Vine Irrigation through Two Shoot Densities in Flavonoid and Non-Flavonoid Compounds in ‘Tempranillo’ Grapes

This study aims to analyze the effects of non-limiting irrigation (I) vs. rainfed (R) through two different shoot densities, high-load (H) and low-load (L), on vegetative growth, agronomic parameters, flavonoid and non-flavonoid polyphenol substances of cv. Tempranillo grown in a semi-arid climate during three consecutive seasons (2014–2016). Under these conditions, in the 2015 and 2016 seasons, irrigation showed significant increases in berry weight (14.7% and 13.4% in H and L, respectively, in 2015, and 35.6% and 23.5% in the same treatments in 2016) and yield (66.7% and 48.5 in 2015; 27.9% and 177.5% in 2016). Additionally, a general decreasing trend is observed in anthocyanins with the exception of peonidin derivates, almost all flavonol compounds, cinnamic acid and resveratrol values with different degrees and statistical significance depending on the shoot density of the vines. A slight variation is observed in 2014 in these parameters. On the other hand, no general trends are established either in flavanol compounds or hydroxybenzoic acid. Thus, the effect of irrigation depends on the parameter considered, the shoot density of the vine and the season considered.


Introduction
A large proportion of vineyards is located in regions with seasonal drought (e.g., Mediterranean-type climates), where soil and atmospheric water deficits together with high temperatures exert large constraints on yield and quality. The increasing demand for vineyard irrigation requires an improvement in water use efficiency.
In arid and semi-arid environments, irrigation is a major tool used to regulate soil water availability to vines. Under these conditions, supplying irrigation to ensure potential vine evapotranspiration increases yield, and sometimes reduces wine quality [1][2][3]. Winter pruning is the most widely used viticultural technique to regulate crop yield and achieve a targeted grape composition, notwithstanding the fact that the number of shoots per vine is not an accurate predictor of harvest yield [4,5]. This technique is often followed by shoot thinning with the removal of fruiting shoots from the vine in order to reduce grape production and plant canopy density [6] and to encourage light and air penetration into the canopy [7,8]; thus, improving the canopy microclimate and grape quality [9][10][11]. Shoot thinning also facilitates more desirable shoot spacing along canes and cordons, and a more even leaf area distribution in the canopy [7,8,12]. Sometimes these conditions can lead to improved bud fertility, fruit set and the mitigation of yield losses. Shoot thinning often increases yield and decreases vegetative growth, leading to a higher ratio of yield to leaf areas, although in other cases, irrigation at 100% of crop evapotranspiration does not  Table 2 shows the results of the ANOVA applied to the results obtained from the vegetative, agronomical and polyphenolic content of cv. Tempranillo grapes at harvest during the three years of the experiment. The table indicates a high effect of year on all those parameters. Furthermore, worth noting is the significant interaction found in most agronomic parameters, as well as in the content of phenolic substances. Thus, the results were analyzed year by year. Table 2. Effect of irrigation (treatment) and experimental season (year) and their interactions on values of vegetative, agronomic and polyphenolic content of cv. Tempranillo grapes at harvest.

Agronomic Parameters
The vegetative growth (leaf area, LA), cluster number (CN), berry weight (BW), cluster weight (CW), yield (Y) and ratio (LA/Y) at harvest from the 2014 to 2016 vintages are reported in Table 3. Although it was significant only in 2015 and 2016, the same trend was observed in the three seasons of the experiment: irrigation stimulated the vegetative growth and vine leaf area. The percentage increases in IH respect to RH were 69.05%, 140.91% and 122.86% in 2014, 2015 and 2016, respectively. These increases were higher than those found when comparing IL and RL in the same years (2.81%, 99.31% and 57.31%, respectively). On the other hand, increases were found in the values of CN, CW and Y in irrigated treatments in respect to non-irrigated ones in L and H treatments in all years. Their significance and extent, however, depended on the parameter, crop load, and year considered. In 2014, irrigation did not have a significant effect on the values of any parameters, and in the following years, the values of these parameters were higher (p < 0.05) in IL than in RL, except for the cluster number in 2015. With regard to BW, with the exception of the decrease in IL vs. RL in 2014, the general trend found was a berry weight increase, which was higher in I treatments (H and L). Due to the importance of berry size for the synthesis and accumulation of phenolic compounds, a decrease observed in the values of all treatments in 2016 compared to the previous years was observed. When analyzed year by year, a decrease in weight of the IL berry (2.07 g) compared to the RL (2.30 g) stood out. In 2015 and 2016, the I berries had higher values (p < 0.05) than the R berries. The highest increases (35.58%) were found when IH and RH were compared in the 2016 season. When all treatments were compared, it was observed that the lowest value was reached in RH in all years, while the highest was in RL in 2014 and in the following years in IL. Furthermore, in 2015 and 2016, the values of berries from irrigated treatments were very close, showing the following sequence in those years: IL > IH > RL > RH. In respect to yield, the lowest and highest values were found in IH and RL for all years. Finally, no clear trend was found in the ratio LA/Y.

