Graphical abstract

Introduction

Viticulture worldwide is threatened by altered patterns in environmental conditions, forcing grapevines to adapt to unprecedented temperatures, rainfall, and CO₂ concentrations (Fraga et al., 2012; Jones & Webb, 2010; van Leeuwen et al., 2024). Grapevine growth and development are highly modulated by environmental conditions, and as such are expected to see repercussions from a changing climate at multiple levels of grape production (van Leeuwen & Darriet, 2016; Mira de Orduña, 2010). Warmer and drier conditions shift grapevine phenology, advancing phenological stages and shortening phenological intervals (Faralli et al., 2024; Fraga et al., 2016; Jones & Davis, 2000; Webb et al., 2007). The ripening process is not exempt from environmental changes, which more often lead to earlier grape maturity (Rogiers et al., 2022; Webb et al., 2011; Webb et al., 2012).

Sugar accumulation in grape berries begins at veraison, at the end of the lag phase and concurrently to the second phase of berry double sigmoidal growth (Coombe, 1992). The physiological processes marking the onset of ripening have been characterised in grapes (Castellarin et al., 2016) and involve a tightly regulated cascade of events, starting with berry softening and abscisic acid (ABA) signalling. This is followed by rapid sugar accumulation and potentially colour development in a second stage. At the transcriptional level, the transition from green to ripening stage is regulated by waves of genes responsible for the activation of ripening-related processes (Fasoli et al., 2018). Early ripeness is seemingly driven by advancing the onset of ripening rather than accelerated ripening rates, the latter being less subject to environmental modulation (Cameron et al., 2021; Sadras & Petrie, 2011). Notably, auxin application delayed and cluster thinning advanced the onset of sugar accumulation, while sugar accumulation rates remained similar to untreated vines (Böttcher et al., 2011; Previtali et al., 2021). One implication of advanced ripening is vintage compression with overlapping harvests across grape varieties or regions, which can result in challenging winery logistics (Jarvis et al., 2019). Additionally, advanced ripening negatively impacts fruit quality due to the decoupling between sugars and metabolites related to acidity, colour, mouthfeel and aroma that are positively associated with wine quality (Previtali et al., 2021; Rienth et al., 2016; Sadras & Moran, 2012; Sadras et al., 2013; Sweetman et al., 2014).

Counteractive measures to early grape maturity have been the subject of thorough investigations in the past decades, with the goal of identifying vineyard practices to delay ripening (Gutiérrez-Gamboa et al., 2021; Novello & de Palma, 2013; Palliotti et al., 2014). Among these techniques, antitranspirant sprays, delayed pruning and late source limitation practices have shown a high degree of success across seasons and regions, alongside reduced cost and ease of implementation commercially (Previtali et al., 2022). Other useful viticultural practices to adapt vineyards to warmer and drier climates have been described, from canopy sprawling to irrigation, crop load management and shading nets for fruit protection (Gutiérrez-Gamboa et al., 2021). While vineyard practices serve as an effective tool for manipulating fruit ripening rates and maturity time in the short term, there are some technical limitations to their implementation. First, their effectiveness plateaus at about a 3-week delay. Second, applications must be planned carefully to avoid multi-seasonal negative effects on yield or vine storage reserves (Poni et al., 2022). A powerful and more long-term option to reduce the pressure exerted by climate change is to leverage natural genotypic variability for the selection of germplasm displaying slower or delayed ripening. Historically, grape varieties and clones have been classified according to their time of commercial maturity, which is typically classified as early-, mid- and late-season ripening. This terminology does not specifically refer to the ripening process (i.e. from veraison to a given sugar ripeness), but more generally to the time of harvest or heat summation required for fruit to ripen (Jones, 2010). As pointed out in previous studies, this approach contains bias as it often reflects developmental differences across the entire phenological cycle and not just related to the onset and rate of fruit ripening. Duchêne et al. (2012) found that genotypic variability in sugar ripeness was reduced when expressing sugars using thermal scales from veraison to a certain level of heat accumulation. The physiological shift from active sugar accumulation to passive sugar concentration should also be considered, as the sugar concentration at maximum berry growth and the rate of berry weight loss thereafter have been shown to be under a certain degree of genotypic regulation (Deloire, 2011; Shahood et al., 2020; Tilbrook & Tyerman, 2008). Natural variation is the primary driver for crop improvement through traditional breeding programs, and screening of newly developed or pre-existing germplasm has been useful to identify selections better suited to climate change. Frioni et al. (2021) and Frioni et al. (2023) screened sugars and organic acid levels at harvest in minor autochthonous and newly bred disease-resistant varieties to identify candidates with optimal sugar-to-acidity ratios, better suited for climate change. To date, little focus has been given to the study of genotypic variation in sugar accumulation traits, namely the timing of veraison and the rate of sugar accumulation following veraison. Although the ripening rate does not seem to be modulated by environmental factors, it has been reported to vary on a genotypic basis (Cameron et al., 2021; Stanfield et al., 2024). Significant progress in the -omics sciences applied to grapevine has allowed functional characterisation of genes and regulatory networks governing berry development transitions (Fasoli et al., 2018; Zenoni et al., 2010). However, genes directly controlling the onset and rate of sugar accumulation remain poorly understood (Duchêne et al., 2012). Only recently, natural variation in the so-called speed of ripening trait (i.e., rate of sugar accumulation) has been studied in a segregating population (Falginella et al., 2025). Ripening speed was introgressed in Vitis vinifera via hybridisation with Vitis riparia and a locus controlling ripening speed was identified. The presence of the RIPENING SPEED locus was associated with a decline in TSS accumulation per day right after veraison, providing a targeted effect on the rate of sugar accumulation without affecting the onset of ripening.

In this study, we characterised physiological differences in the ripening dynamics of two full siblings displaying extreme differences in ripening patterns. A three-year characterisation of sugar accumulation, berry growth and berry softening was performed under field conditions to describe differences in the onset and rate of ripening as well as trait stability over time. In addition, we measured physiological parameters and tested experimental treatments to investigate possible causes for the ripening failure. First, we measured photosynthesis rates to verify potential limitations to carbon assimilation. Second, crop load manipulation was employed to test the possible involvement of limiting source-sink relations in largely delayed fruit ripening. Finally, softening agents, including polyethylene glycol (PEG) and sucrose (SUC), were compared with plant growth regulators [abscisic acid (ABA) and 1-aminocyclopropane-1-carboxylic acid (ACC)] to test whether impeded ripening could be attributed to failing solute accumulation or a lack of hormone signals at the onset of ripening.

