Introduction

The grapevine (Vitis vinifera L.) is one of the most important fruit crops globally, with Europe playing a central role in global viticulture and wine production (Alston & Sambucci, 2019; OIV, 2024). In Brazil, the State of Rio Grande do Sul stands out as the principal region for grape cultivation and wine production. The Serra Gaúcha is the most traditional and representative wine-producing region in Rio Grande do Sul, and ‘Merlot’ stands out as the main cultivar in the Denomination of Origin “Vale dos Vinhedos” (Ibravin, 2016; Mattos Rocha et al., 2022; Susin et al., 2022; Protas et al., 2026).

The expression of a cultivar’s potential is, however, intrinsically linked to vineyard management across all viticultural regions (Reynolds & Vanden Heuvel, 2009; van Leeuwen et al., 2024). In this context, the choice of a specific training system is essential for successful viticulture, particularly in the context of climate change. A training system influences vine productivity and alters the canopy microclimate, so it is a key factor in determining the suitability of grape varieties in a specific growing region (van Leeuwen et al., 2004; Yu et al., 2022; Han et al., 2022; Zohar et al., 2024). Such structural modifications have been shown to impact grape composition and directly influence the biosynthesis of phenolic compounds in grapes (Reynolds & Vanden Heuvel, 2009; González-Neves et al., 2008; Ferrer et al., 2008; Kyraleou et al., 2015; Domingues Neto et al., 2024).

These compounds are key to grape and wine quality, providing health benefits due to their antioxidant and anti-inflammatory properties, in addition to protecting plant tissues from damage caused by abiotic and biotic stresses, such as drought and temperature extremes, and acting as phytoalexins, which are synthesised in response to a pathogen attack (Teissedre et al., 2018; Viret et al., 2018; Hornedo-Ortega et al., 2021; Rao et al., 2025).

The development and adaptation of training systems have led to major advances in viticulture worldwide. Among these, vertical shoot positioning (VSP) is the most widely adopted training system worldwide; however, it typically shows lower productivity than divided canopy systems (Kyraleou et al., 2015; Marcon et al., 2020; Zohar et al., 2024; Campos et al., 2026). Studies demonstrate that divided canopy systems improve fruit composition and subsequent wine quality (Zoecklein et al., 2008; Vilanova et al., 2017).

The traditional lyre training system has been shown to enhance vine productivity while preserving oenological potential and improving vineyard management; however, this system features a fixed canopy opening angle (Carbonneau et al., 2004; Carbonneau et al., 2007; Ferrer et al., 2008; Palliotti, 2011; Filho et al., 2017). Based on these benefits, the foldable lyre concept is being developed in Brazil. This allows the canopy angle to be dynamically adjusted to suit specific vine management and regional conditions.

In terms of canopy architecture, it is hypothesised that choosing the foldable lyre training system, shifting from the fixed aperture of traditional systems to a dynamic adjustment, rather than the traditional vertical shoot positioning system, would effectively improve vine performance while maintaining fruit quality at the same plant density. In this context, this study aimed to evaluate the effects of the training system, among VSP and foldable lyre set at different angles (20°, 30°, and 40°), on vine productivity and biochemical profile of ‘Merlot’ grapes at Serra Gaúcha, Southern Brazil.

Materials and methods

1. Location of the experimental area and meteorological data

The experiment was conducted in an experimental vineyard located at Embrapa Grape and Wine, in Bento Gonçalves, Rio Grande do Sul State, southern Brazil (29°09’53” S, 51°31’43” W and an altitude of 630 m). The vineyard was established in 2018, using cv. Merlot clone 181, grafted onto ‘Paulsen 1103’ rootstock, spaced 3.2 m between rows and 1.5 m between plants, following a North-South row orientation, in a planting density of 2083 vines per hectare.

The region is mountainous, climatically characterised as Cfb (humid temperate), according to the Köppen classification, without a dry season and an annual average temperature of 17.6 °C (Tonietto & Carbonneau, 2004; Junges & Tonietto, 2022).

The experimental vineyard was established on a homogeneous site with a clay loam soil, composed of 36.3 % clay, 30.0 % sand, and 33.7 % silt (w/w). Soil water availability was uniform at 1.23 mm cm^–1, soil pH was 6.5, and organic matter was 5 %.