Polyphenolic Families-Year Effect
As Table 3 shows, the different meteorological conditions in the years of the trial had a strong impact on vegetative growth and yield components. All these facts had an impact on the synthesis and accumulation of phenolic compounds. The results presented in Figure 1 show that the response to these factors was similar for all groups belonging to the same phenolic family. It is worth noting that the weather conditions in 2016 contributed to a better biosynthesis and higher accumulation of anthocyanins (ANs), flavanols (FLAVA), hydroxybenzoic (HB) and hydroxycinnamic acids (HA) and resveratrol (tR) in the berries. However, those of 2015 gave the highest values of flavonols (FLAVO) and the lowest values of tR.

Flavonoid Compounds Anthocyanins
Fifteen anthocyanic compounds were identified, quantified and grouped into anthocyanin monoglucosides (∑Glus), acetyl glucosides (∑Acs) and coumaroyl glucosides (∑Coums) of delphinidin (Dp), cyanidin (Cy), petunidin (Pt), peonidin (Pn) and malvidin (Mv) ( Table 4). The total amount of ANs was given in mg of malvidine-3-glucoside. kg −1 fresh berries. The results show that, regardless of the season and treatment, Glus were predominant forms, followed by Coums and Acs, and Mv-derived compounds were the predominant anthocyanin-derived substances, while Pt and Cy derivatives were the least abundant. Thus, malvidin glucoside was the major individual anthocyanin compound, and cyanidin acetate and peonidin acetate were the minority compounds in both seasons (data not shown). A similar profile was found in the Tempranillo cultivar in different geographic areas in Spain [33,34,38]. When the effect of irrigation was examined year by year, it could be seen that, in 2014, irrigation decreased the ∑Coums, ∑Acs and ∑Mvs, but it had no impact on the total values in either H or in L treatments. Furthermore, the decreases were higher in L than in H. In both H and L treatments, the irrigation reduced the total values of anthocyanin compounds in the following seasons, except for Pn and Cy, and a significant decreasing trend was found in all anthocyanin derivates. These decreases were greater in 2015 than in 2016, and in H in respect to L treatments. When all treatments were compared, IL had the lowest values of total anthocyanin compounds in all the years of the trial, while the highest corresponded to RL in 2014 and to RH in the rest of the seasons. *, ** and *** indicate, for the same crop load (H/L), the significance of rainfed in respect to irrigation at p < 0.05, p < 0.01 and p < 0.001, respectively. For each parameter, means followed by different letters were significantly different between treatments, p < 0.05, Tukey's HSD test.