Materials and methods

1. Plant material and growing conditions

Plant material evaluated in this study was obtained by cross-breeding in the Breeding and Genetics program at E. & J. Gallo Winery in California. A cross-breeding population was created in 2013 between two white-fruited varieties, from which two proprietary seedlings were selected based on phenotypic observations indicating extreme differences in ripening patterns. One seedling was able to fully ripen (> 20 °Brix) before September 1, called normal ripening (NR), which is considered standard for white-fruited varieties in the area. The other seedling, renamed slow ripening (SR), displayed characteristics commonly associated with pre-veraison fruit on September 1, namely green hard berries. Plots of ten vines for each genotype obtained by vegetative propagation were used for a series of experiments described below. In addition, ten vines of cv. Muscat of Alexandria grown at the same site were added as a reference variety commercially grown in the area. NR and SR own-rooted vines were planted in 2016 near Madera (CA) in homogeneous soil conditions. The soil is a Grangeville series (U.S. Geological Survey, 2024), a fine sandy loam with a high presence of sand (40–60 %) and moderate drainage capacity. Vine distance was 2.1 m and row distance 3 m, and vines were grown as a high-wire bilateral cordon at the height of 1.5 m with a sprawling canopy. All management practices were standardised across the two selections and Muscat of Alexandria, with pruning at 2-bud spurs, no canopy manipulation and full irrigation near 100 % crop evapotranspiration (ETc) monitored using Watermark 200SS soil moisture sensors (Irrometer Company Inc., Riverside, CA). Weather data across the four years of study were sourced from a local weather station (~1 km from the vine plots). Hourly air temperature, daily rainfall and reference evapotranspiration were recorded. Growing degree days (GDDs) were calculated from April to October using a base temperature of 10 °C and no upper threshold.

2. Characterisation of ripening dynamics

Replicated (n = 3) berry samples were collected from one vine per genotype over 4 years (2022–2025) to characterise sugar accumulation and berry growth kinetics of NR, SR and Muscat of Alexandria. Preliminary data showed that sampling from a single vine captured the same or more variability than across vines (Figure S1A–B). Each sample consisted of nine berries, collected from three clusters and from the top, middle and bottom positions in each cluster. Samples were collected on a weekly basis, and data collection was intensified when possible, close to veraison. Berry samples were weighed to calculate average berry weight and crushed to measure Total Soluble Solids (TSS) using an ATAGO PAL-1 handheld refractometer (Atago Co. Ltd., Bellevue, WA). Sugar content per berry (mg/berry) was approximated by multiplying average berry weight × TSS × 10.

Berry diameter and firmness were measured in situ using the GrapeGrabber instrument following the non-destructive procedure detailed in Castellarin et al. (2016). Briefly, the elastic modulus of berries was calculated by relating the force applied and the berry displacement measured during berry compression. Firmness calculation, obtained by fitting the Hertz equation in SAS PROC NLIN (version 8.2; SAS Institute), assumes that berries react to compression like a perfectly elastic sphere (Wada et al., 2009). Berry firmness and diameter were measured on six berries per genotype at the same cadence of berry sampling from 2022 to 2024.

3. Experiment 1: Evaluation of differences in leaf physiology and water status

In 2022, vine measurements of leaf water status and gas exchange were taken on Jul 7 (day of the year, DOY 187), Jul 14 (DOY 194), Jul 21 (DOY 201) and Aug 13 (DOY 224). Measurements were taken around solar noon, between 12:30 and 1:30 pm. Leaf water potential (ψ_L) was measured using a 610 PMS pressure chamber (PMS Instruments Co., Albany, OR) on four mature, fully exposed leaves per genotype. On the first three dates, stomatal conductance (gs) was measured on the same leaves using a LI-600 portable porometer (LI-COR Biosciences Inc., Lincoln, NE). On the fourth date, gas exchange parameters (A, net assimilation; gs, stomatal conductance; E, leaf transpiration) were evaluated using a LI-6800 portable photosynthesis system (LI-COR Biosciences Inc., Lincoln, NE), setting reference CO₂ and H₂O conditions to reflect ambient values.

4. Experiment 2: Response to treatments to initiate ripening

Experimental treatments were applied to the SR genotype from 2022 to 2025. The list of treatments applied by year is shown in Figure 1. In 2022, treatments were applied to SR fruit when NR fruit was already ripening to attempt to induce the ripening process. The treatments were named as follows: abscisic acid (ABA), sucrose (SUC), polyethylene glycol (PEG), 1-aminocyclopropane-1-carboxylic acid (ACC) and surfactant (TWEEN). S-abscisic acid (ProTone®, Valent Biosciences, Libertyville, IL) was applied at 400 mg/L (ABA400) and 2000 mg/L (ABA2000) on Aug 4 and Aug 13, respectively. The ABA2000 treatment attempted to increase the effect size observed in preliminary work on ABA400. Sucrose (Sigma-Aldrich Inc., St. Louis, MO) and polyethylene glycol (Sigma-Aldrich Inc., St. Louis, MO) were applied at concentrations of 220 g/L and 380 g/L on Aug 4. These doses were calculated based on berry solute concentrations measured on Aug 2, transformed into solute potential and converted into concentrations needed to reduce osmotic potential by 0.5 MPa, to simulate the onset of sugar accumulation. 1-Aminocyclopropane-1-carboxylic acid (Accede®, Valent Biosciences, Libertyville, IL) was applied at 400 mg/L. Tween® 20 (Caisson Labs, Smithfield, UT) was added to all treatments at 0.1 % v/v as surfactant and applied as a standalone to account for the effect of spraying. For each treatment, twenty clusters were sprayed to runoff with a handheld sprayer early in the morning. Another twenty clusters were marked at the same time as untreated control.

Treatment effects on berry softening, growth and sugar accumulation were assessed following protocols reported in the previous section. For this experiment, all berries were rinsed with de-ionised water, then gently dried prior to crushing for TSS measurement to avoid interference of product residues on the berry surface.

