Agrobacterium tumefaciens (A.t.)-mediated gene transfer was performed on embryogenic calli of “microvine 04C023V0006” (derived from a cross between “Grenache” and the original L1 mutant microvine, Chaïb et al., 2010), “Chardonnay” and “Brachetto” genotypes according to Dalla Costa et al. (2014). Experiments were carried out using A.t. strain EHA105 (Hood et al., 1993) carrying a pK7WG2 plant binary vector (Karimi et al., 2002), with the muscat (M) or the neutral (N) allele of VvDXS1 under the control of the CaMV-35S promoter. The two forms of the gene differ for one nucleotide substitution at position 1822 (G in the neutral allele or T in the muscat allele), which results in the substitution of Lysine with Asparagine at position 284 in the protein sequence. As selectable marker the neomycin phosphotransferase II (nptII) gene was used, which confers resistance to kanamycin.

Transformed and wild-type in vitro plantlets were acclimatized in a growth chamber (94.5 μmol·m⁻²·s⁻¹ cool white light and 16 h-light photoperiod, at 25°C and 70% humidity) and subsequently transferred to the greenhouse.

The PCR was performed in a 20 μl reaction volume containing 100 ng of leaf DNA, 0.25 mM dNTPs, 0.3 μM of each primer (Fw: ATTGCTGTCATAGGTGATGGAG; Rv: CTGTTGTCTTGGTACTCTTAAC), 1X Taq Buffer Advanced (5 Prime, Hilden, Germany) and 1 unit of 5 Prime Taq DNA Polymerase (5 Prime, Hilden, Germany). The initial denaturation at 95°C for 5 min was followed by 35 cycles of 30 s at 95°C, 30 s at 58°C, and 30 s at 68°C, with a final extension of 7 min at 68°C. The obtained amplicon (423 bp) was digested with the FastDigest StyI restriction enzyme (Thermo Fischer Scientific, Waltham, MA, USA) to detect SNP1822 G/T (StyI recognizes the restriction site when G is present) or sequenced with the 3730xl DNA Analyzer (Applied Biosystems, Foster City, CA, USA).

Digoxigenin-labeled probes for VvDXS1 and nptII genes were obtained with the PCR Dig Probe Synthesis Kit (Roche Diagnostics, Indianapolis, IN, USA), using the following primers: VvDXS1-SB-fw = ATGGCTCTCTGTACGCTCTCA and VvDXS1-SB-rv = AGTTGTTTCAGCTCCTTGACAG; nptII-SB-fw = GAAGGGACTGGCTGCTATTG and nptII-SB-rv = AATATCACGGGTAGCCAACG. For each sample, 10 μg of DNA was digested with the Fastdigest® restriction enzymes XbaI and EcoRI (Thermo Fischer Scientific, Waltham, MA, USA) for VvDXS1 probing and with HindIII for nptII probing. Digestion products were precipitated, resuspended in 30 μl Milli-Q water and separated overnight on 0.9% agarose gel (0.5X TBE) at 50 V. Membrane blotting and hybridization were performed following Roche user's manual. The autoradiographic film was exposed overnight before development.

Quantitative real-time PCR (qPCR) amplification was performed on genomic DNA in 96-well reaction plates on the iCycler iQ Thermocycler (Biorad, Hercules, CA, USA) according to the method by Dalla Costa et al. (2009).

Total RNA was isolated from grape leaves, flowers and fruits at different development stages using the Spectrum™ Plant Total RNA Kit (Sigma Aldrich, St. Louis, MO, USA) and quantified with the spectrophotometer NanoDrop ND-8000 (NanoDrop Technologies, Wilmington, DE, USA). Following DNase treatment, 2 μg of RNA were retro-transcribed into cDNA with the SuperScript® III Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA) and oligo(dT) (or random primers for applications with TaqMan array cards). The qPCR was carried out as described in Supplementary Materials.

