Transcriptomic and genomic structural variation analyses on grape cultivars reveal new insights into the genotype-dependent responses to water stress

Grapevine (Vitis vinifera L.) is importantly cultivated worldwide for table grape and wine production. Its cultivation requires irrigation supply, especially in arid and semiarid areas. Water deficiency can affect berry and wine quality mostly depending on the extent of plant perceived stress, which is a cultivar-specific trait. We tested the physiological and molecular responses to water deficiency of two table grape cultivars, Italia and Autumn royal, and we highlighted their different adaptation. Microarray analyses revealed that Autumn royal reacts involving only 29 genes, related to plant stress response and ABA/hormone signal transduction, to modulate the response to water deficit. Instead, cultivar Italia orchestrates a very broad response (we found 1037 differentially expressed genes) that modifies the cell wall organization, carbohydrate metabolism, response to reactive oxygen species, hormones and osmotic stress. For the first time, we integrated transcriptomic data with cultivar-specific genomics and found that ABA-perception and –signalling are key factors mediating the varietal-specific behaviour of the early response to drought. We were thus able to isolate candidate genes for the genotype-dependent response to drought. These insights will allow the identification of reliable plant stress indicators and the definition of sustainable cultivar-specific protocols for water management.


Results
Irrigation treatments and plant physiological parameters. Plants of AR and It were subjected to two different irrigation treatments from fruit set until the harvest: control FI and WD corresponding to 100% and 60% of the net irrigation requirements, respectively. For the It cultivar, an additional point of over-irrigation (OI) was tested, corresponding to an increment of 50% of water supply with respect to FI.
The starting ψ leaf value (around −1,0 MPa) corresponded to that found in grapevines from a region characterized by Mediterranean climate 29 . Irrigation treatment clearly affected vine water status as shown by the seasonal evolution of ψ leaf in the two cultivars (Fig. 1a). During the first irrigation cycle, similar ψ leaf values, ranging from approximately −0.8 to −1.0 MPa, were recorded for the vines treated with a deficit of irrigation. During the successive irrigation cycles, water supplies was reduced and a decrease of the ψ leaf of vines under reduced irrigation was detected, reaching minimum values of approximately −1.6 and −1.5 MPa for AR and It, respectively. Noteworthy, ψ leaf decreased more rapidly in AR than in It and at the end of the treatment differences in ψ leaf were higher in AR.
Excess of irrigation is often practised for table grape production; therefore, a condition of OI was also tested for cultivar It. As expected, the ψ leaf showed higher levels than in the FI (Fig. 1a). For leaf gas exchange, all parameters showed a decline in the thesis with reduced water intake and an increase of the intrinsic water use efficiency (Supplementary Note §1.1), thus confirming the stress status in both cultivars. The comparison of the www.nature.com/scientificreports www.nature.com/scientificreports/ physiological overall data (Supplementary Note §1.1) suggested that WD treatment resulted in better intrinsic water use efficiency in AR with respect to It.
Effect of different irrigation treatments on berry quality and production. We investigated the influence of the applied water stress on the productivity of the two cultivars and fruit quality at harvest time (Supplementary Note §1.2). In agreement with previously reported data 30 , we observed a negative influence of WD on cluster numbers and bunch weight, which resulted in a lower production per vine in both cultivars. Moreover, we also measured a reduction of berry size. Quality parameters such as sugar content, Ph and total acidity did not show significant variations, despite a global reduction of the sugar production per vine.
As previously reported 31 , AR is a black table grape variety mainly characterized by high content of tri-hydroxylated anthocyanins (Ant-3OH), even though consistent levels of di-hydroxylated anthocyanins (Ant-2OH) are also present. We found a significant increment of total anthocyanins content upon the WD application compared to FI (Fig. 1b). This trend was particularly evident either for Ant-3OH and for Ant-2OH anthocyanins (with an Figure 1. Effect of water deficit on physiological and qualitative parameters. (a,b) Evolution of leaf water potential (Yleaf) in Italia and Autumn royal during the entire seasonal irrigation. Arrows indicate sampling date for gene expression assays. Sampling was performed when Yleaf revealed water deficit stress. WD: water deficit; FI: full irrigation; OI: over-irrigation. Data are means ± S.D. *denotes significant (p < 0.01) difference from FI condition, as assessed by t test. (c,d) Anthocyanins content of Autumn royal table grapes experimenting two different irrigation treatments (FI vs WD). Values are expressed as mg/Kg of fresh weight (FW). FI: full irrigation; WD: water deficit. Data are means ± S.E. n = 3. °denotes significant (p < 0.05) difference.
