The Arabidopsis tt19-4 mutant differentially accumulates proanthocyanidin and anthocyanin through a 3′ amino acid substitution in glutathione S-transferase



    Corresponding author
    1. Agriculture and Agri-Food Canada, Saskatoon Research Center, 107 Science Place, Saskatoon, Saskatchewan, S7N 0X2, Canada
    2. College of Plant Science, Jilin University, Changchun, Jilin, 130062, China
    Search for more papers by this author

    1. Agriculture and Agri-Food Canada, Saskatoon Research Center, 107 Science Place, Saskatoon, Saskatchewan, S7N 0X2, Canada
    Search for more papers by this author

    1. Agriculture and Agri-Food Canada, Saskatoon Research Center, 107 Science Place, Saskatoon, Saskatchewan, S7N 0X2, Canada
    2. Department of Biology, University of Saskatchewan, 112 Science Place, Saskatoon, Saskatchewan, S7N 5E2, Canada
    Search for more papers by this author

    1. Agriculture and Agri-Food Canada, Saskatoon Research Center, 107 Science Place, Saskatoon, Saskatchewan, S7N 0X2, Canada
    Search for more papers by this author

    1. Agriculture and Agri-Food Canada, Saskatoon Research Center, 107 Science Place, Saskatoon, Saskatchewan, S7N 0X2, Canada
    Search for more papers by this author

    1. Agriculture and Agri-Food Canada, Southern Crop Protection and Food Research Centre, 1391 Sandford Street, London, Ontario, N5V 4T3, Canada
    2. Department of Biology, University of Western Ontario, London, ON N6A 5B7, Canada
    Search for more papers by this author

    1. Agriculture and Agri-Food Canada, Southern Crop Protection and Food Research Centre, 1391 Sandford Street, London, Ontario, N5V 4T3, Canada
    Search for more papers by this author

    1. College of Plant Science, Jilin University, Changchun, Jilin, 130062, China
    Search for more papers by this author

    1. Agriculture and Agri-Food Canada, Saskatoon Research Center, 107 Science Place, Saskatoon, Saskatchewan, S7N 0X2, Canada
    Search for more papers by this author

    Corresponding author
    1. Agriculture and Agri-Food Canada, Saskatoon Research Center, 107 Science Place, Saskatoon, Saskatchewan, S7N 0X2, Canada
    Search for more papers by this author

X. Li or M.Y. Gruber. Fax: 306 956 7247, e-mail:;


The Arabidopsis transparent testa (tt) mutant tt19-4 shows reduced seed coat colour, but stains darkly with DMACA and accumulates anthocyanins in aerial tissues. Positional cloning showed that tt19-4 was allelic to tt19-1 and has a G-to-T mutation in a conserved 3′-domain in the TT19-4 gene. Soluble and unextractable seed proanthocyanidins and hydrolysis of unextractable proanthocyanidin differ between wild-type Col-4 and both mutants. However, seed quercetins, unextractable proanthocyanidin hydrolysis, and seedling anthocyanin content, and flavonoid gene expression differ between tt19-1 and tt19-4. Transformation of tt19-1 with a TT19-4 cDNA results in vegetative anthocyanins, whereas TT19-4 cDNA cannot complement the proanthocyanidin and pale seed coat phenotype of tt19-1. Both recombinant TT19 and TT19-4 enzymes are functional GSTs and are localized in the cytosol, but TT19 did not function with wide range of flavonoids and natural products to produce conjugation products. We suggest that the dark seed coat of Arabidopsis is related to soluble proanthocyanidin content and that quercetin holds the key to the function of TT19. In addition, TT19 appears to have a 5′ GSH-binding domain influencing both anthocyanin and proanthocyanidin accumulation and a 3′ domain affecting proanthocyanidin accumulation by a single amino acid substitution.


Flavonoids belong to a group of plant natural products with variable phenolic structures and play important roles in protection against biotic and abiotic stress and in agriculture (Aerts, Barry & McNabb 1999; Lepiniec et al.2006). These substances are now known for their positive effect on health, including antioxidant and antitumor properties (Nijveldt et al. 2001; Kris-Etherton et al. 2002), and have been found in fruit, vegetables, grains, barks, roots, stems, flowers, tea and wine. Anthocyanins and proanthocyanidins are two important plant pigments, which share common flavonoid intermediates until the formation of anthocyanidins (reviewed in Marles, Ray & Gruber 2003) (Fig. 1). Anthocyanins are orange red-to-purple pigments, whereas proanthocyanidins (PAs, syn. condensed tannins) are normally colourless compounds and become yellow-to-dark brown after they have been oxidized (Pourcel et al. 2005). In Arabidopsis seeds, upper pathway flavonoids accumulate both in the seed envelope and in the embryo mainly as glycosylated derivatives (Kerhoas et al. 2006; Routaboul et al. 2006). In contrast, PAs accumulate in the inner integument and in the chalaza zone and are oxidized into complexes as the seed dries, causing the typical brown colour of the wild-type testa of Arabidopsis (Pourcel et al. 2005).

Figure 1.

Proanthocyanidin biosynthetic pathway in Arabidopsis. Dotted lines indicate regions of the pathway which are still not fully defined.

Steps involved in proanthocyanidin condensation, transport from their presumed site of synthesis (ER), or uptake into their final destination in vacuoles, are still poorly understood (reviewed in Xie & Dixon 2005; Goodman, Casati & Walbot 2004; Buer, Muday & Djordjevic 2007). Condensation/polymerization reactions to form PA have not been demonstrated enzymatically, although PA dimers and oligomers can be chemically synthesized by combining flavan-3-ols and carbocations or quinone methides (Delcour, Ferreira & Roux 1983; Delcour et al. 1985). Recently, a trafficking pathway for anthocyanins was discovered in Arabidopsis and overlaps with the endoplasmic reticulum and vacuole protein-sorting route (Poustka et al. 2007).

So far, evidence has been reported on three genes with roles in vacuolar transport of flavonoids, although their detailed functions and positions within the pathway are not entirely clear. The Arabidopsis AHA10 encodes an H+-ATPase (Baxter et al. 2005). The Arabidopsis TRANSPARENT TESTA 12 (TT12) encodes a MATE membrane-spanning protein which acts as a flavonoid/H+-antiporter for the vacuole of the seed coat (Debeaujon et al. 2001; Marinova et al. 2007). Glutathione S-transferases (GSTs), such as BZ2 in maize (Marrs et al. 1995), AN9 in Petunia (Alfenito et al., 1998), GST in Vitis vinifera (Conn et al. 2008) and TT19 (Kitamura, Shikazono & Tanaka 2004) in Arabidopsis, also were reported to be involved in flavonoid transport. Among these reported GSTs, the bz2 mutant displays a brown-to-bronze seed pigmentation resulting from disruption of the ability to sequester anthocyanin into vacuoles. The BZ2 protein was proposed to conjugate cyanidin-3-glucoside with glutathione and then the conjugate would be recognized by a glutathione pump for transport into vacuoles (Marrs et al. 1995). The AN9 protein was reported to bind a range of flavonoids directly to itself in vitro as shown by inhibition of glutathione S-transferase activity after binding (Müller et al., 2000). Four anthocyanin-reduced alleles of TT19 are known in Arabidopsis (Kitamura et al. 2004; Hsieh & Huang 2007). The best known allele tt19-1 has a ∼ 1000 bp 5′ inversion ending at the second intron and accumulates mainly unextractable proanthocyanidin in small vacuoles (Kitamura et al. 2004; 2010). The latter authors also provided evidence from tt12-tt19 double mutants that TT19 functions prior to TT12 (Kitamura et al. 2010).

