EFFECT OF VINE TRAINING SYSTEM ON THE PHENOLIC COMPOSITION OF RED GRAPES ( VITIS VINIFERA L . CV . XINOMAVRO )

1: Department of Food Science & Human Nutrition, Agricultural University of Athens, 75 Iera Odos, 11855 Athens, 2: Laboratory of Viticulture, School of Agriculture, Aristotle University of Thessaloniki, 54124 Thessaloniki, 3: University of Bordeaux, ISVV, EA 4577, Œnologie, 210 Chemin de Leysotte, 33140 Villenave d’Ornon, France 4: Department of Animal Sciences and Aquaculture, Agricultural University of Athens, 75 Iera Odos, 11855 Athens 5: Kir-Yanni Estate, 59200 Yianakochori Naoussa, Greece


INTRODUCTION
Phenolic compounds represent one of the most important groups of compounds for red wine as they affect directly wine quality.Proanthocyanidins (condensed tannins) are polymers composed of flavan-3-ol subunits.They are located in grape skins and seeds and are responsible for the stabilization of the colour and the sensory characteristics of the wines due to their astringent and bitter properties (Ribéreau-Gayon et al., 1999;Chira et al., 2009;Sun et al., 2013).
The level of astringency is related not only to the concentration of tannins but also to their degree of polymerization.Astringency increases with increasing tannin size, at least up to 8 or 10 flavanol units (polymerized procyanidins are increasingly reactive with proteins and, therefore, have an increasingly astringent character), and then decreases since they become either no longer soluble or too bulky to bind with the proteins (Chira et al., 2009;Sun et al., 2013).On the contrary, tannin polymerization mediated by ethanol softens their flavour (Chira et al., 2012).In addition, the molecular size of proanthocyanidins affects their bitterness, since monomers are more bitter than polymers (Peleg et al., 1999;Ribéreau-Gayon et al., 1999).Astringency is also reported to increase with the degree of tannin galloylation (Cosme et al., 2009;Chira et al., 2011;Ćurko et al., 2014).Galloylation has been shown to increase tannin interactions with various proteins, suggesting that it could be responsible for increased astringent sensation (Cheynier et al., 1997).
In order to characterize tannin structure, extraction from grapes with the use of specific solvents should be applied, following tannin fractionation in accordance to their molecular size.Since tannin isolation and individual analysis are rather difficult procedures, depolymerization is often employed to facilitate their characterization.Treatment of condensed tannins with acid, in the presence of a nucleophile such as phloroglucinol (Prieur et al., 1994;Souquet et al., 1996), allows subunit profiling by HPLC and calculation of the average molecular mass, expressed as mean degree of polymerization (mDP).
Extensive research has been conducted in order to determine tannin composition and structure using the depolymerization method.The results showed that grape tannins derived from skins and seeds vary in length, subunit composition and sensory properties (Peleg et al., 1999).Seed tannins are shorter, with a lower mDP, and display a higher percentage of subunits bearing gallic acid esters, which is expressed as percentage of galloylation (%G) (Prieur et al., 1994).Skin tannins are generally larger with a higher mDP (Souquet et al., 1996).However, seed and skin proanthocyanidins were found to be equally astringent when tasted at the same concentration in a wine or buffer medium, despite their rather large compositional differences (Brossaud et al., 2001).
Xinomavro (Vitis vinifera L.) is an indigenous Greek red grape variety cultivated almost exclusively in Northern Greece vineyards producing Protected Denomination of Origin (P.D.O.) red wines.Grapes and wines produced from this variety are characterized by high acidity and phenolic richness resulting in a long ageing potential of the wines.However, wines made from Xinomavro grapes are particularly poor in anthocyanins despite being rich in skin and seed flavan-3-ols (Kallithraka et al., 2006).Therefore, young wines are typically pale in colour with «thin» tannins, making Xinomavro vinification a challenging task for winemakers.Although grape and wine tannin composition is of great technological significance for red wines, for the major wine grape varieties cultivated in Greece, such as Xinomavro, tannin mDP has not been examined, although some investigations have shown that grape variety genotype influences proanthocyanidin composition and mDP of skins and seeds (Chira et al., 2009;Cosme et al., 2009;Mattivi et al., 2009).
Several strategies for manipulating the phenolic composition of red wines have been attempted either during grape growing or wine making.The choice of the vine training system, i.e. the arrangement of vine organs in space with the use of an adapted trellis, is one the major techniques applied by grape growers to influence source-sink relationships and fruit-zone microclimate, with the aim to control grape maturity and composition (González-Neves et al., 2004;Reynolds et al., 2004;Pérez Lamela et al., 2007;Reynolds et al., 2009;Río Segade et al., 2009;Mota et al., 2011).In previous studies it was reported that the phenolic composition of grapes and wines differed between vertical shoot positioned (VSP) and divided canopy systems (Katerji et al., 1994;González-Neves et al., 2004;Pérez Lamela et al., 2007;Río Segade et al., 2009) and also among VSP systems (Río Segade et al., 2009).However, to the best of our knowledge, there have been no studies concerning the effect of training system on the structural characteristics of skin and seed proanthocyanidins such as mDP and %G.
Therefore, the aim of the present work was to investigate the effect of training system on the phenolic composition of Xinomavro grapes and wines.For this purpose, a divided canopy system (Lyre) and a vertical shoot positioned trellis with two different pruning systems (Royat and Guyot) were studied in the vineyard.The phenolic characterization of Xinomavro grapes was performed with respect to the following three analytical parameters: anthocyanin content, tannin composition and proanthocyanidin structure.

