The trial was carried out at California State University Fresno (Fresno, CA), in a twenty-five-year-old vineyard planted with Vitis vinifera, L. cv. Cabernet-Sauvignon on Freedom during the 2021 and 2022 growing seasons. Plants were spur-pruned (13.3 two-bud spurs/m) in a bilateral cordon and trained in a California Sprawling system. Cordon wire was positioned at 140 cm from the ground and a foliage support wire was placed at 30 cm above the cordon wire. Rows were oriented east-west and vine spacing was 1.8 m between the plants and 3.6 m between the rows. Plants were drip-irrigated during the whole growing season according to the grower’s practices, corresponding roughly to 3,000 m³ water/ha total (300 mm). The actual crop coefficient of these plants was not measured, but it is estimated to fall between 0.7 and 1 at full canopy size, according to visual assessment of the percent shaded area (Williams & Ayar, 2005). Meteorological data were taken from California Irrigation Management Information System (CIMIS) of the California Department of Water Resources (DWR) network station installed on site.

The trial was conducted using a completely randomised block design, with four treatments and four replicates of 8 vines (32 vines per treatment). Treatments consisted of: i) untreated control (C), ii) late leaf removal above the fruit zone (LR), iii) antitranspirant pinolene spraying (P), and iv) late shoot-topping (T), as shown in Figure 1.

Figure 1. Treatments applied in this study.

All treatments were applied manually at stage 36 of the modified EL scale (Coombe et al., 1995), on 20 August 2021 and 12 August 2022. The intent of the treatments was to reduce photosynthetic activity while maintaining shade. Leaf removal was performed by removing about 30 cm of leaves in the middle portion of each shoot. Pinolene [di-1-p-menthene (C₂₀H₃₄), Vapor Gard, Miller Chemical, USA] was mixed in a 5% solution with water and sprayed until the point of runoff with a backpack sprayer on the whole canopy. Topping was performed by removing the apical portion of the shoots at a height of 60 cm above the cordon wire. With topping the aim was to remove the most photosynthetically active apical leaves at the expense of greater sun penetration. With above-cluster leaf removal (mid-section of the shoots), the aim was to reduce sun penetration in the fruit zone compared to apical topping at the expense of removing less photosynthetically active leaves. The occasional higher cluster on very tall spurs was exposed, but that was a negligible percentage of fruit. With pinolene, the aim was to limit photosynthesis through a reduction in gas exchange while maintaining shading from all the leaves on the canopy.

The total leaf area per shoot was determined using a leaf area meter (LI-3100, LiCOR, NE, USA), on four shoots per replicate sampled from four different vines (for a total of 16 shoots per treatment) on 30 September 2021 and 23 September 2022. Light canopy interception was measured at noon on 10 September 2021 and 9 September 2022 using a line quantum sensor (LI-191R, LiCOR, NE, USA) inserted in the fruit zone and rightly above and parallel to the cordon, on four plants per replicate. Sunlight interception was calculated as % of ambient photosynthetic active radiation (PAR). Prior to data acquisition in each experimental unit, ambient light was acquired in an open road close to the field.

Leaf gas exchange was measured weekly during the central hours of the day (11:30–13:30 solar time) from early September until harvest. In each replicate, four mature and sun-exposed leaves from four grapevines were sampled from the middle section of the main shoot. Measurements were taken using a LI-6800 portable photosynthesis system (LiCOR, NE, USA). Carbon assimilation rate (A_N, µmol CO₂ m^-2 s^-1) and stomatal conductance (g_s, mmol H₂O m^-2 s^-1) were acquired by measuring inlet and outlet CO₂ (set at 410 ppm) and H₂O relative concentration (following ambient relative humidity at the time of the measurement). The intrinsic water use efficiency (WUEi) was derived as the ratio between net assimilation (A_N) and stomatal conductance (g_s).

Stem water potential (Ψ_stem) was measured on one fully expanded leaf per vine on four vines per replicate weekly from the application of the treatments to harvest. One hour before the measurement, each leaf was bagged with a mylar bag, and between 12:30 and 14:30, leaves were sampled and measured for water potential using a portable pressure chamber (PMS Instruments, OR, USA).

For each year of the study, grape ripening was monitored weekly, starting from 27 August 2021, and 12 August 2022, until harvest. In 2021, as the vineyard was commercially hand-harvested, the plants that were not used for yield assessment were held out from harvest and sampled for grape composition two weeks after the harvest date. Two samples of 24 and 48 berries per replicate were taken. The 24-berry set was used for the analysis of anthocyanin, while the 48-berry set was used for the analysis of total soluble solids, pH, and titratable acidity.

