THE EFFECTS OF BERRY SIZE ON ‘TEMPRANILLO’ GRAPES UNDER DIFFERENT FIELD PRACTICES

SUMMARY Small berries are considered to produce the best red wines as berry size determines the skin to pulp ratio and may affect wine composition. However, contrasting results have been reported about this postulate. In this context, the aim of this work was to assess the influence of berry size on grape compositional attributes in ‘Tempranillo’ grapevines under different irrigation, crop load and defoliation regimes. Grapes were collected


INTRODUCTION
Grapevine (Vitis vinifera L.) berry size has key implications for must composition and wine quality, mainly in red wines (Matthews and Anderson, 1988). Historically, the importance of berry size for winemaking was based on the assumption that the skin amount is relatively constant among berry sizes (Singleton, 1972). Hence, skin compounds such as anthocyanins and tannins were assumed to be more diluted in larger berries due to the lower surface area to volume ratio (Matthews and Anderson, 1988). Therefore, grape growers and oenologists believe that the lower the berry size, the better its oenological attributes are. Recently, a study found that Cabernet Sauvignon wines made from small berries presented better oenological attributes (higher total soluble solids and phenolics concentrations) than others coming from large and medium-sized berries (Gil et al., 2015). This might have implications for winemaking, since berry size affected the volatile profiles of Merlot and Cabernet (Xie et al., 2018).
However, other factors, such as crop load, vine water status and leaf surface, might affect grape oenological characteristics (Jackson and Lombard, 1993;Dai et al., 2011;Triolo et al., 2018). In fact, a number of studies contradicted the former theory that smaller berries are of better quality for winemaking; for example, an increase in skin mass proportional to berry size has been observed in Cabernet Sauvignon  and Syrah (Barbagallo et al., 2011). Moreover, skin tannin concentrations were similar among widely different berry sizes on a per berry basis, but they did not decrease with berry size when referred to skin weight . Similarly, skin mass increased with berry size in Syrah (Barbagallo et al., 2011). Therefore, berry composition is not only due to berry size, but also to changes in vine metabolism caused by water status, cultural practices or annual weather conditions . In addition, other authors observed that Cabernet Sauvignon wines from smaller berries did not have better quality than those from larger berries (Holt et al., 2008;Calderon-Orellana et al., 2014). These findings seem to indicate that the influence of other factors such as grapevine water status, solar radiation exposure or weather conditions might play a major role on berry composition. These factors are suffering from alterations due to climate change.
In fact, climate change represents a great challenge for viticulture since grapevines are highly sensitive to atmospheric factors that alter grapevine physiology (Keller, 2010), yield (Bock et al., 2013) and wine quality (Robinson et al., 2014). In this context, climate change is an unavoidable challenge that vine growers must face in the near future (Fraga et al., 2016). According to the International Panel on Climate Change (IPCC), global temperature is expected to rise 1 -5 ºC over the 21 st century (IPCC, 2014). Projections also predict drier conditions in certain regions, such as southern Europe, altering the suitability of certain areas for vine growing (Fraga et al., 2016). Therefore, one of the main challenges posed by this new climate scenario is maintaining yields at a profitable level, which might be achieved by optimizing irrigation (Simmonneau et al., 2017). Finally, climate change induces alterations in grape composition, modifying wine quality and typicity; however, these alterations can be limited through adaptations in the vineyard (van Leeuwen and Destrac-Irvine, 2017).
Among these adaptations, cultural practices such as irrigation, pruning or defoliation largely influence berry size (Matthews and Nuzzo, 2007). Several studies proved that irrigation and pruning affect berry size distribution for Cabernet Sauvignon and Merlot (Holt et al., 2008;Shellie, 2010). This indicates that the way in which berry size differences are achieved is more important in determining berry composition than the actual difference in berry size. Furthermore, grapevine cultivar affects berry size due to genetic factors and most of these studies have been performed on Cabernet Sauvignon, thus, other findings may be obtained for other red cultivars (Dai et al., 2011;Houel et al., 2013).
In this context, the aim of the current study was to assess the interrelations between berry size at harvest and fresh mass distribution between seed, skin and flesh, and fruit composition in the red grapevine cultivar 'Tempranillo'. This study complements previous research were the roles of water deficit, crop load and leaf removal on fruit size and grape composition had been quantified (Intrigliolo and Castel, 2011;Risco et al., 2014) without considering the possible effect of berry size per se within each imposed treatment.

