The ATG autophagic conjugation system in maize: ATG transcripts and abundance of the ATG8-lipid adduct are regulated by development and nutrient availability.

Plants employ sophisticated mechanisms to recycle intracellular constituents needed for growth, development, and survival under nutrient-limiting conditions. Autophagy is one important route in which cytoplasm and organelles are sequestered in bulk into vesicles and subsequently delivered to the vacuole for breakdown by resident hydrolases. The formation and trafficking of autophagic vesicles are directed in part by associated conjugation cascades that couple the AUTOPHAGY-RELATED8 (ATG8) and ATG12 proteins to their respective targets, phosphatidylethanolamine and the ATG5 protein. To help understand the importance of autophagy to nutrient remobilization in cereals, we describe here the ATG8/12 conjugation cascades in maize (Zea mays) and examine their dynamics during development, leaf senescence, and nitrogen and fixed-carbon starvation. From searches of the maize genomic sequence using Arabidopsis (Arabidopsis thaliana) and rice (Oryza sativa) counterparts as queries, we identified orthologous loci encoding all components necessary for ATG8/12 conjugation, including a five-member gene family expressing ATG8. Alternative splicing was evident for almost all Atg transcripts, which could have important regulatory consequences. In addition to free ATG8, its membrane-associated, lipidated form was detected in many maize tissues, suggesting that its conjugation cascade is active throughout the plant at most, if not all, developmental stages. Levels of Atg transcripts and/or the ATG8-phosphatidylethanolamine adduct increase during leaf senescence and nitrogen and fixed-carbon limitations, indicating that autophagy plays a key role in nutrient remobilization. The description of the maize ATG system now provides a battery of molecular and biochemical tools to study autophagy in this crop under field conditions.


INTRODUCTION
The recycling of intracellular constituents is critical to all organisms for optimal growth and development and survival under nutrient-limiting conditions. This recycling is especially important throughout a plant's life cycle, especially as plants scavenge available nutrients from storage tissues and older senescing leaves for the synthesis of reproductive organs and new leaves, and during whole plant senescence as fixed carbon (C) and nitrogen (N) are mobilized to reproductive or storage organs (Vierstra, 1996;Hopkins et al., 2007;Lim et al., 2007). The process is initiated in source tissues by dramatic changes in gene expression where genes involved in basic metabolism, including photosynthesis and protein biosynthesis, are downregulated while those involved in programmed cell death and the stress response and/or encoding various hydrolytic enzymes are up-regulated (Hopkins et al., 2007;Lim et al., 2007).
Nutrient mobilization in plants, as in yeast and animals, involves a battery of catabolic processes, including a sophisticated recycling system called autophagy or "self eating". Here, cytoplasm and organelles are sequestered in bulk into special autophagic vesicles, which then deliver their contents to the vacuole (or lysosome) for breakdown by resident hydrolases (Thompson and Vierstra, 2005;Bassham, 2007). The release products can be either consumed by the host cell or transported to other tissues and organs for reuse.
In yeast (Saccharomyces cerevisiae) and animal cells, autophagy (also referred to as macroautophagy) begins with formation of a sequestering membrane called the phagophore (or isolation membrane), which engulfs portions of the cytoplasm to form the double membranebound autophagosome vesicle (Ohsumi, 2001;Klionsky, 2007;Mizushima, 2007). The outer membrane of the autophagosome fuses with the membrane of the lytic compartment (i.e., vacuole in yeast and lysosome in animals) and the inner vesicle is released into the lytic compartment as an autophagic body. Both the limiting membrane and contents of the The ATG4 processing protease is also essential for ATG8 activity; it exposes the C-terminal glycine in the longer ATG8 precursor and recycles mature ATG8 by releasing it from the ATG8-PE adduct, presumably before autophagic body deposition.
The biochemical roles of ATG8-PE and ATG12-ATG5 conjugates have recently been revealed. Based on cell-free studies, the ATG8-PE adduct appears to promote the expansion of autophagosomal membrane and hemifusion of this membrane with the vacuolar membrane (Nakatogawa et al., 2007). ATG8-PE assembly is stimulated in vitro by a complex containing the ATG12-ATG5 adduct and ATG16, suggesting that the ATG12-ATG5 adduct has an E3-like catalytic role in ATG8 conjugation (Hanada et al., 2007;Fujita et al., 2008). Whereas the ATG12-ATG5 conjugate is released from the autophagosome before deposition in the vacuole, at least some ATG8-PE enters the vacuole attached to the autophagic body and is catabolized.
Over the past few years, obvious orthologs of various yeast ATG proteins have been detected in both higher and lower plants, indicating that a similar autophagic system likely exists throughout the plant kingdom. The Arabidopsis thaliana genome for example encodes proteins related to subunits of the yeast ATG1 kinase and ATG6/Beclin1 phophatidylinositol 3-kinase complexes, and all components of the ATG8/12 conjugation cascades, including small gene families encoding the two tags ((Doelling et al., 2002;Hanaoka et al., 2002;Surpin et al., 2003;Yoshimoto et al., 2004;Liu et al., 2005;Inoue et al., 2006;Fujiki et al., 2007;Xiong et al., 2007;Phillips et al., 2008) and A. Suttangkakul and R.D. Vierstra, unpublished). Using a recombinant system, Fujioka et al. (2008) recently succeeded in assembling a functional Arabidopsis ATG8 conjugation cascade in vitro by adding ATP to reactions containing ATG7, ATG3, and liposomes prepared with PE. Reverse genetic analyses of several Arabidopsis components have linked the ATG system, and by inference autophagy, to senescence, survival under N-and fixed Climiting growth conditions, the stress response, and programmed cell death triggered by www pathogen resistance (e,g., Liu et al., 2005;Xiong et al., 2007;Phillips et al., 2008). For instance, plants missing ATG4a/b, ATG5, ATG7, or ATG10 senesce faster and are hypersensitive to N starvation and extended darkness that limits the availability of fixed C (Doelling et al., 2002;Yoshimoto et al., 2004;Thompson et al., 2005;Phillips et al., 2008).
