Colour variation in red grapevines (Vitis vinifera L.): genomic organisation, expression of flavonoid 3'-hydroxylase, flavonoid 3',5'-hydroxylase genes and related metabolite profiling of red cyanidin-/blue delphinidin-based anthocyanins in berry skin

Background Structural genes of the phenyl-propanoid pathway which encode flavonoid 3'- and 3',5'-hydroxylases (F3'H and F3'5'H) have long been invoked to explain the biosynthesis of cyanidin- and delphinidin-based anthocyanin pigments in the so-called red cultivars of grapevine. The relative proportion of the two types of anthocyanins is largely under genetic control and determines the colour variation among red/purple/blue berry grape varieties and their corresponding wines. Results Gene fragments of VvF3'H and VvF3'5'H, that were isolated from Vitis vinifera 'Cabernet Sauvignon' using degenerate primers designed on plant homologous genes, translated into 313 and 239 amino acid protein fragments, respectively, with up to 76% and 82% identity to plant CYP75 cytochrome P450 monooxygenases. Putative function was assigned on the basis of sequence homology, expression profiling and its correlation with metabolite accumulation at ten different ripening stages. At the onset of colour transition, transcriptional induction of VvF3'H and VvF3'5'H was temporally coordinated with the beginning of anthocyanin biosynthesis, the expression being 2-fold and 50-fold higher, respectively, in red berries versus green berries. The peak of VvF3'5'H expression was observed two weeks later concomitantly with the increase of the ratio of delphinidin-/cyanidin-derivatives. The analysis of structural genomics revealed that two copies of VvF3'H are physically linked on linkage group no. 17 and several copies of VvF3'5'H are tightly clustered and embedded into a segmental duplication on linkage group no. 6, unveiling a high complexity when compared to other plant flavonoid hydroxylase genes known so far, mostly in ornamentals. Conclusion We have shown that genes encoding flavonoid 3'- and 3',5'-hydroxylases are expressed in any tissues of the grape plant that accumulate flavonoids and, particularly, in skin of ripening red berries that synthesise mostly anthocyanins. The correlation between transcript profiles and the kinetics of accumulation of red/cyanidin- and blue/delphinidin-based anthocyanins indicated that VvF3'H and VvF3'5'H expression is consistent with the chromatic evolution of ripening bunches. Local physical maps constructed around the VvF3'H and VvF3'5'H loci should help facilitate the identification of the regulatory elements of each isoform and the future manipulation of grapevine and wine colour through agronomical, environmental and biotechnological tools.


Background
Vacuolar accumulation of anthocyanins in berry skin is responsible for the pigmentation of red/blue grapevine (Vitis vinifera) cultivars. Anthocyanin biosynthesis is developmentally triggered at the onset of berry ripening (8-10 weeks after blooming) and lasts till harvest [1]. Expression profiling of structural genes and transcription factors has claimed a major role for the gene UFGT encoding a UDP-glucose:flavonoid 3-O-glucosyltransferase and its regulatory Myb-type gene VvMybA in conveying the flux of flavonoid intermediates towards the synthesis of anthocyanins [2,3]. The enzyme UFGT is synthesised since véraison under the control of VvMybA and catalyses the 3'-glycosidation of cyanidin-and delphinidin-based anthocyanidins [4]. The resulting stably coloured anthocyanins may undergo methylation upon the activity of methyl-transferases that convert cyanidin into peonidin and delphinidin into petunidin or malvidin, similarly to how it occurs in other anthocyanin-producing species [5,6].
Cyanidin-and delphinidin-derived anthocyanins diverge from each other regarding the number of the substituted groups that are found on the B-ring of the flavonoid skeleton (either two or three, respectively) and have their chromatic properties affected accordingly. Cyanidinbased anthocyanins exhibit a reddish colour whereas delphinidin-based anthocyanins are purple to blue. They jointly determine the tonality of red berry grape varieties and their corresponding wines. The relative proportion of the five anthocyanins is largely under genetic control and unique to each cultivar [7,8]. Sunlight exposition and water deficit have been shown to modulate to some extent the anthocyanin profile [9,10].
In plant species, the biosynthesis of cyanidin-and delphinidin-type precursors is driven, upstream of the enzyme UFGT, by the activity of two flavonoid 3'-and 3',5'hydroxylases (F3'H and F3',5'H) that add either a single hydroxyl group at the 3' position or two hydroxyl groups at the 3' and 5' positions to dihydrokaempferol and/or dihydroquercetin. Once converted into 3'-dihydroquercetin or 3',5'-dihydromyricetin, these intermediates flow through common downstream enzymes to form di-substituted and tri-substituted anthocyanins when the gene UFGT is expressed and to form other polyphenols (flavonols, leucoanthocyanins, catechins, proanthocyanins, tannins) at different developmental stages. F3'H and F3',5'H genes code for CYP75 cytochrome P450 hydroxylases that have been well characterised in the ornamental  plants Antirrhinum majus [11], Campanula medium [12], Dianthus caryophyllus [13], Gentiana triflora [14], Ipomoea purpurea, [15], Petunia hybrida [16], Phalaenopsis [17] and Vinca major [18]. In all these species, F3'H and F3',5'H are found as low/medium copy number genes and have been genetically manipulated to produce novel flower pigmentation. In grapevine, only tentative consensus (TC) sequences for F3'H and F3',5'H genes are available in the TIGR grape gene database http://www.tigr.org. TCs were built based on partially overlapping ESTs from cDNA libraries obtained from different genotypes and tissues. This makes it unfeasible to predict or rule out the presence of different isogenes.
This paper reports (1) the isolation of flavonoid 3'-and 3',5'-hydroxylase gene members in grapevine via sequence homology, (2) the assessment of their chromosomal location by genetic and physical mapping, (3) their transcript profiles in different tissues, (4) the time-course of expression in berry skin of the red cultivar 'Merlot' from prevéraison to harvest compared to the curves of accumulation of anthocyanins, catechins and total polyphenols.

