Vitis vinifera cv. Cabernet Sauvignon grapes were harvested on a weekly basis over the course of berry development from the Shenandoah Vineyard, Plymouth, California during the summer of 2004. Samples corresponding to stages 31 to 38 of the modified E-L system 14 were measured for berry diameter, °Brix (an approximate measure of the mass ratio of dissolved solids, mostly sucrose, to water in fruit juices) and titratable acidity (Figure 1). Berry diameter increased over time with a classical double sigmoid pattern (Figure 1A). Average berry diameter increased during the first 7 weeks of development (E-L stage 31), followed by a cessation of berry expansion at 7 to 8 weeks post-anthesis (E-L stages 32 to 34), and then the increase in berry diameter resumed until maturity (E-L stages 35 to 38). °Brix increased 6 weeks post-anthesis to a peak value of 22 °Brix at 16 weeks post-anthesis (Figure 1B). In contrast, titratable acidity (g/L), which reflects acid accumulation (mainly tartaric and malic acid), increased steadily up to 8 weeks post-anthesis and then sharply declined at the start of véraison between E-L stages 34 and 35 reaching approximately 7 g/L of titratable acids at harvest (Figure 1C).

The mRNA expression profiles of seven time points spanning E-L stages 31 to 38 as indicated in Figure 1 were compared using the Affymetrix GeneChip^® Vitis genome array ver. 1.0. Testing was performed using biological triplicates for each time point. Multiple time points within Stage II (E-L stages 32 to 35) were sampled due to the large number of biochemical changes expected to occur around véraison that affect berry ripening and fruit quality. A visual inspection of the distributions of raw perfect match (PM) probe-level intensities for all 21 arrays showed that the pre-processed and normalized PM intensities using Robust Multi-Array Average (RMA) 21 were consistent across all arrays. Digestion curves describing trends in RNA degradation between the 5' end and the 3' end in each probe set were examined and all 21 proved very similar [Additional File 1A,B]. Correlations among biological replicates were good: Spearman coefficients ranged from 0.977 to 0.997; Pearson coefficients ranged between 0.977 and 0.996.

From the Vitis 16,602 probesets represented on the array [Additional File 1C], an overall mean call rate of 74.5% per array (range 73.5% to 76.2%) was obtained. Data from the 12,596 probe sets that were found to be present in at least two out of the three biological triplicates were retained for further analyses. After performing an ANOVA and a multiple testing correction (Benjamini and Hochberg) 22, we found that 10,068 probesets (60.6%) were differentially expressed (p ≤ 0.05) between two or more E-L stages of berry development [Additional File 2: Table 1]. Because one Unigene can be related to several probesets, the number of Unigenes decreased to 9,143 Unigenes [Additional File 2: Table 2]. These probesets will be hereby referred to as those passing the ANOVA filter. From this set of genes, we extracted a subset of 4,510 probesets that displayed a two-fold or greater change in steady-state transcript abundance over the course of development (i.e., across any two of the seven developmental stages) [Additional File 2: Table 3] representing 4,151 Unigenes (28.3%) in the DFCI Grape Gene Index database VvGI5 23. We refer to this subset of genes as the two-fold ratio (TFR) set [Additional File 2: Table 4].

Principal component analysis (PCA), was used to simplify and define associations between different developmental stages within the global transcriptomic data (Additional File 3). Two principal components explaining 97.4% of the overall variance of transcription profiles (86.8% and 7.6% for axes 1 and 2, respectively) allowed us to clearly differentiate E-L stages 31 and 35 from the other developmental stages analyzed (Additional File 3). It was not possible to clearly separate E-L stages 32 to 34 or 36 to 38 indicating that the transcription patterns occurring at these stages were similar to one another. However, stage 35, which corresponds to early post-véraison, could be distinguished suggesting that transcription patterns at this point in berry ripening are unique to this critical stage in berry development. Further analysis using a third axis explaining 2.7% of the overall variance, confirmed the previous results and slightly improved the resolution among stages 31, 35, and 36 to 38.

