A Trafﬁcking Pathway for Anthocyanins Overlaps with the Endoplasmic Reticulum-to-Vacuole Protein-Sorting Route in Arabidopsis and Contributes to the Formation of Vacuolar Inclusions 1[W][OA]

Plants produce a very large number of specialized compounds that must be transported from their site of synthesis to the sites of storage or disposal. Anthocyanin accumulation has provided a powerful system to elucidate the molecular and cellular mechanisms associated with the intracellular trafﬁcking of phytochemicals. Beneﬁting from the unique ﬂuorescent properties of anthocyanins, we show here that in Arabidopsis ( Arabidopsis thaliana ), one route for anthocyanin transport to the vacuole involves vesicle-like structures shared with components of the secretory pathway. By colocalizing the red ﬂuorescence of the anthocyanins with green ﬂuorescent protein markers of the endomembrane system in Arabidopsis seedlings, we show that anthocyanins are also sequestered to the endoplasmic reticulum and to endoplasmic reticulum-derived vesicle-like structures targeted directly to the protein storage vacuole in a Golgi-independent manner. Moreover, our results indicate that vacuolar accumulation of anthocyanins does not depend solely on glutathione S -transferase activity or ATP-dependent transport mechanisms. Indeed, we observed a dramatic increase of anthocyanin-ﬁlled subvacuolar structures, without a signiﬁcant effect on total anthocyanin levels, when we inhibited glutathione S -transferase activity, or the ATP-dependent transporters with vanadate, a general ATPase inhibitor. Taken together, these results provide evidence for an alternative novel mechanism of vesicular transport and vacuolar sequestration of anthocyanins in Arabidopsis.

The accurate delivery and sequestration of chemically reactive and potentially toxic metabolites pose a significant challenge for plant cells, which can simultaneously accumulate hundreds of different phytochemicals, derived from both primary and secondary metabolism. Establishing the cellular and molecular mechanisms that participate in the trafficking of phytochemicals within and between plant cells poses an important biological problem, with significant implications for the engineering of plant metabolism.
Anthocyanins are one of the major classes of plant pigments and serve multiple ecophysiological functions (Grotewold, 2006). Anthocyanins are synthesized from the general phenylpropanoid pathway by the action of a metabolon loosely associated with the cytoplasmic face of the endoplasmic reticulum (ER) and likely forming a multienzyme complex (Winkel-Shirley, 1999;Winkel, 2004). Once synthesized, anthocyanins accumulate in a large central vacuole; this localization is necessary to prevent oxidation (Marrs et al., 1995) and for anthocyanins to function as pigments. In vivo anthocyanin coloration is significantly affected by factors that influence vacuolar pH (Yoshida et al., 1995), the presence of copigments (Forkmann, 1991), and the formation of anthocyanic vacuolar inclusions (AVIs; Markham et al., 2000). Thus, anthocyanins (or anthocyanin precursors) need to be transported from the cytoplasmic surface of the ER to the vacuole.
Over the past few years, several factors that affect proper sequestration of anthocyanins have been identified. Perturbation in modifications of the core anthocyanidin skeleton required for uptake by the transporters leads to accumulation of the flavonoid in the cytoplasm. In maize (Zea mays), impairment of the UDP-Glc:cyanidin 3-O-glucosyltransferase gene BRONZE1 (BZ1) suppresses anthocyanin accumulation (Larson and Coe, 1977;Fedoroff et al., 1984). Mutations in the maize BZ2 gene, which encodes a glutathione (GSH) S-transferase (GST), prevent vacuolar localization of anthocyanins and brown oxidation products accumulate (hence, the name BZ2; Marrs et al., 1995). Similarly, the petunia (Petunia hybrida) AN9 gene encodes a GST and, despite the low identity between AN9 and BZ2, BZ2 complements AN9 mutants (Alfenito et al., 1998). Interestingly, the GST enzymatic activity of AN9 is not required for the AN9-dependent vacuolar sequestration of anthocyanins, suggesting that AN9/BZ2 serves as ligandins most likely for stabilization, but possibly also for escorting anthocyanins (e.g. cyanidin 3-glucoside) from the ER to the tonoplast (Mueller et al., 2000). Identification of the ZmMRP3 (maize tonoplast-localized multidrug resistance-associated protein), induced by the C1 and R anthocyanin regulators (Bruce et al., 2000), provides an additional player in a model involving carrier and transporter proteins in the trafficking of anthocyanins from the ER surface to the vacuole (Goodman et al., 2004). In Arabidopsis (Arabidopsis thaliana), mutants in TRANSPARENT TESTA19 (TT19) affect both anthocyanin accumulation in vegetative tissues and proanthocyanidin (PA) accumulation in seed coats. TT19 encodes a GST and AN9 complements the anthocyanin, but not the PA defect of the tt19 mutant (Kitamura et al., 2004). Whereas TT19 and AN9/BZ2 may function similarly by stabilizing/ escorting anthocyanins, the TT19 mutant has a distinctive phenotype in the seed coat, where PA precursors accumulate in cytoplasmic membrane-wrapped structures (Kitamura et al., 2004). This contrasts with the phenotype of mutations in the TT12 locus, encoding a multidrug and toxic compound extrusion transporter involved in PA vacuolar sequestration in which the PA precursors are evenly distributed in the cytoplasm (Debeaujon et al., 2001).
