Canopy management through crop forcing impacts grapevine cv. ‘Touriga Nacional’ performance, ripening and berry metabolomics profile

Climate changes are speeding up the maturation of grapes in numerous areas of the world, including in the Mediterranean basin, but warmer temperatures often uncouple technical and phenolic maturity, resulting in unbalanced wines. We tested the efficacy of crop forcing (CF) in delaying the maturation of cv. ‘Touriga Nacional’ vines of the Douro Region, and their impacts on plant performance, berry quality attributes and metabolome were also evaluated. In two consecutive seasons (2019 and 2020), CF was conducted 15 (CF1) and 30 (CF2) days after fruit set by hedging growing shoots to five nodes and removing summer laterals, leaves and clusters. Results showed that while CF2 delayed ripening up to 51 days till first autumn rainfall, which compromised optimal sugar ripeness, CF1 delayed the technical maturation by one month, but both treatments severely impacted the production, mainly CF1, which reduced grapevine yield up to 90 %. The effect of CF in protecting vines against drought stress was not evident, judging by the values of leaf pre-dawn water potential measured along both seasons. CF1 and CF2 resulted in berries with lower pH and higher titratable acidity than controls, while total phenolics content increased by up to 48 % in 2020. A UPLC–MS-based targeted metabolomic analysis showed that CF increased the relative abundance of key metabolites like flavan-3-ols (i.e., catechin gallate increased by up to 661 %), trihydroxylated anthocyanins (i.e., delphinidin-3-O-glucoside increased by up to 656 % after CF2) and stilbenes (resveratrol increased by up to 700 % after CF2) with potential positive impacts in wine quality.


INTRODUCTION
Over a period of more than a century, there has been a 1 °C increase in mean land surface temperature, and climate projections for the end of the 21st century forecast further increases in temperature, depending on the mitigation measures adopted (IPCC, 2014;Martínez-Lüscher et al., 2016).These changes in climate result in the occurrence of extreme weather events, including heat waves and periods of extreme drought (Ramos, 2017;Fraga et al., 2018).Numerous highly important wine regions in the world are located in areas of the planet highly vulnerable to climate change.This is the case of the Mediterranean basin, including the Douro Demarcated Region (DDR), with models for this region predicting significant thermal increases and decreases in rainfall occurrence (Jones and Alves, 2012;IPCC, 2014;Fraga et al., 2016;Costa et al., 2019).
Different management strategies have been tested to mitigate the above-mentioned climate change effects on fruit maturation, including the displacement of vineyards for cooler regions, the selection of better-adapted cultivars, clones or accessions, or the implementation of new viticultural practices (Poni et al., 2018;Martínez-Lüscher et al., 2016).Some of these viticultural practices, including the application of chemical compounds, such as auxins (Böttcher et al., 2011), late winter pruning (Zheng et al., 2017a;Moran et al., 2017), shoot trimming (Martínez de Toda et al., 2014;Zheng et al., 2017b;Santesteban et al., 2017), and minimal pruning (Clingeleffer, 2010;Zheng et al., 2017c;Molitor et al., 2019), have shown a positive delay in sugar ripeness up to 10 to 15 days when performed individually, or even 15 and 45 days when combined.
Forcing bud regrowth (crop forcing) is still a poorly explored viticulture practice that consists of removing growing shoot parts, including tips, summer laterals, clusters and leaves, leaving several nodes, which delays maturation for up two months (Dry, 1987;Gu et al., 2012;Lavado et al., 2019;Martínez de Toda et al., 2019;Pou et al., 2019;Prats-Llinàs et al., 2020).The crop forcing releases the control exerted by the shoot apex and the summer laterals over the outgrowth of the dormant buds that normally produce the inflorescences in the upcoming season (Dry, 1987;Fang et al., 2000;Lavee and May, 1997), enabling the regrowth of new vegetative and reproductive structures (Cline, 1991;Cline, 1994;Martínez-Moreno et al., 2019).
In cv.'Cabernet-Sauvignon' and 'Tempranillo' vines, crop forcing in June shifted fruit ripening from the hot (July and August) to the cool season (October through early November).Smaller and more acidic berries (due to increments of both tartaric and malic acids) were produced, with higher contents of total anthocyanins, tannins, and total phenolics (Gu et al., 2012;Martínez-Moreno et al., 2019).In another study, berries of cv.'Tempranillo' accumulated significantly higher total polyphenols and anthocyanins in response to crop forcing, and wines were notably rich in tannins, but the cluster's weight was compromised, and, therefore, the yield, although a higher number of clusters was produced (Lavado et al., 2019).Results also showed that yield reduction was less pronounced when crop forcing was applied later in the season.Yield reduction was also observed in other studies (Gu et al. 2012;Martínez de Toda et al., 2019), whose intensity depended on the time of crop forcing application (Martínez de Toda et al., 2019).
A bud-forcing technique aimed at obtaining two crops per season was explored in Pinot Noir vines grown in pots (Poni et al., 2021).The primary crop was maintained, but a second, late-ripening crop was ensured through the release of dormancy of the auxiliary buds.The technique proved feasible to warrant two crops in the same season with different features.
Forcing vines to regrowth is an aggressive technique (Lavado et al., 2019), whose advantages need to be carefully validated for each wine region, growing season and cultivar.The present study was implemented in the context of the European project, VISCA (Vineyards Integrated Smart Climate Application, https://visca.eu/), which aimed to make European wine industries resilient to climate changes through the application and optimisation of techniques like crop forcing, shoot trimming and irrigation assisted by medium-long term forecast models.In Douro Demarcated Region (DDR), the three main grapevine varieties, 'Touriga Nacional', 'Touriga Franca' and 'Tinta Roriz' ('Tempranillo'), are cultivated during part of the growing season in marginal weather conditions for agricultural production (low precipitation and high radiation and temperature) that tend to worsen in the future (Gouveia et al., 2011).Models predict advances in budburst of up to 14 days and in flowering and veraison of up to 10 days in 'Touriga Nacional' cultivated at low elevations along the Douro River in a time window from 2051-2080 (Reis et al., 2022).Although this cultivar is considered adapted to warm climates, with a high capacity for heat dissipation and for withstanding high light intensities (Carvalho et al., 2016), elevated temperatures tend to increase the pH and sugar content of the berry and decrease berry weight, titratable acidity and anthocyanins (Costa et al., 2020).Thus, in the present study, we hypothesise that crop forcing may delay the maturation and prevent this uncoupling of technical and phenolic maturity in cv.'Touriga Nacional'.Moreover, we aim to perform a metabolic berry analysis to characterise in detail the modifications of the secondary metabolism in response to crop forcing in this cultivar.