A set of terpenoid biosynthetic genes were selected for loading onto the TaqMan array card based on their expression in aromatic grapevine cultivars. For this purpose, four normalized cDNA libraries were obtained using the extracted RNA from ripening berries of the cultivars Gewürztraminer (TRA), Malvasia di Candia aromatica (MAL), Moscato Bianco (MOB) and Rhein Riesling (RIE). The normalized cDNA libraries were then sequenced with a 454 GS FLX Titanium system (Roche, Indianapolis, IN, USA) and filtered high quality reads were aligned to the 12x V1 version of the Vitis vinifera genome (details are provided in Supplementary Materials). In parallel, a total of 180 genes involved in terpenoid biosynthesis in grapevine were retrieved from KEGG (http://www.genome.jp/kegg-bin/show_organism?org=vvi) by searching for the following ko terms: vvi00900 (terpenoid backbone biosynthesis), vvi00902 (monoterpenoid biosynthesis), vvi00904 (diterpenoid biosynthesis), vvi00905 (brassinosteroid biosynthesis), vvi00906 (carotenoid biosynthesis) and vvi00909 (sesquiterpenoid and triterpenoid biosynthesis). An additional 20 genes were selected based on their co-localization with grapevine QTLs for monoterpenoid content (Battilana et al., 2009) or peculiar expression in ripening berries of aromatic grapes (unpublished data). All the above mentioned genes were checked for their expression in the normalized cDNA libraries and were manually investigated in order to precisely define their gene structure, allelic and possible splicing variants.

TaqMan array (TA) cards (Applied Biosystems, Foster City, CA, USA) are 384-well microfluidic cards with eight ports, each containing 48 connected wells. Primers and probes are preloaded and dried onto the wells by the manufacturer at the following concentrations: 9 × 10⁻⁷ mol/L for primer, 2 × 10⁻⁷ mol/L for probe. All the probes are conjugated at 5′ to a reporter 6-carboxyfluorescein (FAM) and at 3′ to a non-fluorescent quencher (NFQ) with the minor groove binder (MGB) moiety attached to the molecule. In the TaqMan array card developed in our study (Aromix), four samples (for details see Results) can be assessed simultaneously for 83 targets connected with the terpene metabolism (Table S1). The card also features five endogenous genes, as well as the manufacturer's card control PCR. Probes and primers were designed by the manufacturer based on the gene sequences supplied by the authors. All TA cards were run on the ViiA™ 7 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). Each of the eight ports of the card was loaded with a 100 μl solution obtained by mixing 50 μl of 1:20 diluted cDNA with 50 μl of TaqMan Gene Expression Master Mix 2X (Applied Biosystems, Foster City, CA, USA). The cards were centrifuged twice at 1,200 rpm for 1 min and sealed, the loading ports were excised and the cards were placed in the thermal cycler. The following cycling conditions were used: 50°C for 2 min, 95°C for 10 min, and 40 cycles of 95°C for 30 s followed by 60°C for 1 min. Results were processed with qbase^PLUS software (Biogazelle, Zwijnaarde, Belgium; Hellemans et al., 2007) and normalized by the reference genes selected from qbase^PLUS (actin and glyceraldehyde-3-phosphate dehydrogenase).

Approximately 90 grapevine cultivars representing Muscat, herbaceous, or other distinct flavored cultivars, as well as non-aromatic ones were analyzed at technological maturity (harvest time) for monoterpene and sesquiterpene content in whole berries. Berry flavor phenotype as assessed merely by tasting and the VvDXS1 genotype were previously reported for the majority of these accessions (Emanuelli et al., 2014).

For each microvine line, berry number and size, total leaf area and photosynthetic activity were evaluated in four biological replicates 1 month before berry sampling. Targeted cluster thinning or leaf pruning were carried out in order to set a similar ratio between total leaf area and number of berries for all the plants under analysis (Table S2). Clusters were collected at technological maturity (18–20.5 degrees Brix,°Bx) and berry skin was separated from the pulp. Skins were grounded to a fine powder in liquid nitrogen and used for monoterpene and sesquiterpene analysis.

Eleven monoterpenes were extracted from the starting plant material and quantified in their free and glycosidically bound form by solid phase extraction (SPE) and high-resolution gas chromatography-mass spectrometry (HRGC-MS), as previously described by Battilana et al. (2011). A non-targeted analysis was carried out to profile the sesquiterpenes.

