WS imposition and sampling strategy
Montepulciano and Sangiovese potted vines were subjected to a water deficit at 40 % of maximum water availability from fruit-set to veraison (onset of ripening) and were then fully re-watered (Fig. 1a and c). As previously observed by Palliotti et al. [13], water stress conditions induced a faster and more pronounced basal leaf yellowing and shedding in Sangiovese than in Montepulciano vines. Daily minimum and maximum temperatures and rainfall were monitored during the experiment (Fig. 1b). Three pools of fully expanded leaves, sampled between nodes 14 and 16 of the primary shoots of well-watered (WW) and water-stressed (WS) vines, were sourced from both varieties 2, 6 and 27 days after WS was imposed, to assess the short and long-term effect of WS on vine physiology, and were used for the analysis of all physiological, biochemical and transcriptional characteristics.
WS differentially affects leaf gas exchange and ABA content in Sangiovese and Montepulciano
After 2 days of WS, the midday leaf water potential (Ψl) fell to −0.86 MPa in the Sangiovese vines but to only −0.58 MPa in the Montepulciano vines (Fig. 2a). After 6 days of WS, the values fell to −1.20 and −0.91 MPa, respectively, and these values were maintained until 27 days after WS was imposed (Fig. 2a). A similar decreasing trend was observed for the net photosynthesis (Amax) and stomatal conductance (gs) (Fig. 2b and c). Although the Amax and gs rates of the WW vines of both cultivars did not change during the experiment, the WS Sangiovese leaves showed higher Amax and gs values than WS Montepulciano leaves after 2, 6 and 27 days of WS (Fig. 2b and c). The WS Sangiovese vines also showed a significant increase in the intrinsic water use efficiency (WUEi) after 2 and 27 days of WS, whereas there was no increase in the WS Montepulciano vines (Fig. 2d). After 6 days of WS, the WUEi was significantly higher in both cultivars (Fig. 2d).
Regardless of the cultivar, the Area parameter was halved for the duration of WS, indicating a drastic reduction of the plastoquinone pool size on the reducing side of PSII (Fig. 2e). Conversely, the Fv/Fm, Fo and Fm parameters were unaffected in both cultivars during the experiment (Additional file 1). The ABA content of the WS Montepulciano leaves increased significantly compared to WW vines, whereas there was no change in the ABA content of the WS Sangiovese leaves compared to the WW controls (Fig. 2f).
WS affects the leaf pigment content in Sangiovese and the de-epoxidation state in Montepulciano
The total chlorophyll content (Chltotal) of the WS Sangiovese leaves increased significantly compared to the corresponding WW leaves, particularly reflecting an increase in the levels of Chl a, whereas the WS Montepulciano leaves showed no significant changes (Additional file 1).
The analysis of xanthophylls showed no significant variation in the levels of individual carotenoids, the de-epoxidation state (DEPS) or the Cartotal/Chltotal ratio in either cultivar after 2 days of WS (Additional file 1). However, after 6 days the WS Sangiovese leaves showed a significant increase in the violaxanthin, antheraxanthin and zeaxanthin (VAZ) pool and total carotenoids (Cartotal) compared to WW leaves (Additional file 1), due to the accumulation of β-carotene, lutein, antheraxanthin and violaxanthin. The WS Montepulciano leaves showed a significant increase in the DEPS compared to the corresponding WW vines, due to the loss of violaxanthin but the simultaneous accumulation of zeaxanthin (Additional file 1).
After 27 days, the Cartotal and VAZ pool increased significantly in WS Sangiovese leaves due to the accumulation of antheraxanthin, zeaxanthin, neoxanthin, lutein and β-carotene (Additional file 1). Although the DEPS increased by almost 100 % in both cultivars under WS, the Montepulciano cultivar nevertheless achieved full activation of the de-epoxidation process due to the loss of violaxanthin and a concomitant increase in the levels of zeaxanthin and antheraxanthin (Additional file 1).
Finally, the WS Montepulciano leaves displayed a significant increase in both H2O2 concentration and catalase (CAT) activity during the WS period in comparison to the WW vines but no such changes were observed in WS Sangiovese leaves (Additional file 1).