Recently, a large T-DNA activation-tagged Arabidopsis population was developed and screened for seed coat colour and proanthocyanidin variation (Robinson et al. 2009). Within these new variants, we identified three new tt19 alleles with different T-DNA insertion sites. One allele, tt19-4, differed phenotypically from other tt19 alleles by accumulating reduced levels of soluble PA, enhanced unextractable PA, and wild-type levels of anthocyanin due to a 3′-substitution.


Chemicals, E. coli strains, plant materials and plasmids

1-chloro-2,4-dinitrobenzene (CDNB), proanthocyanidin B2, anthocyanins and flavonoid standards were purchased from Sigma-Aldrich (St. Louis, MO, USA) and Chromadex (Santa Anna, CA, USA), respectively. All solvents used in this study were HPLC grade. Escherichia coli strains DH10B and Bl-21 (DE3) (Invitrogen, Carlsbad, CA, USA), expression vector pET28a (Novagen, Madison, WI, USA), and the sequence vector PGEM-T Easy (Promega, Madison, WI, USA) were used for cloning and overexpression of the new TT19 alleles. Arabidopsis thaliana ecotype Columbia (Col-4) from the Arabidopsis Biological Resource Center (ABRC, Columbus OH, USA) was used to develop the sk activation-tagged population (Robinson et al. 2009). The tt19-1 mutant was generously provided by Dr Kitamura (Takasaki, Gunma, Japan). Arabidopsis seeds were cold-treated at 4 °C for 2–4 d and then allowed to germinate and grow in a greenhouse (at 22 °C with 16 h light and 8 h dark supplemented with halogen lamps). Alternatively, seeds were germinated on ½-strength MS medium containing 1% agar and 3% sucrose in a controlled growth cabinet (16 h light and 8 h dark cycle) at 20 °C to induce anthocyanins. All mutant lines, expression lines and wild-type lines were grown in soil at the same time and under the same conditions to reduce variability.

Isolation and genetic analysis of the TT19-4 gene

All mutants were selected by screening T3 seeds of sk activation-tagged Arabidopsis by visual examination, followed by DMACA histochemical staining for proanthocyanidin variation. T-DNA insert loci were identified by TAIL-PCR according to Liu et al. (1995) or plasmid rescue using 10 µg genomic DNA digested by Kpn I, BamH I or EcoRI, followed by self-ligation and cloning as described by Weigel et al. (2000). Positive ‘rescued’ colonies and TAIL-PCR products were sequenced and analysed using Blastn against the NCBI non-redundant database.

For Southern blot analysis, genomic DNA was isolated from 10-day seedlings by the CTAB method (Saghai-Maroof et al. 1984), digested with HindIII, EcoRI and BamHI restriction enzymes, and resolved on 1.2% agarose gels. Southern blot analysis was performed on nylon membranes according to Southern (1975). The TT19 coding region was radiolabelled with α-32P dCTP using Ready-to-Go DNA labelling beads (Amersham Pharmacia Biotech) and was used as a probe. Hybridization was carried out at 65 °C for 2 h with 0.6 M NaCl and blots were washed at high stringency (0.2× SSC, 0.1% SDS).

Tt19-4 lines were crossed reciprocally with Columbia (Col-4) and Landsberg (Ler) for genetic analysis and mapping. When the seed colour pattern of tt19-4 did not co-segregate with BAR herbicide-resistance (T-DNA insertion), molecular mapping was conducted on T-DNA-free tt19-4 segregants. Accordingly, genomic DNA from 50 F2 plants segregating for seed colour pattern was analysed by Simple Sequence Length Polymorphism (SSLP) markers as described by Lukowitz, Gillmor & Scheible (2000). When the pale seed phenotype was mapped between two SSLP markers, each of three TT genes discovered between the markers was sequenced and compared with WT sequence. Amino acid sequences were aligned and compared with other GSTs using the MEGA 4.0 and AlignX program (part of the Invitrogen Vector NTI suite) and fit to AlignX default parameters (gap opening penalty, 10; gap extension penalty, 0.05; gap separation penalty range, 8; identity for alignment delay, 40%) (Lu & Moriyama 2004).

Complementation of the tt19-1 mutant

Full length TT19-WT and TT19-4 cDNAs were amplified from Arabidopsis Col-4 and tt19-4 seedlings, respectively, using primer pairs P3+P4 and high fidelity DNA polymerase Pfu Phusion (Invitrogen) (Supporting Information Table S1). The amplified fragments were sequenced and cloned into XbaI and SmaI sites of the binary vector pBI121 under the control of the 35S promoter. All constructs were first introduced into Agrobacterium tumefaciens strain GV3101 and then introduced into Arabidopsis by floral dipping (Clough & Bent 1998). Transformation-positive T1 transgenic plants were selected by seedling growth on ½-strength MS media with 50 µg/L kanamycin.

RT-PCR and Real-Time Q-PCR

Seedlings of Arabidopsis WT (Col-4), tt19-1, and a T-DNA-free tt19-4 segregant were grown on agar plates (½-strength MS medium containing 1% agar and 3% sucrose) for 10 days to induce anthocyanins (Baier et al. 2004). Three independent RNA preparations were extracted using a commercial RNAEasy mini kit (Qiagen, Valencia, CA, USA), then reverse transcribed by Superscript II (Invitrogen) and amplified with Taq DNA polymerase (Invitrogen) using the Arabidopsis ACT2 gene as an internal control and gene-specific primers for TT19 (P1+P2, Supporting Information Table S1). Primers were designed from the TT19 mRNA sequence deposited in GenBank (Accession No. NM_121728). RT-PCR conditions: 94 °C for 3 min, 25 and 30 cycles at 94 °C for 30 s, 55 °C for 30 s, and 72 °C for 1 min, followed at the end by 72 °C for 5 min.

RNA preparations from seedlings grown under anthocyanin-inducing conditions were also used in quantitative real-time PCR reactions (Q-PCR) with gene-specific primers for TT19 and 21 other flavonoid genes (Supporting Information Table S1). Q-PCR reactions were conducted using SuperScript III First-Strand Synthesis SuperMix, a Platinum SYBR Green qPCR SuperMix-UDG kit (Invitrogen), and the StepOnePlus Real-Time PCR System (Applied Biosystems) following the manufacturers' instructions and as described previously (Gao et al. 2009). For each pair of gene-specific primers, gel electrophoresis and melting curve analyses were conducted to ensure a single PCR amplicon of the expected length and melting temperature. Each RNA preparation was assayed in triplicate and data were analysed using StepOne Software v2.0 (Applied Biosystems). The level of each mRNA was calculated using the mean cycle threshold (Ct) value and normalized to that of endogenous reference genes, ACTIN2 or EF-1α (Supporting Information Table S1). All results were shown as means with corresponding standard errors (SE).