Description of the study site
The experiment was carried out in the same 'Tempranillo' vineyard (Vitis vinifera L.) planted in 1991 on 161-49C rootstock were previously reported irrigation and canopy management trials were carried out (Intrigliolo and Castel, 2011;Risco et al., 2014). In the previous manuscripts, a detailed explanation of the vineyard characteristics is given.
At the experimental site, the average annual rainfall (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012) is 430 mm of which about 65% falls during the dormant period. In addition, the thermal sum (base 10 ºC) from 1 April to 31 October is 1669 ºC and the heliothermal index is 2291 ºC. During the growing seasons in which the experiments reported here were carried out (2004, 2005 and 2008), potential evapotranspiration ranged from 778 to 798 mm (Table I). The year 2005 was very dry, with annual rainfall 35% lower than the long-term average for the site, whereas 2004 and 2008 had rainfall slightly over this long-term average (Table I).

Experimental setup
The results from three different experiments, dealing with irrigation and canopy management practices, are described in the current report. Further details on these experiments (water relations, vegetative growth and yield) can be found elsewhere Castel, 2008, 2011;Risco et al., 2014), where the effects of the different treatments were deeply analysed regardless of the berry size component.
Crop level and irrigation trial carried out in 2004. Four treatments, with three replications each, were established in a randomized-block design. The treatments were: a) Rain-fed (R) with a medium crop level (20.8 clusters/vine); b) Irrigated (I) applying 100% crop evapotranspiration (ET c ) from anthesis till veraison and 50% ET c from veraison to harvest, with a medium crop level (21.9 clusters/vine); c) Low crop load (L), irrigated as I but with 12 clusters/vine; and d) High crop load (H), irrigated as I but with 33 clusters/vine. The ET c was estimated as the product of reference evapotranspiration (ET o ), calculated according to Allen et al. (1998), and a crop coefficient (K c ) that varied depending on the phenological stage of the grapevines. From June to July, K c gradually increased from 0.08 to 0.40. After veraison, the objective was to induce a moderate soil water deficit, therefore the water amounts applied were 0.2 of ET o (Intrigliolo and Castel, 2008). In 2004, the total irrigation depth was 82 mm. Crop level was adjusted by shoot thinning in mid-May, leaving 13-14 shoots per vine in the L treatment, while no shoot thinning was carried out for H. Each plot consisted of five rows with nine vines per row and the surrounding perimeter vines were guards (Intrigliolo and Castel, 2008).
Crop level and irrigation trial performed in 2005. The irrigation treatment consisted of applying water at 50% ET c from anthesis to veraison and then 35% ET c from veraison to harvest (K c increased from 0.16 to 0.70 from June to July, total irrigation depth was 155 mm over the growing season). Crop level treatments had 11 (Low, L), 20 (Medium, M), and 27 (High, H) clusters per vine. In the rain-fed treatment only the M crop level was studied. Crop level was adjusted by shoot thinning in mid-May and by additional cluster removal in early June if needed. Each plot consisted of five rows with nine vines per row and the surrounding perimeter vines were guards. The experiment was laid out in a generalized incomplete factorial block design with three blocks and two replicated combinations per block for the irrigated treatments and a single replicate per block for the rain-fed treatment (Intrigliolo and Castel, 2011).
Defoliation trial: conducted in 2008. The treatments applied were: a) Control -undefoliated; b) ED -All leaves of the first six nodes were removed just before flowering; c) LD -late defoliation, as the former one but applied at fruit-set; and d) EED -east ED, leaf removal was applied just before flowering but only the leaves facing east of the eight first nodes were removed. In the ED and EED treatments, leaves were removed on 29 May, whereas in the LD treatment, leaves were removed on 17 June (Risco et al., 2014). Each treatment consisted of 16 vines randomly chosen within the vineyard. Irrigation (50% ET c from anthesis to veraison and 35% ET c from veraison to harvest, amounting 130 mm over the growing season) and crop load (27-29 clusters per vine) were the same for all treatments (Risco et al., 2014).

Berry sampling and processing
Three clusters per plant from four vines per treatment were collected at harvest. Berries were manually separated from the pedicel. Approximately 1600 berries per treatment were individually weighed in order to obtain the distribution of berry fresh weight per treatment. From this, four size categories were established, representing the four quartiles of the weight distribution.
In the trials of 2004 and 2005, from each size category, 10 berries were randomly selected, their equatorial diameter was measured and skin and seeds were separated. For doing this, berries were sliced in half with a razor blade. Skin was obtained by carefully removing seeds and mesocarp from each berry-half using a small metal spatula and avoiding rupturing of pigmented hypodermal cells. The seeds were carefully separated from remnants of flesh by hand. Both skin and seeds were rinsed in deionised water and weighed after blotting off the excess of water. These analyses were not performed in 2008 due to limited manpower for the fieldwork.