We used the Pavlidis Template Matching (PTM) algorithm 24, to divide the 4,151 TFR Unigenes into twenty gene groups or clusters. Specifically, twenty gene profiles of interest were selected [Additional File 4] to reflect major transcriptional patterns of development across E-L stages 31 to 38 (Figure 2). The PTM algorithm then classified the gene profiles into twenty groups via measurements of Pearson correlation: a correlation coefficient of greater than 0.75 was used to determine cluster membership. Six profiles showed a steady decline (profile groups 1 to 3) or increase (profile groups 9 to 11) in steady-state transcript abundance over time with distinctly different slopes. These six profile groups encompassed 63% of the Unigenes with a majority expressed in profiles 2 and 3 (31.9%) and profiles 9 and 11 (28%; Figure 2). Eight profiles had transient peak increases (profile groups 4 to 8) or decreases (profile groups 12 to 16) in transcript abundance at each of E-L stages 32 to 36. These transient profiles accounted for 22% of the Unigenes. A majority (68.2%) of these transiently expressed genes (profile groups 4 to 8 and 12 to 16) exhibited increased transcript abundance with the highest proportion within profile group 16 (E-L stage 36), followed closely by profile group 15 (E-L stage 35 around véraison), and profile group 12 (E-L stage 32) (Figure 2). Interestingly, genes with transient decreases early in berry development (profile groups 4 and 5) also exhibited large increases in transcript abundance during the later stages (E-L stages 36 to 38). The last four profiles (profile groups 17–20) were selected as having two peaks of expression between E-L stages 32 and 36 (Figure 2). Approximately 4.3% of transcripts had such "up and down" expression patterns (profile groups 17–20). Finally, Unigenes that did not match one of these profiles were grouped into a 21^st cluster (Figure 2), accounting for 11% of the total transcripts considered (profile group 21). Taken together, this analysis revealed that berry development is not only a progressive process, wherein the majority of genes exhibit a steady increase or decrease in expression across all stages of development (profile groups 1 to 3 and 9 to 11), but also a dynamic process, wherein a large number of genes exhibit large, transient changes in transcript abundance at specific times of development. Most notably, the last phase of berry development (Phase III, profile groups 14, 15 and 16) was the time when the largest number of genes (380 transcripts or 9.1%) exhibited transient increases in steady-state transcript abundance.

Functional categories were assigned to Unigenes with two-fold or greater changes in steady-state transcript abundance over the course of the seven developmental stages using the Munich Information Center for Protein Sequences (MIPS, ver. 2.0) catalog with annotations of the top Arabidopsis BLAST hits 25. Because we detected some errors in the functional annotation for some Unigenes, functional categorization of each Unigene were verified manually and corrected if necessary. Corrections were only performed for the 4,151 Unigenes that displayed a two-fold or greater change in expression [See Additional File 2: Table 4]. Functional annotations could be assigned to approximately 64% of transcripts (Figure 3A). An additional 23% of Unigenes had matches to genes with unknown functions or unclear classifications (unclassified), and 13% did not have a BLAST hit (no hit) in public, non-redundant (NR) databases. The relative distribution of Unigenes within each of nineteen functional categories was determined (Figure 3B). To facilitate a functional comparison of the three major stages of berry development, Unigenes from each of the profile groups were regrouped into the three major developmental phases to reflect the greatest degree of transcript abundance changes at each phase: Phase I (profiles 1, 2, and 3), Phase II (profiles 4, 5, 6, 12, 13, and 14), and Phase III (profiles 7, 8, 9, 10, 11, 15, and 16). Statistically significant differences in the distribution of genes within functional categories amongst these developmental stages were observed (Figure 3B; see Additional File 5: Tables 1 and 2). Functional categories that had a large number of transcripts in Phase I followed by a decrease in Phase III included biogenesis of cellular component (42), transport regulation (20), energy (2), and metabolism (1). This is consistent with the developmental aspects of this phase, which are characterized by cell division and expansion, which require a high level of metabolic activity. The process of cell division requires large quantities of structural materials and consumes energy, while cell expansion requires large quantities of solutes and water.

The opposite trend of increasing transcript abundance from Phase I to Phase III was observed for functional groups that included transcription (11), protein synthesis (12), protein fate (14), protein with binding function (16), and to a lesser extent with interaction with cellular environment (34). These trends served to further indicate the complexity of the transcriptional, translational, and interaction-based regulatory processes necessary for berry development.

In order to identify genes with potentially important roles in specific stages of berry development, transcripts with a dynamic pattern were identified from within the first 20 PTM algorithm-defined profile groups. The transcript profiles were examined in further detail (Figure 4).

Polyamines (PAs) are a class of compounds that have plant growth regulator activity. Their roles in cell division 26 and fruit set 27 have been widely investigated. Free, conjugated and wall-bound forms of polyamines accumulate mostly at anthesis before decreasing at fruit set in grapes 28. Two transcripts were detected that belong to profile 4, which are strongly down-regulated at E-L stage 32 (1618814_at, 1616399_s_at; AY174164, TC68466). Both are related to ornithine decarboxylase and arginine decarboxylase, which are involved in polyamine metabolism 29. These two genes located at the start of the PA pathway might play a role in providing precursors that would be used during Phase I of berry development.

