The Application of Ozonated Water Rearranges the Transcriptome of the Vitis Vinifera L. Fruit

Ozonated water has become an innovative, environmentally friendly tool for controlling the development of fungal diseases in the vineyard or during grape postharvest conservation. However, little information is currently available on the effects of ozonated water sprayings on the grapevine physiology. Using the microvine model, we studied the transcriptomic response of leaf and fruit organs to this treatment. The response to ozone was observed to be organ and developmental stage-dependent, with a decrease of the number of DEGs (differentially expressed genes) in the fruit from the onset of ripening to later stages. The most highly up-regulated gene families were heat-shock proteins and chaperones. Other up-regulated genes were involved in oxidative stress homeostasis such as those of the ascorbate-glutathione cycle and glutathione S-transferases. In contrast, genes related to cell wall development and secondary metabolites (carotenoids, terpenoids, phenylpropanoids / avonoids) were generally down-regulated after ozone treatment, mainly in the early stage of fruit ripening. This down-regulation may indicate a possible carbon competition favouring the re-establishment and maintenance of the redox homeostasis rather than the synthesis of secondary metabolites at the beginning of ripening, the most ozone responsive developmental stage. sprayings on the composition of grapes and resulting wines, with phenolic and terpenoid compounds increased or decreased by ozone without showing a linear correlation with the number of applications 10–12, 14 . Such heterogeneous results indicate that more studies in controlled conditions are needed to understand the molecular and biochemical changes induced by O 3 in grapevine organs. Using the microvine model, this study represents the rst transcriptomics analysis exploring the responses triggered by ozonated water spraying on grapevine leaves and fruits.


Introduction
Vitis vinifera gathers most grapevine cultivars used for table grape and wine production. Unfortunately, this species is highly susceptible to a range of fungal diseases such as downy and powdery mildews and the grey mould, respectively caused by Plasmopara viticola, Erysiphe necator and Botrytis cinerea. Moreover, a complex group of pathogenic fungi that attacks perennial organs is responsible for the socalled grapevine trunk diseases. To overcome the negative impacts of these pathogens on plant development and fruit quality, and avoid excessive crop losses, viticulture needs to perform intense fungicide spraying programs, especially in hot and wet weather conditions. Even organic and biodynamic approaches largely require sulfur-and copper-based formulations that may be detrimental to the soil ecosystem in the long term. The ecological and environmental sustainability is an increasing concern for consumers and more generally for society.
One way to reduce the susceptibility of V. vinifera to pathogens is to breed new cultivars introgressing genetic traits of resistance from American and Asian Vitis spp. Several breeding programs are ongoing in Europe and abroad with an increment of new resistant genotypes available. In parallel to introducing these new varieties, which is a long process and often not entirely accepted by traditional viticulturists, the use of ozone (O 3 ) was proposed as a new tool for controlling fungal diseases on susceptible traditional varieties. Indeed, when applied in aqueous solution, ozone has been shown to suppress spore germination of the esca-associated fungus Phaeoacremonium aleophilum and reduce fungal development by 50% on Cabernet Sauvignon cuttings 1 . The use of ozonated water in integrated vineyard pest management appears to be as effective as traditional chemical treatments in reducing fungal populations on leaves and grape bunches 2 . The e ciency of ozone is thought to lie in its oxidising potential, which translates into the ability to react with numerous cellular constituents hence a broadspectrum antimicrobial action 3 .
Its low persistence after application makes ozone particularly attractive from an environmental point of view. This triatomic molecule is highly unstable and spontaneously decomposes into oxygen without leaving hazardous residues, with a shorter half-life in water than in the gaseous state 3 . In aqueous solution, ozone can be broken down via a chain reaction mechanism resulting in the production of reactive oxygen species (ROS), i.e. the hydroperoxide (HO 2 •), superoxide (•O 2 − ) and hydroxyl (•OH) radicals and hydrogen peroxide (H 2 O 2 ), all contributing to the high oxidising power of ozone 4 .
Ozone enters plant tissues through the stomata, lenticels or physical breaks in the cuticle. Then it reacts with molecules present in the apoplastic uid, cell wall and plasma membranes, where it decomposes to produce the ROS mentioned above 5 . Under the oxidative stress induced by ozone and derived products, plants develop defence mechanisms at the genetic, transcriptional and biochemical level, which includes the synthesis of antioxidants such as ascorbate, glutathione, enzymes like superoxide dismutases, catalases and peroxidases, and secondary metabolites like carotenoids, terpenoids and phenolics [6][7][8] .