Effect of water deficit on grapevine gene expression. In order to study the effect on grapevine gene expression of a reduction of irrigation supply in field condition, leaf apexes of plants of AR and It subjected to FI and WD conditions were collected when WD plants showed water deficit stress, two days after the first cycle of the differentiated irrigation treatment. We used these samples for microarray analyses.
Transcripts showing a fold change ≥2 with p < 0.05 were considered as differentially expressed genes (DEGs): (a) in each cultivar under water deficit (Supplementary Dataset 1); (b) between the two cultivars at both FI and WD conditions (Supplementary Dataset 2); (c) between OI and WD in It (Supplementary Dataset 3). Hierarchical clustering and Principal Component Analysis (Fig. 2) of gene expression data showed a wide variation of gene expression between both water conditions and cultivars, evidencing that the variation between genotypes was higher than between conditions. The comparisons between FI and WD conditions indicated a cultivar-specific response to water deficit (Fig. 2). Colors indicate transcriptional activation (red) or repression (yellow). The columns and rows represent samples and genes, respectively, that were grouped based on their expression profile. (b) Principal Component Analysis (PCA) depicting global gene expression profile of AR and It at different water conditions. The analysis highlights that variation between cultivars is higher than between conditions. (c) Venn diagrams show down-and up-regulated genes between different water conditions. The comparison between OI and FI in It does not reveal any significant DEGs. Italia significantly modulated a total of 1037 genes (316 up-and 721 down-regulated), whereas AR, despite an overall variation of expression profile, showed only 29 DEGs between FI and WD (21 up-and 8 down-regulated). Noteworthy, 20 of these genes were common to It responses, whereas, nine genes were modulated exclusively in this cultivar. (d) Venn diagrams show down-and up-regulated genes in AR compared to It. The higher variation of gene expression between cultivars is depicted at WD (the number of DEG is underlined). AR: Autumn royal; IT: Italia; WD: water deficit; FI: full irrigation; OI: over-irrigation.
Differences between cultivars strongly increased at WD compared to FI, as most of the DEGs between the two cultivars at WD are specific to this condition (2474 out of 3352) (Fig. 2).
With regard to the effects of an excess of water on gene expression profile, the comparison between OI and FI in It did not reveal any significant DEGs, whereas we identified a total of 3873 DEGs (1862 up-and 2011 down-regulated) between OI and WD (Supplementary Note §1.3).
We further validated the observed trend on apexes of seven DEGs by real time analyses on leaf and tendril of It at OI, FI and WD conditions and on leaf of AR at FI and WD conditions. The analysis confirmed the expression trend observed by microarray analysis on apexes (Supplementary Note §1.4).
The MapMan analysis provided an overview of the metabolic pathways and regulatory networks affected by water deficit in the two studied cultivars (Supplementary Note §1.5).
We used the Cytoscape plug-in ClueGo to analyze the DEGs in order to identify functionally grouped annotation networks (Fig. 3). Only significant (p < 0.005 or p < 0.001) terms belonging to GO biological process and Kegg ontologies were considered for the network analysis in order to select the most significant biological process and pathways affected by water deficit.
Cultivar-specific response to water deficiency. Networking analyses of the DEGs between FI and WD conditions revealed a cultivar-specific response to WD. In particular, It showed a predominant down-regulation of genes involved in the primary metabolism (DNA replication, carbohydrate metabolism, and cell wall organization), coupled with a modulation (up-or down-regulation) of genes involved in the responses to stress (osmotic stress, response to oxygen containing compounds, inorganic substances, and hormones) (Fig. 3). In contrast, in AR, the modulation of gene expression under WD was limited to a very little number of genes but most of them are specifically involved in the plant response to stress conditions (Fig. 3).
In order to increase the comprehension of the main functional networks affecting the different response to water deficit of It compared to AR, the total number of DEGs between AR and It at WD condition have been analysed. Despite the large number of genes studied (a total of 2474 genes: 1201 up-and 1273 down-regulated in AR compared to It), the analysis showed that most of them do not group in any significant network. Only a few significant pathways of DEGs between the two cultivars were identified: nitrogen metabolism, oxido-reduction process, RNA metabolism and modifications (Fig. 3).