Isolation of flavonoid 3'-and 3',5'-hydroxylase genes in grapevine
Gene isolation was accomplished via sequence homology with the help of newly designed degenerate primers. PCR bands that were amplified from genomic DNA of Vitis vinifera 'Cabernet Sauvignon' were longer than predicted on the basis of plant flavonoid hydroxylase transcripts (    In order to test the phylogenetic relationship between grapevine hydroxylases and the multi-functional family of plant monooxygenases, a Neighbor-Joining tree was constructed using translated sequences of VvF3'H, VvF3'5'H-1, VvF3'5'H-2 and 670 protein entries of plant cyt P450 monooxygenases ( Figure 3). VvF3'H and VvF3'5'H grouped solely with P450 proteins specifically annotated as flavonoid hydroxylases belonging to the family-specific branch of CYP75 monooxygenases. VvF3'H and VvF3'5'H split into two terminal branches, each one hosting either CYP75B or CYP75A monooxygenases. Although most plant P450s are of polyphyletic origin and some P450 families are lineage-specific [19], CYP75 P450s are ancient genes that existed even before the separation between gymnosperms and flowering plants [20]. The ancient origin of CYP75 monooxygenases is also supported by the finding that intron position of VvF3'H and VvF3'5'H was conserved when compared to other plant genes coding for flavonoid hydroxylases such as the Arabidopsis gene TT7 [21].