Using green fluorescent protein-ATG8 fusions as markers, it was also shown that these atg mutants block the vacuolar accumulation of autophagic bodies during N and C stress (Yoshimoto et al., 2004;Sláviková et al., 2005;Thompson et al., 2005;Yano et al., 2007;Phillips et al., 2008).
While our understanding of the Arabidopsis ATG system is progressing, little is currently known about this system or autophagy in other plant species. Given the potential significance of autophagic recycling to many agriculturally relevant processes, including seed germination, N and C mobilization from leaves into seeds, fruits and roots, survival under suboptimal growth conditions, pathogen protection, and senescence (Liu et al., 2005;Thompson and Vierstra, 2005;Bassham, 2007), information is especially needed for crop plants. Such knowledge may reveal important control points in autophagy that could be manipulated in both food and biofuel crops to enhance nutrient use efficiency or to better allocate C and N to specific organs for improved yield. As the first step toward understanding the importance of autophagy in crops, we describe here a collection of components that participate in the ATG8/12 conjugation cascades in both rice (Oryza sativa) and maize (Zea mays), and characterize the expression of the maize Atg genes and the lipidation of ATG8 during development and in response to nutrient limitations. The accumulation of the ATG8-PE adduct in particular correlates with senescence and N availability in maize, suggesting that it could be sensitive indicators of a crop's nutritional status.

Identification of Rice and Maize Atg Genes
Previous genome searches of Arabidopsis identified a suite of proteins structurally and functionally related to many ATG components present in yeast (Thompson and Vierstra, 2005;Bassham, 2007). To facilitate the search for maize orthologs, we first identified a collection of Atg genes from the more complete and better annotated rice (Oryza sativa ssp. japonica) cDNA and genomic databases, using the yeast and Arabidopsis counterparts as queries (Table I).
This rice collection contained easily recognizable relatives (typical E values <e -50 by TBLASTN) of most Arabidopsis ATG genes, with a number of components similarly encoded by small gene families (e.g., 5 rice loci encoding ATG8 versus 9 loci in Arabidopsis). The Arabidopsis and rice collections were then used as queries in TBLASTN searches of the various DNA sequence databases for the maize B73 inbred, including the cDNA, Expressed Sequence Tag (EST), EST assembly (e.g., ZmGI), genome survey sequence, and genomic assembly (e.g., AZM) databases. In most cases, alignments of the cDNA and genomic sequences were used to establish the Zm Atg gene models. However, in the absence of mRNA sequences, further BLASTX alignments with Arabidopsis, rice and yeast ATG proteins were used to predict the fulllength protein sequences in the Zm Atg gene models. As a high-throughput B73 genome sequences became available later, gaps in the Zm Atg gene models were eliminated by sequence analysis of corresponding BAC clones (Table II).
As shown in Fig. 1A and Table II, this survey identified five maize loci encoding the ATG8 protein tag, two encoding the ATG4 processing protease, and one gene each encoding the ATG12 protein tag, the common ATG7 E1 activating enzyme, the ATG3 and ATG10 E2 conjugating enzymes specific for ATG8 and ATG12, respectively, and the ATG5 target of ATG12. For almost all loci, strong amino acid sequence conservation was evident when compared to their rice, Arabidopsis and yeast orthologs (Table II). For example, the five maize Atg8 genes (Zm Atg8a-e), which are 100-95%/100-84% similar/identical to each other, are 94%/86-85% and 92-90%/75-73% similar/identical to Arabidopsis ATG8a and yeast ATG8, respectively ( Fig. 2 and Table II). The Zm Atg8b and Zm Atg8c loci are on different chromosomes but encode identical proteins, suggesting that this pair arose by a recent gene duplication. Furthermore, key amino acids necessary for the function of each component were detected, indicating that the maize conjugation cascades are mechanistically identical to those present in yeast and Arabidopsis.
These residues include: the active-site cysteines in ATG3, ATG7 and ATG10 that form the thiolester adducts with ATG8 and/or ATG12, the lysine acceptor site in ATG5, the positionally conserved glycines in ATG8 and ATG12 that directly participate in covalent binding, and the cysteine essential for the proteolytic activity of ATG4 (Fig. 1A).
Our maize gene models were further validated by sequence analysis of a collection of transcripts for each Zm Atg gene (Table II). With the exception of Zm Atg5, for which we were unable to generate a cDNA by RT-PCR presumably due to the high GC content of its 5' end, cDNAs encompassing the full coding region for each locus were obtained. These cDNA sequences not only confirmed that each Zm Atg gene is transcriptionally active but also identified a number of splice variants for many loci (Figs. 1B and 1C, Table III, and Table SI).
