Transcriptional regulation of anthocyanin biosynthesis in ripening fruits of grapevine under seasonal water deficit

Plant, Cell and Environment (2007) 30, 1381 1399 doi: 10.1111/j.1365-3040.2007.01716.x Transcriptional regulation of anthocyanin biosynthesis in ripening fruits of grapevine under seasonal water deficit SIMONE D. CASTELLARIN 1,2, ANTONELLA PFEIFFER 1, PAOLO SIVILOTTI 1, MIRKO DEGAN 1, ENRICO PETERLUNGER 1 & GABRIELE DI GASPERO 1,2 1 Dipartimento di Scienze Agrarie e Ambientali, University of Udine, via delle Scienze 208 and 2 Istituto di Genomica Applicata, Parco Scientifico e Tecnologico Luigi Danieli, via Jacopo Linussio 51, 33100 Udine, Italy ABSTRACT Anthocyanin biosynthesis is strongly up-regulated in ripening fruit of grapevines (Vitis vinifera L.) grown under drought conditions. We investigated the effects of long-term water deficit on the expression of genes coding for flavonoid and anthocyanin biosynthetic enzymes and related transcription factors, genes sensitive to endogenous [sugars, abscisic acid (ABA)] and environmental (light) stimuli connected to drought stress, and genes developmentally regulated in ripening berries. Total anthocyanin content has increased at harvest in water-stressed (WS) fruits by 37 57% in two consecutive years. At least 84% of the total variation in anthocyanin content was explained by the linear relationship between the integral of mrna accumulation of the specific anthocyanin biosynthetic gene UDPglucose : flavonoid 3-O-glucosyltransferase (UFGT) and metabolite content during time series from véraison through ripening. Chalcone synthase (CHS2, CHS3) and flavanone 3-hydroxylase (F3H) genes of the flavonoid pathway showed high correlation as well. Genes coding for flavonoid 3,5 -hydroxylase (F3 5 H) and O-methyltransferase (OMT) were also up-regulated in berries from dehydrated plants in which anthocyanin composition enriched in more hydroxylated and more methoxylated derivatives such as malvidin and peonidin, the grape anthocyanins to which human gastric bilitranslocase displays the highest affinity. The induction in WS plants of structural and regulatory genes of the flavonoid pathway and of genes that trigger brassinosteroid hormonal onset of maturation suggested that the interrelationships between developmental and environmental signalling pathways were magnified by water deficit which actively promoted fruit maturation and, in this context, anthocyanin biosynthesis. Key-words: abiotic stress; dehydration; drought tolerance; pigmentation; polyphenols; red wine. Abbreviations: ABA, abscisic acid; ACPK1, ABAstimulated calcium-dependent protein kinase1; BAN, anthocyanidin reductase; BR60X1, brassinosteroid-6- oxidase; BRI1, brassinosteroid insensitive 1; CHS1, CHS2, Correspondence: G. Di Gaspero. Fax: +39 0432 558603; e-mail: gabriele.digaspero@uniud.it CHS3, chalcone synthases; CT, control; CYTB5, cytochrome b5; DFR, dihydroflavonol reductase; DWF1, dwarf1; F3H, flavanone 3-hydroxylase; F3 H, flavonoid 3 -hydroxylase; F3 5 H, flavonoid 3,5 -hydroxylase; LDOX, leucoanthocyanidin dioxygenase; FLS1, flavonol synthase; GST, glutathione S-transferase; LAR1 and LAR2, leucoanthocyanidin reductases; NCED1 and NCED2, 9-cisepoxycarotenoid dioxygenases; n.s., not significant; OMT, O-methyltransferase; VvMSA, Vitis vinifera maturation-, stress-, ABA-induced protein; UFGT, UDP-glucose : flavonoid 3-O-glucosyltransferase; WS, water stressed. INTRODUCTION Consumption of anthocyanins and other polyphenols has health benefits associated with scavenging of free radicals and protective effects against cardiovascular diseases. Most red-pigmented fruits, freshly consumed or processed into food and beverages, provide humans with a major source of dietary anthocyanins. Among them, red grapes and wines are particularly rich in bioavailable anthocyanins that are rapidly absorbed as intact molecules (Bitsch et al. 2004) and delivered into the brain within minutes from their ingestion (Passamonti et al. 2005). Anthocyanin pigments redden the berry skin of grapevine cultivars as well as wines fermented in the presence of red skins. With regard to the wine industry, colour is crucial to quality for the production of premium red wines. The first sensorial contact to wine is usually made by visual inspection which starts building up the consumer s perceived quality. This translates wine colour into a value of commercial and economic relevance. Anthocyanins are synthesized via the flavonoid pathway in grapevine cultivars that harbour the wild-type VvmybA1 transcription factor for the expression of UFGT (Kobayashi, Goto-Yamamoto & Hirochika 2004). The encoded enzyme UFGT catalyses the glycosylation of unstable anthocyanidin aglycones into pigmented anthocyanins (Fig. 1). Two primary anthocyanins (cyanidin and delphinidin) are synthesized in the cytosol of berry epidermal cells. Cyanidin has a B-ring di-hydroxylated at the 3 and 4 positions, whereas delphinidin has a tri-hydroxylated B-ring because of an additional hydroxyl group at the 5 position. Flavonoid precursors are initially recruited from the phenylpropanoid pathway by a small family of chalcone synthases (CHS1, CHS2, CHS3) and enter the flavonoid Journal compilation 2007 Blackwell Publishing Ltd 1381