Co-expression analysis of TT19 and flavonoid genes was conducted in silico using the ‘Correlated Gene’ program at RIKEN PRIME ( based on the ATTED-II microarray database of co-expressed genes (Wangwattana et al. 2008).

HPLC UV/MS analyses of flavonoid composition

Flavonoids were extracted from four batches of Arabidopsis seed and each batch analysed twice as described by Kerhoas et al. (2006) with minor modifications. Briefly, 100 mg frozen seeds were ground in liquid nitrogen in a 20-ml Potter (Elvehjem), followed by grinding in 10 mL acetone/water (70:30; v/v) for 10 min using 50 µg of flavone (Sigma) as an internal standard. Following filtration, the pellet was re-extracted overnight at 4 °C in the dark, the two extracts were combined and evaporated at 35 °C under vacuum, and the dried extract was dissolved to 1 mg mL−1 methanol/water (50: 50; v/v). HPLC/UV/MS analysis was conducted using a Hewlett-Packard Agilent 1200 chromatograph, a G1315D diode array detector, a Varian 320 MS/MS detector, HP Chemstation software and a COSMOSIL C18 column (250 × 4.6 mm, 5 µm, NACALAI, Kyoto, Japan) with a 40 min gradient of acetonitrile (solvent A) and water (solvent B) applied at a 0.5 mL min−1 flow rate: 0–5 min, 10% A; 5–30 min, 50% A (linear gradient); 30–40 min, 100% A and an injection volume of 20 µL according to Kerhoas et al. (2006). Flavonoids were quantified by calibration of UV peak areas (λ370) relative to the flavone standard. Flavonoid peaks were detected by absorbance at 370 nm and by both negative and positive electrospray mass spectrometry from 200 to 1200 amu at 350 °C, with a 500 V capillary voltage and a 70 V fragmenter.

Anthocyanin and PA content and composition

Total anthocyanins were induced as described earlier, extracted from seedlings with 70% methanol, and the extract directly injected into an Hewlett Packard Agilent 1100 HPLC and a Zorbax C18 column (150 × 4.6 mm, 5 µm) with a 40 min linear gradient of 5% to 100% MeOH v/v, and an injection volume of 20 µL. Anthocyanins were monitored at UV525nm and quantified using a standard curve of cyanidin chloride (Sigma-Aldrich, St. Louis, MO, USA). For analysis of naturally occurring acetylated anthocyanins, the method of Bloor & Abrahams (2002) was conducted on Col-4, tt19-1, tt19-4, fah, sct and transgene-complemented mutant seedlings.

Extractable seed PAs were quantified after grinding 300 mg seed in 30 mL (10 mg ml−1) of acetone:H2O (70:30, v/v). The extracts were dried, made up to 1 mL, and hydrolysed to anthocyanidins using 2 mL of a n-BuOH/HCl solution (95:5 v/v) and 0.1 mL of FeSO4 in 2 M HCl, for 75 min at 95 °C (Porter, Hrstich & Chan 1986). After cooling and centrifugation, the hydrolysed supernatant was chromatographed on an HP Agilent 1100 HPLC and Zorbax C18 column as described for anthocyanins and extractable PAs were calculated from the absorbance peak (512 nm) and a similarly-treated proanthocyanidin B2 standard curve. Unextractable PA was measured as for extractable PA, but from the supernatant arising after boiling the solid residue (30 mg) remaining after acetone extraction in 1 mL n-BuOH/HCl solution (95:5 v/v) (Porter et al. 1986). Quercetin released from unextractable PA was detected by absorbance at 370 nm, retention time with a standard, and by positive electrospray mass spectrometry.

Subcellular localization of TT19 and TT19-4 fluorescent protein fusions

The cDNA clone (accession no. U70496) encoding smRS-GFP (Davis & Vierstra 1998) was subcloned into the binary vector pCAMBIA (Hajdukiewicz, Svab & Maliga 1994) to generate the construct p35S::smRS-GFP. The entire coding region of TT19 and TT19-4 was amplified by PCR using primers P5 and P6 and inserted between the PstI and BamHI sites of p35S::smRS-GFP resulting in an in-frame translational fusion at the amino-terminus. These fusion constructs, called 35S::pTT19::smRS-GFP or 35S::pTT19-4::smRS-GFP, and the negative ‘empty’ control plasmid p35S::smRS-GFP were used for transient expression by infiltration into N. tabacum leaf epidermal cells as described by Sparkes et al. (2006) with the following modifications. A localization experiment was conducted using a cytoplasmic membrane marker consisting of an in-frame fusion of a red fluorescent protein with the first 2 exons and the first intron of the A. thaliana RCI2 plasma membrane protein (Medina, Ballesteros & Salinas 2007). A TCS SP2 confocal laser scanning microscope (Leica Microsystems, Ontario, Canada) equipped with a 63 × water-immersion objective was used to examine the subcellular localization of GFP in leaf epidermal cells. For the imaging of GFP and RFP expression, excitation with a 488 nm argon laser was used, and fluorescence was detected at 500 to 525 nm and 630 to 690 nm, respectively.

Expression and purification of recombinant TT19 and TT19-4 in E. coli

Full-length TT19 and TT19-4 cDNAs were amplified using primer pairs P7 and P8 (Supporting Information Table S1) after BamHI and SacI digestion. Amplicons were inserted into a pET28a expression vector and used to express recombinant His-tagged pET28a-TT19 and pET28a-TT19-4 in E. coli strain BL21 (DE3). Single colonies were grown to 0.5 (A600) in LB medium containing 0.05 mg mL−1 kanamycin at 23 °C, 28 °C or 37 °C and then induced at the same temperatures by 0.5 mM IPTG (isopropyl-D-thiogalactopyranoside) for 4 h. Induced cells were harvested by centrifugation at 5000 rpm for 15 min at 4 °C. WT TT19 protein was solubilized in its native state and easily extracted using a protein extraction kit (Promega). TT19-4 protein was denatured in 8 M urea and recovered after dialysis 3 times against 1 × PBS (140 mM NaCl, 2.7 mM KCl, 1.4 mM KH2PO4 and 8 mM Na2HPO4, pH 7.4) according to Yoshitaka et al. (2003). In order to compare both proteins, TT19 was also denatured and refolded prior to testing enzyme activity.

His-tagged TT19 and TT19-4 proteins were purified at 4 °C using a His-Bind purification kit (Novagen) following the manufacture's protocol. The concentration of purified proteins was measured by NanoDrop® ND-1000 (NanoDrop Technologies, Inc., Wilmington, DE, USA). Mass spectra of TT19 peptides were recorded in the positive ion reflection mode from a matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) using an API Qstar XL system (Applied Biosystems) at the Structural Sciences Center of the University of Saskatchewan.