The 48-berry samples were crushed. The juice was used to determine total soluble solids (TSS), through digital refractometry (Fisherbrand, Thermo Fisher Scientific Inc.), pH was measured using a pHmeter (Accumet, AB315, Thermo Fisher Scientific Inc.), and the titratable acidity (TA) was measured through automatic titration with NaOH, 0.1N until a pH of 8.2 (Easy Plus, Mettler-Toledo, LLC, OH, USA).

The 24-berry sets were weighted and frozen to be later used to measure anthocyanins as described in Martínez-Lüscher et al. (2019). The skins were separated from the flesh with a scalpel, freeze-dried and then ground. An aliquot of 50 mg of the powder was extracted with methanol: water: 7 M hydrochloric acid (70:29:1, V:V:V). Extracts were filtered (0.45 μm, Thermo Fisher Scientific, San Jose, CA, United States) and analysed using reversed-phase high-performance liquid chromatography (HPLC) coupled to a diode array detector (DAD). The HPLC system was an Agilent 1260 series (Agilent, Santa Clara, CA, United States) with a reversed-phase C18 column LiChrospher® 100, 250 mm × 4 mm with a 5 μm particle size and a 4 mm guard column of the same material. Flow was set to 0.5 ml per minute and column temperature was set to 40 °C. Two mobile phases with a concentration of 5.5 % aqueous formic acid and formic acid in acetonitrile were used to ensure the following proportions (V/V) of acetonitrile were maintained: 0–8 min 8 %, at 25 min 12.2 %, at 35 min 16.9 %, at 70 min 35.7 %, 70–75 min 65 %, and 80–90 min 8 %. Peaks from chromatograms were then identified, observing the absorbance at 520 nm, and the area of each peak was compared to malvidin standards obtained routinely during data acquisition and used to determine the equivalent concentration of the various anthocyanins.

At the time of the commercial harvest (30 September 2021, and 23 September 2022), yield components were assessed as total yield per vine, the number of clusters per vine, and average cluster weight; average berry weight was instead obtained on berry samples as described earlier. In 2021, 100 clusters per treatment were used to quantify visible damages due to excessive solar radiation using a scale of no damage, mild damage, moderate damage, and severe damage as reported in Martínez‐Lüscher et al., 2019 (a picture scale can be found in Figure 2 of this reference and it was used for the current study). In 2022, the same measurement was carried out taking samples of 25 clusters per replicate.

Grapes from each treatment were manually harvested into picking bins of a max. 8 kg. The harvest date corresponded to the commercial picking decision for the field, when grapes reached 24 °Bx. Bins were stored in a cold room at 10 °C for 3 days. Each treatment was crushed separately, and must lots of 38 L were created (3 replicates per treatment, total of 12 independent fermentations). Potassium metabisulfite was added at 30 ppm to every wine lot at crush. All wines were inoculated with the same commercial yeast (Saccharomyces cerevisiae, Enartis WS) according to standard dosage recommendation (25 g/hL). Yeast assimilable nitrogen (YAN) was adjusted in each treatment by adding diammonium phosphate (DAP) to reach a concentration of YAN (in mg/L) equal to ten times the content in soluble solids (in °Brix). Cap management was done by manual punch-downs once a day in the morning, and must samples were collected for monitoring the progress of fermentation by measuring sugar concentration and temperature. At the end of alcoholic fermentation, wines were pressed using a vertical press, racked into 19 L stainless steel kegs, and allowed to settle in a cold room at 50 °F for four weeks. After settling, wines were racked and filtered using a benchtop filter (Buon Vino Super Jet Filter) with cellulose pads of 5–7 microns pore size. Free sulfur dioxide was adjusted in all finished wines to 30 ppm. Wines were bottled in 750 mL bottles with agglomerate corks and kept in storage at 10 °C until analysis.

Must and finished wine analysis include titratable acidity (TA), pH, and °Brix. Fermentation °Brix and temperature were monitored with a hydrometer and digital thermometer (Easy Dens, Anton Paar). Colour intensity (absorbance at 420, 520, and 620 nm) was measured with a Perkin Elmer Lambda 25 Spectrophotometer using quartz cuvettes of 2 mm path length. Wine anthocyanins were measured using HPLC-DAD with the same method described in section 5 of the Materials and methods for grapes.

Statistical analysis was performed using R version 4.3.1–package “agricolae” (de Mendiburu, 2019). Data were subjected to one-way analysis of variance (ANOVA). When the results of ANOVA were significant at p ≤ 0.05, data were then subjected to Tukey’s HSD test. Graphical representations were also obtained with R.