Laboratory determinations
From each size category within each replicate (four repetitions per treatment and berry size category); 50-100 berries were randomly sampled for chemical analyses. Berries (including skin and seeds) were weighed and crushed with a Thermomix blender and hand-pressed through a metal-screen filter. Then, juice was centrifuged at 17608 x g for 10 minutes. Total soluble solids (TSS) were determined using a digital refractometer (PR-32, Atago Co. Ltd., Japan). Juice pH and titratable acidity (TA) were determined using an automatic titrator (Metrohm, Herisau, Switzerland).
In 2004 and 2005, total phenolic index (TPI) was determined by spectrophotometry on berry homogenates and expressed in terms of absorbance units (AU); anthocyanins (OD520 in HCl media, expressed in malvidin equivalents) were also determined by spectrophotometry (Ribereau-Gayon et al., 2000). In 2008, total anthocyanins and TPI were determined by ultraviolet/visible spectrophotometry in samples of 150 berries homogenized (Ultraturrax T25) to a grape paste (Iland et al., 2004). Maceration was not carried out in any of these cases. All determinations were performed in duplicate for each treatment and berry size category.

Statistical analysis
The normality of the berry weight distribution was assessed using the Kolmogorov-Smirnov test. Differences among treatments were assessed using ANOVA and the Tukey's test at p < 0.05. For the compositional attributes and the skin and seed traits, averages for each treatment were calculated as the mean of the values among all size categories, since the number of berries used in each replicate was the same and did not depend on the berry size distribution. Since the treatments imposed in the field exerted a significant influence on berry maturation, altering the concentrations of metabolites in the grape, we accounted for this fact by using TSS as a covariate in the statistical analysis and adjusted means were calculated for each attribute to a given value of TSS (22.7 ºBrix). Consequently, the differences among treatments are reduced and we can hypothesize that these differences are caused by berry size. Relations between berry size and the different studied attributes were assessed through linear regression analysis and the regression coefficient (R 2 ) was calculated. The differences in slopes and intercepts of the fitted lines were assessed through an analysis of covariance (ANCOVA). All statistical analyses were performed using IBM SPSS Statistics for Windows, Version 20.0. (IBM Corp., Armonk, NY).

Berry distribution in size categories
According to the Kolmogorov-Smirnov test, berry size followed a normal distribution in all treatments in 2004 and 2005 (p-values ranging from 0.09 to 0.77), whereas in the defoliation experiment carried out in 2008, berry size distribution did not follow a normal distribution in some treatments (p-values < 0.001).
The smallest berries occurred in 2005 and, thus, size categories corresponded to lighter berries than in the other experimental years. In 2004 and 2008, berry size categories were similar (Table II). The percentages of berries within the middle size classes accounted for 70-80% of the total, except for the defoliated treatments (Table II). The treatments imposing limitations to photo-assimilate supply increased the proportion of berries in the lower size category. The fact that 2005 was less rainy than 2004 and 2008 (Table I) led to smaller berries in that year (Table II)

Skin and seed distribution according to berry size
In 2004 and 2005, seeds of 'Tempranillo' grapes accounted for 2-6% of berry weight, whereas skin represented between 9-16% of total berry mass. These percentages depended on the berry size but also on the irrigation and crop load treatments. In both experimental years, skin weight, seed number and seed weight significantly increased with berry size (Figures 1a and 1d); however, the slope of the fitted regression lines was different depending on the treatment. Skin weight per berry was strongly and positively correlated (R 2 ≥ 0.93) with berry size in both years, except for the irrigated (L) treatment in 2005 (Figure 1d), and it increased more than 200% from the smallest to the largest categories. Seed number was one for small berries and increased up to 5 four seeds per berry in the largest-size category (Figures 1b and 1e). However, the variability in this number increased with berry size independently of the treatment (Figures 1b and 1e). Finally, seed weight (Figures 1c and 1f) increased with berry weight in both years.