Experimental conditions
The study was carried out during the 2010 season in a 7-year-old commercial vineyard in Naoussa, Northern Greece (40°39'32''N 22°4'21''E, 250 m a.s.l.), planted with Vitis vinifera L. cv.Xinomavro (clone V6) grafted onto 110R rootstock.The vineyard was located on a clay loam soil (pH 7.8) and rows were east-west oriented.A divided canopy system (Lyre) and a vertical shoot positioned trellis with two different pruning systems (Royat and Guyot) were applied at 2970 (2.30 × 1.45 m), 3970 (2.30 × 1.10 m) and 3970 (2.30 × 1.10 m) plants/ha, respectively.The Lyre trellis was established with two vertical curtains, 45 and 80 cm apart at the lowest and highest point, respectively; the lowest fruiting wire was set at 55 cm from the ground, and foliage wires at 80, 110 and 145 cm.For the Royat and Guyot trellis, the vines were trained to a single curtain, while the fruiting wire was set at 55 cm from the ground and foliage wires at 85, 110 and 135 cm.In the Lyre treatment, vines were pruned to 10 spurs of 2 buds per vine.The Royat system was composed of a double spur-pruned permanent cordon with 5 spurs of 2 buds per vine, while for the Guyot system, a single cane of 10 buds and one spur of 2 buds were retained.The three treatments were replicated 3 times in a randomized block design, with 20 vines per replication.Canopy management was similarly practiced throughout the growing season and included shoot tucking and positioning and shoot topping to about 30 cm over the top wire, once after fruit set.No leaf removal was performed in any of the three treatments.All treatments were equally irrigated, receiving 42 L per vine during the season.
Vine vigour was assessed by a nondestructive estimation of leaf area per vine at harvest according to the method of Lopes and Pinto (2000).Harvest was conducted simultaneously for the three treatments on 28/08/2010 according to technological maturity surveys (must °brix), and yield per vine was estimated.Samples of 500 berries were collected from each replicate and kept frozen at -20ºC until analysis.

Phenolic content of whole berries
50 berries from each replicate were homogenized using Ultra Turrax T25 at 24,000 rpm for 1 min.Total phenol and anthocyanin content was measured according to Iland et al. (2004).Briefly, 1 g of the homogenate (in triplicate) was transferred into a centrifuge tube and mixed with 10 mL 50% v/v aqueous ethanol (pH 2) for 1 h.After centrifugation at 3500 rpm for 10 min, 0.5 mL of the supernatant was added to 10 mL 1M HCl and mixed thoroughly for 3 h, then absorbance at 520 nm and 280 nm was recorded.
Total and extractable anthocyanins in the juice were determined as described by Ribéreau-Gayon et al. (1999), with slight modifications.An amount of 20 g of the homogenate was macerated for 4 h with two different buffers (pH 1 and pH 3.6).Then, the macerated samples were centrifuged (4000 rpm, 10 min) and the anthocyanin and total phenolic content was measured on the supernatant.