When the detoxi cation capacity of plant cells is overwhelmed, cellular damage can occur.
Most research about the effects of ozone on plants has focused on the physiological changes triggered by ozone as a pollutant. However, ozone applied in aqueous solution and in a timely manner is expected to interact with plants differently than in the gaseous state, with a su ciently high phytotoxic threshold that allows its incorporation in irrigation and spraying treatments in different crop species 9 .
In this study, this model allowed us to characterise the early transcriptome changes triggered in grapevine leaves and berries at different ripening stages after in planta sprayings of ozonated water solutions.

Results
The balance in primary metabolites: an analytical tool to select RNA-Seq samples At the beginning of ripening (BR), soft green berries were sampled while still in the lag phase with no visible anthocyanin accumulation in their skin. These berries just started to accumulate sugar while consuming malate (Fig. 1a). As expected, berries in the mid-ripening stage (MR) showed higher sugar concentrations (close to 1 M) and a lower amount of malic acid (Fig. 1a). Mature leaf samples (L) displayed a comparable amount of soluble sugars to BR, with a two-fold lower malate concentration, indicating strong differentiation between the source (leaves) and sink (berries) organs. Thanks to the measurements of sugars and acids, it was possible to gather synchronised samples 24 for further RNA-Seq analysis with the aim to reduce biases in gene expression caused by the natural developmental asynchrony of grapevine berries and focus only on the early transcriptomic changes triggered by the ozonated water treatment. Indeed, biological triplicates were selected at the same sugar (glucose + fructose) concentrations for control (C) and ozonated water treatment (OW), namely 158 mM in L, 291 mM in BR, and 864 mM in MR (Fig. 1b). Malic and tartaric acids were 184 mEq and 255 mEq for L, 363 mEq and 120 mEq for BR, and 139 mEq and 103 mEq for MR (Fig. 1b), giving an average malate/tartrate ratio of 0.7, 3.0, and 1.3, respectively. No signi cant differences were found between conditions for sugar, acids and sample weight (Fig. 1b).

Transcriptomic overview in leaf and ripening berry
Principal component analyses were performed to visualise the global transcriptome trends (Fig. 2a,b,c). The rst two principal components (PC1 + PC2) explained 65, 74, and 63 % of the variance among samples in L, BR, and MR. C and OW samples were clearly resolved in BR, while the separation was less obvious in L and MR. The hierarchical clustering dendrogram showed the degree of similarity between the transcriptome pro le of all the samples analysed (Fig. 2d). The most striking differences in the transcriptome were determined by the type of organ, i.e. leaf or berry, followed by the berry developmental stage. As before, the dendrogram showed that the C and OW BR samples grouped separately, while for L and MR the three OW replicates clustered conjointly, but one C sample was placed in a different branch than the other two.
Genes differentially expressed according to the ozone treatment were tested in the leaves and two berry developmental stages separately (Fig. 2e,f, and Table S1). In L, the total number of DEGs was 191, with 84 up-regulated genes and 107 down-regulated. The most intense response to the treatment was number of classes were detected in the up-regulated genes (Fig. 3). In common to the three organs, several categories reported enrichment for protein folding and related categories, and response to a plethora of stresses including response to heat, response to hydrogen peroxide, response to reactive oxygen species, and response to oxidative stress (Fig. 3).
Up-regulation of heat-shock proteins and chaperones: a common response of leaves and berries To prioritise the genes whose expression changed the most after a 90-min exposure to ozone, each list of DEGs was ranked according to their respective absolute changes in expression between C and OW ( Table   S1). The overall reaction in both leaves and berries was primarily to activate rather than repress physiological processes. The most changing genes were a conspicuous number of heat shock proteins (HSPs) and chaperones highly up-regulated, with not less than 11 small HSP, or HSP20, up-regulated in L, 36 in BR and 28 in MR (Fig. 4). Among the HSP20 recently identi ed in grapevine 25 , VviHSP20-09, VviHSP20-17, VviHSP20-22, VviHSP20-25, VviHSP20-27, VviHSP20-35, VviHSP20-36, VviHSP20-39, VviHSP20-42, and VviHSP20-44 were commonly highly up-regulated. Other HSP of bigger sizes such as HSP70 and HSP90 were up-regulated as well, together with a series of chaperones, such as the DNAJ homolog, calnexin and calreticulin (Fig. 4).