Plant cells contain at least two different types of vacuolar compartments (Paris et al., 1996), which are most often referred to as the lytic and the protein storage vacuoles (PSVs). PSVs can be compound organelles, evidenced by the presence in tobacco (Nicotiana tabacum) seeds of a subvacuolar membranebound compartment containing organic acids and proteins (Jiang et al., 2001). The secretory pathway is responsible for the vacuolar transport of proteins through the interaction of specific sorting signals in the proteins and vacuolar-sorting receptors. The major route of vacuolar protein transport is from the ER through the trans-Golgi network (TGN) complex, a route that is shared among all eukaryotes (Neumann et al., 2003;Vitale and Hinz, 2005). However, a direct trafficking route from the ER to the vacuole exists in plants, which was first identified for the transport of proteins targeted to the PSV by large vesicles known as precursor-accumulating vesicles (Hara-Nishimura et al., 1998). Spindle-shaped ER bodies (Matsushima et al., 2003) provide additional possible vehicles for the transport of proteins, rubber, or oil from the ER to the vacuole by a mechanism resembling autophagy (Herman and Schmidt, 2004). Whether ER bodies are involved in the transport of PAs or anthocyanins from the ER to the vacuole remains unclear, but the localization of Arabidopsis flavonoid biosynthetic enzymes to large electron-dense cytoplasmic structures and to the tonoplast (Saslowsky and Winkel-Shirley, 2001) suggests that mechanisms other than cytoplasmic flavonoid carrier proteins are at play in the subcellular trafficking of anthocyanins. Most significant in highlighting a vesicular transport for flavonoids is the recent description of the tapetosomes as ER-derived structures that store ER-derived flavonols for their delivery to the Brassica pollen surface upon tapetal cell death (Hsieh and Huang, 2007).
Taking advantage of unique red fluorescent and colored properties of anthocyanins, we describe here the colocalization of anthocyanins with vesicle-like structures containing a protein marker (GFP-Chi) for the PSV in Arabidopsis. Consistent with a TGNindependent ER-to-vacuole vesicular transport of anthocyanins, Brefeldin A (BFA), a Golgi-disturbing agent (Dinter and Berger, 1998), has no effect on the accumulation of anthocyanins and the red fluorescent anthocyanins are detected in ER compartments identified by GFP fused to an ER retention signal (GFP-HDEL). We describe the accumulation of anthocyanins in the vacuole in neutral red (NR)-staining subvacuolar compartments. In sharp departure from what has been observed in other plants, treatment with ATPbinding cassette (ABC) transport inhibitors does not significantly decrease the amount of anthocyanins. However, vanadate, a fairly general inhibitor of ATPases, including ABC transporters, induces a dramatic increase of anthocyanin-filled subvacuolar structures.
Our results indicate that Arabidopsis cells accumulating high levels of anthocyanins utilize components of the protein secretory trafficking pathway for the direct transport of anthocyanin pigments from the ER to the vacuole and provide evidence for the existence of novel subvacuolar compartments for their storage.

Induction of Anthocyanin Accumulation in Arabidopsis Seedlings
To induce high anthocyanin levels in young seedlings, we grow seeds for 2 to 3 d under high light conditions in plain liquid Suc medium without a nitrogen source (anthocyanin inductive condition; see ''Materials and Methods''). If tt5 seedlings are grown in similar conditions (Fig. 1A), no pigmentation is observed because of the absence of the chalcone isomerase (CHI) enzyme encoded by the TT5 locus (Shirley et al., 1992). However, if the product of CHI, naringenin (50-200 mM), is added to the medium, high levels of anthocyanins are observed (Fig. 1A) within 8 to 10 h, reaching a maximum at about 24 h (see below). The addition of naringenin further increases (1.5-to 2-fold) the anthocyanin accumulation levels of wild-type seedlings grown in anthocyanin inductive conditions (Fig.  1B). In addition, because the inductive conditions suppress chlorophyll synthesis in the cotyledons, there was no background color interference with the anthocyanins.
These results indicate that treatment of wild-type or tt5 seedlings grown under anthocyanin inductive conditions with naringenin provides a good system for high levels of anthocyanin production in Arabidopsis.

Novel Fluorescent Properties of Arabidopsis Anthocyanins
The fluorescence provided by the ring-stacking interaction of flavonol and flavone aglycones with diphenylboric acid (DPBA) has been utilized to investigate the localization of several flavonoids (Buer and Muday, 2004;Peer and Murphy, 2006;Vargo et al., 2006;Hsieh and Huang, 2007). However, DPBA does not fluoresce with anthocyanins, prompting us to seek another means for cytoplasmic visualization of these compounds.
