Abscisic Acid Metabolism in Leaves and Roots of Four Vitis Species in Response to Water Decit

Background: Abscisic acid is a phytohormone involved in water decit response. Abscisic acid metabolism is regulated by biosynthesis, conjugation, and catabolism. NCED3 is the rate limiting step of abscisic acid biosynthesis and is a key contributor to plant water decit responses. In this study NCED3 transcript accumulation and abscisic acid metabolism were further characterized as key water decit responses in four Vitis species (Vitis vinifera (Cabernet Sauvignon), Vitis champinii (Ramsey), Vitis riparia (Riparia Gloire), and Vitis vinifera x Vitis girdiana (SC2)) under three levels of water decit in leaves and roots. Results: The concentrations of abscisic acid and derivative metabolites increased with water decit and was dependent upon the species. RNA-Seq and RT-qPCR data were consistent with the changes in abscisic acid metabolite concentrations; the corresponding transcript abundances substantiate NCED3 as a key gene in the water decit response; however, NCED3 protein concentrations assayed in Western Blots were not affected. Major differences in abscisic acid metabolism at the gene, protein, and metabolite levels were detected between leaves and roots in these four species. NCED3 transcript abundance and abscisic acid concentration in drought-tolerant Ramsey increased earlier and more signicantly than the other species during long-term, moderate to severe water decits but were not stimulated as much by short-term, rapid dehydration. In drought-sensitive Riparia, NCED3 transcript abundance and abscisic acid metabolite concentrations increased to a lower extent than in Ramsey during moderate to severe water decits, but short-term rapid dehydration induced a signicantly higher abscisic acid concentration in Riparia than Ramsey. Conclusions: Grapevine have distinct abscisic that highly the severity stress (leaves This study acid metabolism part core water Relative A WD time course experiment was conducted to examine NCED3 transcript abundance, relative NCED3 protein abundance, and [ABA] in response to rapid dehydration WD. As all vines from previous experiments were submitted to the same treatments for the same long-term duration, there may have been a limited range of responses revealed in the previous experiments. To address the limited range of WD and long-term duration in the previous experiments, a third experiment was performed to expose leaves of three of the species (CS, RA, and RI) to a short-term time course of rapid dehydration WD. CS, RA, and RI leaves underwent rapid dehydration or continual petiole irrigation under controlled conditions for 2, 4, 8, and 24 hours.

concentrations are shown to vary between organ, duration of water de cit, and between drought sensitive and tolerant wheat cultivars [87], as was found in this study with grapevines.
In this study, increased ABA metabolism was con rmed as a key response to WD through quantitative analysis of ABA, ABA metabolites, ABA metabolism transcripts (RNA-Seq and RT-qPCR), and NCED3 protein (Western Blots) from three experiments in which the species were exposed to varying intensity and duration of WD. It was hypothesized NCED3 transcripts, NCED3 protein, and ABA concentrations were correlated with each other. The relationship between NCED3, NCED3, and ABA is not well described for grapevine nor other plant species. Commonly, transcript abundance is expected to explain protein abundance. However, this scenario is found to be the exception, not the rule [91][92][93][94][95]. In this study, NCED3 protein did not increase like NCED3 transcript abundance and ABA concentration in leaves and roots in these grapevine species under these WD conditions. The goals of this work were to further establish ABA biosynthesis as a core response to WD in Vitis, examine ABA metabolism in leaves and roots, and compare the ABA metabolisms of the four Vitis species that differ in drought tolerance.

Results
For the moderate and severe long-term WD experiments previously described [8], CS, RA, RI, and SC vines were subjected to one or two week of moderate or severe water de cit, or well-watered daily. Vines were grown in pots under greenhouse conditions. At W1 and W2 of treatment various measurements were taken [8]. Leaves and roots were harvested after W1 and W2 of treatment. The one-week severe water de cit treatment reached a more acute level of stress (lower stem water potentials) than the stress level obtained after two-weeks of moderate water de cit (see Methods).
Grapevine organs and species differ in ABA and ABA metabolite concentrations during one and two-week moderate WD Leaf and root ABA and ABA metabolite concentrations are different between the four species after a one and two-week moderate WD treatment. ABA, ABA-GE, PA, DPA, NeoPA, and 7'OH ABA were extracted and quanti ed (Figs. 1 and 2) as previously described [96]. There were signi cant differences between treatments, organs, species, and weeks for all metabolite concentrations (pmol·mg DW − 1 ; DW refers to dry weight) except for 7'OH ABA. The term "signi cant" will be used in this work to mean statistically signi cant at a p-value of 0.05 or less. To simplify comparisons to WD, metabolite quantities of the Controls of all species were grouped together per organ per week (Figs. 1 and 2) and were referred to as the average of the Controls (Avg Ctrl).
Grapevine organs had different [ABA] under Control and WD treatments. There was a signi cant difference in [ABA] between W1 WD leaves and roots, between W2 Control leaves and roots, and between W2 WD leaves and roots. Per organ per week only RA WD W1 leaves [ABA] was signi cantly increased relative to that of Avg Ctrl in (Fig. 1). WD RA W1 had at least 3-fold higher average [ABA] in the leaves and roots than WD RI W1. At W2, the [ABA] for all species was signi cantly higher in WD leaves relative to Avg Ctrl leaves, and WD RA had a signi cantly higher [ABA] than WD SC (Fig. 1). W2 roots also had signi cantly higher [ABA] for all WD species than those of Avg Ctrl (Fig. 1), and again WD RA had an average [ABA] 2-fold and 3-fold higher than WD RI in the leaves and roots, respectively. However, unlike the leaves there was no signi cant difference between the species in the roots treated with WD at W2 (Fig. 1). As a general rule, [ABA] can be ranked by species during moderate WD as RA > CS > RI > SC in both leaves and roots.
The concentrations of the storage form of ABA, ABA-GE, were signi cantly higher than [ABA] in leaves and lower than [ABA] in the roots (Fig. 1) and in all combinations per treatment per week. At W2, RA and SC WD leaves had signi cantly higher [ABA-GE] than Avg Ctrl, and WD RA leaves had signi cantly higher [ABA-GE] than WD RI leaves (Fig. 1). In the roots, RA was the only species to have signi cantly higher [ABA-GE] than Avg Ctrl, but WD RA W2 roots were only signi cantly different from those of CS (Fig. 1).
In general, ABA catabolism was enhanced by WD. At W1, WD [ABA catabolite] were similar to respective Controls (Fig. 2). W1 WD [DPA] in WD RA roots were signi cantly higher than WD RI roots. By W2, [ABA catabolite] increased by 2-to 10-fold in both leaves and roots relative to W1. Like ABA, ABA catabolites were generally higher in WD W2 leaves than WD W2 roots but were at comparable levels in leaves and roots at W1 (Fig. 2). For example, WD W2 RA had > 10-fold increase in [PA] in leaves relative to roots. In WD W2 leaves, RA was the only species that had signi cantly different concentrations of all four ABA catabolites relative to Avg Ctrl (Fig. 2). RA and CS WD W2 roots generally had higher concentrations of all four quanti ed catabolites than those of the other two species (Fig. 2), but there was no signi cant difference between the WD species for any catabolite in the roots at W2 (Fig. 2). Although not always signi cant, each species displayed unique responses to moderate WD, particularly in W2, via changes in [ABA catabolite] with RA (and CS) generally having higher [ABA catabolite] than the other species in both organs, consistent with the higher [ABA] observed in these species (Fig. 1).