Interestingly, diverse heat-stress transcription factors (HSF) were also modulated by the treatment: VviHSF-A6b was up-regulated in L, BR, and MR. VviHSF-A2 and VviHSF-B2a were up-regulated only in the berry (BR and MR). VviHSF-A3 was up-regulated in BR, VviHSF-B2b in MR, and VviHSF-A1b was the only gene down-regulated in BR. Moreover, the transcription factors multiprotein-bridging factor 1c and 1a (VviMBF1C, VviMBF1A) were up-regulated, with a noticeably strong activation of the rst one in the berries. Lastly, as part of the stress response, two galactinol synthases (VviGOLS) were up-regulated in L, BR, and MR ( Fig. 4).

Antioxidant homeostasis
Other categories of DEGs, identi ed mostly in BR, were related to the antioxidant homeostasis, which involves the scavenging of the reactive oxygen species (ROS). The ascorbate-glutathione cycle (AsA-GSH cycle) detoxi es ROS through the activity of ascorbate peroxidase (APX), monodehydroascorbate reductase (MDAR), dehydroascorbate reductase (DHAR), and glutathione reductase (GR), requiring a pool of ascorbate, glutathione and NADPH. Here, one APX (Vitvi08g01143) in BR and one DHAR (Vitvi13g00241) in L and BR were up-regulated (Fig. 5), indicating an enhanced turnover of the cycle under stress to scavenge O 3 -generated H 2 O 2 into water. Paradoxically, two isogenes encoding VTC2 (GDP-L-galactose phosphorylase), a regulatory step in AsA biosynthesis, were down-regulated in BR (Fig. 5).
Intense down-regulation of the cell wall-related genes and plasma membrane aquaporins Ozonated water sprayed all over the plant surface strongly impacted the cuticle and cell wall-related genes of leaves and berries with more emphasis on the BR berry (Fig. 6). Six cuticle genes were downregulated together with ten expansins, among which the most highly repressed were VviEXPA11, VviEXPA14, VviEXPA18, and VviEXPA19. Four cellulose synthases were up-regulated while seven cellulose synthase-like were down-regulated together with three pectate lyases and other pectinesterases. Only ten xyloglucan endotransglucosylase / hydrolase genes were up-regulated. Interestingly, four aquaporins were modulated by OW at the beginning of ripening: VviPIP1.1 and VviPIP2.3 located on the plasma membrane were down-regulated; on the contrary VviTIP2.1 and VviTIP3.1, whose localisation is the tonoplast, were up-regulated.
Secondary metabolism is affected only at the beginning of ripening Antioxidant secondary metabolites like carotenoids, terpenoids and phenolic compounds can be synthesised in response to the stress. The expression level of several related genes was modulated by OW in BR berries, while no signi cant impact could be detected in L and MR (Fig. 7).
Several genes of the phenylpropanoid and avonoid pathway were also differentially expressed by OW. In particular, two phenylalanine ammonia-lyases (VviPAL) and one trans-cinnamate-4-monooxygenase (VviC4H) were down-regulated in BR. In contrast, three caffeic acid 3-O-methyltransferases (VviCOMT) were differentially regulated with two genes down-regulated and one up-regulated. Other genes involved in the terminal steps of monolignol biosynthesis were also affected; namely, a cinnamoyl-CoA reductase (VviCCR) was down-regulated, while a cinnamyl alcohol dehydrogenase (VviCAD) up-regulated. In the same berries, two iso avone reductases (VviIFR), implicated in the iso avonoid phytoalexin branch pathway, were down-regulated. Lastly, the ozonation of MR berries only reduced the expression of a hydroxycinnamoyl-CoA: shikimate / quinate hydroxycinnamoyltransferase (VviHCT) and a VviIFR. Regarding the avonoid pathway, a avanone 3-hydroxylase (VviF3H), a avonoid 3'-hydroxylase (VviF3'Ha) and a leucoanthocyanidin dioxygenase (VviLDOX) were modulated by the stress in BR: the rst one up-regulated, while the last two down-regulated. A caffeoyl shikimate esterase (VviCSE) and the VviF3'Ha were the only DEGs down-regulated in L. Lastly, relevant transcription factors, such as VviMYB5a, VviMYB5b, and VviMYBF1, controlling different branches of the avonoid pathway, were down-regulated in BR.