In higher plants, the catabolism of asparagine (Asn) occurs by two routes. The first pathway involves the hydrolysis of Asn, releasing ammonia and aspartate by asparaginase activity. L-asparaginase is one of the enzymes for Asn utilization by plants that plays an important role in the nitrogen metabolism of developing plant tissues 30. One Unigene encoding L-asparaginase (1611257_a_at; TC51832) displayed a specific peak during E-L stages 32 (Figure 4A). This last result indicates that this enzyme could play a role during the first phase of berry development as a provider of ammonia for de novo protein synthesis in grape. This result is also supported by the significant transcript abundance of Unigenes encoding glutamate dehydrogenase or glutamine synthetase (data not shown, see Additional File 2: Table 4, 1607579_at, 1613697_at, 1609819_s_at) during the first phase of berry development. These enzymes participate in nitrogen assimilation in plants 31.

In grape berries, fruit softening occurs during Phase III and is largely affected by cell-wall loosening 32 and turgor 33. Xyloglucans account for about 10% of the cell wall composition in berries 32. In fruit, xyloglucan depolymerization is associated with fruit softening 34. Xyloglucan endotransglycosylases, which hydrolyze and transglycosylate xyloglucans, were encoded by multiple isogenes, the majority of which were expressed highly during Phase I in berry development (E-L stage 31), but then declined (data not shown; see Table 1). One Unigene (1618848_at; TC52577), however, which is a xyloglucan endotransglucosylase/hydrolase, displayed a 185-fold increase in expression during Phase II, peaking at E-L stage 33 (Figure 4A). This xyloglucan endotransglucosylase Unigene is closely related to a xyloglucan endotransglucosylase/hydrolase (SIXTH5) that can act reversibly. It has been characterized recently as a tomato xyloglucan depolymerase in vitro in the presence of xyloglucan oligosaccharides (XGOs) 35.

Expansins play important roles in cell wall loosening via non-enzymatic mechanisms and are involved in cell expansion 36. Most expansin genes displayed steadily increasing or decreasing patterns during berry development (see Table 1). Others showed peak expression around E-L stage 34 (α-expansin, 1608074_at, TC62965; α-expansin, 1608191_at, TC64448; Figure 4B). An expansin gene from strawberry, FaExp4, displays exactly the same peak transient expression pattern as these latter two genes at a comparable ripening stage as grape berries, called the White stage in strawberry fruits, just before red fruit color development 37. Thus, these expansins in grape berry may be required during the Phase III of grape berry development, when the second phase of cell expansion occurs.

Terpenes, which are precursors for important aroma compounds 38, accumulate at véraison 3940. One Unigene encoding a limonene cyclase (1613161_at; TC69794; Figure 4B), which is in the monoterpene pathway, is involved in the conversion of geranyl diphosphate into limonene 41. Limonene and some of its derived compounds such as menthol or 1,8 cineol are intimately associated with the "eucalyptus fragrance" of red wine 42. Accumulation of 1,8-cineole in wines is derived from precursors in grape, like limonene. The strong induction of our Unigene related to limonene cyclase (~40 fold from E-L stages 32 to 34) correlates well with the beginning of accumulation of 1,8-cineole in red grape samples 43. One Unigene (1618595_at, TC53841; Figure 4B) belonging to profile 15 and encoding alcohol dehydrogenase exhibited strong homology with an (-) isopiperitenol dehydrogenase, which is involved in the same monoterpene pathway 44. This transcript abundance of this Unigene is correlated to the expression of the limonene cyclase previously discussed above indicating a possible activation of these enzymes in the same metabolic pathway 44.

A number of plant growth regulators including abscisic acid (ABA), auxin (indole-3-acetic acid [IAA], brassinosteroids (BR), ethylene, and gibberellic acid (GA) have been implicated in the control of berry development and ripening. Therefore, steady-state transcript accumulation patterns of Unigenes with functions related to hormone biosynthesis and response were tracked over the course of berry development (Figure 5, Table 2).