Flavonols Compounds
As shown in Table 5, in all seasons and regardless of the treatment, ∑My and ∑Qc were the predominant flavonol compounds, with myricetin-3-glucoside (MyG) being the most abundant, except in H treatment in 2014, in which ∑Kp and ∑Ih were the minor compounds in the profile of cv. Tempranillo.  *, ** and *** indicate, for the same crop load (H/L), the significance of rainfed in respect to irrigation at p < 0.05, p < 0.01 and p < 0.001, respectively. For each parameter, means followed by different letters were significantly different between treatments, p < 0.05, Tukey's HSD test.
As with anthocyanins, the effect of irrigation on flavonol compounds in the H and L treatments was not clear in 2014. For some compounds, a trend opposite to that of the following years was observed and significant increases were found in QcGL and ∑Qc, when IH was compared to RH. However, in 2015 and 2016, irrigation had a clear impact in all flavonol compounds with significant decreases of more than 30%, in both H and L. Moreover, the effect of irrigation was similar for both loads and, in general, the percentage decrease for each compound was similar in the H and L treatments. The decreases in MyG ranged from 36.92% (p < 0.01, RL vs. IL in 2015) to 56.83% (p < 0.001, RH vs. IH in 2016) and in total flavonols from 36.94% (p < 0.001, IL vs. RL in 2015) to 53.23% (p < 0.001 in the same comparison in 2016). In the latter two years, RH and RL values were very close and higher than those found in IH and IL. Table 6 shows the flavanol profile of grapevine cv. Tempranillo. Regardless of the season and treatment, (−)-Epigallocatechin (EGC) and Pro B1 were the most abundant catechin (CAT) and proanthocyanidin (PRO), respectively, while Pro A2 was the least abundant flavanol.

Flavanol Compounds
Overall, the response of these compounds to the water status was low and inconsistent in 2014 and 2015. In 2014, a decrease in catechin (CAT) compounds was observed in only IL vs. RL, and no changes in proantocyanidin (PRO) values were found in the H and L treatments; in 2015, the only change was an increase in CG in RL vs. IL. Finally, in 2016, significant decreases and increases in CAT and PRO compounds were found, but without following any specific pattern. Thus, in this season, there were no changes in total CAT, i.e., decreases (in RH vs. IH) and increases (in IL vs. RL) in total PRO. Finally, the highest and lowest total FLAVO values were achieved in RH and RL, respectively. When comparing all treatments, no clear trend could be established. Only a slight tendency to higher values of CAT in H compared to L treatments was observed in the last two seasons.  Table 7 shows the non-flavonoid polyphenolic profile identified and quantified in cv. Tempranillo grapes grouped into phenolic acids (hydroxybenzoic (HB) and hydroxycin-namic (HA)) and stilbenes (tR). Among the phenolic acids, HA was the predominant one and, more specifically, cinnamic acid (CIN) in all treatments, with the exception of coumaric acid (COU) in I treatments in the 2015 season. The highest values of these compounds were observed in the 2016 season compared to 2014 and 2015. The effect of irrigation on these compounds differed according to the year, crop load and group considered. In respect to GA, no effect was recorded in 2014, an increase in IH vs.

Impact on Agronomic Parameters
Climate conditions are particularly important for grapevine growth. Heat, drought and light intensity are just some environmental stress factors that dramatically affect grape development, primary and secondary metabolism and, consequently, the final content of polyphenols in berries [56] Today, irrigation is widely applied in vineyards located in very hot areas, such as the one where this study was carried out. Irrigation should be optimized to achieve the best results of yield and berry quality at harvest. However, the results of irrigation application depend on a multitude of factors: some are permanent, such as terroir and cultivar, others are not manageable, as is the case with seasons, and, finally, others can be modified, such as the amount and timing of irrigation waters, and agronomical practices such as the crop load and crop level [3,10,11,19,57]. Generally, the number and size of grape clusters formed during grape development determine harvest yield, which is influenced by several key stages of vine phenology and seasonal conditions [58].
From the results of the present investigation, it is noteworthy that in each of the years of the trial, the reduction in shoot numbers achieved via early shoot thinning did not affect vine capacity, given as the total leaf Area (LA). Thus, the LA in H (high crop load) vines was similar to non-thinned vines (low crop load, L). These results were in agreement with previous studies carried out in the field on 'Cardinal' [59], 'Chardonnay' [60] and 'Sangiovese' [61], in which full vegetative growth compensation was achieved in all the years in vines with a low crop level (L) by an increase in the vigor of individual shoots manifested as a similar LA to high crop level vines (H).
It has been reported that water deficits reduce berry size and yield, and some studies have even shown that the decreases are linearly related to decreases in stem water potential [62]. In a recent meta-analysis, Mirás-Avalos and Intrigliolo [63] found that this relationship was variety-dependent. In this regard, Girona et al. [64] reported that 'Tempranillo' berry quality demonstrated great phenological sensitivity to water stress. According to previous work carried out on this cultivar [65][66][67], irrigation caused an increase in LA, BW, CW and Y in H (high crop load) and in L (low crop load). It was observed that the extent of the effect varied with the cultural practices, affecting the crop level, as has been demonstrated in several works [16,17,66]. Vines with a higher crop level seemed to benefit more from a higher amount of irrigation both in terms of yield and of fruit composition [3,66]. In this work, no generalized trend was found for yield; neither the ratio LA/Y nor the largest increases in respect to RH and LH were found in IH or IL (in %), respectively, depending on the seasons. However, every year, the increase (in %) in vine LA and BW in I respect to R vines was more noticeable in H than L vines (decreases were even found in IL in respect to RL in 2014). The relation between grape weight and phenolic substance accumulation has been investigated by several authors [68][69][70], and it has been considered that smaller berries accumulate more phenolic compounds [71][72][73]. Thus, considering the influence of berry weight on the synthesis, accumulation and concentration of phenolic substances at harvest, the effect of irrigation on berry weight in this work is of great relevance.

