Mobilization of Rubisco and stromal-localized fluorescent proteins of chloroplasts to the vacuole by an ATG gene-dependent autophagic process 1

During senescence and at times of stress, plants can mobilize needed nitrogen from chloroplasts from leaves to other organs. Much of the total leaf nitrogen is allocated to the most abundant plant protein, ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco). While bulk degradation of the cytosol and organelles in plants occurs by autophagy, the role of autophagy in the degradation of chloroplast proteins is still unclear. We have visualized the fate of Rubisco, stromal-targeted green fluorescent protein (GFP) and DsRed, and GFP-labeled Rubisco in order to investigate the involvement of autophagy in the mobilization of stromal proteins to the vacuole. Using immuno-electron microscopy (IEM), we previously demonstrated that Rubisco is released from the chloroplast into Rubisco-containing bodies (RCBs) in naturally senescent leaves Plant Cell Physiol 44: When leaves of transgenic Arabidopsis ( Arabidopsis thaliana ) plants expressing stromal-targeted fluorescent proteins were incubated with concanamycin A to inhibit vacuolar H + ATPase activity, spherical bodies exhibiting GFP or DsRed fluorescence without chlorophyll fluorescence were observed in the vacuolar lumen. Double-labeled IEM with anti-Rubisco and anti-GFP antibodies confirmed that the fluorescent bodies correspond to RCBs. RCBs could also be visualized using GFP-labeled Rubisco directly. RCBs were not observed in leaves of a T-DNA insertion mutant in ATG5 , one of the essential genes for autophagy. Stromal-targeted DsRed and GFP-ATG8 fusion proteins were observed together in autophagic bodies in the vacuole. We conclude that Rubisco and stromal-targeted fluorescent proteins can be mobilized to the vacuole through an ATG gene-dependent autophagic process without prior chloroplast destruction. The separated using Each experiment was three using independent and the representative data were shown. immunoblotting with anti-GFP or anti-RBCS antibodies. White arrowheads indicate the mature form of RBCS2B-GFP after cleavage of the RBCS2B transit peptide. C, Non-denaturing PAGE analysis of RBCS2B-GFP fusion. Total soluble proteins (10 μ g for gel stain and GFP fluorescence, 1 μ g for immunoblotting) extracted from fresh leaves expressing RBCS2B-GFP fusion (RBCS-GFP) and from fresh leaves of wild type Colombia (wild type) as a control and 0.5 μ g of recombinant GFP (rGFP) were separated by non denaturing-PAGE, and either stained with Coomassie (gel stain), analyzed by an image analyzer, LAS-3000, to detect GFP fluorescence, or analyzed by immunoblotting with anti-RbcL or anti-RBCS antibodies. native holoenzyme white molecular


INTRODUCTION
In C 3 plants, 75 to 80% of total leaf nitrogen is distributed to mesophyll chloroplasts, and most of this nitrogen is allocated into proteins (Makino and Osmond, 1991). The most abundant plant protein is ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco, EC 4.1.1.39), which catalyzes two competing reactions, photosynthetic CO 2 fixation and photorespiratory carbon oxidation. Rubisco accounts for 12 to 30% of total leaf protein in C 3 species (Evans 1989). The degradation of Rubisco and most other stromal proteins begins at an early stage of senescence, and the released nitrogen can be remobilized to growing organs and finally be stored in seeds (Friedrich et al., 1980;Mae et al., 1983). In addition, these proteins are also degraded under carbon-limited conditions, which are caused by darkness (Wittenbach, 1978), and their carbon is used mainly as substrates of respiration. Despite its important function to allow nutrient recycling, the degradation mechanism is not clearly understood yet (for reviews, see Krupinska, 2006;Feller et al., 2007).
Autophagy is known to be a major system for a bulk degradation of intracellular proteins and the mechanism has been studied in depth in yeast and animals (for reviews, see Ohsumi, 2001;Levine and Klionsky, 2004). In those systems, the cytosol, including entire organelles, is engulfed in membrane-bound vesicles that are delivered to the vacuole (yeast) or the lysosome (animals). These vesicles and contents are then degraded by a variety of resident hydrolases. There are two types of autophagy, called microautophagy and seem to be surrounded by isolation membranes, which are considered to be intermediate structures of autophagosomes, in the cytoplasm. These features indicate that only a part of the stroma is pinched off from chloroplasts and is mobilized to the vacuole for the degradation, possibly by autophagy. This concept is consistent with the evidence that the chloroplast number remains constant until late in senescence but the stromal proteins are gradually degraded soon after leaf maturation (Hörtensteiner and Feller U, 2002). Our observations clearly differ from the autophagy of whole chloroplasts, which has been reported by other researchers (Wittenbach et al., 1982;Minamikawa et al., 2001;Niwa et al., 2004).
