A plant-based biostimulant modulates grapevine susceptibility within a realistic water stress window through priming and phenylpropanoid pathway regulation

22 Fluctuating extreme weather events, coupled with rising average temperatures, can severely 23 impact grapevine physiology and yield. While biostimulants have been gaining acceptance as 24 a short-terms tools to enhance grapevine resilience, their adoption is hindered by inconsistent 25 efficacy, partly driven by unpredictable plant stress levels. Over two contrasting seasons, we 26 integrated physiological, transcriptomic, and metabolomic analyses to investigate how a plant-27 based biostimulant modulates the sensibility of Vitis vinifera under varying intensities of heat, 28 drought, and their combination. This panel of water status , ranging from -0.02 to -1.6 MPa, 29 revealed that the physiological response induced by the biostimulant treatment alleviates water 30 stress within a field-relevant hydraulic window located between -0.4 and -1.2 MPa. Moreover, 31 moderate but constitutive reduction of growth parameters in biostimulant plants, suggest s a 32 trade-off between vegetative development and abiotic stress responses. Accordingly, gene 33 expression analysis revealed an interaction between water availability and the plant response 34 to the biostimulant, which suggest an activation of priming mechanisms. Metabolic profiling 35 supported these findings, highlighting the central role of phenylpropanoid pathway modulation, 36 together with adjustments in ROS dynamics an d stress -related hormone responses, 37 particularly abscisic acid. Overall, this work emphasizes the need for integrating detailed plant 38 water status and leaf gas exchange to accurately evaluate biostimulant performances under 39 abiotic stress. 40 41

Abiotic stress, Biostimulants, grapevine, hydraulic conductance, priming. 42 43 44 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 3

45 Climate change is profoundly reshaping agricultural systems, exposing crops to increasingly 46 frequent and severe episodes of drought, heat waves, and their combination. While each 47 individual stress may have only a minor effect on plant growth and survival, their combined 48 application results in a increasing detrimental impact on the plants 1. Grapevine, one of the 49 most economically valuable crops in the world2 is vulnerable to fluctuations in water availability 50 and rising temperatures, which directly affect photosynthetic efficiency, growth dynamics, berry 51 composition, and ultimately yield stability and wine quality 3. Ensuring grapevine resilience 52 under these conditions is therefore a critical challenge, driving the search for innovative and 53 sustainable strategies that can complement conventional vineyard management. 54 In this context, plant biostimulants have emerged as promising tools to enhance crop tolerance 55 to abiot ic stresses in an environmentally sustainable manner. By stimulating plant growth, 56 physiological and metabolic processes rather than directly supplying nutrients, biostimulants 57 have been shown to improve nutrient use efficiency, modulate hormonal balance, and reinforce 58 tolerance to abiotic stresses such as drought and heat stress4. Commonly, 6 subcategories of 59 non-microbial plant biostimulants are distinguished: chitosan, humic and fulvic acids, animal 60 and vegetal protein hydrolysates, phosphites, seaweed extracts, and silicon 5. More recently, 61 an additional group of plant biostimulant s, the plant -based extracts, has received significant 62 attention. They can be derived from a wide range of plant species 6 and are relatively easy to 63 prepare. Even if this is dependent of the extraction process, these extracts are rich in 64 structurally diverse bioactive molecules, including phenolic acids, flavonoids, te rpenes, 65 alkaloids, and other secondary metabolites. These compounds have been reported to regulate 66 antioxidant activity, modulate signaling pathways, and contribute to enhanced plant resilience 67 under stressful conditions. Several studies in model plants and crops such as tomato, maize, 68 and grapevine among others, have demonstrated their potential in alleviating the detrimental 69 effects of salinity, water deficit or high temperature by improving physiological performance 70 and reducing oxidative damage7,8. Accordingly, a comprehensive meta-analysis of 180 studies 71 worldwide on biostimulants has classified plant extracts as the category with the best 72 performance in terms of crop yield response9. 73 Despite this potential, plant responses to biostimulants, are often highly variable and context-74 dependent. This variability reflects several sources of complexity. Their chemical composition 75 is intrinsically heterogeneous, depending on the botanical species used, its phenological stage, 76 the type of tissue collected and the extraction process 8. Application practices further add to 77 this variability, since plant phenology, dosage, frequency, and delivery mode (foliar versus 78 root) can all lead to contrasting plant responses 9, especially when extracts are applied either 79 as a priming strategy before stress , or as a mitigating treatment during stress. Plant 80 responsiveness itself represents another layer of variability: effects are rarely uniform across 81 species10 and within a single crop, differences among cultivars or rootstocks in basal stress 82 tolerance and physiological plasticity strongly influence outcomes11,12. Beyond these biological 83 and technical factors, the environmental and climatic context exerts a dominant effect11,13, yet 84 our understanding of the precise windows of efficacy of the biostimulant remains limited. Even 85 when multi-year data are available, variability is often reduced to a description of the conditions 86 under which plants developed, without fully characterizing the intensity, duration, or dynamics 87 of the stress to which they were exposed. As a result, integration of stress characterization into 88 experimental designs is still rare, hindering our ability to link biostimulant responses to specific 89 stress scenarios and to decipher the mechanisms underlying their action. 90 In this context, the present study aimed to investigate the effects of a complex biostimulant, 91 composed of multiple plant -derived fractions, on grapevine responses to drought, heat, and 92 their combination. To capture the inherent variability of field conditions, experiments were 93 conducted under semi-controlled conditions over two successive years that differed markedly 94 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 4 in their weather conditions, particularly in terms of temperature, which resulted in contrasting 95 intensities of water deficit. Nevertheless, through systematic monitoring of plant water status 96 in the differe nt combinations of stress , we characterized a continuum of drought intensit y 97 across experiments, providing a robust framework to evaluate treatment effects. This approach 98 not only enabled us to demonstrate that the biostimulant reduces the susceptibility t o water, 99 but also delineate a window of efficacy under semi -realistic conditions. Furthermore, 100 transcriptomic and metabolomic characterizations revealed a priming effect of the treatment, 101 notably through the modulation of the phenylpropanoid pathway in a t issue-specific manner, 102 leading to divergent responses in roots and leaves. These molecular and metabolic 103 adjustments suggest a trade -off between organ growth and stress preparation, ultimately 104 contributing to improved tolerance to water deficit. Together, these findings provide new 105 insights into the mechanisms of action of a biostimulant in perennial crops and highlight the 106 importance of integrating stress characterization into the evaluation of biostimulant efficacy. 107 108 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 5

109 Over two independent experiments performed during 2021 and 2022 seasons, a solid form of 110 a plant -based biostimulant was applied to the root system of grapevine plants at leaf 111 senescence in late fall, while the liquid form of the product was sprayed on shoot three times, 112 when leaves are fully expanded, at floral bud emergence, and at the fruit set stages. Then, 113 control and treated plants were exposed to control conditions (no stress, NS), Heat (H), 114 Drought (D) and combined stress (H:D) under semi-controlled conditions (Fig.1). 115 Figure 1: Procedures and stress monitoring for grapevine experiments conducted in 2021 and 116 2022. Grapevines plants were propagated and kept outdoor. Solid biostimulant treatments were applied 117 before leaf senescence at fall. After reappearance of leaves during spring, three liquid treatments of 118 roots were applied, when leaves are fully expanded, at floral bud emergence, and at the fruit set stages. 119 During summer, plants were transferred to the greenhouse (15:9 h light: dark), under semi -controlled 120 conditions for 29-35 days: No Stress condition (Grey, cooling system/Well watered), Heat stress (yellow, 121 no cooling system/Well watered), Drought stress (blue, cooling/no watering) and Drought combined with 122 Heat stress (orange, no cooling/no watering). Pl ant stress level was monitored through Ψ PD (n=4 ±se). 123 Stress threshold was established between -0.5 and -0.6 Mpa. Green and red dashed lines indicate 124 averaged temperatures in the two different greenhouses with or without temperature cooling system 125 leading to no thermic stress condition (NS and D conditions) and thermic stress condition (H and H:D 126 conditions), respectively. 127 128 129 130 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 6 Treated plants show moderate shoot and root growth reduction under mild water 131 deficit but preserve root biomass under severe stress. 132 Without considering treatment effects, u nder NS conditions and H stress, predawn water 133 potential (ΨPD) was maintained at a constant level (-0.152 MPa ±0.01) for both years, whereas 134 it decreased over the time under drought conditions. Especially under combined stress, t hat 135 decrease was much faster in 2022 (up to -1.6 MPa ±-0.18) compared with 2021 (up to -0.53 136 MPa ±-0.09). Reported temperature in cooled (NS and D conditions) vs no cooled (H and H:D 137 conditions) greenhouses showed longer and higher differences in 2022 compared with 2021 138 year. Overall, such temperature differences, together with lower ΨPD, indicates that plants from 139 drought modalities in 2022 suffered a more intense drought stress compared with 2021 140 respective plants (Fig.1). 