Incidence for the Year
It is known that the values of polyphenols in berries from a specific cultivar and vineyard and similar TSS are highly dependent on the season [32,38,[74][75][76]. The highest values of ANs, FLAVA, HB, HA and tR were reported in 2016 (the rainiest year). Vilanova et al. [39] showed that the composition of grapes was more affected by vintage than by geographic area. These results disagree with those of previous studies carried out in wetter areas, where the highest amounts of polyphenols were detected in the driest years [38,39]. It is possible that the spring rainfall during this last season contributed to a higher synthesis and accumulation of these compounds in the conditions of our trial. On the other hand, these rainfalls caused an oidium (Uncinula necator) infestation, which resulted in a large decrease in yield. Moreover, grapes were smaller in 2016 than in the previous seasons. As the summer temperature rose to atypical values, the anthocyanin biosynthetic genes were downregulated, reducing berry skin anthocyanin biosynthesis [77]. For instance, Tarara et al. [78] showed that high temperatures are associated with decreases in grapevine Dp, Cy, Pt and Pn-based anthocyanin contents, but found no influence on Mv derivative concentrations. However, during 2015, the drier and warmer growing season promoted a higher synthesis and accumulation of FLAVO.

Incidence of Irrigation through Different Shoot Densities in the Phenolic Families
Several specialized (also known as secondary) berry metabolites strongly respond to abiotic stressors, such as water deficits. Among these metabolites, polyphenols are very important, since they contribute to a large extent to grape and wine color, astringency and quality. Of the different classes of polyphenols present in grape berries, the most important are flavonoids (anthocyanins, flavonols and proanthocyanidins (also called condensed tannins)) and stilbenes. Previous works analyzing the effect of water stress on polyphenols report that their effect depends on the season, weather conditions, grapevine variety, magnitude and timing of water deficit and cultivar techniques applied in the vineyard [53,[79][80][81][82][83].
As cited above, many works describe the relationship between berry weight and the values at harvest of components which determine must quality. In the present work, the values of polyphenol families at harvest were generally related to the effect on berry weight. This would explain, in part, the low effect of the treatments in the first year of the trial, and the different effects of irrigation on H and L vines. In this last regard, Walker et al. [84] and Matthews and Nuzzo [85] concluded that the resultant winemaking traits of large or small berries depended more on the factor modifying the weight of the berry (variety, water deficit, etc.) than on the size itself. In this sense, the irrigation on H and L crop loads caused differences (p < 0.01) in stem water potential mean values of vines in 2014-2016 (−0.84 MPa and −0.74 MPa, respectively), yield, LA and ratio LA/Y, and all these factors caused differences in the synthesis and accumulation of phenolic compounds during the vegetative period and, in consequence, in values at harvest. Jackson and Lombard [15] demonstrated that within each genetic material and terroir, vine water status and vine yield (i.e., crop level) or the balance between the sources (vine capacity to produce photoassimilates) and sink demand (grape yield) (i.e., crop load) are probably the major determinants of vineyard performance and fruit composition. Our study focused particularly on profiling the different phenolic families. This is of considerable relevance, as not all phenolic compounds have the same importance in regard to the intensity, tone and color stability of wines [49]. Furthermore, the results achieved show the different sensitivity of the different phenolic compounds to those factors.
In most works, irrigation decreased the total anthocyanin concentration in grapes [53,[86][87][88][89][90]. This decrease was caused, on the one hand, by the indirect effect of berry size variation cited above, and, on the other, by the enhanced accumulation of anthocyanins through the stimulation of anthocyanin hydroxylation, probably by upregulating the gene encoding the enzyme F3050H [47,91]. In addition, F3050H transforms Cy and Dp into Pn, Pt and Mv [92,93]. Our work confirmed the previous findings by Castellarin et al. [93] and Deluc et al. [88]. According to these authors, the principal anthocyanins synthesized in the berries under water deficits are Pn . This latter finding suggests that, under our conditions, the methoxylation of Dp may have occurred in non-irrigated vines. Moreover, according to [78], high temperatures are associated with decreases in grapevine delphinidin, cyanidin, petunidin and peonidin, but they found no influence on malvidin derivatives.