To select, among the DEGs between cultivars, those specifically modulated in responses to WD, we intersected them with the DEGs between FI and WD conditions (Fig. 3). The genes common to these two groups (137 downand 233 up-regulated in AR compared to It) have been used for gene networking analysis. The results indicated that AR at WD showed a predominant up-regulation of genes involved in the following biological processes: cell wall organization, macromolecule metabolic process, amino-and nucleotide-sugar metabolism and a predominant down-regulation of genes involved in the response to oxygen containing compounds. These results agree with the modulation of genes belonging to these pathways in It under WD. Interestingly, the analysis indicated nitrogen metabolism as a significant pathway modulated between the two varieties, confirming its importance in determining the different response to WD of the two cultivars.
Main biological processes characterizing grapevine response to water deficit. Considering the overall results of network analyses, a complete list of DEGs belonging to the main pathways putatively involved in the modulation of grapevine defence responses to water deficit and their cultivar-specificity has been reported (Supplementary Dataset 4).
According to the important role of phytohormones in the regulation of plant stress response we found, respectively, 45 down-and 34 up-regulated genes in response to WD in cultivar It, whereas, four up-and one down-regulated genes were identified in AR (Supplementary Note §1.6). ABA, auxin and ethylene responsive genes were predominant among the DEGs. Three genes characterize the response of AR (up-regulated exclusively in this cultivar) and encode for pathogenesis-related proteins putatively associated to salicylic acid signalling. According to this, we found ABA response elements (ABRE) and ABRE-related motifs in 42 hormone-responsive DEGs (Supplementary Note §1.6). Interestingly, differences between AR and It were identified. In particular, ABRE elements were found in the promoter region of DREB1A (VIT_16s0100g00380) and two ERF5 encoding genes (VIT_16s0013g01060, VIT_16s0013g01050) in cultivar It, but the same regions in AR do not contain these elements.
As expected, under WD condition, we identified 22 up-and 23 down-regulated genes involved in osmotic and water stress response in It (Supplementary Dataset 4). Among them, six genes encode for dehydration responsive proteins (RD22, RD26, XERICO, DRS1) and transcription factors regulated by ABA (DREB1A, MYB102).
Noteworthy, two osmotin-like genes OSM34 generally associated to drought tolerance in other species resulted up-(VIT_02s0025g04340) and down-regulated (VIT_02s0025g04230). In contrast to the large number of genes differentially expressed in It, only two genes involved in osmotic stress response resulted differentially expressed in AR, both up-regulated: b-glucanase (VIT_205s0077g01150) and OSM34 (VIT_02s0025g04340), similarly to It.
Modulation of 41 genes involved in the response to oxygen-containing compounds was observed in It (Supplementary Dataset 4). They include genes involved in the generation of reactive oxygen species (ROS) as peroxidases, in the plant defence from ROS (as catalases, ascorbate peroxidases) and in the cell redox homeostasis (as thioredoxins). Only one gene involved in these pathways was differentially expressed (up-regulated) in AR: the peroxidase 5-like (RCI3: VIT_14s0060g00520).
In addition, a complete list of the transcription factors (TF) differentially expressed under WD, has been also reported (Supplementary Dataset 4). According to the evident differences in the response to WD between AR and It, we found 61 TF differentially expressed (35 down-regulated, 26 up-regulated) in It versus only one TF differentially expressed (up-regulated) in AR. Most of TFs belong to AP2-EREB, bHLH and MYB families (Fig. 4).
www.nature.com/scientificreports www.nature.com/scientificreports/ It only terms containing at least three genes were shown, this restriction was not applied for cv. AR. Nodes with up-or down-regulated genes are shown in red or green, respectively. (c,d) Network of genes differentially expressed between cvs. It and AR at water deficit condition (WD) and regulated under water stress, as indicated by Venn Diagrams. Yellow-and blue-circled numbers represent, respectively, selected down-and up-regulated genes between AR vs It. Nodes with up-or down-regulated genes (AR was used as reference) are shown in blue or yellow, respectively. Data are visualized as clusters distribution network (Cytoscape, ClueGO App). Only significant (p < 0,005) terms belonging to GO biological process and Kegg ontologies were shown. The node size is proportional to the term significance. The colour gradient shows the proportion of up-and down-regulated genes associated with the term. Equal proportions of both clusters are represented in gray. AR: Autumn royal; IT: Italia; WD: water deficit; DEG: differentially expressed genes.