1382 S. D. Castellarin et al. Figure 1. Key steps of the flavonoid pathway leading to anthocyanin biosynthesis. Transcripts of genes coding for all enzymes reported in this picture were analysed in this study. Dashed arrows indicate steps not considered in this study or steps for which the genetic control has not been elucidated in grape. Acronyms of the compounds reported in the picture stand for the following: E, eriodictyol; Nf, naringenin flavanone; Phf, pentahydroxyflavanone; Dhq, dihydroquercetin; Dhk, dihydrokaempferol; Dhm, dihydromyricetin. pathway. Parallel pathways downstream of F3 H and F3 5 H (Bogs et al. 2006; Castellarin et al. 2006) produce either cyanidin or delphinidin. The 3 position of cyanidin and delphinidin and sequentially the 5 position of delphinidin can be methoxylated by OMT that generate peonidin, petunidin and malvidin, respectively. Anthocyanins are delivered into the vacuole where they are visible as coloured coalescences (anthocyanic vacuolar inclusions). It still remains unknown whether anthocyanins enter the vacuole as single molecules and thereafter they aggregate or if cytoplasmic vesicles containing coalesced anthocyanins interact with the tonoplast (Zhang et al. 2006). Whatever the mechanism, some members of the GST protein family are believed to participate in vacuolar trafficking and sequestration of anthocyanins (Marrs et al. 1995; Mueller et al. 2000). Among them, a member of the GST gene family that showed overlapping patterns of expression with anthocyanin accumulation was identified by Ageorges et al. (2006). Total amount of anthocyanins and the relative abundance of single anthocyanins are extremely variable among red- to blue-skinned cultivars. Both traits are under genetic control and are developmentally regulated. However, the production of grapevine secondary metabolites, including anthocyanins, can be impaired or magnified by abiotic stresses such as drought as well as extremes of light exposure and temperature. Physiological studies show that incident radiation on bunches (Downey, Harvey & Robinson 2004; Cortell & Kennedy 2006), plant water status (Esteban, Villanueva & Lissarrague 2001) and exogenous hormones (Roubelakis-Angelakis & Kliewer 1986; Hiratsuka et al. 2001) modify anthocyanin content, expanding the range of metabolite variation beyond that due to the genetic

Water deficit and anthocyanin biosynthesis in grapevine 1383 background of a given cultivar. Growing grapevines under restricted water supply has long been regarded as an agronomic tool for improving polyphenol content in berries. Although the concentration of anthocyanins and other phenolics has consistently increased in response to water deficits (Matthews & Anderson 1988; Ojeda et al. 2002; Roby et al. 2004; and others), it is unclear whether this improvement is only due to inhibited berry growth (see e.g. Kennedy, Matthews & Waterhouse 2002), water loss and concentration of solutes or whether water stress actively triggers biosynthesis of phenolics. In particular, little is known about whether or not, and if so, to what extent the expression of anthocyanin biosynthetic genes is affected by seasonal water availability throughout the progress of ripening. Therefore, we imposed water deficits and evaluated the effect of plant water status on (1) total anthocyanin content, (2) anthocyanin composition and (3) the transcriptional regulation of genes of the flavonoid pathway in berry skin and of key genes of drought-, ABA-, sugar- and brassinosteroid-related signal transduction pathways. The experiment reproduced field conditions of moderate dehydration where vines of drought plots were maintained at a water potential compatible with the yield of a commercial crop. By monitoring flavonoid accumulation and anthocyanin profile at eight stages of berry development from prevéraison until harvest in the seasons 2004 2005 and by comparing metabolite data with transcript levels of 29 genes, we wanted to link transcriptional and metabolite changes in fruits that experienced long-term water deficit. The aims were to unveil the transcriptional regulation of anthocyanin biosynthesis under progressive water starvation and to show how secondary metabolism reacts in response to environmental stress conditions in reproductive organs of a woody perennial. MATERIALS AND METHODS Field plots and physiological measurements Field experiments were conducted in a 10-year-old vineyard of Vitis vinifera Merlot grafted on SO4, with planting density of 4000 vines ha -1 in north-eastern Italy (46 02 N, 13 13 E; 88 m a.s.l.). Vines were trained to spurred cordon system and shoots were positioned upwards. Prior to application of controlled water regime, canopy was pruned to 20 shoots per plant and crop load was normalized to 30 bunches per plant. Every treatment was replicated on four plots of 12 vines each, arranged in a completely randomized design. Sampling was initially scheduled at 2 week intervals to span berry development from 6 weeks after blooming until harvest (Supplementary Fig. S1). At véraison, sampling schedule was adapted to target the stages of 25% coloured berries in 2004, 10 and 50% coloured berries in 2005 and 100% coloured berries in both seasons. Sampling schedule in 2005 spanned the same developmental stages as in 2004, but sampling was more frequent at véraison.during the phase of colour transition, red berries on turning bunches were sampled separately from green berries on the same bunches. Samples of 40 berries each were collected from every plot. Water was supplied by an underground drip system with emitters at 2.5 L m -2 h -1 flow rate. Plant water status was monitored weekly by measuring stem water potential (Y Stem) with a pressure chamber.two leaves per plot were covered on each side of the row with aluminium foil-coated plastic bags at midday for 1 h, in order to allow stem and leaf water potential to equilibrate. Then, leaves were removed and Y Stem was measured according to Scholander et al. (1965). Vines were maintained under a controlled water regime from 15 July in the season 2004 and from 1 July 2005 until harvest by sheltering the rows under a tunnel covered by polyethylene film. CT plants were irrigated approximately once a week according to their water status and kept at a Y Stem between -0.2 and -0.6 MPa. WS plants were left to dehydrate until Y Stem dropped down to -1.2 to -1.4 MPa. Then water stress was partially released by supplying 12 L m -2 water on 13 September 2004 and 6Lm -2 water on 24 August 2005 in order to prevent vines from yield loss. Seasonal water status was expressed by the water stress integral as reported by Myers (1988). Global radiation and air temperature were measured 2 m above ground by a meteorological station positioned at the experimental site. Evolution of canopy architecture was monitored by measuring the point quadrat (Smart & Robinson 1991; Poni et al. 1996). Spatial relationships between bunches and foliage were expressed as leaf layer number (LLN) and percentage of exposed clusters (% EC). Grape phenology overlapped between the two experimental seasons from blooming, which occurred during the first decade of June, to the beginning of July. From the first sampling date (15 July both years, at a mean berry weight of 0.65 and 0.73 g in 2004 and 2005, respectively) to mid véraison, fruit development advanced by 5 6 d in 2005. In 2004, véraison commenced on 10 August and was completed by 24 August (Supplementary Fig. S1). In 2005, véraison commenced on 5 August and lasted until 23 August. At harvest, mean berry weight in CT plants was 1.83 and 1.70 g in 2004 and 2005, respectively, with no substantial difference in the concentration of soluble solids between the years. Transcript profiling Total RNA was extracted from berry skin following the procedure described in Moser et al. (2004) and was treated with 0.5 U mg -1 RQ1 DNase (Promega, Milan, Italy). Firststrand cdna was synthesized using 2 mg of RNA,0.5 mm (dt) 18 primer and 50 U of M-MLV reverse transcriptase (Promega). Quantitative RT-PCR was carried out on a DNA Engine Opticon2 (MJ Research, Waltham, MA, USA). Each reaction (20 ml) contained 200 nm of each primer, 5 ml of diluted cdna, 0.4 U of HotMaster Taq polymerase (Eppendorf, Milan, Italy), 4.0 mm magnesium acetate, 0.4 mm dntps and SYBR solution (Eppendorf). Thermal cycling conditions were 95 C for 3 min followed by 94 C for 15 s, 56 58 C for 20 s and 65 C for 30 s for 40 cycles, followed by a melting cycle from 65 to 95 C. Each