The number of Zm Atg genes with splice variants (11 of 12 genes) was higher than for Arabidopsis (7 of 17) or rice (11 of 14) ( Table III). The average number of splice variants per gene was also higher in maize, suggesting that alternative pre-mRNA splicing is more extensive in this species. Some of the splice variants substantially changed the corresponding reading frames, indicating that the resulting mRNAs would direct the synthesis of altered proteins if translated. For example, several splice variants of Zm Atg4b could affect the activity of the encoded protease by removing a stretch of amino acids near its active-site cysteine (Fig. 1B).
Additionally, several splice variants of Zm Atg10 eliminate the exonic region encoding the activesite cysteine that binds ATG12 during its conjugation cascade (Fig. 1C), thus inactivating the variant ATG10 protein. Whether any of these variant proteins actually accumulate and adversely affect the ATG8 or ATG12 conjugate cascades are not yet known.

Maize and Rice ATG7 has a Duplicated C-Terminal Region
Initially, there was no clear full-length cDNA model for Zm Atg7, even though our BLAST analyses detected ESTs corresponding to the 3' end of the transcript. Fortunately, we were able to generate a putative full-length Atg7 cDNA by RT-PCR using a forward primer designed from a partial 5' genomic sequence and a reverse primer based on the partial 3' EST. This longer Zm Atg7 cDNA matched (i) our partial maize gene models, (ii) a full-length gene model generated from a more recently released BAC sequence of maize, and (iii) a predicted fulllength rice Atg7 transcript. Surprisingly, when aligned with the Arabidopsis ATG7 coding region, we found that the rice and maize transcripts encoding substantially longer polypeptides (1021 and 1042 residues for maize and rice ATG7, respectively, versus 697 residues for Arabidopsis ATG7).
Dot plot amino acid sequence comparisons of the maize and rice polypeptides with Arabidopsis ATG7 revealed that the increased length was created solely by a sizable duplication of the C-terminal region of the maize and rice proteins ( Figure 3A and data not shown). This arrangement has not yet been described in other plants, fungi or animals, implying that it was generated by a cereal-specific tandem duplication. The region involved in the duplication is essential for the E1 activating activity of ATG7 as it contains both the nucleotide-binding pocket and the conserved cysteine that binds ATG8/12 via thiolester bond (Doelling et al., 2002; www.plantphysiol.org on August 27, 2017 -Published by Downloaded from Copyright © 2008 American Society of Plant Biologists. All rights reserved. Fujioka et al., 2008). In maize and rice, the more C-terminal duplicated region retained both the nucleotide-binding pocket and the essential cysteine whereas the more internal duplicated region had the essential cysteine replaced by a serine, implying that the internal segment cannot be catalytically active by itself (Fig. 3A).
To test whether the extended maize and rice ATG7 proteins retained their ability to interact with ATG8/12, we examined by yeast-two hybrid (Y2H) analysis the interaction of Zm ATG7 with Arabidopsis and maize ATG8 and ATG12. As shown in Figure 3B, Zm ATG7 like its Arabidopsis ortholog could bind both tags, including the At ATG8a and Zm ATG8a isoforms, Zm ATG12, and both isoforms of At ATG12. The interactions of the tags with At ATG7 were not abolished by the C558A variant in which the active-site cysteine was substituted for an alanine ( Fig. 3B), indicating that the binding was likely to be non-covalent and did not require formation of the thiolester adduct. Taken together, we conclude that the maize (and likely rice) ATG7 protein retained its ability to interact with its ATG8 and ATG12 substrates despite the C-terminal duplication.

Maize ATG8 is Lipidated in planta
The modification of ATG8 with PE is an important molecular and cytological marker for autophagic compartments (Ohsumi, 2001, Mizushima, 2007. Whereas the free form remains soluble, the lipidated form becomes bound to the limiting membranes of autophagosomes and autophagic bodies via the PE anchor. In an attempt to detect the ATG8-PE adduct in maize, we first confirmed that the antibodies against Arabidopsis ATG8a could recognize its maize orthologs. As shown in Figure 4A, four of the five recombinant Zm ATG8 proteins tagged with 6His were easily detected by immunoblot analysis following SDS-PAGE. (Zm ATG8c was not tested since its amino acid sequence is identical to Zm ATG8b (Fig. 2) their recognition by anti-5His antibodies, it appeared that the sensitivities of the antibodies toward Zm ATG8a-e were not significantly different from that toward the At ATG8a antigen, consistent with the strong amino acid sequence identities among the group (see Fig. 2).
Whereas these antibodies detected several members of the Arabidopsis ATG8 family in crude seedling extracts following SDS-PAGE, (Fig. 4B and Thompson et al., 2005;Phillips et al., 2008), a single species was prominent in crude extracts from maize seedlings (Fig. 4B), which is consistent with the similar size and charge of the maize ATG8 proteins.
To study the lipidation of maize ATG8, we exploited the assay developed by Yoshimoto

Spatial and Temporal Patterns of Zm Atg Gene Expression and ATG8 Lipidation
Previous studies with Arabidopsis demonstrated that the genes encoding the ATG8/12 conjugation cascades are ubiquitously expressed with senescence and nutrient limitations increasing transcript abundance (e.g., Contento et al., 2004;Yoshimoto et al., 2004;Buchanan-Wollaston et al., 2005;Sláviková et al., 2005;Thompson et al., 2005;Rose et al., 2006;Van der Graaff et al., 2006;Osuna et al., 2007;Peng et al., 2007;Phillips et al., 2008). Similar widespread expression was also apparent for the orthologous maize loci. Using semiquantitative RT-PCR, we detected mRNAs for all Zm Atg genes assayed in a variety of tissues, including the shoot apex (Fig. 5A). Slight variations in the expression levels of the five Zm Atg8 loci were observed which could reflect tissue-specific functions of the corresponding isoforms (e.g., Zm Atg8c was best expressed in the shoot apex while Zm Atg8d was better expressed in older leaves) (Fig. 5A).