www.nature.com/scientificreports www.nature.com/scientificreports/ Inter-cultivar comparison of genomic data. In order to find genotype-specific differences putatively associated to the observed different response to WD, we looked for AR and It specific genomic variations among previously published data 19 searching for polymorphic genes between the analysed cultivars involved in abiotic stress. More in detail, we focused our attention on copy number variant (CNV) regions and single nucleotide variants (SNVs) affecting gene transcription. By looking at the annotated copy number for each cultivar, we identified 1249 subregions whose AR and It copy number difference was higher than 0.5 (Table 1, Supplementary Dataset 5). Notably, this indicates that about 5% of the grape genome is variable between the two cultivars. Interestingly, the extracted subregions were found to overlap with 1499 genes. Interestingly, four regions showed very high CN differences between AR and It (>20); these regions had already been described as hyper-duplicated regions by Cardone et al. 19 . Two of these contain genes belonging to NADH dehydrogenases family and for both of them AR showed the higher CN with respect to It. Genes of which the copy number difference was higher than 5 (147 genes corresponding to 111 regions), were mainly found to belong to gene families involved in stress responses (e.g. NBS-LRR and Ankyrin domain proteins) or to relate to transposable elements.
We deeply studied the CNVs and SNVs containing genes involved in pathways of relevance in the response to abiotic stresses and drought in particular (26 CNVs and 49 SNVs) (Supplementary Dataset 6). Among these, HVA22A and RD22 (related to ABA signalling and water stress response) were found more duplicated in It compared to AR. Noteworthy, one RD22 (VIT_04s0008g04120) and two HVA22A (VIT_03s0097g00470 and VIT_03s0038g01650) genes also showed cultivar specific SNVs causing stop codon gains.
In order to gain insights into the genetic bases of cultivar-specific response to water deficit, we compared the so highlighted genomic differences with the DEGs found in our comparisons (Supplementary Dataset 8). We found 158 DEGs showing differences in CN. Looking at the functional annotation, we found enrichment of the following categories: primary metabolism (37 DEGs), stress response (including NBS-LRR and response to stimulus) (32 DEGs), secondary metabolism (terpenoid and phenylpropanoid metabolisms) (14 DEGs), transport (13 DEGs), signalling pathway (12 DEGs). 54% revealed a direct correlation between CNV and expression level (i.e.: the higher CN corresponded to UP regulation). A selection of the genes putatively involved in the cultivar-specific response to water deficit has been reported in Table 2. Interestingly, the different CN of HVA22 and RD22 might explain their differential expression. Of note, a lineage-specific evolution of the RD22 family has been described with the biggest expansion in grapevine 32 . We found 18 paralogues in the reference genome and identified CNVs between AR and It in ten of them 19 . VIT_04s0008g04140 and VIT_04s0008g04150 showed also down regulation in AR with respect to It (Table 2).
Likewise, we compared SNVs with DEGs and we found 336 SNVs affecting the function of 295 DEGs. Among these, to identify candidate genes responsible for the observed genotype-dependent response to water stress, we specifically looked at SNVs corresponding to inter cultivar DEGs and we found 22 genes (Table 3).

Discussion
Water deficit represents the main environmental constraint for growth in grapevine. In response to drought, different cultivars adopt different strategies to limit the effect of water deficit, although the genetic and physiological origins of these differences are still debated 8,18,33,34 .
In the present work, we combined physiological studies with transcriptomics and genomics in order to investigate the different ability of two table grape varieties to respond to water stress and highlighted new clues about the genetic bases of these differences. Most of the previous literature data are based on experiments under controlled conditions, and they are mostly focused on wine grape varieties 35 . We, instead, tested the effects of a reduced irrigation directly on the field in order to investigate how the mentioned different ability reflects an adaptation of the cultivars to the growing conditions.
We demonstrated a negative effect of WD on some important traits for table grape production such as bunch weight and berry size, confirming previous results 30 .
Besides water stress does not induce a constant response in flavonols and flavanols 36,37 , anthocyanins exert homogeneous behaviour increasing under WD 35,38,39 . For this reason, we decided to focus our attention on the different classes of anthocyanins in AR (Fig. 1b) and we demonstrated that the WD affects positively anthocyanin content, especially on anthocyanins-3O-glucosides and 3O-acetyl-glucosides. Moreover, we showed that the relative amount of ant-2OH increased more than the ant-3OH one (2.1 vs 1.2 folds, respectively) (Fig. 1b). Instead, previous studies have described a shift in biosynthesis toward a higher proportion of 3′,4′,5′-trihydroxylated compared to 3′,4′-dihydroxylated anthocyanins through the up-regulation of flavonoid-3′,5′-hydroxylase 37,39,40 .
It has lately been annotated, however, that also phenylpropanoids and terpenoid pathways can take part in the berry response to WD in non-pigmented berries, suggesting that an overproduction of monoterpenes is part of the fruit response to drought 20 . We accordingly found structural variations between It and AR in DEGs belonging to terpenoid and phenylpropanoid gene families (Tables 2 and 3).