1384 S. D. Castellarin et al. cdna sample was analysed at two different dilutions (1:60 and 1:240 of the original cdna), each dilution run in duplicate. Gene transcripts were quantified upon normalization to ubiquitin-conjugating factor (UbiCF, Supplementary Table S1) by comparing the cycle threshold (C T) of the target gene with that of UbiCF. Primers pairs for CHS1, CHS2, CHS3, DFR, LDOX and UFGT were retrieved from literature (Goto-Yamamoto et al. 2002); primers for FLS1 were from Downey, Harvey & Robinson (2003); for F3 H and F3 5 H from Castellarin et al. (2006); for BR60X1, DWF1 and BRI1 from Symons et al. (2006); for LAR1 and LAR2 from Bogs et al. (2005); for CYTB5 from Bogs et al. (2006); for GST from Terrier et al. (2005); for the grapevine rd22 homolog from Hausmann et al. (2003). Primers for F3H (Sparvoli et al. 1994), BAN (Tanner et al. 2003), OMT (GenBank accession no. BQ796057), VvMSA (Cakir et al. 2003), MybA1, MybB, MybC and MybD (Kobayashi et al. 2002, 2004), Myb5a (Deluc et al. 2006), NCED1 and NCED2 (Soar et al. 2004), and ACPK1 (Yu et al. 2006) were newly designed on the original DNA sequences to amplify 150 250 bp gene fragments (Supplementary Table S1). Metabolite profiling Soluble solids were quantified with a refractometer and expressed as Brix. Titratable acidity was expressed as tartaric acid equivalents. Skin phenols were extracted for 4 h in 1/10 (w/v) skin/solvent suspension acidified methanol (1% HCl v/v) then centrifuged at 3500 g for 15 min and filtered with a 0.2 mm polytetrafluoroethylene (PTFE) filter (Chemtek Analytica, Anzola Emilia, Italy). Total flavan- 3-ols (catechins) were quantified as reported in Zironi, Buiatti & Celotti (1992). The calibration curve was constructed with (+)-catechin (Extrasynthase, Genay, France). Absorbance was measured by a UV/VIS LAMBDA 2 spectrometer (PerkinElmer, Milan, Italy). Anthocyanins were extracted for 4 h from 200 mg of berry skin with 2 ml of methanol, then centrifuged and filtered with a 0.2 mm PTFE filter (Chemtek Analytica). Methanol was evaporated under nitrogen flux, and anthocyanins were re-suspended with 100 400 ml of 27:73 methanol : perchloric acid 0.3% (v/v). Anthocyanins were separated in a high-performance liquid chromatograph (PerkinElmer Series 4) using a C18 Purospher RP-18 (5 mm, 250 4 mm) column (Merck, Darmstadt, Germany), according to the procedure reported by Mattivi et al. (2006), and were detected at 520 nm by an LC-95 spectrophotometric detector (PerkinElmer). Total anthocyanins were expressed as malvidin 3-glucoside equivalents and included monoglucoside, acetyl-glucoside and p-coumaroyl-glucoside anthocyanins. The composition of monoglucoside anthocyanins was used for calculating the proportion of 3,4,5 -/3,4 -hydroxylated and of methoxylated/non-methoxylated derivatives. Berry colour was measured with an X-Rite 948 Chromameter (X-Rite, Origgio, Italy) and averaged over 160 berries at each sampling date. Colorimetric specification was referenced to the CIELab scale. Statistical analysis Differences between the treatments were tested for significance by applying analysis of variance (anova). The probability levels used were P 0.05 (*), P 0.01 (**) and P 0.001 (***). Statistical analyses were run using the COSTAT statistical package (CoHort Software, Monterey, CA, USA). For testing correlation between transcriptional profiles of anthocyanin genes and metabolite data, we calculated the cumulative transcription of each gene from the onset of véraison to every stage of ripening as the area below the expression curve. We plotted the cumulative expression at every sampling date (x) against the amount of metabolites accumulated at the same time (Y), and we calculated linear regression between Y and x using biological replicates from two experimental seasons. Regression between climate parameters, such as radiation, day night temperature fluctuations, growing degree days (GDD) and gene expression in CT fruits from the onset of véraison until every ripening stage was also calculated over the two experimental seasons. Based on the coefficient of correlation (r), correlation greater than 0.9 was described hereafter as strong, whereas a correlation less than 0.6 was described as weak.the percentage of the total variation in metabolite content explained by the transcript-metabolite linear relationship was calculated based on the coefficient of determination (R 2 ). RESULTS Plant water status and physiological responses to drying soil Plant water status differed significantly in CT and WS plants throughout the whole course of berry maturation (Fig. 2). Y Stem of CT vines was maintained consistently between -0.2 and -0.6 MPa throughout berry development in both seasons. In 2004, Y Stem in drought plots significantly diverged from controls since 30 July, 15 d after water withdrawal and 11 d before the onset of véraison. In 2005, water supply was interrupted earlier on 1 July, and plant water status significantly diverged between the treatments since 15 July, 21 d before the onset of véraison. During the course of ripening, in 2004, Y Stem constantly decreased in WS plants down to a minimum of -1.35 MPa on 13 September 2004, when water stress was temporarily released by an irrigation that allowed vines to complete maturation 2 weeks later at Y Stem of -1.0 MPa. In 2005, Y Stem of WS vines dropped down faster. Dehydrated vines were irrigated on 24 August, at the time when Y Stem was -1.2 MPa and vines had just completed véraison. This water supply allowed plants to ripen berries at a Y Stem of -1.4 MPa 1 month later. Canopy structure was progressively affected through the course of ripening as the exposure to water shortage progressed. Stratification of foliage was reduced under drought conditions in both years (Supplementary Fig. S2). LLN was significantly lower in WS plants at harvest in 2004 and since the last 2 weeks of ripening in 2005. The difference was greater in 2005, when vines experienced a more intense water deficit in the last part of the season.as a result of that,