5. Experiment 3: Response to crop load manipulation

In 2023, crop load was manipulated in one vine of the SR set. Cluster thinning (CT50) was performed on July 24 (DOY 204), leaving one cluster per shoot. Measurements were collected from twenty clusters randomly tagged following the thinning procedure. A second, more severe thinning was applied to the same vine, removing all clusters except for the twenty tagged clusters on Sep 9 (DOY 248). The number and weight of removed clusters were recorded at each pass.

6. Experiment 4: Effect of treatment timing on ripening initiation

In 2023, a factorial study on the effect of the timing and dose of ABA on the SR genotype was conducted. Two doses were tested based on the outcome of Experiment 2: 400 mg/L (ABA400) and 2000 mg/L (ABA2000). Treatment application was repeated at three stages of development of the NR genotype: pea size (PS, DOY 174), veraison (Ver, DOY 208) and post-veraison (Post-V, DOY 235). Protocols for cluster selection, spraying and measurements were the same as in Experiment 2.

7. Experiment 5: Response to water stress

In 2025, responses of SR clusters to exogenous ABA application were compared to those of severe water stress. Based on previous experiments, ABA was applied at 2000 mg/L (ABA2000) post-veraison (DOY 241) and compared to untreated clusters on a separate vine (CONTROL). On the same day of ABA application, dry down of three SR vines was initiated by stopping drip irrigation. Vines were equally irrigated prior to the beginning of water stress. The middle vine was used for subsequent measurements on water-stressed clusters (WS), leaving the neighbouring vines as buffers. The CONTROL and ABA2000 vines were drip irrigated to approximately 100 % ETc. Soil moisture in WS and CONTROL vines was monitored using Watermark 200SS soil moisture sensors (Irrometer Company Inc., Riverside, CA), logging data with 900M Watermark monitors every 30 minutes. Vine response was evaluated using a LI-600 portable porometer (LI-COR Biosciences Inc., Lincoln, NE), measuring three leaves per treatment at solar noon one week after treatment initiation. Stem water potential (ψ_S) was measured on an additional three leaves after conditioning in aluminium bags for a minimum of 30 minutes. Protocols for cluster selection, spraying and fruit sampling were the same as in Experiment 2.

8. Statistical analysis

Data analysis was performed using R (R Core Team, 2012) version 4.3.3 in RStudio (RStudio Inc., Boston, MA). Genotype differences in vine water status and gas exchange were analysed using one-way ANOVA. Significant factors for temporal changes in TSS, berry weight and elasticity were analysed using linear mixed models with the lmerTest package (Kuznetsova et al., 2017), selecting genotype and sampling date as fixed effects and replicate and vine as a random effect. Vine number was added as a random effect to account for the sampling procedure explained above (an example of the structural variability of random factors is provided in Figure S1C). In case of significant factors in ANOVA or linear mixed models, means were compared using Tukey’s adjusted post-hoc test with an error α = 0.05. Mean separation was obtained using the emmeans package (Lenth, 2020). TSS and softening curves were modelled using the package drc (Ritz et al., 2015).

TSS accumulation curves were fitted using the following equation, which describes a five-parameter logistic regression (Equation 1):

$y \left(TSS, or\frac{mg}{berry}\right)=y_{max}+\frac{y_{min}-y_{max}}{{\left(1+{\left(\frac{x}{x_0}\right)}^{Slope}\right)}^{\phi }}$      (Equation 1)

where y represents the response variable (TSS), x represents the explanatory variable (time in days or thermal time as GDDs), y_min and y_max represent the lower and upper asymptote, respectively, x₀ represents the time at the inflexion point, Slope represents the steepness of the transition between asymptotes and φ represents the asymmetry coefficient. Model fits were evaluated using performance indicators: Akaike Information Criterion (AIC), Residual Sum of Squares (RSS) and the coefficient of determination (R²). Model significance was tested using the log-likelihood ratio test (LRT), which tests the significance of the full model compared to a null model. TSS accumulation rates were calculated as maximum slopes of the sigmoid curves. Additionally, linear regressions between berry TSS and Julian days or thermal time were calculated, only considering samples from veraison to maximum berry weight that resembled a linear trend.

Firmness data were modelled using inverse four-parameter logistic curves, using the following equation (Equation 2):

$y \left(Firmness, MPa\right)=y_{max}+\frac{y_{min}-y_{max}}{1+{\left(\frac{x}{x_0}\right)}^{-Slope}}$      (Equation 2)

where the asymptotes and inflexion point are the same as reported for TSS curves, the slope is negative, and there is no asymmetry coefficient φ. Based on softening curves, the inflection point x₀ was assumed to be the DOY at which 50 % veraison was achieved (Ver50).

All visualisations were produced using the package ggplot2 (Wickham, 2016).

Results

1. Seasonal weather patterns

Differences in seasonal temperature and heat accumulation are shown in Figure 2. According to GDD summation curves (Figure 2A), 2022 and 2024 were hotter seasons (2897 and 2929 total GDD accumulated, respectively) compared to 2023 and 2025 (2646 and 2742 GDD in total). Temporal distribution of heat spikes was season-dependent (Figure 2B). In 2022, there was a sustained period of heat in the middle of the season, starting in July and leading up to September. In 2023 and 2024, temperature peaked in July, with a more pronounced spike in 2024, and started dropping thereafter.