Berry composition: effects of treatments and relations with berry size
Grape chemical composition parameters as function of berry size for each treatment depended on the experiment; therefore, they are described separately.
Crop level and irrigation trial (2004): The irrigation and crop level regimes had a major effect on berry composition, with significant differences detected for all the attributes considered in this experiment (Table III). The treatments that imposed a sink-source limitation (rain-fed and high load treatments) showed lower TSS and anthocyanins concentrations. The interval of TSS ranged between 20.9 (for the high load treatment) and 24.9 (for the low load treatment) ºBrix. Despite this large range, a significant effect of berry weight on TSS across the berry size categories was observed for all the treatments, except for high load (Table III). Moreover, significantly higher TSS was detected for the smallest berries than for the other categories in all the treatments except for the high crop load (Table  III).
On the contrary, pH did not show a clear trend with berry size (Table III). However, titratable acidity significantly increased with berry size (Table III), although no trend was observed when values were adjusted by TSS. Total phenolic index did not significantly vary with berry size for the studied treatments (Table III). Anthocyanins concentration tended to reduce with berry size, even when adjusted by TSS (Table III).
Crop level and irrigation trial (2005): The trends observed in 2004 were similar in 2005. Smaller berries presented a significantly higher TSS content than the rest of the four size classes (Table   IV). In contrast, pH did not vary with berry size, except for the irrigated treatment with low load (Table IV). Titratable acidity increased with berry size, but this trend was not clear when values were adjusted for TSS (Table IV). Color attributes showed a decreasing trend with berry size (Table IV). However, in the case of TPI, this declining was not clear when values were adjusted for TSS.

Defoliation trial (2008):
In 2008, TSS significantly decreased with berry size for all the studied treatments (Table V). Titratable acidity significantly increased with berry size for ED, but decreased for LD, whereas pH did not follow a clear pattern with berry size (Table V). Moreover, TPI and anthocyanins decreased with berry size when defoliation treatments were imposed (Table V).