Discussion
Although the major use of ozone in agriculture lies in its antifungal activities, as con rmed in grapevine 1,2,10 , there is still a lack of information on how ozone can affect grapevine physiology and grape composition. Previous eld studies reported versatile impacts of ozonated water sprayings on the composition of grapes and resulting wines, with phenolic and terpenoid compounds increased or decreased by ozone without showing a linear correlation with the number of applications [10][11][12]14 . Such heterogeneous results indicate that more studies in controlled conditions are needed to understand the molecular and biochemical changes induced by O 3 in grapevine organs. Using the microvine model, this study represents the rst transcriptomics analysis exploring the responses triggered by ozonated water spraying on grapevine leaves and fruits.
BR berries appeared incredibly responsive to ozone exhibiting the highest number of DEGs. The intense transcriptomic reprogramming at the onset of ripening, largely documented in grapevine fruit 26,27 , has also been associated with ROS accumulation 28 , whose synthesis occurs most intensively during the night 16 . Due to the method of monitoring the development of the berries and their sampling, we can reasonably assume that BR berries were very close to the H 2 O 2 and catalase peaks that were spotted in non-developmentally synchronised fruits 28 . The intense transcriptomic changes described here showed that endogenous ROS production previously reported at the onset of ripening is actually far from saturating in standard conditions with no stress. The observed response can also be explained by the greater variety of reactive species formed from aqueous O 3 , including the more potent oxidant and chainpropagating hydroxyl radical 4 , which can differ from the ones endogenously produced. In fact, the endogenous ripening related ROS production does not result in the cell wall and growth inhibition, as this production is suspected to occur just before or at the inception of the second fruit growth phase. Indeed, recent physiological and transcriptomic works evidenced that the less harmful hydrogen peroxide (H 2 O 2 ) accelerated ripening in Kyoho variety 29,30 . The genes suggested by the authors to induce the early ripening were associated with the oxidative stress, photosynthesis, cell wall deacetylation and degradation. More studies are needed to decipher the possible role of ozonated water in grape ripening, knowing that H 2 O 2 is only one of the ROS formed by the decomposition of ozone in aqueous solution 4 .
Berry softening marks the onset of the massive import of sugars in grapevine. Surprisingly, VviSWEET10, which is implicated in the unloading of phloem sucrose inside the berry 23 , was up-regulated in BR together with two TIPs, aquaporins of the vacuole. But the expression of VviHT6, the major sucrose transporter on the tonoplast, was not affected, leaving open the question of a possible enhancement of the ripening program under ozonated water. As ozone decomposition strongly depends on pH, its decay may be faster in the cell wall and cytoplasm than in the acidic vacuole of berries at the beginning of ripening 31 .
In our dataset based on developmentally synchronised berries, some cuticle related genes were downregulated. The degradation of this protective barrier, which leads to greater penetration of ozone into the plant cells, has been reported in growing plants and postharvest fruits exposed to ozone 32 . Moreover, key expansins involved in the cellular expansion and growth 33 were down-regulated with pectate lyases, pectinesterases and cellulose synthases like indicating an immediate multifaceted effect unsettling the cell-wall dynamics, further exacerbated by the down-regulation of two plasma membrane aquaporins suggesting a limited water in ux. Ozone has been shown to modify the composition and mechanical properties of grape skin cell walls 34 , affecting aroma and polyphenols extraction during winemaking 35 .
The lower anthocyanin extractability observed after spraying ozonated water on grapevines 11,13 may originate from the down-regulation of genes encoding pectin-degrading enzymes detected in ozonated berries.
