Vineyard layout

The study was conducted during two consecutive growing seasons (2021 and 2022) on the experimental vineyard of Research Station for Fruit Growing, located 5 km south of the town Sint‑Truiden, Belgium (50.774218 N, 5.155002 E) at 75 m above sea level. A total of 120 six‑year‑old vines of Vitis vinifera cv. Chardonnay (clone ENTAV‑INRA 96), grafted onto SO4 rootstocks (Vitis berlandieri x Vitis riparia) were included in the study. Vines were planted in sandy‑loam soil in a single row within the vineyard, which in total had 22 rows of vines. Row orientation was NNW/SSE, with row and vine spacing 2.0 m and 1.0 m, respectively (S1A Fig). The vines were trained on a six‑wire vertical shoot positioning (VSP) and cane‑pruned single guyot system with two spurs on alternating sides of the crown. On alternating years, a fruiting cane and a backup cane was trimmed down to eight nodes, while on the opposing side the spur was trimmed down to two nodes from which the shoots used in the subsequent year were grown. Canopy management involved minimal leaf removal in the fruiting zone and two instances of topping (S1B Fig).

Experimental design

Vines were divided into four treatment groups consisting of 12 vines each, according to a randomised block design. All four treatments were prepared in demineralised water at a 1:500 dilution with 0.001% Tween‑20 (Merck KGaA, Darmstadt, Germany) added as a surfactant to improve foliar uptake of the treatments, according to supplier specifications. The treatments were (i) a blank treatment containing only demineralised water and Tween‑20 with no additional compounds (Control), (ii) a reference treatment for macronutrient content (NPK‑Ref); (iii) a commercial A. nodosum based seaweed extract (Acadian Plant Health, Dartmouth, Canada); and (iv) a commercial E. maxima‑based seaweed extract (COMPO EXPERT GmbH, Münster, Germany) (S2A Fig). NPK‑Ref was prepared from a mixture of KCl (Honeywell International Inc, Charlotte, NC) and commercial monoammonium phosphate (Haifa MAP™ Haifa Group, Haifa, Israel) to approximate the maximum nitrogen, phosphorus, and potassium content (NPK) in the two biostimulant treatments. While it was possible to mimic the maximum N and K in the extracts, the P‑content was 67% more than that of E. maxima, which contained the highest P of the extracts (S1 Table). The Soil Service of Belgium (Leuven, Belgium) determined the nutrient composition of the seaweed extracts. The contents of P, K, Mg, Ca, Na, S were analysed by inductively coupled plasma atomic emission spectroscopy (ICP‑AES), while total N and total C were determined using spectrometry, according to standard protocols, following the Belgian Accreditation Institution (BELAC) reference methods. Treatments were applied at five specific phenological development stages once vegetation had actively started developing, based on supplier recommendations to enhance fruit sizing. In season 2021 applications commenced every three weeks, starting at 18 days before flowering (11 June), which corresponded to stage E‑L 15 according to the modified Eichhorn‑Lorenz (E‑L) system [58], and every two weeks during season 2022, starting 12 days before the flowering stage (E‑L 18/19) (3 June). The final treatment was applied before véraison (E‑L 34) in both seasons(3 September 2021 and 5 August 2022) (S2A Fig). To ensure consistency in plant response, the same phenological stages were targeted in both seasons. As a consequence, differences in application dates and frequency arose, with shorter intervals between treatments in 2022 due to faster phenological progression under warmer climate conditions, as detailed in the seasonal weather results section. On alternating weeks, the vines were treated with potassium hydrogen carbonate (Karma®, Certis Europe B.V, Brussels, Belgium) as a minimal intervention approach to reduce the pressure of Erysiphe necator (the causal agent of powdery mildew in grapevines) (S2A Fig). Treatments were applied in four blocks, each consisting of 30 vines, to a specific subblock consisting of five vines. Additionally, each block contained a subblock of five vines that remained untreated. Between subblocks there was another vine that remained untreated as a buffer between subblocks (S2B Fig). Treatments were applied by spraying both sides of the canopy using a STIHL SR 430 gas‑powered backpack sprayer (ANDREAS STIHL NV, Puurs‑Sint‑Amands, Belgium) until run‑off. All sprays were administered between 8:00 and 9:00 in the morning on clear days. Samples were collected from, and measurements done on, the three inner vines of each treated subblock (3 plants × 4 blocks per treatment), to minimise the impact of treatment carryover between subblocks (S2B Fig).

Weather data

Environmental conditions were logged via an automated weather station located in the vicinity of the vineyard. The daily maximum (T_max), average (T_avg) and minimum temperature (T_min), precipitation, and relative humidity (RH%) were logged from 1 April to 31 October during the 2021 and 2022 seasons. Cumulative growing degree days (GDD) [59] were calculated using the following formula:

(1)

where n is the number of days, T the daily mean temperature and T_b the base temperature (10 °C) for grapevine growth. Vapour pressure deficit (VPD) was calculated hourly using the following formula proposed by [60]:

(2)

where VPD is in kPa, T is the temperature in °C, and RH% the relative humidity. VPD data were then plotted over the course of the growing season.

Vine vegetative characteristics and leaf starch content

Three representative shoots per vine were tagged on the main fruiting cane at E‑L 12 in the 2021 growing season (21 May 2021). These were measured weekly over three weeks with a measuring tape, after which the average shoot length was determined for each replicate vine. Treatments were then assigned to subblocks to ensure each treatment had vines of a comparable phenological development, with each sampled vine a replicate (n = 12) (S2B Fig). During the 2022 growing season three new representative shoots were tagged on the fruiting cane of vines (receiving the same treatments as in 2021) and measured weekly starting at E‑L 15 (18 May 2022) until one week before the start of flowering (E‑L 21). Following the start of treatment application (E‑L 15 on 11 June 2021 and E‑L 18/19 on 3 June 2022), leaf destructive measurements were done every 2‑3 weeks on a single representative, fully expanded mature leaf, exposed to sunlight, which was sampled from each vine (n = 12). The sampled leaves were kept on ice during sampling in the field and stored in the laboratory at ‑20 °C until analysis. Leaves were scanned in colour at 400 dpi using a Ricoh IM C4500 multi‑function printer (Ricoh, Vilvoorde, Belgium), and the leaf area of individual leaves (hereafter referred to as ‘leaf area’) was determined using the Fiji software package [61]. After scanning, leaf petioles were removed with a scalpel and the fresh weight was recorded using an analytical balance (PIONEER PX 323, Ohaus, Nänikon, Switzerland). Leaves were then dried at 70 °C until a constant dry mass was achieved to calculate the leaf water content, dry matter percentage (DM%), and specific leaf area (SLA) by dividing the leaf area by the leaf dry mass.

