Anthocyanin Vacuolar Inclusions Form by a Microautophagy Mechanism

Densely packed vacuolar anthocyanin bodies form by aggregation of anthocyanins in the cytoplasm and subsequent direct engulfment by the vacuolar membrane.

Abstract

Anthocyanins are flavonoid pigments synthesized in the cytoplasm and stored inside vacuoles. Many plant species accumulate densely packed, 3- to 10-μm diameter anthocyanin deposits called anthocyanin vacuolar inclusions (AVIs). Despite their conspicuousness and importance in organ coloration, the origin and nature of AVIs have remained controversial for decades. We analyzed AVI formation in cotyledons of different Arabidopsis thaliana genotypes grown under anthocyanin inductive conditions and in purple petals of lisianthus (Eustoma grandiorum). We found that cytoplasmic anthocyanin aggregates in close contact with the vacuolar surface are directly engulfed by the vacuolar membrane in a process reminiscent of microautophagy. The engulfed anthocyanin aggregates are surrounded by a single membrane derived from the tonoplast and eventually become free in the vacuolar lumen like an autophagic body. Neither endosomal/prevacuolar trafficking nor the autophagy ATG5 protein is involved in the formation of AVIs. In Arabidopsis, formation of AVIs is promoted by both an increase in cyanidin 3-O-glucoside derivatives and by depletion of the glutathione S-transferase TT19. We hypothesize that this novel microautophagy mechanism also mediates the transport of other flavonoid aggregates into the vacuole.

INTRODUCTION

Anthocyanins are flavonoid pigments produced by most seed plants as part of the phenylpropanoid pathway. They are glycosylated derivatives of anthocyanidins (cyanidin, pelargonidin, delphinidin, and others) that accumulate inside vacuoles and confer deep red and purple colors to flowers, fruits, and vegetative organs (Harborne and Williams, 2000; Grotewold, 2006). Anthocyanins are important for attracting pollinators and fruit dispersal agents; in vegetative tissues, they are believed to act as photoprotectors against high light and UV-B radiation and as free radical scavengers (Koes et al., 1994; Gitz et al., 1998; Gould, 2004).

The most common modifications of anthocyanins are glycosylation, methylation, and acylation (Sasaki et al., 2014). Arabidopsis thaliana produces more than 11 different anthocyanins derived from cyanindin (Bloor and Abrahams, 2002; Tohge et al., 2005; Saito et al., 2013; Kovinich et al., 2014). The conversion of cyanidin into the most highly decorated anthocyanin A11 requires seven modification steps mediated by four glycosyltransferases and three acyltransferases (Yonekura-Sakakibara et al., 2012; Saito et al., 2013). Anthocyanins are thought to be synthesized on the cytoplasmic face of the endoplasmic reticulum (ER) (Hrazdina et al., 1987; Saslowsky and Winkel-Shirley, 2001; Winkel, 2004) from where they are transported to the vacuolar lumen. Vacuolar localization prevents anthocyanin oxidation and the low pH environment confers the typical intense anthocyanin coloration (Marrs et al., 1995; Verweij et al., 2008; Faraco et al., 2014).

Although the enzymes involved in anthocyanin synthesis are reasonably well characterized, the mechanism for trafficking and sequestration of anthocyanins in plant cells remains controversial (Grotewold and Davies, 2008; Zhao and Dixon, 2010). Two main models have been postulated to explain how anthocyanins reach the vacuole. According to the ligandin model, cytoplasmic anthocyanins bind to specific glutathione S-transferases (GSTs), encoded in Arabidopsis by TRANSPARENT TESTA19 (TT19), in maize (Zea mays) by BRONZE2, and in petunia (Petunia hybrida) by AN9 (Marrs et al., 1995; Alfenito et al., 1998; Kitamura et al., 2004; Conn et al., 2008; Sun et al., 2012). These GSTs escort anthocyanins to the vacuolar membrane or tonoplast where some transporters of the ABC (ATP-binding cassette) and MATE (multidrug and toxin extrusion) families transfer anthocyanin molecules into the vacuolar lumen (Goodman et al., 2004; Marinova et al., 2007; Gomez et al., 2009; Francisco et al., 2013). The vesicular transport model postulates that anthocyanins enter the ER lumen and are transported in vesicles and/or membrane-bound organelles to the vacuole. This hypothesis is based on the observation of flavonoid-filled ER-derived vesicles in Brassica napus tapetum cells (Hsieh and Huang, 2007), cytoplasmic anthocyanin-filled vesicles in grapevine (Vitis vinifera) (Conn et al., 2010; Gomez et al., 2011), and the accumulation of anthocyanins in ER bodies in Arabidopsis epidermal cells (Poustka et al., 2007). This model also contemplates the scenario that anthocyanin-containing ER domains could be engulfed by autophagosomes and delivered to the vacuole (Pourcel et al., 2010). Autophagy, either macro- or microautophagy, is the transport of cytoplasmic material to the vacuole or lysosome, and in most studied cases it involves the autophagy-related (ATG) machinery (Müller et al., 2000; Uttenweiler et al., 2007; Krick et al., 2008; Li and Vierstra, 2012). In macroautophagy, double-membrane autophagosomes assemble in the cytoplasm and upon fusion with the vacuolar membrane deliver their contents to the vacuolar lumen in an autophagic body. During microautophagy in animal and yeast cells, the lysosomal or vacuolar membrane deforms locally to engulf cytoplasmic contents directly; however, we know little about how microautophagy proceeds in plants. If anthocyanins are transported by macroautophagy, upon fusion of the autophagosome with the tonoplast, anthocyanins would be surrounded by two membranes, an internal one derived from the ER and an outer one derived from the autophagosome. As in other cases of autophagy, these membranes would be later degraded, releasing anthocyanins into the vacuolar lumen.

Inside the vacuolar lumen, anthocyanins are found either in a uniformly distributed, soluble form or in intravacuolar bodies called anthocyanoplasts (Pecket and Small, 1980) or anthocyanin vacuolar inclusions (AVIs) (Markham et al., 2000). AVIs are frequent in a large number of unrelated flowering plant taxa (Pecket and Small, 1980). As AVIs are highly condensed and deeply colored anthocyanin bodies, they change color properties of flowers such as intensity and hue (Markham et al., 2000). In addition, since anthocyanins are important nutraceuticals with alleged antioxidant properties (Butelli et al., 2008; Horbowicz et al., 2008), AVIs have potential commercial value as densely packed bodies of stabilized anthocyanins to be used as food additives.