141 Physiological parameters were evaluated at the end of stress application. According to 142 ANOVAs, biostimulant treatments significantly impacted shoot length, number of internodes 143 and root weight in both years (Fig.2a). Globally, number of internodes and shoot length were 144 slightly reduced under biostimulant treatment compared with control conditions and 145 independently of drought or heat stress effects. Total leaf area (evaluated in 2022 ), and 146 chlorophyll contents were not significantly affected by the biostimulant treatments. Leaf area 147 was rather stable under NS, D and H stress conditions, but a significant drop occurred under 148 H:D condition. On the other hand, no change of chlorophyll contents was reported in 2021, 149 whereas heated modalities of 2022 (H and H:D) showed higher accumulation. Importantly, we 150 observed a slight but constitutive significant decrease of root biomass in treated plant in 2021, 151 independently of stress modalities. This is illustrated with highly significant Treatment (T) effect 152 according to ANOVA test s. By contrast , root biomass accumulation showed a differential 153 behaviour in 2022, with the appearance of interaction effects between Treatment and Stress 154 variables (TxS). Indeed, in 2022, root biomass was significantly reduced in treated plants under 155 NS condition but remain then stable and similar as control plants under single H and D stress. 156 Nevertheless, the combined stress H:D of 2022, previously identified as the most intense water 157 stress conditions of the experiment, leaded to significant and differential drops of root biomass 158 of 63 % for control plants and only 46 % for treated plants, when compared with other stress 159 conditions (Fig.2a). 160 Picture analysis of cleaned roots, generated 29 descriptive features of root architecture, used 161 for performing PCAs. Most differences in root architectures were observed by comparing NS 162 with H:D stress, as shown in Figure 2b. Mild discrimination of different modalities occurred in 163 2021 through PC2 (16.1 %), whereas clearer separation driven by PC1 (49.2 %) occurred in 164 2022. In the latter, t reated roots were architecturally identical under NS control, but were 165 slightly different from control conditions under combined stress conditions , despite a small 166 overlap of the conditions. This difference is partially driven by PC2 (10.2 %), for which the most 167 contributing variables are “Center of mass” (17.8 %) and “Length distribution” (16.8 %; Fig.S1). 168 As an integrator of plant physiology, visible impact of treatments and stresses was evaluated 169 on the plants (Fig. S2 and Fig2c). We found a significant increase in the number of plants in 170 the biostimulant modality without visible stress in H:D conditions compared to the control, in 171 2021 and in 2022, as well as a reduction or even an absence of plants presenting moderate or 172 high stress. 173 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 7 174 Figure 2: Morphological parameters of Biostimulant (blue) and Control (grey) plants under Heat 175 (H), Drought ( D) and combined stress (H:D) during two consecutive years. a) Number of 176 internodes, shoot length, total leaf area, root weight and leaf chlorophyll. Value represent mean ± se 177 (n=10). Significant differences are shown according to two-way ANOVA (**: p<0.01; *: p<0.05; .: p<0.1) 178 where T indicates biostimulant treatments effect (2 modalities: Biostimulant and Control), S stress effect 179 (4 modalites: NS, H, D, HD) and TxS, interaction effect. b) Principal component analysis (PCA) plots of 180 root architectural para meters of Biostimulant and Control plants under no stress and combined stress 181 conditions (H:D) in 2021 and 2022. PCA was performed from the correlation matrix generated with a 182 total of 29 variables (n=5). Plots on the right show the most contributing param eters to the first and 183 second dimensions of PCAs obtained each year. c) Distribution of visual stress groups under control, 184 thermic, drought and combined stress conditions over two years, after 30 -35 days of stress. Group 185 Percentage represent At least 10 observations per modalities. Treatment effects on group repartitions 186 were tested with Fisher's test with 5% threshold. Significant difference between treatments is indicated 187 with an asterisk (pValue<0.05). Illustrative pictures for different stress levels are shown in Figure S2. 188 189 A panel/Broad range of water stress levels reveals that the biostimulant induces lower 190 susceptibility to water stress. 191 To better understand if softening water stress pressure takes part to the biostimulant 192 mechanisms, plant water status was monitored with leaf stomatal conductance (gs), ΨPD, and 193 midday water potential (Ψl) measurements, over the duration of imposed conditions (Fig.3). gs 194 showed two distinct patterns over the two years (Fig.3a). First, for both years and according to 195 ANOVAs, treatment and time had no significant impact on unstressed plants (NS) and heated 196 plants. However, gs remained consistently high for H plants. Secondly, treatment effects on gs 197 were detected under H:D plants in 2021 and as a Time-Treatment interaction (TxD) under D 198 stress in 2022. In both cases, stomatal aperture of treated plant was maintained opened longer 199 than the control, with significantly higher gs at 15 and 26 days after start of the stress in 2021 200 and 2022 respectively (fig.3a) 201 Although not significant, it is worth noticing that in 2021, under D condition, the gs response of 202 treated plants tended to shift (pT=0.77; pD=0.05; pTxD=0.21)) and raised above that of the 203 2022 ba 2021 2022 Number of internodesTotal leaf area Root weight 0 5 10 15 20 25 0 10 20 30 Biostimulant Control T ** S * TxS ns Shoot length T * S ** TxS ns T ** S ** TxS ns T ns S ** TxS n T ** S ns TxS ns T ns S ** TxS ** Chlorophyll NS H D H:D NS H D H:D 0 5 10 15 20 T ns S ns TxS ns T ns S ** TxS ns NA NA (g)(µg.gFW-1) * 0 25 50 75 5 10 15 20 0 * * * * * (cm)(cm2) 0% 20% 40% 60% 80% 100% Control Biostimulant NS H D H:D NS H D H:D 2021 2022 Dead High stress Moderate stress Low stress No visible stress Control Biostimulant Control Biostimulant Control Biostimulant Control Biostimulant Control Biostimulant Control Biostimulant Control Biostimulant * *c 2021 -5.0 -2.5 0 2.5 -10 -5 0 5 Dim1 (45.1%) Dim2 (16.1%) -4 -2 0 2 -4 0 4 8 Dim1 (49.5%) Dim2 (10.2%) NS Biostimulant NS Control H:D Biostimulant H:D Control -8 Plant proportion .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 8 controls from 15 to 26 days after start of the stress . This trend aligns with a brief but higher 204 significant ΨPD of treated compared with control plants, observed at 15 days after start of the 205 stress. Finally, under the most severe combined heat and drought stress applied in 2022, we 206 observed that the first time point still showed a significantly higher gs in treated plants, before 207 to eventually stabilize at similar low levels in both treated and control plants. 208 For Ψl parameter and according to ANOVAs, control and treated plants followed quiet similar 209 pattern during the different stress applications (Fig.3b). However, s ignificant impact of the 210 biostimulant treatment on ΨPD were detected under combined stress H:D in 2021 and D stress 211 of 2022 (Fig.3c). For these conditions, treated plants exhibited slower decrease of ΨPD over 212 the time when compared with control conditions, leading to significantly higher potentials at the 213 end of the kinetics. It is worth noticing that distinguished potential between control and treated 214 plants start to be visible when the potential of control plants decreases to around -0.4 MPa. 215 Similarly, the lower ΨPD of control plants, significantly compensated by the treatment was -1.25 216 MPa (±1.17), observed at the end of D stress of 2022. Importantly, stomatal regulation was not 217 driving differences observed in gs, since the relationship between gs and ΨPD showed no 218 differential behaviour between control and treated plants (Fig.3d). Moreover, density plots 219 show a significant increase of biostimulated plants with gs ranging from 100-350 mmol.m-².s-1, 220 which correlates with an increased proportion of biostimulated plants with modest but higher 221 ΨPD (from -0.40 to 0 MPa). 222 c 20212022 0 10 20 30 0 10 20 30 0 10 20 30 0 10 20 30 -0.75 -0.50 -0.25 0.00 -1.5 -1.0 -0.5 0.0 Days after start of the stress ΨPD (MPa ) ΨPD * * T ns D ns TxD ns T ns D ns TxD ns T ns D * TxD ns T * D ** TxD ns T ns D ** TxD ns T * D ** TxD ns T ns D * TxD ns T ns D * TxD ns . Biostimulant Control No Stress Heat Drought Heat:Drought * * a gs Biostimulant Control b No Stress Heat Drought Heat:Drought 2021 0 10 20 30 0 10 20 30 0 10 20 30 0 10 20 30 -1.5 -1.0 -0.5 0 0 2022 Days after start of stress Ψl (MPa) -2.0 -1.5 -1.0 -0.5 * * T ns D ns TxD ns T ns D ns TxD ns T ns D ns TxD ns T ns D * TxD ns T ns D ** TxD ns T ns D * TxD ns T ns D ns T xD ns T ns D ns T xD ns Ψl Biostimulant Control No Stress Heat Drought Heat:Drought 20212022 0 10 20 30 0 10 20 30 0 10 20 30 0 10 20 30 0 200 400 600 Days after start of stress gs (mmol.m -².s-1) * * * *T ns D ns TxD ns T ns D ns TxD ns T ns D ns TxD ** T ns D * TxD ns T * D ** TxD ns T ns D ** TxD ns T ns D ns TxD ns T ns D ns TxD ns 0 200 400 600 * d Biostimulant Control 0 100 200 300 400 500 600 700 800 ΨPD (MPa) gS (mmol.m-².s-1) -2 -1 -0.5-1.5 * 350-700100-3500-100 0 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 9 Figure 3: Leaf stomatal conductance and water status of biostimulant treated and control plants 223 under Heat:Drought stress and visual stress scores during two consecutive years. Time-course 224 evolution of stomatal conductance a) g s, b) Midday leaf water potential ( Ψl) and c) Predawn leaf water 225 potential (ΨPD) of control (grey) and biostimulant (blue) treatment throughout stress applications. Value 226 represent mean ± se (n=3 -10). Significant differences are shown according to two -way ANOVA 227 (**:p<0.01; *:p<0.05) where T indicates treatment effect, D days effect and TxD, inte raction effect. d) 228 Relationship between ΨPD and gs under control (grey, n=96) and biostimulant (blue, n=95) treatment. 229 Density plots show the repartition of plant ΨPD and g s on x and y -axis, using all the measurements 230 performed during the stress applicati on in both 2021 and 2022 seasons pooling together the 4 stress 231 modalities (NS, H, D and H:D). Treatment effects on group repartitions (ranges are indicated in grey of 232 y-axis) were tested with Fisher's test with 5% threshold. Significant difference between treatments is 233 indicated with an asterisk (pValue<0.05). 234 235 236 Integration of a continuum of stomatal conductance responses for gene expression 237 analysis reveals an interaction between water availability and biostimulant effects. 