Meteorological data are shown in Figure 2 and Table S1. Considering the period from 1 April to 31 October, minimum temperature averages were slightly lower in 2021 (-0.3 °C), while both mean temperatures and maximum temperatures were higher in 2022 (+0.6 °C and +0.4 °C respectively). In 2022 a higher number of days with temperatures exceeding 38 °C was also registered. Even though both seasons were particularly dry, especially during summer, in 2022 the rainfall from April to October was half that of 2021. Additionally, in 2022, there was less rainfall during the first months of the year compared to 2021, which probably made it more challenging to establish sufficient water reserves in the soil for the growing season. In 2022, the air temperatures recorded during the period between the application of treatments and the harvest were higher compared to 2021. Mean and maximum temperatures surpassed 30 °C and 40 °C respectively, particularly towards the latter part of August. Total ET₀ was of 1,551 mm in 2021, with 1,280 mm from 1 April to 31 October, and 1,632 mm in 2022, with 1,338 mm from 1 April to 31 October.

Figure 2. Climatological trends in 2021 and 2022 in the study area. Red, green, and cyan lines indicate maximum, mean, and minimum air temperatures (°C) respectively. Blue bars indicate daily rainfall (mm). Black arrows and dashed lines indicate the treatment application date. Orange arrows and dashed lines indicate the harvest date.

Figure 3 shows the total leaf area/shoot for both growing seasons, including leaf area on laterals. In both years, C and P showed the highest values that were not different, as they were not subjected to canopy manipulation. Specifically, the highest value was registered for control vines (2,715 cm²) in 2021, and for pinolene (3,054 cm²) in 2022.

In both years, compared to the control, canopy manipulation led to a significant decrease of total leaf area. Amongst all treatments, T registered the lowest values in both years, with 1,049 cm² and 845 cm² in 2021 and 2022 respectively. Similarly, LR caused a reduction of total leaf area, leading to 1,582 and 1,882 cm² for the two years of analysis. Although the differences between the two canopy manipulation practices were not statistically significant, T was a more aggressive practice that reduced shoot leaf area by 61 % and 62 %, while LR reduced it by 42 % and 23 % in 2021 and 2022 respectively. Reductions in total leaf area/shoot with canopy-manipulation practices were similar to those reported in other studies (Palliotti et al., 2013a; Filippetti et al., 2014; Valentini et al., 2018; Lu et al., 2022; Tessarin et al., 2022).

The percentage of light in the fruiting zone with respect to the ambient light was affected by the treatments (Figure 3). As T was the treatment that reduced shoot leaf area, it led to the highest exposure of clusters at noon compared to the other treatments. Specifically, in 2021 light percentage in T vines was significantly increased by 8 % and by 11 % in 2022 with respect to C. Although not significant, LR also exhibited a trend towards higher exposure of the fruiting zone (6 % in 2021 and 11 % in 2022). It's important to note that the measurements were taken at noon, when the sun was perpendicular to the ground, while leaves were removed on the sides (Figure 1), therefore increasing light penetration when the sun was oblique to the ground. It can be assumed that if measurements had been taken in the morning or afternoon, the percentage of light in the fruit zone would have been higher and possibly significantly different from C. C vines exhibited the lowest light percentage in the fruit zone in both years, at 4 % and 6 % respectively. P and C vines were not significantly different, although slightly higher values were recorded for P, but this variation is not directly attributed to the treatment itself, although P increased the glossiness of leaves (Figure 1), which could have an impact on the amount of reflected light.

Figure 3. Total leaf area/shoot and percentage of light in the fruiting zone in 2021 and 2022 for the four treatments: control (C), late above-cluster leaf removal (LR), pinolene (P) and late topping (T). When ANOVA was significant, different letters indicate differences between treatments with p-value ≤ 0.05 (Tukey-HSD test); error bars represent standard error of the mean (n = 16).

Trends for the stem water potential (Ψ_stem) of the four treatments are shown in Figure 4. Values in 2022 were generally lower, indicating an elevated stress condition in the vines because of the higher temperatures and reduced rainfall during that vintage (Figure 2). In 2021, P leaves exhibited the lowest value (-1.77 MPa), which coincidentally is also the minimum value recorded across both years. In 2022, both P and LR treatments recorded the lowest value (-1.65 MPa). In both vintages, T vines recorded higher values compared to the other treatments, especially at the end of the season, and this effect was significant in most of the measurement dates. This effect is related to the strong reduction of the canopy that was induced with the treatment, removing more than 60 % of total leaf area in both years. Poni et al. (2023) already reported that summer pruning practices can reduce single-vine water use, which aligns with our findings. In the case of LR, this trend was observed in 2021 but not in 2022. In this case, the different response is most probably related to the different intensity of the treatment, which was lower in the second year of study, as already discussed. Other authors have shown that in warm climate, reducing leaf area only limits water use in the driest conditions and with high yields (Poni et al., 2006; Abad et al., 2018). Although the second year was warmer, yields were lower. This observation also aligns with Abad et al. (2018), who found that canopy reduction effects on water demand are non-linear, with more substantial reductions yielding the best results. Consequently, our results suggest that it is possible to decrease water stress by reducing the total leaf area by at least 40 %, while reductions of less than one-third do not determine mitigations to water stress conditions. Nevertheless, since stem water potential measurements were taken at midday, this may have affected the accuracy of our results. Accuracy can diminish under extreme temperatures and at elevated levels of water deficit (Deloire et al., 2020); therefore, conducting measurements at alternative times could potentially improve precision and sensitivity (Williams & Araujo, 2002; Heyns, 2017).