Field conditions and experimental site
The experimental trial was conducted during the 2019 and 2020 seasons in a commercial vineyard with sandy loam soil located in the Douro Superior sub-region, Portugal (41º14'36''N,7º06'55''W), at an altitude of about 140 m.The cultivar 'Touriga Nacional' (Vitis vinifera L.) used in the study was planted in 2014 and grafted in 196-17Cl rootstock.Plants were oriented in west southwest to east/northeast, spaced at 2.2 m between rows, 1.0 m along the row and trained on a vertical trellis (4546 plants/ha), uniformly pruned on a unilateral Royat cordon, ca. 10 buds per vine.Eight adjacent rows were selected to implement a randomised block design.Plants were manually winter pruned, when all the leaves had fallen, to four to five spurs and two buds per spur.
The experiment considered three treatments, a control (CF0) with vines grown under conventional practices (only winter pruning) and two crop forcing dates (CF1 and CF2), with four replicates per treatment and five plants per replicate.The crop forcing consisted of hedging the growing shoots to five nodes and removing all the summer laterals, leaves and clusters manually to force the bud break of the dormant bud developed in the current season.Crop forcing was applied 15 days after the fruit set (stage 26-cap-fall complete, according to E-L modified scale) corresponding to CF1 and 30 days after fruit set, which corresponded to CF2 treatment (Figure 1A,B).CF1 was performed on 6th June 2019 and 2nd June 2020, whereas CF2 was performed on 25th June 2019 and 16th June 2020.In both seasons of the study, the experiments were performed in the same vines.Winter pruning and removal of leaves and clusters were performed manually.Desuckering was performed more often in crop-forced vines.
Pre-dawn leaf water potential (Ψpd) was measured 2 h before sunrise with a Schölander pressure chamber (PMS Instruments Co., Model 600, Corvallis, OR, USA) (Schölander et al., 1965).The measurements were carried out in eight plants per treatment (using one well-exposed and fully expanded leaf per plant) every 15 d, from fruit set until harvest.
Phenological stages, including budbreak, fruit set, veraison and harvest, were recorded when 50 % of the plants exhibited each stage, according to the modified E-L system (Coombe, 1995).At harvest, the total leaf area was registered using the method developed by Lopes and Pinto (2005).At the dormancy stage, the number of shoots was recorded, vines were pruned, and pruning mass (kg/vine) was determined for each vine of each replication (32 vines per treatment) (Figure 1A,B).
The budburst rate was calculated by dividing the number of shoots of each vine by the number of buds, and the total fertility index was estimated from the ratio between the number of clusters and the total number of shoots per plant.