Genetic mapping
Genetic loci harbouring genes that code for flavonoid hydroxylases were mapped using the single strand conformational polymorphism (SSCP) analysis of F3'H and Vitis 'breeding line 20/3'. The two genes segregated independently and were mapped to linkage groups LG17 and LG6, respectively ( Figure 4). The linkage groups were annotated according to [22] and included also reference microsatellite markers and SSCP markers generated from BAC end-sequences (BES) of a 'Cabernet Sauvignon' BAC library. The locus containing the gene that codes for flavonoid 3'-hydroxylase was identified by linkage analysis following the segregation of the SSCP of two co-segregating bands amplified from the parent 'Cabernet Sauvignon' by means of the primer pair F3int ( Table 2). The VvF3'H locus mapped to the middle of LG17 and co-segregated with the BES marker 5A23-FM and the microsatellite marker VMC9g4. The locus containing the genes that code for flavonoid 3',5'-hydroxylase was mapped by means of a polymorphic SSCP band amplified from the parent 'Cabernet Sauvignon' using the primer pair F35-1int. This marker mapped to the middle of LG6 and co-segregated with microsatellite marker VrZag30 and BES markers 9C17-FM and 17K4-FM ( Figure 4). Other SSCP bands that were co-amplified by the same primer pair F35-1int could not be scored in the progeny due to faintness or homozy-gosis. SSCP marker F35-2int could not be mapped in the 'Cabernet Sauvignon' parental map, but co-segregated with VrZag30 and SSCP marker F35-1int in the male parent of the same cross (data not shown). The CYP75 clan of plant cyt P450s Figure 3 The CYP75 clan of plant cyt P450s. The branch containing the CYP75 family is focused on a Neighbor-Joining tree constructed using 670 plant cytochrome P450 monooxygenases. Flavonoid 3'-hydroxylases VvF3'H-1a to -1d and flavonoid 3',5'hydroxylases VvF3'5'H-1a to -2b that were identified in grapevine are in bold. They split into two lower hierarchical branches that identified the CYP75B (flavonoid 3'-hydroxylases) and CYP75A (flavonoid 3',5'-hydroxylases) sub-families, respectively. Two flavonoid 3'-hydroxylase genes, VvF3'H-1 and VvF3'H-2, were tightly linked in contig ctg253 that consisted of 28 BAC clones and approximately covered 700 kb ( Figure 5). Six BAC clones contained both isogenes, whilst two BAC clones (6A08 and 5A23) contained only VvF3'H-1. BAC clone 5A23 was initially not included in the ctg253 following computational contig assembly because it did not meet the cut-off due to a low number of fingerprint bands. However, it was later placed in the con-tig upon PCR anchorage to other BAC clones using the BES-derived primers (BES5A23-FM). The same marker BES5A23-FM was used to anchor contig ctg253 to the 'Cabernet Sauvignon' genetic map. The marker based on the SSCP of the VvF3'H gene fragments and the BESderived marker co-segregated at the level of resolution of the genetic map.
Based on BAC order and the number of SSCP haplotypes, at a first approximation we hypothesize the occurrence of four loci containing VvF3'5'H-1-type sequences and seven loci containing VvF3'5'H-2-type sequences over the four contigs ( Figure 6). Two paralogous copies of gene VvF3'5'H-1 were identified in ctg313. The first paralog had two allelic variants, that could be detected by SSCP analysis, one present in three BAC clones 6H20, 9C17, 31P20 (corresponding to the SSCP band genetically mapped from 'Cabernet Sauvignon') and the other in the remaining five clones (data not shown). Two BAC clones (28B15 and 99I17) spanned the region that included both paralogs. A third potential paralogous copy of the gene VvF3'5'H-1 was identified in contig ctg1320 whose BAC clones contained one of the two SSCP haplotypes present in the BACs from contig ctg313. An additional different haplotype of the gene VvF3'5'H-1 was identified in contig ctg2373 that included just three BAC clones. We could detect a potential of seven paralogous copies of the gene VvF3'5'H-2 in contigs ctg313, ctg871, ctg1320 and ctg2373. The primers pair F35-2int identified three types of gene fragments distinguishable by size, hereafter referred to as short, medium and long ( Figure 6). All three gene fragments were found in ctg1320 in the close vicinity of the gene copy of VvF3'5'H-1. Three BAC clones contained all three forms at the same time, confirming the paralogous nature of these VvF3'5'H-2 gene copies. The medium sized gene fragment VvF3'5'H-2 was also amplified in ctg313 and ctg2373 where it showed a SSCP haplotype different from that found in ctg1320 (data not shown). A different locus carrying the longer fragment of VvF3'5'H-2 was present in ctg871 with a SSCP haplotype not distinguishable from that found in ctg1320.