Whole genome transcriptional analysis in leaves subjected to WS
The leaf transcriptome data set of both varieties after 2, 6 and 27 days of WS was initially screened by significance analysis of microarrays (SAM, 12 groups, FDR = 0.1 %) to select genes that were differentially modulated under our experimental conditions (18,413 genes). Analysis of variance (ANOVA, 12 groups, α = 0.01, standard Bonferroni correction) was applied to transcripts positive in the previous SAM experiment in order to skim off the most significantly modulated transcripts (5947 genes, Additional file 2). A PCA was used to verify the consistency of biological replicates and to generally inspect the transcriptomes of the Montepulciano and Sangiovese varieties under WS (Fig. 3a and b). In both PCA plots, PC1 explained ~44 % of the total data set variability and mostly reflected differences among the three sampling points and, within a single sampling point, differences between the WW and WS samples in both varieties. PC1 loadings clearly showed that the dynamics of leaf stress responses are different in the two varieties (Fig. 3c and d). Indeed, the reaction of the Montepulciano transcriptome towards WS started at the second sampling point and continued in the third with gradually increasing intensity, whereas the Sangiovese transcriptome began to react only at the third sampling point, albeit with a stronger shift (Fig. 3c and d).
To evaluate differences in gene expression between Montepulciano and Sangiovese under WS, we focused on changes in expression profiles of genes scoring a fold change (FC) ≥2 between the WW and WS vines of each variety at each sampling point. We identified 1188 genes using this approach (Fig. 3e, Additional file 3). To identify further genes modulated by WS, the same statistical procedure was followed but this time considering only the WS and WW categories regardless of the genotype (six-class SAM and ANOVA). We found 437 modulated genes shared between Montepulciano and Sangiovese, 300 of which (~67 %) were already present among the 1188 genes identified in the initial statistical strategy (Fig. 3e). Hence, we found a total of 1325 genes modulated by WS in at least one of the two cultivars, among which 1034 were functionally annotated (Fig. 3e, Additional file 3). Overall, these 1034 stress-modulated genes were particularly enriched, as expected, in functional categories related to stress such as “Response to abiotic stimulus”, “Death” “Cell death” and “Protein metabolic and modification process” (Fig. 3f).
Differences in the response to WS between Montepulciano and Sangiovese were investigated in more detail by applying the Short Time-series Expression Miner (STEM) clustering method [26] to the 1034 stress-modulated genes. This enabled us to visualize groups of genes whose differential expression between WS and WW samples also differed significantly between the two genotypes (Additional file 4). Figure 3g shows that some Montepulciano transcripts, clustered accordingly with their expression profiles (right side), displayed a significantly different expression profile in the Sangiovese cultivar (left side).
ABA-related genes
Many ABA-related genes modulated by WS were differentially expressed in the leaves of the two varieties, including β-carotene hydroxylase VvBCH1 (VIT_02s0025g00240) which was induced from the onset of WS in Montepulciano but only after 27 days in Sangiovese, and an ABA glucosidase (VIT_17s0000g02680) which was induced 6 and 27 days after the onset of WS in Sangiovese but only after 27 days in Montepulciano (MP 8 → SG 25 in STEM analysis, Figs. 3g and 4). The ABA-responsive bZIP transcription factor VvbZIP25 (ABA Insensitive 5, VIT_08s0007g03420) [27] was upregulated in both varieties but more rapidly in Montepulciano (Fig. 4). Two transcripts encoding membrane proteins of the AWPM-19-like family (VIT_05s0049g02240 and VIT_05s0020g02470) whose levels dramatically increase when the intracellular concentration of ABA increases [28], were upregulated after 6 days of WS in Montepulciano leaves but only after 27 days in Sangiovese (MP 11 → SG 8 in STEM analysis, Figs. 3g and 4). These data agree with the significant increase in the H2O2 accumulation and CAT activity in WS Montepulciano leaves at any sampling point compared to WW controls (Additional file 1). The Sangiovese leaves showed no differences in H2O2 levels or CAT activity under WS (Additional file 1). The main negative regulator of the stomatal closure pathway, HT1 (High leaf Temperature 1, VIT_17s0000g08240) [29] was repressed after 2 and 6 days in Montepulciano, but only after 27 days in Sangiovese leaves.
Abiotic stress-related genes
WS triggered abiotic stress-related transcriptional responses in the leaves of both cultivars, but these genes tended to be regulated more strongly in Sangiovese (Fig. 4). Genes encoding dehydrins, osmotines, thaumatins, chaperones, cold-induced proteins and senescence-associated proteins were strongly upregulated. Eleven genes involved in ROS scavenging during the oxidative burst [30] were upregulated in both cultivars, including those encoding ascorbate peroxidase, glutaredoxin, peroxidase, glyoxal oxidase, peptidil-prolyl-trans isomerase and thioredoxin. Genes involved in the oxidative stress-induced protein damage repair pathway [31] were more strongly induced in Sangiovese leaves. Twenty-four heat shock and heat shock-related proteins (HSPs) were differentially modulated in each variety, although they were upregulated after 27 days in Sangiovese but suppressed after 2 days and upregulated from 6 days onwards in Montepulciano (MP 11 → SG 8 in STEM analysis, Figs. 3g and 4).