In vitro enzyme assays

GST binding activities were determined by measuring the ability of recombinant TT19 and TT19-4 proteins to conjugate 1-chloro-2,4-dinitrobenzene (CDNB, the ‘universal’ substrate of GSTs) with GSH to form dinitro-phenol-glutathione (DNP-GS) (Dixon et al. 2009). Reaction buffer at a final concentration of 97 mM potassium phosphate, 0.97 mM EDTA, 2.5 mM glutathione and 1 mM CDNB (pH 6.5) was equilibrated to 25 °C and monitored at A340 until the baseline was constant. After adding purified TT19 or TT19-4, the increase in A340 for 30 min was recorded and corrected for corresponding non-enzymic and endogenous enzymic reaction rates. After 60 min, 20 µL of the reaction mixture was centrifuged and the supernatant was analysed by isocratic HPLC as for anthocyanin compositional analysis, but monitored at A340.

To test the potential of TT19 and TT19-4 recombinant proteins to catalyse the binding of glutathione with flavonoids and other related natural products, in vitro enzyme assay conditions were altered to include incubation at room temperature for 1 h in 1 mL total volume containing 800 µL Tris-HCl buffer (100 mM, pH 7.0), 100 µL purified recombinant TT19 and TT19-4 proteins (0.6 µg µL−1), 50 µL flavonoid, or anthocyanin, or phenolic acid (Supporting Information Table S2) (1 µg µL−1) and 50 µL GSH (1 µg µL−1). The reaction mixture was centrifuged and 20 µL subjected to HPLC analysis as for the reaction with universal substrate, but monitored at 203 nm to show GSH conjugation. Reactions to show direct binding to TT19 were identical to GSH conjugation reactions to natural products, except that GSH was removed.

Structural modelling of TT19 and TT19-4

The sequences of TT19 and TT19-4 protein were introduced into a protein structure prediction web server ESyPred3D ( (Lambert et al. 2002), using the crystallized Arabidopsis GST protein AtPM24 as a template (Reinemer et al. 1996). Structural simulations were conducted using ICM-pro v3.4 with a MMFF local and global Mol-Mechanics calculation module (MolSoft L.L.C., La Jolla, CA, USA). All sequence alignments were performed using the MEGA 4.0 software.

Statistical analysis

Enzyme activity, compositional and Q-PCR data were expressed as means ± standard error and were analysed by anova using LSD (P < 0.05, SAS 9.0) (SAS Institute, Inc. Cary, NC, USA).


Isolation and molecular genetic analysis of tt19-4, tt19-5 and tt19-6

Approximately 47 000 activation-tagged T3 seed lines of Arabidopsis ecotype Columbia (Col-4) (Weigel et al. 2000; Robinson et al. 2009) were screened for differences in visible seed colour and proanthocyanidin patterning after soaking seeds in DMACA. From this population, we recovered a pale yellow seed coat mutant allele, sk36391, with a novel phenotype (hereafter called tt19-4) and two additional new T-DNA alleles (sk4945 and sk20780) with phenotypes like tt19-1 (hereafter called tt19-5 and tt19-6, respectively) (Fig 2a–l). Backcross F1 (BCF1) seeds and BCF2 seeds of tt19-4 had a wild-type-like medium-brown seed colour phenotype. In the BCF3 generation, 124 plants with the wild-type seed phenotype and 38 plants with the pale mutant seed phenotype indicated a 3:1 segregation ratio (χ2 = 2.78; P < 0.05) and a single recessive seed coat mutation.

Figure 2.

Phenotypes of tt19 mutants. (a, d, g, j) Wild-type Columbia. (b, e, h, k) tt19-1. (c, f, i, l) tt19-4. (a, b, c): 80× magnification, unstained seeds under light microscope. (d, e, f) 16× magnification, seeds stained with DMACA. (g, h, i) 10-day seedlings were grown on agar plates in sucrose to promote anthocyanin accumulation. (j, k, l) Plants growing on soil shortly before harvest. (m) tt19-1 seedling. (n, o, p) Twelve-day-old tt19-1 COMM seedlings (i.e. tt19-1 complemented with the TT19-4 transgene). (q) Major anthocyanins determined by LC-MS/MS in Arabidopsis. Anthocyanin 1 (R1 = H, R2 = Glc): Cyanidin 3-O-[2-O-(xylosyl)-6-O-(4-O-(glucosyl)-p-coumaroyl)) diglucoside] 5-O-[6-O-(malonyl) glucoside. Anthocyanin 2 (R1 = Sin, R2 = H): Cyanidin 3-O-[2-O-(2-O-(sinapoyl) xylosyl)-6-O-(p-coumaroyl) glucoside] 5-O-[6-O-(malonyl) glucoside. Anthocyanin 3 (R1 = Sin, R2 = Glc): Cyanidin 3-O-[2-O-(2-O-(sinapoyl) (xylosyl)-6-O-(4-O-(glucosyl)-p-coumaroyl)) glucoside] 5-O-[6-O-(malonyl) glucoside. (r) Total anthocyanin content in stem and 12-day-old seedlings of Col,4, tt19-1 and tt19-4. (s) Content of individual anthocyanins in tt19 mutant lines and the phenolic mutants fah and sct. Data from anthocyanin analysis was consistent with Bloor & Abrahams (2002). (t) Total anthocyanin content in 12-day-old seedlings of tt19-1, Col-4 WT, and three tt19-1 COMM lines. (r,s,t) Different lettering on anthocyanin bars indicates significant difference of the means of three independent extractions measured at 525 nm (P < 0.05).

Southern blot analysis showed the likelihood of four T-DNA insertions in the tt19-4 mutant genome (Fig. 3a). Using TAIL-PCR and plasmid rescue, four genes were shown to be disrupted (Accession no. At5g54950: Aconitate hydratase-related; At4g21630: Subtilase family protein; At3g57280: Hypothetical protein; and At5g59710: VIRE2 Interacting Protein 2). However, SALK T-DNA mutants defining these genes did not show the tt19-4 seed coat colour phenotype and the mutant seed colour did not co-segregate with herbicide resistance (i.e. T-DNA insertion) (data not shown). Consequently, we mapped the pale seed colour locus using SSLP markers (Lukowitz et al. 2000) and T-DNA-free tt19-4 segregants from an F2 crossed progeny population from tt19-4 X Landsberg. The seed coat colour locus was located between CTR1 and ciw8 markers. Three TRANSPARENT TESTA genes lay in this region of the Arabidopsis genome between the two markers: TT7 (At5g07990, F3'H, Flavonoid 3′-Hydroxylase), TT4 (At5g13930, CHS, Chalcone Synthase), and TT19 (At5g17220, GST, Glutathione S-Transferase) (Fig. 1).

Figure 3.

Molecular analyses of Arabidopsis tt19 mutants. (a) Southern blot of the tt19-4 mutant. Each restriction enzyme cuts within the T-DNA; (b) Alignment of the C-terminal sequence of the mutated 3′ end of the TT19-4 gene and translated protein with those of wild-type TT19. The boxed area indicates the mutation site; (c) Schematic of the Arabidopsis TT19 gene showing the position and structure of all four known tt19 mutations (Kitamura et al. 2004); (d) Expression (real-time QPCR) of the TT19 gene in tt19-1 and tt19-4 relative to Col-4. (e) Expression level (Q-PCR) of TT19 in COMM lines (i.e. mutant line tt19-1 complemented with the TT19-4 transgene) compared with tt19-1 and Col-4. (d, e) Different lettering on bars indicates significant difference of the means of three independent measurements (P < 0.05).