Integration of genetic and physical maps around the locus
VvF3'5'H Contigs ctg313, ctg871, ctg1320 and ctg2373 were anchored to the 'Cabernet Sauvignon' genetic map by using SSR markers present in the BAC clones and single strand conformational polymorphism of the BAC-end sequences (Figure 7). The SSR marker VrZag30 that was identified in five BACs of ctg313 allowed this contig to be anchored to LG6. Map position of ctg313 was confirmed by mapping the SSCP marker BES9C17-FM developed on a BAC clone embedded into ctg313 that co-segregated with SSR VrZag30. The contig ctg871 was anchored to LG6 by the SSR marker VMC3a8. The contig ctg1320 was anchored to LG6 using the SSR VMC3f12 and the BAC end-derived SSCP marker BES38C24-RM. The contig ctg2373 was anchored to LG6 by the BAC end-derived SSCP marker BES17K4-FM. The SSCP band of the marker Physical map of the VvF3'H locus The BES of positive BAC clones provided new markers that were closely linked to the flavonoid 3',5'-hydroxylase gene copies and served to characterize that region more precisely. The marker BES9C17-FM was located on the physical map downstream of SSR VrZag30 and identified only BAC clones belonging to the contig ctg313. BES38C24-RM matched uniquely the BAC clones 38C24 and 21H16 belonging to ctg1320. Conversely, BES40A22-FM and BES38C24-FM primers were designed on BES belonging to ctg313 and ctg1320, respectively, but crossidentify BAC clones assembled in both contigs. SSCP analysis of BES40A22-FM and BES38C24-FM fragments revealed that for each marker two haplotypes are present in ctg313 and two different haplotypes are present in ctg1320. Yet, marker BES17K04-FM matched BACs 6H20, 9C17 and 31P20 of ctg313 and BACs 29J23, 38C24, 4K10, 30K14 of ctg1320. SSCP analysis identified four different haplotypes at the regions amplified by BES17K04-FM primers: two haplotypes were detected in the BAC clones of ctg1320, one more different haplotype was present in ctg313, and yet another SSCP haplotype was found in both BAC17K04 and BAC24N06 belonging to ctg2373. The occurrence of more than two haplotypes of the flavonoid 3',5'-hydroxylase gene fragments and of the BES markers in a diploid genotype confirmed the presence of paralogous copies of the flavonoid 3',5'-hydroxylase genes and rule out the possibility that the individual contigs represent allelic regions that have not been properly Integrated genetic and local physical map spanning the complex locus VvF3'5'H on LG6 Figure 7 Integrated genetic and local physical map spanning the complex locus VvF3'5'H on LG6. Contigs were assembled using FPC and BAC clones were further aligned using BES-derived markers. Contigs were anchored to the 'Cabernet Sauvignon' and the V. vinifera consensus genetic maps using cross-referenced SSR markers. The orientation of each contig is arbitrary. Genetic and physical distances are not drawn to scale. Names written in horizontal refer to BAC clones and contigs; those written in vertical refer to markers and genes. The presence of a given marker on a BAC clone is highlighted with a symbol and different SSCP haplotypes for each marker are shown with different symbols (■ ᭜ • ❍ ᭝). The symbol 'X' stands for lack of amplification of a given marker in the corresponding BAC clone. BAC clones missing any symbols were not tested with the corresponding marker. Physical localisation of VvF3'5'H-1 and VvF3'5'H-2 loci is shown in grey boxes.
assembled. A segmental duplication spanning the markers F35-1int, F35-2int, BES40A22-FM, BES17K04-FM and BES38C24-FM that occurred within 2.2-3.0 cM on LG6 could explain the physical organization of flavonoid 3',5'hydroxylase isogenes as it has been outlined in Figure 7. The contig ctg2373 that comprised only three BAC clones could not be unambiguously positioned in relation to the other contigs. BES markers from the other contigs did not match any BAC clones of ctg2373 but marker BES17K04-FM matched both ctg313 and ctg1320. Marker BES17K04-FM is therefore located within the segmental duplication. It remains uncertain whether ctg2373 partially overlaps the region of ctg313 and split from it due to high heterozygosity or if it actually covers an adjacent locus. Under the assumption of BAC alignment between BAC clones 9C17 and 17K04 that is supported by genetic co-segregation of the corresponding BES and by co-amplification by marker BES17K04-FM, ctg2373 would carry the haplotype allelic to BACs 6H20, 9C17, 31P20 at the BES17K04-FM SSCP marker and the VvF3'5'H-1 gene fragment present in BAC17K04 would miss a counterpart on the homologous chromosome. Otherwise, ctg2373 would identify a forth different locus of VvF3'5'H-1, located in the surrounding region and under-represented among the BAC clones of the library.
Based on the data available to us, we propose that the chromosomal region spanning the VvF3'5H gene cluster includes four copies of VvF3'5'H-1 and five to seven copies of VvF3'5'H-2, all sharing a high sequence similarity. We were able to present a preliminary physical map of this cluster of genes by a combination of computational assembly of fingerprinted BACs and by PCR-based alignment of BAC clones. The region has likely originated from both tandem duplications that generated nearby gene copies (i.e. paralogs within BACs 28B15 and 99I17 in the ctg313) and from segmental intrachromosomal duplications involving a larger block of DNA (i.e. blocks of paralogs duplicated into ctg313 and ctg1320). Segmental duplications are an important source of gene evolution [25] and they have been estimated to cover 1.7-2.0% of the mouse genome [26] and 4% of the human genome [27]. In the rat genome, duplicated genes are largely confined into <1 Mb intrachromosomal duplications rather than widely interspersed through the genome [28]. A similar model of genome dynamics could also have shaped plant genomes as it is emerging for model species [29].