Carbohydrate metabolism-related genes
Drought-stressed plants accumulate a large amount of water-soluble carbohydrates, which are used as osmolytes to maintain leaf cell turgor, protect membrane integrity and prevent protein denaturation [32]. The vacuolar invertase VvGIN2 (VIT_02s0154g00090) was induced in Montepulciano at the onset of WS but was delayed in Sangiovese (Fig. 4). The starch-degrading enzyme α-amylase (VIT_03s0063g00450) was induced in Montepulciano but not Sangiovese, whereas ADP-glucose pyrophosphorylase (VIT_03s0038g04570) was downregulated from the first sampling point in Sangiovese but only at the last sampling point in Montepulciano (MP21 → SG4 in STEM analysis, Figs. 3g and 4). Seven galactinol synthases and one galactinol-raffinosegalactosyl transferase were upregulated in both varieties, but more strongly in Sangiovese (MP25 → SG8 in STEM analysis, Figs 3g and 4). Three trehalose-6-phosphate synthases were upregulated and four trehalose-6-phosphate phosphatases were downregulated in both cultivars (Fig. 4). Finally, a raffinose synthase (VIT_17s0000g08960) was downregulated only in Sangiovese at the first and second sampling points but was induced at the third sampling point, whereas no variation in transcript levels was detected in Montepulciano leaves indicating that the type of water-soluble carbohydrates that accumulate during WS can vary among different cultivars.
Switch genes are putative negative biomarkers of WS
In order to identify putative molecular markers of WS in grapevine, we collected the genes that are differentially expressed between WW and WS leaves at each sampling point for each genotype separately and then applied a t-test (p < 0.01) and filtered the genes with a FC ≥2 when WS and WW leaves were compared. We found that both genotypes were characterized by a small number of differentially expressed genes at the first two sampling points compared with the third point (Fig. 5a and Additional file 5). After 2 days of WS, 181 genes were differentially expressed in Montepulciano leaves and only 59 in Sangiovese leaves. Interestingly, at this time point, both cultivars were characterized by a higher number of downregulated rather than upregulated genes, and only mitogen-activated protein kinase kinase kinase 15 (MAPKKK15; VIT_10s0116g01230) was downregulated in both cultivars. The higher number of differentially expressed genes in Montepulciano leaves provides evidence that this cultivar responds to WS more quickly than Sangiovese at the transcriptomic level, as also revealed by PCA (Fig. 3a–d). After 6 days of WS, 156 genes were differentially expressed in Montepulciano and 386 in Sangiovese. At this time point, there were more downregulated than upregulated genes in Sangiovese leaves (214 vs. 172), but the opposite trend was apparent in Montepulciano leaves (66 vs. 90).
By the third sampling point (after 27 days of WS) there were more upregulated than downregulated genes in both cultivars. Interestingly, the upregulated genes were more strongly induced in Sangiovese compared to Montepulciano leaves, whereas the downregulated genes were more strongly suppressed in Montepulciano compared to Sangiovese leaves. This suggests that the slower response of Sangiovese leaves to WS is balanced by the stronger induction of the response genes. We found 169 differentially expressed genes shared between the two cultivars (Fig. 5a and Additional file 5), suggesting that responses to WS become more aligned between the cultivars at the final sampling point.
WW and WS leaves from each cultivar after 27 days of WS were next analyzed using our recently published integrated approach based on topological co-expression networks to identify common putative key regulators of WS [33]. A comparison of the WW and WS leaf transcriptomes regardless of genotype revealed 1236 genes that are differentially expressed between WW and WS leaves (p < 0.08; FC >1.7). We found that 765 genes were upregulated and 471 were downregulated under WS (Additional file 6), confirming that 27 days of WS predominantly causes gene activation rather than suppression. The coexpression network, based on Pearson correlations, comprised 1236 nodes and 202,422 edges (Additional file 7). By applying the date/hub classification system to define the topological proprieties of the network, we identified 405 Fight-club hubs and 298 switch genes (Fig. 5b and Additional file 6). As previously reported [33], switch genes are characterized by a pronounced negative correlation with the expression profiles of neighboring genes outside their own group in the network, and therefore represent putative key regulators of leaf transcriptome remodeling during the shift from the WW to the WS environment. Interestingly, we found only four genes expressed at low levels in WW leaves but upregulated in WS leaves, whereas most of the switch genes were downregulated in WS leaves. Among the four upregulated genes, we identified heat shock transcription factor B2A (VIT_10s0597g00050), histone H2B2 (VIT_06s0061g00870) and the MYB floral symmetry gene DIVARICATA (VIT_04s0008g00900) whereas the remaining gene did not provide a match (VIT_09s0002g00630) (Fig. 5c).