238 To understand how treated plants physiologically respond to combined thermic and drought 239 stress, the expression profiles of 89 defence and stress-related genes were determined in 240 leaves and roots in the two years. Genes were arranged into 8 categories, namely, signalling, 241 cell wall synthesis, hormones, homeostasis, redox metabolism, growth and primary 242 metabolism, secondary metabolism, a nd defence proteins; giving an overview of the main 243 known regulated pathways of grapevine when exposed to heat and drought (Supplementary 244 table S1). Performing a principal component analysis (PCA) with this dataset shows the first 245 principal component (PC1) accounted for 30.2 % of the total variance and underlines the 246 impact of year of experimentation on gene profiles. Given the strong differences in the stress 247 intensities applied to the plants in the two different years, the impact of the treatment was 248 hardly distinguished when year effect was considered as a single binomial explanatory variable 249 (year 2021 or year 2022, Fig.4a). Actually, when testing the effects of the different modality 250 combinations on the gene expression data with ANOVAs, most of the variance is captured by 251 complex interactions of the year together with other modalities. Instead, transcriptional data 252 from both years has to be integrated and compared through a quantitative variable, reflecting 253 the level of water status at which plants were exposed independently of the year. To do so, the 254 area under the curves (AUC) of stomatal conductance ( gs) parameters was calculated for a 255 total of 16 different AUCs, one for each combination of stress modalities, treatment and year. 256 Among the different measured parameters of plant water status, gs was considered here as 257 the best indicator of drought stress intensity because i) levels and profiles of ΨPD and Ψl is 258 unchanged under NS and H stress modalities in both years , making impossible their 259 differentiation using those parameters. On the other hand, gs conductance indicates a higher 260 stomatal conductance when compared with all other conditions, revealing a different water 261 status of H stress plants. ii) The link between s tomatal conductance and ΨPD is unchanged 262 with the treatment. Therefore, the estimation of the stress intensity through the AUCgs will not 263 be biased by a treatment-related differential behaviour of stomata regulation and still depends 264 of the plant water status. Then, PCA using AUC gs levels for colorations of individuals shows 265 the segregation of plants based on their gradual levels of water status, driven mostly by PC1 266 (Fig.4b). 267 Linear models were built for testing the importance of Heat, AUCgs and treatment effects on 268 the different genes in leaves and roots as well. In total, 64 genes in leaves and 28 genes in 269 roots were significantly affected by at least one of the previous single factors or combination 270 of them (p<0.05). More precisely, 47 genes in leaves and 1 7 in roots were exclusively heat 271 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 10 and/or AUCgs-related genes with no effect of the tre atment (Fig. S4). In l eaves, most of 272 differentially expressed genes were associated with water stress, sometimes with an additional 273 or combined effect of temperature (Fig. S4a). Although fewer genes were differentially 274 expressed in roots overall, temperature , alone or interacting with water d eficit, had a 275 proportionally greater impact compared with leaves (Fig.S4b). For instance, only VvLOX3, and 276 VvPGIP involved in jasmonate biosynthesis and a polygalacturonase inhibiting protein 277 respectively, were exclusively affected by high temperature. Globally, gene expression was 278 negatively correlated with the AUCgs, meaning that lower water status (lower AUC) led to higher 279 expression levels of target genes. Consistent with this pattern, VvPIP2 and VvTIP2.1, involved 280 with hydraulic adjustment s, as well as VvSnRK2.6, VvDREB, VvWRKY30 , considered as 281 central genes of dependent and independent abscisic acid (ABA) signalling, constitute a strong 282 molecular signature of adaptation to water deficit. Transcription factors such as VvABF1 (bZIP 283 family) responded notably to the AUCgs:Heat interaction, particularly in roots. And o ther 284 hormone-related genes were upregulated, including those linked to ethylene signalling (e.g., 285 ERFs) or auxin pathways (VvPIN and VvLax2l). Notably, VvLOX3 in leaves and root appeared 286 specifically linked to the temperature, suggesting a more distinct role for jasmonate signalling 287 in the grapevine heat response. 288 Concerning the impact of the biostimulant , 21 genes were transcriptionally affected by the 289 treatment or a combination effect of treatment with water stress and/or heat effect (Fig.4d). 290 VvLDOX, a gene involved into flavonoid biosynthesis , was the single gene with exclusive 291 treatment effect, as it was constitutively induced in roots in treated plants, when compared with 292 control. 293 In leaves, VvPOX, VvPR8 and VvPR2, all considered as pathogenesis related (PR) genes, 294 showed an interaction effect of the treatment and heat stress. Under control conditions, their 295 relative expression remained low and poorly affected by the temperature whereas the 296 biostimulation triggered significantly higher expression with high temperatures. Fifteen genes 297 were affe cted by the interaction AUCgs:Treatment. Independently of the tissue type, those 298 genes exhibited quiet similar patterns : under control conditions, their expression was rather 299 stable or even increasing with stress intensities, while biostimulation leads to higher expression 300 in NS to mild stress condition, and lower level of expression under hight D stress (low AUCgs). 301 In this way, the treatment significantly affected major stress signalling genes across different 302 water status, all closely linked to ABA and stomatal regulation, namely VvWRKY13, VvNAC17 303 and VvABF2-1 in leaves, and VvSNRK2.6 and VvPP2C4 in roots. Interestingly, VvBES1 and 304 VvDREB were also influenced, the former being tightly associated with brassinosteroid 305 signaling and growth regulation 14, and the latter representing a central regulator of ABA -306 independent abiotic stress responses 15. VvRafS1, known to be involved to raffinose 307 biosynthesis, a trisaccharide largely involved in abiotic stress alleviation as osmoprotector or 308 modulator of redox power among others 16, was also affected in leaves with AUCgs:Treatment 309 interaction. Concerning secondary metabolism, regulation and biosynthesis of flavonoid was 310 significantly affected, with VvCHS2, VvCHI and VvROMT in leaves, and VvCHORM2, VvCHI2 311 and VvLDOX in root. Finally, VvPOD1.1, a gene belonging to a peroxidase family involved in 312 redox metabolism and cellular H₂O₂ detoxification17, exhibited a distinct expression pattern in 313 roots, in response to varying levels of drought stress, but only under mild temperature 314 conditions. Notably, its expression was reduced in treated plants experiencing D stress 315 (Fig.4d). 316 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 11 317 Figure 4: Interaction of biostimulant, heat and AUCg s effect on the expression level of target 318 genes in a two -consecutive year experiment. PCAs (a and b) were performed from the correlation 319 matrix generated with a total of 89 selected gene expressions expressed in root and leaves. Genes were 320 selected according to three -way ANOVA, to evaluate the individual and combined effects of 321 Biostimulant, Heat, and Drought (p<0.05). For Drought effect, AUC (area under the curve) of gs of each 322 modalities was used as a quantitative factor for testing the level of stress that the plants experienced 323 during the trials. Plot (a) shows discrimination by year of experiment and plot (b) shows discrimination 324 by AUC of g S levels. c) Venn diagrams showing the number of genes in leaves or root, significantly 325 ba Biostimulant - NS Biostimulant - Heat Control – NS Control - Heat Heat:AUCgs:Treatment Treatment d AUCgs:Treatment n=24 n=12 n=24 Dim2 (15.7%)-10 0 10 -10 0 10 20Dim1 (30.2%) 2021 2022 20 -20 -20 0 20 -10 0 10 20 30 Dim 1 Dim 2 AUC 5000 10000 PCA - Coloration par AUC AUCgs 1000 500 -20 0 20Dim2 (15.7%) 10 -10 -10 0 10 20 Dim1 (30.2%) -2 0 2 4 0 5 10 VvLDOXRoots NS T 2 4 6 8 0 5 10 -6 -4 -2 0 2 VvPOD1.1 Roots NS T 2 4 6 8 0 5 10 -1 0 1 VvPDV1 Roots -1.0 -0.5 0.0 0.5 VvPR8Leaves -1.5 -1.0 -0.5 0.0 0.5 VvPOXLeav es -1.5 -1.0 -0.5 0.0 VvPR2Leaves Biostimulant Control Biostimulant Control Biostimulant Control Heat:Treatment c AUCgs:TreatmentHeat:Treatment Treatment Heat :AUCgs:Treatment Log2(Relative expression) Log2(Relative expression) Log2(Relative expression) Log2(Relative expression) -5 0 5 0 5 10 VvNAC17Leav es -10 -5 0 0 5 10 VvWRKY13_2 Leav es -6 -4 -2 0 2 0 5 10 VvRafS1 Leav es -5.0 -2.5 0.0 2.5 5.0 0 5 10 VvCHILeav es -2 0 2 0 5 10 VvCHS2Leav es -2.5 0.0 2.5 5.0 0 5 10 VvROMTLeav es -4 0 4 8 0 5 10 VvABF2_1 Leaves -10 0 10 0 5 10 VvBES1_N2 Leaves -4 0 4 0 5 10 VvATPAt Leav es -10 -5 0 5 0 5 10 VvSnRK2.6 Roots -2 -1 0 1 2 3 0 5 10 VvMDHRoots -2 -1 0 1 2 3 0 5 10 VvPP2C4 Roots -4 -2 0 2 0 5 10 VvDREBRoots -6 -4 -2 0 2 0 5 10 VvCHORM2Roots -4 -2 0 2 4 0 5 10 VvCHI2Roots Log2(Relative expression) Log2(Relative expression) Log2(Relative expression) Log2(Relative expression) Log2(Relative expression) AUCgs AUCgs AUCgs AUCgs AUCgs AUCgs .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 12 affected by the treatment, individually or combined with Heat or Drought factors. d) Relative expression 326 profiles of selected genes (p<0.05) significantly affected by the Treatment factor, individually (Treatment, 327 orange dashed square) or combined with heat (Heat:Treatment, green dashed square), drought stress 328 (Drought:Treatment, blue dashed square) or both (Heat:Drought:Treatment, brown dashed square). For 329 Treatment and Drought:Treatment affected genes, plots show treated and control modalities over AUC 330 of gS. For Heat:Treatment affected genes, combined modalities of heat stress and treatment are shown 331 in a barplot . For Heat:Drought:Treatment affected genes, combined modalities of heat stress and 332 treatment are shown over AUC of GSW in a dotplot. 333 334 Metabolic characterisation displays tissue-specific adaptation of biostimulated plants 335 under combined H:D stress associated with differential flavonoid accumulations 336 Transcriptomic data highlighted a treatment -related adaptation of redox and secondary 337 metabolism among others. For this reason, total antioxidant activity assays together with total 338 flavonoid an d polyphenol quantifications were conducted on leaf and root tissues across 339 treatments, stress levels, and both years (Fig.5). All three parameters exhibited similar pattern 340 over different tissues and modalities, which highlights the significant contributi on of those 341 metabolite families to the antioxidant power of grapevine tissues. However, marked differences 342 were observed between the two years. In 2022, leaves exhibited overall lower levels of 343 reducing power and polyphenols, while roots showed higher values compared to 2021. 