Although AVIs have been observed in plant tissues since the early 1900s (reviewed in Pecket and Small, 1980), their nature, function, and structure remain unclear. AVIs were described as being surrounded by a single membrane in grapevine and in red cabbage (Brassica oleracea). In sweet potato (Ipomea batatas), carnation (Dianthus caryophyllus), and lisianthus (Eustoma grandiorum), AVIs appear to lack surrounding membranes and instead consist of a protein matrix or thread-like structures (Small and Pecket, 1982; Nozue et al., 1993; Markham et al., 2000; Zhang et al., 2006; Conn et al., 2010). Besides anthocyanins, AVIs have been reported to contain a metalloprotease called VP24 in sweet potato (Nozue et al., 1997, 2003; Xu et al., 2001) and tonoplast membrane lipids in grapevine cultured cells (Conn et al., 2010). In grapevine cell cultures, AVIs are enriched in acylated anthocyanins (Conn et al., 2003), suggesting that AVIs may sequester specific anthocyanin species. The biogenesis of AVIs and whether they share the same trafficking pathway with the soluble anthocyanins is also controversial. In grapevine and lisianthus, cytoplasmic vesicles or prevacuolar compartments have been reported to enter the main vacuole and undergo intravacuolar fusion to generate AVIs (Conn et al., 2010).

In Arabidopsis, anthocyanins are commonly synthesized in vegetative tissues, mostly in the epidermis, as a response to stressful conditions (Winkel-Shirley, 2002; Kovinich et al., 2014). When grown under anthocyanin inductive conditions (AICs; e.g., 3% sucrose in water), Arabidopsis seedlings accumulate anthocyanins mostly as a soluble pool within the vacuole, and <5% of the cotyledon pavement cells contain AVIs (Pourcel et al., 2010). AVIs also form with low frequency in Arabidopsis tt4 seedlings lacking the chalcone synthase required for anthocyanin biosynthesis, when grown under AIC and supplemented with naringenin, an intermediate in the anthocyanin pathway (Poustka et al., 2007). Conversely, the 5gt mutant, which is unable to glucosylate anthocyanidins at the 5-O position and produces cyanidin-3-O-glucosyde (C3G) derivatives, forms AVIs in almost every cotyledon pavement cell when grown under AIC (Pourcel et al., 2010). How these AVIs form and what their relationship is to other organelles is not known.

Using a combination of anthocyanin intrinsic fluorescence (Pourcel et al., 2010) and fluorescence lifetime imaging (FLIM), electron microscopy, and biochemical approaches, we identified a novel microautophagy mechanism by which AVIs form in somatic tissues of Arabidopsis and purple petals of lisianthus. We also tested cellular and biochemical conditions that promote formation of AVIs in Arabidopsis. We postulate that a similar microautophagy process is likely to participate in the vacuolar uptake of other specialized metabolites.

RESULTS

AVIs Are Surrounded by a Single Membrane

Current models on anthocyanin trafficking predict structurally different AVIs. AVIs should be surrounded by two membranes (one from the ER and one from the autophagosome) if they form by macroautophagy of anthocyanin accretions inside the ER or by no membranes if they are transported in vesicles or formed directly inside the vacuolar lumen by aggregation. To determine the presence of membranes on AVIs, we used wild-type, tt4, and 5gt Arabidopsis seedlings (Supplemental Figure 1) grown under modified AIC (mAIC; see Methods) and supplemented with the membrane dye FM1-43 (Figure 1). We chose these genotypes because the tt4 mutation gives us the opportunity to synchronize anthocyanin synthesis upon incubation with naringenin and the 5gt mutation dramatically increases the density of AVIs. We detected FM1-43 staining around AVIs in the three Arabidopsis genotypes, indicating that AVIs in Arabidopsis are enclosed by membranes (indicated by arrowheads in Figure 1A; Supplemental Figure 2). To determine whether this is also the case in other species, we analyzed purple lisianthus petals, which typically produce large quantities of AVIs (Markham et al., 2000). We incubated lisianthus petals with FM1-43 for 48 h and detected FM1-43 signal around large and rounded AVIs in epidermal cell (Figure 1B), confirming the presence of AVI membranes also in lisianthus.

Figure 1.

AVIs in Arabidopsis and Lisianthus.

(A) and (B) Confocal images of 5-d-old Arabidopsis tt4 cotyledons grown under mAIC and supplemented with naringenin (A) and purple lisianthus petals (B) stained with the membrane dye FM1-43 (green). Arrowheads indicate the FM1-43 signal around AVIs. Anthocyanin autofluorescence is depicted in magenta.

(C) to (E) Electron micrographs of high-pressure frozen/freeze-substituted Arabidopsis 5gt cotyledons showing the presence of a single membrane (arrowheads) around AVIs.

V, vacuole. Bars = 5 µm in (A) to (C) and 100 nm in (D) and (E).

To determine the number of membranes around AVIs, we analyzed high-pressure frozen/freeze-substituted 5gt mutant seedlings grown under mAIC by transmission electron microscopy (TEM). We found that AVIs free in the vacuolar lumen were surrounded by a single membrane tightly pressed against the electron-dense anthocyanin core (Figures 1C to 1E). We measured this membrane in 30 regions of 10 AVIs and found it to be ∼12 nm thick, consistent with the expected thickness of a bilayer unit stained with heavy metals (De, 2000).

Taken together, these results show that AVIs in different Arabidopsis genotypes and lisianthus petals are enclosed by a membrane, suggesting structural similarities among AVIs in different species.

AVI Formation Is Independent of Anthocyanin Accumulation inside the ER and Endosomal/Prevacuolar Trafficking

Previous studies have suggested that the soluble pool of anthocyanins accumulate inside the ER before being transported to the vacuole in ER-derived compartments (Poustka et al., 2007). To test whether AVIs derive from the ER, we analyzed wild-type Arabidopsis seedlings (Col-0) expressing a GFP-HDEL (ER lumen marker) and 5gt and tt4 seedling expressing CALNEXIN-GFP (ER membrane marker) grown under mAIC. We observed AVIs in cotyledon pavement cells but did not detect anthocyanin deposits associated with the ER (Supplemental Figure 3). We further confirmed the lack of association between anthocyanins and ER during AVI formation by calculating the Pearson’s correlation coefficient (PCC) between the ER markers and anthocyanins in AVI-containing cells. In both cases, the PCC values were less than −0.2 (PPC value for GFP-HDEL and anthocyanins in wild-type cells was −0.27 ± 0.06, n = 6 cells; and for CALNEXIN-GFP in 5gt, −0.25 ± 0.05, n = 6 cells), suggesting that anthocyanins were not transported inside the ER during formation of AVIs.

To determine if AVI formation depends on vacuolar trafficking through endosomes or prevacuolar compartments, we tested a collection of 16 mutants known to affect different aspects of endosome-vacuole trafficking and vacuolar dynamics (Supplemental Figure 4) (Uemura and Ueda, 2014). We induced AVI formation in mutant and wild-type seedlings by growing them under mAIC for 5 d and applying sodium orthovanadate (vanadate) for 24 h, a treatment known to trigger AVI accumulation in wild-type seedlings (Poustka et al., 2007). Although this treatment induces AVI formation in low frequencies and, therefore, quantitative comparisons should be interpreted with caution, we found that all mutants were able to form AVIs in comparable densities to those in wild-type seedlings (Supplemental Figure 4), suggesting that fully functional endosomal/prevacuolar compartments are not essential for AVI formation.