The unstained pale seed coat colour of tt19-4 was very similar to tt19-1 and was a much lighter colour compared to the Columbia medium dark brown wild-type seed coat (Fig 2a–c). Crosses between T-DNA-free mutant tt19-4 and mutant tt19-1 showed that tt19-1 could not complement the tt19-4 seed coat phenotype, confirming that the seed phenotype of tt19-4 was caused by a mutation in the TT19-4 gene. Sequence analysis and alignment of the TT19-4 gene revealed that a G-to-T mutation occurred in the 3′-domain, resulting in a missense mutation that converted Trp (amino acid 205) into Leu, whereas the TT19-1 gene had a 1000 bp inversion of the complete 5′ region into the second intron (Fig. 3b). These data suggested that the mutant tt19-4 is allelic to tt19-1. Q-PCR revealed that tt19-1 is a knock-down mutant with decreased GST transcript levels, whereas the expression of the TT19-4 GST was elevated two-fold higher than WT levels both in seedlings and seed of tt19-4 (Fig. 3d). Alignment between TT19, TT19-4 and other flavonoid GSTs showed that amino acid 205 is not conserved in other GSTs, even though it sits in the middle of a small conserved region.

The two additional mutant lines tt19-5 and tt19-6 recovered from the activation-tagged population had different mutation sites in the At5g17220 GST gene (Fig. 3c). Tt19-5 (line sk4945) was a gst mutation missing four base pairs from nucleotide 420-to-423, resulting in an early stop and encoding a 165 AA truncated peptide. Tt19-6 (line sk20780) had a single T-DNA insertion within the first exon of the TT19 gene. These two tt19 mutants showed the same pale unstained seed colour phenotype observed for tt19-1 at the ripening stage, in contrast to the medium brown seeds of Col-4 (data not shown). With increasing length of desiccation or storage (5-6 months), the pale brown unstained seed coat of these and all other known tt19 mutants darkened to a medium brown colour that resembled the wild-type Col-4 seed.

Anthocyanin accumulation is not affected by a 3′ substitution in TT19-4

Tt19-1, tt19-5 and tt19-6 mutant lines did not accumulate red pigment in any vegetative tissues. In contrast, mutant tt19-4 accumulated red pigment in seedlings, young rosette plants, and mature plants to very similar or greater levels than Col-4 (Fig 2g–l, r). Since pigment accumulation in tt19-4 is distinct from other tt19 mutants, the T-DNA-free tt19-4 line was selected for additional investigation. To confirm that anthocyanin accumulated in tt19-4 vegetative tissues and was not reduced as a result of the single nucleotide mutation in TT19-4, constructs containing the TT19-4 or WT-TT19 cDNAs driven by the 35S promoter were introduced into tt19-1 mutants. All T3 homozygous complemented tt19-1 expression lines for both versions of the transgene, i.e. COMm lines with the mutant 35S:TT19-4 gene and COMWT lines with the wild-type gene 35S:WT-TT19 exhibited anthocyanin pigmentation similar to WT levels when seedlings were grown on sucrose-supplemented media (Fig. 2m–p; Fig. 2t; COMm lines shown), and even when expression levels due to the 35S promoter were extremely high (Fig. 3e; COMm lines shown). However, the pale seed coat colour phenotype at the ripening stage remained identical in these COMm lines to the uncomplemented mutant background tt19-1 (data not shown). In general, these data confirm that the pale seed colour and dark DMACA stain reaction in tt19-4, together with the wild-type-like accumulation of anthocyanin in its vegetative tissues, was caused by the single 3′ nucleotide mutation in the TT19-4 gene.

To gain a more detailed understanding of anthocyanin profiles in these mutant lines, we studied individual anthocyanin components of Col-4, tt19-1, and tt19-4 (Fig. 2q). Later, after obtaining recombinant enzyme binding to sinapic acid, we compared them with two phenylpropanoid pathway mutants, fah and sct. Fah is disrupted in F5H (ferulate-5-hydroxylase), while sct is disrupted in SCT (sinapoylglucose:choline sinapoyltransferase) and accumulates disinapoyl glucosides. Tt19-4 and sct showed a similar wild-type pattern of accumulation for the three major anthocyanins (Compounds 1-3 in Fig. 2q, s). In contrast, the fah mutant showed reduced amount of anthocyanin (compound 3) and tt19-1 was reduced in all three anthocyanins (Fig. 2s).

Flavonoid and PA accumulation in tt19 mutants and Col-4

To investigate whether the single 3′ nucleotide mutation in TT19-4 affected flavonoid composition, even though it did not disrupt anthocyanin accumulation, flavonoid and PA profiling was conducted. Distinct flavonoid profiles were observed between the tt19-1 and tt19-4 mutants and Col-4 seeds from a comprehensive analysis of LC mass fragmentation patterns, UV absorption patterns, retention time of commercial standards and a comparison of data with literature values. Seven HPLC peaks clearly appeared in the seed 70% methanol extracts of all three lines in 2-month-old seed (Figs 4a–c). Five of the seven peaks were identified as flavonoid glycosides with characteristics identical to those in Kerhoas et al. (2006) (Table 1). Quercetin-3-O-rhamnoside (peak 7, Q3R) was increased significantly in both mutants tt19-1 and tt19-4 compared with Col-4. In contrast, quercetin 3-O-β-glucopyranoside-7-O-α-rhamnopyranoside (peak 1, Q3G7R) was increased in tt19-1 and not in tt19-4 compared with Col-4. The two kaempferol derivatives (peaks 3 and 5), quercetin 3,7-diO-α-rhamnoside (peak 4, Q37R), and two sinapic esters (peaks 2 and 6) were unchanged between all three lines (Fig. 4d). Isorhamnetin glycoside (traces) and an additional kaempferol glycoside were also detected, but only in newly collected seed (data not shown). (-)-Epi-catechin and dimers of Q-3-O-rhamnoside (biflavonols) detected by Routaboul et al. (2006) were not detected in any of the seed samples under our growth conditions.

Figure 4.

Composition of flavonoids, anthocyanin and proanthocyanidins in Col-4, tt19-1 and tt19-4. (a,b,c) Representational HPLC-UV traces of 70% methanol seed extracts (not containing the usual flavone quantification standard). Flavonoid peaks (orange); sinapic acid peaks (black). (a) Col-4; (b) Tt19-1; (c) Tt19-4; (d) Quantification of flavonoids and PAs in mature seed. Means between lines with significant differences for individual compounds (P ≤ 0.05) are indicated by different letters. Flavonoids were quantified from HPLC-UV peak areas normalized to a co-injected flavone standard. Soluble and unextractable PAs were quantified using the BuOH/HCl assay as described in the Materials and Methods section. (e) Differential patterns of anthocyanidin sub-unit formation (λ512) during acid hydrolysis of unextractable PAs. (f,g) HPLC traces of anthocyanidins (λ512) released after acid hydrolysis. (f) soluble PAs. (g) unextractable PAs. (h) HPLC traces of quercetin (λ370) released after acid hydrolysis of unextractable PAs. (f,g,h) Cy, cyanidin. Q, quercetin.