In this study, we aimed to examine the involvement of the ATG-dependent autophagic process on the mobilization of the most abundant stromal protein, Rubisco, and stromal-targeted fluorescent proteins as a marker for the stromal components to the vacuole via RCBs. Using transgenic Arabidopsis expressing stromal-targeted fluorescent proteins, we first visualized RCBs in living wild-type cells. The atg5 mutation, which compromises the progression of autophagy, disrupted the accumulation of RCBs. In addition, stromaltargeted DsRed and GFP-ATG8 were co-localized in spherical bodies in the vacuole. RCBs could directly be visualized by a RBCS-GFP fusion whose expression was driven by own promoter and which is integrated into a GFP-labeled Rubisco holoenzyme. Thus, Rubisco and likely other stromal-localized proteins can be mobilized to the vacuole by the ATGdependent autophagic process without chloroplast destruction. observed by laser-scanning confocal microscopy (LSCM; Fig. 1A). We treated excised leaves with concanamycin A, an inhibitor of vacuolar H + -ATPase, because it had previously been reported that the interior pH of the vacuole increases in the presence of this drug, resulting in an accumulation of autophagic bodies in the lumen of the vacuole (Yoshimoto et al., 2004). When mature leaves, in which the degradation of Rubisco and major stromal proteins are already progressing, were excised from plants, and then incubated in darkness in the presence of concanamycin A in a carbon-and nutrient-free medium, the GFP signal was also observed in small spherical bodies in the centre of cells, in addition to broad GFP signals in the chloroplast stroma ( Fig. 1, B and E). These bodies are around 1 μm in diameter and do not exhibit chlorophyll autofluorescence. These features are similar to those of the RCBs that were observed in leaf mesophyll cells of wheat fixed for EM (Chiba et al., 2003). The small GFP-labeled bodies exhibit the random motion characteristic of Brownian movement within a cell (see Supplemental Video S1). The movement of the GFPlabeled bodies is clearly different from that caused by cytoplasmic streaming, indicating that the GFP bodies are located in the vacuole. The vacuole is the largest compartment of mature plant cells, such as mesophyll cells, and may occupy as much as 80 to over 90% of the total cell volume (Matile, 1987).
To further confirm the localization of the GFP bodies within cells, cells were released as protoplasts from leaves that were incubated in the same conditions as the leaves described in Figure 1B. GFP bodies were found in the center of a protoplast ( Fig. 2A) and exhibited Brownian movement (data not shown) as previously seen in leaves (Supplementary Video S1). GFP bodies were also found in vacuoles that were isolated from protoplasts (Fig. 2B). Differential interference contrast images clearly indicated that some of spherical bodies, which are localized in the vacuolar sap, are identical to the GFP bodies ( Fig. 2). The GFP bodies were rarely seen when concanamycin A was absent (Fig. 1C). The GFP bodies are likely to be rapidly degraded in the vacuole when concanamycin A is absent. GFP bodies were not seen when leaves were incubated in Murashige and Skoog medium (MS medium), which contains sucrose, even in the presence of concanamycin A (Fig. 1D).
We analyzed the leaves of CT-GFP-containing plants incubated with concanamycin A by IEM with antibodies to RbcL, the large subunit of Rubisco. Spherical bodies containing RbcL, namely RCBs, were found in the vacuolar compartment (Fig. 3A). In addition, double immuno-labeling proved that stromal-targeted GFP was localized in RCBs as well as in chloroplasts (Fig. 3, B and C). When the leaves of CT-GFP-containing plants were analyzed by immunoblotting with anti-GFP antibodies following SDS-PAGE, a single band of 29 kD, which corresponds to the mature form of CT-GFP after cleavage of the transit sequence, was detected irrespective of incubation conditions (Supplemental Fig. S1).
On the other hand, a band of 37-kD that would correspond to the pre-mature form of GFP carrying the RECA-transit peptide was not found even upon overexposure of the immunoblot. These data clearly indicate that GFP bodies found in the vacuole of living cells are RCBs and must be derived from chloroplasts where processing of the CT-GFP transit peptide has occurred.
We analyzed the amount of CT-GFP expressed in order to verify that is does not accumulate at levels unusual for genuine chloroplast proteins (Fig. 4). Although the expression of CT-GFP is driven by a double 35S promoter, it accumulates only to approximately 1% of total soluble protein, which is much less than the amount of Rubisco, which accounts for about 50% of total soluble leaf protein. Taken together with previous observations of RCBs in senescing wild-type wheat leaves (Chiba et al., 2003), these results make it very unlikely that RCBs found in living cells of the transgenic plants could be artifactual products resulting from expression of a fluorescent protein in chloroplasts.