Cytoplasmic Anthocyanin Aggregates Are Engulfed Directly by the Tonoplast

Since both wild-type and 5gt Arabidopsis plants form membrane-bound AVIs but 5gt produces them in much higher density, we analyzed AVI formation in the 5gt mutant. We imaged 5gt seedlings expressing the tonoplast marker γTIP-CFP (Nelson et al., 2007) by confocal microscopy and both single and serial sections of 5gt seedlings grown under mAIC by TEM. Besides the free AVIs in the vacuolar lumen, we detected anthocyanin aggregates outside the vacuole (Figure 2). In most cases, cytoplasmic anthocyanin aggregates were closely associated with the cytoplasmic face of the tonoplast (Figures 2A to 2F). In a first stage, the cytoplasmic anthocyanin aggregates became tightly pressed against the tonoplast, which seemed to wrap around the anthocyanin cores in a double-membrane cup-shaped structure (Figures 2A to 2F). Eventually, the distal open part of the tonoplast anthocyanin-enclosing structure fused and closed and the two previously connected tonoplast membranes lost continuity. Thus, in forming AVIs, each anthocyanin aggregate was surrounded by layers of tonoplast membrane, one completely pressed against the aggregate and the other one, continuous with the main tonoplast, at a regular distance of ∼30 μm (Figures 2C and 2F). Occasionally, the AVI membrane exposed to the vacuolar lumen was partially degraded (Figures 2E and 2F, asterisks).

Figure 2.

Formation of AVIs in 5-d-Old Arabidopsis 5gt Seedlings.

(A), (D), and (G) Confocal images of 5-d-old 5gt seedlings expressing γTIP-CFP. Arrowheads indicate position of the tonoplast domains associated with anthocyanin aggregates. The areas in (G) where the two tonoplast membranes surrounding developing AVIs have become separated from each other are indicated by asterisks.

(J) Confocal image of 5-d-old 5gt seedling stained with FM1-43.

(B), (C), (E), (F), (H), (I), (K), and (L) Transmission electron micrographs of 4-d-old 5gt seedlings showing different stages of anthocyanin aggregate engulfment. White arrowheads indicate tonoplast layer in contact with anthocyanin aggregates, and black arrowheads indicate the outermost tonoplast layer of the microautophagic cup-shaped structure. Note the partially degraded membranes around AVIs in (E) and (F) (asterisks). The areas in (H) and (I) where the two tonoplast membranes surrounding developing AVIs have become separated from each other is indicated by asterisks.

(M) to (P) Immunogold labeling of γ-TIP on AVIs at different stages of formation. White arrowheads indicate labeling of the tonoplast layer in contact with anthocyanin aggregates, and black arrowheads indicate labeling on the outermost tonoplast layer of the microautophagic cup-shaped structure.

Bars = 5 µm in (A), (D), (G), and (J), 1 µm in (B), (E), (H), and (K), 500 nm in (M) to (P), and 100 nm in (C), (F), (I), and (L). V, vacuole; AA, anthocyanin aggregate.

To confirm the number of tonoplast layers associated with forming AVIs, we measured γTIP-CFP fluorescence intensity in different regions of the tonoplast of 5gt cotyledon cells. We found that γTIP-CFP fluorescence intensity around anthocyanin aggregates/forming AVIs was approximately 2-fold higher (average = 2.01; n = 7 cells) than that of other tonoplast areas, confirming the presence of a double-membrane tonoplast domain around anthocyanin aggregates during AVI formation (Figure 3).

Figure 3.

Measurement of γTIP-CFP Fluorescence Intensity in 5gt Seedlings.

CFP intensity from γTIP-CFP was measured in different areas of the tonoplast. Only images with no saturated pixels were used for fluorescence intensity quantification. The intensity of the CFP signal from the tonoplast associated with anthocyanin aggregates is approximately twice as strong as that from tonoplast domains not associated with AVIs. Bar = 5 µm.

In a later stage, the outer tonoplast membrane around the anthocyanin core began to separate from the inner AVI membrane in some areas, generating a lobed profile (Figures 2G to 2I and 4). Finally, when the two membranes separated completely, the resulting free AVI surrounded by a single membrane derived from the tonoplast was released into the lumen (Figures 2J to 2L). We further confirmed the identity of the membranes engulfing anthocyanin aggregates and the AVI membrane as tonoplast by immunogold detection of the endogenous tonoplast protein γTIP in Arabidopsis 5gt seedlings (Figures 2M to 2P).

Figure 4.

Three-Dimensional Serial Section Reconstructions of Forming AVIs.

(A) to (C) Single sections from a 16 serial-section reconstruction spanning a 1.3-μm-thick volume containing two adjacent forming AVIs.

(D) Three-dimensional reconstruction based on 16 serial sections through two AVIs. The AVI membrane is depicted in orange and the tonoplast in light yellow.

The partial detachment of the tonoplast from the AVI membrane generates lobes indicated by asterisks. AA, anthocyanin aggregates. Bar = 1 μm.

To further understand the organization of the tonoplast domains associated with forming AVIs, we analyzed three-dimensional reconstructions from TEM serial sections. We found that the tonoplast membrane not only surrounded the surface of the anthocyanin aggregates but also folded into the anthocyanin core, forming large and complex membrane convolutions inside the forming AVI (Figures 5). These intra-AVI tonoplast domains were tightly pressed against the anthocyanins and continuous with the tonoplast membrane enclosing the anthocyanin core (Figures 5G to 5I). In most AVIs, the anthocyanin core had relatively large internal voids completely lined with tonoplast and occupied by vacuolar lumen (Figures 5A and 5B, asterisks)

Figure 5.

Three-Dimensional Serial Section Reconstructions of Forming AVIs with Internal Membrane Networks.

(A) and (B) Single sections from a 16 serial-section reconstruction spanning a 1.3-μm-thick volume containing a forming AVI. Intra-AVI tonoplast membranes (orange arrowheads) lined voids filled with vacuolar lumen (asterisks) inside of the anthocyanin core.

(C) Corresponding three-dimensional reconstruction based on 16 serial sections such as the ones depicted in (A) and (B). The tonoplast is depicted in light yellow and the AVI membrane is in orange.

(D) and (E) Single sections from a 16 serial-section reconstruction spanning a 1.3-μm-thick volume containing a forming AVI. The intra AVI membranes are indicated by arrowheads.

(F) Corresponding three-dimensional reconstruction based on 16 serial sections such as the ones depicted in (D) and (E). The tonoplast is depicted in light yellow and the AVI membrane in orange.

(G) AVI with a complex internal membrane network. The two boxed regions are shown at higher magnification in (H) and (I) and indicate the sites where the inner membranes are connected to the AVI-surrounding membrane. AA, anthocyanin aggregates; V, vacuole.

