyll and continued to generate new leaves, the apg7-1 seedlings dramatically slowed their growth rate, bleached completely, and accumulated anthocyanins in the apex within 2 weeks on such poor growth conditions (Fig. 7B). This slower growth could be observed by measuring root elongation. After 3 weeks growth on N/C-free agar medium, the primary root of wild-type seedlings averaged 11.6 cm in length as compared with 4.3 cm for apg7-1 seedlings. Wild-type plants could be rescued from such nutrient deprivation by transfer back to N/C-rich medium; even after exposure for up to 40 days to such poor conditions, 80% of the plants resumed growth upon retransfer (Fig. 7C). In contrast, the apg7-1 plants were irreversibly affected, and when transferred back to N/C-rich media, they often did not recover. A time course of the effect showed a sharp transition between survival and death; only 20% of apg7-1 seedlings survived incubations on N/C-free medium longer than 14 days (Fig. 7C).
Given a possible role of APG7 in vacuolar ontogeny and structure through its presumed role in the APG8/12 conjugation system (8,24), we examined microscopically if the apg7-1 mutation affected gross vacuolar morphology. As can be seen in Fig. 8, the size and shape of the central vacuoles of apg7-1 plants were indistinguishable to wild-type as determined by staining protoplasts with two vacuolar dyes, neutral red (NR) and monochlorobimane (MCB) (13,35). Swanson et al. (13) reported the detection of lytic vacuoles as well as the central vacuole when barley aleurone protoplasts were stained with MCB. Here we failed to see these smaller lytic vacuoles and thus were unable to determine whether the apg7-1 mutant affected their production or distribution.
Complementation of apg7-1-Proof that the apg7-1 phenotypes were directly caused by a reduction in APG7 protein activity was obtained by complementation. apg7-1 plants were transformed with a genomic fragment encoding either wildtype APG7 or a mutant where the active-site cysteine was replaced by a serine (APG7C-S) and expressed under the control of the native APG7 promoter. Transformed plants were identified by the hygromycin resistance gene, which was cotransformed with the APG7 transgene. The T1 plants were allowed to self-pollinate, and progenies homozygous for both the apg7-1 allele and the APG7 transgene were identified by PCR. As shown in Fig. 5, C and D, high levels of APG7 mRNAs and proteins were apparent. When transformed plants were examined phenotypically, those harboring the APG7C-S construction behaved like apg7-1 plants (i.e. accelerated senescence, reduced seed production, and hypersensitivity to N/C-free media), whereas those transformed with the unmodified APG7 construction behaved like wild-type plants (Figs. 6 and 7). These combined results not only proved that loss of APG7 was the direct cause of the phenotypes but also implied that the E1 activity of the protein was required for successful complementation.

Molecular Analysis of the Accelerated Senescence
Phenotype of apg7-1 Plants-To help define at the molecular level the accelerated senescence phenotype conferred by the apg7-1 mutation, we used a detached leaf assay (38). Here young leaves were collected from 2-week-old Arabidopsis seedlings and incubated in the dark. As shown in Fig. 9, wild-type leaves senesced within a week, exemplified by the loss of chlorophyll and total protein. Immunoblot analysis of individual proteins showed a concomitant reduction in two photosynthetic markers, Rubisco and magnesium protoporphyrin chelatase (32) (Fig. 9B). Surprisingly, numerous proteins were relatively resistant to such degradation including two subunits of the 26 S proteasome, PBA1-encoding a ␤ subunit of the 20 S proteolytic core, and RPT2-encoding an AAA-ATPase of the 19 S regulatory particle (39,40). The apg7-1 mutant senesced faster, as detected by an accelerated loss of chlorophyll and the two photosynthetic markers (Fig. 9, A and B). Complementation of the apg7-1 plants with wild-type APG7 protein but not the APG7C-S mutant reversed all the effects as defined by a slower loss of chlorophyll, Rubisco, and magnesium protoporphyrin chelatase (data not shown). During the time course of senescence, the levels of APG7 did not decrease and, in fact, appeared to increase slightly in the detached leaves. This effect was seen both for wild-type plants and for apg7-1 plants complemented with the APG7 gene under control of its own promoter (apg7-1/APG7) (Fig. 9B).