To determine whether anthocyanins fluoresce in a spectral range that would allow the visualization of these compounds in the presence of GFP markers of the endomembrane trafficking system, we investigated the fluorescence properties of Arabidopsis anthocyanins. Mutant tt5 seedlings grown in anthocyanin inductive conditions in the absence of naringenin showed no fluorescence in the red channel when excited at 488 and 544 nm of the argon-ion and helium-neon lasers, respectively (emission .565 nm; Fig. 2A). However, when incubated in the presence of naringenin, tt5 seedlings displayed strong fluorescence in the red channel ( Fig. 2A). Two mutants, tt6 and tt3, that block anthocyanin production downstream of the step catalyzed by TT5, were tested for accumulation of fluorescence. The absence of red fluorescence in naringenin-treated tt6 (data not shown) and tt3 seedlings ( Fig. 2A) indicates that the fluorescence was not due to naringenin itself nor to a metabolic byproduct of naringenin, but rather a consequence of the presence of a flavonoid after the enzymatic step catalyzed by dihydroflavonol 4-reductase. Leucocyanidin, however, showed no fluorescence (data not shown). Similar red fluorescence was observed in wild-type seedlings grown in anthocyanin inductive conditions both in the presence or absence (data not shown) of naringenin ( Fig To demonstrate that red fluorescence was due to the anthocyanidins/anthocyanins and not to another pathway intermediate, acid-hydrolyzed methanol extracts from wild-type (Ler) and tt5 seedlings were separated on a cellulose thin-layer chromatography (TLC) plate. As previously described (Dong et al., 2001), a single spot corresponding to cyanidin was observed, which was absent in tt5 seedlings (Supplemental Fig.  S1A). Under UV light (approximately 254 nm), this spot fluoresces red. The cyanidin spot of the TLC plate was imaged using confocal laser-scanning microscopy using the same excitation and emission wavelengths as used for microscopy of the seedlings. Cyanidinloaded cellulose fluoresced red when excited at 488/ 544 nm and visualized using the long-pass emission filter of 565LP. No fluorescence was observed using the 515-to 530-nm emission filter. The blank sample, a cellulose spot below the origin, did not fluoresce (Supplemental Fig. S1B).
To conclusively prove that the red fluorescence observed during microscopy did come from the anthocyanin/anthocyanidin, we measured the emission spectra from the cyanidin spot isolated from the TLC plate using fluorescence spectrophotometry. The cyanidin spot was extracted from the cellulose plate using 95% ethanol. Absorption and fluorescence spectra ( Fig. 2B) were obtained. The absorption maximum of cyanidin in ethanol was established to be 547 nm and the emission maximum was 595 nm at the excitation wavelength of 544 nm. The 595-nm emission maximum was observed regardless of the excitation wavelength, which ranged from 280 to 544 nm (data not shown), with no peaks appearing in the 500-to 530-nm wavelengths, corresponding to the GFP emission spectrum. Finally, to confirm that the main red-fluorescent compounds in cells expressing anthocyanins correspond to anthocyanins themselves, methanolic extracts of PAP1-D (Borevitz et al., 2000) plants were separated by reverse-phase HPLC and the absorption (530 nm) and fluorescence spectra (excitation at 540 nm and emission at 620 nm) were compared (Fig. 2C). The two spectra show very good correspondence, indicating that all major Arabidopsis anthocyanins (Tohge et al., 2005) fluoresce red. Together, these results conclusively prove that anthocyanins fluoresce in a range compatible with the utilization of GFP as a marker to follow subcellular trafficking pathways.
Anthocyanins Share a Golgi-Independent, Vesicular Trafficking Pathway with Proteins Targeted to the PSV The plant secretory system involves multiple pathways for the transport of proteins to the vacuole (Carter et al., 2004), and GFP fusion markers (Chalfie et al., 1994) permit distinguishing between them (Neuhaus, 2000;Di Sansebastiano et al., 2001). To establish whether the ER or ER bodies are a possible initial site of anthocyanin accumulation, as previously suggested for maize (Grotewold et al., 1998) and recently described for flavonols in Brassica and Arabidopsis tapetum cells (Hsieh and Huang, 2007), Arabidopsis seedlings transformed with GFP-HDEL (Haseloff et al., 1997), where HDEL corresponds to an ER-retention signal sequence, were grown under anthocyanin inductive conditions, with (1N) or without (2N) naringenin (Fig. 3). Intact seedlings (Fig. 3, A-H) and protoplasts (Fig. 3, I-P) were directly observed by confocal laser-scanning microscopy. The GFP-HDEL marker provided green fluorescence to spindle-shaped ER bodies in the cotyledonary cells (Fig. 3, A and I). After incubating the seedlings for 24 h with naringenin, the number of green-fluorescent bodies increased and the bodies appeared more dilated (Fig. 3, compare A and E). The red channel showed the presence of red fluorescence occupying most of the cell, which correlated with the vacuolar pigmentation provided by the anthocyanins. In addition, red fluorescence was observed in spindle-shaped bodies, the number and size of which dramatically increased in seedlings treated with naringenin (Fig. 3, compare B and F). These spindle-like structures were clearly outside the large central vacuole, further evidenced by three-dimensional reconstructions from the confocal images (data not shown). Observation under both the red and green channels showed the colocalization of the red and green fluorescence in the ER bodies ( Fig.  3, C, G, K, and O). The very strong red fluorescence of the anthocyanins in the vacuole, which occupy more than 80% of the cellular volume in these cotyledon epidermal cells, made it difficult to establish whether there was any red anthocyanin fluorescence in the cytoplasm that was not associated with the GFP-HDEL marker. Interestingly, however, not all the GFP-HDEL marker was found to be associated with the red fluorescence, suggesting that either the presence of two populations of ER bodies, some filled with anthocyanins and others not, or the levels of red fluorescence in green-but not red-fluorescing bodies was below the level of detection. Taken together, these findings indicate that at least a part of the red-fluorescing anthocyanins colocalize with the GFP-HDEL marker in ER bodies.