Water deficit and anthocyanin biosynthesis in grapevine 1385 Figure 2. Plant water status of vines subjected to differential water regimes during summer 2004 and 2005 in WS and CT vines. Water status was expressed as stem water potential. Bars represent SE. Asterisks indicate significant differences between the treatments at the same date using the Student Newman Keuls test (*P 0.05, **P 0.01, ***P 0.001). White background represents pre-véraison; light grey background represents the phase of bunch turning colour (véraison) from 10 to 100% of coloured berries; deep grey background represents the stage of post-véraison maturation until harvest. Dashed boxes highlight the period when water deficit was partially released in drought plots. the percentage of sunlight-exposed clusters increased in WS vines in both years, but differences were significant only at harvest in the season 2005. The yield per vine was significantly reduced by drought in both seasons, but technological parameters of juice quality were not significantly affected (Table 1). The yield of the crop, if projected to the surface unit of a commercial vineyard, decreased from 18.4 to 15.0 tons ha -1 in the season 2004 and from 16.4 to 12.9 tons ha -1 in the season 2005. Because the number of clusters per vine was normalized in each treatment by cluster thinning prior to the application of drought and no fruit abscission occurred later on, yield reduction in WS plants was exclusively due to lower berry weight (Fig. 3). Curves of sugar accumulation and total acidity overlapped throughout ripening in CT and WS berries (Fig. 3). The observed range of ph variation (0.15 and 0.09 in 2004 and 2005, respectively) was negligible in absolute terms but was statistically significant between the treatments at harvest. Grape phenology suggests that water deficit has slightly advanced anthocyanin biosynthesis. In 2004, on 12 August when sampling at approximately 25% véraison was scheduled, 18% berries actually turned red in CT plants and 34% inws plants.in 2005,on 5August,6.4% berries turned red in CT plants and 10% in WS plants.at the sampling date as of 11 August, which was aimed at pointing 40% véraison, 33.5% berries actually turned red in CT plants and 48% in WS plants. Two genes of the brassinosteroid synthetic pathway (BR60X1 and DWF1) and one gene coding for a brassinosteroid receptor (BRI1), which are implicated in hormonal control of ripening (Symons et al. 2006), were used as molecular markers of the progress of ripening (Fig. 4). DWF1 was up-regulated inws vines throughout véraison.transcripts of BR6OX1, which is under a feedback negative regulation by brassinosteroids, as well as of the grapevine homolog of brassinosteroid receptor insensitive BRI1 were lower in WS plants just before and at the beginning of véraison in both years. Altogether, these genes required for biosynthesis and sensing of brassinosteroids suggest a transcriptional induction of ripening-related processes in WS fruits. All of the genes tested in this paper were expressed in berry skin at some or all developmental stages from 0.6 0.7 g berry weight until full maturation. Genes CHS1, CHS2, CHS3, F3H, F3 H, LDOX, DFR, CytB5, four Mybtype transcription factors (MybB, MybC, MybD and Myb5a) and drought-, ABA- and sugar-related genes were expressed throughout the sampling period. Among them, genes of the basal flavonoid pathway showed high expression levels at the first sampling date in both years and treatments (Fig. 5). At that phase of berry development (3 4 weeks before véraison), berries had not yet entered the lag phase of expression of flavonoid genes which spans approximately 1 2 weeks before véraison (Boss, Davies & Robinson 1996). A set of genes more strictly committed to the synthesis (UFGT and MybA), modification (F3 5 H and OMT) and storage (GST) of anthocyanins was steadily induced at véraison. Flavonol and proanthocyanidin genes were, in general, weakly expressed. Transcripts of FLS1 slightly increased from véraison to harvest, unlike LAR1, LAR2 and BAN, which decreased as ripening progressed. Transcriptional changes of flavonoid and ripening-related genes were also examined at the onset of véraison in asynchronous red and green berries sampled on the same bunch. Most genes of the anthocyanin pathway were up-regulated in coloured berry skin and, among them, CHS3, GST, UFGT, OMT, F3 5 H and CYTB5 showed the most remarkable induction (Table 2). Anthocyanin accumulation and related transcriptional changes in WS fruits Anthocyanin content was higher in berry skin of WS vines in both years (Fig. 6). Differences between WS and CT plants were significant since the commencement of colouring through all sample dates in summer 2004. In 2005, while

1386 S. D. Castellarin et al. Yield Soluble solids Titratable acidity (kg vine -1 ) (kgm -2 ) ( Brix) (g L -1 ) a ph Table 1. Yield and juice composition at harvest of grapes from WS and CT vines in the seasons 2004 and 2005 CT 2004 4.59 1.83 21.5 6.7 3.35 WS 2004 3.75 1.50 21.0 7.1 3.20 * * n.s. n.s. ** CT 2005 4.10 1.64 20.2 5.5 3.30 WS 2005 3.22 1.29 19.6 5.1 3.21 * * n.s. n.s. *** Asterisks indicate significant differences between the treatments using the Student Newman Keuls test (*P 0.05, **P 0.01, ***P 0.001). a Expressed as tartaric acid equivalents. WS and CT vines did not differ significantly until the end of véraison, from that moment onwards, anthocyanin accumulation was abruptly induced in WS vines. In vintages 2004 and 2005, grapes harvested from WS vines accumulated 37 and 57% more anthocyanins, respectively, than those from CT plants, if anthocyanin content was expressed as concentration in fresh tissues (mg g -1 of skin fresh weight). We calculated also the amount of anthocyanins in full ripe grapes on per organ basis (mg berry -1 ) and per plant basis. Average berry content of anthocyanins was 4.1 mg in WS plants versus 3.3 mg in CT plants in 2004, and 1.7 mg in WS plants versus 1.3 mg in CT plants in 2005. The average amount of grape anthocyanins synthesized per plant was 9.5 g in WS plants versus 8.3 g in CT plants in 2004 and 4.0 g in WS plants versus 3.1 g in CT plants in 2005.A correlation was found between the cumulative water stress, expressed as the summation of Y Stem (S Y) the vines experienced from 2 weeks before véraison on, and the steady-state anthocyanin content (Fig. 6). Anthocyanin content was linearly correlated with S Y in (2004) (R 2 = 0.80) and in 2005 (R 2 = 0.82). Most anthocyanin genes were up-regulated in WS plants. The expression profile of UFGT, the anthocyanin-specific biosynthetic gene, was higher in WS fruits than CT fruits at all sample dates in both seasons (Fig. 6). The expression (a) (b) Figure 3. Evolution of berry weight (a) and soluble solids and titratable acidity (b) in WS and CT vines in the seasons 2004 and 2005. Bars represent SE. Asterisks indicate significant differences between the treatments at the same date using the Student Newman Keuls test (*P 0.05, **P 0.01, ***P 0.001). White background represents pre-véraison; light grey background represents the phase of bunch turning colour (véraison) from 10 to 100% of coloured berries; deep grey background represents the stage of post-véraison maturation until harvest. T.a., Titratable acidity; S.s., soluble solids.