There were clear differences in the use of the catabolite pathways in grapevines. The major catabolic pathway was through 8'OH ABA with [DPA] and [PA] being much higher than [7'OH ABA] at W1 and W2 (Fig. 2). NeoPA, which also goes through the 8'OH ABA catabolism pathway, had very low concentrations relative to the other catabolites (Fig. 2). Nevertheless, these concentrations also increased with [ABA] induced by moderate WD with RA > CS > RI; SC had very low concentrations and showed no increase due to WD (Fig. 2). PA, an intermediate in the 8'OH ABA catabolic pathway had much higher concentrations than DPA in WD leaves. However, the reverse was true in WD roots, where [DPA] was higher than [PA].
To further understand ABA metabolism in the plant as a whole, the distribution of ABA and ABA metabolites was examined for the whole plant (the sum of leaf and root total ABA and ABA metabolites, to be de ned as Total ABA Metabolites). ABA and all ABA metabolites were summed to estimate the total ABA produced by the plant (Additional File 1A) or per organ (Additional File 1B). On a whole plant basis, there were signi cant increases in [Total ABA Metabolite] by W2 (Additional File 1A) providing evidence for an increase in ABA biosynthesis along with other metabolism pathways. Total ABA Metabolite distribution in W1 was less clear and may indicate that changes in ABA metabolite concentrations were the result of redistribution of ABA (e.g. conjugation, catabolism and transport).
[ABA] and each [ABA metabolite] per organ was also divided by the summed Total ABA Metabolite concentration (Additional File 1A) to determine the distribution of each ABA metabolite (Additional File 2). This approach revealed ABA-GE represented the major portion of the leaf ABA metabolites in W1 Controls (0.38 ± 0.05), followed by PA (0.23 ± 0.03), ABA (0.17 ± 0.03), and DPA (0.13 ± 0.02) (Additional File 2). However, RA WD W1 leaves had a signi cant decrease in the proportion of ABA-GE from 0.48 ± 0.15 to an average 0.20 ± 0.08, indicating ABA metabolism was shifted from ABA-GE to other ABA metabolites such as ABA and DPA. The reduction of ABA-GE pools was speci c to W1 RA leaves; there was no signi cant difference in RA Control vs. WD W2 leaves indicating this shift in ABA metabolism was an early response to WD (Additional File 2). The other species did not show the same shift in ABA-GE distribution in WD leaves. CS ABA-GE (0.2 ± 0.02) remained constant in W1 and W2 and did not change signi cantly in response to WD in the leaves (Additional File 2). Likewise, the proportion of ABA-GE in WD W2 roots decreased relative to Control in all of the species except for CS, which again was proportionately low relative to the other three species (Additional File 2). No such changes were observed in WD W1 roots. ABA and ABA-GE in the roots in general were a much smaller proportion of Total ABA Metabolites than DPA. were investigated (Fig. 3). ABA and 7'OH ABA had the highest z-scores of all the metabolites quanti ed (Fig. 3). Such changes were not obvious for 7'OH ABA in Total ABA Metabolites (Additional File 2). RA W1 leaves had the highest score for ABA (Fig. 3). RI W1 leaves had the highest score for WD:C ABA-GE (Fig. 3), indicating ABA-GE may have been accumulating in RI W1 WD leaves. The WD:C ABA-GE score for RA W1 leaves was about two STD lower than the average WD:C ABA metabolite average (Fig. 3), indicating ABA-GE may be deconjugated into active ABA and/or downstream catabolites, supporting similar observations from Additional File 2. ABA-GE also had a negative score for the roots of all species in W1 and W2 (Fig. 3), possibly re ecting ABA activation through this pathway.
The ABA catabolites have different scores in the different species and organs. 7'OH ABA had the highest scores for RA W1 roots and leaves and RA and CS W2 roots and leaves (Fig. 3) despite 7'OH ABA having lower concentrations than metabolites representing catabolism through 8'OH ABA (PA and DPA). The high z-scores for 7'OH ABA in these select species and organs may indicate these species redistribute a higher portion of ABA catabolism through the 7'OH ABA pathway than the other species in response to WD, but this pathway either catabolizes lower [ABA] than that through 8'OH ABA, PA, and DPA ( Fig. 2) or catabolites are further degraded into compounds not quanti ed here. Alternatively, the high z-scores for 7'OH ABA may be attributed to the large difference between W2 root average RA and CS Control [7'OH ABA] (0.14 ± 0.03 pmol·mg DW − 1 ) and average WD RA and CS (5.31 ± 1.67 pmol·mg DW − 1 ) that result in a greater fold difference (38-fold difference) than those of PA (6-fold difference) and DPA (3-fold difference); this difference is re ected in the z-score despite differences in [catabolite]. RA W1 roots and RA W2 leaves had the highest score for NeoPA and DPA, which were also about one STD above the average ABA metabolite ratio (Fig. 3). RA was the only species to have a positive score for PA and DPA in W2 leaves (Fig. 3). The differences between the catabolites indicate ABA may be preferentially deactivated by different mechanisms in each species in each organ and with increasing WD stress. For example, at W2, RA was the only species to have positive scores in leaves and roots for PA (~ 0.5 STD from mean), while all other species had negative scores. Combined with the positive scores for RA W2 leaves and roots 7'OH ABA and RA W2 leaves NeoPA scores, RA is clearly catabolizing ABA relative to the other species. WD RA W2 [7'OH ABA] has the greatest change relative to Control [7'OH ABA] and the average [ABA metabolite]. NeoPA, PA, and DPA had the highest z-scores in W1 across various organs and species, but the scores were either lower or negative for W2 indicating these catabolism pathways may be favored at different times during WD. At W1, RA and SC had positive scores for DPA in the roots, but not the leaves, hinting these catabolism pathways may be utilized differently across organs. The different species, organs, and times have different scores for the ABA metabolites.
One and two-week moderate WD signi cantly increases ABA biosynthesis, conjugation, and catabolism transcripts abundance ABA metabolism gene transcript abundance signi cantly increase in response to WD. Differences in many ABA metabolite gene transcript abundances mirrored [ABA metabolite] and may partially explain differences observed between species. Previously, the ABA biosynthesis gene, NCED3, was identi ed as a potential hub gene in response to WD in this W1 and W2 moderate WD experiment [8]. Additional works indicate NCED3 as a potential WD hub gene [97]. To support this hypothesis, the RNA-Seq series, PRJNA516950, was analyzed with the more accurate CS clone 8 v1.0 reference genome [98] as opposed to the PN40024 reference genome [99] as was done previously [8]. The goal of this analysis was to investigate the transcript abundance of ABA metabolism genes in response to WD. CS ABA metabolism genes were identi ed with protein basic local alignment search tool (BLAST) using known ABA metabolism PN40024 protein sequences as a query. Orthologs were con rmed based on the highest total score, E-value, and length. When two BLAST hits were highly similar for the same query, the hits were identi ed as alleles of the same gene. In total, 46 ABA metabolism genes were identi ed from the primary (alternative (alt)1) and/or secondary (alt2) CS haplotig sequences (Additional File 3). Alleles are referred to as alternatives without a designated haplotype because the phased chromosome-scale assembly was not complete for the CS clone 8 v1.0 genome. Detailed explanation of alt assignment is in Additional File 3. In total, allele(s) for three NCED genes (NCED3, NCED5 alt1 and 2, and NCED6 alt1 and 2) were identi ed in CS (Additional File 3). Both alleles were identi ed for BETA-OHASE 2, ABA ALDEHYDE OXIDASE 3 (AAO3), ABA1, ABA2, ABA3, BETA-1, 3-GLUCANASE 1 (BG1), ATP-BINDING CASSETTE G25 (ABCG25), and ATP-BINDING CASSETTE G22-LIKE (ABCG22-like). Several BGLU, BG, UGT, and CYP707A paralogs were also identi ed. Only one ABCG40 and NPQ1 (Violaxanthin de-epoxidase or NON-PHOTOCHEMICAL QUENCHING 1) allele were annotated in this version of the CS genome.