To investigate the possible transport route of anthocyanins from the ER bodies to the vacuole, we utilized intact plants (Fig. 4, A-D) or isolated protoplasts (Fig.  4, E-H) of transgenic Arabidopsis lines expressing vacuolar-sorting signals fused to GFP. The GFP-Chi marker, corresponding to a fusion of the GFP to the C-terminal vacuolar-sorting determinant from the barley (Hordeum vulgare) chitinase A protein, is targeted to pH-neutral PSVs directly from the ER in a Golgiindependent manner (Di Sansebastiano et al., 1998Sansebastiano et al., , 2001Fluckiger et al., 2003). Arabidopsis seedlings transgenic for 35STGFP-Chi grown in anthocyanin inductive conditions for 3 d show green fluorescence provided by GFP-Chi in discrete structures that could  correspond to the ER and to small peripheral vacuoles (Fig. 4A). These seedlings accumulated anthocyanins in the epidermal cells of the cotyledons in the form of uniform vacuolar red fluorescence and in discrete cytoplasmic structures (Fig. 4B, arrow) that often colocalized with GFP fluorescence (Fig. 4, A-C). Consistent with previous findings (Di Sansebastiano et al., 1998;Fluckiger et al., 2003) that showed that chloroplastpoor cells failed to accumulate GFP-Chi in the large central vacuole, we did not observe a colocalization of red and green fluorescence in the central vacuole (Fig.  4, C and G). However, this could also be a consequence of the more acidic pH of the vacuole affecting the GFP fluorescence and not necessarily that the GFP-Chi does not accumulate there. Similar to what we observed for GFP-HDEL (Fig. 3), not all the small structures that accumulated GFP-Chi accumulated fluorescent anthocyanins.
Consistent with this, we observed that d-TIP-marked vacuoles accumulated anthocyanins as seen in the colocalization of red anthocyanin fluorescence in vacuoles with d-TIP-GFP in epidermal cells (Fig. 4, I-L) and protoplasts (Fig. 4, Q-T).
To explore whether anthocyanins would also colocalize with components of the secretory pathway that utilize the TGN for transport from the ER to the vacuole, we utilized Arabidopsis lines expressing an N-terminal vacuolar-sorting determinant from the barley aleurain fused to GFP (Ale-GFP; Di Sansebastiano et al., 2001). Epidermal cells of 35STAle-GFP-expressing seedlings grown under anthocyanin inductive conditions (Fig. 4, M-P) showed small green-fluorescent bodies, likely corresponding to lytic vacuoles (Fluckiger et al., 2003), and smaller punctuated structures marked with Ale-GFP peripheral to the large central vacuole (Fig. 4M). No colocalization of Ale-GFP and red fluorescence was observed (Fig. 4O), suggesting that anthocyanins follow the direct ER-to-vacuole route, rather than going through the Golgi pathway. To conclusively establish that the observed vesicular trafficking of anthocyanins did not involve the TGN, we investigated the effect of BFA, a Golgi-disturbing agent (Driouich et al., 1993;Satiat-Jeunemaitre et al., 1996), on the accumulation of anthocyanins and the formation of AVIs. After incubating 2.5-d-old tt5 seedlings with BFA (10 mg/mL) for 1 h, we added 100 mM naringenin and measured the amount of anthocyanins that accumulated after 24 h. No difference was observed in the levels of anthocyanins when comparing BFA-treated and nontreated seedlings, nor did we observe any effect of BFA on the formation of the AVIs (Fig. 5A, yellow line). Consistent with the BFA treatment affecting the TGN-dependent transport, and providing evidence that BFA was effective in disturbing the TGN under the conditions tested, we observed that the green fluorescence furnished by Ale-GFP was significantly different after BFA treatment, indicating a likely retention in ER-like structures (Supplemental Fig. S2). In addition, the protein-sorting inhibitor Sortin 1, which interferes with the TGN-dependent vacuolar transport of proteins (Zouhar et al., 2004), had no effect on the ability of tt5 seedlings grown under anthocyanin inductive conditions to accumulate pigments or form AVIs when complemented with naringenin (data not shown). Taken together, these results indicate that anthocyanins can utilize a TGNindependent vesicular transport from the ER to the vacuole that at least in part overlaps with proteintrafficking pathways to the PSV.

Anthocyanin-Accumulating Subvacuolar Structures in Arabidopsis
The normally low anthocyanin pigment accumulation of Arabidopsis vegetative green tissues is dramatically enhanced in PAP1-D plants, resulting from the overexpression of the PAP1 R2R3-MYB anthocyanin regulator (Borevitz et al., 2000;Tohge et al., 2005). Yet, the PAP1-D pigmentation phenotype is usually not observed until plants are 2 to 3 weeks old. The microscopic observation of pigmented tissues in the PAP1-D plants revealed, in a fraction of the pigmented epidermal cells, the presence of small anthocyanin inclusions that appeared as rounded spherical structures, apparently within the large central vacuole (Fig.  6A, PAP1-D). At a much lower frequency, similar structures were also observed in mature wild-type Ler plants grown under high light conditions (Fig. 6B, Ler).