Bars = 1 μm in (A) to (G) and 500 nm in (H) and (I).

Our observations show that anthocyanins aggregate in the cytoplasm in close association with the tonoplast, which then directly engulfs the aggregate in a process that resembles microautophagy. As other cases of microautophagy in yeast and animal cells rely on the ATG machinery (Müller et al., 2000; Uttenweiler et al., 2007; Krick et al., 2008), we tested whether the core autophagy protein ATG5 was necessary for AVI formation. For this, we made a double atg5 5gt mutant and quantified AVIs in 5-d-old cotyledons grown under mAIC (Figure 6). We found no statistically significant differences between the density of AVIs in 5gt and atg5 5gt (Figure 6A).

Figure 6.

Effects of the atg5 Mutation and Wortmannin on AVI formation.

(A) AVIs in cotyledon pavement cells from atg5 5gt seedlings grown in mAIC. Graph shows AVI density measured in 6 to 20 cotyledons of each genotype grown side-by-side in the same conditions.

(B) AVI formation kinetics in tt4 5gt cotyledons from seedlings grown in mAIC supplemented with naringenin.

(C) tt4 5gt seedlings grown under mAIC and supplemented with both 100 µM naringenin and 30 μM wortmannin (WM) for 14 and 15 h, respectively. The wortmannin treatment did not affect AVI formation in cotyledon pavement cells. AVI density was measured in 38 cotyledons and compared with control seedlings grown side-by-side in the same conditions but without wortmannin. To check for the effective penetration and action of wortmannin, we grew together with the tt4 5gt seedlings, wild-type seedlings expressing the endosome-localized protein RABF2A-YFP. Upon treatment with 30 μm wortmannin, the endosomes drastically enlarged as describe before (Haas et al., 2007), indicating that wortmannin is able to penetrate into cotyledon cells grown under mAIC. Arrowheads indicate RABF2A- YFP-positive endosomes.

Error bars indicate sd. Bars = 10 µm.

Vacuolar morphology and tonoplast remodeling are sensitive to wortmannin, an inhibitor of phosphatidylinositol 3-kinase (Zheng et al., 2014). We next tested whether wortmannin blocks AVI formation in a tt4 5gt double mutant, for example, by not allowing the tonoplast to deform around anthocyanin aggregates. The tt4 5gt seedlings grown under mAIC produced detectable AVIs 2 h after incubation in naringenin; after 6 to 7 h, AVI density remained constant (Figure 6B). We treated tt4 5gt seedlings grown under mAIC with both naringenin and 30 μm wortmannin for 14 and 15 h, respectively. We did not detect changes in the number of AVIs or anthocyanin aggregates per area in the wortmannin-treated seedlings (Figure 6C).

Taken together, our results show that AVIs form by a microautophagy mechanism that does not depend on either ATG5 or phosphatidylinositol 3-kinase activity.

AVIs and Diffuse Vacuolar Anthocyanins Exhibit Different Fluorescence Lifetimes

To analyze possible differences between the two pools of vacuolar anthocyanins (diffuse and AVIs) we analyzed their fluorescence properties. As previously reported for extracted soluble anthocyanins in in vitro conditions (Poustka et al., 2007), we found that both cellular anthocyanin pools in wild-type cotyledon pavement cells emitted in the 600- to 660-nm range when excited with green light, indicating that subcellular compartmentalization does not affect their spectral properties (Supplemental Figure 1C). We then analyzed their fluorescence lifetime (Figure 7), which is an intrinsic property of a fluorophore independent of its concentration, sample thickness, or fluorescence intensity. However, fluorescence lifetime is sensitive to the molecular microenvironment, protein and lipid associations, molecular modifications and conformational changes, and changes in pH (Chang et al., 2007). FLIM generates images based on the differences in the excited state decay rate of a fluorophore, reflecting differences in the fluorophore microenvironment or molecular binding state.

Figure 7.

Fluorescence Lifetime Analysis of Vacuolar Anthocyanins in Arabidopsis.

(A) Fluorescence lifetime of isolated wild-type anthocyanins. τm, weighted mean lifetime. τ1 and τ2 are the lifetimes in picoseconds of the two identified fluorescence lifetime components.

(B) Relative contributions of the two identified components.

(C) FLIM analysis of vacuolar anthocyanin pools, soluble versus AVIs (white arrowheads), in cotyledon pavement cells of different Arabidopsis genotypes. Images are pseudo-colored according to the anthocyanin fluorescence lifetime values in picoseconds. Bars = 20 µm.

(D) Quantitative analysis of fluorescence lifetime. We analyzed 13 ROIs from six wild-type cotyledons, 51 ROIs from 8 5gt cotyledons, and 62 ROIs from 12 tt4 cotyledons.

Bar graph represents average, and error bars represent se. The asterisks indicate a significant difference between the two values compared by two-tailed Student's t test (P < 0.05).

We found that the fluorescence lifetime of anthocyanins measured both in vitro and in vivo conditions was best modeled as a double exponential function, with a majority of the photons exhibiting a shorter fluorescence lifetime distinct from the minor component of longer fluorescence lifetimes (Figures 7A and 7B). The two fluorescence lifetime components were detected in both in vivo and in vitro conditions, suggesting that they are an inherent characteristic of the anthocyanins. Comparing the weighted mean (τm) of the two fluorescence lifetime components for anthocyanins in tt4, 5gt, and wild-type Arabidopsis seedlings, we determined that the two dominant anthocyanin pools, soluble and AVIs, exhibited statistically significant differences in their fluorescence lifetimes (Figures 7C and 7D). In the three genotypes, the mean fluorescence lifetime of the AVIs was shorter than that of the soluble pool. These results are consistent with the two pools having different types of anthocyanins and/or differences in microenvironmental conditions, packing, and interactions with other cellular factors.

AVIs Are Enriched in Decorated Cyanidin-3-O-Glucoside Derivatives

Our results indicate that the AVIs in the wild type and 5gt mutants are similar from the perspective of their biogenesis. To investigate possible mechanisms mediating the aggregation of anthocyanins in the cytoplasm, we first analyzed whether anthocyanins within AVIs were chemically different from those of the soluble pool uniformly distributed in the vacuolar lumen. The 5gt mutant produces AVIs in much higher frequency than the wild type and accumulates C3G derivatives normally present at very low levels in wild-type plants (Pourcel et al., 2010; Supplemental Figure 1B). We sequentially extracted anthocyanins from 5gt seedlings removing first the diffuse, soluble vacuolar fraction by heating the samples in water at 70°C for 3 s followed by a methanolic extraction of the remaining anthocyanin pool represented mostly by AVIs (Figures 8A and 8B). By analyzing these fractions by HPLC followed by identification of the major peaks by mass spectrometry, we determined that the soluble pool in 5gt is enriched in cyanidin-3-sambubiose (3-O-(2-O-(β-d-xylopyranosyl)-β-d-glucopyranosyl) cyanidin), whereas that in AVIs are enriched in coumarylated cyanidin-3-sambubiose derivatives (Figure 8B). We measured the fluorescence lifetime of HPLC-fractionated anthocyanin species from 5gt seedlings (Figures 8C and 8D) and found that cyanidin-3-sambubiose purified from 5gt showed a longer fluorescence lifetime than the fraction containing the coumarylated derivatives, indicating that the shorter fluorescence lifetime of 5gt AVIs compared with that of the soluble pool (Figure 7) could be partially due to their enrichment in decorated (coumarylated) cyanidin-3-sambubiose species.