Organs showed signi cant differences in the genes expressed and gene expression levels (Additional File 4) [8]. There were signi cant differences in ABA metabolism genes comparing WD to Control in all four species per organ (Additional File 4). CS (4 and 0 genes in leaves and roots, respectively), RA (1 and 1 genes), and RI (1 and 0 genes) had the most differentially expressed ABA metabolism genes out of all WD vs. Control W1 contrasts (Additional File 4). RA W2 leaves and roots had the most differentially expressed ABA metabolism genes (21 and 17 genes in leaves and roots, respectively) in response to WD followed by CS (10 and 12 genes), SC (10 and 7 genes), and RI (4 and 8 genes) (Additional File 4). The transcripts of NCED3 (Fig. 4) and at least one putative BETA-OHASE 2 allele were signi cantly increased in response to WD in W2 leaves and roots of all four species (Additional File 4). However, the other differentially expressed ABA metabolism genes varied between species and organs.
RA has distinct ABA metabolism differentially expressed genes (DEGs) compared to the other species. In the previous analysis of this experiment, RA stood out transcriptionally and physiologically by outperforming the other species, having higher photosynthesis and greater and earlier transcriptomic responsiveness to WD [8]. To better understand the role each pathway of ABA metabolism (biosynthesis, conjugation, or catabolism) contributes to the transcriptomic responsiveness of RA and the other species, differential expression analyses (DEAs) contrasting WD to Control per time point per organ were focused upon using the 46 CS ABA metabolism genes. To identify speci c genes differentially expressed in WD RA leaves and roots relative to WD organs from the other three species, DEA was performed as previously described [8] using RA WD contrasts (Additional File 5). The majority of WD responsive DEGs related to ABA metabolism in RA and the other species in either leaves or roots were involved in biosynthesis and conjugation (Additional File 5).
In this section, ABA metabolism DEGs that were unique to WD RA (relative to the WD of all three other species) will be discussed in order of ABA metabolism pathway (biosynthesis, (de)conjugation, and catabolism) sequentially in W1 leaves then roots and nally in W2 leaves followed by W2 roots. No gene was signi cantly differentially expressed in RA relative to all three of the other species in WD W1 leaves. NCED3, NCED5 alt2, and ABCG25 alt1 had signi cantly higher transcript abundance in RA WD W1 roots than those of the other species (Figs. 4 and 5 and Additional File 5). NCED3 was the only ABA metabolism DEG in RA W1 WD roots relative to Control (Additional File 4), but NCED3 was not a DEG in W1 roots of any other species at W1 in response to WD ( Fig. 4 and Additional File 4). BG2 alt1 and UGT71C3 alt1 were also differentially expressed in RA W1 WD roots relative to the other three species, but these genes were not DEGs W1 root WD vs. Control contrast for any species.
There were nine ABA metabolism DEGs in RA leaves after W2 of WD relative to those of the other three species that were also DEGs in RA WD W2 leaves relative to RA Control W2 leaves (Figs. 4 and 6 and Additional Files 5 and 6). Four of these genes were involved in ABA biosynthesis: NCED3 alt1, NPQ1 alt1, ABA1 alt1, AAO3 alt1. NCED3 had signi cantly higher transcript abundance in RA WD W2 leaves than those of all other species (3.7-, 9.8-, and 6-fold difference for CS, RI, and SC respectively) ( Fig. 4 and Additional File 5). NCED3 had the highest expression level of the ve annotated NCEDs in both leaves and roots in W1 and W2 of WD in all species with the highest level of expression in RA (Fig. 4). The other NCEDs were lowly expressed (Fig. 4). NCED5 alternatives had higher expression in CS, while both NCED6 alternatives had higher expression levels in RA after W2 of WD compared to the other species (Fig. 4). NCED3 was also the ABA metabolism DEG that had the highest average TPM over all WD species and times compared to the other ABA metabolism DEGs that were signi cantly differentially expressed in RA WD W2 leaves vs. those of the other species and in RA Control vs. RA WD W2 leaves. Overall, NCED3 appears to be the major NCED contributing transcripts to downstream ABA biosynthesis. NPQ1 alt1 was signi cantly ~ 4-fold lower in RA WD W2 leaves than those of the other species (Additional Files 5 and 6). ABA1 alt1 was also signi cantly lower in RA WD W2 leaves than the WD W2 leaves of the three other species (Additional Files 5 and 6). Overall, average ABA1 alt1 transcript abundance over all treatments, times, and species was signi cantly higher in the leaves than the roots. AAO3 alt1 had signi cantly lower transcript abundance in RA than CS (16-fold difference), RI, and SC (both about 56-fold difference) (Additional Files 5 and 6) in W2 WD leaves. Although lowly expressed, average AAO3 alt1 transcript abundance was signi cantly higher in roots than that in leaves across all species, times, and treatments.
There were four ABA metabolism DEGs in RA leaves after W2 of WD that were also DEGs in W2 RA leaves in response to WD involved in deconjugation. The deconjugation DEGs were BG1 alt1, BG1 alt2, BG3 alt1, and BG3 alt2. BG1 alt1 transcript abundance was signi cantly lower in RA WD W2 leaves than CS (6.5fold difference), RI, (26-fold difference), and SC (64-fold difference) ( Fig. 5 and Additional Files 5). BG1 alt1 and alt2 were DEGs in RA WD leaves and roots relative to respective Controls, but neither BG1 allele was a DEG in the organs of any other species (Additional File 2). BG1 alt1 had higher transcript abundance in roots than in leaves across all species, treatments, and times. BG1 alt2 also had signi cantly higher average transcript abundance in all Control and WD species roots relative to all Control and WD species leaves. In WD W2 roots, BG1 transcripts alt1 and alt2 were the only signi cantly differentially expressed ABA metabolism DEGs in RA WD vs. the other three species WD contrasts with ~ 16-40-fold lower abundance in RA than the other species ( Fig. 5 and Additional File 5). Control RA and SC had the highest BG3 alt1 expression in the leaves and roots in W2, but RA WD W2 had the lowest BG3 alt1 expression in leaves and roots. Interestingly, per treatment, BG3 alt1 transcript abundance was comparable in the leaves and roots for each species. BG3 alt2 had the lowest transcript abundance in RA for all treatments and organs in W2 relative to the other species. CYP707A1 alt1 was the only ABA catabolism gene signi cantly differently expressed between RA and all three other species in W2 WD leaf contrasts and Control vs. WD contrasts (Additional Files 4-6). CYP707A1 alt1 had signi cantly higher transcript abundance in RA WD W2 leaves than those of the other three species (Additional Files 4 and 6), and like AAO3 alt1 and BG1, CYP707A1 alt1 average transcript abundance was higher in roots than leaves as an average over all species, treatments, and times.