ABA amounts in berries decrease after anthesis, but then increase significantly at véraison 45. External applications of ABA to ripening fruit can accelerate berry development (see 13 and references therein). The transcript abundance of 9-cis-epoxycarotenoid dioxygenase (NCED), which encodes the rate limiting step in ABA biosynthesis, increased during the lag phase and peaked at stage 35 around the start of véraison (Figure 5A). Both NCED1 (1608022_at, TC57089) and NCED4 (1607029_at, TC55541) had similar expression patterns, but differed significantly in their relative trancript abundance. A transcript (1614892_at, TC54474) encoding ABI1 (protein phosphatase 2C) showed an expression pattern like that of the NCED genes, but was more highly correlated with NCED4 than NCED1. The RD22 gene (1619802_at, TC67323), a dehydration-responsive protein, displayed a very large increase in abundance at véraison that continued to increase during berry maturation, whereas another transcript (1621346_at, TC65114) encoding an ABI3/VP1 (ABscisic acid Insensitive 3/ViviParous 1) transcription factor showed highest transcript abundance during the lag phase.

Traditionally, wine grape has been considered a non-climacteric fruit, however, there are studies that indicate that ethylene plays an important role in berry development and ripening 13 and is required for increased berry diameter and ripening processes, such as anthocyanin biosynthetic gene expression and accumulation 4647. In addition, ethylene appears to be involved in controlling the expression of an alcohol dehydrogenase gene from grape 48. Furthermore, some inhibitors of ethylene biosynthesis can delay berry ripening 49. Ethylene-related transcripts displayed some very unique and intriguing patterns of expression (Figure 5B) indicating that this signaling pathway is differentially expressed along berry development. One transcript (1617012_at, TC68057) encoding a putative Ethylene Response Factor 1 (ERF1), a putative ethylene output gene, displayed a steady increase in abundance with maximal expression at ripening (Figure 5B) indicating a potential post-véraison role for this signaling pathway. An ethylene-induced transcription factor (1619585_at, TC62897) exhibited transcript accumulation during the lag (E-L stages 32 to 34) and early véraison (E-L stage 35) stages of development. A putative ethylene co-activator (1621552_at, TC66829) protein displayed biphasic peak transcript abundance at E-L stages 32 and 35. The transcript abundance of ACC oxidase (1615952_s_at, TC56709), the enzyme responsible for the last step in ethylene biosynthesis, was highest at E-L stage 32, the start of the lag phase, and then declined throughout the remainder of berry development. Interestingly, the transcript abundance of an ethylene receptor ERS1 (1622402_at, TC62349) and EIN4/ETR5 (1618518_at, TC55908) were at their lowest during E-L stages 32 to 33 until véraison, but then increased at a later stage (E-L stage 38). Ethylene pathway activation in grape berry appears to occur within a three week period of berry development (weeks 6 to 8 after anthesis; E-L stages 30 to 33) when the highest ethylene (ACC) content and transcript abundance of ACC oxidase were detected in Cabernet Sauvignon 46. This hypothesis is supported by the observation that application of exogenous ethylene 8 weeks after anthesis hastened the ripening of the grape berries and resulted in a decrease in average cell size. In contrast, if the same ethylene treatment was applied during earlier stages of berry development (at 4, 5, 6 or 7 weeks), maturation was delayed 47.

According to the Arabidopsis model of ethylene signaling, reduced expression of transcripts and activity of receptors increases the sensitivity to ethylene, whereas increased receptor expression and activity decreases sensitivity 50. In tomato, the expression of most genes encoding ethylene receptors increases during fruit development. In parallel, high levels of ethylene are expressed to counterbalance the negative effect of increased receptor expression on the ethylene signaling pathway 51. In grape berry, the slight decreases observed in ethylene receptor transcript expression occurring between E-L stages 31 and 32 and the peak of ethylene accumulation during this same period, indicate a higher sensitivity to ethylene during the early stages of berry development. This would be expected to lead to a greater activation of the ethylene signaling pathway prior to véraison.

As in grape berry, strawberry is able to produce significant levels of ethylene during fruit development, but not to the same extent as climacteric fruits. Recently, three ethylene receptors have been identified in strawberry 52. Two of them (FaEtr1 and FaErs1) display the same pattern of expression during fruit development as those observed for ERS1 ethylene receptor during grape berry development. In addition, the highest rates of ethylene production in strawberry were detected in very young green fruits. Following this, the hormone decreases continuously until the White stage of fruits. Following this stage, ethylene showed a slight but steady increase for the remainder of development. When considered together, the similarities of expression of ethylene receptors during fruit development for both grapes and strawberries coupled with the concomitant ethylene production during the early steps of fruit development indicate a conserved mechanism for ethylene perception between these fruits prior to ripening.