Among switch genes downregulated in WS leaves we found many representing the flavonoid biosynthesis pathway, including VvMYBPA1 (VIT_15s0046g00170) and its target VvANR (VIT_00s0361g00040), VvCHS3 (VIT_05s0136g00260), Cytochrome b5 DIF-F (VvCytoB5; VIT_18s0001g09400), and a flavonoid 3′-5′-hydroxylase (VIT_08s0007g05160). We also found genes related to cell wall metabolism, including cellulose synthases and a xyloglucan endotransglucosylase/hydrolase, and many genes related to biotic stress responses, such as R protein, Avr9/Cf-9 induced kinase and a TIR-NBS-TIR type disease resistance protein. Interestingly, four calmodulin proteins were identified among the switch genes, including CAM5 (VIT_11s0016g05740), which is downregulated in response to heat stress in Arabidopsis [34]. Finally, the two-pore potassium channel KCO1 and the AKT1 channel, described, were also found among the switch genes which are downregulated in WS leaves (Fig. 5c and Additional file 6).
Characterization of the physiology and transcriptome of leaves following stress recovery
After 46 days of WS, the vines were re-watered and allowed to return to ~90 % of maximum water availability (Fig. 1c) which was achieved after another 24 days. Leaves were sampled from the WW plants, which had received an uninterrupted water supply throughout the experiment, and the revived WS plants (RWS).
Physiologically, the two cultivars behaved differently after rehydration (Fig. 6a–d). Although both cultivars promptly recovered to a non-limiting Ψl (approximately −0.6 MPa) (Fig. 6a), the RWS Sangiovese leaves reached a higher Amax value than corresponding WW vines, whereas RWS Montepulciano leaves only achieved a partial Amax recovery, setting at 79 % of the corresponding WW vines (Fig. 6b). The gs rates of RWS leaves from both cultivars returned to values similar to WW vines (Fig. 6c). Therefore, differences in Amax following rehydration were also reflected in the WUEi values, which were significantly higher than WW controls in RWS Sangiovese leaves but significantly lower than WW controls in RWS Montepulciano leaves (Fig. 6d).
Re-watering also had a differential impact on the transcriptomes of the two cultivars. Far more genes were differentially expressed between WW and RWS leaves in the Montepulciano vines (2381 genes) compared to Sangiovese vines, where only 197 were identified (Fig. 6e and f; Additional file 8). There was also a greater FC between WW and RWS gene expression levels in Montepulciano leaves compared to Sangiovese leaves (Fig. 6g and h). The majority of the Montepulciano transcripts modulated by RWS reversed the expression profiles observed during WS, i.e. transcripts downregulated by WS were upregulated by re-watering and vice versa (Fig. 6e and h), whereas this reversal was not observed in the Sangiovese leaves (Fig. 6f). Interestingly, many of the genes most strongly induced by recovery in the Montepulciano leaves were involved in protein regulatory activities, suggesting that adjustments following the release of stress are achieved predominantly through the management of existing protein pools. Nine HEAT-repeat-containing proteins mainly involved in cargo transport and in protein translation [35] were strongly upregulated in the Montepulciano RWS leaves, as well as three bromo-adjacent homology (BAH) domain-containing proteins and six CCR4-NOT transcription factors, both with roles in gene expression regulation [36]. Six E3 ubiquitin protein ligases with a well-known role in protein degradation [37] were also upregulated in Montepulciano RWS leaves. Many transcripts for HSPs and galactinol synthases that were induced by WS were among the most strongly suppressed genes in the RWS Montepulciano leaves. Taken together, these results indicate that the recovery process in Montepulciano leaves actively counteracts the negative effects of prolonged WS, whereas there is no equivalent process in Sangiovese leaves (Fig. 6h).