The continued accumulation of APG7 protein in senescing leaves of wild-type plants despite an overall loss of chlorophyll and total protein suggested that the APG8/12 conjugation pathways are up-regulated during senescence. To test this possibility, we extracted total RNA from detached WT and apg7-1 leaves undergoing dark-induced senescence and subjected it to RNA gel blot analysis. To better reflect the actual levels of RNA in the leaves, we equalized the loads of total RNA based on the same number of leaves. As can be seen in Fig. 10, rRNAs (as detected by EtBr staining) were degraded during senescence, and this breakdown was exacerbated by the apg7-1 mutation. In fact by day 7, most of the rRNAs were converted into smaller degradation intermediates in the apg7-1 leaves. A parallel loss of CDC2a mRNA, (encoding a cyclin-dependent kinase (41)) was also observed. This suggested that other mRNAs were also degraded as senescence progressed, the rate being enhanced by the apg7-1 mutation (Fig. 10). As expected CAB mRNA also rapidly disappeared following the dark incubation consistent with the instability of the CAB mRNA and the light requirement for continued CAB transcription (42). Conversely, levels   FIG. 7. Enhanced sensitivity of apg7-1 plants to low nitrogen  and carbon (N and C) growth conditions. Plant lines include wildtype (WT), apg7-1, and apg7-1 complemented with either APG7 or the APG7 active-site mutant (C-S). A, plants were grown for 1 week in Nand C-rich liquid medium (0 days) and then transferred to N-and C-free liquid medium for an additional 32 days. B, plants were grown for 1 week on N/C-rich agar medium and then transferred for 2 weeks growth on N/C-free agar medium. C, survival of plants grown for 1 week on N/C-rich agar medium and then transferred to N/C-free agar medium for various lengths of time before transfer back to N/C-rich agar medium. The percentage of plants that survived this regime as determined by resumption of growth is plotted versus days grown on N-and C-free medium. FIG. 8. Vacuoles appear normal in the apg7-1 mutant. Protoplasts from young leaves of 2-week-old wild-type (WT) and apg7-1 seedlings were either stained with NR and visualized by light microscopy or with monochlorobimane (MCB) and visualized by fluorescence microscopy. The light microscopic images of the MCB-stained protoplasts are shown to the left.
of the senescence-associated SEN1 mRNA increased during the senescence time course. Similar to the observations of Weaver et al. (38) SEN1 transcript levels increased substantially in wild-type leaves well before detectable chlorosis (2.5 days) and then decreased gradually as senescence progressed. For the apg7-1 plants, a higher basal level of SEN1 mRNA was detected before leaf detachment. This level increased further during the senescence process followed by a substantial loss of transcript that paralleled the loss of total RNA. Like SEN1, transcript abundance for APG7 and Ϫ8 increased markedly during detached leaf senescence although not as rapidly (Fig. 10). (Given the high nucleotide identity among the nine Arabidopsis APG8 genes, we assume that the APG8a probe used here recognized most if not all of the tran-scripts from this gene family.) Although the increase in APG7 mRNA was evident by day 5, the increase in APG8 mRNAs was evident as early as day 2.5, suggesting that APG8 mRNA levels respond more rapidly to the dark incubation. The highest amounts of APG7 and Ϫ8 transcripts were detected between days 5 and 7 of the senescence time course of wild-type leaves. For apg7-1 leaves, the maximum for the APG8 mRNAs was at day 5, which was then followed by a rapid loss that paralleled the catabolism of rRNAs and the CDC2a mRNA.