In V. vinifera L., ANs are present as mono-glucoside forms of methoxylated and/or hydroxylated anthocyanidins. The number and type of substituents in the B ring of the AN molecule identify different ANs: Cy and Pn with two substituents (3 -substituted), and Dp, Pt and Mv with three substituents (3 ,5 -substituted) [76]. Acylated derivatives are considered to be among the most stable compounds [48]. Furthermore, it is known that Mv-3-glucoside is a less reactive AN than Pn-3-glucoside [76], which is known for being highly reactive [94]. On the other hand, the adjacent hydroxyl groups of o-diphenols are more sensitive to enzymatic oxidation (except for laccase) and non-enzymatic oxidation (catalyzed by copper or iron ions) to produce o-diquinones, or even o-diphenol dimmers. Therefore, Cy, Dp and Pt, which contain the o-diphenol structure on the B ring, are more sensitive to oxidation. However, neither Mv nor Pn possess ortho-positioned hydroxyl groups, which results in their comparatively higher resistance to oxidation [95]. Thus, the AN composition of the grape determines the stability of the color of future wine, and the characterization of the AN profile would presumably permit to know which varieties have a more stable color than others, given that their stability is different.
However, it is not only the anthocyanin concentration and profile that is responsible for wine color; co-pigmentation phenomena can account for between 30 and 50% of color in young wines [49]. Some authors suggest that the co-pigmentation reactions of anthocyanins are the first phase in the formation of stable polymeric pigments during wine aging [96]. Co-pigmentation in wine results from molecular interactions between AN pigments and other organic molecules, called cofactors, forming molecular associations or complexes. The most common cofactors include a variety of compounds, such as phenolic acids, flavonoids and, particularly, derivatives of flavonol and flavone subgroups [50]. In the co-pigmentation phenomenon, which contributes to the stabilization of the color of red wines [49,75], Baranac et al. [97] reported that flavonol substances were among the best co-pigments, especially quercetin [98]. The profile found was similar to the 'Tempranillo' grapes reported in previous works with Qc and My-based flavonols, MyG, QcG and QcGR, dominating [33,34,99]. Given the importance of these phenolic substances on the stability and intensity of the color of red wines through co-pigmentation phenomena, the study and monitoring of agronomic factors and viticultural practices that increase their content and improve the cofactor concentration are major objectives for the production of highquality wines, especially in terms of their color. Studies have shown that a water deficit has a moderate effect on flavonol synthesis, and the effect of the irrigation application period is almost negligible [92,100]. When Cabral et al. [101] investigated the impact of deficit irrigation on grapevine cv. Touriga Nacional during three seasons in the Douro region, they observed increases in KpG, My-hexoside, QcG, KpR and QcGR in two of the three vintages, while MyG decreased only in two of them. Gamero et al. [83], in wines elaborated from vines grown in similar edaphology and climatic conditions to those in the present work, reported a different irrigation effect depending on whether the wines were created from thinned or control vines. However, little or no change was noticed for other flavonoid compounds. The synthesis of these substances is light dependent, so the higher leaf area observed in IH and IL treatments may also have contributed to the decrease. Additionally, in this work, a similar response of ANs to irrigation was found in FLAVO compounds. This may be because flavonol biosynthesis is closely related to that of anthocyanins [102] (Jeong et al., 2006), even though it has been suggested that those phenolic family compounds share the same biosynthetic enzymes [91,100]. On the other hand, flavonol occurrence can be considered as a biomarker for a sun exposure regime achieved in a bunch area within the canopies following microclimate manipulation management. LA and the LA/Y ratio were higher in H than in L in 2015. Thus, as expected, the decrease in FLAVO was greater in H than in L.
The major flavonoids synthesized in the grapevine berry, anthocyanins and tannins (also known as proanthocyanidins), strongly impact the quality of red wines via their contributions to wine color and astringency [103][104][105]. In this work, in contrast with anthocyanin and flavonols, the differences between treatments were rarely significant, and a clear trend caused by irrigation was not observed in catechins or proantocianidins, regardless of the crop level. The impact of irrigation on these substances still remains unclear, as contrasting results have been reported among studies [43,90,106,107]. The pre-and post-veraison application of water deficit increased proanthocyanindin levels in Syrah and Cabernet Sauvignon berries, but only transiently, and at harvest, no differences were observed [43,106]. In grapes from cv. Graciano grown in plastic pots, irrigated with sustained deficit irrigation (SDI), Niculcea et al. found registered decreases in FLAVO and increases in CA in respect to non-irrigated grapes [40]. In Tempranillo