NR provides a vital vacuolar stain that diffuses through membranes, but is trapped in the acidic vacuolar compartment by protonation (Ehara et al., 1996;Di Sansebastiano et al., 1998). Staining of PAP1-D leaves with NR showed the presence of NR-staining bodies in over 70% of epidermal cells. These NRstaining bodies were similar in shape and size to the anthocyanin inclusions, but were present in wild type in a larger number of cells (Fig. 6B, Ler). These NRstaining structures (but not anthocyanin inclusions) were also found, although at a lower frequency, in tt5 plants grown under either normal or anthocyanin inductive conditions (Supplemental Fig. S3; Fig. 6B, tt5).
To determine whether the anthocyanin inclusions were inside the vacuole or whether they corresponded to a separate NR-staining acidic compartment, vacuoles were isolated from PAP1-D plants (see ''Materials and Methods''). The NR-staining and anthocyaninaccumulating bodies were always observed inside the large central vacuole (Fig. 6, C and D), indicating that they most likely correspond to subvacuolar structures. For clarity purposes and to avoid introducing one additional name for these structures, we will refer to them here as AVIs. The anthocyanin pigmentation of the AVIs was more intense than in the rest of the vacuole (Fig. 6C), indicating that anthocyanins, although present in the vacuolar sap, were enriched in the AVIs. Similarly, NR was preferentially sequestered in these subvacuolar compartments, staining these structures darker than the surrounding vacuole (Fig.  6D). Taken together, these results demonstrate the presence in Arabidopsis of novel AVI-like structures that accumulate anthocyanins and suggest that they are either more acidic than the rest of the vacuolar sap  (and hence are likely membrane bound), contain compounds with affinity for NR (such as other phenolics [Stadelmann and Kinzel, 1972]), or a combination of both.

Participation of ABC Transporters and GSTs on AVI Formation
Vanadate significantly reduces anthocyanin accumulation in maize cells (Marrs et al., 1995). To investigate the effect of vanadate in the accumulation of anthocyanins and in the formation of AVIs in Arabidopsis, tt5 seedlings were grown in anthocyanin inductive conditions for 2.5 d and treated with 1 mM vanadate 1 h prior to the addition of 100 mM naringenin. Whereas anthocyanins take longer to accumulate in the vanadate-treated seedlings compared to the untreated control (Fig. 5A), a delay explained by reduced naringenin uptake in the first 15 h (Fig. 5B), there was little difference in anthocyanin accumulation between vanadate-treated and control tt5 seedlings with naringenin after 24 h (Fig. 5A). However, when seedlings were observed under the microscope, a dramatic increase in the number of AVIs was noticed in vanadate-treated compared to control seedlings (Fig. 7, compare A and B and C and D). Nearly every cell contained AVIs, clearly visible even in the absence of NR. The bathochromic shift of the AVIs (from purple red to bluish; Fig. 7D) reflects the alkalinization of the vacuole, and the incubation of the vanadatetreated seedlings for a short time in diluted acid conditions rapidly restores a bright pink color to the vacuole (data not shown). The addition of vanadate did not result in immediate alkalinization of the medium in which the seedlings were grown. In contrast, when vanadate was added to tt5 seedlings in the absence of naringenin, no significant difference in the number of subvacuolar structures staining with NR Figure 7. AVIs are formed in the presence of transport inhibitors. A to H, AVI formation in tt5 seedlings treated with naringenin, together with inhibitors affecting ABC transporters (vanadate, Na 3 VO 4 ; C and D), cellular GSH levels (BSO; E and F), or GST enzymatic activity (CDNB; G and H). Similar areas of the cotyledon (A, C, E, and G) were observed to avoid variance due to positional and development effects. The vacuoles of vanadate-treated cells were more alkaline, reflected in the bluish hue of the anthocyanins, whereas cells treated with BSO or CNDB accumulate more AVIs without the same effect on pH. Seedlings were observed 24 h after the addition of naringenin and the various inhibitors. Scale bars 5 10 mm.  Figure S3 for images representing the cells from which the data were obtained. C, AVIs in isolated vacuoles from PAP1-D plants. D, Isolated vacuoles of PAP1-D showed strong NR uptake by subvacuolar structures when compared to the sap, suggesting that they corresponded to acidic, membrane-bound subvacuolar compartments. Scale bars 5 10 mm. was observed (data not shown), suggesting that the observed increase in subvacuolar structures by vanadate is dependent on the presence of anthocyanins. From these results, we conclude that anthocyanins can accumulate in Arabidopsis even in the presence of inhibitors of ABC transporters and that the inhibition of ABC transporters results in the increased number of AVIs, suggesting that their formation (or filling) does not require ATP-energized transporters.