Meteorological data were collected from an automatic weather station, situated 542 m from the vineyard (Figure 1).

The season 2021/2022 exhibited a higher average temperature (20.2 °C) and lower accumulated rainfall (520.2 mm), characterising a dry cycle, with reduced water availability in the summer due to below-average precipitation in January and February, due to a weak La Niña. In contrast, the 2022/2023 season had a lower average temperature (18.3 °C) and higher rainfall (700.0 mm) but was unevenly distributed. The data correspond to the period from pruning to harvest.

2. Experimental design and treatments

The experiment followed a Randomised Complete Block Design with four replications. In the vineyard layout, each training system was distributed across two non-adjacent 45-metre rows. To eliminate potential interference between rows, such as shading or canopy overlap from the lyre systems, a guard row trained in a VSP system was maintained as a buffer between every treatment row.

Each experimental plot (the experimental unit) consisted of five vines, amounting to 20 vines per training system. Within each plot, the vines located at the boundaries of each replication were treated as border plants to minimise edge effects.

Treatments consisted of four training systems (Figure 2A): Vertical Shoot Positioning (VSP), foldable lyre at 20°, foldable lyre at 30°, and foldable lyre at 40°.

The experiment was carried out during two consecutive vintages, 2021/2022 and 2022/2023 vine growing seasons, in a non-irrigated vineyard, with vines pruned in double spur cordon in all training systems (Figure 2B).

The foldable lyre consists of two opposite and oblique canopies through a mechanism of modulation that can adjust the angles (Figure 2A), transforming these systems into a VSP (5º angle). In this study, the foldable lyre remained open throughout the seasons, maintaining the angles defined above. The spatial architecture of the foldable lyre enables a spurred cordon length of 1.5 metres, which is double that of the VSP system (Figure 2B). This is due to the use of two support wires for the spurred cordons, which also permit the establishment of two fruiting zones. At the same planting density, the Lyre system can support a greater bud load (48 buds per m²) than the VSP system (28 buds per m²).

3. Vineyard management

The pruning dates were September 9th, 2021, and August 19th, 2022. Short pruning with 3 buds per spur was adopted in both seasons, with 14 spurs per plant (42 buds) in the VSP system and 24 spurs per plant (72 buds) in the lyres system, evenly distributed along the bilateral cordons. Pruning was followed by the application of 4 % hydrogen cyanamide (Dormex®, BASF, Ludwigshafen am Rhein, Germany) on the same day. The same cultural management was adopted in all the training systems and cycle production.

4. Evaluations

4.1. Determination of photosynthetic photon flux density (PPFD) in the cluster zone

PPFD was measured close to the fruit zone (above, below, east, and west sides), positioning the sensors at a perpendicular angle to the length of the cluster during the ripening period, in January 2021 and 2022. Measurements were performed in two vines per training system, in triplicate, with a line quantum sensor (Apogee SQ-311-SS, Logan, USA; 10 sensors) at 8:00, 10:00, 13:00, 15:00, and 17:00 h. PPFD values were obtained by multiplying the signal (mV^–1) by the calibration factor (5 µmol m^–2 s^–1 mV^–1), and the daily mean PPFD was calculated from the instantaneous measurements. Representative vines were selected and marked to ensure measurement consistency and spatial reproducibility, allowing for precise and consistent positioning of the equipment throughout the season.

4.2. Characterisation of the agronomic performance

During the growing season, the budburst percentage was calculated as the ratio of budburst to the number of buds retained at pruning. Leaf area was determined by sampling 75 leaves per training system with a Leaf Area Meter (LI-COR 3100, Lincoln, USA). Canopy leaf area was estimated from the number of shoots per vine, leaves per shoot, and mean leaf area. The leaf area-to-yield ratio was obtained by dividing total leaf area by vine production weight. At harvest, the following variables were assessed: the number of shoots, clusters and production per vine (kg vine^–1).

4.3. Grape must composition at harvest

The grapes were harvested according to weather conditions and the evolution of ripening, monitored weekly by measuring soluble solids content, titratable acidity, and pH, as well as tasting skins and seeds.