Recently, many comparative studies have addressed the topic of water stress response in grapevine at a molecular level, using different experimental approaches 8,14,[16][17][18]20,41 . With respect to previous data, we focused our attention on the early response to water deficit. Besides differences in experimental settings, pedo-climatic and growing conditions, other than tissues analysed, our data confirmed that WD induces modulation in genes related to response to stimuli, response to abiotic stress, ABA response, protein and carbohydrate metabolisms, nitrogen metabolism, and ROS response, thus revealing the importance of such pathways in the response to water stress. Notably, both the analyzed varieties showed modulation of genes related to osmotic stress response and those related to the primary immune plant system such as the defence proteins (PR1). This finding demonstrated that the mechanisms acting in the primary response to both biotic and abiotic stresses are shared as already highlighted for other kind of stresses 18 .
However, comparing our data with those already available 42 we found qualitative and quantitative differences in the genes specifically modulated in response to WD. Indeed, only 8-10% of the DEGs found in our experiments overlap with those recently selected as responsive genes in Montepulciano and Sangiovese 8 . This strongly supports the genotype-dependent response to WD.
In accordance with this, we found strong differences in the WD response between AR and It. AR showed a limited and specific response, involving the modulation of genes specifically related to plant defence mechanisms, including drought-responsive genes such as desiccation proteins.
The strong differences observed between AR and It under WD stress might also depend on a different timing of response between the two cultivars: AR could activate later a more extensive response to WD, similarly to what found for the anisohydric cultivar Sangiovese 8 . This suggests that the genotype-specific responses to WD need to be investigated at the early phases after WD. (2019) 9:2809 | https://doi.org/10.1038/s41598-019-39010-x www.nature.com/scientificreports www.nature.com/scientificreports/ This behaviour could probably reflect a better adaptation of AR to the WD conditions. Indeed, adaptation and resilience to water stress, such as the extremely limited response in the early phase found in AR, could be considered more advantageous. In this way, the plant could activate its defence responses more gradually -only if the WD condition is prolonged -and this could avoid investing much resources and energy if not strictly necessary.
In order to understand the molecular basis of such kind of genotypic -specific response, we deeply analyzed NGS data belonging to the studied cultivars and we identified structural variants in stress-related genes. Many genomic variations were also correlated to the expression differences, and thus putatively associated to the different genotypic-specific behaviour observed in response to WD. Genomic data obtained from different plant species have revealed that plant genomes are highly plastic as a result of different mechanisms such as genome duplication, segmental duplications, and transposable elements (TEs) mobility. Moreover, contrary to what was previously thought, it is now clear that the genetic plasticity is useful for the adaptation to a changing environment 43 .
Among the most polymorphic genes between AR and It, we found that the higher CN differences affected some well-known stress-related gene families, such as ankyrin repeat proteins belonging to the RING finger family, recently described as specifically related to drought response in Arabidopsis thaliana 44,45 . Indeed, the Arabidopsis thaliana ankyrin DRA1_(At4g03500) was identified as a negative regulator of drought tolerance 45 . Interestingly, there are 18 orthologs of this gene annotated in grape (Gene Tree EPlGT00940000163197, Ensembl Plants release 41) and 12 of these had shown a differential expression in response to WD between AR and It ( Table 2, Supplementary Dataset 2). Of note, the gene VIT_12s0059g00050, down-regulated in AR in response to WD, showed a lower CN in AR with respect to It. Additionally, we also found five SNVs (of which four specifically in AR) causing a gain of a stop codon, in three genes coding for ankyrin repeat proteins (Table 3). These data, in agreement with what previously known about ankyrin proteins in Arabidopsis thaliana 44,45 , reveal the role of these proteins in the genotype-dependent response to drought. Stress conditions could have induced gene duplications and these events could create genome plasticity leading to a different ability to respond to the changing environment.
Antioxidant enzymes, metabolites, transcription regulators, and cross-talk with hormones prompted by abiotic stress conditions are crucial to ensure the right antioxidant homeostasis, achieving a positive balance between photosynthesis and respiration 33 . According to this, among the genes showing CNVs and SNPs directly related to the expression modulation in early stress responses we also found genes involved in photosynthesis, energetic metabolism, electron transport and ROS scavenging pathways as NADH dehydrogenases and quinone oxidoreductases (Supplementary Dataset 8).  www.nature.com/scientificreports www.nature.com/scientificreports/ Of note, we also found more than 200 TEs showing different CN in AR and It. It is well known that stresses induce activation of TEs in plants and that the resulting genome plasticity is fundamental to survive in adverse environments 46 . Rocheta and colleagues 18 described differential expression of TEs in grapevine in a stress-specific manner, suggesting a role of TEs in grapevine stress response and adaptation response to abiotic stresses.