To gain further insights into the regulation of ATG8 conjugation during development, we examined the levels of the free ATG8 protein and its lipid adduct in selected maize tissues at various stages of development. In most cases, enrichment of the ATG8-PE adduct by membrane isolation was not needed to enable detection of the faster migrating lipidated form, indicating that some maize tissues accumulate high levels of this species. For example, both free ATG8 and the ATG8-PE adduct were easily detected during seed/seedling development in crude extracts prepared from both the endosperm and embryo at increasing times of maturity and several parts of the seedling 5 days after germination (DAG) (Fig. 5B). In young embryos, almost all ATG8 was present in the free form, but as they developed, the ATG8-PE adduct appeared around 22 days after pollination (DAP). In a mature embryo at 30 DAP, the level of free ATG8 was substantially reduced, while the level of ATG8-PE remained the same. The endosperm also accumulated higher levels of the ATG8-PE adduct as they matured, suggesting that autophagy may participate in programmed cell death during endosperm development. A high level of adduct was evident in the scutellum, but not the plumule and radicle, during germination, which could reflect the autophagic mobilization of stored reserves in the scutellum to support new growth (Fig. 5B). Leaves harvested from soil-grown, well-fertilized seedlings also contained both the free and lipidated form of ATG8, with a similar ratio evident in green leaves of various ages as categorized by their positions on the plant (Fig. 5C).
Strikingly, we found that both the total amount of ATG8 and the abundance of the ATG8-PE adduct in maize leaves was markedly affected by senescence. For example, when we compared the yellow and green sectors in an individual senescing leaf from a mature flowering plant, the yellow sectors had substantially more total ATG8 protein ( Fig. 5D). Moreover, while the green sectors had a near equal ratio of lipidated ATG8 to free ATG8, similar to that found in the adjacent non-senescing leaf positioned on the plant axis immediately above, the yellow sectors contained almost exclusively the ATG8-PE adduct, suggesting that high rates of autophagy were underway in these areas. This senescence-dependent effect could also be induced when the plants were grown hydroponically in a defined solution containing inorganic nutrients, a growth condition that accelerates the developmental senescence of leaves even though the plants appear well fertilized. As shown in Figure 5E for plants grown hydroponically for 10 DAG, the oldest leaf (L1) contained only traces of ATG8-PE, the next oldest (L2) contained much higher levels but almost all was lipidated, while younger leaves (L3) contained a mixture of the free and lipidated form, and the shoot apex (SA) contained almost exclusively the free form. Eventually, both forms of ATG8 disappeared along with most other proteins as the older leaves completed senescence (see L1 and L2 from 24-DAG seedlings (Fig. 5E)).

Zm Atg Gene Expression and ATG8 Lipidation Increase During Nutritional Stress
The age-dependent accumulation of the ATG8-PE adduct in the hydroponically grown seedling leaves (Fig. 5E)  gene expression and the lipidation of the ATG8 protein. One obvious candidate is N, since its availability is often limiting to plant growth and protein synthesis. To test this possibility, we germinated and grew maize seedlings hydroponically for 15 days on deionized water, Murashige and Skoog (MS) medium lacking nitrogen (MS-N), or on MS-N medium supplemented with 0.5, 1, or 2 times the normal concentration of nitrate (see Materials and Methods). As expected, the growth of seedlings treated with only water or MS-N was significantly retarded 15 DAG as compared to N-fertilized plants (Fig. 6). Concomitant with this growth reduction was an increase in the oldest leaves (L1) of both the amount of total ATG8 and the proportion converted to the lipidated form. As increasing amounts of N was added to the MS-N medium, the amounts of the ATG8-PE adduct and total ATG8 dropped in the L1 leaf ( leaves (Fig. 7B). In contrast to the effects on ATG8 levels in leaves, both the total amount of ATG8 and the ratio of lipidated ATG8 to the free form was unaffected by N limitation in the roots and shoot apex (Fig. 7C).
The increase in maize ATG8 protein levels during N starvation was consistent with previous data in Arabidopsis showing that the expression of the ATG system is up-regulated by nutrient stress (Contento et al., 2004;Yoshimoto et al., 2004;Buchanan-Wollaston et al., 2005;Sláviková et al., 2005;Thompson et al., 2005;Rose et al., 2006;Van der Graaff et al., 2006;Osuna et al., 2007;Peng et al., 2007;Phillips et al., 2008). To test whether the suite of Zm Atg genes was affected similarly, we analyzed the transcript levels for each during N starvation along with several senescence-associated transcripts. Here, total RNA was isolated from maize leaves of various ages (L1 to L3) from 15-DAG seedlings fertilized with either MS or MS-N, and the levels of individual mRNAs were determined by semi-quantitative RT-PCR.