2. Berry ripening dynamics: berry growth, softening and sugar accumulation

Dynamics of berry growth, TSS accumulation and berry softening for the genotypes over the three years of study are shown in Figure 3. The cadence of TSS measurements in 2022 was not enough to fully determine the initiation of ripening in NR (Figure 3A). In 2023 and 2024, data collection was started earlier, capturing ripening earlier and at a higher resolution across the three genotypes (Figure 3B–C). Based on these measurements, ripening was most advanced in the NR genotype, followed by Muscat of Alexandria. SR had a consistent delay in the onset of sugar accumulation until after the NR individual had reached its plateau at about 20 °Brix. Using 8–9 °Brix as a proxy for veraison, TSS stalled below this threshold for a long period in SR all three years. TSS curves of Muscat of Alexandria were intermediate between SR and NR in both 2022 and 2023 but became much closer to NR than SR in 2024. The results of models fitted to TSS data are shown in Table 1. Compared to three- and four-parameter models, five-parameter sigmoidal curves resulted in improved fits (Figure S2) and model performance metrics (Table S1) for TSS observations. Five-parameter models well described relationships between TSS data and DOY or GDD, with R² ≥ 0.965 across the board. Lower and upper asymptote estimates were the same for both models. There was a consistent difference in the upper asymptote by genotype, representing the maximum TSS levels reached at the end of the ripening curve, which never surpassed the threshold of 18 °Brix across the three years in SR fruit while being well above 20 °Brix in NR. For the model of TSS by chronological time (DOY), all parameters were influenced by the interaction between genotype and year (p = 0.019 or lower) except for the curve slope and asymmetry coefficient φ, for which the genotype was the only significant factor (p < 0.001). The inflection point x₀, representing the midpoint of the active sugar accumulation phase, was largely and consistently delayed in SR compared to NR (76 days on average) and Muscat of Alexandria (73 days on average). Fruit ripening rate, expressed as the TSS/day ratio at the inflexion point, was impacted by the genotype x year interaction (p = 0.007), and it was higher in NR compared to SR and Muscat of Alexandria (both p ≤ 0.001) and unchanged between the latter two genotypes (p = 0.476) in 2023 and 2025. In 2024, differences in TSS/day were not significant between the three genotypes (all p ≥ 0.134). In the model of TSS by GDD, there was only a significant effect of the genotype on curve parameters, namely x₀, slope and the GDD-based ripening rate (all p ≤ 0.002). The rate of TSS increase per 100 GDD accumulated was lowest in Muscat of Alexandria (1.21 °Brix/100 GDD) and significantly increased in both NR (2.53 °Brix/100 GDD, p = 0.002) and SR (2.24 °Brix/100 GDD, p = 0.015). The GDD-based ripening rate was unchanged between NR and SR fruit (p = 0.657). Statistical outputs of five-parameter sigmoid models applied to sugar content estimates per berry are shown in Table S2. Sugar content models displayed slightly lower R² values compared to TSS ones, but the overall performance of the fits was high (≥ 0.914 for the DOY model, except for Muscat of Alexandria in 2025; ≥ 0.921 for GDD models). Similar to the TSS models, parameters extracted for sugar content estimates modelled against DOY were affected by the interaction of genotype and year, while expressing sugar content by GDD resulted in significant effects only for the genotype factor (p < 0.001 for x₀, slope and GDD-based ripening rate). According to the GDD-based model, Muscat of Alexandria had the highest sugar loading rate (52.6 mg/100 GDD), compared to which NR and SR rates were significantly reduced to 29.8 (p = 0.021) and 27.2 mg/100 GDD (p = 0.013), respectively. NR and SR sugar loading rates were not significantly different (p = 0.940). Linear regressions in the range from Ver50 to maximum sugar content expressed by the DOY and GDD scales were explored for TSS (Figure S3) and sugar content estimates (Figure S4). Corrections to the GDD scale were also applied to test the effect of capping GDD calculations at 30 °C; however, this did not affect the results. Results of linear models were in line with the outputs of the sigmoid curves.

Sampling was started late in 2022, and full curves of berry growth were not obtained (Figure 3D). In 2023 and 2024, berry growth (Figure 3E–F) followed a double sigmoid curve, and the shape of this curve was impacted by both genotype and year. Berries of Muscat of Alexandria quickly became significantly larger than NR and SR irrespective of the year. There were evident differences between 2023 and 2024, with higher maximum berry weight and higher levels of late-season dehydration occurring in 2023. The effect of the year on TSS and berry growth dynamics was more marked in NR and Muscat of Alexandria compared to SR.

Changes in firmness values by genotype from 2022 to 2024 are depicted in Figure 3G–I. In 2022, measurements were started when berries of the NR genotype and Muscat of Alexandria were already soft (firmness < 1 MPa). In SR, there was an early drop in firmness in 2022 (Figure 3G), then typical pre-veraison firmness was recovered (> 4 MPa). The full transition from hard (> 4 MPa) to completely soft (< 1 MPa) started around DOY 240. Increased frequency of sampling in 2023 (Figure 3H) and 2024 (Figure 3I) allowed us to fully describe the timing and kinetics of softening for all genotypes. In both years, there were fluctuations in firmness earlier in the season preceding the final drop in firmness to fully soft berries. Inverse sigmoid fits are shown in Figure S5 and softening parameters extrapolated from the fits are reported in Table 2. Ver50 occurred on the same DOY 187 in NR in 2023 and 2024 and was delayed by 68 and 66 days in SR, respectively. For Muscat of Alexandria, Ver50 occurred 17 and 8 days after NR in 2023 and 2024, respectively. The curve slope, representing how quickly the berry population completed softening, was highest in Muscat of Alexandria in both years when full curves were obtained. In 2023, the slope was lower in NR compared to SR, while the opposite trend was observed in 2024.

3. Experiment 1: Exploration of genotype-specific vine physiology traits

Midday stomatal conductance (gs) and leaf water potential (Ψ_L) measured for the three genotypes at three dates are shown in Figure 4A–B. At all July dates (DOY 187, 194 and 201), there were no differences in gs values between NR and SR, while gs was consistently lower in Muscat of Alexandria. In turn, water status was higher in Muscat of Alexandria (Ψ_L: –1.02 MPa) compared to both NR (–1.40 MPa) and SR (–1.31 MPa) at the first sampling date. At subsequent dates, Ψ_L values declined in both Muscat of Alexandria and NR, while remaining relatively constant in SR. As a result, NR vines were significantly more stressed than SR and Muscat of Alexandria at DOY 194 and 201. According to gas exchange measurements taken on DOY 224 (Figure 4C), differences in A_net between SR and NR were barely insignificant (p = 0.055), while both gs and E rates were increased in SR (gs: p = 0.037; E: p = 0.039). Muscat of Alexandria vines displayed the lowest levels of A_net, gs and E across the genotypes investigated. In 2024 (DOY 227), the separation between SR and NR was more evident: the highest A_net, gs and E values were recorded in SR, with a significant decline in both NR and Muscat of Alexandria.