344 Notably, ANOVA s revealed significant Treatment effects in leaves of 2021 , with a slight 345 decrease of antioxidant power and polyphenol content under biostimulation conditions . No 346 significant changes of leaf Flavonoid contents were detected in leaves. However, in roots, 347 Drought:Treatment interactions were detected in 2021 for FRAP values (near from 348 significance, Pvalue=0.068), and total polyphenol and flavonoid contents (Fig.5). Such 349 interaction effects are explained by a significant increase of all three parameters in treated 350 plant under no stress conditions when compared with the control, while their levels remained 351 unchanged when single or combined stress were applied. 352 2021 2022 0.0 FRAP 0.0 0.1 0.2 Polyphenols 0.1 0.2 0.3 0.4 Flavonoids 0 1 2 3 4 2021 2022 FRAP 0 0.03 0.06 0.09 0.0 Polyphenols 0.1 0.2 0.3 Flavonoids 0 1 2 3 NS H D H:D NS H D H:D T * S ns T*S ns T ns S ** T*S ns a NS H D H:D NS H D H:D T ns S ns T*S * T ns S ns T*S ns T ns S ns T*S ns T ns S * T*S ns T ns S ns T*S * T ns S ns T*S ns b * * * g equiv gallic acid/g FW-1 mg Equiv catechin/g FW -1 mg Ascorbic a Equiv g FW-1 * T * S * T*S ns T . S ** T*S ns T ns S ns T*S . T ns S . T*S ns Leaves Roots * * Biostimulant Control .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 13 Figure 5: Total polyphenols and Flavo noid content and free radical scavenging activities of 353 treated (blue) and control (grey) plants under Heat (H), Drought (D) and combined stress (H:D) 354 during two consecutive years. a) Leaves and b) roots. Flavonoid, polyphenols contents and FRAP, 355 DPPH scavenging activities. Value represent mean ± se (n=6). Global effect are shown according to 356 two-way ANOVA (**:p<0.01; *:p<0.05; .:p<0.1) where T indicates treatment effect, S stress effect and 357 TxS, interaction effect. Significant differences between contro l and treated plants are highlighted with 358 asterisks (* p<0.05). 359 360 Most contrasting effects between treated and control plants regarding antioxidant metabolites 361 occurred under non-stress conditions. In addition, physiological differences were statistically 362 detectable under mild stress, particularly for water-related parameters. For those reasons, the 363 no-stress and combined heat –drought (H:D) conditions in 2021 were selected for further 364 metabolic characterization. Metabolic profiling was performed through a LC-MS analysis 365 pipeline, providing relative abundance information of 3601 metabolites. In leaves and root s, 366 700 and 1363 metabolites were respectively detected, as significantly affected by treatment, 367 Heat or Drough t, (ANOVA, p<0.05). This indicated a good robustness of the metabolomic 368 dataset with a clear phytochemical diversity shift . Then, unsupervised PCAs and a heatmap 369 analysis (Fig. 6a and 6b) were performed to evaluate overall differences in metabolic profiles 370 of both organs. 371 In leaves, the first principal components (PC1) accounted for 48.8% of the total variance and 372 clearly separated stressed from non-stressed control plants (Fig.6a). In contrast, treated plants 373 displayed highly similar metabolic profiles regardless of stress conditions. Notably, the treated 374 group clustered more closely with the stressed controls along PC1, but were distinguished 375 from them along PC2 (13.8% of total variance) . Clustering analysis indicates that such 376 difference between non -stressed control s and biostimulated plants is largely driven by a 377 pronounced reduction in the abundance of numerous metabolites (cluster 1 and 2, from left to 378 right; Fig. S 4a). To highlight treatment -related metabolites, pairwise comparison was 379 performed between treated vs control plants from NS and HD modalities. Under NS conditions, 380 biostimulant treatment induced the significant accumulation of 6 metabolites and the depletion 381 of 191 metabolites, while only 6 features were significantly depleted under HD conditions 382 (Fig.6c). Subsequent analysis of the chemical groups of down-accumulated compounds under 383 non-stress (NS) conditions revealed that a substantial proportion belonged to antioxidant -384 related classes. A large fraction was assigned to flavonoids, including intermediary compounds 385 and derivatives, such as epicatechin (ID:4973), cianidanol (ID: 1968), rutin (ID: 4678), 386 procyanidin B1 (ID: 5482), B2 (ID: 5480) and C1 (ID:4887). In addition, coumarin derivatives 387 such as columbianetin (ID: 4911) was identified, together with trihydroxy benzoisochromen 388 (ID: 2102), an anthraquinone derivative formed from oxidized coumarins. Moreover, tannins 389 such as gallic acid hexoside (ID: 5630) and vulpinic acid (ID: 4241), belonging to furanones, 390 were also detected (Fig.6e). 391 In roots, the first two principal components (PC1 and PC2), which together explained 53.8 % 392 of the total variance, clearly segregated the treated from control plants under non -stress 393 conditions. Under combined stress, roots showed clear different patterns distinguished from 394 no stress profile. However, biostimulant treatment had no visible impact. Hierarchical clustering 395 showed that treated plant under no stress condition exhibited a marked accumulation of many 396 compounds in the second cluster (from left to right) when compared with all other modalities 397 (Fig. S4B). This cluster was particularly enriched with flavonoid family compounds or derivates 398 (17 %). Overall, 48 features were identified as differentially accumulated, when comparing 399 control and treated plant under no stress conditions (Fig.6d), and 10 of them were classified 400 as flavonoids compounds, derivates or biosynthesis intermediary. Notably, putative flavonoids 401 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 14 of particular interest include procyanidin B2 and C1, cianidanol, prunin (ID:3273) or demethyl 402 medicarpin (ID:1353). In addition, other putative polyphenols, such as gallic acid hexoside and 403 the stilbene pinosylvin (ID: 1017) , were also differencially accumulated, along with 404 columbianetin and trihydroxy benzoisochromen. Several metabolites of interest for their red-405 ox buffering role, including cianidanol, procyanidin B1, columbianetin, and gallic acid hexoside, 406 were detected in both leaves and roots (Fig. 6e and f), and illustrate the depletion of antioxidant 407 potential in leaves and activation of redox-related pathways in roots, highlighting organ-specific 408 modulation of the antioxidant metabolism in response to biostimulant treatment under non -409 stress conditions. 410 Finally, as major regulators of growth -defence trade-offs and given the previous detection of 411 differentially expresse d genes involved in phytohormone biosynthesis pathways, relative 412 abundances of putative salicylic acid, ABA (ID:1410) and jasmonic acid (ID:1012) were 413 specifically examined in the present dataset (Fig. S7). In leaves, the treatment under non -414 stress conditions triggered a marked increase in salicylic acid levels and a reduction in ABA. 415 On another hand, in roots, treatment had no effect under NS condition but showed lower levels 416 under combined stress conditions, with a significant reduction of putative salicylic acid forms 417 (ID:464). Finally, in both organs, jasmonic acid was no affected by the treatment despite a 418 marked decrease of abundance in roots under stress. 419 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 15 420 Figure 6: Metabolic profiling analysis of roo ts and leaves treated with biostimulant under No 421 Stress (NS) and combined Heat:Drought (H:D) conditions in 2021 . PCAs were performed from the 422 correlation matrix generated with a total of 1364 metabolites in leaves (a) and 956 metabolites in roots 423 (b), selected according to two-way ANOVA, to evaluate the individual and combined effects of treatment 424 and H:D stress. Volcano plots of metabolic markers that were accumulated (red) or depleted (blue) 425 between biostimulated vs control plants under NS conditions and H:D conditions for Leaves (c) and 426 Roots samples (d). The criteria used for metabolite abundance filtering were α ≤ 0.05, p ≤0.01 and fold 427 change ≥1.5 for upregulated (Red) and ≤0.6 for downregulated (Blue ). Sunburst plot of the global 428 Roots -30 -20 -10 0 10 20 -25 0 25 Dim1 (39.2%) Dim2 (14.6%) -20 0 20 40 -50 -25 0 25 50 Dim1 (48.8%) Dim2 (13.8%) H:D Biostimulant H:D Control NS Biostimulant NS Control Leavesa g b h 0 100000 200000 300000 400000 Columbianetin Cianidanol Procyanidin B2 Gallic Acid hexoside Ave peak area g FW-1 0 100000 200000 300000 400000 500000 Columbianetin Cianidanol Procyanidin B2 Gallic Acid hexoside 800000 Ave peak area g FW-1 H:D BiostimulantH:D ControlNS BiostimulantNS Control 0 1 2 3 -2 0 2 Log2 Fold Change -Log10 P-value H:D Biostimulant vs H:D Control Depleted Accumulated H:D Biostimulant H:D Control NS Biostimulant NS Control 0 1 2 3 4 -4 -2 0 2 4 Log2 Fold Change -Log10 P-value NS Biostimulant vs NS Control 0 1 2 3 -2.5 0.0 2.5 5.0 Log2 Fold Change -Log10 P-value H:D Biostimulant vs H:D Control Depleted Accumulated c d e f Leaves Roots * * * * * * * * 0 1 2 3 4 -2.5 0.0 2.5 5.0 Log2 Fold Change -Log10 P-value NS Biostimulant vs NS Control .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 16 metabolome showing the classification (Superclass and Class) of all the putatively annotated 429 metabolites and unknown metabolites depleted under biostimulation in leaves (e) and accumulated 430 under biostimulation in roots (f). Refer to Supplementary Tables 1 for the full names and classification 431 of the metabolites shown in those sunburst plots. Differential quantitative profiles of some of the 432 representative metabolites within main classes of phenylpropanoid pathways impacted by the 433 biostimulant treatment (Tannins, flavonoids and coumarins) and commonly detected in leaves (g) and 434 roots (h). Significant differences between control and treated plants are highlighted with asterisks (* p≤ 435 0.05). 436 437