Table 1.  Flavonoids identified by LC-MS/MS and UV spectra
PeakCompound nameRT (min)[M+H]+Fragments
1Quercetin 3-O-β-glucopyranoside-7-O-α-rhamnopyranoside3.83611449, 303
3Kaempferol 3-O-β-glucopyranoside-7-O-α-rhamnopyranoside5.42595433, 287
4Quercetin 3,7-di-O-α-rhamnopyranoside6.17595449, 303
5Kaempferol 3,7-di-O-α-rhamnopyranoside7.98579433, 287

Proanthocyanidins are known to be less extractable and bound tightly with cell wall components when highly polymerized (Fournand et al. 2006). When extractable and unextractable proanthocyanidins were profiled in our Arabidopsis lines by the butanol:HCl hydrolysis method, Col-4 showed 7-fold higher extractable seed PA (released anthocyanidins) than present in either the tt19-1 or tt19-4 mutants in 70% MeOH extracts (Fig. 4d). In contrast, both mutants showed higher unextractable seed PA (anthocyanidins released from the extraction residue) than Col-4, such that total PA was equivalent for all three lines. Even so, unextractable PA was ∼1.3-fold higher in seeds of mutant tt19-1 than in seeds of tt19-4. A time-course also showed that unextractable PA in tt19-1 and tt19-4 seeds reached maximum peaks of hydrolysis by 40 min and 30 min, respectively, while unextractable PA hydrolysis from Col-4 seeds proceeded steadily and much more slowly (Fig. 4e). After 60 min, all three lines continued at the same rate of hydrolysis (measured up to 90 min; data not shown).

Extractable PA hydrolysates from Col-4 seed were composed primarily of anthocyanidin subunits (Fig. 4f). In contrast, hydrolysates from the unextractable fraction (residue) were composed of a mixture of anthocyanidins and a larger proportion of quercetin, which showed the same retention time as an authentic standard and [M + H]+ of 303 (Fig 4g, h). Hydrolysates of unextractable residues from tt19-1 and tt19-4 were composed mainly of anthocyanidin, with ∼4-fold less quercetin than residue hydrolysates in Col-4 (Fig 4g–h).

Expression of flavonoid pathway genes in WT, tt19-1 and tt19-4

TT19 was proposed as a flavonoid lower pathway gene by Kitamura et al. (2004), but its relationship with other flavonoid genes was not clearly defined. To gain insight into the effect of mutations in TT19 on flavonoid metabolism, the expression patterns of 21 flavonoid genes from available microarray data were initially evaluated in silico for co-expression with TT19 using the ‘ATTED-II’ database ( Eleven of these flavonoid genes had similar expression coefficients and were co-ordinated with TT19 using microarray data from ATTED-II (Supporting Information Fig. S1a). QPCR analysis on these co-expressed genes in tt19-1 and tt19-4 mutants showed that RNA for ANS, DFR, PAP1, F3OGT, TDS6 and an A5MAT (malonyl CoA anthocyanin acyl transferase) was increased in 10-day-old seedlings of both mutants after growth under conditions which induced anthocyanins. In contrast, F3H and F3'H RNAs were up-regulated only in tt19-4 seedlings and slightly down-regulated in tt19-1 seedlings (Supporting Information Fig. S1b). Several other flavonoid pathway genes usually showing coordinated expression similar to WT TT19 (i.e. A5GT: anthocyanin 5-O-glucosyltransferase, a second 5MAT homologue, and SCPL10: sinapoyl-Glc:anthocyanin acyltransferase) were unchanged in tt19-4 seedlings when tested by Q-PCR, but their expression was reduced in tt19-1 seedlings (Supporting Information Fig. S1b).

Seedlings of the two mutant lines also were tested for expression of four flavonoid regulatory genes (TT2, TT8, TTG1, and TTG2) and six additional flavonoid biosynthetic genes normally not found to be co-expressed with TT19 in WT Arabidopsis using the ‘ATTED-II’ microarray database. Transcription of TT2 and TT16 was reduced 2-4-fold in mutant lines, with tt19-1 being most extreme, while TT8 was enhanced almost 2-fold in both mutants (Supporting Information Fig. S1c). The upper pathway genes CHS, CHI and FLS1 and the lower pathway gene AHA10 had higher expression in mutant tt19-4 (Supporting Information Fig. S1c). In contrast, Col-4 and the mutant tt19-1 had similarly low expression for these genes.

Purification, localization, and enzyme activity of recombinant TT19 and TT19-4

Although the TT19 gene has been genetically characterized and localized to the cytoplasm (Dixon et al. 2009 and Kitamura et al. 2010) and its function has been related to proanthocyanidin and anthocyanin accumulation, enzyme activity of the protein and purification of a recombinant protein has not been achieved to date (Kitamura et al. 2010). To characterize the function of the TT19 and TT19-4 proteins, we introduced their cDNAs into a pET28a expression vector and expressed them as his-tagged recombinant proteins in E. coli. Cells containing wild-type TT19 protein were lysed after growth and the recombinant protein was easily extracted to yield a single distinct 28 kDa band after SDS-PAGE (Fig. 5a, lane 2). In contrast, TT19-4 protein was unextractable under similar conditions, and required denaturation first, followed by extraction and purification after re-naturation (Yoshitaka et al. 2003). Purified TT19 and TT19-4 proteins (Fig. 5a, lanes 6–7) were quantified at 0.6 and 0.3 µg µL−1, respectively. TT19 was further subjected to MALDI/TOF-MS/MS. A total of 9 peptide fragments matching the calculated molecular weight of GST-TT19 peptides were detected and gave a reconstructed molecular weight of 28 310 Da (Table 2). This was consistent with the predicted molecular weight of TT19 (24, 581.0 Da) and the pET28a His-tag (3 563.9 Da), respectively.

Figure 5.

Enzyme activities of purified recombinant TT19 and TT19-4. (a) SDS-PAGE analyses of TT19 and TT19-4 purification by Ni-NTA conjugation. Lane 1, protein molecular weight marker; lane 2, supernatant of TT19 cell lysate; lane 3, supernatant of TT19-4 cell lysate; lane 4-5, column wash; lanes 6-7, purified TT19 and TT19-4 proteins, respectively. (b) Typical DNP-GS conjugation product (squares) formed by TT19 (9 µg) binding the universal substrate CDNB to GSH. Reaction was monitored at A340. Triangles represent a control reaction without enzyme. TT19-4 showed the same pattern of DNP-GS production as TT19.

Table 2.  TT19 peptide fragment assigned by MALDI-TOF MS/MS
m/z (Da)Assignments

The recombinant TT19 and TT19-4 proteins were tested for glutathione S-transferase (GST) activity using the universal substrate 1-chloro-2,4-dinitrobenzene (CDNB). Since TT19-4 could not be easily solubilized, both recombinant proteins were denatured and refolded to ensure consistent enzyme activity data. HPLC analysis clearly showed both proteins formed the reaction product, a 1-GS-2,4-dinitro-benzene conjugate, with identical UV spectra and retention time as literature values of DNP-GS (Fig. 5b) (Farkas, Berry & Aga 2007). This confirmed that both the mutant TT19-4 and wild-type TT19 recombinant proteins had glutathione S-transferase activity.