Each dried leaf per replicate vine (n = 12) was ground to a powder using a mortar and pestle, and the soluble sugars were extracted using an ethanol extraction step, after which the leaf total starch content was determined using a commercial kit for total starches, K‑TSTA‑100 (Megazyme International, Bray, Ireland). Firstly, the soluble sugars extraction was done according to [62], with the following modifications: for the ethanol extraction, 400 μL of 80% ethanol was added to 50 mg of dried leaf powder in a 2 mL screw cap tube. Then, samples were vortexed and boiled for 15 minutes at 80 °C in a heating block (Model 2053‑1, Lab‑Line Instruments Inc., Melrose Park, IL). Samples were then vortexed again and centrifuged at 12 500 RCF for 10 minutes and the supernatant was discarded, retaining the leaf powder pellet. These steps (adding ethanol to the pellet, vortexing, centrifuging, and discarding the supernatant) were repeated another two times to ensure most of the soluble sugars were removed. Then the remaining pellet was dried overnight at 70 °C and the starches were extracted using the aforementioned starch extraction kit (with the protocol optimised for use with microtiter plates [63]). Briefly, this involved adding 100 μL of 80% ethanol and 250 μL of ice‑cold 1.7 M NaOH (VWR, Leuven, Belgium) and two 2 mm glass beads to the screw cap tube containing the dried leaf pellet. Tubes were then vortexed, and the content was homogenised in a bead mill homogeniser (Bead Ruptor Elite, Omni International, GA, USA) using a speed of 4.5 m s^‑1 with five cycles of 30 seconds each. Next, 1000 μL of a 100 mM sodium‑acetate buffer at pH 3.7 was added and the tubes were vortexed. Then, 12.5 μL of undiluted thermostable α‑amylase was added, followed immediately by 12.5 μL amyloglucosidase, after which the tubes were incubated at 50 °C for 60 minutes to break down and convert the starch to D‑glucose. Tubes were then centrifuged at 13 000 RCF for three minutes and the retained supernatant transferred to new 1.5 mL microcentrifuge tubes. The sample supernatants were then diluted with 100 mM sodium‑acetate buffer at pH 5.0, depending on the expected final concentrations of starch (e.g., 1:4 dilution early season and 1:7 late season). Subsequently, following brief vortexing, 20 μL of each sample was transferred in duplicate to a flat bottom 96‑well plate (96 MicroWell™ Plates, Nunc™, VWR, Leuven, Belgium), to which 300 μL glucose oxidase/peroxidase (GOPOD) reagent (from the K‑TSTA‑100A extraction kit) was added. Plates were closed and incubated at 50 °C for 20 minutes, after which the absorbance was read using a SpectraMax® ABS spectrophotometer (Molecular Devices, San Jose, CA, USA) with the wavelength set to 510 nm. The total starch was then calculated using a standard curve of glucose as specified in the kit protocol.

Chlorophyll fluorescence indices and stomatal conductance

Two mature, fully expanded leaves, exposed to full sunlight, and of a similar age (corresponding to leaves between the third and fifth node of the shoots) were tagged per replicate vine (n = 12) for leaf physiological measurements. The leaf chlorophyll content index (CCI) was measured in duplicate on each tagged leaf (giving four readings that were averaged per replicate vine) using a CCM‑200 chlorophyll content meter (Opti‑Sciences, Hudson, NH) between 10:00 and 11:30 in the morning. The CCI is a transmittance‑based measure of the relative chlorophyll content in leaves with a species‑dependent power‑law relationship with respect to absolute chlorophyll concentration [64]. Next, on the same leaves and within the same time interval, the dark‑adapted fast chlorophyll‑a fluorescence induction (OJIP) transient was measured and recorded using a handheld Pocket PEA chlorophyll fluorimeter (Hansatech Instruments, Norfolk, United Kingdom). From the OJIP induction curves the following photosystem II (PSII) parameters were derived: the reaction centra density per excited cross section (RC∕CS), the maximum quantum yield of PSII photochemistry (F_v∕F_m)—often used as a measure of plant performance or photosynthetic potential [65], and the quantum yield for electron transport beyond QA⁻ (PSII electron transport efficiency, Φ_E0). For reference, in healthy, unstressed leaves, the maximum quantum yield (F_v∕F_m) typically approaches 0.83 [66]. Before measuring, leaves were dark‑adapted for 30 minutes to ensure all PSII RCs were open, allowing for estimation of the minimal fluorescence (F₀). Leaves were then subjected to a saturating actinic light flash (627 nm, 3500 μmol m^‑2 s^‑1), to determine maximum fluorescence (F_m). The maximum quantum yield of PSII photochemistry (F_v∕F_m) was calculated as (F_m–F₀)∕F_m [67]. The two leaf readings were averaged per replicate vine.

Additionally, leaf stomatal conductance was measured between 11:30 and 14:00 on the abaxial (under) side of the first tagged leaf (as measured from the crown of the vine) for each replicate vine (n = 12). Measurements were done for 30 seconds until steady state using an SC‑1 leaf porometer (METER Group, Pullman, WA, USA). Parameters were measured weekly (between 15 June and 25 August in 2021; and 15 June to 12 August in 2022), except for days with rain or full cloud cover, where the measurement was taken the following week on a cloudless day.