438 Plant-based extracts are promising solutions for mitigating abiotic s tresses such as salinity, 439 heat, and drought7. In this study, we investigated the mode of action of a biostimulant derived 440 from plant-based ethanolic extracts of 9 plant species. The biostimulant solid form was applied 441 to grapevine plants as a root treatment at leaf senescence before winter and the liquid form as 442 three foliar spray s at three different phenological stages prior to exposure to individual or 443 combined heat and drought stress (Fig.1). Our results revealed constitutive effects of the 444 treatment on grapevine physiology, independent of the imposed stress conditions. In particular, 445 treated plants displayed a slight reduction in the number of internodes, shoot length, and root 446 biomass (Fig.2). Although these observations may appear as negative growth effects, they did 447 not translate into any reduction in total leaf area, as tested in 2022. Moreover, chlorophyll 448 content and stomatal regulation remained similar between treated and control plant s across 449 stress conditions. This suggest that the photosynthetic and transpiration potential of 450 biostimulated plants appeared unaffected. 451 Importantly, in treated plant the simultaneous maintenance of higher ΨPD and stomatal 452 conductance (Fig.3d) indicates an improved ability to preserve plant water status and sustain 453 gas exchange under mild water-limited conditions. This suggests that the biostimulant tend to 454 hamper drought susceptibility, likely by supporting root water uptake and/or lower decrease of 455 ΨPD. In the current study , when mild water stress was applied (H:D 2021 and D 2022), we 456 observed the maintenance on average of -0.18 MPa of biostimulated plants when compared 457 to the control . Yet, previous meta-analyses have established a linear relationship between 458 water deficit (expressed as water potential Ψ) and reductions in berry size and yield. For this 459 reason, according to Williams and Phene (2010) and Mirás-Avalos and Intrigliolo (2017), such 460 decrease observed in this study, might on average prevent for loss of around 10% yield3. 461 Accounting for abiotic stress levels reveals a stress window with optimal biostimulant 462 effects. 463 Variation in the effectiveness of biostimulants is commonly reported, as their impact depends 464 not only on crop species and genotype but also on environmental conditions 9,20. Over two 465 successive years, monitoring leaf stomatal conductance and hydraulic parameters (gs, ΨPD 466 and Ψl, Fig.3) enabled to characterize the plant’s water status precisely across experiments, 467 and to consider a range of stress intensities for defining the window within which the 468 biostimulant was effective. This window first emerged under moderate drought in 2021, when 469 the effects of treatment on gs were close to significance. Then, gs and ΨPD remain significantly 470 higher under both H:D in 2021 and D in 2022. However, when the intensity of stress further 471 increased, as in the severe H:D conditions in 2022, the beneficial effects of the treatment was 472 not detectable , suggesting that the product is effective only within a restricted range of 473 moderate water limitation. It is worth noticing that, although the gs values were similarly low at 474 midday at the end of the stress period, the significantly higher ΨPD observed in treated plants 475 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 17 suggest that they could have maintained a better water status for a few more days if the stress 476 had continued. 477 Numerous studies reported a reduced susceptibility of grapevine when treated with 478 biostimulants21. Daler and Kaya (2024) showed that the effects of α-lipoic acid were mainly 479 detectable under drought and varied strongly between rootstocks, highlighting the combined 480 influence of stress intensity and genotype on biostimulant efficiency. Whey protein hydrolysate 481 was also showed to efficiently alleviate combined heat and water stress, by preventing 482 dramatic decrease of ΨPD to an average of -1.75 MPa experienced under control conditions 483 and maintained to -1.23 Mpa when treated22. More precisely, Frioni et al (2021) reported that 484 a seaweed extract efficiently maintained higher leaf transpiration and water use efficiency 485 when Ψl values were comprised between -0.6 and -0.8 MPa, while no effect of the treatment 486 could be detected under more severe water deficit. However, such studies were conducted in 487 climatic controlled chamber s and compared plant responses to the same progressive and 488 intensifying stress. Alternatively, some grapevine studies include multi-years or environment 489 dimension to their design, and year-to-year variability was acknowledged but rarely linked to 490 plant water status24,25. Evidence from tomato further reinforces this view, as multi -year open-491 field trials consistently reported that the effects of protein hydrolysa tes or other plant-based 492 biostimulant were strongly modulated by the growing season and irrigation regime11,13. In these 493 studies, biostimulants either enhanced productivity or improved fruit metabolic profil es 494 depending on the prevailing weather conditions, while in other years no significant effect was 495 detectable. Such findings underline that biostimulant performance cannot be generalized but 496 must be interpreted in the light of the abiotic stress intensity context. 497 Irrigated vineyards typically function within a safe range of water potentials, and non-irrigated 498 vineyards rarely reached ΨPD values inferior to −1.5 MPa, which do not lead to cavitation, 499 turgor loss and/or plant death26. Since the efficien cy of the biostimulant was observed to be 500 roughly in the range of -0.4 to -1.2 MPa, we can consider that the window of effectiveness of 501 the botanical extract falls within the physiologically relevant range of water potentials for 502 grapevine. By contrast, no impact of the biostimulant was detectable on grapevine physiology 503 when plants exhibited lower ΨPD values, close to the −1.5 MPa threshold, which represented 504 a situation rarely encountered in vineyard condition. This indicates that the biostimulant 505 remains efficient within a restricted range of water deficit, beyond which the physiological limits 506 of the plant override any potential benefits. 507 Biostimulant applications cope with a trade -off between growth and physiological 508 functions under stress through priming mechanism 509 A significant proportion of genes selected for transcrip tomic analysis followed similar and 510 consistent pattern, according to plant’s stomatal aperture, in leaves and to a lesser extent in 511 roots (Fig.S4). Typically, those genes, involved in water and heat stress management, are well 512 known (Table S1 for references ) and showed an increasing expression when AUC gs 513 decreased, which is considered as markers of water availability. 514 When combined with different stresses, biostimulant treatments caused a subset of key genes 515 to appear , which differed from other marker genes: higher expression under high water 516 abundancy but lower levels when water was lacking and AUCgs was reduced (Fig. 4). In this 517 way, VvNAC17 and VvWRK13, belonging to two families of transcription factors usually 518 considered as priming markers involved with drought adaptation4,27, were significantly affected 519 by the biostimulant. They were often observed with biostimulation-related priming in tomato28, 520 grapevine with seaweed extract 29,30 and cotton with γ -poly-glutamic acid enhancing drought 521 tolerance31. In line with these observations, several genes typically associated with stomatal 522 closure under drought stress were down-regulated in biostimulated plants. Notably, VvRafs2, 523 a raffinose synthase whose protein activity contributes to drought tolerance in maize and 524 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 18 Arabidopsis32 transcriptionally linked with ABA levels 33, and VvABF2, a bZIP transcription 525 factor implicated both in stomatal regulation and in enhancing cellular drought tolerance 34, 526 showed reduced expression under stress in treated plants. These transcriptional changes are 527 consistent with the lower ABA levels detected in leaves of treated plants ( Fig.6d) and may 528 explain the maintenance of stomatal aperture over the stress periods 35. In roots, no clear 529 differences in ABA levels were detected between treated and control plants (Fig. S7) . 530 Nevertheless, key regulators of ABA signalling such as VvPP2C and VvSnRK2 were affected 531 by the treatment. Notably, the up-regulation of VvSnRK2 under optimal conditions may prime 532 the root by activating downstream transcription factors 4, while the concomitant induction of 533 VvDREB highlights the reinforcement of both ABA -dependent and ABA -independent stress 534 pathways15. 535 When investigating potential priming effect at the metabolic level , non-stressed (NS) plants 536 cannot be considered strictly naïve . N evertheless, we can hypothesize that they only 537 experienced the treatment phase, and that they can be considered as a proxy of primed but 538 not triggered plants when compared with stressed conditions 27. Data indicates that leaves 539 developed a new metabolic profile, which was closer to the stressed state , yet still clearly 540 distinct, independently of cultivation conditions (Fig.6a). This shift was partly driven by the 541 depletion of putative phenolic compounds (Fig.5) including powerful antioxidant molecules 542 such as putative gallic acid s, coumarins and various putative flavonoids (Fig.6a). This is 543 consistent with both the reduced total flavonoid content measured in leaves and the modulation 544 of flavonoid biosynthesis genes such as VvCHI and VvCHS2 (Fig.5d). A comparable decline 545 in foliar flavonoids has also been observed in grapevines treated with kaolin 36. Flavonols are 546 known to exert profound effects on leaf development and physiology; for example, the absence 547 of flavonols in the tt4-1 (CHS) mutant of Arabidopsis results in smaller leaf area and slower 548 inflorescence growth 37. In line with this, we observed a constitutive and slight decrease in 549 certain biomass traits, such as internode number and shoot length, in treated plants, but no 550 reduction in leaf surface area was detected, at least during the 2022 season (Fig.2). Notably, 551 the observed depletion of flavonoids seems contradictory with their reported role in 552 antagonizing ABA-induced stomatal closure in both Arabidopsis and tomato38, a discrepancy 553 that warrants further investigation in grapevine. 554 By contra st, i n roots, the biostimulant induced a significant transcriptional and metabolic 555 modulation of the phenylpropanoid pathway , leading mainly to the accumulation of probable 556 flavonoids when no stress was applied ( Fig.5 and Fig.6b). S uch divergence in the root 557 metabolome was only observed under no-stress conditions. This may suggest that treated 558 roots undergo a preparatory adjustment prior to stress, which is later overridden once stress 559 is imposed. Accordingly, root metabolic profiles visualized through PCA ( Fig.6b) showed a 560 clear dissociation only under no stress condition driven by putative flavonoid and precursor 561 accumulations among others. Congruent with metabolic data, VvCHORM2, and VvCHI2, two 562 genes initiating the biosynthesis pathway of flavonoids through the conversion of ch orismate 563 to phenyalanine and p -coumaryl CoA to leucoanthocyanidin respectively , showed higher 564 expression with high AUC gs compared with control plants, and remain ed similarly lower with 565 the control under low AUC gs. Similarly, VvLDOX, which catal izes the conversion of 566 leucoanthocyanidin to epicatechin showed constitutive up-regulation when only exposed to the 567 biostimulant. Such priming-related reprogramming of flavonoids in absence of stress was 568 commonly reported in leaves of grapevine treated with seaweed extract 29,30 or oregano 569 essential oil vapours for instance 39. Enhanced flavonol accumulation has been previously 570 associated with improved drought tolerance, as demonstrated in Arabidopsis lines 571 overexpressing the transcription factors AtMYB12 and AtPAP140, as well as in a maize mutant 572 naturally enriched in flavonols41. In the Col-0 Arabidopsis background, flavonols modulate root 573 development, acting as negative regulators of root hair formation42, lateral root initiation43, and 574 root gravitropism through the modulation of auxin transport 44, most likely via scavenging of 575 reactive oxygen species. However, mutant characterization in other genetic backgrounds 576 revealed a broad spectrum of additional architectural phenotypes in both roots and shoots. In 577 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 19 this context, the differential accumulation of flavonoids observed in grapevine roots may have 578 reshaped root system architecture, leading to a moderate reduction in biomass, that remained 579 difficult to quantify in our study ( Fig.2b). Crucially, this potential architectural adjustment did 580 not appear to increase susceptibility to stress. On the contrary, it may reflect an adaptive trade-581 off, where a more compact yet metabolically buffered root system is compensated by 582 enhanced antioxidant capacity, as observed in treated plants, at least during the 2021 season 583 (Fig.5). Unfortunately, the approach used in this study to assess architectural changes did not 584 provide sufficient resolution to clearly conclude on root structural modifications (Fig.2b). For 585 this reason, further targeted studies focusing specifically on root systems will be require d to 586 determine whether such architectural changes, potentially driven by phenylpropanoid pathway 587 reprogramming, actively contribute to improved water uptake and, ultimately, enhanced 588 drought adaptation. 589