4. Experiment 2: Response to hormone treatments

Softening and TSS accumulation curves for treated SR clusters in comparison to untreated ones are shown in Figure 5. According to TSS accumulation curves, the only treatment that was able to advance the onset of sugar accumulation in SR was ABA2000 (Figure 5A). The first significant increase in TSS for ABA2000 compared to the control happened on DOY 236. The TSS value at this date (9.2 °Brix) indicated the rapid phase of sugar accumulation had started. The onset of ripening did not happen until around DOY 262 (8.3 °Brix) for untreated clusters. This represents a 26-day delay in the onset of sugar accumulation. After DOY 257, the rate of TSS accumulation in ABA2000 became much slower, with differences from the control disappearing from DOY 278. Comparing softening patterns in ABA2000 and untreated clusters (Figure 5B), there was a significant reduction in berry firmness in ABA2000 berries shortly after the treatment on DOY 236 (1.2 MPa), concomitant with the increase in TSS values. Firmness levels in the control remained in the range typically associated with hard pre-veraison berries (3.3 MPa). TSS and softening curves in ABA2000 were parallel to those drawn for untreated clusters, while the slopes were equal, indicating an effect largely on the onset and not on the rate of ripening. When a lower dose of ABA was applied (ABA400), the advancing effect was not as evident as in ABA2000. TSS values in ABA400 were significantly higher at two dates preceding the onset of ripening in untreated clusters, namely DOY 257 (9.4 vs 7.8 °Brix, p = 0.008) and 262 (10.1 vs 8.3 °Brix, p = 0.002). In the case of ACC, there was a transient effect on TSS values around veraison (DOY 262), which were significantly higher (10.2 °Brix) than the control (8.3 °Brix, p < 0.001). Softening curves of berries treated with ABA400 and ACC were largely unaffected by the treatment applied, except for a few early dates showing transient differences. PEG application did not affect the onset of ripening, but it led to consistently higher TSS values compared to the control during the ripening stage, starting from DOY 271. When SUC was applied, there were also negligible effects on TSS accumulation and on berry softening. Lastly, the application of Tween did not alter berry firmness patterns but led to small significant increases in berry TSS at the end of the sugar accumulation process, starting from DOY 299.

5. Experiment 3: Response to crop load manipulation

On the first cluster thinning pass (DOY 204), 52 clusters were removed, equalling to 13 kg of fruit. The second thinning exercise executed on DOY 248 exported an additional 45 clusters for an additional 12 kg of fruit removed from the vine. The final vine yield for the two vines was 30.5 kg (98 clusters) for the untreated control and 3.5 kg (10 clusters) in CT50. The effect of this two-step cluster thinning on the kinetics of sugar accumulation and berry firmness is represented in Figure 6. Crop removal did not advance the onset of sugar accumulation nor did it accelerate ripening (Figure 6A). Similarly, except for the very first sampling date, severe cluster thinning did not affect softening curves (Figure 6B). TSS values measured in CT50 berries at the last four sampling dates were significantly lower compared to the control (between 2 and 3 °Brix on average), displaying a lower TSS plateau compared to the control.

6. Experiment 4: Effect of ABA application timing on ripening kinetics

The response of the SR genotype to ABA treatments applied at three different times based on the development of the NR sibling, namely pea-size, veraison and post-veraison, is illustrated in Figure 7. Neither of the applied ABA doses (400 vs 2000 mg/L) was able to trigger ripening at the first two timings, namely when the NR genotype was at the pea size (Figure 7A–B) and veraison stages (Figure 7D–E). When ABA400 was applied at pea size (Figure 7A), there were transient significant differences compared to the control at DOY 271 (–1.9 °Brix, p = 0.022) and 285 (–1.8 °Brix, p = 0.027). TSS accumulation curves of ABA2000 and Tween applied at pea size (Figure 7B–C) were comparable and appeared to be delayed compared to the control, as indicated by significantly lower TSS values compared to the control starting from DOY 264. Sugar accumulation curves for the three treatments when applied at the veraison stage of NR completely overlapped with those of untreated clusters (Figure 7D–F). Post-veraison treatments (Figure 7G–I) had the largest effect on sugar accumulation, leading to a significant advancement of ripening. Starting from 256, TSS values were consistently higher in ABA400 compared to the control (2.5 °Brix on average, p ≤ 0.001), an approximate ripening advancement of 7 to 10 days. The effect of ABA was even more marked with the highest dose applied (ABA2000, Figure 7H), which caused an earlier onset (from DOY 249) and a larger (5.3 °Brix on average, p < 0.001) increase in TSS accumulation. Tween application post-veraison did not affect the onset or rate of sugar accumulation but appeared to increase the plateau of TSS accumulation, still to a lower extent compared to ABA2000.

Softening curves measured in ABA treatments and control clusters confirmed TSS results (Figure 7J–K). In ABA400, there was a trend for lower firmness following treatment application; however, no significant differences were found. In contrast, the application of ABA2000 caused a quick significant drop in firmness, indicated by large significant drops in firmness values compared to the control, allowing ABA2000-treated berries to reach full softness earlier.