Extraction of phenolic compounds from grape seeds and skins
Seeds and skins of 150 berries per replicate were removed manually from grapes, freeze-dried and finally ground to powder.The extraction of skin and seed tannins was carried out according to previously reported methods (Chira et al., 2011;Lorrain et al., 2011).Briefly, 3 g of the obtained powder was first extracted with 25 mL of acetone/water (80:20, v/v) for 3 h and then with 25 mL of methanol/water (60:40, v/v) for 2.5 h.The supernatants were combined and evaporated under reduced pressure at 30ºC to remove organic solvents; the residue was dissolved in water and lyophilized to obtain a crude tannin extract.

Grape total phenolics, tannin content and antioxidant activity
Part of the crude extract was re-dissolved in a model solution (12% ethanol; 5 g/L tartaric acid; pH 3.5 adjusted with 1N NaOH) for the determination of total phenol content (TPC) by the Folin-Ciocalteu method (Waterman and Mole, 1994), antioxidant activity (Brand-Williams et al., 1995), and chemical astringency estimation through the protein precipitation assay using BSA (Harbertson et al., 2003).Absorbance measurements were recorded on a Jasco V-530 UV/VIS spectrophotometer.

Preparation and analysis of skins and seeds
The remaining crude extract of seeds and skins was dissolved in a 5% aqueous ethanol solution and extracted thrice with chloroform to remove the lipophilic material (Chira et al., 2011;Lorrain et al., 2011).Then the aqueous phase was extracted thrice with ethyl acetate.Organic and aqueous fractions were collected separately; the organic fraction was concentrated under reduced pressure at 30ºC and lyophilized to obtain a dry powder.The organic fraction contained monomeric and oligomeric proanthocyanidins, while the aqueous fraction contained polymeric tannins (Chira et al., 2011;Lorrain et al., 2011).
The organic fraction of seeds and skins was analyzed according to Kallithraka et al. (2006) for the determination of (+)-catechin (C), (-)-epicatechin (EC), epicatechin gallate (ECG), epigallocatechin gallate (EGCG), gallic acid (GA) and procyanidins B1, B2 and C1.The equipment used for HPLC analysis consisted of a Jasco AS-1555 intelligent sampler, a Jasco PU 2089 Plus quaternary gradient pump, a Jasco MD-910 multiwavelength detector and a Jasco LC-Net II / ADC.The column was a Waters Nova-Pak C18 (150 x 3.9 mm, 4 μm).Identification was based on comparing the retention times of the peaks detected with those of original compounds, and on UV/VIS spectral data.Quantification was performed by establishing calibration curves for each compound determined, using the standards.Results were expressed as mg per g dry weight.All analyses were performed in triplicate.
Tannin mDP and %G were determined in both organic and aqueous phases of seed and skin extracts.Tannin extracts were re-dissolved in methanol and were left to react with phloroglucinol solution (50 g/L phloroglucinol, 10 g/L ascorbic acid, 0.1N HCl, in methanol) for 20 min at 50ºC.The reaction was ended by addition of aqueous sodium acetate (40 mM).LC/MS and HPLC analysis were performed for the identification and quantification of phloroglucinol adducts and terminal units (Kennedy and Jones, 2001;Chira et al., 2009;Lorrain et al., 2011).Reaction products were analyzed by LC/MS on a Shimadzu LC/MS 2010A coupled to a single quadrupole mass spectrometer equipped with an electrospray ion source.The mass spectrometer was operated in positive-ion mode.The source's temperature was set at 70ºC, the capillary voltage at 3.5 kV and the cone voltage at -30 eV.The absorbance was recorded at 280 nm and mass spectra were recorded in the range of 50-1500 amu.Separation was performed on a reversed-phase Waters XTerra RR C18 (100 x 4.6 mm, 3.5 μm) column at a flow rate of 0.5 mL/min, using a 20-μL injection volume and the following elution program: eluent A from 80% to 40% in 20 min, which was kept isocratic for further 10 min and then from 40% to 80% in 2 min.Eluent A was 0.1% aqueous acetic acid and eluent B methanol.All analyses were performed in triplicate.The following extension proanthocyanidin units were identified: (-)epigallocatechin-phloroglucinol, EGCe-P; (+)catechin-phloroglucinol, Ce-P; (-)-epicatechinphloroglucinol, ECe-P; and (-)-epicatechin gallatephloroglucinol, ECGe.(+)-Catechin, Ct; (-)epicatechin, ECt; and (-)-epicatechin gallate, ECGt were present as terminal tannin units.