Yield and berry composition
To evaluate the plant yield at harvest, the number of clusters was counted, and cluster weight (g) was measured in each vine.Average cluster weight was calculated from the ratio between the total cluster weight and the number of grapevines under study.
Ten berries were collected from each plant, being selected from different cluster positions.The berries were crushed, and several biochemical quality parameters were determined: pH, total soluble solids (TSS; ˚Brix), total acidity (g L -1 ), malic acid (g L -1 ), total phenolics (Absorption Unit-A.U.) and yeast assimilable nitrogen (YAN; mg L -1 ) using OenoFoss TM (FOSS Analytical, Hilleroed, Denmark) according to manufacturer's protocol equipment and by official methods of the Organisation Internationale de la Vigne et du Vin (OIV, 2021).

Data mining
UPLC-MS analyses were achieved using selected ion monitoring (SIM) mode and the resulting SIM chromatograms were integrated using the subroutine QuanLynx 4.1 for data mining.A pool of all samples was prepared to obtain a quality control sample (QC), and the samples were randomly injected independently from treatment conditions.Three QC samples were injected at the beginning of the sample set, and one QC sample was injected every 8 samples to check for potential analytical drifts.QC samples were analysed by Principal Component Analysis to evaluate the reproducibility of the UPLC-MS method (Fiehn et al., 2008).

Statistical analysis
All data are presented as mean values ± standard deviation (SD) of four biological replicates in each assay.One-way ANOVA, using Prism 7.0 (GraphPad Software, Inc., La Jolla, CA, USA), was performed for data analysis.The Principal Component Analysis was made using SIMCA P+ version 12.0 (Umetrics AB, Umeå, Sweden) and Heatmap metabolomics was performed with the ComplexHeatmap package (v1.18.1) on Bioconductor v3.9 after values normalisation by using the R centre and scale functions.

Evolution of phenological stages and pre-dawn water potential in 'Touriga Nacional' after Crop Forcing
The season of 2019 in Douro Valley was characterised by low levels of precipitation throughout the year (Figure S1).The highest monthly average maximum temperature was observed in July (34.7 °C), and the lowest was in January (1.1 °C).In the 2020 season, precipitation was similar to the previous year.The highest monthly average maximum temperature was 41 °C measured in July and August, while in January, a below-zero minimum temperature (-2.2 °C) was registered (Figure S1A).In both seasons, potential evapotranspiration (ETp) showed an inverse correlation with precipitation and a positive correlation with temperature and radiation (Figure S1).
A representative vine subjected to a second pruning is shown in Figure 1A.New vegetative and reproductive structures developed mainly from the apical and sub-apical dormant buds on the trimmed shoots, while the more basal nodes stayed dormant (acrotony) (Figure 1B,C).Figure 1D shows a representative result of the number and size of bunches per vine at the mature stage.calculated in both vintages from April until October were as follows: 2201.7 °C (2019) and 2335.9 °C (2020); thus, 2020 was warmer than the previous season.
Differences in GDD between veraison and harvest were also significant in both seasons and followed the trend CF0 > CF1 > CF2: in CF0, 708.75 (2019) and 867.50 °C (2020); in CF1, Compared with control plants (CF0), CF1 and CF2 in cv.'Touriga Nacional' resulted in a delay in all phenological events in both seasons, as follows (in 2019 and 2020, respectively): second budburst in 48 and 66 days in CF1, and 70 and 74 days in CF2, and harvest in 27 and 41 days in CF1, and 41 and 51 days in CF2 (Figure 1E).GDD  Pre-dawn leaf water potential was measured in both seasons, from the beginning of June until harvest, and no significant differences were observed between control plants and the ones subjected to CF treatments (Figure 2).