7. Experiment 5: Effect of water stress on ripening kinetics

Soil water tension patterns for water-stressed and normally irrigated SR vines are shown in Figure 8A. Once watered at DOY 244, water tension in CONTROL vines promptly decreased to close to zero and then gradually increased over the following two weeks. In contrast, soil remained dry under WS vines where irrigation was not applied. When shrivelling symptoms started to appear, irrigation was supplied to WS vines on DOY 252, as observed in the decrease of soil tension values below the irrigation threshold set at 60 cbar. A second dry down was applied from DOY 254 to 265, resulting in a pronounced water tension increase in WS vines compared to the irrigated control. The effect of ceased irrigation on SR vine performance is shown in Figure 8B. At DOY 247, there was a significant decrease (p = 0.0001) in midday Ψ_S in WS vines (–1.43 MPa) compared to the CONTROL (–1.02 MPa). Water status differences between WS and CONTROL vines were maintained one week later (ΔΨ_S = 0.3 MPa, p = 0.001) (data not shown). Gs was not significantly different in CONTROL and ABA2000 vines at DOY 247 (p = 0.998); in contrast, WS vines had significantly lower gs values (p = 0.013 or lower). Responses of fruit TSS levels to ABA2000 and WS treatments are presented in Figure 8C. In both treatments, TSS levels were significantly increased at each of the 11 sampling times following treatment application. As for the comparison between ABA2000 and WS fruit, TSS means were unchanged only at the first sampling point after treatment initiation (DOY 248, ΔTSS = 0.87 °Brix, p = 0.459), and TSS were higher in ABA2000 than WS at all following dates (p = 0.017 or lower). Changes in BW and sugar per berry for the three treatments are compared in Figure S7. Sugar content estimates reflected trends observed for TSS values. BW sharply increased in ABA2000-treated clusters compared to CONTROL, but the same effect was not as evident nor always significant for WS. The results of five-parameter logistic regression curves on TSS and sugar content observations by treatment are shown in Table S11, and individual fits per replicate and treatment are shown in Figure S7. SR treatments significantly affected the inflection point x₀ in both TSS and sugar content curves (p ≤ 0.004). The TSS/day ratio was higher in ABA2000 (0.83 °Brix/day) compared to both CONTROL (0.22 °Brix/day) and WS (0.24 °Brix/day).

Discussion

1. Onset and rate of ripening in the slow ripening genotype

Berry growth, sugar accumulation and berry firmness data were collected at high temporal resolution over four years for the SR and NR genotypes plus a reference variety, Muscat of Alexandria. Temporal changes in these parameters are the cornerstone for the understanding of physiological processes associated with the beginning of ripening, as described in Castellarin et al. (2016). When comparing sugar accumulation curves, overall trends were consistent across the four years, with the NR genotype displaying the earliest ripening, SR being the most delayed and Muscat of Alexandria falling in the middle (Figure 3A–C). Softening curves built from berry firmness measurements (Table 2) were used to calculate the timing of 50 % veraison (Ver50, i.e., the inflexion points of the firmness curve). As shown in Figure S6, expressing TSS values as a function of the days post Ver50 resulted in aligned curves for 2023, 2024 and 2025. By comparing Ver50 across genotypes, differences in ripening initiation were anchored to a fixed physiological parameter, namely, half of the berries softened. This approach overcomes the effect of longer or shorter transitions from pre- to post-veraison stages, driven by the highly asynchronous development of berry populations across genotypes and possibly introducing bias in the interpretation of temporal changes (Shahood et al., 2020; Tavernier et al., 2025). Over the two years when this approximation could be made, Ver50 was delayed 68 and 66 days in SR compared to NR. Firmness data were not collected in 2025, but an approximation of the Ver50 value based on relationships between TSS and firmness in previous years resulted in 71 days of delay, which is consistent with the previous two seasons (Table 2). Extensive research has been dedicated to the identification of strategies to delay ripening, to counteract advanced grape maturity driven by warmer and drier conditions (Gutiérrez-Gamboa et al., 2021; Palliotti et al., 2014; Previtali et al., 2022). At the study location, classified as a hot and dry Mediterranean climate, delaying veraison by over two months signifies that ripening starts at the end rather than the start of summer. Delayed ripening therefore offers benefits in terms of reduced organic acid degradation, slower sugar accumulation and improved colour, phenolic and aroma compounds that are sensitive to high temperature (Mori et al., 2007; Previtali et al., 2021; Sadras & Moran, 2012; Sweetman et al., 2014). It is important to note that the delay observed in the SR genotype is much greater than the effect obtained using vineyard practices to delay ripening, where the maximum delay is about three weeks (Previtali et al., 2022). Additionally, the effect size of these management practices is variable, and yield declines are sometimes a trade-off for the ripening delay (Poni et al., 2022). Exploring genetic material could not only offer a longer-term solution, but also a potentially more stable solution since slow ripening is de-associated from other effects observed when the delay is achieved by in-season manipulation of grapevines.