Analysis of anthocyanin extracts
Anthocyanins were extracted with acidified methanol (0.1% HCl 12 N) from 1 g of dried skin powder at three successive times (4, 18 and 24 h).After centrifugation, the supernatants were combined and analyzed for total anthocyanins (Ribéreau-Gayon et al., 1999) and antioxidant activity (Brand-Williams et al., 1995).HPLC analysis was carried out for the determination of monomeric anthocyanins on a Restek pinnacle II C18 (250 x 4.6 mm, 5 μm) column at a flow rate of 1 mL/min, using a 10-μL injection volume, detection at 520 nm, and the following elution program: 90% eluent A for 1 min, then from 90% to 50% in 22 min and from 50% to 5% in 10 min, which was kept isocratic for further 2 min.Eluent A was 10% aqueous formic acid and eluent B methanol.Identification was based on comparing retention times and UV spectra of the peaks detected with those of original compounds.Malvidin-3-Oacetylmonoglucoside (MlvAc) and malvidin-3-(6-Op-coumaroyl) monoglucoside (MlvCoum) were tentatively identified based on previous observations (Arnous et al., 2002;Kallithraka et al., 2005).
Results were expressed as mg Mlv per g dry skin weight.All analyses were performed in triplicate.

Vinification and wine analysis
About 20 kg of the total harvested grapes per vineyard cell were vinified separately.For the winemaking trials, grapes of each replicate were crushed, destemmed and supplemented with 50 mg/L SO 2 (as potassium metabisulfite).Pectolytic enzymes (Safizym Colour, Fermentis, France) at 3 g/hL as well as lyophilized yeasts of the commercial strain SC 22 (Fermentis, France) at 20 g/hL, previously hydrated in water (15 min, 38ºC), were added.Beginning on the second day of fermentation, and for the following days, two punching downs per day were conducted to extract phenolic compounds.After 5 days of maceration, the wines were drained and transferred to other tanks and spontaneous malolactic fermentation was completed after approximately 3 weeks.The wines were racked, supplemented with 50 mg/L SO 2 (as potassium metabisulfite), filtered and bottled until analysis.

Statistics
Data were subjected to one-way analysis of variance (ANOVA), using Statistica V.7 software (Statsoft Inc., Tulsa, OK).Comparison of mean values was performed using Tukey's HSD test when samples were significantly different by ANOVA (p < 0.05).