Fluorescence of stromal-targeted GFP was obvious in atg5-1 (Fig. 5A) as well as in wildtype plants (Fig. 1A). Immunoblot analysis revealed that there is no difference in the expression of GFP between wild-type plants and atg5-1 (data not shown). When mature leaves of atg5-1 were incubated in the presence of concanamycin A, the accumulation of spherical bodies labeled with GFP was not observed (Fig. 5B). Many stromules (stromafilled tubules that extend from the surface of plastids; for reviews, see Kwok and Hanson, 2004a;Natesan et al., 2005) appeared in mesophyll cells after incubation irrespective of the addition of concanamycin A (Fig. 5, B-E). Such stromules were not found in similar leaves of wild-type plants (see, Fig. 1, B and C). Stromules are much less common in leaf tissue than in tissue from other plant organs (Köhler and Hanson, 2000).
To elucidate the effect of the atg5 mutation and the leaf age on the appearance of RCBs more clearly, a statistical analysis was performed (Fig. 6). Under the growth conditions used, wild-type plants started to bolt at around 25 days after sowing and finally had 10 rosette leaves on average. The atg5-1 plants grew normally as previously shown ( Fig. 6A, Thompson et al., 2005), but they started to bolt around 2 days earlier than wildtype plants and finally had eight to nine rosette leaves under our growth conditions. In atg5-1 plants, the enhanced chlorosis of leaves was not observed until 30 days. We monitored the expression of two genes in order to verify that leaf senescence is progressing similarly in wild-type and in atg5-1 plants in the leaves used for examination of RCB formation. In lower leaves of atg5-1 plants at 30 days, the decline of RBCS2B transcripts, a marker for the young, photosynthetically active stage of leaf development (Hensel et al., 1993), and the increase of SAG13 transcripts, a marker for early stage of leaf senescence (Weaver et al., 1998), was similar to that observed in wild-type plants (Fig. 6B).
In wild-type plants, the appearance of RCBs was highly related to the leaf age ( Fig.   6C). In mature and early-senescent leaves (leaf 6 and 4 at 30 days), in which the level of RBCS2B has obviously declined compared to the uppermost leaves (leaf 10) or expanding leaves (leaf 6 at 20 days), most mesophyll cells accumulated RCBs, as shown in the image analysis ( Fig. 1B). In contrast, RCBs were rarely seen in expanding (leaf 6 at 20 days) and the young uppermost (leaf 10 at 30 days) leaves. In atg5-1 plants, RCBs were not seen in any stage of leaf development examined (Fig. 6C).
Autophagy in plants can be monitored with the use of transgenic Arabidopsis expressing a GFP-ATG8 fusion protein (Yoshimoto et al., 2004;Contento et al., 2005;Thompson et al., 2005). Spherical bodies labeled with GFP-ATG8, namely autophagic bodies, accumulate in the vacuole when roots are incubated with concanamycin A (Yoshimoto et al., 2004). We examined leaf cells to determine whether autophagic bodies were also present in leaves in the same transgenic plants (Fig. 7). GFP-ATG8 was predominantly cytoplasmic in both epidermal and mesophyll cells of excised leaves ( We next produced transgenic Arabidopsis expressing both stromal-targeted DsRed and GFP-ATG8 and observed the behavior of those proteins (Fig. 8,Supplemental Video S3). DsRed fused to the C-terminal end of the RECA transit peptide correctly labeled the stroma of chloroplasts (Fig. 8A). When the leaves were incubated with concanamycin A, spherical bodies labeled with DsRed, namely RCBs, were found in the vacuole (arrowheads in Fig. 8B). Most of RCBs were also labeled with GFP-ATG8, but there were also some autophagic bodies that are not labeled with DsRed (Fig. 8B, Supplemental Video S3). In addition, a strong signal of GFP-ATG8 was found on a chloroplast protrusion (arrows and insets in Fig. 8B). Some cells contained aggregated structures that are strongly labeled with GFP-ATG8 and DsRed (dashed-line circles in Fig. 8). The aggregated structures were moving slowly in a cell (Supplemental Video S3), indicating that these are located in the vacuole. As the aggregated structures did not exhibit chlorophyll fluorescence, these were not whole chloroplasts. These structures are probably created by the aggregation of autophagic bodies after their transportation into the vacuole. Accumulation of an aggregated form of autophagic bodies in the vacuole has also been reported previously (Liu et al., 2005).