Effect of Cytokinins on the apg7-1 Phenotype-Cytokinins have been well described as inhibitors of leaf senescence which can effectively block the chlorosis of detached leaves at micromolar concentrations (43). To test if cytokinins can antagonize the effects of the apg7-1 mutation, we treated detached leaves senescing in the dark with various concentrations of kinetin, and we examined the rate of senescence by chlorophyll and protein measurements. For wild-type leaves, 1 M kinetin retarded the loss of chlorophyll and Rubisco but had a minimal effect of the magnesium-protoporphyrin chelatase ( Fig. 11 and data not shown). For the apg7-1 leaves, the inhibitory effect of kinetin was substantially diminished but still evident. Although delayed slightly, detached leaves treated with 1 and 10 M kinetin still lost chlorophyll and both photosynthetic proteins, suggesting that cytokinins cannot fully override the effects of the apg7-1 mutation (Fig. 11 and data not shown). Kinetin treatment had little effect on the levels of PBA1 and RPT2 and APG7 (Fig. 11B and data not shown). Only a slight decrease in APG7 protein was evident when wild-type plants were treated with 1 M kinetin. DISCUSSION Vacuolar autophagy has long been considered to be an important route for recycling various cellular constituents in yeasts and animals, especially during times of starvation and programmed cell death (6 -8). Ultrastructural analyses suggest a similar role for autophagy in plants (10 -16), but its exact functions have remained elusive given the paucity of molecular markers and mutants currently available. One of the essential components of the autophagic system in yeasts is a pair of conjugation pathways that attach the polypeptide tags APG8 and Ϫ12 to the lipid PE and the protein APG5, respectively. By an unknown mechanism these modifications trigger autophagosome assembly and/or fusion of the vesicles with the vacuole.
Here, we show that plants likely have a conjugation system related to yeast APG8 and Ϫ12 with the discovery of a similar set of APG genes in Arabidopsis. This set includes small gene families encoding the APG8 and Ϫ12 tags and one or more genes encoding the target of APG12 (APG5) or the enzymes required for the following: (i) processing of the APG8 precursor (APG4), (ii) the ATP-dependent activation of both APG8 and Ϫ12 (APG7), and (iii) conjugation of the tags to their respective targets (APG3 and Ϫ10) (see Fig. 1). All important catalytic domains in the yeast APG components are conserved in their Arabidopsis counterparts, strongly suggesting that the Arabidopsis pathways have similar mechanisms of catalysis. This possibility is further supported by our observations that complementation of the apg7-1 phenotype was successful with wild-type APG7 but not with a mutant version predicted to be enzymatically inactive (APG7C/S). Final proof will require demonstration that the various APG components have the proposed enzymatic activities and can produce the expected APG8-PE and APG12-APG5 conjugates in planta. Unfortunately, we have been unable thus far to generate enzymatically active Arabidopsis APG7 by recombinant methods to begin this analysis (data not shown). Attempts are now underway to detect the APG8 and Ϫ12 conjugates in crude plant extracts.
The comparable phenotypes of Arabidopsis, S. cerevisiae, and P. pastoris mutants disrupted in their respective APG7 loci (see Refs. 21, 22, and this report) suggest that the encoded proteins and by inference the associated APG8/12 conjugation pathways function in a similar fashion in plants. Thus, it is reasonable to propose that the Arabidopsis apg7-1 plants like their yeast counterparts are compromised in autophagic process(es) activated under nutrient limiting conditions (6,8). Clearly, proof of this connection will require an ultrastructural analysis of autophagy in apg7-1 and other Arabidopsis apg mutant lines. In both yeasts and Arabidopsis, APG7 is not essential. Provided the apg7-1 plants were grown in nutrientrich conditions, their phenotype was indistinguishable from wild type for most of their life cycle. Thus under normal physiological conditions, the Arabidopsis APG8/12 system likely contributes little to the N/C supply or that the plants can recycle cellular constituents by other routes when the APG8/12 system is compromised. These other routes could include the Ub/26 S proteasome pathway, other cytosolic and organellar proteolytic pathways (2), and/or possibly chaperone-assisted vacuolar import (9), a pathway that remains to be demonstrated in plants. In this regard, we note that levels of two 26 S proteasome subunits (PBA1 and RPT2) remain stable during Arabidopsis leaf senescence, suggesting that the entire proteolytic complex remains active throughout the process.