Water deficit and anthocyanin biosynthesis in grapevine 1387 Figure 4. Expression of the genes DWF1, BR6OX1 and BRI1 related to the synthesis and sensing of brassinosteroids in berry skin of WS and CT vines in the seasons 2004 and 2005. White background represents pre-véraison; light grey background represents the phase of bunch turning colour (véraison) from 10 to 100% of coloured berries; deep grey background represents the stage of post-véraison maturation until harvest. profile of all other anthocyanin biosynthetic and storage genes is reported in Fig. 5. Genes CHS1 and CHS2 showed similar expression patterns in WS and CT plants during ripening in 2004, while CHS3 was transiently up-regulated in WS vines at the beginning of véraison. In the season 2005, transcript profiles of CHS2 and CHS3 were constantly higher under drought conditions, while transcripts of CHS1 were higher from post-véraison until harvest. The gene F3H showed an expression pattern similar to that of CHS2, with slight differences between treatments in 2004 and a constantly higher expression under drought conditions in 2005. Transcription of DFR and LDOX was promoted by water deficit at the beginning of véraison in both years, but as ripening progressed, these differences were not consistently maintained. The difference between treatments was confirmed in 2005 when expression of DFR and LDOX remained higher in WS plants from véraison until harvest at all but one sample date. Expression of GST showed a profile that resembled that of UFGT, with an up-regulation in WS berries at all but one sample date over 2 years. Among the transcription factors (Fig. 7), transcripts of MybA were constantly more abundant in WS vines. Differences in gene expression were maintained at all ripening stages in both years. Expression of Myb5a was higher in WS fruits during véraison in both years then declined to barely detectable levels in both treatments, while MybC was constantly up-regulated in WS fruits from post-véraison until harvest in both years. Expression profiles of MybB and MybD did not show a consistent difference in CT and WS plants among the years.

1388 S. D. Castellarin et al. Figure 5. Expression of genes of the basal flavonoid pathway (CHS1, CHS2, CHS3, F3H, CYTB5, DFR, LDOX) and a GST in berry skin of WS and CT vines in the seasons 2004 and 2005. White background represents pre-véraison; light grey background represents the phase of bunch turning colour (véraison) from 10 to 100% of coloured berries; deep grey background represents the stage of post-véraison maturation until harvest.

Water deficit and anthocyanin biosynthesis in grapevine 1389 Figure 5. Continued The cumulative transcription of each gene from the onset of anthocyanin biosynthesis until every stage of ripening was calculated as the area below the expression curve (x) and plotted against the amount of anthocyanins accumulated at the same time (Y). We calculated linear regression between Y and x over the treatments and the experimental seasons (Table 2). The integral of the expression of UFGT, CHS2, CHS3 and F3H through the time series correlated strongly (r > 0.9) with anthocyanin content and explained a large part of the increase in anthocyanin accumulation in WS fruits in both years (R 2 = 0.82 0.95). Cumulative expression of a couple of genes (CHS1 and MybC) explained the observed variation to a smaller extent (R 2 = 0.66), while DFR, LDOX, GST and four Mybs showed a weak correlation with anthocyanin content among the seasons (R 2 = 0.01 0.43), even though a couple of them (MybA and GST) were constantly up-regulated in WS berries in either season. Anthocyanin profile and transcriptional changes of F3 5 H and OMT in WS fruits The relative abundance of 3,4,5 -hydroxylated (delphinidin, petunidin, malvidin) to 3,4 -hydroxylated (cyanidin

1390 S. D. Castellarin et al. Gene Ratio of gene expression in red to green berries Linear regression (R 2 ) Total anthocyanins Radiation CHS1 5.7 0.66 0.96 0.97 CHS2 14.2 0.95 0.40 0.39 CHS3 39.9 0.94 0.78 0.76 F3H 28.8 0.82 0.27 0.27 F3 H 11.1 0.43 0.69 0.68 F3 5 H 74.1 0.74 0.93 0.89 CYTB5 51.6 0.50 0.92 0.51 DFR 7.3 0.09 0.32 0.31 LDOX 12.1 0.40 0.80 0.77 UFGT 46.6 0.84 0.91 0.88 GST 146.6 0.24 0.61 0.59 OMT 38.4 0.92 0.44 0.43 MybA 13.9 0.34 0.71 0.69 MybB 6.0 0.03 0.28 0.23 MybC 4.9 0.66 0.93 0.90 MybD 0.8 0.11 0.24 0.23 Myb5a 1.6 0.01 0.10 0.70 Day night temperature fluctuation Table 2. Transcript abundance of anthocyanin genes in red and green berries collected on the same bunch at the beginning of cluster pigmentation, correlation between the integral of gene expression during time series throughout ripening (x) and total anthocyanin content (Y) as well as between climate parameters (x) and cumulative transcript level throughout the course of ripening (Y) The percentage of the total variation of Y explained by x was calculated based on the coefficient of determination (R 2 ). All data were calculated over two experimental seasons. and peonidin) monoglucoside anthocyanins was compared between the two treatments. The percentage of trihydroxylated anthocyanins was higher in WS fruits in both years. Because total anthocyanin content was also higher in WS fruits, the accumulation of di- and tri-hydroxylated anthocyanins was also compared in terms of amount per berry (Fig. 8). At harvest 2004, content of di-hydroxylated anthocyanins was 0.55 mg berry -1 in both CT andws grapes, while content of tri-hydroxylated anthocyanins increased from 1.70 mg berry -1 in CT plants to 2.16 mg berry -1 in WS plants. In the following year, content of di-hydroxylated anthocyanins was 0.38 and 0.37 mg berry -1 in CT and WS vines, respectively, while tri-hydroxylated anthocyanins increased from 0.53 mg berry -1 in CT plants to 0.78 mg berry -1 in WS plants. The increase of trihydroxylated anthocyanins in WS fruits was statistically significant in both years. Transcription of F3 5 H was up-regulated under drought conditions since the completion of véraison onwards, peaking up in concomitance with the increased biosynthesis of 3 4 5 -OH anthocyanins in WS plants (Fig. 8). The expression of F3 H did not show a clear pattern associated with either treatment in 2004 when also the pattern of accumulation of 3 4 -OH anthocyanins overlapped in CT and WS plants (Fig. 8). In 2005, expression of F3 H was down-regulated in WS plants throughout the phase of colour transition. From full véraison onwards, F3 H was up-regulated by drought.the pattern of accumulation of 3 4 -OH anthocyanins varied accordingly. Altogether, the ratio of transcript levels of F3 5 H to F3 H was consistently higher in WS berries throughout the period of anthocyanin biosynthesis (Fig. 8). We also compared anthocyanin composition between treatments in terms of level of methoxylation (Fig. 9). The proportion of the cyanidin-derived peonidin to its nonmethoxylated precursor was constantly higher in WS fruits during the course of ripening in both years. Similarly, the proportion of the delphinidin-derived di-methoxylated malvidin to its non-methoxylated precursor increased in WS fruits. Differences were more remarkable and statistically significant in 2005 when water deficit was more intense. The mono-methoxylated form of delphinidin, petunidin, showed overlapping patterns of contribution to the anthocyanin profile in CT and WS vines in both years. The pattern of OMT gene expression was continuously higher in WS plants throughout maturation, consistently with a more intense anthocyanin methoxylation.the largest differences between CT and WS fruits were observed in season 2005. Colour of berry skin turned darker in WS plants in accordance with an increased contribution of more hydroxylated and more methoxylated anthocyanins (Supplementary Table S2). Using the CIELab scale, we found significant decreases of the a* value in WS plants, which means a departure from red magenta, and significant decreases of the b* value, which means shifting towards blue. Transcriptional control of branching points of the flavonoid pathway leading to the synthesis of catechins and flavonols Vines under water stress did not show significant modifications of the amount of skin catechins (Supplementary Fig. S3), even though catechin biosynthetic enzymes (LAR1, LAR2 and BAN) might compete with the anthocyanin enzymes LDOX and UFGT for common precursors. At transcriptional level (Supplementary Fig. S4), the magnitude of expression of LAR2 was much higher than LAR1