Vine growth and yield components
The Lyre system significantly increased all vegetative growth components, suggesting that the increased number of retained buds resulted in a higher number of shoots and increased leaf area at the end of canopy growth, compared to the other systems, whether expressed on a per vine, per meter of row or per ha basis (Table 1).Canopy division has been reported to increase canopy volume and surface area compared to non-divided VSP systems in previous studies (Smart, 1985).difference in vigour components between Royat-and Guyot-trained vines.
Bud fruitfulness (clusters per bud) was higher in Royat and lowest in Lyre vines (Table 1).Larger systems (i.e.Lyre) usually tend to induce higher fruitfulness (i.e. higher number of bunches per shoot) due to improved microclimate in the fruiting zone (Swanepoel et al., 1990).Nevertheless, fruitfulness also depends on the balance between shoot and fruit growth components, primarily controlled at winter pruning by retaining a fixed number of nodes according to the vines' vigour potential.The lower fruitfulness of the Lyre system compared to Royat in our experiment may be related to the higher (i.e.double) number of buds left at pruning in the Lyre system, which could have affected source-sink relationships and, thus, flower initiation in developing latent buds.Decline in fruitfulness in response to increased node number has also been reported by Archer and Fouché (1987).
Lyre vines exhibited higher yield per vine, per meter of row and per ha than the Royat and Guyot vines (Reynolds et al., 2004) (Table 1).Changes in yield can be driven by a combination of changes in the number of clusters per vine and cluster weight.In the conditions of our study, the number of clusters per meter of row and per ha was not affected by training system, while cluster weight was higher in the Lyre system.It is therefore possible that the higher weight of the Lyre clusters could be mostly responsible for the higher yield observed in the Lyre system compared to the other treatments.However, our oneyear data are not sufficient to draw a more robust conclusion on the effects of training system on yield components and their relative contribution to yield variations.
Berry component weights were similar among the three treatments (Table 1).Berry weights ranged from 1.58 g to 1.68 g, while the skins and seeds represented 10.26% and 2.66% of total grape weight, respectively.This finding suggests that the higher cluster weight in the Lyre system was probably the result of a higher number of berries per cluster.
Previous studies also report that changes in cluster weight are mostly a response to changes in berry number rather than in berry weight (Bennett et al., 2005).
Despite of the increased average yield for Lyre, crop loads (i.e.leaf area to yield ratio) were highest in Lyre vines by 18% and 10% compared to Royat and Guyot vines, respectively (Table 1).It has been established that a range of 0.7-1.4m 2 of exposed leaf area per kg of fruit weight is necessary for full ripening (Howell, 2001)  though only total leaf area was measured, the divided canopy of the Lyre system would be expected to have a higher proportion of exposed leaves than the other two systems (Reynolds and Vanden Heuvel, 2009) therefore canopy surface area to yield ratio would be further improved as compared to VSP systems.

Anthocyanin composition of grape skins
Grapes derived from the Lyre training system contained significantly higher amounts of total and extractable anthocyanins, compared to the other treatments (Table 2).However, no significant differences were recorded for anthocyanin extractability values (%AE), which are an index of grape phenolic maturity in red grapes (Saint-Cricq de Gaulejac et al., 1998).Significant differences were also found regarding the levels of individual anthocyanins in skin extracts, with highest values observed in grapes originating from the Lyre vines (10.5 mg/g dw in sum compared to 8.3 mg/g dw for Royat and Guyot) for all anthocyanin-3-Omonoglucosides except MlvAc (Table 3).
It has been reported that anthocyanin profile may be affected by grape variety (González-Neves et al., 2004).For the Xinomavro variety, Mlv and MlvCoum contents were found to be the most important anthocyanins in grape skins in all training systems studied.Mlv represented an average content of 67%, 61% and 64% of total anthocyanin-3-Omonoglucosides in Lyre, Royat and Guyot, respectively, while MlvCoum represented 32%, 38% and 36%, respectively, with minor contributions from the other anthocyanins (Table 3).These results do not confirm previous studies reporting that vine training and pruning method may modify grape anthocyanin profile (González-Neves et al., 2004).
-77 -   The results of the comparison between training systems showed that Lyre trellis resulted in an improvement of anthocyanin composition of Xinomavro grape skins, compared to the two VSP systems.This effect could be attributed to multiple factors.First, the higher leaf area per fruit weight of the Lyre vines (Table 1) suggests a better source-sink ratio and a more favourable assimilate distribution during ripening, as also reported for Tannat grapes (González-Neves et al., 2004).Moreover, divided canopies (as in the Lyre trellis) achieve a better ratio of canopy surface area to canopy volume, increasing the proportion of exterior to interior leaves, thereby positively affecting vine carbon balance (Gladstone and Dokoozlian, 2003).Furthermore, divided canopy systems are reported to improve the light microclimate during the day due to decreased canopy density, especially in the fruiting zone (Reynolds and Vanden Heuvel, 2009), as compared to VSP systems.The reduction of shading around the clusters in red varieties was demonstrated to enhance anthocyanin synthesis in previous studies (Bergqvist et al., 2001).