Aberrant phenotypes did become evident when the apg7-1 plants were starved for nitrogen and carbon, leaves of apg7-1 plants were detached and incubated in the dark or after bolting when the mature rosette leaves of apg7-1 plants senesced prematurely. Ohsumi and co-workers (44) recently observed similar effects for an Arabidopsis tDNA mutant of APG9, 2 an autophagy-associated protein of unknown function, suggesting that mutations in many of the APG loci will elicit a similar phenotype. All of the apg7-1 phenotypes can be explained by a common response, i.e. attempting to maintain the supply of nitrogen and carbon for continued survival. They are also consistent with the hypersensitivity to nutrient limiting conditions exhibited by yeast apg/cvt/gsa mutants blocked in microautophagy, macroautophagy, and pexophagy (18). Based on the premises that chloroplasts can be removed by autophagy (45) and that apg7-1 plants have impaired autophagy, it was possible that the apg7-1 plants would exhibit a slower rate of chlorosis. Instead we observed more rapid chlorosis and loss of chloroplast enzymes, indicating that chloroplast protein turnover was accelerated. This suggests that chloroplasts are not substrates of a proteolytic system dependent on APG7. In support, Ono et al. (46) recently showed that during wheat leaf senescence, the number and size of chloroplasts did not change dramatically until the last stage of senescence just before cell autolysis. A more likely scenario is that the apg7-1 plants accelerate chloroplast protein turnover by other mechanisms including the activation of Clp and other chloroplast proteases, some of which are known to be up-regulated during leaf senescence (38,47).
The preferential accumulation of APG7 and Ϫ8 transcripts during detached leaf senescence further supports a role for the APG8/12 system in the senescence syndrome. In fact, their accumulation would have appeared even more dramatic than shown with the RNA gel blots in Fig. 10 if the amount of RNA loaded for each time point was equal instead of adjusted for an equal amount of tissue. As a consequence, the APG system can now be added to a growing list of "senescence-associated" processes up-regulated during senescence (38,48). Compared with the accumulation of SEN1 mRNA, accumulation of APG7 and Ϫ8 mRNAs was delayed, suggesting that the APG system functions later than other senescence-associated components. In yeast, the levels of the APG8 transcript increase dramatically during starvation, but those of the APG7 transcript are 2 H. Hanaoka, T. Noda and Y. Ohsumi, unpublished observations. unaffected (49,50). Cytokinins, known regulators of the senescence program in leaves (43), were unable to fully reverse the effects of the apg7-1 mutation as measured by retention of chlorophyll and the two photosynthetic proteins, Rubisco and magnesium protoporphyrin chelatase. Therefore, cytokinins likely function either upstream of the APG8/12 pathway or in a parallel pathway that ultimately promotes leaf survival. Cytokinin treatment slightly decreased the levels of the APG7 protein; it will be interesting to see if a concomitant decrease in the APG7 and Ϫ8 mRNAs occurs as well.
Compared with the APG8/12 pathways in yeasts, their counterparts in Arabidopsis may be more elaborate given the presence of multiple loci encoding many components in the predicted conjugation cascade. Especially interesting is the presence of multiple isoforms of APG8 that phylogenically cluster into two distinct clades. An intriguing possibility is that these isoforms have distinct functions, targets, and/or expression patterns. For example, because APG8h and -i lack extra amino acids beyond their C-terminal glycines, they presumably can enter the conjugation cascade without processing by APG4 (see Fig. 1). In mammalian species, three different APG8 proteins have been identified that work with the E2 APG3 (51). Interaction studies showed that each can bind specifically to a different set of proteins, suggesting that specific APG8 isoforms can perform unique functions (52)(53)(54).
The identification of Arabidopsis orthologs for the yeast APG8/12 conjugation system now expands the potential repertoire of small polypeptide tags operating in plants to five, with Ub, RUB1/NEDD8, and SUMO being the others (3,55). In all these cases, the respective conjugation pathways appear more complex than in yeast, implying that plants have increased their reliance on these post-translational modifiers. Coincidentally, most if not all likely participate directly or indirectly in protein turnover. By using the APG genes and proteins identified here, a detailed genetic, molecular, and cytological understanding of the APG8/12 conjugation system and its proposed role in plant autophagy is now possible.