590 Overall, our findings reveal the effect of the biotimulant AXIOMA Vine on plant physiology 591 within a moderate field-realistic drought scenario, where treated plants maintained higher 592 stomatal conductance and ΨPD than controls. Under well -watered or extremely stressed 593 conditions, however, its effects were negligible. This context-specific efficiency aligns with the 594 view that biostimulant action is conditional on both environmental constraints and crop 595 physiology, and emphasizes the need for detailed monitoring of plant water status in evaluating 596 their effectiveness. Nevertheless, the study also reveals a potential trade-off, manifested as a 597 moderate reduction of some growth parameter , which all together, appears to confer the 598 substantial benefit of limited susceptibility to moderate water deficit conditions. Finally, analysis 599 of gene expression and metabolism highlight a high representation of ROS and stress 600 hormone-responses such as ABA, during the priming phase wit h the biostimulant. Given the 601 central role of these particular pathway , considered as critical when plants face multifactorial 602 stress1, raises the question of the efficiency of the biostimulant when plants are exposed to 603 other fluctuating weather events, such as flooding or cold stress. 604 605

606 Plant based biostimulant 607 The biostimulant AXIOMA Vigne (MA 1190734) is an assemblage of individual plant extracts, 608 all manufactured by Axioma Biologicals (Brive -la-Gaillarde, France). Extracts were obtained 609 by hydroalcoholic extraction (55,7 % v/v) at room temperature, from aerial parts of plants 610 belonging from families with potential biostimulant properties, namely Thuya occidentalis L., 611 Trigonella foenum -graecum L., Solanum dulcamara L., Capsicum annum L., Matricaria 612 chamomilla L., Euphrasia officinalis L., Laurus nobilis L. Sambucus nigra L., Urtica urens L. 613 and Equisetum arvense L45–51 and diluted with deionized water to reach 0.12% of dry weight. 614 NMR analysis detected sugars (sucrose, arabinose and fructose), organic acids (acetic acid, 615 malonic acid and formic acid), amino acids (glutamine and tyrosine), phenolic compounds, 616 choline, trimethylamine, an isopropylamine -type compounds as major components of the 617 extract and traces of SO3 (0.039 %), Fe (0.039 %), ZN (0.000022 %), Cu (0.0000047 %); Mn 618 (0.000012 %), Co (0.00025 %), B (0.039 %), Mo (0.00035 %), MgO (0.099%). In addition, for 619 solid application a concentrated liqu id biostimulant form (973 mL of hydroalcoholic plant 620 extracts and co-formulants per litre) was integrated into Lithothamnium calcareum of 2-4 mm 621 size containing 95 % of dry matter to a concentration of 50 mL of concentrated biostimulant 622 per kg of Lithothamne. 623 624 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 20 Plant material, growing conditions and treatments 625 The experiment was conducted in 2021 and 2022 following the same protocol. Grapevine 626 plants ( Vitis vinifera cv. Cabernet Sauvignon) were propagated in a greenhouse from 627 woodcuttings. After 35 days, rooted cuttings were transferred to 2.8 L pots of sand-soil mixture 628 (Klassman RHP 15 commercial potting mix with70 % fair peat of sphaine, 15 % cold black 629 peat, 15 % pearlite and Danish clay). The plants were watered by subirrigation and fertilised 630 twice a week (nutrient solution N/P/K 20/20/20). Eighty -two-month-old plants with 10 –12 631 leaves were placed outside and were divided into four randomized blocks. During this period, 632 plants were exposed to natural fall and winter conditions and were irrigated continuously with 633 drip irrigation. Biostimulant treatments were performed according to the manufacturer 634 instructions for field applications. At leaf senescence (about two weeks after the plants were 635 transferred outside), the solid treatment was applied to the root system of 40 plants by 636 depositing 1.5 g of the solid treatment on the surface of each pot (equivalent to 10 kg.ha -1). 637 Then, during the next growing season, the liquid treatment of an equivalent of 2 L.ha -1 was 638 applied with three successive foliar treatments at three phenological stages: at fully expanded 639 leaves (BBCH 11-19), at floral bud emergence (BBCH 57-60), and at the fruit set (BCCH 71) 640 stages. At the same time, 40 other plants received deionized water as a control. 641 Application of stress conditions 642 Two weeks after the last treatment (the 7 th of June in 2021 and 14 th of June in 2022), the 643 treated and untreated plants were transferred to greenhouses and watered to field capacity by 644 flooding them to their maximum capacity. The pots were then left to drain excess water for half 645 a day . Four environmental conditions were then imposed by combining two temperature 646 regimes and two watering regimes. Plants were equally separated either in a temperature -647 controlled greenhouse maintained at 27–30 °C or in a non–temperature-controlled greenhouse 648 (40 plants per greenhouse), where the cooling system was turned off, resulting in ele vated 649 temperatures. The greenhouse temperature of control and Heat stress plants were 650 continuously recorded. Simultaneously (in each greenhouse system), plants were either 651 irrigated daily (well -watered with a drip irrigation system) or subjected to water d eprivation 652 (water deficit). These combinations resulted in four distinct stress conditions: plants grown 653 under controlled temperature and well -watered conditions served as the no -stress control 654 (NS); plants exposed to elevated temperatures while remaining well-watered constituted the 655 heat stress treatment (H); plants maintained at controlled temperature without irrigation were 656 assigned to the drought stress treatment (D); and plants exposed to both elevated 657 temperatures and water deficit represented the com bined heat and drought stress treatment 658 (H:D). Plants treated with the biostimulant and control (untreated) were exposed to the four 659 abiotic conditions leading to eight modalities (n=10). Plants were maintained under these 660 conditions for the duration of th e stress period prior to physiological measurements and 661 sampling. To prevent an excessive water loss by soil evaporation, pots subjected to water 662 deprivation (D and H:D) were placed into dark plastic bags well fixed around the trunk. To 663 prevent plants under combined stress from reaching the wilting point, pots were weighed every 664 two days from day 10 after stress onset (when Ψ PD<-0.6MPa) and until the end of the 665 experiment to allow precise management of water lost through evapotranspiration by supplying 666 the same amount of water lost daily. 667 Water status measurements 668 Grapevine water status parameters were monitored every 5 days using two to four plants per 669 modality (control and treated). Predawn water potential (ΨPD) and midday water potentials (Ψl 670 ) were measured in MPa using a pressure chamber (Model 1000, PMS Instrument Co.) 52. 671 Specifically, ΨPD measurements were completed before sunrise, whereas Ψ measurements 672 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 21 were taken between 12am and 2:00pm. Leaves used at predawn and at midday were fully 673 expanded mature leaves, and at midday exposed to direct solar radiation. After being removed 674 from the branch with a razor blade, leaves were immediately inserted into the DG Meca 675 pressure chamber, with the petiole protruding from the cover. Slow pressure was applied, and 676 the potential value (MPa) was determined using a pressure gauge at the appearance of the 677 water meniscus at the e nd of the petiole. Stomatal conductance ( gs, mmol.m -2.s-1) was 678 measured between 12am and 2:00pm using a LI -600 leaf porometer (LI -COR) operated in 679 auto gsw mode. The instrument was calibrated every 30 measurements. Measurements were 680 performed on the abaxial leaf side of the uppermost fully expanded leaf of 10 plant biological 681 replicates. All measurements were performed in mixed sequence to avoid systematic e rrors 682 between treatments caused by temporary environmental changes. 683 Plant physiology monitoring 684 At the end of the stress period, visual impact of different modalities and stress was assessed 685 at least twice by establishing a visual score from 1 to 5 based on visual cues such as leaf 686 discoloration, dried leaves and apex state (1: No visible stress; 2: Low stress; 3: Moderate 687 stress; 4: High stress; 5: Dead plant, Fig. S1). At the end of the experiment, shoot length was 688 measured. In 2022, the number of intern odes and total leaf area was determined by using a 689