The rst coordinated response to the ozonated treatment was the induction of a plethora of HSPs and other chaperones. HSPs are involved in the cellular response to a diverse array of stresses, including oxidative 36 . They act mainly as molecular chaperones, participating in protein folding, assembly, translocation and degradation in many normal cellular processes and maintain proteins in their functional conformations under stress conditions, preventing their aggregation and denaturation, and assisting in protein refolding 37 . The induction of HSP transcripts in plants fumigated with ozone was rst described in parsley 38 and then con rmed in other plants such as Arabidopsis thaliana and Medicago truncatula 39,40 . Using proteomic approaches, the increased expression of these proteins under ozone stress was also detected in poplar, bean, maize and rice [41][42][43] . The induction of HSPs is under the tight control of an HSF network 44 , with signi cant player VviHSF-A2 and VviHSF-A6b already reported intensi ed in grapevine under stress 17,18,45 , often together with VviGOLS 46 . Moreover, transgenic Arabidopsis thaliana plants constitutively expressing the transcriptional coactivator AtMBF1c showed enhanced tolerance to environmental stresses 47 . Here these genes were strongly up-regulated, possibly cross-regulating several plant response mechanisms to various stresses.
Plants submitted to different abiotic or biotic stresses typically produce ROS, triggering oxidative stress 48 . AsA is the most abundant antioxidant in plant cells, found in all subcellular compartments, including the apoplast, and therefore representing the rst line of defence against ozone 49 . AsA can directly scavenge ozone and different ROS 50 and, along with glutathione in the AsA-GSH cycle, is the primary H 2 O 2 reducing substrate operating in cytosol, chloroplasts and mitochondria of plant cells 51 . It has been shown that the antioxidant response to the stress is genotype-dependent, with grape varieties such as Touriga Nacional able to boost the cell redox-buffering capacity with the existing AsA and GSH pools, while other varieties, like Trincadeira, need to synthesise both metabolites because of its incapacity to keep the cellular redox state at working levels 52 . Therefore, it is not surprising that VviVTC2, the central regulator of the AsA biosynthetic pathway 53 , was down-regulated in BR, indicating a non-need for resynthesis but a buffering capacity of the microvine coping with oxidative stress. Similarly to our results, OsVTC2 was downregulated in ozone-exposed rice, attributing the changes in total and reduced AsA concentration to AsA turnover rather than biosynthesis, with a parallel increase of OsAPX, OsDHAR, and OsGR 54 . Also in our dataset, VviAPX and VviDHAR were up-regulated under ozone. Elevated expression of these two genes in response to ozone has already been detected in Arabidopsis thaliana 55,56 , and DHAR-overexpressing plants have shown increased tolerance to ozone by incrementing foliar AsA level 57 . In grapevine, AsA is also a precursor for the synthesis of both tartaric and oxalic acids. The down-regulation of VviVTC2 in BR berry under ozone stress could indicate a switch from the Smirnoff-Wheeler (SW) pathway to the alternative AsA biosynthetic pathway, knowing that the rst one supports AsA biosynthesis in immature berries, while the alternative synthesises AsA from a methyl derivative of D-galacturonic acid released during pectin degradation as fruits ripen 58 . Given that GDP-D-mannose and GDP-L-galactose, intermediates of the SW pathway, are also precursors of the non-cellulosic components of the plant cell wall 59 , we can speculate that the inhibition of enzymes involved in cell wall synthesis and growth would lead to AsA sparing and in turn to reduced AsA synthesis, materialised through the down-regulation of VviVTC2.
Other critical antioxidant enzymes such as CAT, POD, SOD, RX, and GST were modulated by the stress indicating an intense redox homeostasis activity to prevent ozone and derived byproducts damages 48 . In particular, the treatment induced the expression of six out of eight GSTs detected in BR berries. This elicitor effect was also observed in MR berries, con rming previous results in ozone-exposed Arabidopsis and rice seedlings 39,43,60 . Thiols such as GSH are versatile targets for most oxidants, including ozone 61 , so we hypothesise that GST activity increased in order to counterbalance reduced substrate availability, allegedly enhanced in BR berry by VviDHAR up-regulation. GSTs are also necessary for the transport of anthocyanins from the cytosol to the vacuole. Consequently, a strong correlation between these proteins and anthocyanin accumulation has been found in V. vinifera 62 , indicating a possible involvement in the increased phenolic content under ozonated water treatments.
Although secondary metabolites are important antioxidants whose synthesis is typically induced in plants as a defence mechanism against ozone 6,7 , in the early transcriptional response to the ozonated water application their pathways were generally unaffected in leaves and mid-ripening berries, with some genes down-regulated in berries starting to ripen.