The wild-type TT19 recombinant protein was also tested with 17 flavonoid monomers and glycosides, two anthocyanins, two proanthocyanidin dimers, and 10 phenolic acids as substrates (Supporting Information Table S2), since it had been suggested earlier that TT19 may glutathionate flavonoids and transport them into the vacuole to form anthocyanin and PAs (Kitamura et al. 2004). GSH conjugates were not detected in vitro for the wild-type TT19 protein in any reactions with flavonoid, anthocyanin, proanthocyanidin and most phenolic acid substrates; hence we did not test these substrates with the TT19-4 protein. However, binding of sinapic acid and three coumaric acid isomers to the TT19 protein itself did occur, as measured by the removal of substrate from the reaction and its reappearance after heat denaturation post-reaction. Crude protein extracts from 35S::TT19 over-expression in a WT background were also tested in the enzyme assay with all substrates in Supporting Information Table S2. As with purified proteins, these cell-free extracts also did not glutathionylate flavonoids.

Confocal microscopy of tobacco cells transiently transformed with TT19-1-GFP and TT19-4-GFP clearly revealed green fluorescence from both fusion proteins 48 h after infiltration. Fluorescence was observed in cytoplasmic strands distinct from cytoplasmic membrane, as well as in the nucleus (Supporting Information Fig. S2). The localization pattern for both proteins in the cytoplasm was also distinct from the localization of a red fluorescent cytoplasmic membrane fusion marker only at the periphery of the cell (Supporting Information Fig. S2).

Structural model and GSH binding-site of TT19-GST

In eukaryotes, the structure of several GST enzymes has been determined by X-ray crystallography (Reinemer et al. 1996). Most GSTs share a similar three-dimensional structure and possess a well-defined glutathione-binding domain in their active sites (Farkas et al. 2007). This existing protein structural data provided an opportunity to compare and predict the three-dimensional structure of TT19 and TT19-4 using the homology modelling web-service, ESyPred3D (Lambert et al. 2002). The herbicide-induced Arabidopsis GST AtPM24 (AtGSTF2, At4g02520) with 39.6% AA sequence identity to TT19 was used as a structural template, since it was the closest aligned GST that had been fully analysed by X-ray crystallography (Reinemer et al. 1996). The results showed that the predicted TT19 protein has structural features similar to AtPM24, including a characteristic βαβαββα folding pattern in domain I and α-helices primarily in domain II (Fig. 6a). Based on TT19 protein sequence alignments to AtPM24, petunia AN9 (the most closely aligned anthocyanin-related GST; Müller et al., 2000), maize IAW9 (maize GSTIII) and maize BZ2, amino acid residues Lys41, Pro50, Gln53, Glu66, Ser67 and Arg68 in TT19 should correspond to the site of GST binding activity. Each of these amino acids is located in a 5′ region conserved in both Arabidopsis, petunia, and maize (Figs 6a, c). However, the Trp to Leu transition, which distinguishes TT19-4 from the wild-type TT19 is in a non-conserved position found within a very short conserved region in the 3′ end of these translated sequences (Fig. 6c). Comparison of predicted 3D structures from the translated TT19 and TT19-4 sequences suggested these two proteins should fold similarly within domain I, but differently within the C-terminal (Fig. 6b).

Figure 6.

Amino acid alignment and structural modelling of the predicted 25 kDa Arabidopsis TT19-GST. (a) Predicted 3D structure of TT19-GST. Stick model indicates location of GSH binding. α-Helices (12) and β sheets (4) are labelled. Red represents domain 1, consisting mainly of α-helices. Blue represents N-terminal GST-binding domain 2, consisting of βαβαβßα. (b) Simulations of the predicted folding of TT19 and TT19-4 proteins by Molsoft ICM 3.4 software; C: C-terminal. Green represents WT TT19. Yellow represents TT19-4. (c) Amino acid alignment of the translated TT19 sequence with the herbicide-induced AtPM24-GST from Arabidopsis thaliana, the anthocyanin-related 1AW9-GST from Petunia, AN9 from Petunia and BZ2 from maize. The active residues for the binding of GSH are indicated by arrows. The red box at amino acid 205 indicates the position of the Trp→Leu transition in TT19-4. Coloured highlights indicate amino acid identity among all five translated protein sequences.


The flavonoid pathway is responsible for red–blue pigments present in flowers and yellow-brown pigments in seed (Tian, Pang & Dixon 2008). Although knowledge of the pathway has advanced substantially since the function of the BANYULS gene (anthocyanidin reductase) was determined (Xie et al. 2003), the last steps in polymerization and transport are still poorly defined. The TT19 protein in Arabidopsis (Kitamura et al. 2004), together with its homologues An9 in petunia (Müller et al., 2000) and Bz2 in maize (Marrs et al. 1995; Goodman et al. 2004), are considered to be GST proteins involved in anthocyanin transport, and the TT19 protein has also been linked to proanthocyanidin accumulation (Kitamura et al. 2004). However, the full mechanism of action of flavonoid GSTs has not been proven.

In our study, we highlight a new induced Arabidopsis GST mutant allele, tt19-4, which has a novel phenotype that can assist in our understanding of the function of TT19. The mutated GST gene TT19-4 specifies wild-type levels of vegetative anthocyanin, a pale seed colour, reduced soluble seed PA, enhanced unextractable seed PA, and enhanced seed Q3R in mutant tt19-4 compared with the wild-type Col-4 phenotype. This new tt19-4 allele was similar to mutant tt19-1 in its pale seed colour, Q3R, and PA profile, but distinct in that tt19-4 has wild-type (i.e. lower) levels of Q3G7R and wild-type (i.e. higher) vegetative anthocyanins compared with tt19-1. Thus tt19-4 appears to have a mutant phenotype intermediate between Col-4 and tt19-1. Two other new alleles from the SK population, tt19-5 and tt19-6 also had reduced levels of vegetative anthocyanin and a pale seed colour similar to known tt19 alleles.

Sequence and genetic analysis showed that the TT19-4 gene has a one base pair change in the C-terminal; otherwise the sequence is identical to wild-type TT19 in the Arabidopsis genome. Molecular complementation of tt19-1 with the TT19-4 gene could recover the wild-type vegetative anthocyanin phenotype, but not the wild-type seed coat colour. Solution structure analysis of the herbicide-conjugating GST protein from Arabidopsis has suggested that GSH interacts with GST proteins in domain I at Lys41, Pro50, Gln53, Glu66, Ser67 and Arg68 (Reinemer et al. 1996). Our GST universal substrate enzyme activity and GFP fusion data confirm that the change in amino acid 205 in domain II of the C-terminal from TT19 (Trp) to TT19-4 (Leu) does not disrupt cytoplasmic localization nor its ability to glutathionylate.

Unextractable PA cannot be extracted with neutral solvents, but portions (subunits) can be released as anthocyanidins after acid treatment (Porter, Hrstich & Chan 1986). PA analysis supports our finding that pale seeds of tt19-1 and tt19-4 each contain substantially greater amounts of unextractable PA than Col-4, while soluble PA is decreased proportionately in these two mutants. This is consistent with a recent report on tt19-1, which also reported increased unextractable PA in tt19-1 seeds as well as proanthocyanidin localized in small vacuoles (Kitamura et al. 2004; 2010). Our findings, together with these reported data, suggest that the visible seed colour differences between the mutants and WT seed are due to changes in the proportion of unextractable and soluble PAs and that dark seed coat colour is a direct consequence of the presence of soluble PAs in a large central vacuole, not unextractable PAs.