Grapevine yield, vigour and berry quality attributes after crop forcing
Both CF1 and CF2 induced a substantial increase in the number of leaves (data not shown), but no differences were observed in the total leaf area at harvest compared with controls (Figure 3A) because the leaf size was substantially decreased in plants subjected to CF (data not shown).Differences in the primary leaf area were only observed in 2020 between CF1 (2.59 m 2 ) and CF0 (1.52 m 2 ) (Figure S2A).
Figure 3B shows that second pruning decreased the bud burst rate in both seasons by up to 94 %, but while no significant differences were observed between CF1 and CF2 in 2019, CF2 induced a higher decrease in bud burst rate than CF1 in 2020.Crop forcing also negatively impacted the fertility index in both seasons by up to 800 % in 2020, but the fertility index was higher in vines subjected to CF2 than in CF1 (Figure 3C).
Crop forcing treatments also induced an increase of up to 114 % in the number of shoots and by up to 393 % in the number of buds in both growing seasons (Figure 3).On the contrary, the pruning weight was significantly lower in CF vines in both seasons, although more evident in 2020 (Figure 3D).The yield (kg/m 2 ) of vines subjected to CF substantially decreased in both seasons by up to 92 % (Figure 4A).Values of yield in kg per linear meter are shown in Figure S2D.
The number of clusters decreased by up to 81 % after CF1 and by up to 55 % after CF2 (Figure 4B).Thus, in both seasons, CF2 treatment resulted in a higher number of clusters than CF1.Average cluster weight per plant also registered a significant reduction in CF (Figure 4C), with CF2 246 % and 379 % and CF1 55 % and 214 % lower compared to CF0 in both years of the experiment.Figure 4C shows a negative impact of CF2 treatment stronger than CF1, especially in 2019.The cluster size, measured at harvest, was significantly lower after CF treatments in both seasons (up to 50 %) (Figure S2E), and berry weight also decreased by up to 38 % (Figure 4D).CF treatments induced significant modifications in berry quality attributes at harvest in both seasons (Figures 5 and 6): a decrease in the pH of up 0.7 units (18 %) (Figure 5A; differences among CF treatments being observed only in 2019), associated with an increase in the titratable acidity, more evident after CF2, with a ca.140 % increase over the control in both seasons (Figure 5B).Tartaric acid was variable among seasons, showing higher values in CF in 2019, whereas in 2020, CF1 registered the least content (Figure 5C).The content in malic acid (Figure 5D) increased substantially in response to CF in both seasons: by up to 176 % after CF1 and by up to 421 % after CF2, while tartaric acid slightly increased only in 2019 by ca. 9 % in response to CF1 or CF2.Total soluble solids (TSS) showed no differences in berries Supervised classification using OPLS-DA with crop forcing (CF1 and CF2) as dependent variables on metabolomic data from grape berries cv.'Touriga Nacional' at harvest in the vintages of 2019 (A and B) and 2020 (C and D).Variables in the score plots were coloured according to the treatment (CF0-green, CF1-blue and CF2-red).
from vines subjected to CF1 in both seasons, but a substantial decrease by 14 % and 45 % was observed after CF2 in 2019 and 2020, respectively (Figure 6A), where oenological maturation was not accomplished, since the TSS did not reach the optimal conditions (22-24 °Brix).In 2020, total phenols increased by up to 48 % after CF1 and CF2 treatments, but in 2019 no differences were observed (Figure 6B).Noteworthy, this increase did not translate in the colour intensity of the must that was similar in control and CF-treated plants in both seasons (Figure 6C).CF2 treatment induced a significant reduction by 51 % in yeast assimilable nitrogen (YAN) in 2019 but this effect was not observed in 2020, while CF1 did not change YAN in both seasons (Figure 6D).