Berry classical parameters and morphology

Berries were sampled from véraison (E‑L 34) until harvest (E‑L 38) every 14 days for the 2021 growing season, and every 10‑14 days for the 2022 season, with delayed sampling of at least two days following rainfall to reduce the impact of sugars dilution (S2C Fig). A total of about 80 representative berries were sampled from healthy clusters across the inner three vines of a treated panel (subblock) (n = 4), using sterile nitrile gloves, into resealable bags, sampling from the top, middle, and bottom of the cluster, and alternating between sun‑exposed and shaded sides. Berries were kept on ice during sampling and transport until further processing. Thirty representative berries that were used for morphological measurements (n = 4) were stored in resealable bags at ‑20 °C, while the remainder were immediately processed at the laboratory for juice analyses. Briefly, the remaining berries were gently crushed by hand, externally through the bag, to extract a representative juice sample for each subblock (n = 4), and the juice was decanted and centrifuged at 4000 RCM for five minutes. The clear must was used to determine total soluble solids (TSS in °Brix), pH, and titratable acidity (TA in g L^-1). TSS was determined with a digital refractometer (HI 96801, Hanna Instruments, Temse, Belgium) in duplicate and averaged per replicate. A pH probe (HI 1131, Hanna Instruments, Temse, Belgium) was used to determine the pH, while TA was measured using the same probe in conjunction with an autotitrator (HI 902, Hanna Instruments, Temse, Belgium). Each sample consisted of 20 mL juice and 20 mL demineralised water, titrated with 0.33 M NaOH (VWR, Leuven, Belgium) to a pH of 7.01, with results reported as tartaric acid equivalent in g L^-1. For berry morphological measurements, 20 undamaged berries were randomly selected from each sampled subblock (n = 4). The berry length from pedicel to the bottom and berry diameter were measured using a vernier calliper (HOLEX 412821_100, Hoffman Group, Borne, The Netherlands). Individual berry weights were also recorded on an analytical balance (PIONEER PX 323, Ohaus, Nänikon, Switzerland), taking care that the berries remained frozen and surface ice was removed. The berry volume was calculated from the berry length and diameter, assuming a perfect spherical shape. Individual berry densities were calculated by dividing the berry mass by the berry volume. Following this, the 20 berries were dried at 75 °C until a constant mass was observed to determine the berry dry mass and DM%. The morphological measurements of the 20 berries were averaged per replicate subblock (n = 4).

Harvest data

At technological maturity (TSS = 21 °Brix) all berry clusters were harvested using sterile nitrile gloves into separate bags for each of the inner three vines per subblock, yielding 12 bags per treatment. The clusters were counted and weighed for each replicate vine (n = 12), to obtain an average number of clusters and average cluster weight for each replicate vine. Following this, the clusters were pooled per treated block, yielding four replicates per treatment. From each replicate, 100 berries were randomly sampled, of which 30 were frozen at ‑20 °C for morphological measurements and the remainder were immediately crushed to juice for analysis (as described previously).

Statistical analysis

A series of linear mixed models (LMMs) were fit against the longitudinal vegetative and berry quality data. A significance level of α = 0.05 was used to determine significant differences in all analyses. For the vegetative dataset, an initial global model was constructed using the ‘lmer’ function from the R lme4 package [68]. Fixed effects included season, treatment, and time (with all interactions among these factors), while environmental covariates (temperature, rain, humidity, VPD, and GDD) were incorporated as additive predictors. Each replicate vine (n = 48 per timepoint, with 12 per treatment) was included as a random factor nested within block (n = 4). This was done to account for the repeated measures taken on each vine and for the non‑independence of observations, thereby preventing pseudoreplication [68]. To identify the most parsimonious model, the ‘dredge’ function from the MuMIn package was used to compare candidate models starting from the global model estimated via restricted maximum likelihood (REML), while constraining all models to include the fixed effects season, treatment, and time. The most parsimonious model was selected based on the lowest corrected Aikaike Information Criterion (AICc) [69]. Subsequently, separate analyses were performed for each year by fitting optimal LMMs via ‘dredge;’ in these within‐year models the design was balanced, as each treatment was associated with identical levels of time, with the same number of replicates per timepoint (n = 12).

In all cases, model assumptions were thoroughly checked. Specifically, we examined residual versus fitted value plots and Q‑Q plots to assess normality and homogeneity of variances, while simulated residuals were generated using the DHARMa package, to further assess model fit and detect potential issues such as non‑uniformity, over‑ or under‑dispersion, and zero inflation. When evidence of heteroskedasticity was observed, models were refitted using the ‘lme’ function from the nlme package with an appropriate variance structure (e.g., using weights = varIdent(form = ~ 1 | Time:Treatment)) and maximum likelihood estimation (ML). These models were compared to the previous optimal LMM using AICc and R² to determine if model fit was significantly improved (ΔAICc > 10) [69]. Final models were refitted using REML, and an analysis of variance (ANOVA) type III test was performed with Kenward‑Roger degrees of freedom estimation to assess the significance of the fixed effects. The analysis was followed by a post hoc pairwise comparison using estimated marginal means (EMM) with Satterthwaite degrees of freedom adjustment, via the ‘emmeans’ package [70]. In cases where the model included a significant interaction term, EMMs were evaluated at each level of the interacting factor using the Šidák confidence level adjustment for multiple comparisons. In the absence of significant interaction effects, pairwise comparisons among treatment levels were conducted using EMMs with Tukey adjustment for multiple comparisons, implemented via the ‘emmeans’ package.

For the berry quality data, a similar modelling approach was followed with season and treatment as factors, including the aforementioned environmental covariates as additive predictors in the global starting model. The exception was that ‘time’ was modelled using a 2^nd order polynomial of the means‑centred DAAs, to account for slowing ripening dynamics over time. Furthermore, as berry sampling was done per treatment per block, the total replicates per timepoint were reduced to n = 16, with four per treatment. Data were first assessed for normality and homogeneity of variances. Continuous response variables (e.g., yield, average cluster mass) were analysed using LMMs, whereas count data (e.g., number of berry clusters) were modelled with a Generalised Linear Model (GLM) assuming a negative binomial distribution with a log link (using the ‘glm’ function from the lme4 package) [68]. Following significant associations of season or its respective interaction terms with a response variable, each season was modelled separately—with treatment, time, and their interaction as fixed effects, and block as a random factor. For each season, the most parsimonious model was selected (based on AICc) and ANOVA was conducted on that model to report overall effects. A priori planned pairwise comparisons were performed using EMMs [70] on the model that retained the treatment × time interaction—even if that term was excluded from the final parsimonious model. This allowed us to analyse harvest‑specific contrasts rather than merely reflecting the overall treatment effect. All contrasts (including sequential differences between timepoints, the overall treatment effect, and the treatment effect at harvest) were adjusted with the Šidák correction for multiple comparisons. For yield parameters, a priori analyses compared the change from 2021 to 2022 within each treatment (i.e., using EMMs and excluding cross‑treatment interaction contrasts). Correlations between variables were assessed via a repeated‑measures correlation using the R package ‘rmcorr,’ to account for the repeated measures. Results were visualised using the ‘ggplot2’ package [71]. All analyses and visualisation of the data were performed in R version 4.2.3 [72].