A major function of plant ABC transporters, particularly from the MRP family, is to pump conjugates of potentially toxic compounds with GSH to the vacuole (Klein et al., 2006). To establish the participation of GSH or GSTs in the accumulation of anthocyanins and in the formation of AVIs, we treated tt5 seedlings grown in anthocyanin inductive conditions with 100 mM naringenin and with 1 mM buthionine sulfoximine (BSO), which depletes cellular GSH levels, or with 0.1 mM 1-chloro-2-4-dinitrobenzene (CDNB), a common GST substrate that saturates the enzymes, decreasing the activity on other substrates. Similarly, as observed with vanadate, both treatments resulted in significant increase in the accumulation of AVIs, but without the bathochromic shift (Fig. 6, E-H). Neither CDNB nor BSO treatments resulted in a significant effect in the total levels of anthocyanins (data not shown). Taken together, these findings indicate that the inhibition of the synthesis or transport of glutathionated compounds to the vacuole results in an increase in the formation of AVIs without an obvious effect on total anthocyanin accumulation.

DISCUSSION
Despite the fundamental importance for plants to properly transport and sequester phytochemicals, little is known about the molecular and cellular mechanisms involved in these processes. Taking advantage of novel anthocyanin red autofluorescence properties in combination with protein markers for the secretory pathway, we describe here a TGN-independent ERto-vacuole vesicular anthocyanin-trafficking route shared with proteins targeted to the PSV. We also uncover the presence of novel Arabidopsis anthocyaninaccumulating subvacuolar structures that resemble the anthocyanoplasts/AVIs present in the pigmented tissues of many other plant species.
Establishing trafficking pathways for anthocyanins has been complicated by the fact that the color of the compounds depends on the proper conditions (pH and modifications) furnished by the vacuole. Anthocyanin extracts from red cabbage (Brassica oleracea) were previously shown to fluoresce with peaks at 363, 434, and 519 nm (Drabent et al., 1999). Our studies, however, identified significant fluorescence in vivo for total anthocyanins and for individual pigments above 565 nm (Fig. 2), making this fluorescence compatible with the visualization of GFP. The difference in our results with those previously reported is likely a consequence of the red cabbage extracts containing complex mixtures of anthocyanins with other phenolics and proteins. Indeed, when the red cabbage extracts were subjected to chromatographic separation, one of the peaks (peak 10; Drabent et al., 1999) displayed significant fluorescence increase in the 550-to 650-nm range (Drabent et al., 1999). Our results, demonstrating that anthocyanins can have fluorescence properties compatible with GFP visualization, pave the way for similar colocalization studies to be carried out in other plants. Autofluorescence provides a significant advantage over the use of flavonoid stains such as DPBA because it can be visualized in vivo, without disturbing the cellular organization.
Taking advantage of the fluorescent properties of anthocyanins, we exposed a trafficking mechanism for these compounds from the ER to the vacuole that involves membrane-bound structures that initially contain the ER marker GFP-HDEL (Fig. 3). The shape and induction of these structures in GFP-HDELexpressing plants make them likely candidates for being ER bodies (Matsushima et al., 2003), which correspond to ER-derived cytoplasmic structures proposed to be transferred to the vacuole by mechanisms that include autophagy (Herman and Schmidt, 2004). We established that the red-fluorescing anthocyanins colocalized with the PSV-targeted marker GFP-Chi (Fig. 4), which uses a TGN-independent ER-to-vacuole trafficking mechanism. The TGN-independent vesicular trafficking of anthocyanins was further confirmed by the observation that anthocyanin accumulation is insensitive to BFA (Fig. 5A) and that the red fluorescence did not colocalize with a marker (Ale-GFP) that utilizes a TGN-dependent pathway (Fig. 4). These results suggest that anthocyanins may hitchhike on the protein secretory pathway for transport from the ER to the tonoplast. It is, however, unclear whether the accumulation of anthocyanins in GFP-HDEL-containing structures precedes their localization in the GFP-Chi vesicles or whether these reflect two separate mechanisms by which anthocyanins can reach the vacuole in membrane-bound structures. However, the colocalization of the red-fluorescing anthocyanins with the GFP-Chi marker (Fig. 4), which did accumulate in the ER and in ER-derived structures (Fluckiger et al., 2003), highlights that the presence of the pigments in ER bodies was unlikely driven by the expression of GFP-HDEL, which sometimes results in ER body formation, possibly because of ER retention or retardation of the fusion proteins (Herman and Schmidt, 2004). The visualization of anthocyanins in the ER bodies could have been furnished by a higher concentration of the pigment in the dilated ER. Anthocyanin fluorescence was not detected in the thin reticulate cortical ER, which could be a consequence of either low signal, below the detection limits, or anthocyanins accumulating in only specific domains of the ER. This latter possibility would be consistent with the apparent exclusion of anthocyanins from some of the ER bodies (Fig. 3).