Polyphenolic composition and antioxidant activity of skins and seeds
As illustrated in Table 2, seed extracts were richer in phenolic compounds and showed higher antioxidant activity than skin extracts in all samples examined.Total phenolics of whole berries were higher in the Lyre grapes (Table 2) as opposed to Royat and Guyot ones.The results agree with a previous study reporting higher levels of total phenolics in Sangiovese grapes trained to a modified Y-shaped open trellis as compared to a traditional VSP system (Palliotti, 2012).However, training system did not significantly affect skin and seed individual phenol content.Training system affected the antioxidant activity of the skins, with higher values for Lyre.
Concerning the astringency sensation, seeds from the Lyre system contained significantly higher contents of precipitable with BSA tannins (Table 2).
The concentration of flavan-3-ol monomers (C, EC, ECG, EGCG) and oligomers (B1, B2, C1) in Xinomavro skins and seeds showed significant differences among the training systems (Table 3).In general, grape skins of the Royat grapes were richer in monomeric flavanols (C, EC), while a different trend was recorded for seeds, where highest C values were measured in the Guyot grapes.Regarding skin oligomeric proanthocyanidin (B1 and C1) concentration, no difference was found among the three treatments.Galloylated monomers (ECG, EGCG) were not detected in any skin extract.Proanthocyanidin B2 was the dominant oligomer in Royat skin extracts, while proanthocyanidin B1 predominated in the skin extracts of the other two systems (Table 3).Regarding seed extracts, Lyre system was characterized by significantly lower concentrations of gallic esters (EGCG), while seeds from the Guyot system were significantly richer (p < 0.05) in proanthocyanidin B1.It was also observed that seeds were almost 100 times richer in oligomeric proanthocyanidins as compared to the skins.

Structural characterization of skin and seed proanthocyanidins
In all skin samples analyzed, polymers represented 98% and oligomers 2% of the total pool of skin proanthocyanidins, while in seeds, polymers represented 77-81% and oligomers 19-23% of total seed proanthocyanidins.Oligomeric fractions were characterized by a lower mDP but a higher %G than polymeric fractions, with relatively similar trends for seed and skin fractions (Figure 1).This observation
does not confirm previous works in other varieties (Prieur et al., 1994;Sun et al., 2013;Ćurko et al., 2014) reporting that lower mDP values corresponded to low %G.
Concerning the effect of training system, significant differences were found regarding mDP and %G among the three systems, in both skins and seeds of Xinomavro grapes (Figure 1).Grapes harvested on Royat vines had a higher mDP in the polymeric fraction of both skins and seeds.Regarding oligomeric tannins, Royat showed significantly higher mDP only in berry skins.There were no significant differences in the %G of oligomeric fractions of skins and seeds among the three systems studied.However, in the polymeric fractions, Royat seeds and skins were characterized by the lowest %G compared to the other two treatments.
Figure 2 illustrates the percentage contribution of terminal (t) and extension (e) subunits in polymeric and oligomeric tannin fractions of skins (Figure 2A) and seeds (Figure 2B).In the skins, the main terminal subunit was (-)-epicatechin (%ECt) in both polymer (60-74%) and oligomer fractions (48-61%), while the main extension subunit was (+)-catechin (%Ce).These data are not in agreement with the results obtained by Souquet et al. (1996) and Cohen et al. (2008), who reported that %Ct was the main terminal and %ECe the main extension unit in Merlot skin tissues.These differences could be due to genotype but might as well be attributed to the different tannin extraction or depolymerization methods employed (phloroglucinolysis vs thiolysis).An interesting finding was the absence of %Ct from the polymeric skin fractions, accompanied by a relatively high (-)epicatechin gallate (%ECGt) content (24-29%).%ECGe presented the lowest contribution to skin procyanidin extension units.
Regarding the effect of training system on the structural properties of proanthocyanidins, Royat was characterized by higher %ECt values in the terminal polymeric skin fractions accompanied by lower %ECGt, while the opposite was observed for oligomeric skin fractions (Figure 2A).In skin extension subunits, both polymeric and oligomeric fractions of the Royat grapes contained higher %Ce and lower %ECe than Lyre and Guyot.In the seeds, no significant differences were found regarding terminal procyanidin subunits of polymeric and oligomeric fractions among the three systems (Figure 2B).Regarding seed extension subunits, polymeric seed fractions of the Royat grapes contained higher %Ce and lower %ECe, similarly to skin proanthocyanidin structure.
Sensory analyses of flavan-3-ol monomers and oligomers have shown that astringency increases with mDP (Clifford, 1986;Robichaud and Noble, 1990).It has also been reported that astringency increases with both %G and molecular weight (Ricardo-da- Silva et al., 1991).Moreover, the structural characteristic of the molecules might also exert a significant influence on their astringency as it has been reported by Kallithraka et al. (1997) and Thorngate and Noble (1995) who demonstrated that (-)-epicatechins are significantly more astringent than (+)-catechin when tasted in model wine solutions.
However, it is difficult to relate the differences in astringency and bitterness of grape phenolic extracts to their compositional differences.In the present study, the higher mDP values of Royat seed and skin tannins might result in more astringent wines, but the lower %G values might counteract this effect.Moreover, the lower (-)-epicatechin and the higher (+)-catechin contents of the tannin extension units in the skin and seeds of Royat grapes might also result in a lower astringency sensation potential of the grapes derived from this training system, as compared to Lyre and Guyot (Kallithraka et al., 1997).The higher astringency of seed tannin extracts in the Lyre berries (Table 2) is supportive of this conclusion.