Transcripts encoding the Ub-conjugating enzyme (Ubc9) and the glyceraldehyde-3-phosphate dehydrogenase (Gapdh) were used for normalization and as an internal control, respectively (Czechowski et al., 2005). As can be seen in Figure 7D, the Gapdh/Ubc9 transcript ratio was not significantly affected by leaf age or N availability under these growth conditions. In contrast, the senescence-associated See1a transcript (Smart et al., 1995) was significantly upregulated by the MS-N growth condition, confirming that premature senescence was induced in these N-limited leaves. Accelerated leaf senescence in N-limited leaves was also evident from shows a light-dependent expression pattern (Willingham et al., 1994). The Sbp transcript levels in N-limited L2 leaves were also higher than in L2 leaves fertilized with MS. The reason for this accumulation of these photosynthesis-related transcripts in N-limited middle leaves is not clear.
When the Zm Atg genes were analyzed similarly, we found that the transcript abundance of almost all were up-regulated by the absence of N and further up-regulated as the leaves aged (Fig. 7D). In most cases, the highest transcript abundance was found in the L1 leaves from MS-N fertilized plants. This trend was particularly obvious for Atg3, Atg7, Atg10, and four of the five isoforms of Atg8 (Atg8b, c, d and e). Interestingly, in addition to Atg7, those genes encoding components required for ATG8 conjugation were more strongly affected by the combined action of N availability and leaf age (Atg3, Atg4b and Atg8b-e) than those encoding components for ATG12 conjugation (Atg5, Atg10 and Atg12) (Fig. 7D). This difference suggests that formation of the ATG8-PE adduct is more limiting to autophagic recycling during senescence.
Fixed-C limitations has also been shown to increase the abundance of various ATG transcripts in Arabidopsis using either the withdrawal of sucrose from the growth medium of cell cultures (Contento et al., 2004;Rose et al., 2006), or placement of whole plants in extended darkness to deplete the availability of soluble sugars (Sláviková et al., 2005;Thompson et al., 2005;Phillips et al., 2008). To test the importance of fixed-C availability, we wrapped a 7.5-cm band of aluminum foil around individual maize leaf blades from plants grown in constant light to subject the region to extended darkness (Weaver and Amasino, 2001; Van der Graaff et al., 2006). At various times, the darkened sectors and the corresponding sectors from lightexposed control leaves of the same developmental age were harvested and total RNA was isolated. The abundance of individual transcripts in each RNA pool was then determined by semi-quantitative RT-PCR as described above using the signal from the Ubc9 mRNA for normalization. While short treatments (1-3 days) did not detectably affect the color of the darkened sectors, prolonged treatments up to 10 days induced obvious senescence as indicated by leaf sector yellowing/browning. Transcript analysis of Gapdh, See1a, Lhcb5 and Sbp demonstrated that this covering affected the photosynthetic machinery well before senescence became apparent. While the mRNA levels for Gapdh were unaffected and those for the See1a transcript were only modestly up-regulated in the darkened and presumably fixed C-limited sectors relative to illuminated sectors, the levels of both Lhcb5 and Sbp dropped dramatically consistent with their light requirement for transcription (Willingham et al., 1994;Van der Graaff et al., 2006). As with N limitation, almost all the Atg transcripts were up-regulated in the darkened sectors with the maximal levels appearing after 2 days in the dark (Fig. 8A). Atg7 displayed the strongest upregulation with an ~ 8-fold increase in transcript levels in the darkened sectors. Only the abundance of the Atg8a mRNA appeared to be unaffected by darkness with Atg8c and Atg10 exhibiting only small increases (Fig. 8A). As with N limiting growth condition (Fig. 7C), the genes associated with ATG8 conjugation were in general more strongly affected as compared to those involved with ATG12 conjugation.
Despite the increases in Atg mRNA abundance by darkening the leaf, the levels of the free ATG8 and the ATG8-PE adduct were not markedly altered. The steady-state levels of the free and conjugated forms were similar in the darkened and illuminated sectors even after 4 days of treatment (Fig. 8B). This lack of effect on ATG8 levels during fixed-C limitation as opposed to the strong effect induced by N limitation (see Fig. 7B) could indicate that autophagy is more sensitive to N starvation. However, a more plausible interpretation is that our experiments with N starvation likely caused the entire plant to experience N deficits while those testing fixed-C limitation subjected only part of the leaf to fixed-C deficits. Consequently, it is possible that the neighboring illuminated areas became a source of soluble sugars to the darkened areas, thus mitigating the drop in fixed-C in the absence of photosynthesis.

DISCUSSION
A number of genetic studies with the ATG8/12 conjugation cascades in Arabidopsis have demonstrated the importance of autophagy to nutrient recycling and survival of plants under N/C-limiting growth conditions (see above). In an effort to determine the importance of autophagy to crop plants, we completed a molecular description of an orthologous ATG system in the cereals rice and maize. We then assayed in maize their expression patterns and the accumulation of the lipidated ATG8 adduct during senescence and in response to N or fixed-C limitations. All components required for ATG8/12 conjugation in yeast and Arabidopsis (Ohsumi, 2001;Thompson and Vierstra, 2005) were identified in both rice and maize, indicating that the pathway is highly conserved among plant species. As in Arabidopsis, four maize components are encoded by single genes (Atg3, Atg5, Atg7 and Atg10), whereas two are encoded by small gene families (two for Atg4 and five for Atg8). This increased diversity for the ATG8 protein tag and the ATG4 processing protease required for ATG8 maturation suggests both subfunctionalization of the ATG8 isoforms and distinct roles for the two ATG4 isoforms in processing the ATG8 precursor and/or recycling ATG8 modified with PE. As reported for the Our description of the maize Atg genes is one of the earliest comprehensive genomic analyses on a specific metabolic pathway since the release of draft genome sequence (http://maizesequence.org). Despite difficulties in determining the exact number of genes encoding each Zm ATG protein given the incomplete annotation of the sequence and the large number of discontinuous segments still remaining, we are certain about the gene identifications described in Fig. 1 and Table II for the following reasons. First, we focused on genes involved in the ATG8/12 conjugation cascades. They appear to be the most conserved members in the Atg system, which in turn simplified the identification of the rice and then the maize orthologs.