Weather data

Whereas the nine‑year average of cumulative growing degree days (GDDs) between 2013 and 2021 was 1114, during the 2021 growing season (April to October) 1001 GDDs were accumulated, while 2022 saw a total of 1230 GDDs. In 2021, temperatures were cooler than average in April (‑2.8 °C) and May (‑0.9 °C), warmer in June (+2.0 °C), and cooler in July (‑1.9 °C) and August (‑1.8 °C) (Fig 1A). Rainfall was significantly higher in July (231 mm vs 53 mm) and August (136 mm vs 62 mm), with lower‑than‑average rainfall in September (28 mm vs 45 mm) (Fig 1B). Vapour pressure deficit (VPD) exceeded 3 kPa only on three days before anthesis (flowering) (Fig 1C). In 2022, compared to the nine‑year average (2013 to 2021) April was slightly cooler (‑0.7 °C), followed by a warmer May (+2.4 °C), average June and July, and warmer August (+2.1 °C), with a cooler September (‑1.1 °C) (Fig 1A). Rainfall was above average in April and June, but below average in July (9 mm vs 53 mm) and August (8 mm vs 62 mm), with higher‑than‑average rainfall in September (116 mm vs 45 mm) (Fig 1B). The 2022 season had elevated VPD values post‑anthesis (flowering), exceeding 3 kPa on 18 days, peaking at 6.7 kPa at 34 days after anthesis (Fig 1C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. (A) Mean daily temperature and cumulative growing degree days (GDD), (B) monthly rainfall, and (C) vapour pressure deficit (VPD) for the 2021 and 2022 seasons.

(A) Mean daily temperatures are indicated by blue circles for 2021, red triangles for 2022, and a purple dashed line for the average of the preceding eight years (2013‑2020). The purple shading indicates the temperature range over this period, and solid lines represent the GDDs. The maximum daily VPD is shown with blue circles for 2021 and red triangles for 2022 (C).

https://doi.org/10.1371/journal.pone.0331039.g001

Seaweed extract effect on grapevine vegetative growth and physiology

In 2021, vines were treated with a negative control (water), two seaweed extracts based on A. nodosum and E. maxima, and an NPK‑Ref treatment. During this season, there were no significant differences in shoot lengths between the different treatment groups (P = 0.107) (S4 Fig). In the following season (2022) the average shoot length of three tagged shoots emerging from the fruiting cane was found to be significantly influenced by treatment (P = 0.008) and time (date of measurement) (P < 0.001). Specifically, the NPK‑treated vines had significantly longer shoots on average compared to water treated vines (EMM: 36.8 ± 2.2 cm vs 29.7 ± 2.0 cm, P = 0.006) over the 2022 growing season, while no differences were observed for the seaweed extract treated vines relative to control vines (S4 Fig).

For vegetative growth parameters, results showed strong main effects of time (development stage) on the leaf area, leaf dry matter (DM%) and leaf starch content, with less pronounced effects on leaf dry mass (S2 Table). Season showed a strong effect on leaf area, dry mass, and leaf starch content. Treatment only had a significant effect on leaf area (F_3,50.9 = 5.524, P = 0.002), but not on leaf dry mass, specific leaf area, leaf DM% and leaf starch content (S2 Table).

When analysing each season independently, treatment also showed no significant association with leaf dry mass in 2021 (F_3,44 = 5.340, P = 0.149; Fig 2A), and 2022 (F_3,44 = 7.611, P = 0.055; Fig 2B). Leaf area, in contrast, was associated with improvements in both seasons, with vines treated with the A. nodosum extract having larger leaves in 2021 (+12%, P = 0.009; Fig 2C) and 2022 (+15%, P = 0.010; Fig 2D) compared to control vines. While the effect was stronger (+8%) relative to the NPK‑Ref vines, it was not statistically significant (P = 0.355; Table 1). However, by 2022 leaf area was 18% larger in NPK‑Ref vines (P = 0.002) compared to control vines (Table 1, S3 Table), reaching a comparable size to A. nodosum vines (P = 0.905). While there was no evidence that E. maxima increased leaf area relative to control vines in either 2021 or 2022, there were also no significant differences observed relative to A. nodosum and NPK‑Ref vines in both years (Fig 2C and 2D, Table 1). Overall, leaves were significantly smaller in 2022 compared to 2021 (F_1,270 = 45.70, P < 0.001; S2 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Comparison of treatment effects on vegetative growth and physiological parameters of V. vinifera cv. Chardonnay averaged over the 2021 and 2022 seasons.

https://doi.org/10.1371/journal.pone.0331039.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. The evolution of leaf dry mass (2021, A; 2022, B) and leaf surface area (2021, C; 2022, D) of V. vinifera cv.

Chardonnay treated with a water control, an A. nodosum extract, an E. maxima extract, and an NPK‑Ref treatment. Each time‑series point represents the mean ± standard error (n = 12). Violin‑boxplots (central panels) show parameter distributions averaged across timepoints per season, with boxplot lines marking the 1^st quartile, median, and 3^rd quartile, the diamond indicating the mean and whiskers extending to 1.5 × the interquartile range. Different uppercase letters indicate significant differences based on EMMs between treatments during a season, averaged over Time (P < 0.05). Treatment application days are shown as dark blue bars on the x‑axis (DAA), the first treatment at 18 and 14 days before anthesis is omitted. Samples were taken on dry and clear days as far as possible; blue shading indicates periods with cool/cloudy conditions, and red shading denotes VPD stress (>3 kPa).

https://doi.org/10.1371/journal.pone.0331039.g002

To evaluate the effect of the seaweed extracts on vine physiology, the leaf chlorophyll content index (CCI; proportional to the absolute amount of chlorophyll in the leaf) and the stomatal conductance of the vine were measured. Additionally, fluorescence transients were measured and analysed to obtain the reaction centra density (RC∕CS), the maximum quantum yield (F_v∕F_m), and the electron transport efficiency of photosystem II (Φ_E0). The global models showed strong main effects of season and time (development stage) on CCI, RC∕CS, F_v∕F_m, Φ_E0, and the stomatal conductance of the vines (S2 Table). A modest effect of treatment was observed only for CCI (F_3,44 = 3.556, P = 0.022), however, strong season × treatment interactions were detected for CCI (F_18,689 = 10.54, P < 0.001) and RC∕CS (F_18,689 = 6.213, P < 0.001; S2 Table). The remaining results of the global models are summarised in S2 Table. Given the clear differences between seasons and multiple interactions affecting photosynthetic parameters, separate analyses were done for 2021 and 2022.