According to the importance of phytohormones in the regulation of plant responses and adaptation to abiotic and biotic stresses [47][48][49][50] , and the key role of ABA in the modulation of the complex hormonal network in response to WD 51,52 , we identified numerous differences between AR and It not only at transcriptomic but also at genomic level. Indeed, we revealed CNVs in 30 genes of phytohormone signalling and perception, most of them ABA-dependent (Supplementary Dataset 6). Our results confirmed that ABA-mediated perception and response might be the major responsible for the varietal-specific behaviour observed applying water stress.
As an example, higher CN in the gene RD22, in It, is also coupled with a significantly higher expression of these genes in cultivar It compared to AR ( Table 2). The Responsive to Dehydration 22 (RD22) has been recently described as a link between ABA signalling and abiotic stress responses 32 by maintaining cell integrity under stress conditions 53 . Notably, a big cluster of paralogous copies of RD22 on chromosome 4 has been described 32 and it is a hot spot for subsequent duplication/deletion events mediated by unequal crossing-overs, leading to CNV in this region 19 . According to Matus and colleagues, we found that different members of the grape VvRD22 group present a different expression during the early response to water stress. In addition to this, we also found CNVs in this gene family between AR and It, thus supporting the role of RD22 in ABA-mediated response in a genotype-dependent way.
Moreover, we found that the ethylene responsive factors (ERFs) resulted the most represented TF family differentially expressed under WD, counting 12/62 genes. Nine of them, possess ABA-responsive elements (ABRE or ABRE-related) in their promoters. Eight of these ABA-responsive ERFs resulted down-regulated in response to WD stress, whereas, one was up-regulated (Supplementary Dataset 4). This is in agreement with literature data on different plant species, showing that ERFs might regulate the drought stress responses in both directions: overexpression of some ERFs enhance the tolerance to water stress, whereas others could act as repressors in drought stress responses [54][55][56][57][58] .  Table 3. SNPs occurring in genes differentially expressed between AR and It selected as candidate genes for grapevine response to water deficit stress.
www.nature.com/scientificreports www.nature.com/scientificreports/ Other important players in the regulation of ABA responses to WD include the type 2 C protein phosphatases as ABI1. These proteins regulate the sensitivity of guard cells K+ channels to this phytohormone, indeed abi1 and abi2 arabidopsis mutants are insensitive to ABA and unable to close stomata 59 . We showed that WD stress increases ABI1 transcription only in It, however, it is not possible to depict the consequent effect on ABA regulation, considering the complex regulation occurring at this stage of ABA signalling.
ABF2, an ABRE-binding bZIP factor, is another component of ABA signalling likely involved in the adaptive processes to abiotic stresses as drought. Overexpression of this gene enhances resistance to water stress and induces stomatal closure 60 . Interestingly, our results indicate an up-regulation of ABF2 in It in response to WD condition. The increase of its expression level might be responsible of the transcriptional activation of other ABA-responsive genes related to stress response. According to this hypothesis, we found ABRE motifs in the promoter of drought related genes as RD22, ERF, DREB1, DDF2. Interestingly, differences in the presence/absence of ABRE motifs were also found between AR and It, they might affect the ABA-mediated response of the two cultivars under WD.
A hypothetical scheme of the ABA-mediated mechanisms involved in responses to WD stress in cultivars AR and It is depicted in Fig. 5. Our results suggest that the increase of ABA and/or of ABA perception in cultivar It could be responsible for the transcriptional induction/repression of signalling genes and transcription factors, such as those belonging to AP2/AREB and MYB families. They might affect the transcriptional regulation of drought-related genes. In contrast to the 25 ABA-responsive genes differentially expressed in It in response to WD, only two genes resulted differentially expressed in AR highlighting that ABA perception is strongly genotype-dependent.

Conclusion
In conclusion, our analysis confirmed that drought stress response is strictly genotype-dependent and we can infer that the observed responses are dependent on the specific adaptation of each cultivar to environmental conditions. The very limited reprogramming in AR suggests it is able to react more rapidly and efficiently than It limiting the energy dissipation, and thus confirming a better adaptability.
For the first time, by investigating genomic variants, we were able to identify candidate genes related to genotype-specific response to water deficit in grapevine. We highlighted that structural variants and TEs are some of the primary sources of genomic plasticity, strongly affecting genotype adaptation ability.