curve linking the size of the central vein of Cabernet -Sauvignon leaves to the leaf 690 area measured on all leaves with a planimeter (R² = 0.9486) (n=10). After 29 and 35 days of 691 stress applications in 2021 and 2022 respectively, roots were carefully cleaned with deionized 692 water and photographed for root architecture analysis, performed with “ARIA” software 53. 693 Finally total root fresh weight (FW) was determined. Leaf chlorophyll was measured at the end 694 of the experiment on three leaves per plant using the Dualex Force A clamp. 695 Sampling 696 For sampling, 3rd or 4th leaves from the apex and the roots cleaned with water were harvested 697 between 10 am and 1 pm. The central vein of the leaves was removed with a scalpel, and the 698 remaining lamina immediately frozen and stored and a t −80 °C. Five replicates were 699 constituted by pooling the leaves or the roots of two different plants. Leaf and roots pools were 700 ground in powder in liquid nitrogen in a cryogrinder (SPEX FreezerMill 6875). 701 Biochemical Characterization and Antioxidant Potential Assay 702 For biochemical determination, 100 mg aliquots of fresh leaf and root powder were extracted 703 with 704 10 mL of absolute methanol and agitated for 30 min at 4°C. After centrifugation, the 705 supernatant is removed through speed -vac and pellet was solubilized with 2 mL absolute 706 ethanol. Total flavonoid, were determined by adding 200 µL of methanolic to 60 µL NaNO 2 7 707 %. After 6 min at room temperature 120 µL AlCl 3 à 10 % is added followed with 120 µL AlCl3 708 à 10 % and 250 mM NaOH. The absor bance was measured at 765 nm and compared to a 709 catechin acid standard curve. Total soluble polyphenols was determined by adding 200 µL of 710 methanolic extract to 1 mL of Folin Ciocalteau reagent (Sigma -Aldrich) diluted 10 times, to 711 which was added 800 µL 7.5 % (w/v) aqueous Na2CO3. After 30 min at room temperature, the 712 absorbance was measured at 765 nm and compared to a gallic acid standard curve. 713 Antioxidant capacity was assessed using Ferric Reducing Antioxidant Power (FRAP) assay. 714 Briefly, 200 µL of extract were mixed with 1 mL of phosphate buffer 0.2 M (pH 6.6) and 1 mL 715 potassium ferricyanide 1 %. After 20 min at 50 °C, trichloroacetic acid 10 % is added. After 716 centrifugation, 1 mL of supernatant is mixed with 1.2 mL of FeCl 3 0.016 %. The absorbance 717 was measured at 700 nm. 718 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 22 Metabolomic analysis 719 For metabolomic analysis, soluble metabolites were extracted from 10 mg aliquots of 720 lyophilised leaf and root powders. Two successive extractions were performed at 4 °C with a 721 buffer composed of ethanol (80 %), 10 mM Hepes–KOH (pH 6) with formic acid 0.1 % and the 722 obtained supernatants were pooled together. 723 Untargeted metabolite profiling of those extracts was carried out using ultra-high-performance 724 liquid chromatography coupled to high -resolution mass spectrometry ( UHPLC-HRMS). 725 Specifically, analyses were performed with an Ultimate 3000 UHPLC system 726 (ThermoScientific, Bremen, Germany) connected to an LTQ-Orbitrap Elite mass spectrometer 727 equipped with an electrospray ionization (ESI) source. Data acquisition was condu cted in 728 negative ionization modes, following the methodology described 54. Chrom atographic 729 separation was achieved on a C18 Gemini column (2.0 × 150 mm, 3 µm particle size, 110 Å 730 pore size; Phenomenex, Torrance, CA, USA). Full -scan high -resolution MS data were 731 collected at a resolving power of 240,000 (measured at m/z 200). In addition, tandem MS 732 (MS/MS) spectra were obtained in higher -energy collisional dissociation (HCD) mode, using 733 normalized collision energies of 60 %. A pooled Quality Control (QC) sample was generated 734 by combining 20 µL from each individual sample and biological standard. QC injections were 735 performed every fifth run and served to monitor analytical consistency throughout the 736 untargeted metabolomics workflow. These QC data were used to compute the coefficient of 737 variation (CV) for each detected metabolite feature, ensuring that only the most reliable signals 738 were included in subsequent chemometric analyses. Raw LC -MS datasets were processed 739 using MS-DIAL version 4.855 After curation steps—including blank subtraction, signal-to-noise 740 ratio (SN) filtering above 10, and exclusion of features with CVs greater than 30% in QC 741 samples—a total of 3,594 features were retained. Among these, 316 were annotated at Level 742 2 confidence with both MS1 and MS2 spectral matches, 2,264 features showed MS1 -only 743 matches (Level 3 identification), and 1,716 features remained unannotated. For the annotated 744 compounds, InChiKeys were submitted to ClassyFire to assign chemical class ification using 745 automated structural ontology. 746 RNA extraction and reverse transcription 747 748 Three biological replicates of leaf and root per modality were selected randomly for RNA 749 extraction. Fresh leaf powder at 80 mg was added to 12 µL of TCEP and 1.4 mL of extraction 750 buffer preheated to 56 °C (300 mM Tris -HCl, pH 8.0, 25 mM EDTA, 2 M NaCl, 20 g/L 751 cetrimonium bromide (CTAB), 20 g/L polyvinylpolypyrrolidone (PVPP), 500 μL/L tri Spermidine 752 hydrochloride (0.05 %) (≥ 98 % Sigma) and 10 g/L β -mercaptoethanol ad ded 753 extemporaneously). The mixture was shaken and centrifuged at 5600 g for 5 min. The following 754 steps of total RNA extraction were performed using the NucleoMag RNA kit from MACHERY-755 NAGEL following the manufacturer’s instructions. Total RNA was reverse transcribed using 2 756 μL-Oligo(dT)12-18 (Invitrogen), ribonuclease inhibitor and M -MLV reverse transcriptase 757 (Invitrogen) following the manufacturer’s instructions, cDNAs were stored at −20 °C. 758 759 Gene Expression 760 Gene expression was assessed using quantitative polymerase chain reaction (qPCR), 761 following the method previously described with the BioStim96 microarray 29. Building on the 762 existing chip design and incorporating additional genes associated with water stress tolerance 763 in grapevine, a new chip named “VitiSummerGen” was developed. This chip includes a 764 selection of genes from various biological processes: pathogen-related (PR) proteins (n = 12), 765 genes fr om secondary metabolism including the phenylpropanoid pathway (n = 11), the 766 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 23 salicylic acid pathway (n = 2), the mevalonate pathway (n = 1), and the shikimate pathway (n 767 = 4). Additional genes are involved in redox metabolism and homeostasis (n = 10), signa ling 768 pathways responsive to heat and drought stress (n = 13), and phytohormonal signaling (auxin, 769 jasmonic acid, abscisic acid; n = 10). Genes related to structural adaptation, such as those 770 involved in cell wall remodeling (n = 8), were also included. Fin ally, the array encompasses 771 genes linked to physiological functions such as water transport (n = 4), energy metabolism (n 772 = 8), nutrient acquisition (n = 6), and cell/plastid division (n = 5). A full list of genes and 773 associated functions is provided in Table S1. 774 The specificity of each primer pair was confirmed by checking the size of the amplified product 775 (data not shown) and by the presence of a single melting peak in qPCR analyses. Primer 776 efficiency values ranged from 0.8 to 1.2, allowing relative gene expressi on to be calculated 777 using the simplified ΔΔCq method (2−ΔΔCq) derived from the Pfaffl model. Three housekeeping 778 genes ( VvGAPDH, VvTIP41, VvTHIORLYS8 ) were used as internal standards for 779 normalization RT -qPCR was conducted according to the MIQE (minimum inf ormation for 780 publication of quantitative real -time PCR experiments) guidelines 56, and final relative 781 expression was the ratio from the sample and the average of relative expression control NS 782 from of its respective year and organ. 783 Statistical analyses 784 Statistical analyses were performed using R Software (R Core Team, Vienna, Austria, 2025). 785 Normality and homogeneity of variance were analysed by Kolmogorov–Smirnov and Levene’s 786 tests. The significance of the results was assessed (p -value < 0.05) using independent 787 samples Student’s t-tests, one-way analysis of variance (ANOVA) followed by Duncan’s tests, 788 or two-way ANOVA, as described in figure legends. Density plots were performed with ggExtra 789 packages. For principal component analysis (PCA) the FactoMineR package was used57. Data 790 visualization was performed with Excell and ggplot2 package 58. Determination of Area Under 791 the Curve (AUC) was performed by Trapezoidal Integration with PRACMA package 59. Venn 792 diagrams were plotted using the online software https://www.interactivenn.net/. Treatment 793 effects on group repartitions for visual stress were tested with Fisher's test with a 5 % threshold. 794

795 The authors gratefully acknowledge the technical assistance provided by Pierre Gastou and 796 Fanny Pinoteau. We also thank Akinao staff for their valuable help in conducting field trials and 797 laboratory analyses. This work received financial support from the French government in the 798 framework of the IdEX Bordeaux University "Investments for the Future" program / GPR 799 Bordeaux Plant Sciences. 800 Author contributions 801 MD, AMB, GC and C ELD conceived and designed the experiments. MD and JB performed 802 field and laboratory work. MD and JB carried out transcriptomic and metabolomic analyses. 803 TP, MD and JB performed data analysis and statistical modelling. TP and MD wrote the 804 manuscript with the contribution from GC and CELD . All authors read and approved the final 805 version of the manuscript. 806 Data availability statement 807 All data can be found online in the main text and supporting information materials (Table S2). 808 The metabolic dataset and all metadata will be deposited online after journal acceptance for 809 publication using Dataverse INRAE. 810 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 24 811 Conflict of interest 812 The authors declare that they have no conflict of interest. 813 Supplementary information 814 Fig. S1: Contribution of different root architectural parameters to PC1 and PC2 in Principal 815 component analysis (PCA) of biostimulant treated and control plants under no stress and 816 combined stress conditions in 2021 (a) and 2022 (b). 817 Fig. S2: Illustrative pictures for different stress classification estimated at the end of the stress 818 phase. 819 Fig. S3: Hierarchical clustering analysis of genes exclusively affected by Heat, AUC gs and/or 820 a combination of them in leaves and roots. 821 Fig. S4: Hierarchical clustering analysis from the correlation matrix generated with a total of 822 1364 metabolites in leaves and 956 metabolites in roots. 823 Fig. S7: Putatively annotated phytohormone related-metabolites under biostimulant (Blue) or 824 control conditions (Grey) under No Stress and Heat:Drought conditions in roots and leaves. 825 Supplementary Table S1 : Names, pathway categories and references of the 96 genes 826 used for the VitiSummerGen microarray chip. 827 Supplementary Table S2 : Data file 828 829