In 2021, the main effects of treatment and development stage were associated with differences in CCI. Specifically, between berry set (E‑L 27) and the onset of berry softening (E‑L 34) the average CCI decreased by 14% from a peak of 18 units to 15.6 (P < 0.001; Fig 3A). Regarding treatment, A. nodosum vines maintained a 12% higher CCI compared to negative control and E. maxima vines (P = 0.019 and P = 0.029), while compared to NPK‑Ref vines there was no significant difference (P = 0.144; Fig 3A, Table 1). In contrast, E. maxima vines maintained a comparable CCI to both negative control (P = 0.998) and NPK‑Ref vines (P = 0.892) (Fig 3A, Table 1). Subsequently, in 2022, only the main effect of development stage was associated with differences in CCI, declining steadily from a peak of 13.2 units at E‑L 31 to 11.1 units at E‑L 35 (‑19%, P < 0.001; Fig 3B). Furthermore, CCI values were significantly lower compared to 2021 which was characterised by cloudier and cooler conditions (Fig 1A, Fig 3A and 3B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. The evolution of the leaf chlorophyll content index CCI (2021, A; 2022, B) the density of reaction centra per cross section RC∕CS (2021, C; 2022, D), the maximum quantum yield F_v∕F_m (2021, E; 2022, F) and the electron transport efficiency of photosystem II Φ_E0 (2021, G; 2022, H) of V. vinifera cv. Chardonnay treated with a water control, an A. nodosum extract, an E. maxima extract, and an NPK‑Ref treatment.

Each time‑series point represents the mean ± standard error (n = 12). Asterisks indicate significance of the season level factors based on two‑way ANOVA (0.01 < P ≤ 0.05: *; 0.001 < P ≤ 0.01: **; P ≤ 0.001: ***; ns = not significant; see S4 Table for means separation by EMMs). Violin‑boxplots (central panels) show parameter distributions averaged across timepoints per season, with boxplot lines marking the 1^st quartile, median, and 3^rd quartile, and whiskers extending to 1.5 × the interquartile range. Different uppercase letters indicate significant differences based on EMMs between treatments during a season, averaged over Time (P < 0.05). When no letters are shown, a significant Treatment × Time interaction was present. Treatment application days are shown as dark blue bars on the x‑axis (DAA). Measurements were taken on dry and clear days as far as possible; blue shading indicates periods with cool/cloudy conditions, and red shading denotes VPD stress (>3 kPa).

https://doi.org/10.1371/journal.pone.0331039.g003

Similarly, in 2021, treatment and development stage were also significantly associated with reaction centra density. Ascophyllum nodosum‑treated vines had a 6‑7% higher RC∕CS compared to control (P = 0.009) and E. maxima‑treated vines (P = 0.003), with no difference compared to NPK‑Ref vines (P = 0.357) (Fig 3C, Table 1). In 2022, however, a significant interaction effect between treatment and development stage (P = 0.006) on RC∕CS was observed. While most stages showed similar responses across treatments, at véraison (E‑L 35), A. nodosum vines had a 21% lower RC∕CS compared to control vines (P = 0.007; Fig 3D, S4 Table), with no significant difference relative to NPK‑Ref (P = 0.633) or E. maxima‑treated vines (P = 0.260) (Fig 3D, S4 Table). Overall, RC∕CS declined by 15% in 2021 and 25% in 2022 over the season (both with P < 0.001; Fig 3C and 3D, Table 1). Decreases were particularly sharp in 2022 (‑23%, P < 0.001) following sustained periods of high vapour pressure deficit (VPD > 3 kPa; Fig 3D) and varied widely between treatments, ranging from ‑20% (control) to ‑48% (A. nodosum) between E‑L 23 and E‑L 35 (S4 Table).

Furthermore, A. nodosum‑treated vines had significantly higher F_v∕F_m values on average during the 2021 season compared to NPK‑Ref‑treated vines (+1.5%, P = 0.020), but not compared to negative control‑treated vines (Fig 3E). This contrasts with 2022, where only development stage (serving as proxy for climatic conditions) was associated with differences in F_v∕F_m (P < 0.001; Fig 3F, Table 1). Meanwhile, E. maxima treatments were not associated with significant differences in the average F_v∕F_m over either season, compared to control (P = 0.566) and NPK‑Ref vines (Table 1). Notably, however, was an observation following a heat stress event in 2022 at 34 days after anthesis (E‑L 32). During this time a maximum temperature of 39.5 °C was reached, while the VPD exceeded 3 kPa for 21 hours over the preceding two days, peaking at 6.7 kPa (Fig 1C). Fourteen days following this event, maximum temperatures remained well below 30 °C and VPD rarely exceeded 2 kPa (Fig 3F, S3 Fig). Vines showed a significant increase in F_v∕F_m values compared to the control vines when treated with the E. maxima extract (+3.0%; P = 0.002) and the NPK‑Ref treatment (+2.6%; P = 0.019), while vines treated with the A. nodosum extract did not show a significant improvement (S4 Table).