Grape berry metabolome under crop-forcing treatments
Targeted metabolomics identified a total of 45 metabolites (Table S1), including 6 amino acids, 5 phenolic acids, 5 stilbenoids, 7 flavonols, 11 flavan-3-ols and 11 anthocyanins.Unsupervised PCA analysis showed that the vintage has a stronger effect on variability than the cropforcing treatments.The effect of crop forcing treatments on grape berries metabolites was particularly evident when a supervised OPLS-DA analysis was performed, showing score plots with a clear separation of treatments versus control (Figure 7 and Figure S3).This separation was driven by the metabolites projected in the PC1 positive quadrant of the vintage of 2019 (flavan-3-ols and anthocyanins di-OH on CF0 vs CF1; anthocyanins di-OH, anthocyanins tri-OH and amino acids on CF0 vs CF2) and vintage of 2020 (amino acids, flavonols, phenolic acids and flavan-3-ols on CF0 vs CF1; anthocyanins tri-OH, anthocyanins di-OH and flavan-3-ols, flavonols and stilbenoid DP1 on CF0 vs CF2).
A heatmap representation of each metabolite change after CF treatments compared with controls is shown in Figure 8. Results showed that forcing buds to regrowth clearly and consistently changed the biosynthesis of grape berries' secondary metabolites in two consecutive vintages.Most of the identified phenolic acids showed an increase in both treatments and vintages.For instance, caftaric acid increased by 24 % and 63 % after CF1 and CF2 in 2019 and by 67 % and 115 % after CF1 and CF2 in 2020.Flavonols detected showed an even greater increase in grapes from treated vines.This is the case of quercetin-3-O-glucoside, that increased by 99 % and 185 % after CF1 and CF2 in 2019 and 95 % and 278 % after CF1 and CF2 in 2020 (Figure 8).The eleven flavan-3-ols identified showed a similar pattern, with substantial increases being observed in 2020, particularly after CF2 (e.g.catechin gallate increased by 661 %, while procyanidins' increased only in the 2020 vintage after CF1 and CF2, but decreased in 2019 (Figure 8).The three anthocyanin di-OH identified showed also a strong increase in grapes from vines subjected to CF.This is the case of cyanidin-3-O-glucoside, that increased by 155 % and 580 % after CF1 and CF2 in 2019 and 506 % and 1073 % after CF1 and CF2 in 2020.Interestingly, non-acylated forms of anthocyanins tri-OH revealed a strong increase in response to CF (e.g., delphinidin-3-O-glucoside increased by 215 % and 503 % after CF1 and CF2 in 2019 and 315 % and 656 % after CF1 and CF2 in 2020), while most of the acylated (except the acylated petunidin) forms showed a consistent decrease, particularly in CF1 and CF2 of 2020 season.CF treatments affected amino acid levels mostly in the 2020 vintage in response to CF2, but L-phenylalanine revealed a significant increase after CF2 in both vintages.CF treatments affected stilbenoids, particularly in the 2020 vintage, when resveratrol increased by 214 % and 700 % after CF1 and CF2, respectively (Figure 8).