The small unextractable seed fraction of Col-4 appears to be different in composition and structure compared with the large unextractable fraction present in tt19-4 and tt19-1. The Col-4 unextractable seed fraction releases a greater proportion of quercetin (potentially but not necessarily from biflavonoids) and is comparatively slow (difficult) to hydrolyse compared with the large unextractable fraction from both mutants. In turn, the unextractable fraction of tt19-1 released a greater proportion of anthocyanidin units (A512) with a different bimodal distribution pattern compared with tt19-4, and its major peak (A512) occurred later in time compared with its counterpart in tt19-4 seeds. The distinct kinetics of hydrolysis for the unextractable residue in these three lines suggest that unextractable PA may be more tightly bound (hence released more gradually) in Col-4 seeds compared with either mutant and in tt19-1 seeds compared with tt19-4 seeds. Proanthocyanidins are less extractable and bound more tightly with cell wall components when highly polymerized (Fournand et al. 2006). The characteristics of PA in the tt19 mutants may be due, at least in part, to PA accumulating in small vacuoles (as shown in Kitamura et al. 2010) that would contain more membrane than in the large vacuole of wild-type plants. This hypothesis could be tested by determining total vacuolar membrane volume and the titre of vacuolar membrane marker proteins in the tt19 mutants and by quantifying PA binding to vacuolar membrane in vitro.

Glutathione S-transferases are a family of multifunctional enzymes involved either in secondary metabolism accumulation or in the detoxification of diverse exogenous substrates by glutathione (GSH) conjugation (Dixon, Lapthorn & Edwards 2002). Earlier, it was suggested (but not proven) that anthocyanin is glutathionated by the TT19 homologue Bz9 maize (Marrs et al. 1995). Later studies could not confirm these results in a detailed analysis with the petunia homologue An9 (Müller et al., 2000). To date, no reports have actually presented clear evidence of flavonoid-GSH conjugates or anthocyanin-GSH conjugates arising from plant extracts or recombinant GST proteins.

In our hands, recombinant Arabidopsis TT19 and TT19-4 proteins each possess the ability to conjugate GSH to the universal substrate CNDB; definitely, both can be classified as GST-binding proteins based on this evidence. However, we were unable to detect any GSH conjugation products when TT19 and GSH were reacted together with a wide variety of natural flavonoids, proanthocyanidin dimers, certain phenolic acids and anthocyanins. The exception was sinapic acid and three coumaric acid isomers, which bound to the TT19 protein and could be released when the protein was heat-denaturated. Binding of coumaric acid and other phenylpropanoids to GSTs has been observed by others (Dean et al. 1995), but specific in vivo substrates for TT19 glutathionylation in Arabidopsis are still unknown. Since the small unextractable seed fraction of Col-4 seeds includes a greater proportion of quercetin than the equivalent fraction in tt19-1 or tt19-4 seeds and the extractable seed fractions of tt19-1 and tt19-4 have enhanced levels of Q3R and Q3G7R (tt19-1 only), we suggest that quercetin holds the key to understanding the function of TT19 and we highly doubt that cyanidin conjugates with glutathione in Arabidopsis. There may be an intermediate interacting with quercetin in the upper flavonoid pathway that is glutathionated or bound by TT19 and impacts vacuole formation. Analysis of vacuole ultrastructure and extractable and unextractable PA in the fah mutant, which also accumulates Q3R and is reduced in a sinapoyl anthocyanin (Huang et al. 2009), may reveal whether sinapic acid actually plays this role in TT19 function.

Q-PCR analysis of flavonoid genes for Col-4 and the two tt19 mutants was conducted to show differences in seedling anthocyanin gene expression, since the mutants differ in anthocyanin phenotypes. The patterns of many genes changed similarly for seedlings of both mutants compared with Col-4, i.e. ANS, DFR, F3OGT, A5MAT, and PAP1 from anthocyanin biosynthesis, as well as TT8, TT10, TDS6, and AHA10 from proanthocyanidin synthesis, even though seedlings do not accumulate proanthocyanidins. Other flavonoid genes were differentially expressed between these two mutants. These included F3H, F3'H, CHS, CHI, TTG1, and FLS1 (which were ∼1.40-fold higher in tt19-4 than in tt19-1); A5GT, 5MAT, TT2, TTG2 (which remained at WT levels in tt19-4 and were reduced in tt19-1); and TT19 (which was reduced in tt19-1 and two-fold greater than WT in tt19-4). Our seedling flavonoid expression data is the first reported for tt19 mutants and like the phytochemical data suggests changes related to flavonols. Nevertheless it is puzzling, since the TT19-1 and TT19-4 genes are alleles with lesions in a biosynthetic gene, not a regulatory gene, and expression levels differ for the two mutated genes. They suggest some form of post-translational control which is affected when TT19 is mutated and indicate a more complicated control of flavonoid biosynthesis than previously known.

Our results suggest that the Arabidopsis TT19-GST has two functional domains. The first is an N-terminal domain, which contains a GSH binding site, specifies anthocyanin and soluble proanthocyanidin accumulation and dark seed coat colour in the WT state, and which is fully functional in TT19-4 for anthocyanin accumulation. The second is a C-terminal domain specifying dark seed coat colour and soluble proanthocyanidin accumulation in the WT state and a change to pale seed colour, very little soluble PA, and substantial unextractable PA in the tt19-4 mutant state. Structural models and difficulty with protein solubilization provided us with evidence that the point mutation in the C-terminal of TT19-4 will likely change its 3D protein structure compared with the WT-TT19 protein. They suggest that the resulting Trp → Leu transition at amino acid 205 is at the root of the changes in phytochemical composition in the tt19-4 mutant line.

Alignment with other GSTs shows that amino acid position 205 is not necessarily conserved, but it lies within a small conserved motif. Among the aligned proteins, AN9 shows the same Trp in this position as TT19 and also could not conjugate anthocyanin with glutathione (Müller et al., 2000). This may indicate that TT19 and AN9 share a similar function distinct from other GSTs. Knockout of this amino acid in AN9, followed by ultrastructure and analysis of the resulting anthocyanin and PA phenotypes would provide an opportunity to determine its importance in PA synthesis and extractability. Moreover, the variation that exists at amino acid 205 for GST homologues in other species can be exploited to determine whether the proanthocyanidin-related function of the 3′ domain is universal or only relevant to Arabidopsis.


X. Li, P. Gao and L. Wu were recipients of Visiting Fellowships to a Government of Canada Laboratory. This work was partially supported by the Program for New Century Excellent Talents in University of Ministry of Education of the People's Republic of China (NCET-09-0423). We also thank the Arabidopsis Biological Resources Center (ABRC) at the Ohio State University for providing the T-DNA lines: SALK_118408 (At5g54950); CS89452 (At4g21630); SALK_016650 (At3g57280); SALK_011858; CS6172 (fah) and SALK_018120 (sct). Dr Kitamura is gratefully acknowledged for providing the tt19-1 mutant line.