RCBs can directly be visualized by GFP-labeled Rubisco
To monitor the mobilization of Rubisco to the vacuole in living cells directly, we There was no obvious GFP fluorescence in non-green plastids of leaf epidermis and root (data not shown). These results confirmed that RBCS2B-GFP fusion is correctly expressed under the control of the RBCS2B promoter.
When mature leaves of RBCS-GFP transgenic plants were incubated in the presence of concanamycin A, accumulation of small spherical bodies having GFP fluorescence, i.e., similar to the RCBs seen in CT-GFP leaves (Fig. 1B), was observed (Fig. 10B). In contrast, the accumulation of GFP bodies was not observed in leaves of atg5-1 plants expressing the RBCS2B-GFP fusion (Fig. 10C). These results further support the mobilization of Rubisco to the vacuole via RCBs by ATG-dependent autophagy.

DISCUSSION
In this study, we succeeded in visualizing RCBs in the vacuoles of living cells. The atg5 mutation, which compromises the progression of autophagy (Suzuki et al., 2005;Thompson et al., 2005;Phillips et al., 2008), disrupted the accumulation of RCBs. GFP-ATG8 was co-localized with stromal-targeted DsRed in spherical bodies in the vacuole.
Taken together, these data clearly indicate that Rubisco and stromal-targeted fluorescent proteins are partially mobilized to the vacuole in vivo and that this process requires ATG-dependent autophagy.
Based on the experiments reported in this study as well as results from our previous electron-microscopic study (Chiba et al., 2003), and a number of related reports concerning the autophagic process, a proposed mechanism of the mobilization of Rubisco and possibly other stromal proteins to the vacuole for degradation is presented (Fig. 11). The electronmicroscopic study showed that RCBs in the cytoplasm are surrounded by membrane structures, which are morphologically considered to be isolation membranes characteristic of macroautophagy (Chiba et al., 2003). RCBs in the cytoplasm contain stromal proteins, RbcL, RBCS, and glutamine synthetase, but do not contain thylakoid proteins, lightharvesting chlorophyll a/b protein of PS II, α,β-subunits of coupling factor 1 of ATPase, and cytochrome f (Chiba et al., 2003), suggesting that only the stromal portion of chloroplasts is incorporated into RCBs. The mobilization of RCBs by an ATG-dependent system to the vacuole is supported by our finding that GFP-ATG8, which is a marker for isolation membranes, autophagosomes, and autophagic bodies, was co-localized with stromaltargeted DsRed in the autophagic body (Fig. 8) and that the atg5 mutation disrupted the accumulation of RCBs (Figs. 5, 6, and 10).
The RCBs accumulated in the vacuole only when the vacuolar lytic activity was suppressed by the addition of concanamycin A (Fig. 1). In the ATG-dependent system, the outer membrane of an autophagosome fuses with the vacuolar membrane and the innermembrane structure, the autophagic body, is released to the vacuolar lumen and is rapidly disintegrated (Ohsumi, 2001). Therefore, under the usual cellular status in which the vacuolar lytic activity functions normally, RCBs should be rapidly destroyed and the stromal protein contents will be released and degraded by vacuolar proteases as previously shown (Thayer and Huffaker, 1984;Bhalla and Dalling, 1986;Yoshida and Minamikawa, 1996).
In addition to localization on autophagic bodies found in the vacuole, GFP-ATG8 fluorescence was observed on a chloroplast protrusion (Fig. 8), which might be an incipient stromule, such as those occasionally seen in wild-type mesophyll cells. We propose that a chloroplast protrusion or stromule might be sequestered from the main chloroplast body by peroxisomes in yeasts are entirely engulfed by the vacuole in microautophagy, or entirely sequestered into autophagosomes, which then fuse with the vacuole in macroautophagy (Takeshige et al., 1992;Tuttle and Dunn, 1995). In cotyledons of germinated Vigna mungo seeds, microautophagy of a whole starch granule and macroautophagy of a whole mitochondrion were observed (Toyooka et al., 2001). It was reported that abnormal plastids can be removed, possibly by autophagy, in cotyledon cells of Arabidopsis mutants lacking AtTIC40 (Niwa et al., 2004).