Expression analysis of VvF3'H and VvF3'5'H along with major structural genes of the flavonoid pathway in different grape tissues
The genes encoding both flavonoid 3'-hydroxylase and flavonoid 3'-5'-hydroxylases are widely expressed in grapevine organs (Figure 8). Transcripts of gene VvF3'H-1 were detected in all tissues tested. VvF3'H-1 was strongly expressed in green apical leaflets and flowers but it was only slightly detectable in fully expanded leaves. It was also present in all berry tissues. The other sequence VvF3'H-2, which translated into a truncated flavonoid 3'hydroxylase due to a 23-bp frameshift deletion, was apparently not expressed and is likely to be a pseudogene.
Transcripts of VvF3'5'H-1 were present in all tissues. High levels of expression were detected in pigmented apical leaflets, flowers, seeds and berries but not in green fully expanded leaves and roots. Differential expression could be detected between green berry skin and red berry skin, the expression being higher in pigmented berry skin. The flavonoid pathway does not exclusively lead to the synthesis of anthocyanins and several flavonoid hydroxylase isogenes were therefore expressed also in non-pigmented tissues. Berry flesh and seeds accumulate nonanthocyanin flavonoids such as proanthocyanidins and catechins. Their synthesis is catalysed downstream of the flavonoid hydroxylases by the genes DFR, LDOX and BAN. Gene BAN was strongly expressed in seeds, berry flesh and berry skin at pre-véraison, whilst was almost undetectable in post-véraison berry skin at the stage when the rate of accumulation of proanthocyanins and catechins is declining (Figure 9). Sunlight exposed tissues expressed the gene coding for flavonol synthase (Fls1). Flavonol synthase converts the products of flavonoid hydroxylases into the UV-protecting flavonols. Inner tissues (seeds and berry flesh) and underground organs (rootlets) did not exhibit detectable levels of Fls1.

Expression profiling of flavonoid hydroxylase genes in skin of ripening berries
The time-course of expression of flavonoid hydroxylase genes in the red-skinned cultivar 'Merlot' revealed that gene VvF3'H-1 is expressed throughout ripening. High transcript levels were detected from the first sampling date at the stage when catechins are rapidly synthesised and peak at 5.6 mg g -1 of berry skin (Figure 9). VvF3'H-1 expression decreased 8-10 weeks after blooming, at the ripening stage when most genes committed to the flavonoid pathway have been reported to level off [1]. VvF3'H-1 transcripts increased soon after, at the onset of véraison (August 12 th ), which were at that point in time 2-fold higher in red berries than in green berries collected from the same bunch. VvF3'H-1 expression lasted till harvest, with a secondary peak of expression one week before harvest, concurrently at the peak of anthocyanin content.
The gene VvF3'5'H-1 was weakly expressed during the earliest ripening stages. At the onset of véraison, transcripts of VvF3'5'H-1 were poorly detectable in green berries, while they increased 50-fold in berries that had already turned red within the same bunch. The peak of VvF3'5'H-1 expression corresponded to August 24 th , at the stage when all berries had turned red. As described for VvF3'H-1, a weaker peak of expression was found one week before harvest. Expression of VvF3'5'H-2 was checked by semiquantitative PCR because the different isoforms of this gene were distinguishable by size ( Figure 10). Transcripts of the isoforms VvF3'5'H-2-long and VvF3'5'H-2-medium did appear at the onset of véraison and lasted till full ripening, whilst the isogene VvF3'5'H-2-short was constitutively expressed throughout the ripening.