Wine composition
The wines from the different trellis treatments showed similar alcohol contents (Table 4).However, Lyre wines were less acid, as depicted by their lower titratable acidity and higher pH compared to Royat and Guyot.According to a previous study, cordon trained Riesling grapes had lower acidity than divided trellis grapes (Reynolds, 1988).In accordance with skin anthocyanin levels, Lyre wines displayed the highest concentration in total and individual anthocyanins (except malvidin-3-Oacetylmonoglucoside) among the three systems (Table 4), thereby leading to a higher colour intensity, despite their lower acidity and ionization index.Previous studies in Cabernet Sauvignon vines trained to Lyre and traditional VSP cordon also showed that wines of the Lyre system were characterized by higher levels of anthocyanins (Katerji et al., 1994).
Similarly to whole berry phenolic potential, wines produced from grapes of the Lyre treatment showed higher total phenol contents, although no differences were recorded for total tannins.Νο differences were observed in the antioxidant capacity of wines among studied systems.Moreover, differences observed in the structural properties of skin and seed proanthocyanidins among training systems (Figures 1 and 2) were not confirmed by the astringency measured in the experimental wines, which showed similar values.

CONCLUSIONS
Phenolic compounds play a vital role in the quality of red wines.They are responsible for some positive tasting characteristics but also for some unpleasant, negative aspects.Body, fullness and roundness are all organoleptic qualities which characterize premium red wines.On the other hand, bitterness, roughness, harshness, astringency and thinness are faults that  must be avoided.The overall organoleptic impression is based on a harmonious balance between these two types of sensations, directly related to the type and concentration of various phenolic compounds, especially tannins.
Xinomavro grapes are rather poor in anthocyanins and, as a consequence, the resulting wines typically have low colour intensity.On the other hand, Xinomavro grapes are rich in skin and seed tannins, which are characterized by increased astringency and dryness.The results of the present study could be of importance to both vineyard managers and winemakers.By adopting Lyre as a training system, higher anthocyanin levels could be achieved in both grapes and wines.It is also noteworthy that the higher anthocyanin content of Lyre berries was accompanied by a significant gain in yield.On the other hand, Guyot grapes were richer in tannin monomers, while Royat resulted in grapes with relatively higher skin tannins which possibly presented a less pronounced astringent character due to their structural properties (lower %G, lower (-)epicatechin and higher (+)-catechin contents of the tannin extension units).
With this knowledge, the winemaker could decide on the maceration process parameters (i.e.intensity and length) in order to optimize the quality of Xinomavro wines.In this view, grapes from the Lyre system would be seemingly more suitable for prefermentation skin contact in order to extract more anthocyanins, while those of Royat might be more adapted for longer maceration duration to produce full-bodied, balanced wines with longer ageing potential.

Table 2 . Phenolic composition of whole berries and skin and seed extracts of Xinomavro berries at harvest
Values with different letters within a row are significantly different (Tukey's test, p < 0.05).au, absorbance units; dw, dry weight.