Finally, when evaluating PSII electron transport efficiency Φ_E0, neither 2021 nor 2022 showed a significant main effect of treatment on Φ_E0 when modelled without interaction (P > 0.05). However, including a Treat × Time interaction in 2022 (at a large cost to model parsimony, ΔAICc ≈ 230), yielded a borderline significant interaction effect (P = 0.040). Post hoc comparisons indicated modest differences at the final two stages (E‑L 34 and E‑L 35). At E‑L 34, A. nodosum vines had a higher Φ_E0 than control vines (EMM: 0.398 ± 0.017 vs. 0.352 ± 0.017, P = 0.033), while E. maxima and NPK‐Ref were intermediate (S4 Table). By E‑L 35, NPK‐Ref had the highest Φ_E0 (EMM: 0.444 ± 0.017), differing from A. nodosum (EMM: 0.390 ± 0.017, P = 0.009) and E. maxima (EMM: 0.399 ± 0.017, P = 0.038), but not from control (EMM: 0.414 ± 0.017, P = 0.293). While no consistent overall pattern emerged, results from a repeated‐measures correlation analysis (rrm) revealed a significant positive relationship between F_v∕F_m and Φ_E0 in 2021 (rrm = 0.68, P < 0.001, 95% CI [0.61, 0.75]) and 2022 (rrm = 0.75, P < 0.001, 95% CI [0.71, 0.79]).

Seaweed extract effect on berry development and quality

The results of the regression analyses showed strong main effects of season and time (development stage) on berry volume, berry mass, berry density and berry DM% (S5 Table). Furthermore, a significant main effect of VPD and an interaction effect between season and time, was observed for berry density and berry DM% (S5 Table). Treatment did not show a significant association with any of the measured parameters, while interaction terms with treatment were not included in the most parsimonious models (P > 0.05; S5 Table). Due to the significant effect of season and the comparatively lower P‑values, separate analyses per season were also done for berry volume and mass.

Berry volume was not significantly associated with treatment in 2021 (P = 0.082), while a treatment effect was observed in 2022 (P = 0.005). In particular, compared to control, vines treated with A. nodosum and NPK‑Ref had an increased berry volume averaged over the season (+8%, P = 0.022 and P = 0.037), with no significant difference detected between them (P = 0.999; Fig 4B). However, at harvest differences were not significant compared to control for either A. nodosum or NPK‑Ref (P = 0.787 and P = 0.987; Fig 4A and 4B, Table 2). A similar trend was observed for berry mass, with no significant treatment association in 2021 (P = 0.085), in contrast to 2022 (P = 0.001), where vines treated with A. nodosum and NPK‑Ref had on average an increased berry mass (+8%, P = 0.008 and P = 0.009; Fig 4D). Again, at harvest, differences compared to control were not significant for either A. nodosum or NPK‑Ref vines (P = 0.738 and P = 0.954; Fig 4C and 4D, Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Comparison of treatment effects at harvest on berry size and quality metrics of V. vinifera cv. Chardonnay for the 2021 and 2022 seasons.

https://doi.org/10.1371/journal.pone.0331039.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Berry volume (2021, A; 2022, B) and berry mass (2021, C; 2022, D) of V. vinifera cv. Chardonnay treated with water, an A. nodosum extract, an E. maxima extract, and an NPK‑Ref treatment.

Each time‑series point represents the mean ± standard error (n = 4). Asterisks indicate significance of the season level factors based on two‑way ANOVA (0.01 < P ≤ 0.05: *; 0.001 < P ≤ 0.01: **; P ≤ 0.001: ***; ns = not significant. Boxplots (central panels) show parameter distributions at harvest, with boxplot lines marking the 1^st quartile, median, and 3^rd quartile, and whiskers extending to 1.5 × the interquartile range. Different lowercase letters indicate significant differences between treatments at harvest, based on EMMs (P < 0.05). Different uppercase letters indicate significant differences based on EMMs between treatments during a season, averaged over Time (P < 0.05). Samples were taken on dry and clear days as far as possible; blue shading indicates periods with cool/cloudy conditions with rainfall.

https://doi.org/10.1371/journal.pone.0331039.g004

Additionally, the berry classical parameters of titratable acidity–TA, pH, total soluble solids–TSS (°Brix), and berry sugar content were measured to evaluate the effect of treatment on ripening dynamics. The global model showed strong main effects of season and development stage (time) on all classical parameters, with a significant season × time interactions for TA, TSS, and berry sugar content. However, neither treatment, nor its interaction terms, had a significant influence in the global models (S5 Table), while the effect of the environmental covariates VPD and rain is shown in S5 Table. Given the significant effect of season, and its interaction with time, subsequent analyses were conducted separately for the 2021 and 2022 datasets to better resolve treatment effects, particularly at harvest.

While treatment was not associated with differences in TA in 2021 (F_3,12 = 0.80, P = 0.519) or 2022 (F_3,12 = 0.74, P = 0.548), the model-estimated mean (EMM) for TA at harvest decreased from 10.3 ± 0.3 g L^-1 in 2021 to 7.7 ± 0.2 g L^-1 in 2022 (‑25%, P < 0.001; Fig 5A and 5B). The raw means for each treatment are shown in Table 2. In contrast, TSS differed significantly among treatments over the 2021 season (P = 0.045), but not 2022 (P = 0.346), while the EMMs for TSS at harvest were comparable in 2021 and 2022 (EMM: 20.7 ± 0.2 and 21.0 ± 0.2 °Brix, P = 0.999; Fig 5C and 5D). Main effects analysis revealed that A. nodosum extract–treated vines had an average increase of 6% in TSS relative to control vines (P = 0.028; Fig 5C) over the season, and were not different from NPK‑Ref vines (P = 0.999; Fig 5C). However, by harvest, the effect of A. nodosum was no longer significant in relation to control vines (P = 0.999), with the TSS for all treatments having converged (Fig 5C). See Table 2 and S7 Table for raw means.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Development of titratable acidity–TA (2021, A; 2022, B), total soluble solids–TSS (2021, C; 2022, D) and berry sugar content (2021, E; 2022, F) of V. vinifera cv.