Although unexpected, pinolene did not alleviate water stress in the treated vines. However, examining the gas exchange values, which will be discussed further, P only showed a tendency to reduce transpiration and stomatal conductance, with minimal to no significant difference compared to C, in contrast to findings from other studies (Palliotti et al., 2013a; Di Vaio et al., 2019). Therefore, it is likely that in this trial, the modulation of gas exchange was not sufficient to significantly reduce water loss and prevent the onset of stress conditions.

Figure 4. Trends for stem water potential (Ψ_stem) in 2021 and 2022. When ANOVA was significant, different letters indicates differences between treatments with p-value ≤ 0.05 (Tukey-HSD test). Shaded regions indicate the standard error of the mean. C indicates control, LR indicates leaf removal, P indicates pinolene and T indicates late topping. Treatments were applied on 20 August 2021 and 12 August 2022. Harvest took place on 30 September 2021 and 23 September 2022.

Figure 5 shows leaf gas exchange data in the four treatments in 2021. Considering net assimilation (A_N), T showed the highest values compared to the other treatments, with the maximum value registered at the beginning of the measurements (10.54 µmol CO₂ m²s^-1). On the other hand, P reduced the photosynthetic activity of the treated vines compared to C, even though the observed difference was not statistically significant. In the case of stomatal conductance, g_s, the highest value was 0.21 mol H₂O m² s^-1, registered by T on the first sampling point date, while the lowest value was 0.08 mol H₂O m^-2 s^-1, registered by both P and LR on the last date. No treatment showed significant differences when compared to C vines. However, T maintained the greatest g_s in all the observation dates, while P the lowest, and this difference was statistically significant on one of the dates. Both g_s and A_N decreased across all treatments as the season progressed; this was a result of increased water stress, as reported in Figure 4, and because of leaf aging, which reduced photosynthetic functionality, consistent with findings by Lupo et al. (2023). Further analysis via principal components analysis (PCA), shown in Figure S1, supports these results. As expected, physiological parameters like A_N and g_s were directly correlated with stem water potential. Consequently, treatments that mitigate stress conditions, such as T, demonstrated improved vine performance. In this context, T was the most effective treatment, showing distinct clustering from the control in PCA results.

Pinolene, known for its antitranspirant properties due to its film-forming capacity (Fahey and Rogers, 2018), exhibited similar effects after we applied it to Cabernet-Sauvignon. However, contrary to findings in other studies (Palliotti et al., 2013a; Brillante et al., 2016; Di Vaio et al., 2019; Silvestroni et al., 2020), water use efficiency in treated plants was not enhanced. This may be due in part to the heightened climatic stress during our study period compared to those prior investigations. It is worth noting that the observed reduction in g_s was not statistically significant compared to C, probably due to the more stressful conditions of our trial that diminished C functionality, which might explain the absence of differences in intrinsic water use efficiency. Although the concentration rate was already higher than that reported in previous studies (Brillante et al., 2016; Di Vaio et al., 2019), it is possible that using a higher concentration of pinolene could yield different results.

Several studies have reported evidence of increased photosynthetic activity as a form of compensation following a reduction in the source-sink ratio (Petrie et al., 2003; Kliewer & Dokoozlian, 2005; Bobeica et al., 2015). In this study, a similar trend was observed in T vines throughout the season, but not in LR vines. We believe that the absence of this result in LR vines may be linked to the smaller reduction in the source (leaves) that these vines underwent, as discussed earlier. Hence, this reduced reduction in the source may not have been sufficient to trigger the overcompensation mechanisms at the leaf level in the plants. This interpretation aligns with the PCA results, where LR showed a weaker correlation with physiological parameters, such as A_N and g_s compared to T.

Figure 5. Trends for net photosynthesis (A), stomatal conductance (g_s), transpiration (E) and intrinsic water use efficiency (WUEi) in 2021. When ANOVA was significant, different letters indicates differences between treatments with p-value ≤ 0.05 (Tukey-HSD test). Shaded ribbons indicate the standard error of the mean. C indicates control, LR indicates late above-cluster leaf removal, P indicates pinolene and T indicates late topping. Treatments were applied on August 20, 2021. Harvest took place on September 30, 2021.