vines grown in the same edaphoclimatic conditions as the present work, Gamero et al. [83] found increases in Pro B1 and Pro B3 dimers in irrigated grapes. When regulated deficit irrigation (RDI) was applied to 'Monastrell' grapevines grown in eastern Spain under semiarid conditions, the content of EC in these grapes was mainly flavanol, and epicatechin-3-gallate decreased [108]. Thus, according to Bucchetti et al. [79] in their work with Merlot, our results with Tempranillo indicate that tannin accumulation is less sensitive to water deficits than anthocyanin accumulation, and is largely unaffected by late-season water deficits. The different methods used to determine these phenolic substances could be in part the cause of these results. Moreover, a gene expression study undertaken by Zarrouk et al. demonstrated differential expression during the grape berry development of the ANR gene in grape seeds and a slight downregulation under water stress (cited by [109] of these substances.
Regarding phenolic acids, previous works report different profiles and effects of irrigation depending on the cultivar considered. In their work, cited above, Niculcea et al. [41] found that t-caftaric acid was the main HA, accounting for 55% in Tempranillo and Graciano in well-watered plants, in respect to the irrigation effect, while in Tempranillo, the SDI treatment resulted in increased t-caftaric and caffeic acids, which in Graciano did not alter the total HA at harvest, but its individual composition was modified. Thus, the SDI treatment reduced the concentrations of t-caftaric, c-coutaric and t-coutaric and increased the concentrations of coumaric and ferulic acids. In their triannual work, Cabral et al. registered increases and decreases in gallic and coutaric acids in berries irrigated with 30% and 70% ETc depending on the season considered [101]. Finally, gallic acid was the most abundant phenolic acid in 'Monastrell' grapes, and only the content of total hydroxybenzoic acids and trans-fertaric acid was affected by the watering regime, increasing in grapes from rainfed grapevines in respect to those in RDI [108]. Valdés et al. [110] found decreases in hydroxycinnamic acids in berries in respect to early defoliated vines. These authors explained these results on the basis of the lower canopy porosity and decreased cluster exposure of the defoliated vines. Thus, the decreases in these substances in IH and IL grapes vs. RH and RL could be explained by the increase in LA in irrigated vines. However, the value of the LA increase did not correlate with the decreased values of these acids. These acids react with monomeric anthocyanin-forming pyranoanthocyanins, which are more stable compounds [111] and, thus, stabilize wine color. Therefore, in addition to winemaking techniques, viticultural practices that increase the content of these substances in the berry must be applied. In this sense, irrigation would not be a recommendable technique.
Generally, stilbenes are considered as phytoalexins, and their formation in grape leaves correlates with disease resistance. It accumulates mainly in the grape skin and seeds of red and white grapes at a wide range of concentrations, depending on biotic and abiotic conditions [112]. Resveratrol is the most bioactive stilbene in grapevines [113] and its synthesis is catalyzed by stilbene synthase (STS) [114]. The highest values of tR were found in 2016, and in this season, the vineyard was infected by oidium. In our work, a general, significant decrease in tR with similar extent was found in H and L treatments in 2015 and 2016. However, conflicting results have been reported in the literature regarding the effect of irrigation on this substance. While a short drought effect was observed in the tR concentration in grape berry skins in the Barbera cultivar [115] and no significant differences were found by [108] in 'Monastrell' grapes, a substantial increase in STS was observed in the Cabernet Sauvignon cultivar [100]. According to the work of Molero, PCAs demonstrated that seasons had a great effect on polyphenol values under our conditions [83,116]. The season effect was even greater than treatment effect. When the PCA plots were examined year by year, anthocyanin and flavonol values were associated with low values of berry weight and yield. Thus, the highest berry weight and lowest antocyanin values were observed in I treatments. Molero et al. [117] reported different values of anthocyans in rainfed and irrigated treatments, but they did not differentiate between H and L treatments. In this work, Cat and Pro permitted a good separation between H and L treatments. The vineyard was drip-irrigated with pressure-compensating emitters of 4 L.·h −1 located in a single row 60 cm apart. Irrigation was managed uniformly, as follows: water consumption was calculated with a weighing lysimeter located at the experimental vineyard [116]. Irrigation started when the stem water potential (SWP) reached a level of −0.6 MPa [118,119]. The SWP measurements were determined at midday using a pressure chamber (Model Soil Moisture Corp, Santa Barbara, CA, USA).