The samples were collected from both sides of the plant and from different positions in each third, apical, medium, and basal of the cluster to ensure the fruit was sampled at all stages of ripeness and to capture variability within each treatment. Two samples of 50 berries were collected per plot, 10 berries per marked vine. The berries were crushed to extract the must and analysed in duplicate according to the methodology described by OIV (2021). The harvest dates were February 17th, 2022, and February 28th, 2023.

4.4. Determination of phenolic compounds

At harvest, for phenolic composition analysis by UPLC/MS, two samples of 30 berries were collected per plot, 6 berries per marked vine. Extraction of skin + pulp compounds was performed according to the methodology validated in previous studies (Pereira et al., 2005; Pereira et al., 2020; Pereira et al., 2021). Extractions were carried out using 96 % ethanol under cold conditions (5 °C) for 1 hour. The extracts were centrifuged for 15 minutes, and 1.5 mL of samples were dried using a concentrator at 30 °C for 3 hours. The samples were then resuspended using 5 mL of mobile phase, filtered through a 0.45 µm filter and stored in vials.

Analyses were carried out in triplicate per sample, according to the methodology described by Canedo-Reis et al. (2020) with adaptations described in Campos et al. (2024), using a Waters Acquity UPLC system (Milford, MA, USA), equipped with a single quadrupole mass detector (MS), Waters QDa (Milford, MA, USA), with an electrospray ionisation (ESI) source.

The following phenolics compounds were analysed individually: cyanidin 3-O-glucoside, delphinidin 3-O-glucoside, peonidin 3-O-glucoside, petunidin 3-O-glucoside, malvidin 3-O-glucoside, cyanidin 3.5-O-diglucoside, malvidin 3.5-O-diglucoside, catechin, epicatechin, epicatechin gallate, epigallocatechin, epigallocatechin gallate, procyanidin B1, and procyanidin B2, rutin (quercetin-3-O-rutinoside), kaempferol 3-O-glucoside, isoquercetin (quercetin-3-O- glucoside), myricetin 3-O-glucoside, taxifolin, and resveratrol. Quantification was carried out using the standard curve for each individual compound, and the concentration of individual and total phenolic compounds was determined and expressed in mg Kg^–1.

5. Statistical analysis

Data were analysed using a two-way ANOVA in a 4 × 2 factorial scheme (training systems × harvest years). As the same vines were evaluated over two consecutive seasons, Levene’s test was used to confirm the homogeneity of variances and validate the subsequent pooled analysis. The statistical model included the fixed effects of training systems (TS), harvest (H), and their interaction (TS × H). The block effect was included to ensure local control and isolate spatial variability from the residual error. This approach enabled a robust evaluation of the stability of the training systems against inter-annual climatic variability. All analyses were performed using Sisvar 6.0 (Ferreira, 2011), and means were compared using a Tukey’s HSD test (p < 0.05). The results are presented as means, and the data dispersion is expressed as standard deviation (SD), which is derived from the model’s residual variance.

Multivariate statistical analysis was performed using principal component analysis (PCA) with Minitab^® 19.01 Software (2019), as an exploratory synthesis to characterise the correlations between the variables.

Results and Discussion

1. PPFD in the cluster’s region

No significant interaction (p > 0.05) was detected between training systems and seasons for PPFD in the cluster zone of ‘Merlot’. Thus, comparisons of the main effects were based on the marginal means of each factor (Table 1).

Notably, the 30° and 40° lyre training systems provided higher shading in the bunch zone and lower light interception, while the 20° lyre and VSP systems allowed greater solar interception, providing greater cluster exposure. Compared to the VSP system, the Foldable Lyre system showed a 9.74 % reduction in solar interception at 20°, and 33.98 % and 32.87 % reductions in solar interception were observed at 30° and 40°, respectively. These variations are directly attributable to canopy architecture (Campos et al., 2026).

2. Agronomic performance

A significant interaction (p < 0.05) between training systems and seasons was detected for the following characteristics of ‘Merlot’ vines: shoots and clusters per vine, budburst percentage, leaf area, and yield (Table 2). This indicates the combined influence of both factors. The effects of bud load and canopy architecture were evaluated collectively within an integrated framework that considered the influence of training systems.