Figure 8.

Analysis of Anthocyanins from the 5gt Mutant.

(A) Five-day-old 5gt seedlings grown under AIC during the sequential extraction of anthocyanins.

(B) HPLC-DAD analysis of sequentially extracted anthocyanins from 5-d-old 5gt seedlings. Profiles of three different pools were analyzed: total anthocyanins, pool extracted with 70°C water (enriched in anthocyanins uniformly distributed in the vacuolar lumen), and pool extracted with methanol-formic acid (enriched in AVIs). The hypothetical structures of anthocyanins in peaks A, B, and C based on their fragmentation patterns after liquid chromatography/mass spectrometry analysis are shown on the right.

(C) Chromatograms of fractions containing either cyanindin-3-sambubioside (C3S) or C3S coumarylated derivatives from 5gt seedlings. Fractions were collected using an HPLC-DAD with a fraction collector.

(D) Quantitative FLIM analysis of the 5gt anthocyanin fractions shown in (C). Bar graph represents fluorescence lifetime mean values (τm), and error bars represent se. The asterisks indicate a significant difference between the two values compared by two-tailed Student's t test (P < 0.05).

Flavonoids can bind to membranes (Oteiza et al., 2005), and the in vitro interaction of anthocyanins with neutral lipids results in the formation of AVI-like structures (Zhang et al., 2010, 2012) To investigate the possibility that neutral lipids participate in the aggregation of anthocyanins into rounded bodies in the cytoplasm, we tested for the presence of neutral lipids inside 5gt AVI cores using specific fluorescent dyes. Nile Blue was used to detect mostly neutral lipids (triglycerides and steroids) and Bodipy for oils and nonpolar lipids (Cooper et al., 2010; James et al., 2010). Neither dye stained the AVI anthocyanin cores in 5gt, although both dyes stained lipid bodies in the same cells (Supplemental Figure 5), suggesting that anthocyanins inside AVIs are not associated with neutral lipids.

TT19 Negatively Regulates AVI Formation in Wild-Type but Not in 5gt Mutant Seedlings

To further investigate the cause for anthocyanin aggregation and AVI formation, we explored the possibility that a binding partner could affect anthocyanin solubility. GSTs have long been implicated in anthocyanin binding and trafficking (Marrs et al., 1995; Alfenito et al., 1998; Xiang et al., 2001; Kitamura et al., 2004; Conn et al., 2008; Sun et al., 2012). Arabidopsis TT19 can bind cyanidin and C3G in vitro, increasing their solubility in water (Sun et al., 2012), and the tt19 mutants fail to accumulate anthocyanins in the vacuole (Kitamura et al., 2004). Thus, we investigated a possible role of TT19 in preventing anthocyanin aggregation in the cytoplasm.

We first analyzed anthocyanin distribution in the tt19 mutant (Figure 9). Although tt19 seedlings have greatly reduced anthocyanin contents, when grown under mAIC and supplemented with naringenin, they produced low but detectable amounts of pigments with an anthocyanin profile similar to that of the wild type (Figures 9A and 9B). When we analyzed tt19 cotyledons from seedlings grown under these conditions, we observed cytoplasmic anthocyanin aggregates and AVIs surrounded by a membrane in a large proportion of pavement cells but very little soluble vacuolar anthocyanins (Figure 9A). These results suggest that TT19 is required for maintaining wild-type Arabidopsis anthocyanins soluble in the cytoplasm, and its absence leads to the formation of anthocyanin aggregates and AVIs.

Figure 9.

TT19 in AVI Formation.

(A) tt19 mutant seedlings grown under mAIC and supplemented with naringenin and the membrane dye FM1-43. White arrowheads indicate FM1-43 signal around AVI. Anthocyanin autofluorescence is shown in magenta.

(B) Chromatographic measurements of anthocyanins extracted from Arabidopsis wild-type Col-0 (WT) and tt19 seedlings grown in mAIC.

(C) Micrographs showing merged transmitted and green fluorescence channels of 5gt and 5gt Pro35S:TT19:GFP-expressing cotyledons.

(D) Quantification of AVI density in 5-d-old 5gt and two independent 5gt Arabidopsis transgenic lines overexpressing TT19-GFP. Bar graph represents average number of AVIs per cell, and error bars represent se.

Bars = 5 μm in (A) and 20 μm in (C).

To test whether TT19 overexpression can reduce anthocyanin aggregation and AVI frequency in 5gt, we overexpressed a functional GFP-tagged TT19 (Supplemental Figure 6) with the constitutive CaMV35S promoter in 5gt seedlings. The overexpression of TT19-GFP in 5gt did not reduce the density of anthocyanin aggregates and AVIs (Figures 9C and 9D), suggesting that excess TT19 does not mitigate aggregation of the anthocyanins in the 5gt mutant.

DISCUSSION

We have found that anthocyanins can form cytoplasmic aggregates that are incorporated into the vacuole by microautophagy (Figure 10); that is, cytoplasmic material is directly engulfed by the tonoplast or lysosomal membrane. This microautophagy process leads to the release into the vacuolar lumen of anthocyanin inclusions surrounded by a single membrane derived from the tonoplast. Based on the evidence discussed here and even though little is known about plant microautophagy, it is reasonable to propose that AVI formation involves an unconventional microautophagy process. First, it does not appear to depend on the ATG core machinery, as evidenced by our results from atg5 5gt mutants (Figure 6A). Second, it is independent of phosphatidylinositol 3-kinase activity (Figure 6C), which is required for other local tonoplast deformation events reported previously (Saito et al., 2011a). Third, the binding of the tonoplast to the surface of the anthocyanin aggregate itself seems to drive the deformation of the tonoplast to the point of creating an intricate network of tonoplast membrane lining the internal voids of the anthocyanin cores (Figures 5). Our results also showed that in different Arabidopsis genotypes, both anthocyanin composition and the availability of a flavonoid-related GST affect the frequency of AVIs.

Figure 10.

Model of AVI Formation by Microautophagy.