Figure 6 displays the evolution of total soluble solids and must pH for 2021 and 2022. In 2022, ripening occurred more rapidly than in the previous year, leading to harvest one week earlier. The highest sugar concentration between both years was 27.7 °Brix, recorded for P in 2022.

In 2021, no significant difference in total soluble solids was observed from ripening through harvest for all treatments. T was the only treatment capable of significantly decreasing sugar accumulation, although only at the end of the season (post-harvest date) when more pronounced effects of the applied treatments were observed. Specifically, TSS content was lower in P and T compared to C vines (reductions of -1.4 and -2.1 °Brix respectively), with a significant difference in the case of T. In 2022, sugar concentration in T was the lowest compared to the other treatments, similar to the previous year. However, in contrast to the prior year, P exhibited the highest sugar concentration throughout the entire second half of the season. It was interesting to note that in 2021 sugar accumulation followed a linear pattern with a slow rate, in contrast with 2022, where a fast accumulation rate and a maximum sugar concentration was reached one week prior to harvest, followed by a decrease of approximately 2 to 3 °Bx in LR and P respectively, but not in C and T. Despite at harvest the difference was no longer significant, T reduced the pace of sugar accumulation in the berries. This trend is in accordance with findings from other studies (Filippetti et al., 2014; Valentini et al., 2018; Lu et al., 2023; Poni et al., 2023), which already reported the effectiveness of late trimming in delaying berry ripening.

Several studies have already proven that late leaf removal can delay ripening by slowing down sugar accumulation (Lanari et al., 2013; Caccavello et al., 2017; Poni & Gatti, 2017). However, our results did not show any effect on berry ripening, which is similar to the findings of Tessarin et al. (2022). Palliotti et al. (2013b) observed that to determine a significant reduction in sugar accumulation, vine leaf area should be reduced by 30–35 %. In our case, this reduction was achieved only in 2021 but not in 2022, as previously discussed. Additionally, to fully understand the mechanisms governing the effects of defoliation, the age and the functionality of the leaves must be considered. As indicated by Poni et al. (2023), throughout the season leaf functionality shifts towards the top of the shoots, as leaves at the bottom tend to be older. Therefore, we suggest that the lack of observed effects on grape ripening may be due to the removal of the leaves from the middle section of the shoots, and not from the upper section, thus limiting the impact on vine photosynthesis.

In 2021, differences in must pH between treatments were significant only in the first half of the season. At harvest, none of the treatments produced a significant modification of pH compared to control. The lowest pH value was 4.16, recorded for P, while the highest was 4.74 for LR, the highest value observed in both years. Starting from mid-September, LR consistently had the highest pH among all treatments, while although not statistically significant anymore, both P and T lowered the pH compared to C. In 2022, a similar pattern emerged for P, primarily resulting in the lowest pH compared to the other treatments, with a minimum value of 3.75, also the lowest over both years. However, topping, which had restrained pH in the previous year, showed higher values, while LR, which exhibited the highest values in 2021, led to a lower must pH. Overall, none of the treatments were able to mitigate the drop in pH which is typically associated with the respiration metabolism of organic acids (notably malic acid) in grapes. The high pH values shown across all treatments at harvest would typically require a significant addition of tartaric acid for winemaking purposes in order to decrease pH (ideally below 4.0) and increase titratable acidity.

Figure 6. Trends for total soluble solids (TSS) and pH in 2021 and 2022. When ANOVA was significant, different letters indicates differences between treatments with p-value ≤ 0.05 (Tukey post hoc test). Shaded regions indicate the standard error of the mean. C indicates control, LR indicates leaf removal, P indicates pinolene and T indicates late topping. Black arrow indicates the harvest date.

Yield components and grape composition at harvest are shown in Table 1. Crop level was higher in 2021 than in 2022, with an average yield per vine of 12 kg and 8.3 kg respectively. Similarly, average berry weight in 2022 was lower by 0.14 g, possibly due to the hotter and drier climate of this year, resulting in increased water stress. Martínez-Lüscher et al. (2017b) already reported similar impact on yield in the San Joaquin Valley of California when precipitation before budbreak is reduced, because of reduced cluster number. In 2022, the average cluster number across all treatments was reduced by 33 % with respect to 2021.