The percentage of budburst indicates that the VSP system promotes more uniform and efficient budburst, while the foldable lyre systems exhibited a reduced budburst percentage; this was offset by the higher bud load, leading to a net increase in the number of shoots per plant. Several studies with Vitis vinifera L. cultivars have demonstrated that increasing the bud load results in a lower bud burst percentage (Bonin et al., 2024; Farias Bonin et al., 2025). Nevertheless, while burst rates may decline, a higher bud load effectively increases the number of clusters per vine, thereby enhancing overall yield. The higher total number of productive buds compensates for lower burst percentages by increasing cluster count per vine. According to Al-Saif et al. (2023), this reduction can be explained by limited nutrient availability per bud, despite the general increase in budburst being linked to the release of stored nutrients in the vines.

In terms of leaf area, vines trained on lyre systems at 20°, 30°, and 40° achieved a level about 36.40 % higher than that observed in the VSP system.

The foldable lyre systems also significantly increased vine production. Firstly, the increased production can be attributed to the higher bud density of lyre systems. 20°, 30°, and 40° foldable lyre systems increased yield per vine by 51.81 %, 50.39 %, and 56.99 %, respectively, in 2022 and by 18.65 %, 49.09 %, and 65.97 %, respectively, in 2023 compared to VSP. These gains, particularly for the 30° and 40° systems, are consistent with reported increases of 36.4–66.3 % for ‘Merlot’ under the lyre system by González-Neves et al. (2008) and Ferrer et al. (2008).

The evaluation of leaf area-to-production ratio revealed no significant interaction, so the main effects were based on marginal means between seasons; the foldable lyre at 20° presented 1.27 m² kg^–1, reflecting a high leaf area to fruit load, reducing canopy/production efficiency. By contrast, the 30° and 40° lyres showed a more favourable balance. Bud load management optimises the vigour-to-production ratio by regulating vegetative growth and canopy efficiency (Farias Bonin et al., 2025). Balanced vine ratios are typically in the range of 0.80–1.20 m² kg^–1 for single canopies and 0.50–0.80 m² kg^–1 for divided canopies in warm climates. However, colder regions require slightly higher ratios due to thermal and light limitations on photosynthesis (Kliewer & Dokoozlian, 2005).

3. Grape must composition at harvest

No significant interaction (p > 0.05) was observed between seasons and training systems for SS, TA, and pH in both seasons. Thus, comparisons of the main chemical parameters of ‘Merlot’ grape must were based on the marginal means of each factor (Table 3).

The soluble solids content, TA, and pH in the grape must had an average of 21.95 °Brix, 4.40 g L^–1, and 3.50, respectively. Implementing the optimal training system can increase yield without compromising fruit quality (Reynolds & Vanden Heuvel, 2009; Filho et al., 2017). Despite the higher production per vine, bud load and leaf area of the foldable lyre systems, the values of SS, TA, and pH remained similar between the training systems. As suggested by Orlandini et al. (2015), the improved shoot distribution enhanced light penetration and canopy ventilation, ensuring proper cluster maturation and a higher yield.

Balance was also preserved between vegetative growth and fruit load, which had a positive effect on grape ripeness and quality at harvest. Furthermore, maintenance or degradation of TA and the dynamics of pH stability in grape berries are influenced by multiple interacting factors, such as concentration of inorganic cations, acid-ion interactions, cultivar and environmental conditions (Zohar et al., 2024; Ramos, 2025; Plantevin et al., 2024).

4. Phenolic compounds of grapes

Significant differences and interactions (p < 0.05) between seasons and training systems were observed for phenolic compounds in the skin + pulp of ‘Merlot’ grapes.

4.1. Anthocyanins

The malvidin 3-O-glucoside concentrations in grapes produced in VSP, 20°, 30°, and 40° treatments represented 54.14 %, 52.54 %, 34.75 %, and 56.31 % of the total anthocyanin content in 2022 and 25.33 %, 27.74 %, 25.18 %, and 41.56 % in 2023 harvests, respectively (Table 4).

Malvidin 3-O-glucoside is the main and prominent anthocyanin found in grapes and has anti-inflammatory properties, with cardiovascular health benefits (Decendit et al., 2013; Delić et al., 2024). Mori et al. (2007) observed that higher temperatures favour the accumulation of malvidin while tending to degrade other anthocyanins. This could explain higher levels of malvidin across all training systems during the warmer 2021/2022 season than in 2022/2023.