Brassinosteroids (BR) have recently been implicated in playing an important role in berry development 53. Castasterone concentrations are low during the early stages of berry development and then increase at véraison 53. Brassinosteroids have been shown to increase cell size 54 indicating that berry enlargement may be affected by castasterone levels. BRH1 RING finger protein (1617572_at, TC66046) transcript abundance, which is known to be down-regulated by exogenous application of BR, decreased during E-L stages 31 to 35, but increased in fully mature berries (Figure 5C). The transcript abundance of the Brassinosteroid Receptor 1 gene (BRI1, 1612516_at, TC56501) peaks at the start of the lag phase (E-L stage 32) and then declines thereafter. The transcript abundance of BRU1 (1608945_at, TC54729), which is a BR-responsive transcript encoding a xyloglucan endotransglycosylase (XET), showed a transient increase in abundance at véraison. In the same family, transcripts for another BR-responsive protein (1619068_at, TC60314) declined with berry development. Clearly, there are many significant changes in transcript abundance that are associated with brassinosteroid responses during berry development.

Very little is known about the role of gibberellin (GA) in grape berry development except a possible role in cell enlargement. Biologically active concentrations of GA are high in flowers and in fruits just after anthesis, but then drop to lower levels over the course of berry development 5355. There is a second peak of active GA at the start of the lag phase and it is 77 times higher in the seed compared to the berry mesocarp 56. The transcript abundance of two putative GA receptors, GIDL1 and GIDL2 (1618181_at, TC67464; 1620071_at, TC56624, respectively), increased during berry development (Figure 5D). Interestingly, the transcript abundance of the GA signaling pathway repressor, GAI1 (1606777_s_at, TC56894), declines transiently at véraison. The transcript abundance of a putative GA β-hydroxylase (1610610_at, TC66284) declines over the course of berry development (Figure 5D) more or less coincident with the known accumulation pattern of GA₁ in developing berries.

The mechanisms by which the phytohormone indole-3-acetic acid (IAA) regulates berry development are complex and not fully understood. Increased auxin production produced through the action of an ovule-specific auxin-synthesizing transgene enhanced fecundity in grapes 57. Earlier reports indicated that auxin concentrations were high during early Phase I and declined following véraison 55 consistent with the role of this phytohormone in promoting cell division and expansion during Phase I. Treatment of grape berries with synthetic auxin-like compound, benzothiazole-2-oxyacetic acid (BTOA) delayed ripening 45. A more recent study showed that auxin concentrations remain relatively constant over the course of berry development 53.

Our data indicate that there are numerous transcript responses to auxin (Figure 5E). The transcript abundance of Aux22 (1614660_at, TC53887), which forms heterodimers with auxin response factors (ARF) in order to repress auxin responses, increased after véraison (Figure 5E). Transcripts for both Auxin Response Factor 2 (ARF2, 1613813_a_at, TC65541) and a Small Auxin Up RNA protein (SAUR) (1609591_at, TC63193) increased after véraison, whereas transcripts for a different SAUR transcript (1606566_at, TC62299) and an Auxin-induced Response Factor, ARF18 (1616225_at, TC52772) both declined in a very similar pattern during berry development. A transcript (1619610_at, TC56575) encoding IAA-amino acid hydrolase, which is involved in IAA homeostasis, was highly expressed during the later stages of berry development (Figure 5E). The synthesis and hydrolysis of IAA conjugates, which function in both permanent inactivation and temporary storage of auxin 58, may play an important role in the control of IAA concentrations as berry development progresses. IAA-amino acid hydrolase may provide for local concentrations of auxins within the berries to promote mesocarp cell enlargement. Several transcripts (1611479_at, CD799903; 1617179_at, CF414958; 1610034_at, TC59892) related to auxin transport and perception also displayed increased abundance at the onset of véraison. Given the importance of auxin-mediated processes in developing berries, more research needs to be conducted to elucidate the mode of action of auxin signaling and response pathways.

Methyl jasmonate (MeJA) is known to promote the synthesis and accumulation of terpenes and resveratrol in berry cell cultures 5960, however, its effects in vivo are not well understood. The transcript abundance of 12-oxophytodienoate reductase (12-OPR) (1607601_at, TC61395), which is involved in jasmonate biosynthesis 61, and a constitutive pathogen-response 5 protein (1614324_at, CF213899), both decreased with berry development (Figure 5F). Less is known about the role of cytokinins in berry development. The transcript abundance of cytokinin oxidase (1620306_at, TC69712), which degrades cytokinin 62, decreased over berry development, whereas a known cytokinin-response regulator, a Type-A response regulator (1612955_at, TC52530), showed a steady increase in transcript abundance over berry development (Figure 5F).