Whole genome transcriptional analysis in berries subjected to WS
We also compared the transcriptome dynamics of Montepulciano and Sangiovese berries sampled at the same time as the leaves. These sampling points span the herbaceous growth phase, corresponding to BBCH 69 (end of flowering with all flowerhoods fallen), BBCH 71 (fruit set: young fruit begin to swell, remains of flower lost) and BBCH 77 (berries begin to touch) as described by Lorenz et al. [38].
As per transcriptomic analysis, the same statistical workflow as applied to the leaves was applied to the berries. SAM (12 groups, FDR = 0.1 %) revealed that 23,464 genes were differentially modulated under our experimental conditions, and ANOVA (12 groups, α = 0.01, standard Bonferroni correction) retrieved the 11,839 most significantly modulated transcripts (Additional file 9). Genotype-specific PCA (Fig. 7a and b) revealed consistency among the biological triplicate samples. This statistical approach also suggested that most of the data set variability explained differences among the phenological stages and not between the WW and WS samples.
We identified 354 differentially expressed berry genes by comparing WW and WS plants (FC ≥2) at each time point in each variety, among which 269 were already annotated (Fig. 7c and Additional file 10). These transcripts, 48 (~18 %) of which were also found among the stress-modulated transcripts in the leaves, were particularly enriched in the functional categories “Response to stress”, “Response to abiotic stimulus” and “Catabolic process”, as highlighted by the BiNGO overrepresentation analysis (Fig. 7d). Overall, the response of berries to WS involved fewer genes than the leaves, and these genes were generally subject to weaker modulation (Additional file 10).
We next used the STEM clustering approach to determine the extent to which the WS response differs between berries of the two varieties (Fig. 7e and Additional file 11). Five galactinol synthases, a galactinol-raffinose galactosyltransferase and a raffinose synthase were upregulated more strongly and rapidly in Sangiovese compared to Montepulciano berries (SG25 → MP8 and SG25 → MP23 in STEM analysis, Fig. 7c and e). With the exception of the raffinose synthase (VIT_14s0066g00810), these genes were also expressed in WS leaves (Additional file 10). Three β-expansin-like genes (VvEXLB2–4) were upregulated by WS but the profile differed between the cultivars (SG25 → MP8 and SG25 → MP23 in STEM analysis, Fig. 7c and e). Ten of the 12 of heat shock and heat shock-related proteins were also commonly expressed in berries and leaves but the expression profiles were distinct (Fig. 7c and Additional file 10). In the berries, these genes were downregulated after 6 days in Sangiovese vines and almost no modulation was observed at the other time points or in the Montepulciano berries (SG14 → MP29 in STEM analysis, Fig. 7c and e). The difference between the varieties in terms of heat management in the leaves therefore appeared to be lost in the berries.
Three important ABA-related transcripts were significantly modulated in berries subjected to WS. The 11,12,9-cis epoxycarotenoid dioxygenase VvNCED3 (VIT_19s0093g00550), which catalyzes the last step in ABA biosynthesis [39], was upregulated in both varieties albeit with minor differences in the expression profile (SG 25 → MP 23 in STEM analysis, Fig. 7c and e; Additional file 11). The β-carotene hydroxylase VvBCH1, which was expressed in WS leaves, was upregulated at the final sampling point in Montepulciano berries but was not modulated in Sangiovese berries. Furthermore, the ABA-degrading enzyme ABA 8′-hydroxylase CYP707A2 (VIT_07s0031g00690) was downregulated at the final sampling point in Montepulciano berries but not in Sangiovese berries. These findings suggest that ABA is synthesized in berries after prolonged WS in both varieties, but only in Montepulciano berries is the ABA level maintained by downregulating the enzyme responsible for ABA degradation. Interestingly, genes related to auxin metabolism and signal transduction were more strongly repressed in Sangiovese than Montepulciano berries, e.g. indole-3-acetic acid amidosynthetase (VIT_01s0150g00300), (SG4 → MP18 in STEM analysis, Fig. 8c and e).
Finally, the metabolism of volatiles was remarkably impaired in the berries of both varieties under WS, although the volatiles that were affected differed in each cultivar. Six linalool synthases were downregulated in Sangiovese berries whereas three pinene synthases were repressed in Montepulciano berries. Furthermore, two germacrene-D-synthases (VIT_19s0014g04880 and VIT_19s0014g04900) were downregulated solely in Sangiovese berries but another (VIT_19s0015g02070) was strongly downregulated only in Montepulciano berries (Fig. 7c).