Crop forcing delays the vegetative cycle and alters production in cv. 'Touriga Nacional'
Previous reports have carefully addressed the potential advantages but also the limits and limitations of crop-forcing (Pommer, 2006;Martínez-Moreno et al., 2019;Lavado et al., 2019;Poni et al., 2021).Our results agree with previous reports in ' Cabernet-Sauvignon', 'Chardonnay', 'Tempranillo' and 'Maturana Tinta' cultivars, showing that the crop forcing after fruit set shifted fruit ripening and consequently harvest stage to the October to early-November period, which has been reported as the main hallmark of crop forcing (Gu et al., 2012;Lavado et al., 2019;Martínez de Toda et al., 2019;Martínez-Moreno et al., 2019;Pou et al., 2019;Tian and Gu, 2019;Prats-Llinàs et al., 2020).The majority of these studies also report a strong decrease in the vine yield, much like it was observed in the present study in 'Touriga Nacional', where yield reduction after CF1 and CF2 was accompanied by a decrease in the number of clusters, cluster size (cluster length) and berry weight.The number of clusters produced by vines subjected to CF1 were lower than after CF2, most likely because the buds that produce the inflorescences in the upcoming season (Lavee and May, 1997) were forced to break when they were still developing and were then less fertile (Lavado et al., 2019).Flower disorder, when initiated flowers do not develop, turning into tendrils due source to sink limitation (Bessis, 1967), may also have accounted for the observed reduction.
CF2 was applied later in the season when the buds of the new shoots already differentiated well, as previously observed (Smith and Holzapfel, 2009;Sadras and Moran, 2013;Prats-Llinàs et al., 2020).Apical dominance, the control exerted by the shoot apex or summer laterals (and/or leaves) over the outgrowth of the lateral latent buds, is thought to play a primary role in the paradormancy of these buds, which normally do not grow out until the following season after over-wintering (Cline and Deppong, 1999;He, 2008;Lavee and May, 1997).However, the observed strong reduction in plant yield after crop forcing can also be related to a reduced source-sink, due to a lower number of fully expanded photosynthetically active leaves in CF conditions (Gu, 2012;Martínez de Toda, 2019) and with different meteorological conditions observed when maturation is postponed till October or November (Martínez-Moreno et al., 2019).
The development of secondary (smaller) leaves after CF may have accounted for the observed similar values of total leaf area between treatments in both seasons.As the total leaf area in plants subjected to crop forcing and control plants did not change, the overall available surface area for the loss of water through transpiration was similar (Hochberg et al., 2017), thus, explaining why leaf pre-dawn water potential was similar between treatments at the end of the season.
The observed decrease in pruning weight in plants subjected to CF, more evident in 2020, is apparently in contradiction with the higher number of shoots in response to the crop forcing.Most likely, the amount of stored starch in grapevine canes is also compromised after crop forcing, which is a good topic for future studies, as these reserves are of pivotal importance for winter acclimation (Noronha et al., 2021a) and bud burst in the upcoming season (Noronha et al., 2021b).Thus, the need for longer-term studies and detailed analysis of the implications of these extreme canopy management strategies on carbohydrate reserves, bud fertility and future yield are still needed, as recently discussed (Santesteban, 2022).

Crop forcing changes berry qualitative traits in 'Touriga Nacional'
As aforementioned, an important consequence of global warming in grapevine cultivation is the uncoupling of technical and phenolic maturation.In 2019 and 2020, the desired sugar concentration, between 23 and 25 °Brix, was attained in September in the control plants and in October in grapevines subjected to CF1; therefore, CF1 effectively delayed the technical maturation by one month.However, grapes from plants subjected to CF2 were harvested before optimal sugar ripeness was attained, in early November (2019) and late October (2020), to prevent the loss of the harvest, due to the increase in rot incidence, by the autumn rainfall.Indeed, the risks of delaying the ripening to Autumn, especially in CF2, may compromise the optimum ripening (lower temperatures, precipitation) and the production of high-quality wines, as observed before (Prats-Llinàs et al., 2020).
The delay in the harvest in 'Touriga Nacional' promoted by CF was also reflected in the increased acidity of the grape juice, which correlated with increased levels of malic acid upon CF1 and CF2 treatments, which is in agreement with previous observations in other cultivars (Gu et al., 2012;Palliotti et al., 2014;Martínez-Moreno et al., 2019).Most likely, the cooler conditions occurring during the ripening of plants subjected to CF slowed down the degradation of malic acid, which is normally stimulated at high temperatures (Buttrose et al., 1971;Ford, 2012).
As suggested before (Dry, 1987;Lavado et al., 2019), the reduction in bud fertility after crop forcing may explain the production of smaller berries.Other studies have shown that these smaller berries have higher contents in secondary compounds.This is the case of 'Cabernet-Sauvignon' and 'Tempranillo' cultivars, where higher contents of total anthocyanins, tannins and total phenolics were measured at harvest (Gu et al., 2012, Martínez-Moreno et al., 2019;Lavado et al., 2019;Tian and Gu, 2019).However, in the present study in 'Touriga Nacional', the effect of crop forcing in the maturation delay was not so obvious in the stimulation of the synthesis of secondary compounds, although important qualitative changes were observed.Indeed, total phenolics at harvest only increased after crop forcing in the 2020 season, and this increase was not accompanied by an increment in the colour intensity of the musts.It is known that temperature accumulation plays an important role in the synthesis of phenolics.Maximum phenolics accumulation occurs when the daily temperature is between 15 and 25 °C and the night temperature is between 10 and 20 °C (Kliewer and Torres, 1972;Azuma et al., 2012).In the present study, the observed lower cumulative temperature (GDD) from veraison to harvest after CF1 and CF2, when compared to CF0, could counteract the expected increase in total phenolics in response to CF.During October and November, monthly minimum temperatures did not reach 10 °C, which could impair phenolic synthesis.Besides TSS and total acidity, phenolic content is important to schedule harvest date because it determines astringency, ageing, colour and body of the wine (Hidalgo, 1999).