Calcium has many essential roles in plant growth and development 63, however, the role of calcium signaling in grape berry development is largely unexplored. Recently, an ABA-responsive calcium-dependent protein kinase (CDPK) was described that was specifically expressed in the seed and flesh of berries with increased transcript abundance over berry development and ripening 64. In the current study, a large number of genes with functions related to calcium sequestration, transport and signaling were found to display developmentally regulated expression patterns (Figure 6A; Table 3).

Calcium homeostasis within the cytosol is tightly controlled by membrane spanning Ca²⁺-ATPases and H⁺/Ca²⁺ exchangers, which typically maintain low concentrations of Ca²⁺ in the cytosol and restore this concentration following signaling-related transient changes in calcium levels. Transcripts encoding plasma membrane Ca²⁺-ATPase genes (1614028_at, TC62785; 1622073_at, CF404214), which are closely related to ACA8 and ACA13, respectively, in Arabidopsis thaliana, showed increased transcript abundance during E-L stages 33 and 34 and in later developmental stages. Interestingly, ABA markedly and rapidly stimulates the expression of the ACA8 gene in cell cultures of Arabidopsis thaliana 65. A tonoplast Ca²⁺/H⁺ exchanger (1617237_s_at, TC66680), which is a close homolog of CAX3 from A. thaliana and plays a key role in cytosolic Ca²⁺ homeostasis 66, showed a transient increase in transcript abundance at E-L stages 34, indicating a possible role for calcium signaling around véraison.

ABA accumulates until two weeks after the beginning of véraison before decreasing later in berry development 67. Thus, it is likely that ABA is directly or indirectly involved in the control of Ca²⁺ signaling and homeostasis events, particularly around véraison.

The increased expression of several Unigenes encoding calmodulin or calcium interacting protein kinases (see Table 3) supports this hypothesis 68. One Unigene encoding a calmodulin-related suppressor of gene silencing (1618587_at, TC64370) displayed a pronounced pattern with two peaks of expression at E-L stage 32 and at E-L stage 35 corresponding to two transitions of berry development (Phases I to II and Phases II to III). This Unigene displayed a 10-fold change in its transcript abundance across berry development and may be involved in the suppression of posttranscriptional gene silencing (PTGS) by interacting with a proteinase known to suppress PTGS in plants 69. This correlation indicates a possible role for calcium in regulating the activity of the PTGS mechanisms. To date, only one paper reported the possibility of the involvement of PTGS in the regulation of gene expression during plant development 70. Further investigations are necessary to evaluate the real impact of this Unigene in the triggering of véraison.

Phenolic compounds, derived from flavonoids (anthocyanins, tannins and flavonols), are the major wine constituents responsible for organoleptic properties such as color and astringency. Twenty-one Unigenes encoding biosynthetic enzymes of the general phenylpropanoid and flavonoid pathways were found to exhibit differential mRNA expression patterns across berry development (Table 4). The vast majority of these genes are expressed predominantly in the skin 71.

The mechanisms by which anthocyanins accumulate in the vacuole of grape berry skin cells during Phase III are not fully understood. These compounds must be transported from the site of synthesis in the cytosol to their final destination, the vacuole. Several models have been proposed for sequestering anthocyanins in the vacuole in Arabidopsis thaliana. One model 72 indicates the action of a glutathione-S-transferase (GST) in facilitating the transfer of anthocyanins into the vacuole. Another model indicates that a transporter of the multidrug-resistance-associated protein family could facilitate the transport of an anthocyanin-GST complex into the vacuole 73. Here, the Unigene transcript encoding a GST (1619917_s_at, TC69505; Figure 6B) displayed a 63-fold increase in abundance during the stages in berry development in which flavonoids accumulate (Figure 6B). This Unigene is closely related to a GST homolog known to be involved in anthocyanin sequestration 74. This Unigene also displays a skin-specific expression pattern 71, which is consistent with the tissue localization of anthocyanins. A Unigene homologous to a glutathione-S-conjugate transporting ATPase (1609870_at, TC58286) showed a peak of expression at véraison (E-L stage 34). While not yet characterized in detail, this Unigene belongs to the ABC transporter sub-family, members of which are known to transport anthocyanins 74. The putative multi-drug transporter (1607560_at, TC62162), which is known to be involved in the sequestration of tannins into vacuoles 75, exhibited peak transcript abundance at E-L stage 32 followed by a decline, and is consistent with the pattern of maximal tannin accumulation that occurs a few weeks before véraison.