In many plant species, anthocyanins accumulate in the vacuole in discrete structures described by a variety of names (Pecket and Small, 1980;Nozzolillo and Ishikura, 1988;Nozue et al., 1993;Kubo et al., 1995;Markham et al., 2000;Conn et al., 2003;Irani and Grotewold, 2005;Zhang et al., 2006). We found here that intravacuolar anthocyanin-accumulating inclusions are also present in Arabidopsis, particularly in cells induced to accumulate high anthocyanin levels, either as a consequence of the expression of the PAP1 regulator or by the addition of the pathway intermediate, naringenin. These inclusions stained heavily with NR (Fig. 5), a vital dye that gets trapped by protonation into acidic compartments. Generally, the number of NR-staining intravacuolar bodies present was larger than the structures heavily pigmented with anthocyanins (Fig. 6B). This may indicate that NR stained all the AVIs, but only those with high levels of anthocyanins were visible in the absence of NR. Alternatively, there might be different types of subvacuolar structures, only some of them capable of accumulating anthocyanins. Interestingly, however, in vanadate-, BSO-, or CDNB-treated seedlings, most of these subvacuolar structures were filled with anthocyanins ( Fig. 7; see below). NR-staining structures were also found in plants lacking anthocyanins, such as, for example, tt5 mutants. This suggests that the formation of these structures may not be triggered by the accumulation of anthocyanins. More likely, once anthocyanins reach the vacuole, they enter preexisting NR-staining bodies, resulting in the characteristic coloration of AVIs.
To investigate the possibility that an autophagic mechanism (Marty, 1978) is involved in the formation of the subvacuolar structures, we looked into whether a mutation in the ATG7 locus (atg7-1 in the Wassilewskija [Ws] genetic background) affects the formation of the NR-staining structures or the formation of AVIs. ATG7 encodes the Arabidopsis E1-like ATP-dependent activating enzyme required for autophagy (Doelling et al., 2002), previously known as APG7 (Klionsky et al., 2003). We could not detect any significant difference in the number of NR-staining subvacuolar structures or AVIs (under anthocyanin inductive conditions) between atg7-1 and Ws (data not shown). However, Ws seedlings accumulated less anthocyanins and had a significantly lower number of AVIs when compared to Ler or Columbia seedlings (data not shown), indicating that natural variation among accessions influences the physiology of these subvacuolar compartments, something that needs to be taken into consideration when comparing mutants. To further eliminate a possible role of autophagy, we investigated the effect of 3-methyladenine (3-MA), a potent autophagy inhibitor in animal (Seglen and Gordon, 1982) and plant (Takatsuka et al., 2004) cells, on anthocyanin accumulation and AVI formation. 3-MA functions by inhibiting the PI3K enzyme necessary for the nucleation of preautophagic structures (Thompson and Vierstra, 2005). Treatment of 3-d-old tt5 seedlings grown in anthocyanin inductive conditions with 10 mM 3-MA and 100 mM naringenin resulted in similar anthocyanin levels and number of AVIs (data not shown), yet affected, as expected, the distribution of the GFP-Chi marker (Supplemental Fig.  S3). These results led us to conclude that a classical autophagic mechanism is unlikely to be involved in the formation of AVIs.
The existence of a vesicular-type transport of anthocyanins from the ER to the vacuole provides an alternative to models that involve AN9/BZ2-like GST carrier proteins and/or tonoplast transporters for the cytoplasmic and tonoplast trafficking of these compounds, respectively (Alfenito et al., 1998;Mueller et al., 2000;Mueller and Walbot, 2001;Goodman et al., 2004). Interestingly, whereas the tt19 mutation completely abolishes anthocyanin and PA accumulation (Kitamura et al., 2004), perturbing the formation or vacuolar uptake of GSH conjugates (GS-X) with CDNB or BSO or inhibiting ABC transporters with vanadate increases the number of AVIs (Fig. 7) without a significant effect on anthocyanin accumulation (Fig. 5). This is in sharp contrast to what has been previously found in maize, where vanadate treatment phenocopies the bz2 mutation with respect to anthocyanin accumulation (Marrs et al., 1995), suggesting distinct mechanisms of action of TT19 and ATP-energized transport mechanisms. Whereas bz2 mutants accumulate brown pigments in the cytoplasm (Marrs et al., 1995), tt19 mutants lack significant amount of pigments, even under anthocyanin inductive conditions in the presence of 200 mM naringenin (data not shown). Similarly, the petals of petunia an9 mutants are colorless (Mueller et al., 2000). These differences could suggest that distinct biochemical products result in maize, petunia, and Arabidopsis from the blockage in the BZ2/AN9/TT19 steps. However, this possibility is unlikely, given that the proteins seem to be largely exchangeable between different plant species with regard to anthocyanin accumulation (Alfenito et al., 1998;Mueller et al., 2000;Larsen et al., 2003;Kitamura et al., 2004). Alternatively, it is possible that TT19 has additional functions than those proposed for BZ2. This could explain the ability of BZ2 to complement the anthocyanin deficiency phenotype of tt19 mutants, but not PAs, and the presence of vesicles filled with PA precursors in seed coat endothelial cells (Kitamura et al., 2004;Kitamura, 2006). Further highlighting an additional role of TT19, flavonoids could not be transported across the ER membrane to then form part of the tapetosomes in tt19 Arabidopsis tapetum cells (Hsieh and Huang, 2007).
The results presented here provide a new perspective with regard to ER-to-vacuole trafficking and vacuolar sequestration of anthocyanin pigments, and maybe of other vacuole-targeted phenolic compounds as well. Whereas our results do not rule out the existence of other mechanisms for transport of anthocyanins to the vacuole, such as the interplay of GSTs and tonoplast transporters (Goodman et al., 2004), they highlight the existence of vesicular transport of anthocyanins with properties shared with the secretory pathway. Cellular, molecular, and genetic tools are becoming increasingly available in Arabidopsis to further dissect the mechanisms by which anthocyanins are transported and sequestered in the vacuole.