Stage I: Cytoplasmic anthocyanin aggregates associate with the cytoplasmic face of the tonoplast and become surrounded by double membrane tonoplast protrusions. The tonoplast membrane not only binds closely to the surface of the anthocyanin aggregates but also lines its internal voids. The distal domains of the tonoplast protrusions eventually fuse. Stage II: The two membranes surrounding the aggregate start to separate and bulges filled with vacuolar lumen form around the forming AVI. Stage III: The two membranes separate completely and the newly formed AVI is released into the vacuolar lumen, tightly surrounded by a membrane derived from the tonoplast. AA, anthocyanin aggregate.

AVIs Arise from Microautophagy of Cytoplasmic Anthocyanin Aggregates

Although the final outcome of both macro- and microautophagy is the same, i.e., the release of an autophagic body with cytoplasmic content into the vacuole, the two events involve rather different membrane dynamics. Whereas the two membranes that form an autophagosome in macroautophagy are assembled in the cytoplasm around the cargo, microautophagy requires the local deformation, either invagination or evagination, of the vacuolar membrane. Thus, the resulting autophagic body inside the vacuole is surrounded by the inner membrane of the autophagosome in macroautophagy and by a portion of the tonoplast in microautophagy.

Compared with macroautophagy, microautophagy is poorly characterized in terms of molecular mechanisms, regulation, and cargo selection. Different microautophagy modalities have been characterized primarily in yeast and mammals. Similar to macroautophagy, microautophagy can be selective or nonselective, depending on cargo specificity (Mijaljica et al., 2011). In Saccharomyces cerevisiae, nonselective microautophagy mediates the uptake of cytosolic contents into long tubular invaginations of the vacuolar membrane. These tubes form as a consequence of the lateral segregation and redistribution of lipids and transmembrane proteins, allowing for sharp kinks in the vacuolar membrane (Müller et al., 2000). Larger cargo like mitochondria, peroxisomes, endosomes, or nuclear fragments are directly engulfed by vacuoles and lysosomes through selective microautophagy (Sakai et al., 1998; Nowikovsky et al., 2007; Kawamura et al., 2012). For example, in piecemeal microautophagy of the nucleus in carbon and nitrogen-starved S. cerevisiae cells, the vacuolar membrane connects to nuclear protrusions through nuclear-vacuolar junctions. Once these junctions are formed, the vacuolar membrane invaginates, the nuclear envelope pinches off, and the vacuolar membrane fuses, leading to the release of a microautophagic body surrounded by the vacuolar membrane that is later degraded by the vacuolar hydrolases (Roberts et al., 2003). In mammalian cells, lysosomes protrude arm-like projections that wrap and engulf cytoplasmic content (Sakai et al., 1989) in a process morphologically similar to the formation of AVIs. The few studied examples of both nonselective and selective microautophagy depend on specific proteins as well as the core ATG machinery that also regulate macroautophagy (Müller et al., 2000; Uttenweiler et al., 2007; Krick et al., 2008). However, based on our analysis of the atg5 5gt double mutant, we concluded that ATG5, which is a key component of the ATG core machinery, is not necessary for AVI formation.

Microautophagy has not been characterized in plants at a cellular or molecular level. However, highly dynamic tonoplast deformations have been reported in many different developmental contexts and environmental conditions (Saito et al., 2002, 2011a, 2011b; Reisen et al., 2005; Oda et al., 2009; Zheng and Staehelin, 2011; Segami et al., 2014; Zheng et al., 2014). Vacuolar “bulbs” are defined as complex local invaginations of the tonoplast that could serve as membrane reservoirs for rapid vacuolar volume changes. However, a recent study showed that although plants can naturally form double-membrane tonoplast invaginations called intravacuolar spherical structures, the more complex bulbs are likely artifacts induced by the dimerization of the fluorescent tags on tonoplast proteins (Segami et al., 2014). AVIs and the native intravacuolar spherical structures described by Segami et al. (2014) and Saito et al. (2002) have in common their double-membrane structure and the separation of the two membranes at a regular distance of 30 to 45 nm. However, different from intravacuolar spherical structures and bulbs, AVIs form from evaginations, not invaginations, of the tonoplast, indicating that the underlying membrane deformation mechanisms for these processes are different. In addition, whereas both mutations in the SNARE VTI11 and wortmannin interfere with the formation of tonoplast bulbs in Arabidopsis (Saito et al., 2011a), they do not affect the formation of AVIs in Arabidopsis seedlings (Supplemental Figure 4; Figure 6C).

The close association between the vacuolar membrane and the cargo seems to be a common theme in many cases of microautophagy. Consistently, we observed a Velcro-like association between the surface of the anthocyanin aggregates and the cytoplasmic face of the tonoplast. Indeed, this association is so tight that in some AVIs the tonoplast forms a convoluted network closely adhered to crevices inside the anthocyanin core (Figure 5). Currently, it is unclear what kind of molecules mediates this interaction. Flavonoids have been reported to bind membranes, so they could potentially associate with the tonoplast directly (Sengupta et al., 2004; Oteiza et al., 2005; Verstraeten et al., 2013; Pawlikowska-Pawlega et al., 2014). Some of the enzymes in the anthocyanin pathways have been shown to partially localize to the tonoplast (Saslowsky and Winkel-Shirley, 2001; Toda et al., 2012). It is possible that, in species that produces AVIs, anthocyanins aggregate and bind the surface of the vacuolar membrane as they are being synthesized, inducing local deformations of the tonoplast.

The mechanism of AVI formation that we present here reconciles many apparently contradicting observations found in the literature. Dense deposits of anthocyanins have been observed inside prevacuolar compartments, small vacuoles, and vesicles in grapevine (Gomez et al., 2011), Arabidopsis (Poustka et al., 2007), and lisianthus (Zhang et al., 2006). Many of these structures could correspond to early intermediates in AVI formation when anthocyanins are enclosed into tonoplast cup-shaped domains. In addition, AVIs have been reported to have or lack a membrane (Small and Pecket, 1982; Nozue et al., 1993; Markham et al., 2000; Zhang et al., 2006; Conn et al., 2010). We have observed that the tonoplast-derived membrane that surrounds free AVIs in the vacuolar lumen can partially degrade; therefore, depending on the stage, AVIs can partially lose their membranes. Consistent with our results, the identification of tonoplast phospholipids in grapevine AVIs confirms that the AVI membrane derives from the tonoplast (Conn et al., 2010). In addition, our study is the first to employ high-pressure freezing/freeze substitution to analyze AVI structure by electron microscopy. High-pressure freezing followed by freeze substitution is much a better technique to preserve cellular membranes that could have been lost in other AVI studies based chemical fixation.

What Factors Trigger Anthocyanin Aggregation in the Cytoplasm?