At the leaf stage used in this study, autophagy of whole chloroplasts, i.e. mobilization of them to the vacuole, which is monitored by DIC imaging and chlorophyll autofluorescence, was not observed even in the presence of concanamycin A. Why would chloroplasts be partially rather than fully degraded by autophagy? One possible explanation is that the autophagic machinery induced by starvation does not have sufficient size capacity to sequester whole chloroplasts. Chloroplasts are normally above 5 μm in size and are much larger than mitochondria or peroxisomes. Plant autophagic bodies found in roots were approximately 1.5 μm in size and contained mitochondria, endoplasmic reticulum, and Golgi bodies (Yoshimoto et al., 2004). Autophagic bodies found in leaf cells were also this size (Fig. 7, C and D) and therefore not large enough to engulf a 5 μm chloroplast. It could also be advantageous for a starving plant to re-utilize materials from chloroplast proteins without destroying whole chloroplasts. If environmental conditions improve, then a chloroplast that has undergone only loss of some protein content could be rejuvenated and resume normal function. Furthermore, when only a portion of the content of a chloroplast is removed by autophagy, some basal functions of the chloroplast could be maintained under starvation conditions.
In the atg5 mutants, a gradual loss of Rubisco subunits occurred during the extended dark treatment (Thompson et al., 2005). We found that Rubisco also decreases during natural senescence in the atg mutants (Ishida et al., in preparation). These results indicate that chloroplast proteins can be degraded by other mechanisms, such as plastid-localized proteases (Sakamoto, 2006), even if autophagy is disrupted. We observed that selective attenuation of the Rubisco degradation occurs in senescent leaves of the RBCS-antisense rice (Ishizuka et al., 2004). In addition, we observed specific degradation of Rubisco by reactive oxygen under chilling-light conditions (Nakano et al., 2006). These previous findings as well as our current results indicate that chloroplast protein degradation in vivo is a complex and highly regulated phenomenon that merits further investigation.
RCBs were originally identified in naturally-senescing wheat leaves by IEM (Chiba et al. 2003). To visualize RCBs accumulated in the vacuole, we used a 20-hour dark treatment of detached leaves with concanamycin A in the absence of sugar. The accumulation of RCBs in the vacuole was mainly found in mature and senescent leaves and it was rarely observed in young or expanding leaves under these conditions (Fig. 6). These observations are in agreement with prior findings that the degradation of Rubisco mainly occurs in mature and senescent leaves (Mae et al., 1983). RCBs started to accumulate in the vacuole after around 8 hours in this condition (data not shown) while at night, leaves attached with seedlings are often in the dark for 10 hours in the growth conditions used here. Therefore, our present studies are also suggestive of a role of RCBs in natural senescence of Arabidopsis, but further careful experimentation will be needed to understand the physiology of chloroplast stromal recycling by autophagy during the normal life cycle of the plant.
Even when mature leaves were incubated with concanamycin A, few RCBs were detected in the presence of sucrose-containing MS medium (Fig. 1D). It has previously shown that the rate of the Rubisco degradation is accelerated under conditions of carbon or nutrient deficiency (Wardley et al., 1984;Ferreira and Teixeira, 1992). Similarly, plant autophagy is also stimulated under carbon-or nutrient-deprived conditions (Aubert et al., 1996;Moriyasu and Ohsumi, 1996;Yoshimoto et al., 2004;Thompson et al. 2005). In fact, several Arabidopsis mutants with disruptions in the autophagic pathway have shown accelerated leaf senescence and could not survive under severe starvation conditions (Hanaoka et al., 2002;Doelling et al., 2002;Thompson et al., 2005;Xiong et al., 2005;Phillips et al., 2008). It was reported that atg5 mutants die after 6 days of darkness, which is tolerated by wild-type Arabidopsis (Thompson et al., 2005). These results suggest that the degradation of stromal proteins by the ATG-dependent autophagic pathway via RCBs makes a contribution both to the recycling of nutrients for growing tissues and to the retention of chloroplast function under such stress conditions, i.e. carbon or nutrient deficiency rather than natural senescence. A reutilization of nutrients from chloroplast proteins without destroying whole chloroplasts is probably essential for light-and nutrientlimited plants. Examining the effect of nitrogen limitation and other stresses on whole plants with regard to RCB formation and mobilization will be valuable to understand the relative importance of the degradation pathway we have described.

Plant materials and growth conditions
Transgenic Arabidopsis (Arabidopsis thaliana) Columbia ecotype expressing stromal-targeted GFP (CT-GFP) was created as previously described (Köhler et al., 1997;Holzinger et al., 2007). The construct consists of the transit sequence from Arabidopsis      Plastid-targeted DsRed appears red and GFP-ATG8 appears green. In merged images, overlap of DsRed and GFP-ATG8 appears yellow. Stromal-targeted DsRed and GFP-ATG8a were co-localized in a chloroplast protrusion (arrows), and free spherical bodies