Our study represents a step toward the definition of a more rational and efficient use of water resource in viticulture for table grape production. Vines were cane pruned (two canes every 12-15 buds per vine) with free-growing shoots (complete overhead canopy separated from fruit). Row orientation was North-South.

Material and
The climate is typical Mediterranean semi-arid, characterized by hot dry summers (although short periods of heavy rainfall may occur) and mild rainy winters. The data were collected during the year of the experiment as reported by 30 . Irrigation treatments. Two irrigation treatments, based upon a percentage of the net irrigation requirements [NIR = ET c (crop evapotranspiration) -Effective rainfall] from fruit set till harvest, were applied: control full irrigation (FI) and water deficit irrigation (WD) at 100 and 60% of NIR, respectively. For the It cultivar an additional point of over-irrigation (OI) was tested, corresponding to an increment of 50% of water supply with respect to FI. ETc was estimated using varying crop coefficients According to the typical practice adopted in the Apulian region, the vines were drip-irrigated by means of irrigation lines installed 180 cm above the soil surface with drippers spaced 70 cm apart and set to supply water at a constant pressure with two 8 L h −1 drippers vine −1 . Except for the irrigation treatments, all the other standard cultural practices in the vineyard were applied equally to all vines.
Vine water status and leaf gas exchange. During the steady period of the water potential diurnal curve (generally between 12.30 and 13.30 h), the midday leaf water potential (Ψ leaf ) was measured seven times during the irrigation period. Measures were made one day before the irrigation application (lowest water availability) and at the mid-cycle, 2-3 days after irrigation (greatest water availability). Two mature exposed leaves per vine (opposite to the cluster, in the middle shoot of the fruit cane) were selected from the canopy, enclosed in plastic bags, and quickly sealed; then, the petioles were cut within 1-2 s and their Ψ leaf was measured immediately in the field by a model 600 pressure chamber instrument (PMS Instrument Company, Albany, USA). The time between leaf excision and chamber pressurization was generally <10-15 s 62 . Gas exchange (leaf photosynthesis rate, stomatal conductance to water vapor, and transpiration rate) was measured on healthy, fully expanded mature leaves (2019) 9:2809 | https://doi.org/10.1038/s41598-019-39010-x www.nature.com/scientificreports www.nature.com/scientificreports/ exposed to the sun (one leaf on each of 5 vines per treatment), from main shoots located on the exterior canopy (see more details in Supplementary note).
Fruit quality analyses. Four 7-bunch samples of AR and It for each treatment, respectively, were randomly harvested at commercial maturity (September 8, 2013) according to a sugar-acid ratio >25. Twenty berries from each bunch were collected, weighed, and their firmness was measured using a deformation tester (Digital Fruit Firmness Tester, Forlì, Italy). Finally, juice was extracted from each sample and used to measure pH, total soluble solids (TSS) as °Brix, and titratable acidity as described in detail in Supplementary note.
Moreover, in the case of AR samples, the anthocyanins profile was also determined. Anthocyanins extraction procedure was adapted from the method previously reported 63 as described in the Supplementary note.
Three individual vine replicates were assigned to each experimental treatment using a randomized block design. Differences in the quality parameters between the irrigation treatments (FI vs WD) of the two cultivars were tested through pairwaise Student's t-tests by using STATISTICA 8.0 (StatSoft Inc., Tulxa, OK) package and the statistical tools available in excel.

Microarray analyses and validation of expression data.
Total RNA was extracted from 0.1 g of shoot apex collected at 10% veraison with Total RNA Isolation Mini Kit (Agilent Technologies). RNA integrity was assessed by automated gel electrophoresis on 2100 Bioanalyzer (Agilent Technologies, Amstelveen, Netherlands). cDNA synthesis, labelling and hybridization were performed according to the manufacturer's instructions (version 6.9.1, Agilent Technologies). Hybridization was carried out on an Agilent custom array. Starting from the assembled V. vinifera L. genome sequence (http://www.genoscope.cns.fr/externe/GenomeBrowser/Vitis/), we  70 . The table compares copy number variation (CNVs) and mRNA expression of ABA-responsive genes. + indicates higher number of CNVs or mRNA expression in It compared to AR. The ABA-responsive genes HVA22A (VIT_03s0132g00070, VIT_03s0132g00080) and RD22 showed higher CNs in It compared to AR, according to their higher expression in It, whereas GTG2 (VIT_07s0005g06120) and ABI1 showed higher CNs in It but did not resulted differentially expressed between cultivars. Genes whose promoters contain ABRE or ABRE-related motifs are underlined. In parentheses it is indicated the number of genes. downloaded annotated transcripts sequences, and using the online tool eArray provided by Agilent Technologies S.p.A., we designed a custom array containing 44 K probes, corresponding to about 26k annotated genes plus EST and transcripts reported in literature as candidate genes for important traits and QTLs available at NCBI databases (www.ncbi.nlm.nih.gov/gene/). The array images were analysed using Agilent Feature Extraction software version 12.0 (Agilent Technologies, Santa Clara, CA).