Accumulation, composition of anthocyanins and colour evolution in berry skin during the progress of ripening
Biometric and analytical parameters have been used to follow the progress of ripening (Figure 9). The onset of ripening began on August 3 rd when sugars started to accumulate and acidity commenced to decline. The onset of véraison (August 12 th ) was considered as the time when 10% of the berries had turned red (H° = 102.98, C* = 18.27). Bunches went through this phase of colour transition until August 24 th when all berries had turned red (H°= 20.46, C* = 7.43). Anthocyanins have been detected by HPLC analysis since the onset of véraison on August 12 th (0.4 mg g -1 ). The accumulation of anthocyanins lasted until September 20 th , one week before harvest (19.8 mg g -1 ), slightly decreasing at harvest (17.1 mg g -1 ).
At the earliest stage of the véraison, cyanidin-based anthocyanins (cyanidin and peonidin) were roughly as abundant as delphinidin-based anthocyanins (delphinidin, malvidin and petunidin) (42% versus 58%). Then, on August 24 th the percentage of delphinidin-based anthocyanin (79%) peaked simultaneously with the highest level of expression of VvF3'5'H-1. The fraction of trisubstituted anthocyanins slightly decreased in the following weeks (77-78%), then remaining substantially unchanged until harvest (75%). Berry skin colour moderately varied from H° = 331.55, C* = 3.83 to H° = 306, C* = 2.59 from September 2 nd to September 28 th . Anthocyanin composition in berry skin changed over the period of Expression of structural genes of the flavonoid pathway in different tissues Figure 8 Expression of structural genes of the flavonoid pathway in different tissues. Expression profiles of flavonoid hydroxylase genes were analysed alongside nine major structural genes of the flavonoid pathway in the following tissues: 1, apex and apical leaflet (pale green); 2, apex and apical leaflets (reddish pigmented shoot tips); 3, fully expanded leaves; 4, immature flowers; 5, rootlets; 6, seeds; 7, berry flesh; 8, berry skin (pre-véraison, green); 9, berry skin (post-véraison, red). The actin gene was used as a constitutive gene for the normalisation of cDNA samples.

Conclusion
We have shown that genes coding for flavonoid 3'-and 3',5'-hydroxylases are present as low/medium copy members in the grapevine genome and are organised in complex gene clusters that most likely resulted from gene duplication and intra-chromosomal segmental duplication. These genes are expressed in coordination with other genes of the flavonoid pathway in any tissues of the grape plant that are known to accumulate flavonoids, such as leaves, flowers, roots, seeds and, particularly, in skin of ripening berries. RT-PCR showed a strong relationship between the expression of VvF3'H-1 and VvF3'5'H and the kinetics of accumulation of cyanidin-and delphinidinbased anthocyanins in the red-skinned variety 'Merlot'. The presence of multiple putative 3',5'-hydroxylase isogenes which emerged from fine structural analysis at physical level, has first been shown for plant hydroxylases. This complexity was not unexpected for species like grape which synthesize large amounts of different classes of polyphenols and display a continuous variation of berry skin colour from brick red to dark blue among the so called red-skinned cultivars.

Plant material
Genomic DNA was extracted from young leaves [30]. 'Cabernet Sauvignon' was used for cloning flavonoid 3'and 3',5'-hydroxylase genes. Forty six full-sibs of the cross 'Cabernet Sauvignon' × Vitis 'breeding line 20/3' were used for genetic mapping. Gene expression was analysed in the red variety 'Merlot'. Apical leaflets of 1 cm 2 surface area, fully expanded leaves (>10 cm 2 ) and adventitious rootlets were collected from canes grown in hydroponic culture for two months. Flower buttons were collected one week before blooming, seeds and flesh were extracted from berries collected two weeks after véraison. Transcript and metabolite profiling in berry skin was performed on samples of 40 berries collected at ten ripening stages from pre-véraison (July 15 th , 2004) till harvest (September 28 th , 2004) from plots of twelve vines replicated four times.