Carotenoids contribute to light harvesting and protect the photosynthetic membrane against photooxidative damage, not only by quenching the triplet states of chlorophyll but also by scavenging ROS 63 . The fact that the ozonated water treatment impaired the synthesis of carotenoids through the downregulation of VviZISO1, VviZDS1, VviCISO1 and VviLBCY2 in the early ripening berry seems counterintuitive, however, similar observations were made in different rice genotypes 54 . The regeneration of carotenes and xanthophylls from their oxidised radicals relies on AsA 50 and, in addition, the violaxanthin de-epoxidase enzyme requires AsA as a cofactor 64 . Here, the higher expression of VviVDE2 in the ozonated BR berry indicates an activation of the de-epoxidation in the xanthophyll cycles, which protects against ROS-generating stresses 65 . This mechanism is expected to be also activated in ozone-treated leaves as they often undergo a reduction in photosynthetic rates and need to dissipate the excess excitation energy absorbed by the antennae 8 . However, here, no sign of photosynthetic apparatus damage was observed in leaves (data not shown). The activation of the xanthophyll cycles in BR berry may respond to the zeaxanthin and lutein roles in ROS scavenging and preventing membrane lipid peroxidation 66,67 .
Terpenoids have been shown to improve the ability of plants to cope with internal oxidative changes 68 , reduce ozone damage and quench ozone and ROS 7 . However, ozone has been shown to stimulate and reduce the biosynthesis and emission of these volatiles depending on the severity and duration of the exposure and the plant species sensitivity 69 . Here, the overall down-regulation of the genes involved in their synthesis in the treated BR berries, such as VviDXS and VviTPS31, key determinants in the production of monoterpenes in grapevines 70,71 , and VviGGPPS, the precursor of diterpenes and carotenoids, contrasts with the higher terpenoid content found in berries from Bobal and Vermentino grapevines subjected to ozonated water treatments 10,11 . Nevertheless, this increase was detected in berries at the end of the ripening period and not after each ozone exposure, and was much less pronounced when the treatment implied an application at the onset of veraison 11 . By contrast, our ndings refer to the early response to the treatment and in a longer-term -such as the time of harvestthe expression of the affected genes could vary. In this line, an immediate depression of isoprene emission was reported in Quercus pubescens leaves exposed to ozone, attributed to a temporary inhibition of photosynthesis, but a subsequent fast recovery and even stimulation 12 days after fumigation 72 .
Plants exposed to ozone often respond with increased transcription and activities of enzymes involved in the phenylpropanoid, lignin and avonoid pathways because of their barrier and antioxidant roles 8,73 .
However, this response may not be immediate: for example, the induction of genes involved in the avonoid synthesis in Arabidopsis was part of the later response to two days of ozone exposure, with chalcone synthase, dihydro avonol reductase and leucoanthocyanidin dioxygenase being the most responsive 39 . In Melissa o cinalis L., an ozone treatment (5 h) initially impaired PAL activity, the rst enzyme in the general phenylpropanoid pathway, followed by a subsequent increase 7 h after the end of the exposure 74 . Similarly, our results showed that the early response to the ozonated water treatment, mainly in BR berries, consisted of an overall down-regulation of several genes involved in these pathways. Whether these genes are reactivated later is presently unknown.
Plants are sessile organisms that produce metabolites as an adaptive strategy to cope with challenging and changing environments 75 . Secondary metabolic routes are highly demanding for energy and carbon compounds, including the metabolites synthesis, their transcriptional regulation, and transport in subcellular compartments 76 . On the urgency to respond to the stress in the short-time, grapevine vegetative and reproductive organs apparently prefer to allocate carbon and energy to immediate defence response (HSPs, chaperones, AsA-GSH cycle). We can speculate that multiple treatments and/or a longer span between ozone exposure and sampling could lead to adaptation mechanisms triggering cascades of signal networks ending with the synthesis of stress-related genes and secondary metabolites accumulation, as often observed in grapes at harvest. This study is an original contribution performed with a perennial fruit crop. The goal was to characterise the rst responses of both vegetative organs and eshy fruits to ozonated water treatments. Therefore, further studies will be needed to get a comprehensive understanding of the long-term effects on plant physiology and especially on fruit composition. Based on this rst study and previous experiences 15 , we propose the microvine as a relevant perennial eshy fruit model to perform such investigations.