One important aspect that needs clarification for a comprehensive understanding of the slow ripening selection evaluated in this study is the relative importance of ripening rate and ripening onset in determining the final delay at harvest. Data mining studies (Cameron et al., 2021) and manipulatory studies (Böttcher et al., 2011; Previtali et al., 2021) provide evidence to support the hypothesis that the rate of ripening is controlled at the genotypic level. Recent studies have explored the genetic determinants controlling the rate of ripening (Falginella et al., 2025). As for the genotypes investigated in the present study, the horizontal separation between TSS curves of SR and NR (Figure 3A–C) indicates that the predominant explanatory variable is the shift in the onset of ripening. To understand the contribution of ripening rate to the SR trait, TSS and sugar content observations were modelled to extract ripening rates as TSS/day. In previous studies, sigmoid functions have been successfully used to characterise sugar accumulation in grapes, with good results when using TSS or sugar concentrations in g/L (de Rességuier et al., 2024; Sadras et al., 2008). Three-parameter sigmoid functions have been typically used to describe sugar changes over time (i.e., DOY). These models set a fixed lower asymptote (i.e., pre-veraison TSS values) and return three parameters, namely the upper asymptote, inflexion point and slope of the curve. In the present study, a five-parameter sigmoid curve that allows for a flexible lower asymptote and includes an asymmetry coefficient (Table 1) provided more satisfactory results for two reasons. First, it was observed that the lower asymptote varied both year-to-year and according to the genotype. Second, the asymmetry coefficient provided greater flexibility to capture differences in the transitions between the lower or upper asymptote and the inflexion point. The asymmetry was especially evident in the case of the SR genotype, where the three- and four-parameter fits did not converge or provided the worst fits across the three years (Figure S3). This was due to slow but steady increases from 5 to about 8.5 °Brix leading up to veraison (TSS curves for SR genotype in Figure 2A–C). TSS/day values extracted from five-parameter models were significantly higher in NR compared to SR across two years (Table 1 and Figure S3), suggesting ripening may be delayed and slower at the same time in the SR genotype. Similarly, different rates of ripening could be observed when aligning TSS curves based on the time of veraison (Figure S6). The lack of differences in TSS/day across genotypes in 2024 may be explained by the exceptionally hot and dry July (Figure 2, 15 days with Tmax ≥ 40 °C in 2024 compared to 6 and 0 days in 2023 and 2025, respectively), which corresponds with the time of ripening of NR fruit and likely provided supra-optimal conditions for sugar accumulation (Greer & Weston, 2010). When expressed as TSS/day, ripening rates cannot fully describe genotype-specific variation in sugar accumulation, as the latter is mixed with environmental factors (Duchêne et al., 2012). This represented an important obstacle in our study, where it was found that sugar accumulation started and was completed by the end of July in NR, while SR started to ripen mid-September, under much cooler and shorter days. To account for these differences, ripening data were also modelled based on thermal time (GDD), and this approach was applied to TSS data (Table 1 and Figure S3) and sugar content estimates (Table S2 and Figure S4). GDD-based models allowed the removal of the interference of environmental conditions and focused on genotypic differences, removing the highly significant effects of the Year and Genotype x Year interaction factors observed in models by DOY (Table 1). Ripening rates were essentially unchanged in GDD-based models, signifying that the NR and SR genotypes display inherently equal sugar loading capabilities. Modelling estimates of sugar content per berry by DOY and GDD led to a similar result, with unchanged sugar loading rates (mg of sugar/berry) between NR and SR fruit (Table S2). Enhanced sugar loading rates were observed for Muscat of Alexandria, driven by large berries and possibly skewed by the methodology adopted to estimate sugar content. Our simplified estimation of sugar content per berry assumed constant sugar concentration across the entire berry weight, which does not account for seed weight and differences in sugar concentration by berry tissue (Possner & Kliewer, 1985; Ristic & Iland, 2005). Validation of sugar content results requires further work with measurements conducted on deseeded berries, given the strong genotypic determination of factors which could confound final sugar content estimates. Another result of interest that may be partly linked to environmental differences is the lower levels of final TSS measured in SR berries at the end of the season, well below 20 °Brix in all years (Table 1). A combination of less favourable temperatures for plant metabolism and senescence may play a role in the plateauing of sugar accumulation around 15 to 17 °Brix. Whether this is the maximum berry sugar accumulation in the SR genotype or ripening could continue in the presence of favourable conditions, it remains unsubstantiated, and it could be the object of future research in controlled conditions. The maximum sugar ripeness measured in SR fruit is likely to be suboptimal for winegrapes, except for low alcohol wines. The development of other fruit quality traits would also need to be screened to verify the oenological aptitude of this material.

2. First insights on slow ripening mechanisms

A series of targeted manipulatory experiments were conducted to gather initial insights regarding potential reasons behind the large lag and slower ripening described in detail above for SR. Three hypotheses were formulated as possible causes for this unique phenotypic variant. In screening different rescuing treatments, the metric of success was to capture a rapid initiation of ripening in treated SR fruit compared to the untreated control.

First, we tested the possibility of carbon constraint at the leaf level by analysing physiological processes in leaves of the NR and SR. Carbon limitation may occur because of damaged photosynthesis machinery or a structural blockage of vascular systems in charge of carbon translocation (i.e., phloem). Physiological measurements at the leaf level, taken during the long lag phase of SR over two years, showed no limitation in terms of carbon assimilation (Figure 4). Similarly, no colour or textural symptoms associated with sugar build-up in leaves were observed in the field until senescence, such as leaf reddening and curling (data not shown). These symptoms are often reported for virus-infected vines due to the pathogenic agent blocking the sugar transportation mechanism (Song et al., 2021; Sudarshana et al., 2015).

The second hypothesis was that the ripening delay could be attributed to source–sink regulation, ultimately through improper balance between fruit weight (sink) and leaf area (source). This ratio, more often called crop load, has been widely studied in grapes and was shown to be tightly related to sugar maturity at harvest (Kliewer & Dokoozlian, 2005). Crop load was also found to be a main driver for ripening rate (Cameron et al., 2021). Ripening is reported to be delayed in overcropping conditions, and cluster thinning can advance the onset of ripening (Petrie & Clingeleffer, 2006; Previtali et al., 2021). Severe cluster removal applied at key stages did not advance the onset of ripening in SR (Figure 6). This result is in contrast with outcomes of this practice in the literature (reviewed in Vanderweide et al., 2024) and provides experimental evidence that the slow ripening is not the result of crop load limitations.

Another plausible hypothesis was that there is a failure in the processes leading to the initiation of ripening. Lack of inactivation of one or more processes required for berries to start ripening may explain the extended lag phase observed for SR (Figure 3). Experimental treatments were applied to compensate for physiological events that were potentially latent in SR fruit (Castellarin et al., 2016). The same dose of a surfactant compound (Tween20^®) was applied in addition to each softening agent or plant growth regulator. When applied as a standalone, the Tween treatment was not significantly different from untreated fruit in terms of both the onset and rate of ripening, and there were only transient differences at later stages, likely due to vine and cluster variability (Figure 5 and 7C,F,I). The effect of adding a surfactant was absent in both the 2022 and 2023 experiments, excluding the possibility of artefacts when interpreting the effect of target compounds. Specific agents were used to induce softening (PEG), sugar accumulation (SUC) and signalling (ABA and ACC). Softening of grape berries is hypothesised to occur due to a shift from symplasmic to apoplasmic phloem unloading, causing a reduction in turgor and the start of rapid sugar accumulation (Zhang et al., 2006). Failure to soften, as characterised by berry firmness measurements (Figure 3G–I), may be the limiting factor preventing rapid sugar accumulation and interfering with the transition of phloem unloading routes. PEG and SUC applications were aimed at inducing berry softening and subsequent sugar accumulation by increasing solute concentrations in the apoplast. PEG was selected as a softening agent that would remain confined in the apoplast given its inability to cross cell membranes, avoiding possible confounding effects of sucrose uptake into the cells. Both PEG and SUC treatments were ineffective (Figure 3), aside from transient significant differences that are explained by vine-to-vine variation since treatments were applied on adjacent vines. PEG caused an apparent increase in TSS; however, this was only true at the end of the season, and there was no sign of advanced ripening. These discrepancies are to be attributed to two main factors. First, PEG treatment was highly concentrated, and the compound itself resulted in TSS increases (100 g/L PEG solution = 9.0 °Brix). Even though berries were rinsed prior to TSS analysis, a fraction of the agent was expected to penetrate berries from the time of treatment, interfering with grape-derived sugars. As for the late season TSS effect, high concentrations of PEG spray led to crystallisation on berries, possibly causing more rapid dehydration and increased sugar concentration in the berries. The inefficacy of PEG and SUC may be explained by insufficient solute concentration reaching the apoplast, but based on the results observed for ABA treatments, it is more likely that the failure in ripening observed in the SR selection is related to disrupted hormone signalling.