Some flavonoids self-aggregate via interaction of their glycosyl moieties (Hoshino, 1991) and 3-deoxyanthocyanidins in Sorghum bicolor self-organize in rounded aggregates that bind and destabilize the plasma membrane upon fungal attack (Nielsen et al., 2004). In general, decorations on the anthocyanin backbones can increase anthocyanin solubility in water (Aksamit-Stachurska et al., 2008). However, the molecular mechanism that leads to anthocyanin aggregation within AVIs cannot be fully explained by either the solubility or the type of decorations on the different anthocyanin species. In grapevine cultured cells, AVIs are enriched in acylated anthocyanins with no preference for a specific anthocyanidin backbone (Conn et al., 2003), whereas in lisianthus, AVIs preferentially accumulate diglucosidic anthocyanins with a preference for cyanidin over delphinidin (Markham et al., 2000). In Arabidopsis, an increase in the accumulation of C3G and C3G derivatives correlates with AVI occurrence in both wild-type and 5gt mutant seedlings supplemented with naringenin (Pourcel et al., 2010). Similar to grapevine but different from lisianthus, the 5gt mutant forms AVIs that are enriched in acylated anthocyanins, suggesting a rather complex and likely species-specific connection between anthocyanin structure and sequestration into AVIs. The size and number of AVIs in grapevine cultured cells positively correlate with anthocyanin content (Conn et al., 2010), suggesting that anthocyanin concentration may also influence AVI formation.

The role of TT19 and functionally related GSTs (e.g., petunia AN9 and maize BZ2) in anthocyanin accumulation has remained controversial. Whereas initially proposed to catalyze the conjugation of glutathione to C3G to facilitate transport by tonoplast pumps (Marrs et al., 1995; Alfenito et al., 1998), subsequent studies demonstrated that they may function as flavonoid binding proteins (Mueller et al., 2000; Sun et al., 2012), escorting and/or stabilizing anthocyanins until their vacuolar uptake. Arabidopsis tt19 seedlings synthesize low levels of anthocyanins (Kitamura et al., 2004; Sun et al., 2012). Our results demonstrate that the pattern of anthocyanins in tt19 mutants is equivalent to that of the wild type (Figure 9). However, tt19 seedlings produce AVIs in high density (Figure 9) despite the low levels of anthocyanins, indicating that the lack of TT19 triggers the aggregation of anthocyanins that otherwise are soluble in wild-type plants.

What aspect of TT19 is important for AVI formation? GST activity inhibitors such as buthionine sulfoximine and 1-chloro-2-4-dinitrobenzene induce high AVI accumulation without a significant effect on anthocyanin levels (Poustka et al., 2007), suggesting that either GST activity is necessary for proper anthocyanin localization or that these inhibitors interfere with TT19 binding to flavonoids. It is interesting that TT19 partially localizes to the tonoplast and that, similar to AN19, it binds in vitro cyanidin and C3G (Sun et al., 2012), increasing their solubility. Taken together, these findings suggest that TT19 functions by increasing the solubility of anthocyanin, preventing the formation of anthocyanin aggregates. If anthocyanins do aggregate like in the tt19 mutant, they are sequestered into the vacuole as AVIs by microautophagy.

A Common Mechanism for the Formation of Other Flavonoid and Polyphenol Intravacuolar Inclusions?

Anthocyanins are not the only polyphenolic compounds reported to localize to vacuolar inclusions. Tannin inclusions resembling AVIs have been shown to accumulate in the vacuole in white spruce and Pinus elliotti (Chafe and Durzan, 1973; Baur and Walkinshaw, 1974). In Arabidopsis, vesicle-like structures very similar to AVIs but containing proanthocyanidins (condensed tannin) precursors were detected in the seed coat of tt19 (Kitamura et al., 2010) and tds4-1, which lacks the enzyme leucoanthocyanidin dioxygenase for both anthocyanin and proanthocyanidin synthesis (Abrahams et al., 2003). In fact, AVIs in grapevine cells have been reported to contain long-chain proanthocyanidins besides anthocyanins (Conn et al., 2010). The coumarin aglycone escutelin and its glycoside esculin are polyphenols with potential roles in photoprotection; both have been reported to occur in intravacuolar inclusions in Fraxinus ornus leaves (Tattini et al., 2014). Thus, although we lack detailed structural information on these other intravacuolar inclusions, it is likely that all these polyphenolic compounds are sequestered into intravacuolar inclusions by microautophagy as described for AVIs.

METHODS

Plant Material and Conditions

Arabidopsis thaliana seeds were surface-sterilized in 70% ethanol for 10 min and washed three times with autoclaved water. Seedlings were grown in either AIC (3% [w/v] sucrose in water) or mAIC (half-strength liquid Murashige and Skoog medium supplemented with 5% [w/v] sucrose), in a rotary shaker with constant light. Both conditions give similar results in terms of anthocyanin induction.

The following seed stocks were obtained from ABRC at The Ohio State University: 5gt (SALK_108458C), tt4 (SALK_020583), tt19 (SALK_105779C), Col-0 expressing 35S:γTIP:CFP (CS16256) and 35S:GFP-HDEL (CS16251).

The mutant lines analyzed in Supplemental Figure 4 have been previously characterized (Koizumi et al., 2005; Ebine et al., 2008, 2011, 2014; Hashiguchi et al., 2010; Uemura et al., 2012; Asaoka et al., 2013; Tanaka et al., 2013; Uemura and Ueda, 2014).

The mAIC medium was supplemented with 100 µM naringenin (Sigma-Aldrich), 1 mM sodium orthovanadate (Enzo Life Science), or 30 µM wortmannin as described in the text. Orthovanadate and wortmannin were added 1 h before naringenin induction.

For construction of the Pro35S:TT19:GFP expression cassette, the TT19 cDNA was isolated from the RNA of seedlings grown in AIC using primers TT19f (CACCATGGTTGTGAAACTATATGGACAGG) and TT19r (GTGACCAGCCAGCACCA) and was cloned into pENTR D-TOPO and then LR recombined into the vector pGWB5 prior to transformation into tt19 using the floral dip method (Clough and Bent, 1998).

Quantification of AVIs

Images of cotyledons were taken with an Olympus BX60 epifluorescence microscope using a 40× objective. For AVI density quantification, a region of interest (ROI) of known area was overlaid on each image and the AVIs contained in it counted using the Cell Counter plug-in in Fiji (Schindelin et al., 2012).

Confocal Imaging

Arabidopsis seedlings grown under mAIC were imaged 5 d after germination unless otherwise stated. All images were taken in a Zeiss LSM 510 Meta except Figure 1A and Supplemental Figures 2 and 3, which were captured with a Zeiss LSM 780 system. Anthocyanins were imaged using 514-nm excitation; emission was collected at 610 to 670 nm.

For detection of membranes around AVIs, all samples were treated with 4 µM FM1-43 (Molecular Probes). tt4 and tt19 seedlings grown in mAIC with 100 µM naringenin were incubated in FM1-43 for 44 and 16 h, respectively, and lisianthus purple petals for 48 h.

For lipid staining, seedlings were incubated in media supplemented with 4 μg/mL Bodipy (Molecular Probes) or 1 mg/mL Nile Blue (Sigma-Aldrich) for 4.5 h and 5 min, respectively. Bodipy was imaged using 488-nm excitation and a 494- to 526-nm emission filter; Nile blue was excited at 458 nm and emission collected at 558 to 612 nm.