One and two-week moderate WD signi cantly increases ABA transport gene transcript abundance ABA transport genes were signi cantly differentially expressed in response to WD. ABA transporter genes ABCG25 alt1 and alt2 and ABCG40 had signi cantly higher average transcript abundance in roots than leaves in all species, weeks, and treatments. Neither ABCG25 alt2 nor ABCG40 were higher in CS, RA, and RI WD W2 roots or RA WD W2 leaves relative to respective Control ( Fig. 6 and Additional File 4). On average ABCG25 alt1 transcript abundance was ~ 4-fold higher in WD than Control W2 leaf and root samples (Additional File 4). ABCG25 alt2 transcripts were signi cantly higher in RA WD W2 leaves (~ 32fold increase) and RA WD W2 roots (~ 8-fold increase) relative to respective Control ( Fig. 6 and Additional File 5).
RA had signi cantly increased ABCG25 and ABCG40 transcript abundance relative to other species after W2 of WD in leaves and roots (Additional File 5). ABCG25 alt1 was signi cantly (~ 3-fold) higher in RA than CS WD W2 leaves, and about ~ 4-fold change higher in RA W1 and W2 roots than those of the other three species. ABCG25 alt2 transcript abundance was also signi cantly higher in RA WD W2 leaves than those of RI and SC (Additional File 5). After W1 of WD, RA roots had ~ 8-fold change increase in ABCG40 transcripts relative to those of CS, but ABCG40 was not signi cantly different in Control vs. WD contrasts for either RA or CS in W1 (Additional File 4).
ABA metabolite gene transcript abundance may partially explain ABA metabolite concentrations Multiple ABA metabolism genes had signi cantly increased transcript abundance similar to the increases observed in ABA metabolites. To more easily compare species response to WD and link [ABA metabolite] to upstream transcripts, the average ratio of WD:C transformed TPM were expressed as a z-score per ABA metabolism gene group (e.g. NCED3, NCED5 alt1 and 2, and NCED6 alt1 and 2 are included in the NCED gene group) with darker colors indicating greater difference from the mean ratio of the ABA metabolism genes (Fig. 7). RA clearly stands out in this comparison, having the highest score for ß-carotene hydroxylases (roots W1 and W2), zeaxanthin epoxidases (roots W1 and W2), NCEDs (leaves and roots W1 and W2), AAO3s (leaves W1), UDP-glucose glucosyl transferases (leaves and roots W2), ß-d-glucosidases (roots W1), and various ABA hydroxylases (leaves W1 and roots W1 and W2) ( Fig. 7) that may partially explain the high [ABA] and [ABA metabolite] observed in RA. RA also had the lowest scores for violaxanthin de-epoxidases (leaves W1), xanthoxin dehydrogenases (roots W2), ß-d-glucosidases (leaves W2), and ABA aldehyde oxidases (leaves and roots week 2) (Fig. 7). The low score (~-2 STDs from the mean) and low expression (Fig. 7) of ß-d-glucosidases as well as the high z-score for UDP-glucose glucosyl transferases (~ 1.75 STD from mean) in RA W2 leaves may explain the high [ABA-GE] observed in RA WD W2 leaves; ABA may be conjugated into ABA-GE but not deconjugated allowing [ABA-GE] to increase. The higher [ABA-GE] in RI W1 WD leaves relative to the other species (Fig. 7) and the higher score for ß-d-glucosidase (~ 2 STD from mean) may indicate ABA-GE is an important source of ABA for RI in earlier WD response. RI and SC had the lowest [catabolite] in the leaves, which is paralleled in the score of the ABA hydroxylases.
ABA metabolism genes were correlated with multiple WGCNA modules WGCNA was performed on leaves and root samples separately with ABA and the ABA metabolites as additional traits (Figs. 8 and 9 and Additional File 7). Gene association to each module was calculated between each gene and the eigengene of each module [100] (Additional File 7). ABA metabolism genes were spread across multiple gene modules. In the leaves, 30 WGCNA modules were identi ed (Fig. 8). WD was positively correlated with ve modules. These modules included lightyellow, darkgreen, brown, saddlebrown, and green (Fig. 8). Generally, the ABA metabolites were positively correlated with the same ve modules as WD (Fig. 8). The 46 ABA metabolism genes were spread across 20 different modules in leaves. However, within these modules only lightyellow overlapped with WD, ABA, and the ABA metabolites. The lightyellow module contained the greatest number of ABA metabolism genes (seven) including NCED3 (0. 91 In the roots, 34 WGCNA modules were identi ed (Fig. 9). In total, ten modules were positively associated with WD (Fig. 9). Among the WD positively correlated modules, six were also positively correlated with 7'OH ABA, ABA, ABA-GE DPA, NeoPA, and PA ( Fig. 9 and Additional File 7). Common positively correlated modules between the metabolites included royalblue, lightgreen, midnightblue, pink, and paleturquoise.
The ABA metabolism genes were spread across 20 modules in the roots. The midnightblue module contained the greatest number of ABA metabolism genes (ten) including NCED3 (0.95), CYP707A2 alt1 Genes in the ABA metabolite and WD modules have roles in response to stress and stimulus. Gene ontology enrichment was performed using biological process terms for the brown, green, and lightyellow modules in the leaves and the midnightblue and pink modules in the roots (Additional File 8). These modules were selected for high correlation to WD and the ABA metabolites as well as for the number of ABA metabolism genes contained in each module. The gene ontology (GO) term corresponding to response to endogenous stimulus was enriched in all these modules (Additional File 8). All of these modules were enriched for the response to stress and response to stimulus GO terms (Additional File 8).
All modules except green leaf included the response to abiotic stimuli GO term (Additional File 8). The lightyellow leaf module was enriched for the biosynthetic process GO term (Additional File 8), and lightyellow had the highest correlation with BETA-OHASE 2, NCED3, and NCED6 in the leaves (Additional One-week severe WD signi cantly increases ABA biosynthesis transcripts and metabolites, but does not change NCED3 protein A second experiment was performed with a more severe WD over W1 described previously [8]. Brie y, the same four Vitis species underwent a natural dry-down over the course of a week that achieved a stem water potential lower than that of vines that experienced the two-week moderate WD treatment. RT-qPCR was reported previously, and NCED3 NRQ were signi cantly increased in response to WD in all species except for RI; RA WD and CS WD had the highest NCED3 NRQ in the leaves and roots [8]. Fold difference (FD) of NCED3 protein abundance was quanti ed from western blots that were made relative to a CS Control leaf sample run on every gel as an inter-run caliber (IRC) as previously described [101] (Additional File 9). There was no signi cant difference of relative NCED3 between the treatments or species (Additional File 10). However, CS Control leaves and RA Control and WD leaves had signi cantly higher relative NCED3 protein than respective roots (Additional File 10).