Cluster number per vine and average berry weight were not significantly different in 2021, nor was the total yield per vine. In 2022 we observed a higher number of clusters per vine for plants subjected to LR and T and a significantly lower berry weight for P vines, reduced by 0.26 g when compared to the control. A possible explanation regarding the increase in cluster number in LR and T, which lacks data to be confirmed, could be an increase in shoots from latent buds as a result of the increased light in the cordon zone following the first year of treatment. Because these vines are not shoot-thinned, more secondary shoots could have increased the cluster number. This hypothesis is corroborated by the fact that a substantial increase in cluster number did not correspond to a significant increase in yield, which remained similar to C in LR and T, as secondary shoots produce small and light clusters.

Although the total yield per vine in 2022 was not significantly different between treatments, the reduced average berry weight observed for pinolene resulted in the lowest crop level registered in both years, just 8 kg per vine. This was over 1 kg more compared to C in the same year, as well as when compared to the other treatments. In contrast to Fahey and Rogers (2018), who reported that pinolene can reduce berry dehydration in treated plants, our observations showed only a slight increase in berry weight in 2021 and a notable reduction in 2022. As previously discussed, the 2022 vintage was exceptionally hot and dry, certainly impacting the water status of the plants and leading to increased stress conditions compared to the previous year. Therefore, we hypothesise that the reduction in berry weight observed in 2022 is due to pinolene's removal of waxes on the berry surface, thus exposing the grapes to higher dehydration in vintages where plants are particularly susceptible to water stress conditions (example pictures in Figure S2).

Grape must composition at harvest was not affected by the applied treatments. Considering all treatment together, the average total soluble solids content was of 24.5 °Brix in 2021 and 24.0 °Brix in 2022. In both years, the highest total soluble solids content was 24.8 °Brix, measured in LR in 2021 and in P in 2022, while T always showed the lowest values: 23.8 and 22.9 °Brix for 2021 and 2022 respectively. The average must pH was 4.53 in 2021 and 4.39 in 2022, with the greatest value observed in LR in 2021 (4.56) and the lowest in P in 2022 (4.31). Considering the two vintages altogether, titratable acidity was slightly higher in 2022 (4.00 g/L) than in 2021 (average of 3.27 g/L). This result was unexpected, as higher temperatures are known to increase respiration rates and the metabolism of malic acid (Keller, 2020). The lowest value was 2.98 g/L for T in 2021, while the highest was 4.39 g/L for C in 2022. From a winemaking perspective, the possibility of preserving organic acids in grapes would be a desirable attribute; nevertheless, none of the treatments influenced pH or TA.

*Grape composition at harvest for all treatments in 2021 and 2022. TSS indicates total soluble solids, TA indicates titratable acidity, abw indicates average berry weight. In case of significant ANOVA, different letters across a column indicate significant difference between treatments with a p-value < = 0.05 (T-test). C indicates control, LR indicates leaf removal, P indicates pinolene and T indicates late topping. Total anthocyanins in 2021 were measured on a sampling date after harvest, on October 18, 2021.

Pinolene already proved to effectively delay ripening in other studies (Brillante et al., 2016; Di Vaio et al., 2019; Silvestroni et al., 2020). In this trial, a slight decrease in TSS accumulation was observed in 2021 but not in 2022. In our opinion, the difference found in the cluster number per vine between the first and the second year could play a key role in the explanation of the different outcomes observed. As a matter of fact, in 2022 we found a significantly lower number of clusters in pinolene-treated vines. The sugar content per berry values supports this theory, as it was significantly lower in P vines compared to the control (0.18 g compared to 0.24 g), indicating that if not for the difference in the number of clusters, we might have seen a lower TSS concentration in P vines.

Grape production was also assessed, considering the extent of sun damage on clusters (Figure 7). In 2021 we observed a higher frequency of clusters with no visible damages compared to 2022. At the same time, we also found the highest percentage of severe damage in 2021, with almost 60 % of clusters from the LR treatment seriously compromised by sun damages. In both years, the assessment of berry damage from C and P grapes did not show differences, and both treatments had the best quality among all the treatments under analysis. While in 2021 the extent of the damages on T was similar to C and P, in 2022 we registered a higher frequency of mild and moderate damage (classified as B and C), while fewer clusters had severe damages. In the case of LR, we found the highest percentage of damaged clusters in both years, especially with severe intensity.

As expected, the canopy manipulation operated with both leaf removal and shoot topping resulted in an overexposure of the clusters to high solar radiation, which is detrimental for grape quality and leads to a higher risk of sunburn and damage (Poni et al., 2023). Damages in T were overall milder than LR, and this is a consequence of the different exposures determined by the two treatments. By removing leaves, which were operated on both sides of the canopy, clusters were exposed to direct sunlight for an extended duration during the morning and the afternoon. On the contrary, shoot topping concentrated the exposure of the fruiting zone mainly during the central hours of the day.