Cloning of flavonoid 3'-and 3',5'-hydroxylase genes with the help of degenerate primers
Degenerate primers were designed on two specific and conserved regions of plant F3'H and F3',5'H genes (Table 1, Figure 1). PCRs were performed in a 40 µl volume containing 200 µM of each dNTP, 2.5 µM of each primer, 1 U of HotMaster Taq polymerase (Eppendorf) with 2.5 mM MgCl 2 buffer, and 40 ng of template DNA. The denaturation step was 95°C for 1 min, followed by 40 cycles of 92°C for 45 s, 60°C for F3'H primers and 56°C for F3',5'H primers for 50 s, 65°C for 90 s. Amplified bands were extracted from agarose gel with Nucleospin Extract (Macherey-Nagel) and cloned into pGEM-T vector (Promega). Unique clones were identified by means of PCR with M13 universal primers and SSCP-fingerprinting of the insert. Selected clones were sequenced with BigDye terminator chemistry and a MegaBace500 sequencer (Amersham Biosciences).

Sequence analysis
Nucleotide and amino acid sequences were compared with GenBank accessions, with EST singletons and tentative consensus sequences (TCs) held at the TIGR grape gene index. Conceptual translation was obtained with DNAclub software. Amino acid entries of 670 plant cytochrome P450 monooxygenases were downloaded from GenBank as of November 30 th , 2004. Cluster analysis of plant cyt P450s and the candidate grape flavonoid hydroxylases was done with the Neighbor-joining method.

Genetic mapping
Nested primers were designed to amplify an exon of grapevine flavonoid 3'-and 3',5'-hydroxylase genes (F3, F3-1, F35-1, F35-2, see Table 1 and 2). Additional primer pairs (F3int, F35-1int and F35-2int) were designed in the exon sequences flanking an intron for detecting SNPs use-Expression of VvF3'5'H-2 during the progress of ripening Figure 10 Expression of VvF3'5'H-2 during the progress of ripening. Length polymorphic transcripts of VvF3'5'H-2 were detected semi-quantitatively after cDNA samples had been normalised upon actin gene expression, at the sampling dates reported in Fig 9. The short, medium and long transcripts were named according to the fragment size amplified from genomic DNA. At sampling 5, green berries (5g) were kept separated from red berries (5r). PCR amplification on 'Merlot' genomic DNA was loaded in the right extreme lane.
ful for mapping in untranslated regions (Table 1 and 2). Flavonoid hydroxylase gene fragments containing the intron were amplified in a 10 µl volume containing 200 µM of dNTPs, 0.4 µM of each primer, 0.5 U of Taq polymerase, 2.5 mM MgCl 2 and 20 ng of template DNA. The PCR profile was 35 cycles of 92°C for 50s, annealing temperature as reported in Table 2 for 50s, 65°C for 1 min. Gene fragments were separated on 4% glycerol and 0.5 × MDE polyacrylamide gel (BioWhittaker Molecular Applications) for SSCP analysis, run at room temperature for 16 hours at 6 W and stained with silver nitrate. The segregation of F3int, F35-1int, F35-2int and 335 SSR markers was followed in the 'Cabernet Sauvignon' × Vitis 'breeding line 20/3' full-sib family [G. Di Gaspero, unpublished data]. Linkage groups of markers segregating from the 'Cabernet Sauvignon' parent were identified using Carthagene with a LOD of 5.0 [31]. Marker order was calculated using the command Build 3.

Local physical maps
A 6X BAC library of 'Cabernet Sauvignon' was screened by PCR with the F3, F35-1 and F35-2 primer pairs (Table 2) using a 3D pooling strategy [23]. Then, positive individual clones were pre-cultured in 2X LB medium for 20 hours and 2 µL of cell suspension were inoculated into the same medium and grown for 16 hours. Cultured clones were directly used as PCR template. Positive BACs were confirmed using the corresponding F3int, F35-1int and F35-2int primer pairs (Table 2). PCR products were then run into a SSCP gel (as above) for identifying the haplotype of each BAC clone. A set of 30,151 BAC fingerprints of the 'Cabernet Sauvignon' BAC clones [S. Paillard, unpublished data] were assembled into contigs using FPC v8.0 [24]. The first assembly was made at a cut-off value of 10 -50 and with a tolerance of 0.4 bp. After several steps of dQing and merging, 2,035 BAC contigs were obtained [S. Paillard, unpublished data]. BAC clones that were positive at the PCR screening were localized over the contigs. Clone order was determined by running the consensus band (CB) algorithm on each individual positive contig. Contig size was estimated by converting CB units into physical length, under the assumption of a band mean size of 1,100 bp. Positive contigs were anchored to the genetic map by means of already mapped SSR markers that were present in the assembled BACs [ [23], Adam-Blondon, unpublished data]. Further SSCP markers (38C24-RM, 17K14-FM, 9C17-FM, 40A22-FM and 5A23-FM, Table 1 and 2) were developed on the BAC-end sequences (BES) [A.F. Adam-Blondon and S. Paillard, unpublished data] of the positive clones and enabled new genetic markers to be identified that were closely linked to VvF3'H and VvF3'5'H loci (Table 1 and 2). BES-derived markers (Table 1 and 2) were also used to cross-reference the BAC clones within a contig, validate the clone order and identify the BAC haplotypes in regions where allelic BACs were incorporated into a given contig.