ABA was quanti ed, and the [ABA] after W1 of severe WD were comparable to those of W2 moderate WD (Figs. 1 and 10). All WD treated leaves, except those of RI were signi cantly different from respective Control (Fig. 10), which was paralleled in NCED3 NRQ [8]. RA WD leaves had a signi cantly higher [ABA] than those of RI and SC, which was also observed in NCED3 NRQ. In the roots, no WD treated species had a signi cant difference from respective [ABA] Control or to other WD treated species (Fig. 10). The similarity in [ABA] between the W2 moderate and one-week severe WD experiments indicated that ABA metabolism was dependent on severity and duration of WD stress.
Exposing leaves to short-term WD with increasing severity signi cantly increases NCED3 transcripts and ABA concentration, but it does not impact NCED3 protein abundance A WD time course experiment was conducted to examine NCED3 transcript abundance, relative NCED3 protein abundance, and [ABA] in response to rapid dehydration WD. As all vines from previous experiments were submitted to the same treatments for the same long-term duration, there may have Over the course of the experiment WD leaves had signi cantly lower stem water potential and lost signi cantly more water than Control leaves (Additional File 11). RI WD leaves had signi cantly lower water content at 2 hours of rapid dehydration relative to respective Control leaves, and CS and RA WD leaves had a signi cant change in water content by 8 hours of treatment, but there was not a signi cant difference between WD species at any timepoint. Both stomatal conductance (Gs) and photosynthesis (Ps) were signi cantly reduced over the course of the WD treatment (Additional File 11). Stem water potential, osmotic potential, and calculated turgor pressure [102] were also signi cantly reduced in the WD treated leaves (Additional File 11). There were signi cant differences between treatments and species for most physiological measurements, but there was not a signi cant difference for time in most cases (Additional File 11). For this reason, Control and WD physiological measurements were shown as an average of all species and time points (Additional File 11). Per time point per measurement there was no signi cant difference between WD species, but all WD species were signi cantly different from Control by two hours of treatment.
Each species increased NCED3 transcript abundance in response to the rapid dehydration (Fig. 11). There were no signi cant differences in NCED3 transcript abundance between the Control leaves of the species for any time point. NCED3 transcript abundance was signi cantly higher in CS rapid dehydration leaves at all time points relative to CS Control leaves with a general trend of NCED3 transcripts increasing with rapid dehydration time ( Fig. 11 and Additional File 12). NCED3 transcript abundance was signi cantly higher in RA rapid dehydration leaves at all time points except 24 hours relative to RA Control leaves, but NCED3 NRQ was much lower in RA WD leaves than CS WD leaves ( Fig. 11 and Additional File 12). Average RA NCED3 transcript abundance was highest after two hours of rapid dehydration and decreased over time, but RA rapid dehydration NCED3 transcripts at two hours of treatment were not signi cantly different than any other time point of rapid dehydration ( Fig. 11 and Additional File 12). Surprisingly, NCED3 transcripts stayed constant in RI rapid dehydration leaves, and RI WD was only signi cantly different from RI Control after two hours of treatment ( Fig. 11 and Additional File 12). Interestingly, there were no signi cant differences in NCED3 transcript abundance between the rapid dehydration leaves of the species per time point (Fig. 11 and Additional File 12).
[NCED3 protein] was not signi cantly different for any species in Control or rapid dehydration at any time point (Additional File 10). Western blots were performed as in the one-week severe water de cit experiment using a CS Control two-hour sample as an IRC [101]. RI rapid dehydration two-and four-hour leaves had the greatest variability in NCED3 protein sample between replicates. The NCED3 relative abundance similarity between Control and WD leaves in the rapid dehydration was comparable to those observed in the one-week severe WD.
[ABA] increased in response to short term rapid dehydration WD (Fig. 11). There was no signi cant difference in Control [ABA] between the species at any timepoint ( Fig. 11 and Additional File 12).
[ABA] in CS WD 24 hours was signi cantly different from RA WD 24 hours, but this was the only signi cant difference between WD species at any time ( Fig. 11 and Additional File 12). CS WD [ABA] was signi cantly different from that of CS Control at two and 24 hours of treatment ( Fig. 11 and Additional File 12). CS WD experienced a general increase in ABA with time like CS WD NCED3 transcript abundance (Fig. 11). RA WD had the lowest [ABA] of the WD treated species at all times; RA WD was only signi cantly different from RA Control after 24 hours of treatment ( Fig. 11 and Additional File 12). RA [ABA] paralleled RA NCED3 transcript abundances (Fig. 11), which surprisingly did not increase as much as CS during this short-term WD treatment. RI WD had the highest [ABA] of the WD species at four and eight hours of treatment and steadily increased with time ( Fig. 11 and Additional File 12). However, RI WD [ABA] did not follow the same trend as RI WD NCED3 transcript abundance (Fig. 11), which remained relatively constant throughout the stress. This observation may indicate RI is relying on a different source of ABA (like ABA-GE deconjugation) more than the other species under short-term rapid dehydration. Although the WD species were experiencing the same level and duration of stress and having similar physiological responses (Additional File 11), each species displayed unique ABA metabolism responses via NCED3 transcript abundance and [ABA] ( Fig. 11 and Additional File 12). NCED3 transcript abundance and [ABA] during short term rapid dehydration did not display the same responses as the longer-term moderate and severe WD, indicating ABA metabolism is highly dependent not only on organ and species but also on stress severity and duration.

Discussion
Increases in NCED3 transcript abundance, ABA, and ABA metabolite concentrations were a core WD response in Vitis Overall, this work highlights differences in transcripts, a protein, and metabolites in the ABA metabolism . Throughout these experiments, NCED3 stood out as a key WD response gene in both leaves and roots of all species, which was supported by WGCNA (Figs. 8-9), GO enrichment analysis (Additional File 8), and RT-qPCR (Fig. 11). From the WGCNA, the lightyellow module in the leaves and the midnightblue module in the roots contained the most ABA metabolism genes in each respective organ (Additional File 3). Both of these modules identi ed NCED3 as a hub gene (Additional File 7), meaning NCED3 was among the top 20 most connected genes to other genes in these modules. Not only was NCED3 a hub gene for WD and ABA metabolism associated modules in both the leaves and roots, but NCED3 was the only ABA metabolism hub gene (Additional File 7) of the 46 ABA metabolism genes. No other ABA metabolism gene was close to being a top gene in the lightyellow, midnightblue, or any other module (Additional File 7), further highlighting the importance of NCED3. NCED3 was closely connected to other ABA signaling genes in these modules. The lightyellow leaf and midnightblue root modules did not contain the same ABA metabolism genes, indicating a difference in ABA metabolism and signaling gene transcription in leaves and roots. Only BETA-OHASE2 alt1 and atl2, CYP707A4 alt1_1, NCED6 alt1, and NCED3 were in the lightyellow modules in the leaves and the midnight blue module in the roots. The commonality of NCED3 and other NCED genes correlating with modules associated to WD and corresponding to the most ABA metabolism genes in both organs support the hypothesis that NCED3 acts as a core gene in response to WD in both leaves and roots. In addition to the most ABA metabolism and signaling genes, the lightyellow and midnightblue modules include numerous genes involved in plastid function like plastid transkelotases and the chloroplastic 50S ribosome subunit (Additional File 7). Other genes in these modules include aquaporins, ion transporters, ascorbate oxidase, galactinol synthase, and cysteine and sulfur regulatory genes, which have known associations to ABA [38,[103][104][105] (Additional File 7). GO enrichment analysis further supported NCED3 as a hub gene with biosynthetic and abiotic stress response terms linked to the lightyellow and midnightblue modules. Additionally, NCED3 transcript abundance increased in all WD experiments described here regardless of severity, duration, organ, and even species (although there were distinct responses to each WD treatment for each organ and species). The increase in NCED3 transcripts and [ABA] detailed here from three WD experiments was re ected in downstream signaling and physiological responses (Additional File 11) [8].