Compared to VSP, the 20° lyre system increased total anthocyanin content by 16.14 %, reinforcing the findings of Kyraleou et al. (2015), who found that lyre enhanced the anthocyanin composition of ‘Xinomavro’ grapes. The low concentrations of cyanidin and malvidin 3,5-O-diglucosides align with the prevalence of 3-O-glucosides in V. vinifera L., contrasting with the higher 3,5-O-diglucoside levels typical of V. labrusca L. and hybrids (Hornedo-Ortega et al., 2021; Silva et al., 2022).

The variable anthocyanin profiles observed in different studies (Bergqvist et al., 2001; Downey et al., 2006; Pereira et al., 2006; Tarara et al., 2008) highlight the complexity of the relationship between temperature and solar radiation in the regulation of flavonoid biosynthesis.

4.2. Flavonols, flavanonol, and stilbene

For resveratrol (Table 5), the predominant compound for flavonols, flavanonol, and stilbene in the 2022 harvest, the highest concentration was observed in grapes from the 40° lyre (2.88 mg kg^–1, 0.60 % of total phenolics), while the lowest concentration was observed in VSP (1.56 mg kg^–1, 0.29 %). In the 2023 harvest, the highest levels were found in grapes from the 20° lyre (2.18 mg kg^–1) and 30° lyre (2.29 mg kg^–1). Resveratrol has been linked to a reduced risk of cardiovascular disease, improved antioxidant and anti-inflammatory effects (Chedea et al., 2021; Sy et al., 2023).

Isoquercitin was the second most predominant flavonol, flavanonol, and stilbene compound, with the 30° lyre system showing the highest concentrations in both seasons (1.50 mg/kg in 2022 and 2.07 mg/kg in 2023). Da Costa et al. (2020) evaluated ‘Chenin Blanc’ under different training systems and found that vines trained to the lyre system had higher isoquercitin levels. They also observed interannual variations in phenolic composition driven by environmental conditions across production cycles.

Grapes from foldable lyre systems exhibited higher resveratrol accumulation, which is likely due to the specific microclimate within the cluster zone. This indicates that temperature has a greater impact on resveratrol synthesis than UV radiation, which is consistent with the findings of Oliveira et al. (2019). For total flavonol, flavanonol, and stilbene content, grapes grown on 40° lyres in the 2022 harvest had the highest total content, while grapes from 30° lyres had lower content. Compared to VSP, grapes produced on lyre systems showed increases of 25.89 %, 41.12 %, and 9.51 % in total flavonol, flavanonol, and stilbene content at 20°, 30°, and 40° lyres, respectively, based on the average contents from both seasons.

4.3. Flavanols and procyanidins

The majority compound in grapes from both seasons was Catechin (Table 6). In the 2022 harvest, grapes from the 20° and 40° lyre systems had the highest levels, at 1.86 mg kg^–1 and 1.91 mg kg^–1, respectively. In the 2023 harvest, the 20° lyre had the highest content, with 3.22 mg kg^–1.

Procyanidins B1 and B2 were present in small amounts in grape skin and pulp in both seasons. In 2022, the highest B1 levels were found in the 30° and 40° lyres, whereas in 2023, the 40° lyre reached 2.02 mg/kg, with higher values overall. The average B2 level was 0.23 mg kg^–1 in 2022, increasing to 1.14 mg kg^–1 in the 40° lyre in 2023.

The difference in the accumulation of flavanols may be related to the higher light intensity combined with reduced precipitation during the ripening period, which can accelerate the oxidation of these compounds and result in lower concentrations at harvest (Shi et al., 2023). Comparing the average contents for 2022 and 2023 harvests, relative to the VSP, the 20° and 40° lyres showed increases in total flavanol content of 1.39 % and 17.03 %, respectively.

4.4. Caftaric acid and total phenolic compounds

Comparing the caftaric acid concentration between the training systems, an increase of 26.39 % and 20.83 % was observed in grapes grown on 30° and 40° lyres, respectively, while grapes on 20° lyres showed a reduction of 76.39 % compared to grapes trained on VSP (Table 7).