Treatment Application
The experimental design was a split plot with four replicates. Irrigation was the whole plot factor, and crop load the subplot factor. Two irrigation treatments were established in relation to crop evapotranspiration (ET c ): rainfed (0% ET c ) (R) and non-limiting irrigation (100% ET c ) (I) throughout the season. The value of ET c was determined with a weighing lysimeter installed in the experimental vineyard. Initially, the crop load was adjusted by winter pruning, and, then, within each irrigation regime, two different shoot densities were adjusted by pruning: high crop load (H) (12 shoots per vine) and low crop load (L) (6 shoots per vine) at stage 12 (phenological development stages according to [120]). Both practices were combined and, thus, four experimental treatments were established: RH, RL, IH and IL. Therefore, the study was carried out in 16 experimental plots, each one consisting of 6 rows of 18 vines (108 vines per experimental plot).

Environmental Conditions
Meteorological data were obtained during the experiment from an agro-climate station (Network of Extremadura Advice to Irrigation; REDAREX) located 100 m from the vineyard. We obtained agrometeorological data from a station close to the vineyard (100 m) with the characteristics described in [121]. Growing degree days (GDDs) were calculated using 10 • C as the base temperature, recorded during the vegetative-productive period [122,123]. Maximum (T Max ), minimum (T Min ) and mean temperatures (T Mn ) and rainfall were registered (Table 1)

Agronomic Determinations
When vegetative growth ceased, the mean leaf area per shoot was estimated destructively by measuring the area of all leaves of 10 shoots in each treatment using a LAI-Licor 3100 canopy analyzer (LI-COR Inc. Lincoln, NE, USA). This average leaf area per shoot was multiplied by the total number of shoots of 10 marked vines per experimental plot (40 vines per treatment) to estimate the total leaf area per vine (LA).
Yield at harvest was calculated by weighing ten control vines for each experimental plot, in accordance with [67]. All clusters per vine were weighed and counted in 10 marked vines of each experimental plots to determine the number of clusters and the yield per vine. Leaf area and yield relationship was estimated as the ratio between LA and yield in each of the 10 marked vines in each experimental plot.