Only one NCED3 allele was detected in the CS v1.0 and later in the CS v1.1 genome using available protein BLAST. However, a second allele (VvCabSauv08_v1_Primary000127F) was found with manual curation using only the alternative contig in the CS v1.1 genome. The two NCED3 alleles share 99% identity determined by Clustal Omega alignment, and therefore the transcript abundance for this gene was representative of both alleles.
Leaves and roots differ in NCED3 transcript abundance, NCED3 protein levels, and ABA and ABA metabolite concentrations The primary site of ABA biosynthesis (leaves versus roots) as well as the initial site and signal to trigger a stress response during WD has been a controversial topic [106][107][108]. Recently, a small root sourced signaling peptide, CLE25, was identi ed in Arabidopsis thaliana that controls stomatal closure via ABA biosynthesis in the leaves [109]. However, there is no orthologous peptide in many species including grapevine, and the subtleties of WD detection in roots and shoots as well as the identities of longdistance signaling molecules (including ABA itself) remains elusive. These experiments characterize ABA metabolism in leaves and roots of grapevine. However, it is important to consider whole organs were used for these analyses. Speci c cell types like the guard cell [110] or root endodermis [111] have specialized responses to ABA and likely have unique regulation of NCED3 and ABA metabolism each of which require further investigation in these grapevines.
In these experiments, leaves and roots had different levels of ABA metabolism related transcripts (Figs. 4-7 and 11), NCED3 protein (Additional File 10), and ABA metabolites (Figs. 1-3 and 10-11). In both the moderate and severe WD experiments, leaves generally had higher abundance of NCED3 transcripts and [ABA] than roots. Leaves also had higher [NCED3 protein] than roots in the severe WD experiment. The differences in NCED3 transcripts, NCED3 protein, [ABA] and [ABA metabolite] between leaves and roots may re ect different sensitivity to ABA between the organs. It is likely the [ABA] threshold to illicit a speci c ABA response is organ speci c [86] and may involve ABA transport between organs. Previously, exogenous application of ABA was demonstrated to greatly affect transcriptomic signaling within berries, shoot tips, leaves, roots, and cell culture with no one gene demonstrating the same change in transcript abundance across all organs [86]. For example, ABCG22-like and ABI3 transcripts uniquely increased in the berries while MYB121 and DREB2H had speci c transcript abundance accumulation in the roots. Organ speci c sensitivity to other hormones has also been previously described. For example, exogenous [µM] of auxin stimulates leaf and shoot expansion but inhibits root elongation [112,113].
Gibberellins similarly demonstrated organ speci city with high concentrations biosynthesized in the stamen from which lower concentrations are transported out to support other oral organ development like the petals [114].
ABA, like gibberellins is biosynthesized [115][116][117] and transported throughout a plant [118]. ABA transport from vascular cells into guard cells is well documented with several identi ed transporters (ABCG40 [119], ABCG25 [120], ABCG22 [121], ABCG30, ABCG31 [122], and more recently NPF4.5 (AIT2) and NPF4.6 (AIT1) [123]). However, despite growing evidence that root ABA may be in part shoot sourced [83,124,125], a pathway for ABA transport into the root remains poorly described compared to that of ABA-vascular unloading and transport into guard cells [120] and endosperm-ABA unloading into the seed embryo [122]. In the moderate WD experiment, ABCG25 and ABCG40 had higher transcript abundance in roots than leaves for all species, treatments, and time points (Fig. 6), indicating possible ABA phloem unloading and ABA transport into the roots. Literature further supports the possibility of shoot-sourced ABA accumulating in the roots [120,122,126]. ABCG25 is predominantly expressed in phloem companion cells in Arabidopsis thaliana [120,127], indicating ABA is transported down through the plant.
ABCG25 transcripts accumulated in Arabidopsis thaliana roots in response to salt stress [126], and when nGFP was expressed under a 2.0 kb AtABCG25 promoter, a signal was observed in the root in the presence of 10 µM ABA after 20 hours of treatment [128]. AtABCG40, responsible for importing ABA into guard cells, also plays a role in lateral root development and is expressed in primary and lateral roots in addition to being crucial for stomatal function [119].
ABA-GE is abundant during WD [82] and demonstrated to be the most abundant ABA metabolite in the moderate WD experiment (Fig. 1). However, ABA-GE transport regulation remains unresolved. Leaves had signi cantly higher [ABA-GE] than roots in the moderate WD experiment (Fig. 1), but the organs may have different requirements for this metabolite. BG1 had signi cantly higher transcript abundance in roots than leaves for the majority of species for both treatments and times in the moderate WD experiment (Fig. 5), indicating the possible importance of ABA-GE deconjugation in the roots. Despite low membrane permeability [129] and attempts to identify an ABA-GE transporter [119], this potential ABA transport pathway remains unsolved. ABA-GE transport may be one source of shoot derived ABA in roots that has not been characterized [118]. ABA transport into the roots warrants further elucidation.

Grapevine species differ in physiological and biochemical responses to WD
In these experiments, species had distinct biochemical regulation of ABA. CS generally had the most variable physiological measurements [8] in both leaves and roots relative to the other species (Figs. 1-3 and 10). RA also had the highest NCED3 transcript abundance in moderate and severe WD treatments (Fig. 4) [8]. Additionally, RA had the highest levels of NCED3 transcripts (Fig. 4) and [ABA] (Fig. 1) at the earliest time point, indicating RA may be the most sensitive (earliest to respond) of the species investigated for long-term WD detection. RA maintained higher physiological function during the moderate and severe WD [8] despite higher levels of ABA than the other species (Figs. 1 and 10). These observations indicate RA may be less sensitive to [ABA] over time in terms of a physiological response (like stomatal closure) or RA may be enacting changes that better allow RA to function under WD (e.g. suberization or modi ed hydraulic conductance) relative to the other species. A number of aquaporin genes display differential and unique responses to water de cit in RA leaves (Additional File 4). Genes involved in cysteine biosynthesis and metabolism are also constitutively higher in the roots of RA [8]; thus, linking the gene expression and physiological responses to known ABA effectors that were not investigated here. The physiological and sensitive transcriptomic response that occurs at milder longer-term WD in RA indicates RA may be able to take advantage of moderate or longer-term WD by maintaining open stomata longer than the other species despite decreasing water availability. In the rapid dehydration WD, RA had lower [ABA] and NCED3 NRQ Drought tolerance is the ability of a vine to sustain physiological activity while minimizing or repairing damage during WD [130]. The drought tolerance of plants is associated with stomatal behaviors in response to WD and resultant use or preservation of available water. Grapevine responses to WD are characterized as belonging to a spectrum of stomatal reactions in the face of WD ranging from isohydric to anisohydric [131]. Isohydric species maintain a relatively constant leaf water potential through early stomatal closure during WD [132] like RI. Anisohydric species experience decreasing leaf water potential and maintain open stomata during WD [132] like RA. It is possible the iso-and aniso-hydric behaviors are controlled by different mechanisms (chemical and/or hydraulic regulation) altogether, by different mechanisms for different WD severities (moderate versus severe), or by different mechanisms at different time points (initial versus long-term) during a drought [132]. Isohydric behavior is often considered advantageous for conferring drought tolerance [131][132][133][134]. However, by the de nition of drought tolerance [130], it may be worth reconsidering the association of isohydric grapevines with drought tolerance in favor of connection to drought avoidance. A vine that can maintain stomatal aperture (i.e. anisohydric), photosynthesis, and other physiological functions under decreasing water availability like RA may be considered more drought tolerant than a vine that shuts down early when experiencing stress like RI.