For example, like its structural relative Ub, ATG8 is arguably one of the most conserved proteins in eukaryotes (Fig. 2, Fig. S1 and (Doelling et al., 2002;Xie and Klionsky, 2007) (Table SI)), it is likely that rice Atg3b and Atg10b loci are minor contributors to the pools of each enzyme. And lastly, the rice Atg8e gene is currently without any EST or cDNA data, thus leaving its gene model and transcriptional activity unsupported (Table I).
While almost all of the rice and maize ATG proteins are structurally orthologous to their Arabidopsis and yeast counterparts, ATG7 is notably distinct. The rice and maize polypeptides contained an approximately 325-amino-acid duplication of their C-terminal half that replicates the region encompassing the active-site cysteine and nucleotide-binding pocket necessary for ATG8/12 activation. While the nucleotide-binding pocket sequence is evident in both repeats, the more N-terminal repeat had its active-site cysteine replaced with a serine. Despite this duplication, maize ATG7 remains functional as judged by the ability of the full-length protein to interact by Y2H analysis with ATG8 and ATG12 and by the presence of its ATG8-PE product in planta. At present, it is unclear how the catalytic regions of rice and maize ATG7s are organized and whether the more N-terminal predicted nucleotide-binding pockets can actually participate in ATG8/12 activation.
An interesting aspect of the cereal Atg gene models is that most of the genes are associated with alternative splicing of pre-mRNA. Current estimates of genes with multiple splice variants are approximately 21% and 22% for Arabidopsis and rice, respectively (Wang and Brendel, 2006). Although the number of maize genes subjected to alternative slicing is not yet available, it appears to be similar to that of rice (Chung, T. and Larkins, B.A., unpublished data increasing the gene number or providing a sophisticated means to control the abundance of functional Atg transcripts in response to various developmental and environmental cues. The former scenario could explain why rice and maize have fewer ATG8 isoforms than Arabidopsis despite having larger proteomes overall. The latter scenario is also reasonable given the importance of the ATG autophagic system to cell survival and its likely regulation by a number of internal and external signals. Curiously, all the splice variants from the Arabidopsis ATG genes are predicted to encode proteins that retain their active site and/or other amino acids essential for function (Table III). On the contrary, many splice variants of the rice and maize Atg genes encode proteins lacking one or more essential residues and are therefore likely to be nonfunctional. The possibility that this splice variant distinction reflects unique methods employed by dicotyledonous and monocotyledonous plants to control the ATG system is intriguing.
As observed previously with Arabidopsis (Contento et al., 2004;Sláviková et al., 2005;Thompson et al., 2005;Phillips et al., 2008) lipidated ATG8, which is consistent with its role in absorbing nutrients from the endosperm for seedling development. (iii) And finally, both natural and hydroponics-induced senescence and N starvation dramatically increased the relative abundance of the ATG8-PE adduct. This effect was particularly striking in senescing leaves where the yellow sectors had high levels of total ATG8 with almost all of the pool converted to the lipidated form.
What drives this substantial increase in the steady state levels of the ATG8-PE adduct during seed development, leaf senescence and N starvation is not yet known. In yeast, this lipidation is controlled by a set of nutrient sensing systems (e.g., Target of rapamycin, Snf1, and cAMP-dependent protein kinases) that suppress the activity of the ATG1 kinase complex, whose action ultimately impacts the ATG8/12 conjugation machinery (Klionsky, 2007;Mizushima, 2007). Assuming that the ATG12-ATG5 adduct represents the E3 activity that transfers activated ATG8 from the ATG3 E2 to PE (Hanada et al., 2007;Fujioka et al., 2008), the assembly of this adduct in particular could be the rate-limiting step. Since the pool of the ATG8-PE adduct is also modulated by turnover, the steady-state increase could also reflect a decrease in delivery of the adduct along with the autophagosome to the vacuole, a decrease in the recycling of ATG8 by the ATG4 protease, and/or an increased half-life of the adduct once in the vacuole (Mizushima and Yoshimori, 2007). Clearly, reverse genetic analyses of the two conjugation cascades and upstream regulators and a detailed understanding of the intracellular dynamics of ATG8 will be needed to identify these control point(s) in maize.