The goal of treating SR fruit with ABA and ACC was to evaluate whether hormonal mechanisms beyond solute gradient dynamics impeded ripening in the SR genotype. Although grapes are non-climacteric fruit, there is evidence for ethylene biosynthesis and crosstalk with other phytohormones at the onset of grape ripening (Böttcher et al., 2013; Chervin & Deluc, 2010; Chervin et al., 2004). Extensive work on the involvement of ethylene in fruit ripening-related processes has been conducted using ethylene-releasing compounds (2-choroethylphosponic acid [CEPA]) or precursors of ethylene inhibitors (1-methylcyclopropane [1-MCP]). ACC is an intermediate for ethylene biosynthesis, and its application has been recently shown to positively influence colour development in table grapes, with negligible effects on sugar accumulation (de Aguiar et al., 2024). Exogenous ACC applications conducted using the same commercial product as in de Aguiar et al. (2024) were not effective at advancing ripening of SR fruit in the present study (Figure 5A). These results add to existing work on ethylene in grapes and point to its far from fully understood involvement in berry ripening. Due to the lack of effects observed in our first trial, larger attention was given to the highly significant ABA treatments that resulted in the desired outcome of rescuing ripening. The role of ABA in grape ripening has been thoroughly investigated, and its involvement in triggering malate and sugar transport to the apoplast has been demonstrated (Castellarin et al., 2011; Pilati et al., 2017). ABA is also hypothesised to contribute to the regulation of grape berry ripening in other ways that are still not fully elucidated. Over three years, substantial evidence was collected on the efficacy of ABA to rescue ripening in the SR genotype (Figure 3 and Figure 7), which points to incomplete signalling being the most likely reason for the delayed sugar accumulation of the SR genotype. The limited effect of the lower dosage (400 ppm) compared to the prompt response to the highest rate (2000 ppm), suggests suboptimal endogenous levels of ABA in the SR fruit. The timing effect observed in experiment 4 (Figure 7) further confirms that SR fruit had reached a state of proneness to ripen, which was only realised when appropriate levels of chemical signals were delivered to the berry via exogenous treatments. Similarly, we observed that seed maturity in the SR genotype is reached at the end of the green phase, excluding improper seed maturation from being the main factor (data not shown). The hypothesis of limitations in ABA-related pathways is further corroborated by the effect of drying down SR vines in experiment 5, where the WS treatment promptly triggered ripening at the same time as ABA2000 exogenous application (Figure 8). Notably, ABA2000 and WS applied to SR fruit caused TSS accumulation to start at the same time; however, with a more marked accumulation in ABA2000. It is possible that photosynthesis rates were reduced in WS vines concurrently with the measured drop in gs, as previously reported (Gambetta et al., 2020). Such limitation to water and photosynthate allocation to the fruit is the likely cause behind the differences in berry weights between ABA2000 and WS fruit, affecting the final sugar content per berry (Figure S7). The pivotal role of ABA in the reorganisation of berry processes marking the transition from pre- to post-veraison stages is well established (Castellarin et al., 2016; Pilati et al., 2017). As to the SR genotype, a mechanistic explanation for impeded ripening remains unsubstantiated. Given the interplay between ABA and other hormones (Böttcher et al., 2013; Böttcher et al., 2010; Dal Santo et al., 2020; Gouthu & Deluc, 2015), insufficient auxin sequestration and improper ABA biosynthesis or their interaction are all possible reasons. ABA catabolism may also be enhanced in the SR genotype, leading to lower levels of active ABA than those required to counterbalance auxins (Owen et al., 2009). Since natural ABA was used in our trials, experimenting with ABA analogues with longer residual activity in plants and/or analysis of ABA catabolites would allow us to investigate the direct effect of ABA and its fate in SR fruit. To clarify the interactive role of ABA and auxins and the mode of action of exogenous ABA, hormone profiling represents a necessary next step for this research. Based on recent advancements in grape functional genomics, differential expression analysis of genes involved in ripening signalling or regulation would provide important insights into the mechanisms behind the SR trait described in this study. Alongside fundamental work, segregation studies are required to understand the genetic basis of the slow ripening trait and develop tools to incorporate it through traditional or new-generation breeding techniques.

Conclusion

Screening of material displaying phenotypic variation in the dynamics of sugar accumulation represents the basis for developing long-term strategies to cope with a warmer climate and advanced grape maturity. We characterised physiological aspects of ripening in fruit of a slow ripening genotype and found that ripening was delayed by more than 65 days over three years. The rate of sugar accumulation was slower, but to a lower degree. Although this delay may be excessive for commercial applications, this material offers a unique opportunity to explore mechanisms of slow ripening in grapevines. Preliminary experiments on carbon limitation at the leaf level and using severe crop removal to address potential unbalance in source-sink relationships suggested that impediments at these levels are unlikely. Conversely, sugar accumulation was triggered when exogenous ABA was applied to the fruit. These insights are the foundation for future studies on regulatory mechanisms, which could lead the way for modulating this trait for vines better adapted to current climatic conditions.

Acknowledgements

The authors would like to acknowledge the Viticulture Research team at E. & J. Gallo Winery for their precious support with the experimental trials: Nona Ebisuda for field work support; Don Katayama for help with chemical application; Cella Bioni, Miriam Villa and Jared Nicholson for help with sample collection and processing. We appreciate Lindon Inouye from Valent Biosciences for generously providing chemical products used in the trial.

Author contributions

PP, PC, MB, and ND conceptualised and designed the study; PP, EG and OB collected data in the field and performed data analysis; KS provided the Grape Grabber and assistance with data processing and interpretation; MF and SZ helped with data interpretation. All authors have contributed to data interpretation and manuscript writing and have read and approved this manuscript before its submission.