Specific members of the MYB transcription factor family play critical roles in the regulation of flavonoid metabolism during grape berry development 76. We detected four transcripts encoding MYB transcription factors that have been previously characterized in grape berry (see Table 4) 77787980. VvMYBPA1 (1611091_s_at, TC54724) regulates the proanthocyanidin (condensed tannins) pathway in the grape berry 77. In the Shiraz cultivar, VvMYBPA1 peak expression appears to occur during E-L stages 34 and 35 in the skin and seeds, whereas, in Cabernet Sauvignon this gene is expressed at an earlier developmental stage (E-L stages 32) (Figure 6B). Such differences are likely to be cultivar-dependent. In the same way, the MYC family of transcription factors also plays a key role in regulation of the anthocyanin pathway. One MYC transcription factor transcript (1618504_at, TC61713), which shares strong amino acid sequence identity with MYC genes known to be involved in the regulation of anthocyanin production 81, displayed a pattern of transcript accumulation that decreased from the beginning of berry development until E-L stage 35 and then increased for the remainder of fruit development (Figure 6B). Furthermore, this Unigene is preferentially expressed in the skin 71. These expression patterns correlate well with the accumulation of anthocyanins and proanthocyanins.

In grape berries, volatile aroma compounds, such as terpenes, benzenoids, and phenylpropanoids, accumulate in exocarp and mesocarp tissues following the initiation of berry ripening 3882. Three transcripts (Figure 6C) encoding O-methyltransferases, which may participate in the biosynthesis of volatile compounds, were also detected 83. The first Unigene (1608603_at, TC56956), which encodes a putative phloroglucinol O-methyltransferase, is involved in the biosynthesis of volatile 1,3,5-trimethoxybenzene, a compound not previously described in grape 83, displayed a very high transcript abundance at the beginning of berry development (E-L stage 31) before decreasing after véraison until E-L stage 36 and then increasing again in mature berries (Figure 6C). The second Unigene (1613542_at, TC62584) was expressed at E-L stage 31, but then declined. The third Unigene (1620469_at, CF209780) displayed very low transcript abundance with a slight increase following véraison (Figure 6C). Finally, two S-adenosyl-L-methionine (SAM):salicylic acid carboxyl methyltransferases were identified with developmentally-induced expression patterns. The first Unigene (1616348_at, TC52353) showed a broad peak of expression between E-L stages 32 to 35, whereas the second Unigene (1612552_at, TC57170) showed increased transcript abundance after véraison (E-L stage 34) (Figure 6C). Such genes are thought to play important roles in scent production or plant defense 84. Little correlation between the level of sequence similarity and the structural similarity of their substrates has been observed for most of these protein families, so that gene functions have to be assigned following detailed biochemical testing 85.

The acid:sugar balance at harvest is an important factor of wine quality as it affects important sensory attributes 15. Two major organic acids that contribute to titratable acidity, tartrate and malate, are the most abundant organic acids in grapes and reach maximal concentrations around the end of Phase I (E-L stage 32; see Table 5). Tartrate concentrations were found to peak at E-L stage 32 and then declined steadily until harvest, E-L stage 38 (Figure 7A). Tartrate concentrations decreased in parallel with three different transcripts encoding L-idonate dehydrogenase (1622252_at, TC52651; 1613165_s_at, TC52651; 1612918_at, TC52651), a key enzyme in tartrate biosynthesis 86. The innermost region of the berry pulp surrounding the seed has been shown to contain the highest tartrate concentrations 87. Consistent with this observation, tartrate synthase transcripts have been shown to be more abundant in seeds than in outer mesocarp and skin tissues 71.

Like tartrate, malate concentrations peaked around E-L stage 32, but then declined more rapidly than tartrate during berry ripening (Figure 7B). In contrast to the good correlation between tartrate and L-idonate dehydrogenase transcript abundance, there is a less obvious correlation between malate concentrations and the transcript abundance of Unigenes encoding malate dehydrogenases (Figure 7B). Transcript abundance for two isogenes encoding cytosolic NAD-dependent malate dehydrogenases (1612546_at, TC68207; 1609147_at, TC55437), which catalyze the interconversion of malate to oxaloacetate, increased during ripening. Transcripts for mitochondrial isoforms of the enzyme (1622059_at, TC60439; 1617448_at, TC54982) also increased over this same time period. In contrast, the transcript abundance of a NADP-dependent malic enzyme (1609345_s_at, TC57092), which catalyzes the oxidative decarboxylation of malate to pyruvate, declined slightly from E-L stages 34 to 36, but then increased by stage 38 (Figure 7B). The slight increase in the expression of all of these enzymes together may contribute to the declining concentrations of malate during ripening. Very little is known about the mechanisms of malate transport processes in the phloem/xylem and within developing grape berries. The regulation of malate concentrations in berries appears to be quite complex. More research is needed to elucidate this well known developmental process.