Whatever the mechanisms, the strong correlations between the steady-state abundance of the ATG8-PE adduct and leaf senescence and growth under N-limiting conditions suggest that the levels of the adduct represents a sensitive indicator of both the amount of autophagic recycling and the nutrient status of the plant/organ/tissue. For specific tissues, it could also be a measure of their source/sink status. For example, in hydroponically grown www.plantphysiol.org on August 27, 2017 -Published by Downloaded from Copyright © 2008 American Society of Plant Biologists. All rights reserved. plants, older leaves, which are expected to be important nutrient sources for the growing regions as they senesce, contained mostly the conjugated form of ATG8, while the shoot apex, comprised of rapidly expanding young leaves that are strong nutrient sinks, contained mostly the free form. Taking this possibility one step further, our analysis of ATG8 lipidation also implies that autophagic recycling is regulated in a tissue/organ-dependent manner in maize and likely other plants. In N-fertilized plants, the high ATG8-PE to free ATG8 ratio in roots relative to shoots and leaves would imply that roots are relatively more active in autophagy under N-rich growth conditions. When depleted of N, root autophagy appeared unaffected as this ratio remained unchanged. Instead, a dramatic increase in the ATG8-PE adduct was observed in leaves, suggesting that autophagy in this tissue was specifically induced. This N starvationinduced increase further underscores the notion that senescing leaves are a major source of remobilized nutrients during nutrient starvation (Lim et al., 2007;Munné-Bosch, 2008). While more data are clearly needed to connect ATG8 lipidation to autophagic recycling rates and plant nutrition, our data suggest that the formation of the ATG8-PE adduct could be a simple indicator for the source/sink status of individual tissues and thus could provide a useful marker for the nutritional state of crops grown in the field.

Plant Materials and Growth Conditions
All experiments were performed with the maize (Zea mays) inbred B73. Seeds were germinated in pots (125 cm 3 ) with 1:1:1 mixture of compost, peat moss and vermiculite, and were grown at 21 °C in growth chamber with a 16 hr-light/8hr-dark photoperiod. For darkinduced fixed-C limitation, the basal part of the second leaf blade was wrapped with aluminum www.plantphysiol.org on August 27, 2017 -Published by Downloaded from Copyright © 2008 American Society of Plant Biologists. All rights reserved. foil (75 mm in length) at 15 DAG. At various times, the wrapped part of the leaf blade was harvested along with a corresponding region from a light-exposed leaf blade of the same developmental age, which served as the control. For N limitations, seedlings were grown hydroponically in pots (450 cm 3 ) with vermiculite. The pots were sub-irrigated with 10 mL of various liquid media every 5 days, starting at 1 DAG. The media included: (i) MS medium (MS basal salts), (ii) MS-N medium (MS micronutrient solution plus 3 mM CaCl 2 , 1.5 mM MgSO 4 , 1.25 mM KH 2 PO 4 , 5 mM KCl (pH 5.7)), or (iii) MS-N medium supplemented with either 0.825 g/L NH 4 NO 3 and 0.95 g/L KNO 3 (MS-N + 0.5X N medium), 1.65 g/L NH 4 NO 3 and 1.9 g/L KNO 3 (MS-N + 1X N medium) or 3.3 g/L NH 4 NO 3 and 3.8 g/L KNO 3 (MS-N + 2X N medium). When needed, seedlings were grown to maturity in a greenhouse to allow for natural senescence and to obtain developing kernels following self-pollination. Extra lighting was provided to maintain a 14-hr light/10-hr dark photoperiod.

DNA and Protein Sequence Analyses
Maize Atg genes were identified by TBLASTN (Altschul et al., 1990) using the orthologous protein sequences from Arabidopsis and rice (Oryza sativa spp. japonica) as queries. The rice Atg sequences were first identified in the GenBank rice NR and TIGR pseudomolecule (Ouyang et al., 2007) databases by TBLASTN using the yeast (Saccharomyces cervisiae) and Arabidopsis thaliana orthologs as queries. Corresponding rice Atg genomic sequences were subsequently used as queries against rice NR and EST databases. Rice Atg gene models and potential splice variants were identified by manual inspection of the resulting sequence alignments. The Arabidopsis and rice ATG protein sequences were then used to search by TBLASTN the ZmGI EST contig (Lee et al., 2005) and TIGR AZM5 genomic assembly (Chan et al., 2006)  were also used as BLASTN query sequences against the AZM5 genomic sequence database to help assign each cDNA to a genomic locus.
Sequence alignment between individual cDNAs and genomic sequences were manually inspected for consensus coding regions, intron splice sites, and potential splice variants.
Sequence gaps were filled in by BLASTN of the resulting gene models to maize HTGS BAC sequence database (http://blast.ncbi.nlm.nih.gov/Blast.cgi). Dot plots revealing local sequence similarities were generated by PLALIGN (Pearson, 2000). Full-length cDNAs for each Zm Atg gene were generated by RT-PCR of total RNA isolated from leaves, using primers that bracketed the translation start and stop codons (see Table SII

RT-PCR Analyses
RNA was extracted from three biological replicates of various maize tissues using Trizol (Invitrogen). First-strand cDNA synthesis was achieved with the Superscript II (Invitrogen) reverse transcriptase and an oligo-(dT) 18 primer (Fermentas Inc., Glen Burnie, MD). The amount of RT product in each sample was standardized by a second round of PCR using primers against Zm Ubc9 (named because of its closest amino acid match to Arabidopsis UBC9 (Czechowski et al., 2005)). After appropriate dilutions, the RT products were subjected to semi-quantitiative PCR using Zm Atg gene-specific primers ( cycles was adjusted beforehand to obtain a near-linear amplification of each transcript. The amounts of PCR products were quantified using GelDoc imaging software (Biorad Laboratories, Hercules, CA) following agarose gel electrophoresis and ethidium bromide-staining of the product. Normalized expression values were generated using RT-PCR products generated from Zm Ubc9 as the standard.