Crop forcing modifies the metabolomic profile of 'Touriga Nacional' berries
Metabolomic analysis showed that despite total phenols only increasing after CF2 in 2020 and no differences were observed between treatments and control regarding pigments, important qualitative changes in key metabolites were observed in plants subjected to CF.Indeed, results from the targeted metabolomics analysis clearly discriminated between CF treatments, with specific variations in each polyphenol class highly consistent among vintages.
Amino acids are particularly prone to variations triggered by exogenous stimuli (Martins et al., 2014) and, in the present study, also suffered important changes in response to CF in 'Touriga Nacional'.The amino acids content in berries may account for up to 90 % of the yeast assimilable nitrogen (YAN) in grape juice, which is responsible for the success of fermentation, fermentation rate and aromatic profile of the wine (Carrau et al., 2008).Thus, the consistent decrease observed after CF1 treatments may negatively impact wine quality.
The management of the canopy substantially affected stilbenoid production (stilbenoids DP1 and DP2), particularly CF2 in 2020, which goes in agreement with previous observations that stilbenes production can be induced by abiotic stress or mechanical injuries (Bavaresco et al., 1997;Bavaresco and Fregoni, 2001;Pezet et al., 2004).
The increase in catechin, epicatechin and catechin gallate was generally observed after CF1 and CF2 in both seasons and may have important repercussions in wine quality as they are the most abundant flavan-3-ols in the berry and well known for their high antioxidant capacity, besides conferring bitterness and astringency to wines (reviewed by Teixeira et al., 2013).Conversely, there was a consistent decrease in procyanidins after CF1 and CF2, particularly in 2019, which are built from flavan-3-ols (+)-catechin and (−)-epicatechin (Bittner et al., 2013).The low temperatures that occurred later in the season may explain, at least in part, this observation because it was shown low diurnal temperatures reduce the degree of polymerisation of proanthocyanidin (Cohen et al., 2008).
It was observed that flavonoid biosynthesis-related genes in grape skins are differentially regulated by temperature and light conditions (Azuma et al., 2012), which could explain our results showing that canopy management through CF1 and CF2 positively affected the content in flavonols in grapes from 'Touriga Nacional' whose maturation occurred later in the season, at lower temperatures.CF clusters also had a higher incidence of radiation since they were more exposed to direct sunlight.Although, flavonols are known to contribute to wine colour as co-pigments (Asen et al., 1972, Boulton, 2001) and act as internal regulators, antioxidants and UV screeners, among others (Kaffarnik et al., 2005;Agati et al., 2013;Martinez et al., 2016).
Anthocyanins are the main compounds involved in the colour of red grapes; in Vitis vinifera L., the monoglucosides of delphinidin, cyanidin, petunidin, peonidin and malvidin but also their acetyl-, p-coumaroyl-, and caffeoyl derivatives are present in quantities varying among varieties, climate, and viticulture practices (Crupi et al., 2012, Coletta et al., 2013).Our results showed that although no differences were observed between must colour from grapes harvested from treated vines and control, di-OH and tri-OH anthocyanins substantially increased after CF, while acetylated and p-coumaroylated anthocyanins decreased.Trihydroxylated anthocyanins (delphinidin, petunidin, and malvidin-3glucosides) were described to be more stable in wines than dihydroxylated ones (cyanidin and peonidin-3-glucosides) (Hernández-Jiménez et al., 2013).Grape anthocyanin profile changes during ripening, with an increase in the proportion of malvidin 3-glucoside and its p-coumaroyl derivatives at the late ripening stages (25 and 28 °Brix) (Guidoni and Hunter, 2012).The lower maturation degree (lower ºBrix) reached by grapes under CF2 treatments may have accounted for the observed lower levels of acetylated and p-coumaroylated anthocyanins measured in juices.