For the colocalization of ER markers and anthocyanins in AVI-containing cells, images of wild-type seedlings expressing CFP-HDEL (Nelson et al., 2007) and 5gt and tt4 5gt seedlings expressing CALNEXIN-GFP (Irons et al., 2003) were acquired in a Zeiss LSM780 system. Anthocyanins, GFP, and CFP were excited at 514, 488, and 458 nm, and emission was collected at 606 to 668, 455 to 544, and 464 to 508 nm, respectively. Twelve ROIs were defined within AVI-forming cells and analyzed using the Coloc2 plug-in in Fiji (Schindelin et al., 2012) to obtain PCCs.

Electron Microscopy and Immunolabeling

Arabidopsis cotyledon and lisianthus petals were high-pressure frozen in a Baltec HPM 010 and freeze-substituted in 2% OsO₄ in acetone for 4 d. Samples were embedded in Eponate 12, sectioned, and stained with 2% uranyl acetate in 70% methanol and lead citrate (2.6% lead nitrate and 3.5% sodium citrate, pH 12). Serial sections of 80 nm were imaged and used to reconstruct the 3D architecture of forming AVIs using the program MIDAS of the IMOD software package (Kremer et al., 1996).

For immunolabeling, high-pressure frozen samples were freeze-substituted in 0.2% uranyl acetate and 0.2% glutaraldehyde in acetone at −90°C for 4 d and embedded in Lowicryl HM20 (Electron Microscopy Sciences). Immunolabeling using primary antibody anti-γTIP was performed as described before (Reyes et al., 2011), with the modification that sections were incubated with the primary antibody overnight at 4°C.

FLIM

All images were collected with a multiphoton optical workstation at the Laboratory for Optical and Computational Instrumentation at the University of Wisconsin, with a 40× water immersion objective. The excitation wavelength was 890 nm. A 632- to 60-nm band-pass emission filter (Chroma Technology) was used to selectively collect anthocyanin fluorescence. Fluorescence intensity and lifetime data were collected sequentially with a Hamamatsu GaAsP PMT (gallium arsenide phosphide photomultiplier tube; H7422). FLIM acquisition times ranged from 180 to 240 s. Time-resolved fluorescence emission was collected via time-correlated single-photon counting electronics (Becker and Hickl; SPC-830). The second harmonic from a urea crystal was collected using a 445/20-nm filter and used as the instrument response function in the lifetime fit model.

SPCImage 5.1 software (http://www.becker-hickl.com/software.htm) was used to analyze the fluorescence lifetime decay curves. The lifetime decay curve of each pixel was fit to a double-exponential decay model, where τ₁ and τ₂ are the short and long lifetime components, respectively, and α₁ and α₂ their relative contributions (where α₁ + α₂ = 100%). A Student’s t test was used to determine significant differences between samples.

Anthocyanin Extraction and Analysis

Seedlings were collected and stored at −80°C until lyophilization. Dry weight of lyophilized seedlings was measured; 50 µg/μL extraction solution consisting of 50% (v/v) methanol and 3% (v/v) formic acid was added and incubated at room temperature overnight on a rotary shaker. Samples were centrifuged at 13,500g for 2 min and the supernatant passed through 0.2-µM filters (Nanosep ODM02C35). Total anthocyanins in the resulting filtrates were measured in a spectrophotometer (Nano Drop ND-1000). Anthocyanin composition was analyzed using a Waters Alliance 2695 HPLC equipped with a photodiode array detector. Twenty microliters of plant extract was injected onto a Symmetry C18 column (4.6 × 75 mm, 100 Å, 3.5 µm) at 35°C. The mobile phase flow rate was 1 mL/min and consisted of buffers A (5% [v/v] formic acid in water) and B (5% [v/v] formic acid in acetonitrile), with the following elution profile (0 min 100% A, 20 min 75% A, 22 min 20% A, 22.1 min 100% B, 25 min 100% B, 25.1 min 100% A, and 32 min 100% A) using a linear gradient between time points. Absorbance spectra were collected with the photodiode array detector from 200 to 700 nm. Anthocyanin species from 5gt seedling extracts were identified by mass spectrometry as described previously (Pourcel et al., 2010) using an electrospray probe interfaced with a quadrupole-time-of-flight mass spectrometer (QTof Premier; Micromass) operated in positive ion mode. Probe and source conditions included capillary voltage of 3.2 kV, 400°C desolvation temperature, 50 L h⁻¹ cone gas, 400 L h⁻¹ desolvation gas, and 110°C block temperature. Cyanindin-3-sambubiose and other C3G derivatives from 5gt seedlings were purified by solid-phase extraction. Briefly, 5 mg of dried powder was dissolved in 100 μL methanol:water:glacial acetic acid (99:1). Ten volumes of water:glacial acetic acid (99:1) was added and vortexed vigorously for 1 min. Three volumes of methanol:water:glacial acetic acid (10:89:1) was added and vortexed vigorously for 1 min. A cartridge (Oasis HLB; 60-µm particle size) preconditioned with 2 mL of methanol was equilibrated with 2 mL of methanol:water:glacial acetic acid (10:89:1). After loading the sample, the cartridge was washed with methanol:water:glacial acetic acid (10:89:1). Anthocyanins were eluted with a series of methanol:water:glacial acetic acid (15:84:1 to 80:19:1, at 5% methanol increments). Fractions <40% methanol and >60% methanol containing cyanindin-3-sambubiose and coumarylated cyanindin-3-sambubiose-derivatives, respectively, were pooled, lyophilized, and used for analysis.

Accession Numbers

Sequence data from this article can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: TT4 (AT5G13930), 5GT (At4g14090), and TT19 (AT5G17220).

Supplemental Data

• Supplemental Figure 1. Imaging and HPLC Analysis of Anthocyanins in Different Arabidopsis Genotypes.

• Supplemental Figure 2. Detection of Membranes on Arabidopsis AVIs with the Membrane Dye FM1-43.

• Supplemental Figure 3. Anthocyanins Do Not Aggregate inside the ER during AVI Formation.

• Supplemental Figure 4. Induction of AVI Formation in Vacuolar Trafficking Mutants.

• Supplemental Figure 5. Staining of AVI-Containing Cells with Lipid Dyes.

• Supplemental Figure 6. Rescue of Anthocyanin Synthesis in tt19 Mutant Seedlings Expressing Pro35S:TT19:GFP.

Glossary

ER

endoplasmic reticulum

GST

glutathione S-transferase

AVI

anthocyanin vacuolar inclusion

AIC

anthocyanin inductive condition

C3G

cyanidin-3-O-glucosyde

mAIC

modified AIC

TEM

transmission electron microscopy

PCC

Pearson’s correlation coefficient

FLIM

fluorescence lifetime imaging

ROI

region of interest