Figure 7. Percentage of no damages (A), mild damages (B), moderate damages (C) and severe damages (D) on the clusters at harvest. When ANOVA was significant, different letters indicates differences between treatments with p-value ≤ 0.05 (Tukey-HSD test). C indicates control, LR indicates leaf removal, P indicates pinolene and T indicates late topping.

Berry anthocyanins evolution is shown in Figure 8. The highest concentration recorded was 357.4 mg/kg of berry in 2021 and 343.5 mg/kg of berry in 2022, both for P. Except for one date in 2021, no significant difference was observed between the treatments in both years. On this occasion, T exhibited the lowest concentration, with just 127.8 mg/kg of berry. After an initial accumulation, anthocyanin concentration tended to decrease in the latter part of the season, independently from the treatment applied. P showed the highest concentration in 2021 and throughout 2022, except for the harvest date, when it was the lowest, although the difference was not significant. At harvest, total anthocyanins (Table 1), as well as berry anthocyanin profile (Table S2), were unaffected by the treatments.

Di Vaio et al. (2019) already reported an increase in anthocyanins concentration thanks to the application of pinolene, while other studies have highlighted a negative effect on the phenolic substances (Palliotti et al., 2013a; Brillante et al., 2016; Di Vaio et al., 2020; Silvestroni et al., 2020). Our results do not align with any of the cited theories, rather suggestting there is no effect in this regard.

LR and T exhibited the lowest concentration in 2021. In 2022 this trend was confirmed for LR but not for T, which ended up with the highest content at harvest after C vines. The role of light and temperature on anthocyanin content has been largely described. While increased exposure can promote their accumulation (De Toda et al., 2015; Tessarin et al., 2018), a reduction in their content may derive from both under and overexposure (Mori et al., 2007; Chorti et al., 2010). On the other hand, other studies did not highlight a direct effect of both leaf removal (Filippetti et al., 2014; Valentini et al., 2018) and shoot topping (Poni & Gatti 2017; Tessarin et al., 2022) on berry anthocyanins concentration. It appears that the level of exposure we obtained was not high enough to determine an increase or a decrease of these compounds at the berry level, rendering our results more in line with these latter findings.

Figure 8. Berry anthocyanins evolution in 2021 and 2022. When ANOVA was significant, different letters indicate differences between treatments with p-value ≤ 0.05 (T-test). Shaded regions indicate the standard error of the mean. C indicates control, LR indicates leaf removal, P indicates pinolene and T indicates late topping. The black arrow indicates the harvest date. In 2021, the experimental vines that were not used for yield assessment were sampled two weeks after the harvest date to observe the effect of treatments with a longer hang time.

Wine produced from the different treatments was analysed and results are reported in Table 2. Overall, alcohol content was higher in 2022 than in 2021, In 2022, the lowest alcohol content was 15.2 % for the C treatment. In both years, LR exhibited the highest alcohol concentration, with a maximum of 18.8 % in 2022. This was likely due to a higher incidence of sun damage and raisining as shown by the results from Figure 7. Berry shrivelling and sun damages are well known to increase the concentration of soluble solids in must. These berries were likely excluded from the grape composition samples shown earlier.

In 2022, T and LR treatments reduced wine acidity (4.81 and 5.01 g/L respectively), even though each treatment had higher content when compared to the previous year. LR also exhibited a higher pH in the same year. For this study, acidity in wines was not modified in any treatments. However, conventional winemaking practices would typically necessitate the addition of tartaric acid to counterbalance the high pH in the musts. Standard targets for titratable acidity exceeding 5 g/L, and pH values below 3.8 are generally preferred to optimise wine colour and enhance the efficacy of sulfur dioxide against spoilage microorganisms and oxidation (Ribéreau-Gayon et al., 2006).

Optical density was measured at 420, 520, and 620 nm to assess wine colour. In 2021, significant differences were found only at 620 nm, and in this case, LR resulted in the highest value of 0.11, which corresponds to nearly 3 % of the total colour intensity (sum of the three optical densities). In 2022 a significant drop of all colour parameters was observed in comparison with 2021 (~70 % less colour), which is related to the decrease of anthocyanins on berries. Regardless of this difference, we observed significant differences in all wavelengths, where LR had the highest colour intensity followed by T. Wine from the control always showed the lowest colour values. In 2022, LR confirmed the trend observed in 2021, maintaining the highest values at each wavelength with significant differences compared to most treatments. An improvement in wine colour was observed by Tessarin et al. (2022). Our results over two consecutive years confirm that leaf removal can effectively improve wine colour. However, we also believe that in the second year, the higher alcohol content in the wine played a key role in facilitating colour extraction, thus intensifying the effects of the defoliation. In addition, as presented in Figure 7, LR had a higher number of dehydrated berries that increased the amount skin/juice ratio in the musts, while they were excluded from analysis of anthocyanins reported on berry analysis (Table S2).