830 1 Zandalinas SI, Sengupta S, Fritschi FB, Azad RK, Nechushtai R, Mittler R. The impact 831 of multifactorial stress combination on plant growth and survival. New Phytologist 832 2021; 230: 1034–1048. 833 2 Sambucci JM, Alston O. The Grape Genome-Grapes in the world economy. 2019 834 doi:10.1007/978-3-030-18601-2_7. 835 3 Gambetta GA, Herrera JC, Dayer S, Feng Q, Hochberg U, Castellarin SD. The 836 physiology of drought stress in grapevine : towards an integrative definition of drought 837 tolerance. Journal of Experimental Botany 2020; 71: 4658–4676. 838 4 Gechev T. Genomics control of biostimulant-induced stress tolerance and crop yield 839 enhancement. The Plant Journal 2025; 123: 1–10. 840 5 Du Jardin P. Plant biostimulants : Definition , concept, main categories and regulation. 841 Scientia Horticulturae 2015; 196: 3–14. 842 6 Yadav A, Singh S, Yadav V. Screening herbal extracts as biostimulant to increase 843 germination, plant growth and secondary metabolite production in wheatgrass. 844 Scientific Reports 2024; 14: 1–14. 845 7 Martínez-Lorente SE, Martí-Guillén JM, Pedreño MÁ, Almagro L, Sabater-Jara AB. 846 Higher plant-derived biostimulants: mechanisms of action and their role in mitigating 847 plant abiotic stress. Antioxidants 2024; 13: 2–27. 848 8 Ahmad A, Blasco B, Martos V. Combating salinity through natural plant Extracts based 849 biostimulants: A review. Frontiers in Plant Science 2022; 13: 1–19. 850 9 Li J, Van Gerrewey T, Geelen D. A meta-analysis of biostimulant yield effectiveness in 851 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 25 field trials. Frontiers in Plant Science 2022; 13:836702: 1–13. 852 10 Herrmann MN et al. A global network meta-analysis of the promotion of crop growth, 853 yield, and quality by bioeffectors. Frontiers in Plant Science 2022; 13: 1–15. 854 11 Patanè C et al. A study on the effect of biostimulant application on yield and quality of 855 tomato under long-lasting water stress conditions. Heliyon 2025; 11: e41187. 856 12 Daler S, Kaya O. Exogenous alpha-lipoic acid treatments reduce the oxidative 857 damage caused by drought stress in two grapevine rootstocks. Physiologia Plantarum 858 2024; 176: 1–17. 859 13 Fusco GM, Burato A, Pentangelo A, Carillo P, Parisi M. Processing tomato responses 860 to plant-based biostimulants are modulated by environmental conditions. Physiologia 861 Plantarum 2025; 177: 1–13. 862 14 Cao X et al. Grape BES1 transcription factor gene VvBES1-3 confers salt tolerance in 863 transgenic Arabidopsis. Gene 2023; 854:147059: 1–19. 864 15 Agarwal PK, Agarwal P, Reddy MK, Sopory SK. Role of DREB transcription factors in 865 abiotic and biotic stress tolerance in plants. Plant Cell Reports 2006; 25: 1263–1274. 866 16 Chai F et al. Comparative metabolic profiling of Vitis amurensis and Vitis vinifera 867 during cold acclimation. Horticulture Research 2019; 6:8: 1–12. 868 17 Pérez FJ, Noriega X, Rubio S. Hydrogen peroxide increases during endodormancy 869 and decreases during budbreak in grapevine (Vitis vinifera L .) buds. Antioxidants 870 2021; 10: 1–14. 871 18 Williams LE, Phene DWGCJ. The effects of applied water at various fractions of 872 measured evapotranspiration on reproductive growth and water productivity of 873 Thompson Seedless grapevines. Irrigation science 2010; 28: 233–243. 874 19 Mirás-Avalos JM, Intrigliolo DS. Grape composition under abiotic constrains: Water 875 stress and salinity. Frontiers in Plant Science 2017; 8: 1–8. 876 20 Tinte MM, Masike K, Steenkamp PA, Huyser J, van der Hooft JJJ, Tugizimana F. 877 Computational metabolomics tools reveal metabolic reconfigurations underlying the 878 effects of biostimulant seaweed extracts on maize plants under drought stress 879 conditions. Metabolites 2022; 12: 1–24. 880 21 Monteiro E, Gonçalves B, Cortez I, Castro I. The role of biostimulants as alleviators of 881 biotic and abiotic stresses in grapevine : A review. Plants 2022; 11: 1–18. 882 22 Alfonso E, Andlauer W, Brück WM, Rienth M. Whey protein hydrolysates enhance 883 grapevine resilience to abiotic and biotic stresses. Frontiers in Plant Science 2025; 884 6:1521275: 1–15. 885 23 Frioni T, VanderWeide J, Palliotti A, Tombesi S, Poni S, Sabbatini P. Foliar vs. soil 886 application of Ascophyllum nodosum extracts to improve grapevine water stress 887 tolerance. Scientia Horticulturae 2021; 277: 109807. 888 24 Boselli M, Bahouaoui MA, Lachhab N, Sanzani SM, Ferrara G, Ippolito A. Protein 889 hydrolysates effects on grapevine (Vitis vinifera L., cv. Corvina) performance and 890 water stress tolerance. Scientia Horticulturae 2019; 258: 108784. 891 25 Bavaresco L, Canavera G, Parisi MG, Lucini L. Role of foliar biostimulants (of plant 892 origin) on grapevine adaptation to climate change. In: 43rd World Congress of Vine 893 and Wine. 2023, pp 1–5. 894 26 Charrier G et al. Drought will not leave your glass empty: Low risk of hydraulic failure 895 revealed by long-term drought observations in world’s top wine regions. Science 896 Advances 2018; 4: 1–10. 897 27 Martinez-Medina A et al. Recognizing plant defense priming. Trends in Plant Science 898 2016; 21: 818–822. 899 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 26 28 Kanojia A et al. Plant Stress Physiological and molecular insights into the effect of a 900 seaweed biostimulant on enhancing fruit yield and drought tolerance in tomato. Plant 901 Stress 2024; 14: 2–17. 902 29 Bodin E, Bellée A, Dufour M-C, André O, Corio-Costet M-F. Grapevine stimulation: a 903 multidisciplinary approach to investigate the effects of biostimulants and a plant 904 defense stimulator. Journal of Agricultural and Food Chemistry 2020; : 15085–15096. 905 30 Zarraonaindia I et al. Holistic understanding of the response of grapevines to foliar 906 application of seaweed extracts. Frontiers in Plant Science 2023; 14: 1–17. 907 31 Wang Z et al. Unveiling the molecular mechanisms of γ-polyglutamic acid-mediated 908 drought tolerance in cotton through transcriptomic and physiological analyses. BMC 909 Plant Biology 2025; 25: 2–14. 910 32 Li T et al. Raffinose synthase enhances drought tolerance through raffinose synthesis 911 or galactinol hydrolysis in maize and Arabidopsis plants. Journal of Biological 912 Chemistry 2020; 295: 8064–8077. 913 33 Wang H, Blakeslee JJ, Jones ML, Chapin LJ, Dami IE. Exogenous abscisic acid 914 enhances physiological, metabolic, and transcriptional cold acclimation responses in 915 greenhouse-grown grapevines. Plant Science 2020; 293: 110437. 916 34 Yoshida T et al. Four Arabidopsis AREB/ABF transcription factors function 917 predominantly in gene expression downstream of SnRK2 kinases in abscisic acid 918 signalling in response to osmotic stress. Plant, Cell and Environment 2015; 38: 35–49. 919 35 Jones RJ, Mansfield TA. Suppression of stomatal opening in leaves treated with 920 abscisic acid. Journal of Experimental Botany 1970; 21: 714–719. 921 36 Dinis LT et al. Kaolin exogenous application boosts antioxidant capacity and phenolic 922 content in berries and leaves of grapevine under summer stress. Journal of Plant 923 Physiology 2016; 191: 45–53. 924 37 Buer CS, Kordbacheh F, Truong TT, Hocart CH, Djordjevic MA. Alteration of flavonoid 925 accumulation patterns in transparent testa mutants disturbs auxin transport , gravity 926 responses , and imparts long-term effects on root and shoot architecture. Planta 2013; 927 238: 171–189. 928 38 Watkins JM, Chapman JM, Muday GK. Abscisic acid-induced reactive oxygen species 929 are modulated by flavonols to control. Plant Physiology 2017; 175: 1807–1825. 930 39 Fandino ACA et al. Priming grapevines with oregano essential oil vapour results in a 931 metabolomic shift eliciting resistance against downy mildew. BMC Plant Biology 2024; 932 24. doi:10.1186/s12870-024-05875-y. 933 40 Nakabayashi R, Mori T, Saito K. Alternation of flavonoid accumulation under drought 934 stress in Arabidopsis thaliana. Plant Signaling & Behavior 2014; 9: e29518. 935 41 Li B et al. Flavonoids improve drought tolerance of maize seedlings by regulating the 936 homeostasis of reactive oxygen species. Plant Soil 2021; 461: 389–405. 937 42 Gayomba SR, Muday GK. Flavonols regulate root hair development by modulating 938 accumulation of reactive oxygen species in the root epidermis. Development 2020; 939 147: dev185819. 940 43 Chapman JM, Muday GK. Flavonols modulate lateral root emergence by scavenging 941 reactive oxygen species in Arabidopsis thaliana. Journal of Biological Chemistry 2021; 942 296: 100222. 943 44 Buer CS, Djordjevic MA. Architectural phenotypes in the transparent testa mutants of 944 Arabidopsis thaliana. Journal of Experimental Botany 2009; 60: 751–763. 945 45 Ismaiel SA, Salama HM. Allelopathic effects of black nightshade (Solanum nigrum l.) 946 on germination, growth and yield of broad bean (vicia faba l.) and common bean 947 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint 27 (phaseolus vulgaris l.). Applied Ecology and Environmental Research 2021; 19: 3431–948 3441. 949 46 Bibi S et al. Allelopathic effects of thuja orientalis L. and melia azedarch L. on seed 950 germination of Penisitum americanum L. International Journal of Basic and Applied 951 Sciences 2016; 5: 140–143. 952 47 Salam SGA, Rashed MM, Ibrahim NA, Rahim EAA, Aly TAA, AL-Farga A. 953 Phytochemical screening and in-vitro biological properties of unprocessed and 954 household processed fenugreek (Trigonella foenum-graecum Linn.) seeds and leaves. 955 Scientific Reports 2023; 13: 1–13. 956 48 Przybylska-Balcerek A et al. Sambucus nigra extracts–natural antioxidants and 957 antimicrobial compounds. Molecules 2021; 26: 2–17. 958 49 Boukhari M, Asencio-Vicedo R, Cerdán M, Sánchez-Sánchez A, Jordá JD, Ferrández-959 Gómez B. Foliar application of Equisetum arvense extract enhances growth, alleviates 960 lipid peroxidation and reduces proline accumulation in tomato plants under salt stress. 961 Plants 2025; 14: 6–10. 962 50 Qasem A et al. Determination of chemical compounds and investigation of biological 963 properties of Matricaria chamomilla essential oils, honey, and their mixture. Molecules 964 2022; 27: 2–18. 965 51 Bab T, Pall E. Revealing the phenolic composition and the antioxidant, antimicrobial 966 and antiproliferative activities of two Euphrasia sp. extracts. Plants 2024; 13: 1–16. 967 52 Scholander PF, Hammel HT, Bradstreet ED, Hemmingsen EA. Sap pressure in 968 vascular plants. Science 1965; 148: 339–346. 969 53 Pace J, Lee N, Sing Naik H, Ganapathysubramanian B, Thomas Lübberstedt. Analysis 970 of maize (Zea mays L.) seedling roots with the high-throughput image analysis tool 971 ARIA (Automatic Root Image Analysis). PLoS ONE 2014; 9: e108255. 972 54 Luna E et al. Metabolomics to exploit the primed immune system of tomato fruit. 973 Metabolites 2020; 10: 1–19. 974 55 Tsugawa H et al. data-independent ms/ms deconvolution for comprehensive 975 metabolome analysis. Nature Methods 2015; 12: 523–531. 976 56 Bustin SA et al. The MIQE Guidelines : Minimum Information for publication of 977 quantitative real-Time PCR experiments. Clinical chemistry 2009; 55: 611–622. 978 57 Lê S, Josse J, Husson F. FactoMineR: An R package for multivariate analysis. Journal 979 of Statistical Software 2008; 25: 1–18. 980 58 Wickham H. ggplot2: Elegant Graphics for Data Analysis. Journal of Statistical 981 Software 2010; 35: 1–3. 982 59 Gander W, Gautschi W. Adaptative quandrature—Revisited. BIT 2000; 40: 084–101. 983 984 Figure legends 985 986 Tables 987 988 Figures 989 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 24, 2026. ; https://doi.org/10.64898/2026.02.23.707262doi: bioRxiv preprint