Mature berries contain unusually high concentrations of free proline; proline being the most abundant amino acid in Cabernet Sauvignon 8889. Proline concentrations increased significantly at véraison and remained high until berries were fully ripe (Figure 7C). Transcripts encoding pyrroline-5-carboxylate synthetase (1619565_at, TC52705), the key regulatory enzyme in proline biosynthesis, remained relatively constant with a small peak of expression occurring at E-L stage 35 (Figure 7C). Proline dehydrogenase transcripts (1617293_s_at, BQ792635), which encode the first enzymatic step in proline catabolism, increased only during the latter stages of berry development. These mRNA expression patterns are consistent with earlier reports and with protein expression patterns of these enzymes 88. Proline accumulation correlated poorly with steady-state transcript and protein abundance changes for these two enzymes indicating that proline production is regulated by posttranslational mechanisms 88. Steady-state transcripts encoding a proline transport protein (1610800_at, CK906448) also increased in conjunction with proline abundance.

Sugar accumulation in grape berries has been well studied because sugar content is a key factor in producing wine. In contrast to organic acids, hexose sugars (i.e., Glc and Fru) begin to accumulate substantially in the lag phase (Phase II) and continue thereafter. In grapevines, carbohydrates produced during photosynthesis are exported from the leaf as sucrose and transported in the phloem to the berry cluster 9091. Prior to véraison, most sugars imported into the berries are metabolized with little if any storage of these compounds. Following véraison, however, sugars accumulate in the vacuole to high levels in the form of glucose and fructose following the enzymatic cleavage of sucrose (mainly in the apoplast, but also in the cytoplasm and vacuole). Monosaccharide transporters direct the transport of these sugars through different organelles 92.

In the berries in this study, fructose was more abundant than glucose; in contrast sucrose concentrations remained relatively low and constant throughout berry development (Figure 8A). Transcript abundance for the Unigene encoding sucrose synthase (1609402_at, TC62599), increased gradually over berry development consistent with increased hexose accumulation in the berry. This Unigene has high homology with the sucrose synthase (CiTSUSA) in Citrus unshiu 93. CiTSUSA also increases with fruit development and catalyzes the reaction in the cleavage direction (sucrose to UDP-glucose and fructose). Komatsu et al. 93 suggest that the action of this gene may be important for sink strength.

Sucrose-6-phosphate phosphohydrolase (SPP) (1608257_at, TC68135), which catalyzes the last step in sucrose synthesis, showed a slight increase in transcript abundance after E-L stage 32 and then remained relatively constant throughout the remainder of berry development (Figure 8B). In contrast, the transcript abundance of two vacuolar invertases, GIN1 and GIN2 (1611613_at, TC60693; 1612836_at, TC57719), which catalyze the catabolism of sucrose to fructose and glucose, declined over the course of berry development (Figure 8B), consistent with an earlier report 94. The mRNA expression of these two invertases is consistent with the early increases in sugar accumulation during Phase II (E-L stages 32 to 34). On the other hand, transcript abundance for a neutral invertase (1620628_at, TC67908) and a cell wall acid invertase (1611027_at, TC56057) remained relatively constant during the course of berry development consistent with earlier reports on the amount and activity of these enzymes in developing berries 95. In grape berries, sucrose cleavage is largely catalyzed by cell wall bound invertases 95. Sucrose cleavage is usually associated with cell wall invertase activity at the onset of ripening, together with a shift towards apoplastic phloem unloading of sugars in berries during this same period of time 95. Finally, transcripts encoding fructokinase (1616255_at, TC57339), which catalyzes the formation of fructose-6-phosphate and may regulate starch formation, declined in abundance in a similar manner as GIN1 and GIN2 following a peak of expression at E-L stage 32.

In most sink cells, sucrose is either cleaved by invertase into glucose and fructose or degraded by sucrose synthase into uridine-5'-diphosphate (UDP) glucose and fructose for subsequent metabolism and biosynthesis 9697. Cell wall invertases appear to play the main role in the cleavage of sucrose during Phase III of berry development 95. However, the increase in sucrose synthase during Phase III of berry development indicates that this isogene may participate in the catabolism of sucrose to fructose and glucose. Alternatively, this sucrose synthase isogene may play a critical role in cellulose synthesis associated with Phase III cell expansion similar to its role in cotton fiber elongation 98. Two cellulose synthase isogenes (1607069_at, TC53461; 1611149_at, TC56091) displayed increased transcript abundance during Phase II and III, consistent with this hypothesis (see Table 1). Additional developmentally regulated transcripts related to carbohydrate metabolism and transport are summarized in Table 6.