Overall, ABA biosynthesis appeared to be a WD response that differentiated the species. The time frames investigated in these experiments emphasize the ABA metabolism and WD response of the different species depends on duration and severity of stress. WD experiments must be designed and compared to each other very carefully for this reason. More time points should be investigated in the future to better understand differences in short-and long-term WD and the transition between them in terms of ABA metabolism.
The rapid dehydration response varies and is inconsistent with long-term slower dehydration responses Previously, a similar rapid dehydration experiment was performed [135]. The average water lost at two (-0.139 ± 0.022 g) and four (-0.122 ± 0.012 g) hours of rapid dehydration treatment lost were comparable to the previous experiment after two (~ -0.1 g) and three (~ -0.14 g) hours of a similar rapid dehydration treatment. However, in another similar rapid dehydration experiment where microarray transcript quanti cation was performed, but water loss was not quanti ed, NCED3 transcript abundance showed different expression patterns in CS, RA, and RI than were observed here [136]. These differences may be real or due to different transcriptomic technologies. Microarray data are often subject to crosshybridization of the probes and can be less reliable, such as the probe for NCED3 (VIT_19s0093g00550), which may cross-hybridize [137].The [ABA] curve of RI in the current rapid dehydration assay mirrored that of the NCED3 transcript curve that was previously observed and was expected in this experiment.
The minor increase in RA and RI NCED3 transcript abundance at 24 hours of treatment was the most distinct difference between this and the previous experiment. In addition to the differences in technology, this distinction may be a result of differences in the age of leaves selected, the time of year, time of day, or even different experiences the plants had with WD in the past [138]. Control NCED3 transcript abundance was comparable for all species and time points between this and previous experiments. NCED3 protein concentration was not changed by WD Despite differences in ABA biosynthesis transcripts and [ABA metabolite], [NCED3 protein] was generally constant for all species, treatments, and time points. Few studies consider NCED3 protein levels, although different Arabidopsis thaliana species have demonstrated similar NCED3 protein levels in response to 10 hours at low water potential, and differences in ABA accumulation were resultant of variations in protein sequence [139]. It is possible NCED3 protein sequence variation (Additional File 9) contributes to the observed differences in ABA accumulation in the species or NCED3 may have distinctive activity in the different species. In peanut, authors brie y mention increased ABA with a parallel increase in NCED3 protein [140], but this correlation was not observed in grapevine under these WD experiments. It is possible this relationship may be organ, stress, or time speci c. While NCED3 protein was not found to increase with ABA in these experiments, other proteins have been found to directly correlate with both NCED3 transcript abundance and [ABA] in plants. For example, overexpression of ATAF1 in Arabidopsis thaliana resulted in proportional increase in both NCED3 transcripts and ABA abundances in response to abiotic stress [141]. Another study in Arabidopsis thaliana examined transcript, [protein] and [metabolite] in response to illumination, which can be an abiotic stress, but NCED3 protein was not detected, indicating it was not an abundant protein under these conditions [142]. NCED3 activity has previously been screened [143], but a correlation between transcripts and protein was not made. Recently, an Arabidopsis thaliana proteome database was released examining gene expression and [protein] across organs [144] and development [145], but no such resource has been developed for abiotic stress response, making comparing this work across plant species di cult. It is also possible post-translational modi cations play an important role in NCED3 activity, but none are well described at this time [146]. To the best of the author's knowledge, no other study examining transcript, protein, and metabolite concentrations has been performed that includes NCED3 and ABA at this time.
Other factors and regulatory mechanisms may impact ABA metabolism Three levels of ABA regulation were examined in this study. However, numerous other steps of regulation may impact [ABA], [ABA metabolite], and the physiological responses a plant has to WD. For these reasons, the response of a plant to a short, long, moderate, or severe WD may be vastly different. Many levels of the biochemical regulation of WD response and ABA biosynthesis require characterization.
NCED3 has uncharacterized phosphorylation and other potential post-translational modi cation sites that may affect activity or localization [147]. NCED3 may also interact with other proteins with unknown consequences. Alternative splicing and alternative 3'UTR usage may affect transcript function, lifetime, and localization may potentially signi cantly impact or optimize a WD response based on the severity and duration of the stress [12,148,149]. Signal peptides like CLE25 [109], miRNA [14], siRNA [15], and other regulatory molecules likely play important roles in ABA metabolism and WD response and require further investigation.
Lastly, grafting interactions may affect ABA metabolism, signaling, sensitivity, and physiological responses to WD. Traditionally, grapevines are cultivated by grafting a desirable fruit bearing scion onto an adventitious rootstock selected based on environmental conditions. Grafting enables desirable traits of a rootstock like disease resistance or salt tolerance to be conferred to the scion [150]. Previously, grafting was not found to affect the physiological response of grapevine to a short WD [151]. However, grafting has indisputable effects on both the rootstock and scion [14,117,152,153], and the molecular effects grafting may have on ABA metabolism and WD response remain unresolved. Further experiments are underway to better understand the complex effect grafting has on ABA metabolism and WD response in grapevine under long-term and cyclic WD.

Conclusions
ABA biochemistry was investigated at three levels of regulation (transcript, protein, and metabolite) across three experiments of varying severity and duration in the leaves and roots of four grapevine species. RNA-Seq analysis and metabolite quanti cation demonstrated ABA metabolism was a major WD response. ABA metabolites could in part be explained by upstream ABA metabolism gene transcript abundances. Gene expression pro ling, DEA, WGCNA, GO enrichment analysis, and RT-qPCR supported NCED3 as a key WD response gene. NCED3 was highly expressed and signi cantly differentially expressed in response to WD in leaves and roots of all four species. NCED3 was also the only ABA metabolism gene identi ed as a hub gene in WGCNA modules corresponding both to WD and the ABA metabolites. Finally, NCED3 was supported as a hub gene in GO by the presence of biosynthetic and abiotic stress response GO terms in the WD and ABA metabolite correlated modules. Western blots demonstrated NCED3 protein abundance did not appear to change in response to WD indicating that protein activity may be different from protein abundance. There was unique and speci c ABA metabolism regulation and WD response that occurred in the leaves and roots. Grapevine species demonstrated a spectrum of physiological, biochemical, and metabolic responses to different WD conditions. These responses depended on the duration and severity of WD. RA responded earlier to long-term WD with higher NCED3 transcripts and [ABA] while maintaining physiological activity longer at the cost of water availability. During long-term WD, RI generally experienced a smaller spike in NCED3 transcripts and [ABA], which were maintained at a constant level throughout the stress in parallel with quick closing stomata. NCED3 transcript abundance did not mirror [ABA] in all species during short-term WD, indicating the species have distinct ABA metabolism responses to different WD severities and durations. Thus, this study shows that ABA metabolism and regulation in grapevine species is variable and complex.