CONCLUSION
Increasing temperatures during the growing season tend to damage wine typicity since early veraison dates lead to earlier maturation of the grapes.These grapes tend to have higher sugar content, less acidity and phenolics.Our results showed that crop forcing (CF) shifted fruit ripening in 'Touriga Nacional' cultivated in the Douro region for up to 1.5 months, from September up to October/November.This delay in sugar ripeness was accompanied by a substantial increase in total phenols in 2020 and malic acid in both seasons.Thus, CF may have positive impacts on sugar/acid balance of the wine and be useful to synchronise technical and phenolic maturity, depending on the vintage.However, new refinements of the technique are still necessary because yield potential was severely compromised through a strong reduction in bunch number and bunch weight.A metabolomic analysis showed that CF increases the relative abundance of key phenolics like flavan-3-ols, well known for their high antioxidant capacity besides conferring bitterness and astringency to wines, or trihydroxylated anthocyanins, more stable than dihydroxylated ones in wines, but their potential benefits in wine quality deserve further research and attention of the winemakers.

FIGURE 1 .
FIGURE 1.Time of crop forcing in cv.'Touriga Nacional'.Photographs of grapevines subjected do crop forcing (A), grapevine shoot with a second bud burst (B), inflorescence elongating (C) and harvest (D).(E) recorded dates of different phenological stages indicated by coloured arrow heads.The scissors symbol indicates the dates of crop forcing (CF1 and CF2).Asterisk (*) indicate harvest before technical mature stage.

FIGURE 3 .
FIGURE 3. Effect of crop forcing on grapevine vegetative growth.(A) leaf area; (B) budburst rate; (C) total fertility index and (D) pruning weight, in 2019 and 2020 vintages of grape cv.'Touriga Nacional' in different crop forcing conditions.Pruning conditions: CF0 = plants with no crop forcing; CF1 = plants submitted to CF 15 days after fruit set; CF2 = plants submitted to CF 30 days after fruit set.Results represent mean ± SD of four replicates.Asterisks indicate ANOVA statistical differences: ****p ≤ 0.0001; n.s.= non-significant.Lowercase letters indicate differences between treatments.

FIGURE 4 .
FIGURE 4. Effect of crop forcing on grapevine productivity.(A) yield; (B) number of clusters per plant; (C) average cluster weight per plant and (D) berry weight, in 2019 and 2020 vintages of grape cv.'Touriga Nacional'.Crop forcing conditions: CF0 = plants with no crop forcing; CF1 = plants submitted to CF 15 days after fruit set; CF2 = plants submitted to CF 30 days after fruit set.Results represent mean ± SD of four replicates.Asterisks indicate ANOVA statistical differences: ****p ≤ 0.0001; Lowercase letters indicate differences between treatments.
The only exception was observed in 2020 when pre-dawn leaf water potential in response to CF decreased by up to 15 % in August and increased by up to 13 % in September.Still, vines were under higher stressful conditions in 2020 than in 2019, judging from the values of GDD.

FIGURE 8 .
FIGURE 8. Heatmap representation of metabolome of cv.'Touriga Nacional' berries.Modifications observed in amino acids, anthocyanins, flavan-3-ols, flavonols, phenolic acids and stilbenoids in berries from vines cv.'Touriga Nacional', subjected to crop forcing.Each row represents a different metabolite and each column represents the crop forcing treatments in two consecutive years.Values were expressed as the logarithm of the ratio between treatments and the control (log10(CF1/ CF0; log10(CF2/CF0)) and presented in virtual colours as indicated by the colour key, in which the offset was determined by the average values found within the four biological replicates of each sample type.Metabolites were labelled according to their metabolic class.Asterisks indicate statistical significance between crop forcing treatment and the respective control within each sample year following Student's t-test: *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.