Water deficit and anthocyanin biosynthesis in grapevine 1391 (a) (b) R 2 = 0.80 y = 2.94x 19.26 R 2 = 0.80 y = 0.56x 1.75 R 2 = 0.82 y = 0.29x 3.85 R 2 = 0.82 y = 0.70x 4.25 (c) Figure 6. (a) Accumulation of anthocyanins in berry skin of WS and CT vines in the seasons 2004 and 2005; (b) correlation between long-term exposition to water shortage, expressed as integral of stem water potential calculated at each sampling date during the period of sampling, and steady-state anthocyanin content in each plot; (c) mrna abundance of UFGT. Bars represent SE. Asterisks indicate significant differences between the treatments at the same date using the Student-Newman-Keuls test (*P 0.05, **P 0.01, ***P 0.001). White background represents pre-véraison; light grey background represents the phase of bunch turning colour (véraison) from 10 to 100% of coloured berries; deep grey background represents the stage of post-véraison maturation until harvest. before the inception of anthocyanin biosynthesis. The expression of LAR2 levelled off in bunches that were turning red, and it was insensitive to water regimes. Transcript levels of LAR1 were low throughout maturation but were slightly higher in WS plants at all sample dates and in both years. Expression of FLS1 correlated strongly (r = 0.91) with water stress integral between treatments and was consistent among seasons. FLS1 was constantly even though slightly up-regulated in WS plants in 2005, and differences between WS and CT plants were more evident in the last 4 weeks of ripening, when flavonol biosynthesis was more intense (Downey et al. 2004). When plotted versus total radiation during the ripening period and among the two seasons, FLS1 expression was linearly correlated with total radiation (r = 0.88). Expression of genes involved in drought-, ABA- and sugar-signal transduction pathways Genes NCED1 and NCED2 were transiently up-regulated in WS fruits at the very onset of véraison in both years (Supplementary Fig. S5). However, expression of ACPK1,

1392 S. D. Castellarin et al. Figure 7. Expression of R2R3-Myb type transcription factors in berry skin of WS and CT vines in the seasons 2004 and 2005. White background represents pre-véraison; light grey background represents the phase of bunch turning colour (véraison) from10to100% of coloured berries; deep grey background represents the stage of post-véraison maturation until harvest.

Water deficit and anthocyanin biosynthesis in grapevine 1393 (a) (b) (c) (d) Figure 8. Evolution of anthocyanin profile of di- and tri-hydroxylated anthocyanins (a) in berry skin of WS and CT vines in the seasons 2004 and 2005; transcript abundance of (b) F3 5 H, of (c) F3 H and (d) transcript ratio of the two flavonoid hydroxylase genes. Content of monoglucoside 3,4 -hydroxylated anthocyanins (3,4 -OH), monoglucoside 3,4,5 -hydroxylated anthocyanins (3,4,5 -OH) and total anthocyanin content (Tot anth) are expressed on per berry basis. Bars represent SE. Asterisks indicate significant differences between the treatments at the same date using the Student Newman Keuls test (*P 0.05, **P 0.01, ***P 0.001). White background represents pre-véraison; light grey background represents the phase of bunch turning colour (véraison) from 10 to 100% of coloured berries; deep grey background represents the stage of post-véraison maturation until harvest. Tot anth, total anthocyanins; anth, anthocyanins.