Additional studies are needed to further elucidate this interesting and important topic in order to produce better more drought tolerant crops.

Small pot experiments
After pots were prepared, 100% relative soil water content (RSWC) was measured as the weight of the individual pot two hours after irrigation to the point of saturation (water owing from bottom of pot).
Each pot was covered with aluminum foil to minimize evaporation. Experiments performed in the small pots included 1) water de cit and control treatments for one-and two week moderate WD used for RNA-Seq, ABA metabolite derivative quanti cation, and physiological measurements 2) one-week severe WD and Control treatments to support the original RNA-Seq experiment and quantify NCED3 transcripts and NCED3 protein as well as ABA (in addition to original measurements reported elsewhere [8]).
For the original moderate WD RNA-Seq experiment previously described [8], 3-5 individual CS, RA, RI, and SC vines were subjected to one or two week of moderate water de cit, or well-watered daily complete nutrient solution. At W1 and W2 of treatment various measurements were taken [8]. Total plant leaves (excluding petiole) and roots were harvested after W1 and W2 of treatment. Sand was removed from root samples by brie y washing in room temperature tap water for 10 seconds and patting dry on paper towels for an additional 10 seconds. Samples were immediately frozen in liquid nitrogen and stored at -80°C.
For the one-week severe WD experiment, 3-5 individual CS, RA, RI, and SC vines were subjected to W1 of severe water de cit or watered daily to saturation with a complete nutrient solution. At W1 of treatment various measurements were taken [8], and total plant leaves (excluding petiole) and roots were harvested. Sand was removed from root as in the moderate WD experiment. Samples were immediately frozen in liquid nitrogen and stored at -80°C.

Rapid dehydration experiments
Page 23/41 The day before the experiment, the rst mature leaf (5-6th node from the apical meristem) of CS, RA, and RI vines were measured from petiole attachment point to the tip of the leaf down the midvein and marked with a tag to ensure leaves of similar developmental stage and leaf area were used for this experiment.
Also, on the day before the experiment, dehydration chambers were prepared. Air tight dehydration chambers (1.2 L Rubbermaid Takealong © container, Newell Rubbermaid, Atlanta, GA, USA) were prepared as before [135], containing 50 mL 333 mM NaCl or DI water (for rapid dehydration and Control treatment, respectively) and a wire support grid. Dehydration chambers were placed in a 27 ˚C growth chamber (~ 200 µmol m − 2 s − 1 ) to equilibrate overnight. The following day, two hours before solar noon, the marked leaves were cut under water, weighed, and quickly placed in a dehydration chamber (~ 30 secs). WD leaves were placed abaxial side up with no contact to the salt solution. Control leaves were placed adaxial side up with petioles dipping into the water to prevent dehydration and provide petiole Turgor was calculated as the difference between leaf water potential and osmotic pressure [102]. A fth leaf was used to measure leaf water content. Water content was calculated as the difference between the leaf fresh weight at harvest and the leaf dry weight divided by the difference between turgid leaf weight and the leaf dry weight [154]. All leaves were weighed before nal measurements were performed to determine water loss. This experiment was performed ve times with each leaf coming from an individual vine. Three leaves were combined for NCED3 transcripts, NCED3 protein, and ABA quanti cation per round to have enough material for all measurements.

RT-qPCR
RT-qPCR was performed as previously described [8]. Brie y, all samples were ground with a mortar and pestle under liquid nitrogen. RNA extraction was performed with a previously described CTAB based extraction and LiCl precipitation following cleanup with the Spectrum™ Plant Total RNA kit (Sigma-Aldrich). All RNA extractions were treated with RNAse-free DNase I (Qiagen) to remove genomic DNA contamination. Samples were quanti ed on a nanodrop, and RNA quality was con rmed with gel electrophoresis by loading 250ng RNA on a 1.2% gel as well as checked for the presence of gDNA with a GoTaq Green-LAR based PCR analyzed with electrophoresis on a 2% gel employing 400ng of RNA.
Primers were designed using NCBI Primer-BLAST. Primer sequences were previously described [8]. for 15 s. Fluorescence was recorded after each cycle and melting curve analysis was performed from 65°C to 95°C. Reference genes were selected based on a low coe cient of variation of expression reported in literature and uniform expression for all cDNA samples for each of the above-described experiments. NCED3 transcript abundance for the W1 severe WD experiment was previously reported [8].

Western Blots
The IRC was spiked with 2 kDa peptides the antibodies were designed against (Additional File 10) to act as a positive control. Several postulated NCED3 degradation products or subunits were also detected (Additional File 10), but only the 67 kDa band was used for relative quanti cation.
Relative NCED3 protein abundance was quanti ed using Western Blots with internal NCED3 peptide standards and StainFree membrane normalization using an IRC loaded on each gel [101]. Proteins were extracted as previously described with chloroform/methanol [170] for one-week severe WD samples and in 2x laemmli buffer for rapid dehydration samples. peptide standard (Paci c Immunology) that acted as a positive control. Proteins were transferred onto a PVDF membrane (BioRad) using a Trans-Blot Turbo (BioRad) and imaged with a Chemidoc (BioRad) for total protein [101]. Membranes were blocked and probed with NCED3 (rabbit) primary antibodies (each 1:1000) (a 1:1 mix of antiNCED3.1 and 3.2 (Paci c Immunology)) (Additional File 9) followed by goat anti-rabbit-HRP (BioRad) secondary antibody (1:000). ECL-Spray (Advansta) was applied uniformly to the membrane, and the membranes were imaged with a Chemidoc (BioRad) for 12 secs. NCED3 protein in samples was quanti ed in ImageLab (BioRad) relative to the IRC ran on each gel and normalized to a fragment of total protein free of transfer artifacts from a StainFree membrane image [101]. Antibody (antiNCED3.1, antiNCED3.2, and antiNCED3.3) target sequences were determined to be NCED3 speci c relative to the other CS NCEDs via BLAST, unique from each other, and present in available Vitis genomes (PN40024 [99], CS [171], and RI [158]) (Additional File 9). Antibody detection was linearly related to the amount of protein loaded and deemed to be su cient for relative quanti cation of protein amount (Additional File 9).

Statistical analysis
Statistical analysis was performed comparing multiple means including one-, two-, three-, and four-way ANOVAs after assumptions were met. Non-normal data was box-cox transformed to meet normality and homoscedastic assumptions [172]. Post Hoc tests were performed with Tukey's Test HSD for comparisons between species, treatments, and time points after assumptions were met. Asterisks

Consent for publication
Not applicable.
Availability of data and material RNA-Seq data were deposited in the Sequence Read Archive (SRA) database with the accession number PRJNA516950 in a previous work [8]. All other data available upon request to corresponding author.