As previously reported by Price et al. (1995) and Karaoğlan et al. (2015), the caftaric acid content was inversely proportional to cluster exposure to solar radiation. This study observed a similar tendency, with bunches exposed to the VSP and 20° lyre system showing lower levels of caftaric acid, likely due to their greater exposure to sunlight.

The highest total phenolic content was found in grapes from vines managed on 20° lyres, representing a 16.66 % increase compared to VSP. This variation can be attributed to the fact that anthocyanins represented more than 97 % of the total phenolic compounds analysed.

Katerji et al. (1994) reported that the increased radiation exposure associated with the lyre system induces mild physiological stress in grapevines, which leads to the accumulation of phenolic compounds. Similarly, Palliotti (2011) demonstrated that divided canopy architecture modulates the microclimate of clusters. They found that Y-shaped ‘Sangiovese’ vines had higher levels of total phenolics than VSP vines during the hottest periods of the day due to favourable canopy angulation.

This study shows a ‘vintage effect’ on phenolic composition, influenced by weather conditions (van Leeuwen et al., 2004; González-Neves et al., 2008; Liu et al., 2018; Da Costa et al., 2020).

5. Principal component analysis (PCA)

A principal component analysis (PCA) on major phenolic classes (total phenolics, total anthocyanins, total flavonols, flavanonol, and stilbene, total flavanols, and procyanidins) accounted for 87.5 % of the variance (PC1: 62.0 %, PC2: 25.5 %), demonstrating clear discrimination by harvest season and training system (Figure 3).

PC1 (explaining 62.0 % of the total variance) captured a dominant vintage effect, clearly separating the two growing seasons. All the 2022 training systems clustered on the negative axis of PC1 (–2.7 to –0.2), indicating lower overall phenolic levels. In contrast, the 2023 samples shifted positively (up to +4.1). The 2023 season showed a strong correlation with vectors for total anthocyanins, total phenolics, total flavanols, and total procyanidins, reflecting an overall increase in these metabolites across all treatments.

PC2, which explains 25.5 % of the variance, revealed micro-discrimination among the training systems. These differences became more pronounced in 2023. In 2022, the systems showed limited dispersion. FL30, FL40, and FL20 occupied the upper left quadrant, while VSP was confined to the lower left negative quadrant (PC2 at approximately –1.0). In 2023, the effects of canopy architecture intensified. FL30 (2023) moved into the top right quadrant (PC2 > +1.0) and exhibited a positive correlation with total flavonols, flavanonols, and stilbenes. Meanwhile, FL40 (2023) occupied the top right quadrant (PC2: –1.5 to –2.2). VSP (2023) and FL20 (2023) showed broader dispersion across the central right area.

The PCA highlight the significant impact of vintage on grape phenolic profiles, revealing distinct patterns of discrimination across seasons and training systems. This variability highlights the sensitivity of metabolic composition to changing climatic conditions and management practices (Pereira et al., 2005; Pereira et al., 2006; Oliveira et al., 2019; Da Costa et al., 2020; Minnaar et al., 2022; van Leeuwen et al., 2024; Faralli et al., 2024).

Conclusion

The Foldable Lyre system significantly increases bud load by up to 70 % and production by over 50 %, without compromising oenological quality. The system's primary advantage relates to its metabolic versatility: the 20° configuration maximises total anthocyanins, while the 30° and 40° configurations promote the accumulation of flavonols, stilbenes, and procyanidins.

In humid, temperate climates, the 30° and 40° configurations optimise canopy architecture, increasing leaf area by 36 % and enhancing light interception. These configurations offered the most balanced performance in terms of vegetative development, yield, and phenolic synthesis.

These findings provide the wine sector with a strategic tool to maximise profitability and land-use efficiency. Winegrowers can ensure higher production and economic sustainability as compared to the VSP system by adjusting canopy angles to specific oenological objectives.

Acknowledgements

The authors would like to thank the São Paulo State University (UNESP), School of Agricultural Science, Botucatu, for the opportunity to pursue the master’s degree in Agronomy/Horticulture, and the Brazilian Agricultural Research Corporation (EMBRAPA), National Centre for Grape and Wine Research (CNPUV), for allowing the implementation and execution of this research. This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) – Finance Code 001, CNPq (Productivity Grants 304258/2024-5 and 302930/2022-1; Project 405775/2021-0), and EMBRAPA (Project 20.22.00.050.00.00).