1394 S. D. Castellarin et al. (a) (b) Figure 9. (a) Methoxylation of primary anthocyanins (cyanidin and delphinidin) in berry skin of WS and CT vines in the seasons 2004 and 2005; (b) transcript abundance of OMT. Percentage of mono-methoxylated form of cyanidin (peonidin, peo) is referred to the y-axis on the right. Percentage of mono-methoxylated form of delphinidin (petunidin, pet) and di-methoxylated form of delphinidin (malvidin, mal) are referred to the y-axis on the left. White background represents pre-véraison; light grey background represents the phase of bunch turning colour (véraison) from 10 to 100% of coloured berries; deep grey background represents the stage of post-véraison maturation until harvest. di-, di-substituted anthocyanins; tri-, tri-substituted anthocyanins. which positively correlated with ABA content in grape berries and behaved as a key component of the ABA-signalling pathway in other experiments (Yu et al. 2006), was not consistently enhanced in WS fruits. The gene rd22, which responded to dehydration stress in an ABA-mediated manner in other plants (Yamaguchi- Shinozaki & Shinozaki 1993), was up-regulated in WS vines at most sample dates in 2004, but this trend was not confirmed in 2005. In addition, the expression of the gene VvMSA, which was induced by sugars and synergistically modulated by ABA (Cakir et al. 2003), was triggered in WS fruits only in the late season. None of these genes correlated well with plant water status at the seasonal time scale, when water stress integral was plotted versus cumulative expression of each gene at every ripening stage. Linear correlation calculated over two experimental seasons ranged from r = 0.41 to r = 0.55 for NCED1, VvMSA, ACPK1 and rd22, while it was a bit higher (r = 0.81) for NCED2. DISCUSSION Modulation of anthocyanin biosynthesis by long-term water deficit Water stress deeply modified secondary metabolism and transcriptional reprogramming of anthocyanin genes in ripening red berries. WS fruits accumulated more anthocyanins than CT fruits on per gram of fresh skin basis but also on per berry basis and per plant basis. The higher content of anthocyanins in WS berries was therefore not only due to reduction in berry growth, partial water loss or concentration of dry matter. Anthocyanin biosynthesis was actively induced under drought conditions. The majority of genes committed to the flavonoid pathway showed patterns of increased transcript accumulation in WS plants. The expression of the following genes was investigated by quantitative RT-PCR. Expression of genes CHS1, CHS2, CHS3, F3H, LDOX and DFR was determined to monitor the overall activity of the flavonoid pathway. The corresponding enzymes produce precursors for the synthesis of flavonols, catechins, proanthocyanins and anthocyanins. Expression of UFGT and GST was measured because these genes are strictly required for anthocyanin biosynthesis and storage. Transcripts of two flavonoid hydroxylases (F3 H and F3 5 H, Castellarin et al. 2006) and CYTB5, which is a candidate for modulating F3 H and F3 5 H activity (Bogs et al. 2006), were analysed to monitor gene regulation at the branching point of the pathway that addresses precursors to the parallel synthesis of di- or tri-substituted anthocyanins. Transcripts of OMT were quantified to assess whether an association does exist between the expression of this gene and the degree of anthocyanin methoxylation. The expression of five Myb transcription factors was measured to assay the role of regulatory genes in modulating flavonoid and anthocyanin biosynthesis. The expression of FLS1 was determined to monitor flavonol pathway activity, as well as genes BAN, LAR1 and LAR2 for the activity of catechin/

Water deficit and anthocyanin biosynthesis in grapevine 1395 proanthocyanin pathway. All structural and regulatory genes tested were selected based on literature reports. Selected genes had to match the criteria that they were involved to some extent in flavonoid biosynthesis and they were specifically expressed in anthocyanin-pigmented fruit tissues at some stages of ripening. Water deficit magnified the expression of genes UFGT, CHS2, CHS3 and F3H which explained well the increase of anthocyanin content in WS fruits during both experimental seasons. DFR and LDOX that act as key points of transcriptional control in the Caryophyllales (Shimada, Otsuki & Sakuta 2007) and onions (DFR, Kim et al. 2005) were up-regulated at most sample dates in WS berries. However, the integral of expression of neither DFR nor LDOX was as strongly correlated as UFGT, CHS2, CHS3 and F3H with anthocyanin content. LDOX and DFR expression peak up at blooming and fruit set, respectively, rather than 2 months later during fruit ripening (Boss et al. 1996). Altogether, these features would suggest that neither of these genes is a key point in transcriptional regulation of the anthocyanin pathway in grape, consistently with what hypothesized by Boss et al. (1996). The enhanced expression of Myb regulatory genes in WS berries suggested a coordinate up-regulation of structural genes of the general flavonoid pathway and of anthocyanin-specific genes. The mrna level of the MybA transcription factor controlling the expression of the anthocyanin-specific gene UFGT was constantly higher in WS fruits throughout ripening. The Myb5a transcription factor which has been shown to generally affect the expression of several structural genes of the flavonoid pathway (Deluc et al. 2006) and the transcription factor MybC (Kobayashi et al. 2002) were up-regulated in WS berry skins at véraison and soon after véraison until harvest, respectively. Enhanced accumulation of anthocyanins in WS fruits was not attained at the expense of catechin content, even though they share common precursors. Catechin accumulation was almost completed at the time when vines entered the phase of most severe water deficit, and expression of the catechin biosynthetic genes LAR1, LAR2 and BAN had already declined when anthocyanin biosynthetic genes were triggered. The anthocyanin composition was modified under drought conditions. Water deficit stimulated hydroxylation and methoxylation of the flavonoid B-ring. We observed a significant increase of 3,4,5 -hydroxylated anthocyanins which shifted the anthocyanin profile of WS grapes towards an enrichment of purple/blue pigments. These differences in berry coloration were also detectable by chromameter reads. Expression of F3 5 H was up-regulated in WS berries in a manner that could explain the increase of the 3,4,5 - hydroxylated anthocyanins. Bogs et al. (2006), Castellarin et al. (2006) and Jeong et al. (2006) showed that transcript abundance of F3 5 H and the level of anthocyanin hydroxylation during berry maturation are positively correlated. This experiment showed that this correlation is maintained when different water regimes and different years are compared. Drought conditions also promoted the conversion of hydroxylated anthocyanins (cyanidin and delphinidin) into their methoxylated derivatives (peonidin, petunidin and malvidin). Methoxylation of delphinidin rarely occurred only at the 3 position since its 3 -methoxylated derivative, petunidin, did not increase in WS fruits. Most frequently, delphinidin was methoxylated at both 3 and 5 positions, which led to a significant increase of malvidin. A candidate gene coding for an OMT (Ageorges et al. 2006) was up-regulated in WS berries in a manner that was consistent with the pattern of increased accumulation of peonidin and malvidin. It is particularly fascinating that fruits responded to environmental stress conditions (water deficit and increased sunlight radiation) by enriching the anthocyanin profile in peonidin and malvidin, the two 3-monoglucoside anthocyanins to which human gastric bilitranslocase displays the highest affinity (Passamonti, Vrhovsek & Mattivi 2002). Large seasonal variation of anthocyanin content was observed in two consecutive years. CT plots were subjected to similar water regimes (192 mm versus 211 of water supply in 2004 and 2005, respectively, over the sampling period), which resulted in a comparable water status throughout the maturation. CT plants were grown following the same agronomical practices and had similar crop load over the 2 years. Upon reproducible experimental conditions, CT grapes accumulated 3.3 mg berry -1 anthocyanins in 2004 versus 1.3 mg berry -1 anthocyanins in 2005, and this difference was maintained also if normalized to yield per plant. This extent of variation is consistent with other studies in which anthocyanin content had a two- to threefold variation in different vintages (Downey et al. 2004). Anthocyanin content was proportionally higher in 2004 versus 2005 also in WS berries, even though an additional effect due to a different water regime between the years must be considered for the WS treatment. In order to find out which environmental factor(s) could have caused the observed increase of total anthocyanins in 2004 under both water regimes, we compared meteorological data between the years. Total radiation decreased from 842 776 kj m -2 in 2004 to 782 162 kj m -2 in 2005 ( 7%) over the period of anthocyanin biosynthesis. Mean temperature was exactly the same in the 2 years (21.4 C), but day night fluctuations were higher in 2004 than in 2005 with average minimum temperature 0.8 C lower and average maximum temperature 0.7 C higher in 2004 than in 2005. The number of days with maximum temperature higher than 30 C was four in 2004 and three in 2005. These climate factors are in agreement with the higher anthocyanin content observed in vintage 2004. Several studies have demonstrated positive correlation between radiation, day night temperature fluctuation and anthocyanin content, while excessive heat (>30 C) adversely affected coloration (Mori et al. 2005; Yamane et al. 2006). In our experiment, radiation, day night temperature fluctuation, GDD from the onset of véraison until every ripening stage were plotted versus cumulative gene expression at each sampling date and were used for calculating linear regression (Table 2). UFGT, CHS1, F3 5 H, MybC and DWF1 correlated well with radiation