Yeast Two-Hybrid (Y2H) Assays
Y2H assays were conducted with full-length cDNAs as previously described with minor modifications (Gingerich et al., 2007). Using the LR Clonase reaction (Invitrogen), the At ATG8a, At ATG12a, At ATG12b, Zm Atg8a, and Zm Atg12 cDNAs were recombined into the pDEST22 activation domain plasmid, whereas the cDNAs encoding At ATG7, At ATG7 (C558A) and Zm ATG7 were recombined to the pDEST32 binding-domain plasmid (Invitrogen). The plasmids were then introduced into the haploid yeast strains YPB2a and LB41α, respectively, and diploid yeast strains were then generated by mating. Protein-protein interactions were identified by testing for growth of the diploid strains on selection medium lacking histidine, leucine and tryptophan, and containing 5 mM 3-amino-1',2',4'-triazole (3-AT). The At ATG7(C558A) mutant in which the Cys-558 codon was replaced for that encoding alanine was generated by the PCR-based QuickChange method (Stratagene, La Jolla, CA) with appropriate mutagenic primers.

SDS-PAGE Analysis
In most cases, maize tissue was harvested, immediately frozen to liquid nitrogen temperatures, and homogenized in SDS-PAGE sample buffer. The extracts were clarified by centrifugation at 16,000 X g and then heated before use. Protein amounts were determined by spotting aliquots of the extracts onto Whatman 3MM filter paper (Fischer Scientific, Pittsburgh, PA) and comparing the Coomassie blue staining intensity of the spots with bovine serum albumin standards (Ghosh et al., 1988). In most cases, equal amounts of total protein were subjected to SDS-PAGE with or without 6 M urea in the separating gel. Immunoblot analyses were performed with affinity-purified anti-At ATG8 antibodies (Thompson et al., 2005) or anti-5His antibodies (Qiagen, Germantown, MD) following electrophoretic transfer of the proteins onto Immobilon PVDF membranes (Millipore, Bedford MA). Cross-reactivity of anti-At ATG8 antibodies with Zm ATG8 protein isoforms was checked using recombinant 6His-tagged fulllength versions expressed in E. coli BL21(DE3) cells. The pDEST17 plasmids expressing each of the Zm ATG8a, -b, -d and -e isoforms were generated as described (Invitrogen) by a series of LR reactions with appropriate pENTR/D-TOPO clones containing the full-length cDNAs. The pET28 plasmid expressing At ATG8a was previously described (Thompson et al., 2005).
To characterize the ATG8-PE adducts, various maize tissues were homogenized in TNPI buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM phenylmethanesulfonylfluoride, and 10 mM iodoacetamide). The extract was filtered through four layers of cheesecloth and clarified by centrifugation at 2,000 X g for 5 min, and the supernatant was centrifuged at 100,000 X g for 1 hr. The resulting 100-k pellet was resuspended in TNPI buffer, solubilized by adding Triton X-100 at a final concentration of 0.5 %, and clarified at 16,000 X g for 10 min to obtain the solubilized membrane fraction. This membrane fraction was incubated at 37 °C for   and SV6 in Atg4b, and for SV2, SV5 and SV6 in Atg10) splice sites, or sites of intron retention (for SV2 and SV3 in Atg4b and for SV4 and SV6 in Atg10). Refer to Table SI for     Detection of the ATG8-PE adduct in maize. Crude extracts (CE) prepared from maize roots were centrifuged at 100,000 X g to collect the soluble (S) and membrane fraction (100k P). Chung et al. 39 The 100k P was solubilized with Triton X-100 and incubated for 1 hr with or without phospholipase D (PLD). The samples were then subjected to SDS-PAGE in the presence of 6 M urea. The free and lipidated form of ATG8 are indicated by the arrowheads. Atg transcripts are expressed ubiquitously in various maize tissues. The shoot apex (SA), the L4, L3 and L2 seedling leaves (leaf numbers correspond to the order of appearance) and roots were collected from seedlings grown hydroponically for 20 days with MS medium. The nonpollinated ears and tassels were collected from a mature plant grown in soil. Total RNA was subjected to semi-quantitative RT-PCR using primers specific for various maize Atg genes or the Ubc9 control gene. B-E, Analysis of free ATG8 and the ATG8-PE adduct. Crude extracts were prepared from various tissues and subjected to SDS-PAGE with 6 M urea followed by immunoblot analyses with anti-At ATG8a antibodies. B, Endosperm and embryos at various days after pollination (DAP) from a soil-grown plant and tissues dissected from seedings germinated for 5 days on MS medium. Sc, scutellum, Pl, plumule, Rd, radicle. C, Leaves from seedlings grown in soil for 20 DAG. D, Naturally senescing leaves. Samples were collected from a non-senescing leaf (NSL) and yellow (Yl) and green (Gr) sectors from a subtending senescing leaf (SL, see picture for the sampled area) from a flowering plant grown in soil. E, Shoot apex (SA) and leaves of various ages (L1 older to L4 younger) from seedlings grown hydroponically on MS medium for the indicated DAG. Panels C and D contain 20 μg of total protein per lane. The lanes in Panels B and E were loaded on an equal fresh-weight basis except Pl, Rd and SA. Protein profiles in the lower panels of B and E were detected by Ponceau S staining.   Expression levels were normalized by dividing the amounts of PCR products with those produced from the Zm Ubc9 control transcript. Each bar represents the average of three biological replicates (± SD). Note that expression values among genes cannot be compared, due to the different amplification conditions used for each gene.