Transcript profiling
Total RNA was extracted from berry skin and flesh, leaves, flower bunches, rootlets and seeds following the procedure described in [31]. Total RNA was treated with 0.5 U/ µg RQ1 DNase (Promega) in presence of an RNAse inhibitor (RNAguard, Amersham Biosciences). First strand cDNA was synthesised using 2 µg of RNA, 0.5 µM (dT) 18 primer and 50 U of M-MLV Reverse Transcriptase (Promega). Quantitative real-time PCR was carried out on a DNA Engine Opticon2 (MJ Research) using SYBR Green Jump Start Taq ReadyMix (Sigma). The primer pairs F3-1, F35-1 and F35-2 ( Table 2) were used for expression analysis of VvF3'H-1, VvF3'5'H-1 and VvF3'5'H-2, respectively. Each cDNA sample was analysed at two different dilutions and with two replicates per dilution. Gene transcripts were quantified comparing the threshold cycle (C T ) with that of the constitutive actin gene [32]. Semi-quantitative PCR was performed on different tissues upon cDNA normalisation based on the expression of the actin gene and visualised on agarose gel stained with ethidium bromide. Primers pairs for the genes of the flavonoid pathway were retrieved from the literature: primers for Chs1, Chs2, Chs3, DFR, LDOX, UFGT were from [33], primers for Fls1 were from [34]. Primers for F3H and BAN were newly designed on the original gene sequences [35,36] to tag 150-200 bp gene fragments.

Metabolite profiling during berry development
Total soluble solids were quantified with a refractometer (Analytical control) and expressed as °Brix. Titratable acidity was measured with an automated titrator (Crison) and expressed as tartaric acid equivalents. Skin phenols were extracted for 4 hours in 1/10 (w/v) skin/solvent suspension of 100% methanol and 1% HCl, then centrifuged at 5,000 rpm for 15 min and filtered with a 0.2 µm PTFE filter (Chemtek Analytica). Total phenols were determined following the procedure of [37]. The calibration curve was obtained with (+)-gallic acid (Sigma). Total flavan-3-ols (catechins) were quantified as reported in [38]. The calibration curve was constructed with (+)-catechin (Extrasynthase). Absorbance was measured by a UV/VIS LAMBDA 2 spectrometer (Perkin-Elmer). Anthocyanins for the HPLC analysis were extracted for 4 hours from 200 mg of berry skin with 2 mL of methanol, then centrifuged and filtered with a 0.2 µm PTFE filter (Chemtek analitica). Methanol was evaporated under nitrogen flux and anthocyanins were re-suspended with 100-400 µL of 27:73 methanol:perchloric acid 0.3% (v/v). Anthocyanins were separated in a high performance liquid chromatograph (Perkin Elmer Series 4) using a C18 Purospher RP-18 (5 µm, 250 mm × 4 mm) column (Merck) protected by a C18 guard column, according to the procedure reported in [39]. Anthocyanins were detected by a LC-95 spectrophotometric detector (Perkin Elmer) operating at 520 nm. Anthocyanin content was expressed as mg/L of malvidin 3-glucoside upon the construction of a standard curve. The composition of monoglucoside anthocyanins was used for calculating the ratio of tri-/di-substituted derivatives. Berry colour was measured with an X-Rite 948 Chromameter (X-Rite) and averaged over 160 berries at each sampling date. Colorimetric specification was referenced to the CIELab scale. Hue angle (H°) and chroma (C*) were calculated according to [40].