A selection of seven DEGs was further validated by real time assays. cDNA was prepared as reported in the Supplementary note. Three biological replicates (different plants) were analysed for each sample. All reactions were performed in triplicate. Relative amounts of all mRNAs were calculated using the 2 −ΔΔCt method 64 , where ΔCt = Ct(target gene) − Ct(reference gene). The housekeeping gene actin was used as an endogenous reference for normalization.
Analysis of transcriptomic data. Microarray expression data were processed and analysed using the R package limma (R version 3.1.2, limma version 3.23.2) 65 . Transcripts showing a fold change ≥2 with p < 0.05 were considered as differentially expressed. Hierarchical clustering on both entities and conditions was performed using Euclidean distance metric and Ward's linkage rule.
Overviews of metabolic and regulatory pathways were obtained using MapMan software v3.6.0. Gene networking analyses were performed using Cytoscape platform v3.4.0 66 and ClueGo plug-in v2.3.3 67 . The analysis was performed using Vitis vinifera as reference. Enrichment/depletion of terms and groups was performed by two-sided hypergeometric test, corrected with Bonferroni step down method. Kappa Score Threshold of 0.4 was applied. Only significant (p < 0.005 or p < 0.001) pathways belonging to the gene ontology (GO) biological process and Kegg ontologies were considered. GO terms from level 3 to level 8 of GO hierarchy were selected. Kappa Score grouping was applied. Data were visualized as clusters distribution networks. V1 version (http://genomes.cribi.unipd.it/grape/) was used for the annotation of Vitis vinifera genes as the most diffused in public databases.
In order to have a comprehensive view on the DEGs belonging to specific pathways of interest, an integrated approach based on multiple annotation methods was used to identify significant genes belonging to target pathways: response to hormone and oxygen-containing compounds, osmotic stress, nitrogen metabolism and photosynthesis. In particular, GO based annotation of Cytoscape was enriched using both Vitis vinifera and Arabidopsis thaliana as references. In addition, the annotation of MapMan based on GO and enzymatic code (EC) was also included.
The Vitis vinifera section of the PlnTFDB -Plant Transcription Factor database 68 was queried for the correct annotation of the transcription factors differentially expressed in our experiments. Their assignment to specific families was performed after conversion to GSVIVP annotation codes.
Mining of genomics data. Copy number and single nucleotide variations between AR and It cultivars were retrieved from data produced by Cardone and co-authors 19 (76-bp paired-end libraries sequenced using the Illumina GAIIx platform) (Sequence Read Archive, ID: SRP009057).
To calculate and compare the copy number of each region of the grape genome between the two studied cultivars, NGS reads were aligned to the reference using the mrFAST aligner (Alkan et al., 2009). The absolute CN of non-overlapping windows of 1 Kbp unmasked sequence (KbUS) was then calculated by using mrCaNaVaR version 0.31. Duplicated and deleted segments were predicted based on 5 KbUS sliding windows 69 : regions with at least five consecutive windows having a CN > 2.5 were identified as segmental duplications, while regions with low read-depth of coverage (CN 1.5 and below) were identified as deletions. Large CNVRs were identified as regions > 10 kbp showing gain or loss using a threshold of L2R > 0.25 for amplifications and L2R < 0.25 for deletions. Here, all CN differences between AR and It were selected as significant if showing a variation of at least 0.5 copies.
The SNV analysis was performed starting from the data collected in the Supporting Information published by 19 (SnpEff output). After removing the low-quality calls, we searched for SNVs present in both AR and It when compared to pinot noir reference genome 19 and showing a different genotype. The genotype is defined as 0/0 for homozygous reference, 0/1 or 1/0 for heterozygous, 1/1 for homozygous alternative.

Data Availability
All the data produced and described in this manuscript are fully available as Supplementary Datasets. The associated plant material and protocols are fully available in the section Materials and Methods and in the Supplementary note. The gene expression data have been submitted to GEO. The sequencing data used for the genomic analysis are available at SRA under the ID: SRP009057.