Relationships between berry size and berry composition
Although there was a clear trend to lower TSS with increasing berry size in all treatments from all the experiments (Figure 2), these trends were not always significant. For instance, in 2004 the relationship between TSS and berry size was not significant for the high load treatment, whereas it was for the other treatments (R 2 between 0.95 and 0.99). In 2005, significant correlations were detected only for the irrigated treatments with medium and high loads.
Finally, in 2008, the only non-significant correlation was observed for LD. In 2004, the decreasing rate was rather similar between treatments (Figure 2a), although slightly more pronounced for the low crop load when compared to the rain-fed treatment. In 2005, the decreasing trend in TSS with increasing berry size was slightly higher for the irrigated treatment with high load (Figure 2b). In 2008, the declining rate was lower for the undefoliated control than for the other treatments (Figure 2c).    'Tempranillo' berry size followed a normal distribution, similarly to the findings from other studies (Shellie, 2010;Calderon-Orellana et al., 2014). However, this distribution was not observed in 2008 under the defoliated treatments, likely because of the modification in the source-sink ratio of the vines and, maybe, reducing the assimilate supply for the clusters. Berry size variation is established early in the first developmental stages of grapevine (Gray and Coombe, 2009). However, the treatments imposed here exerted a significant influence on berry size distribution since they varied the proportion of berries within each size class in all years. Our results suggest that high load and rain-fed treatments posed limitations to berry growth; in the case of the rain-fed treatment due to the effect of water stress, whereas in the high-load treatment, the limitation in berry growth was caused by a greater competition for the photoassimilates (Hunter and Ruffner, 2001). A similar behavior in relation to water stress was previously observed in Cabernet Sauvignon, where berries coming from a deficit-irrigated treatment showed smaller sizes than those from medium and highly irrigated treatments . In accordance with our results, a recent study on Cabernet franc showed that water stress was the factor that affected largely fresh berry mass, sugar content and malic acid concentration (Triolo et al., 2018).
As previously found for Cabernet Sauvignon and Shiraz Walker et al., 2005;Calderon-Orellana et al., 2014), flesh and skin growth for 'Tempranillo' appeared to be coordinated since skin weight significantly increased with berry size both in 2004 and 2005. Independently of the treatment, seed number and weight increased with berry size, as previously reported for Cabernet Sauvignon and Merlot Shellie, 2010;Calderon-Orellana et al., 2014;Gil et al., 2015). These differences in seed weight might be relevant for tannin concentrations in wines because seeds and skins are the most important sources of tannins in red wines (Harbertson et al., 2003).
Sugar concentration in grapes plays a major role in shaping its sensory properties, determining alcohol content after fermentation, and providing precursors for the synthesis of aroma compounds (Dai et al., 2011). Total soluble solids concentration in the berries decreased with berry size independently of the treatment (irrigation, crop load or defoliation) imposed each year. Grapes from treatments involving some limitations (rain-fed, high crop load) showed a lower capacity for accumulating solutes. These results agree with previous reports for other varieties and indicate that berry size is not the only factor affecting sugar accumulation, but also environment and viticultural practices (Jackson and Lombard, 1993;Clingeleffer, 2010;van Leeuwen and Destric-Irvine, 2017;Triolo et al., 2018). However, it is relevant to notice the fact that the weather conditions occurring each year also altered the TSS concentrations in the grapes; for instance, the control treatment in 2005 (a dry year) showed a higher level of TSS than that of the control treatments in 2004 and 2008.
In contrast to TSS, we did not detect a clear relationship between titratable acidity and berry size in any of the studied years due to the significant effect that treatments exerted on maturation. The concentration of organic acids in berries is influenced by those environmental parameters or viticultural practices that affect source-sink relationships and cluster microclimate (Jackson and Lombard, 1993), as crop load and irrigation in our study. The apparent incoherence observed for the effect of the early and late defoliation on berry acidity might have occurred because each defoliation treatment modified the cluster microclimate differently, altering the concentrations of the main organic acids in the berries (Risco et al., 2014). As for the year effect, similarly to the case of TSS, the control treatment in 2005 showed the lowest titratable acidity.
Phenolic attributes tended to decrease with berry size, although their values depended on the cultural practice imposed. In the case of TPI, high crop load caused significant reductions in their concentration for all berry sizes when compared to grapes from the other treatments, suggesting that photo-assimilate availability is a key factor controlling TPI synthesis. Irrigation treatments also diminished TPI content in berries; however, this reduction was not always significant in relation to the rain-fed control, in contrast with previous observations (Matthews and Anderson, 1988). In addition, early defoliation increased the TPI content of the berries, whereas the other defoliation treatments caused a decrease in TPI when compared to the undefoliated control.
Anthocyanins concentration clearly decreased with berry size in 'Tempranillo', as previously observed for other cultivars such as Cabernet Sauvignon and Shiraz Walker et al., 2005). Cultural treatments and year exerted a significant effect on the anthocyanins concentration in 'Tempranillo' berries. For instance, in 2004, the high load treatment presented the lowest anthocyanins concentration for all berry sizes, indicating that photo-assimilate availability plays a major role on the synthesis of these compounds. Moreover, regulated deficit irrigation altered the concentrations of anthocyanins in 2004 and 2005, proving that this practice causes genetic changes in the expression of certain enzymes for most of the secondary metabolic pathways in Vitis vinifera, including those for anthocyanins synthesis (Castellarin et al., 2007;Santesteban et al., 2011). In the case of the defoliation treatments imposed in 2008, our results suggest that the modifications of the light environment at the cluster zone altered anthocyanins synthesis in small berries.
The changes in berry composition over size reported here are in accordance with previous findings by other authors in different cultivars (Shellie, 2010;Calderon-Orellana et al., 2014;Gil et al., 2015). They support the hypothesis that berry mass per se is unlikely to be the main influence on the solute concentration in grapes and that the cultural treatments used to induce those small berries exert a major effect on berry composition . The mechanisms by which water deficit, photo-assimilate availability or light exposure increased the concentrations of anthocyanins and soluble solids in the berries are probably the differential growth responses of skin and inner mesocarp tissue to these constraints .
In view of these results, growers might try to control seasonal and within-vineyard variability and modulate berry size and composition through cultural practices (Harbertson et al., 2003). Depending on the objectives of the winery, growers may decide to favor small berries for increasing TSS and color attributes in order to obtain red wines with deep color, full body, soft tannin and fruity aromas, which are preferred nowadays for international markets (Gil et al., 2015). In addition, they could use larger berries for entry-level wines after a process of berry classification. Nevertheless, our results proved that berry composition can be modulated through varying crop load, irrigation management and defoliation; and larger berries obtained under several of the treatments imposed possessed interesting oenological properties, hence, they could be used for achieving high-quality 'Tempranillo' wines under Mediterranean conditions. Further research is needed to understand the effects of other cultural practices and on other cultivars, as well as about the economic feasibility of these management practices (Clingeleffer, 2010).
Finally, winegrowers must consider that most of the berries would fit into the middle classes of the size distribution. Medium-size berries account for up to 80% of the total yield, whereas small berries only represent 10%. Hence, the improvements in quality observed for small berries may not influence the total production of the vineyard and not be of economic importance. Cultural practices can be used for obtaining medium-size berries with optimal compositional attributes, as reported here.

CONCLUSIONS
This study provided further evidence supporting the essential influence of agronomical practices on berry size and composition by analyzing separately the attributes of grapes from different sizes in a red grapevine cultivar widely grown in Spain. In our study, smaller berries had higher sugar and anthocyanins concentrations than larger berries. However, irrigation, crop load and defoliation affected these compositional traits, producing greater berries with similar traits than those smaller but coming from rain-fed and not defoliated treatments, suggesting that grape composition, for a given berry size, can be modulated through agricultural practices. Therefore, wineries can sort berries for selecting those from a given size and agricultural treatment in order to make different wine styles.