Plant material
Two-year-old ML1 microvines were grown in 3 l pots under semi-controlled conditions in a greenhouse (Montpellier SupAgro-INRAe campus, France) with day/night temperature 25/15°C, 1 kPa of VPD, and 12 h photoperiod. Microvines were managed for eight months to display all fruit developmental stages from owering to ripe stages 15 . Plants were maintained at full ETP (EvapoTranspiration Potential), thus avoiding water stress issues, and no fungicide sprayings were performed. At the beginning of the experiment, the developmental stage of single green berries was checked by visual inspection and rmness assessment 21 , in order to detect the rst softening signs as the onset of sugar storage. Single berry growth was weekly monitored by image analysis of clusters taken 30 internodes below the apex, with a Lumix FZ100 camera (Panasonic).

Ozonated water treatment and sampling
Before treatment, plants were randomly divided into two groups: four plants for the control (C) and four plants for the ozonated water treatment (OW) (Fig. 8a). To have ventilation representative of eld conditions, plants were brought outside the greenhouse for the entire duration of the experiment (9 am − 12 pm). Ozonated water was prepared extemporaneously using an ozone generator (Cosemar Ozono S.L., Spain) connected to a sprayer containing Milli-Q water at a temperature of 15 ºC and a conductivity of 18.2 MΩ/cm. A redox meter (PCE-228-R, PCE Ibérica S.L., Spain) was used to continuously measure in millivolts (mV) the oxidation-reduction potential (ORP) of the aqueous solution. One hundred fty ml of ozonated water was sprayed on the entire surface of each OW plant once its ORP reached 1000 mV (Fig. 8b). The four C vines were sprayed with the same amount of Milli-Q water used for the treatment. Right after the spraying, plants were enclosed in plastic bags to prevent drift and avoid too rapid ozonated water evaporation (Fig. 8c). Ninety minutes after the start of the treatment, 15 single green berries at the beginning of ripening (BR) (+ 3 days after softening), 15 single berries in the mid-ripening stage (MR) (+ 18 days after softening), and two young adult leaves per plant (L) located between the 30th and 40th nodes were sampled for both C and OW. Single berry samples (pericarp and seeds) and leaves were wrapped separately in aluminium foils and immediately frozen in liquid N 2 . Each sample was weighed and ground into liquid N 2 using a ball mill (Retsch, Germany). The resulting powder was stored at − 80°C, and used for primary metabolites and RNA analyses.

Primary metabolites analysis
Sugars and acids were analysed by high-performance liquid chromatography (HPLC), according to Rienth et al. 18 . Brie y, 100 mg of leaf or berry frozen powder was 5x diluted in HCl 0.25 N and left overnight at room temperature after shaking. Samples were then centrifuged at 15000 g for 10 min, and a supernatant aliquot was diluted 10x with a solution of H 2 SO 4 5 mM containing 600 µM acetic acid as internal standard, before injection into the HPLC system. The statistical analysis of the data was performed with SPSS statistics software (version 23.0 for Windows, Chicago, IL, USA). The mean values of the selected samples were compared using the independent samples t-test, and the differences were considered statistically signi cant when the p-value < 0.05.

RNA extraction and sequencing
Three samples per treatment (C and OW) and organ (L, BR and MR) were selected for individual RNA extraction and library preparation as described in Rienth et al. 77 . Samples were sequenced on an Illumina HiSeq3000 in paired-end mode, 2x150 bp reads, at the Genotoul platform of INRAe-Toulouse (France).

Data analysis
Raw reads were trimmed for quality and length with Trimmomatic, version 0.38 78 . Reads were aligned against the reference grapevine genome PN40024 12X2 79 , using the software Hisat2, version 2.1.0 80 with standard parameters, yielding an average of 25.3 M sequence per sample (Table S2). Aligned reads were counted using the VCost.v3 annotation with HTSeq-count (version 0.9.1) 81 , in union mode, mRNA type, nonunique all, and stranded options. Only genes with RPKM > 1 were kept for further analysis (Table S3).
Overrepresented gene categories were identi ed with the gPro ler web-server (version 101_eg48_p14_baf17f0) with a signi cance threshold of 0.001.