Grape Samples
Grapes (V. vinifera, L. cv. Tempranillo) were collected at harvest from CICYTEX (Badajoz, Spain) vineyard during 2014, 2015 and 2016 growing seasons. To improve the physiological homogeneity of the different samples, berries were picked from the top, central and bottom parts of the cluster, following a zigzag path between two marked rows of 18 vines. These berries were calibrated according to their density [124]. Density was estimated by flotation of berries in seven different salt solutions from 115 to 175 g L −1 NaCl. Total soluble solids were measured on the several density ranges with a digital ATR ST plus refractometer (Schmidt + Haznsch, Berlin, Germany). At harvest, the berries selected were those which floated between 150 and 170 g L −1 , (TSS: 22.5-24.5 • Brix). Berry weight was determined according to the official methods of the International Organisation of Vine and Wine [125].

Extraction of Phenolic Compounds and Determination of Total Phenolic Content
Phenolic compounds were extracted from grape berries following the methodology previously described by [126] with some modifications. A total of 300 g of selected berries was crushed and homogenized in a blender for 1 min (speed 3, Worwek Model TM-31, Hamburg, Germany). Of the homogenate obtained, 50 g was macerated with 50 mL oxalic acid buffer 0.3 M (pH 1.00) during 16 h at 22-24 • C, and then centrifuged at 21,952× g, 4 • C for 10 min (Allegra 25R Beckman Coulter, Brea, CA, USA).

Analysis of Phenolic Compounds by HPLC
HPLC separation, identification and quantification of phenols were performed on an Agilent 1200 Series system (1200 LC; Agilent Technologies, Palo Alto, CA, USA) equipped with a degasser, quaternary pump, column oven, 1290 infinity autosampler, UV-VIS diodearray detector (DAD) and the Chemstation software package for LC 3D systems (Agilent Technologies) to control the instrument and for data acquisition and data analysis. Separation was performed in a Kromasil ® column 100-5-C18 250 × 4.6 mm, (Akzonobel, Bohus, Sweden). The supernatant of samples previously obtained was filtered (0.20 µm, Chromafil PET 20/25, Macherey-Nagel, Düren, Germany) and injected directly into the HPLC.
For identification and quantification of compounds, the analysis was carried out as described in [127]. The column was maintained at 40 • C. The mobile phase consisted of a gradient mixture of a solvent A (0.85% phosphoric acid solution) and solvent B (acetonitrile), with a flow rate of 1 mL.min −1 .

Statistical Data Analysis
The effect of treatment, year and treatment × year interaction was evaluated by a one-(treatment) or two-way (treatment, year) ANOVA. For the same crop load level, statistical comparisons between mean values were established with Student's t-test. The mean values of each experimental treatment were compared using Tukey's HSD test (p < 0.05). Finally, phenolic profile data were submitted to principal component analysis (PCA) with the aim of discriminating treatments based on association of the studied variables. The data analyses were performed using XLSTAT-Pro (Addinsoft, Paris, France, 2009).

Conclusions
Shoot thinning did not alter the effect of irrigation on the vegetative or agronomic parameters: the applied water stimulated vegetative growth, yield and berry weight in the three seasons of the study, both in shoot thinned and un-thinned vines. In regards to the impact on the phenolic content of the berries, it should be noted that for the majority of substances and phenolic groups, a similar behavior in both crop levels was found: irrigation caused in berries from both thinned and unthinned vines decreased in flavonoid compounds such as anthocyanins, flavonols and epigallocatechin, and between non-flavonoid compounds in t-resveratrol and cinnamic acid. However, the degree and statistical significance of the effect depended in some cases on the shoot density of the vine (i.e., anthocyanins and t-resveratrol), and in others it was similar (i.e., flavonols). In addition, it is noteworthy that the season effect was similar in both shoot densities and the impact of irrigation was greater in 2015 and 2016 than in 2014. Our findings are essential to vine growers and should help to decide the most appropriate crop level if full irrigation is applied. In addition, these results provide deeper information on the phenolic profile of cv. Tempranillo. Given the importance of these substances in organoleptic attributes of red wines, detailed knowledge of this profile would enable vine growers to adapt most adequate wine-making techniques to elaborate red wines of high quality.