Chardonnay treated with a water control, an A. nodosum extract, an E. maxima extract, and an NPK‑Ref treatment. Each time‑series point represents the mean ± standard error (n = 4). Asterisks indicate significance of the season level factors based on two‑way ANOVA (0.01 < P ≤ 0.05: *; 0.001 < P ≤ 0.01: **; P ≤ 0.001: ***; ns = not significant. Boxplots (central panels) show parameter distributions at harvest, with boxplot lines marking the 1^st quartile, median, and 3^rd quartile, and whiskers extending to 1.5 × the interquartile range. Different lowercase letters indicate significant differences between treatments at harvest, based on EMMs (P < 0.05). Different uppercase letters indicate significant differences based on EMMs between treatments during a season, averaged over Time (P < 0.05). Samples were taken on dry and clear days as far as possible; blue shading indicates periods with cool/cloudy conditions with rainfall.

https://doi.org/10.1371/journal.pone.0331039.g005

In the model of sugar content per berry, a significant association with treatment was observed in 2021 (P = 0.018) and 2022 (P = 0.004), with no significant interaction with time. Main effects analysis of berry sugar content revealed that, following correction for multiple comparisons, none of the treatments differed significantly during ripening or at harvest, relative to control vines in 2021 (Fig 5E, Table 2). In contrast, for 2022, NPK‑Ref vines had a higher sugar content during the season compared to control vines (EMM: 306 ± 9 vs 276 ± 9 mg, P = 0.012; Fig 5F). The EMM for A. nodosum (300 ± 9 mg) did not differ from control vines (P = 0.099) or NPK‑Ref vines (P = 0.999; Fig 5F). Similarly, E. maxima vines did not differ significantly from control or NPK‑Ref vines in terms of berry sugar content. Finally, at harvest, EMMs indicated no significant differences between treatments (P = 0.531; Fig 5F). The corresponding raw means at harvest ranged from 373 ± 16 mg (control) to 400 ± 13 mg (NPK‑Ref); see Table 2 and S7 Table for full values.

Seaweed extract effect on harvest parameters

At harvest, the treatments were not significantly associated with differences in the number of berry clusters per vine (Χ²(1, N = 32) = 2.10, P = 0.552; Fig 6A), cluster mass (F_3,85 = 0.15, P = 0.929; Fig 6B) or yield per vine (F_3,85 = 0.31, P = 0.816; Fig 6C), with season being the main driver of differences for these parameters (P < 0.001). Although treatments were not associated with differences in yield in 2021 (P = 0.820) or 2022 (P = 0.652), a priori specified comparisons revealed that the increase in yield and number of clusters from 2021 to 2022, varied by treatment. The number of berry clusters was significantly higher in NPK‑Ref vines (+46%, P = 0.010), but not in A. nodosum, E. maxima, nor control vines (P > 0.05; Fig 6A). On the other hand, vines treated with E. maxima and NPK‑Ref had an increased yield per vine of 60% (P = 0.040), and 80% (P = 0.007) on average respectively, while in A. nodosum and control vines there was not a significant increase (P > 0.05; Fig 6C). No significant increases in cluster mass were observed for any of the treatments (Fig 6B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Total number of clusters (A), average cluster mass (B) and average yield per vine (C) of V. vinifera cv.

Chardonnay treated with a water control treatment, an A. nodosum extract, an E. maxima extract, and an NPK‑Ref treatment (n = 12) for 2021 (empty violins) and 2022 (striped violins). Samples were taken at technical maturity (average of 21 °Brix). Shaded violin plots show the distribution of data, bar hinges represent quartile values, while whiskers extend to 1.5 × the inter quartile range and the sample mean is indicated with a white diamond. Treatments marked with an asterisk differed significantly from 2021 to 2022 based on their EMMs (P < 0.05).

https://doi.org/10.1371/journal.pone.0331039.g006

Seaweed extracts show environment‑dependent improvements in leaf anatomy and vine physiological parameters

As for many other plant species, grapevine leaves are crucial for photosynthesis, supporting various growth aspects such as new shoot growth and berry ripening [73]. In our study, foliar applications of an A. nodosum extract increased leaf area of vines by +12% and +15%, comparable to the NPK‑Ref treatment (Fig 2C and 2D). In both seasons, foliar treatments were applied during rain‑free periods of at least 24 hours to ensure effective foliar absorption. Similar results have been reported on other white winegrape cultivars such as Narince [74], and white table grape cultivars such as Perlette [42], under warm climate conditions. Enhanced vegetative growth following A. nodosum application likely resulted from its rich composition of growth‐stimulating phytohormones (cytokinins and auxins), which drive cell division and leaf cell expansion [75,76]. This suggests that A. nodosum may be a viable substitute or adjunct to synthetic foliar NPK treatments to increase canopy density. Additionally, A. nodosum‑treated vines showed increased chlorophyll content index (+12% CCI; Fig 3A), and increased active photosystem II reaction centre density (+6% RC∕CS; Fig 3C), under the cooler, cloudier climate conditions in 2021. Betaines present in A. nodosum likely modulate the preservation of chlorophyll by delaying chlorophyll degradation and preserving leaf greenness, especially under suboptimal conditions [77]. These effects corresponded with a modest, yet significant increase (+1.5%) in the maximum quantum yield of photosystem II (F_v∕F_m; Fig 3E) relative to the NPK‑Ref treatment, without significant changes PSII electron transport efficiency (_ΦE0; Fig 3G). However, in the sunnier 2022 season, these effects were less pronounced, with significantly lower CCI and RC∕CS compared to 2021 (Fig 3A and 3B). This is consistent with plant responses involving chlorophyll degradation and activation of their thermal energy‑dissipation and antioxidant systems [78,79], to avoid damage to the photosynthetic apparatus under high irradiance. In contrast, E. maxima showed limited effects overall, likely due to comparatively higher levels of auxins relative to cytokinins (360 × higher), primarily enhancing root growth and elongation [80]. However, an increase in F_v∕F_m was observed two weeks after a heat stress event at the start of véraison in 2022 (T_max of 39.5 °C and VPD of 6.7 kPa; Fig 3F, Fig 1A and 1C). Gibberellins and brassinosteroids, present in large amounts in E. maxima [39], elicit a wide range of physiological responses—promoting cell division, flower and fruit development, and abiotic stress protection [81]. Brassinosteroids in particular modulate the production of reactive oxygen species (ROS), as well as proline, a key osmolyte with protective effects against abiotic stressors (e.g., oxidative stress, drought, salinity, nutrient limitation, and extreme temperatures) [39,82]. Proline also enhances photosynthesis by protecting RuBisCO activity [83] and stabilising membrane structures to reduce dehydration damage [82]. Although hormonal differences likely explain the observed differential effects between extracts, the hormone concentrations of our specific seaweed batches were not quantified. Differences in nutrient content between A. nodosum and E. maxima (~59 times more sulphur and ~73 times more magnesium; S1 Table) further support varying physiological effects, given their roles in chlorophyll synthesis and activation of important enzymes like RuBisCO [47]. While previous studies suggest that A. nodosum extracts could maintain plant water status while retaining stomatal opening under water‑stressed conditions (allowing for increased carbon assimilation) on some grape varieties [48,50], our study showed no significant differences in stomatal conductance under drier conditions in 2022. This is potentially due to Chardonnay’s anisohydric nature, which stabilises photosynthetic performance through lower stomatal sensitivity, at the cost of more variable midday leaf water potentials [84,85]. However, it could be indicative of ample soil moisture due to high rainfall from the previous season, though leaf water potentials were not measured in this study to confirm this hypothesis.