Plant Materials and Growth Conditions
GFP-HDEL (Haseloff, 1999), GFP-Chi (Di Sansebastiano et al., 1998), and Ale-GFP (Di Sansebastiano et al., 2001) were used as GFP-expressing lines. Arabidopsis (Arabidopsis thaliana) CHI (tt5-1), flavanone 3-hydroxylase (tt6), dihydroflavonol reductase (tt3), and PAP-1D seeds were obtained from the Arabidopsis Biological Resource Center. For induction of anthocyanins in seedlings (anthocyanin inductive conditions), seeds were surface sterilized and plated in water containing 3% Suc. After 2 d of stratification at 4°C, seeds were germinated for 2 to 4 d at 25°C 6 2°C in continuous cool-white light (GE F30T12-CW-RS) at approximately 100 6 10 mmol m 22 s 21 on a rotary shaker at 100 rpm. For naringenin treatments, seedlings were allowed to grow for 2.5 d and then naringenin (Aldrich) was added to a final concentration of 100 or 200 mM from a 100 mM stock (in ethanol). Treatments with various chemicals were carried out after seedlings were germinated for 2.5 d (unless otherwise indicated). Seedlings were preincubated with each inhibitor (Sigma) for 1 h at 25°C 6 2°C before the addition of naringenin. Only 3-MA was added 12 h before the naringenin treatment. The final concentrations were 1 mM for vanadate (stock solution 1 M sodium orthovanadate in water), 10 mg/mL for BFA (stock solution 10 mg/mL in ethanol), and 10 mM for 3-MA (stock solution 1 M in water). Each treatment was done at least in triplicate. Soil-sown seeds were stratified at 4°C for 2 d and transferred to a growth chamber at 22°C 6 2°C with a 16-h dark/8-h light photoperiod.

Anthocyanin Extraction, Analysis, and Quantification
After different treatments, seedlings were harvested, rinsed with water, and lyophilized for 2 d. Dry weight was measured and 50% methanol was added to get a final suspension of 50 mg/mL (w/v). Two volumes of acidic methanol (1% HCl in 50% MeOH) were added and absorption read at 530 nm using a Cary 50 UV-VIS spectrophotometer (Varian) in 40-mL quartz microcuvettes. The fluorescence spectra of anthocyanins were determined on a Flex station spectrofluorimeter (Molecular Devices), with readings taken at 10-nm intervals. Aglycones were obtained by boiling the methanolic extracts containing 1 M HCL for 20 min. For TLC experiments, anthocyanidins were extracted by adding one-fourth of the original volume of isoamyl alcohol and separated on cellulose TLC plates (5,730/6; Merck) in a presaturated chamber with water:formic acid:HCl (10:30:3).
HPLC analysis of flavonoids and anthocyanins was carried out by separating 20 mL of the methanolic extract on a C-18 column using a Waters Alliance 2695 separations module equipped with a 2996 photodiode array detector and a fluorescence detector (Waters Corporation). Flavonoids were separated using solvent A: 5% formic acid in water; solvent B: 5% formic acid in acetonitrile; 0 to 30 min, 95% A to 70% A, linear gradient; 30 to 35 min, 70% A to 95% A, linear gradient. Chromatograms and spectra were extracted and analyzed with Empower software (Waters Corporation).

Protoplast and Vacuole Isolation
Plant tissue (0.4 g) was sliced into pieces with a razor blade and incubated for 2 h at 25°C in the solution containing 2% (w/v) cellulase Onozuka R-10 (KARLAN) and 1% (w/v) macerozyme R-10 (KARLAN) dissolved in 4 mL of enzyme incubation medium (0.8 M mannitol, 60 mM MES, and 5 mM MgCl 2 , pH 5.5). Digested tissues were filtered through one layer of Miracloth (Calbiochem). Protoplasts were centrifuged at 600 rpm in a swing bucket centrifuge (Marathon 21000R; Fisher Scientific) for 10 min at 4°C. Vacuole isolation was then performed as previously described (Di Sansebastiano et al., 1998).
Microscopy NR (Sigma) was dissolved in water and used at a final concentration 1 mg/ mL. Seedlings, protoplasts, and vacuoles were incubated with NR for 20 min at room temperature. For quantifying the number of AVIs, the same area of abaxial epidermal cells of cotyledons was always observed, or cells were counted in the entire abaxial surface. Samples were examined using a PCM-2000/Nikon Eclipse 600 laser-scanning microscope (Nikon) equipped with an argon and helium-neon laser (Ex 488, 544). To visualize GFP and anthocyanins, a 515/30-nm band-pass emission filter (EM515/30HQ) and 565-nm longpass filter (E565LP) were used, respectively. Light microscopy observations were made with a Nikon Eclipse 600 microscope equipped with Nomarski differential interference contrast optics. Images were captured and processed with a SPOT 2 slider CCD camera and the associated software (Diagnostic Instruments). All images were further processed using Adobe Photoshop software (Adobe Systems).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Fluorescent properties of anthocyanins.
Supplemental Figure S2. Effects of 3-MA and BFA treatment on secretory pathway markers.