Total anthocyanins content was significantly higher in 2021 compared to 2022 in all treatments. In 2021, C had the highest total anthocyanin concentration, whereas in 2022, T had the highest concentration. Although the berry content at harvest did not differ between treatments, LR resulted in the lowest concentration of total anthocyanins in the wine in both years. As expected, the major contributor to total anthocyanins were glucosides in both years (average 57 % across all treatments in 2021 and 50 % in 2022). Acylated anthocyanins (acetylated, and coumarylated) in C were significantly higher than all treatments, whereas in 2022, T was the treatment with the highest concentration.

Interestingly, in 2022, the percentage of acetylated anthocyanins was higher (approximately 46 %) compared to 2021 (37 %). These ratios are relevant for winemaking, because acylated anthocyanins are known to contribute colour stability in wines by preventing bleaching by sulfur dioxide and preventing pigment degradation by means of copigmentation (Brouillard & Dangles, 1994; Giusti & Wrolstad, 2003; Waterhouse et al., 2016). In the case of our results, while the concentration of acylated anthocyanins were lower for all treatments in 2022, the colour is expected to be more stable during wine ageing. Similarly, T showed a higher rate of 3’4’5’ hydroxylated compounds than the other treatments. In the hottest 2022, the amount of these compounds was 50 % higher in T than in C, although this difference was not statistically significant. The 3’4’5’ hydroxylated anthocyanins are more stable in wine.

An interesting aspect of the data is the lack of strong correlation between anthocyanins in the berries and in the wine. Principal Component Analysis (PCA) data from both 2021 and 2022 indicate a clear separation between berry and wine anthocyanins (Figure S1), showing no strong correlation between their distributions. This suggests that the anthocyanin content in the wine is influenced by factors beyond berry anthocyanin concentration, such as alcohol content, which may enhance their extraction (He et al., 2012). Additionally, the abundance of raisins and dehydrated berries in the winemaking process may have further skewed the final anthocyanin profile in wine (Tarara et al., 2008; Torres et al., 2020). Our results suggest a strong correlation between water stress and the final anthocyanin content in the wine. Treatments that induced more significant water stress, like P and C in 2021, resulted in a notably higher concentration of total anthocyanins in the wine. This could be due to the increased concentration of anthocyanins that develop as a physiological response to water stress, a phenomenon well-documented in various studies (Mirás-Avalos & Intrigliolo, 2017).

Depending on the level of exposure and bunch temperature, studies on the behaviour of anthocyanins in response to increasing light exposure can be quite contradictory (Mori et al., 2007; Chorti et al., 2010; Alatzas et al., 2023). Our results suggest that increased exposure to sunlight in LR might lead to a lower concentration of total anthocyanins in the wine due to their degradation. This is consistent with results from many other trials in California (Martínez‐Lüscher et al., 2017a; Martínez‐Lüscher et al., 2020; Torres et al., 2020; Torres et al., 2021). In the case of T, the low concentration observed in 2021 may also be related to a reduced sugar accumulation, therefore to a delay in bio-synthesis and not to degradation.

While the other treatments showed more inconsistency over the two years, LR consistently affected the concentration of all anthocyanin groups. LR resulted in an increased content of total delphinidins and cyanidins in 2021. When we consider the relative abundance of these components, it was increased significantly by LR in both years. LR also decreased the concentration in 2021 and their relative abundance in both years. Unexpectedly, LR had the highest colour intensity (with T) but the lowest concentration of anthocyanins each year.

These results suggest that anthocyanin concentration in the wine is subject to more complex dynamics, influenced by environmental factors like light exposure and water stress. However, the lack of a strict correlation between berry and wine anthocyanins implies that the winemaking process plays a crucial role in determining wine color quality and stability.

*Wine composition in 2021 and 2022. In case of significant ANOVA, different letters across a column indicate significant differences between treatments with a p-value < = 0.05 (Tukey post hoc test). TA indicates titratable acidity, OD indicates the optical density of the wines measured at 420 nm, 520 nm, and 620 nm. Delphinidin, Cyanidin, Petunidin, Peonidin, and Malvidin are the sum of all the corresponding compounds, including glucosides, acetylated and coumarilated (e.g., Delphidin is the sum of Delphinidin-3-glucoside, Delphinidin-3-(acetyl)-glucoside, and Delphinidin-3-(cou)-glucoside). Total Anthocyanins are reported as the sum of Delphinidin, Cyanidin, Petunidin, Peonidin, and Malvidin in all forms. C indicates control, LR indicates leaf removal, P indicates pinolene, and T indicates late topping.