1396 S. D. Castellarin et al. (r = 0.95 0.98) and the summation of day night temperature fluctuation (r = 0.94 0.98) over the two experimental seasons. The expression pattern of most flavonoid genes also showed footprints of short-term climate fluctuations which occurred during véraison 2005. These genes were coordinately induced at the beginning of véraison in parallel with the activation of anthocyanin biosynthesis. Then, transcript levels transiently decreased at the subsequent sampling date, following an abnormal decrease of mean and minimum temperatures in mid-august (minimum temperature 9 C on 8 August 2005) which delayed the completion of véraison in that year. Which signal transduction pathways enhanced expression of anthocyanin biosynthetic genes under drought conditions? In this experiment, fruit trees grown under field conditions were gradually exposed to water deficit as the soil dried out under a rain-off shelter. The slow acclimatation to water depletion affected canopy architecture and stimulated ripening-related processes. Drought reduced foliage in the basal part of shoots and enhanced exposure of bunches to sunlight, particularly during the latest stages of maturation. As a result of the restricted water supply to bunches, berry weight decreased significantly. The cost of physiological adaptation to drought was not detrimental to sugar metabolism and fruit sink strength, because parameters of juice quality (soluble sugars and acidity) were not affected by the level of water deficit applied in this experiment. Early transition to ripening in WS plants was witnessed by a higher percentage of coloured berries in turning bunches compared with CT plants at the same date. In addition, at the mrna level, genes committed to brassinosteroid control of maturation (Symons et al. 2006) were strongly triggered in WS plants at the inception of véraison. Long-term water shortage and slow acclimatation under field conditions may have impacted anthocyanin biosynthesis through signal transduction pathways that are different from those triggered by a sudden imposition of water deficiency to plants under a simplified water relationship system (Okamoto, Kuwamura & Hirano 2004).The role ofaba, the hormone committed to the signalling of water stress in leaves and roots, might have played a negligible role in skin of WS fruits. Okamoto et al. (2004), even following an abrupt exposition to water depletion, found that ABA level enhanced only transiently in a whole berry assay and the effect disappeared if the treatment was applied during the last stages of ripening. While it is proven that (1) anthocyanin accumulation and expression of anthocyanin biosynthetic genes are enhanced by exogenous application of ABA to berries (Ban et al. 2003; Peppi, Fidelibus & Dokoozlian 2006) and (2) ABA is a water stress-responsive signal in roots and leaves (Stoll, Loveys & Dry 2000), there is no direct evidence for causality between whole plant water starvation, ABA levels in berry skin and anthocyanin transcriptional regulation in the site of synthesis. We monitored directly in the tissue of anthocyanin biosynthesis the expression of four genes related to ABA metabolism, two of them (NCED1 and NCED2) coding for the ABA biosynthetic enzymes 9-cisepoxycarotenoid dioxygenases and two others (rd22 and ACPK1) induced in the presence ofaba.at transcript level, none of these genes were consistently induced inws skins as one would expect if ABA synthesis is transcriptionally induced or ABA level increases in ripening fruit exposed to progressive water deficit. We observed a coincidence between the beginning of sugar accumulation and the increase in transcripts of anthocyanin biosynthetic genes in both treatments, which provided further hints that anthocyanin biosynthesis is developmentally triggered in a sugar-dependent manner in grapevine (Gollop et al. 2002), like in Arabidopsis (Solfanelli et al. 2006). However, the expression of the sugar-sensitive VvMSA gene (Cakir et al. 2003) did not differentially react between WS and CT fruits consistently with a lack of significant differences in sugar content between the treatments. Hence, we did not find any hints that long-term adaptation to a progressive water depletion consistently perturbed transcriptional levels of the ABA- and sugar-sensitive genes tested here. Water deficit might have magnified anthocyanin biosynthesis through additional signalling pathways or by modifying transcript and/or protein turnover and/or protein activity. Because drought progressively modified canopy architecture by a reduction of basal foliage of shoot where vines bore bunches, fruits were more exposed to solar radiation in WS plants. One could argue that the altered anthocyanin biosynthesis was partly due to light-mediated effects. However, WS and CT berries differed significantly for anthocyanin content and composition, and showed remarkable differences in transcript levels of anthocyanin biosynthetic genes long before the leaf layer number and the percentage of sun-exposed clusters were significantly different in the two treatments. We also used FLS1 gene expression in berry skin as an indicator of cluster light exposure. Previously, FLS1, the gene required for flavonol biosynthesis, strongly correlated with the degree of fruit exposure in grape (Price et al. 1995; Downey et al. 2004; Fujita et al. 2006), likely connected with the UV protectant properties of flavonols (Kolb et al. 2003). FLS1 was not consistently up-regulated in WS fruits during the most intense phase of anthocyanin biosynthesis and substantially increased only in the last part of season 2005.The effect of light shading on flavonoid accumulation was studied by several authors who found that light had minimal effects on total anthocyanin content (Downey et al. 2004; Jeong et al. 2004; Cortell & Kennedy 2006), even though anthocyanins dramatically decreased in response to complete darkness. By contrast, light exposure consistently affected anthocyanin composition by increasing the ratio of tri-hydroxylated to dihydroxylated and methoxylated to non-methoxylated anthocyanins, in a way that was similar to what we observed in WS fruits. In conclusion, to determine the mechanism of enhanced accumulation and of altered profile of anthocyanins in the