Berry morphology and classical parameters showed season‑dependent responses to seaweed extracts

Berry morphology and ripening were significantly influenced by seasonal variation between 2021 and 2022, with modest treatment effects. No effects were seen in 2021, while in 2022 A. nodosum‑treated vines had an 8% increase in volume and mass relative to control vines when averaged over the season, similar to the NPK‑Ref treatment (Fig 4B and 4D). In contrast, E. maxima vines did not differ from either A. nodosum and NPK‑Ref, or control vines, aligning with a study on three red winegrape varieties [54]. While leaves were significantly smaller in 2022 compared to 2021 (Fig 2), A. nodosum and NPK‑Ref vines had relatively larger leaves in each season compared to control vines. Further, more optimal photosynthesis conditions were measured at the start of ripening in 2022 (higher F_v∕F_m and PSII electron transport efficiency, Φ_E0; Fig 3E to 3H). These factors may have facilitated enhanced carbon assimilation, contributing to increased sugar loading in NPK‑Ref and A. nodosum vines (306 and 300 mg vs. 276 mg control, Fig 5F). and higher osmotic pressure in berries, promoting water uptake and expansion (Fig 4B) [86]. Previous studies corroborate these findings, showing that A. nodosum extracts enhance berry size and mass in white wine‑ and table grape varieties [42,74,87]. The reported presence of auxins and gibberellins may drive early stage cell division [88], with the comparatively high cytokinin content [38,76] driving later berry expansion [88]. Glycine‑betaine, an osmolyte in A. nodosum, may further enhance cell expansion through osmotic regulation [77]. The comparatively limited effects of E. maxima might be due to its different hormonal balance, notably a higher auxin‑to‑cytokinin ratio [39]. Additionally, A. nodosum’s substantially greater potassium content (≈144‑fold higher than E. maxima; S1 Table), essential for osmotic balance and water transport, could partly explain its superior performance. Furthermore, when comparing the sugar loading curves for each treatment to the sugar content of the control vines at harvest, it appears that NPK‑Ref, A. nodosum, and E. maxima treated vines may have reached these levels earlier in both seasons (ordered from earliest to latest). These results have implications for wine style, as titratable acidity and pH showed no significant differences between treatments during the season and at harvest, consistent with existing literature [42,44,54,74,87,89–92]. However, by harvest the differences in berry size, mass, TSS, and sugar content were no longer significant. The later half of the ripening period was characterised by increased rainfall (Fig 1B), which may have resulted in a dilution of sugars and acids, alongside an increase in berry volume. However, the lack of significant differences in berry morphology and ripening parameters at harvest likely reflects limited statistical power due to low biological replication (n = 4), potentially masking small‑to‑medium effect sizes.

Yield indicators may be improved by seaweed extract applications

In our study, inter‑annual climatic differences were the primary driver of differences (P < 0.001). While there were no differences in yield components at harvest between treatments within both seasons (Fig 6A to 6C), a priori within treatment contrasts between years showed vines treated with E. maxima had a significant 60% increase year‑on‑year, alongside the NPK‑Ref vines (+80%). Conversely, the observed yield difference of 53% for A. nodosum‑treated vines was not significant in this study, likely due to high inter‑vine variation in 2022. These findings are consistent with literature, which maintain that yield components, like inflorescence formation and cluster number, contribute 60% to final yield and are heavily influenced by the prior year’s environmental conditions and interventions (such as foliar sprays in this study) [93,94]. Treatment effect on yield seemed to be driven mainly by a higher cluster number in 2022 (Fig 6A) combined with a potential increased berry mass (Fig 4C and 4D), from treatment‑enhanced photosynthetic capacity (larger leaves, higher CCI, RC/CS and F_v∕F_m). While previous studies using A. nodosum showed significant yield improvements in hot semi‑arid climates following treatment for Chardonnay [52], Thompson seedless [87], Narince [74], and Merlot [91], and in humid subtropical climates for Niágara Rosada [95]. studies in more comparable climate regions showed no significant yield improvements relative to control vines in red winegrape varieties [44,54,89,90].

While our study provides valuable insights into the effects of seaweed extracts on vine photosynthetic performance, influencing berry ripening and yield indicators [96] under cool‑climate field conditions, several limitations warrant consideration. First, the study design could be improved with more replicate blocks per treatment for berry classical parameters (increasing n = 4 to n ≥ 8), to improve statistical power to detect medium effect sizes. Next, potential drought stress conditions could be better monitored by quantifying soil moisture content or through evaluating plant water status by pre‑dawn leaf water potentials, in particular in 2022 which had drier conditions between E‑L 23 and E‑L 34. Furthermore, while seaweed extract nutrient content was quantified in 2021, it would have been useful if these measurements were repeated in 2022, while also quantifying the secondary metabolite and hormonal profiles of the extracts. By having a more expansive chemical profile, stronger links can be made between the extract effects under different climate conditions, while also accounting for variation due to storage time of the extracts. Future research could explore the use of individual nutrients or hormones in comparison to the seaweed extracts, to gain a clearer understanding of the modes of action. While previous research comparing foliar and soil irrigation applications have demonstrated that foliar treatments yield the strongest effects [97], further research into the application method is warranted. Specifically, different concentrations and application frequencies, as well as more targeted application timepoints, could provide valuable insights for agronomists [98]. Future research should also investigate the effect of different extraction methods on the composition of the seaweed extracts, and the stability of growth promoting compounds over different storage periods. Finally, although outside of the scope of this study, the eventual wine quality could be impacted by seaweed extracts by, for example, impacting key metabolic pathways during ripening, or altering the composition of the microbiome on the berries and leaves, which could be important aspects for future research.