Genetically modified plants were obtained from microvine (Mi), Chardonnay (C) and Brachetto (B) genotypes (Figure 1, Figure S1). Microvines were chosen, since they have short generation cycles and continuous flowering, which ensure fruit production and significantly reduce the time required for genetic studies in grapevine (Chaïb et al., 2010). The identity of the VvDXS1 integrated allele (neutral or muscat form) was confirmed by sequencing. The number of T-DNA integration copies as calculated by real-time PCR and Southern blot is shown in Figures 1A,B, Figure S1A. One T-DNA integration copy was detected for Mi-N4, two for C-M5, C-M6 and B-N8, while Mi-M1 and the remaining lines transformed with the neutral allele presented a multi-copy insertion.

The qPCR expression analysis proved that VvDXS1 transgene was transcribed in the leaves of all transformed lines, although the level of expression was variable across the lines. Conversely, endogenous VvDXS1 was stably expressed (Figure 1C, Figures S1B, S2). After this evaluation, Chardonnay and Brachetto plants—which were not expected to produce fruits at least in the short-term—were not further investigated and were kept for future experiments. Several biological replicates of each transformed and WT microvine line were transferred to soil and successfully acclimatized in the greenhouse in 2013, 2014, and 2015 (Figure 2). Data on flowering time and berry development were collected for Mi-M1, Mi-N3, Mi-N4, and Mi-WT acclimatized in January 2014 (Table 1). The first microvine plants began to bloom about 150 days after acclimatization and anthesis occurred on average 197 days after acclimatization in 11 out of the 14 observed plants. Eight microvine plants reached veraison and maturity, respectively 281 and 309 days (on average) after acclimatization. Modified plants did not show any evident phenotypic difference in vitro or in greenhouse conditions, compared to the control plants (Table S2).

The expression of both endogenous and transgenic VvDXS1 in different plant tissues and during fruit development was investigated in Mi-N4, the microvine line with the highest transcription of the single copy transgene in leaves, and in Mi-M1, the only microvine line transformed with the muscat allele (Figure 1). The mRNA level of VvDXS1 transgene was significantly higher in the leaf in comparison to flower at anthesis (stage E-L 21 of the modified E-L system by Coombe, 1995) and berry at pre-veraison (E-L 32), veraison (E-L 35) and maturity (E-L 38). Following berry development, the transgene transcript level decreased from pre-veraison to veraison and returned to increase after veraison. A similar trend was observed for the endogenous VvDXS1 mRNA, even though the recovery following veraison was not detected (Figure 3, Figure S3). A comparable time-course profile, with the minimum point of the convexity corresponding to the veraison stage, was also obtained from the total VvDXS1 expression analysis in three microvine lines (Figure S4).

In order to define a minimal gene set that could describe the transcriptional perturbation of terpenoid biosynthesis, a manual selection of candidate genes actually expressed during berry ripening in four aromatic grapevine cultivars was performed. By combining the cDNA libraries of the four varieties a total of 21246 genes was found to be expressed (17603, 18032, 17724, and 18198 in MAL, MOB, RIE, and TRA libraries, respectively), of which 14875 were common to all four cultivars (data not shown).

Two-hundred genes involved in terpenoid biosynthesis were checked for their expression in the normalized cDNA libraries and a final set of 78 genes (including five endogenous genes) was selected to be included in the TaqMan array card. For 10 of these genes two specific probes were designed in order to discriminate between splicing variants, which resulted in a total of 88 probes (Table S1).

The relative expression of these genes was evaluated in Mi-M1, Mi-N4, and Mi-WT plants during the 2014 growing season at four phenological stages (Table S3). In addition to the five endogenous genes, 64 terpenoid biosynthetic genes were expressed in the plants under study, of which 47 were expressed in all the investigated phenological stages while 17 were expressed only in some stages. Finally, nine genes on the TaqMan array could not be amplified, the majority of which (six) coded for terpene synthases.

Overall, the genes belonging to the MEP and the MVA pathways showed an evident up-regulation in the transformed lines at anthesis and veraison and a moderate down-regulation in the pre-veraison stage compared to WT. This general trend was also observed for the genes involved in the metabolism of carotenoids while the monoterpene and sesquiterpene biosynthetic genes were more highly expressed at anthesis. Based on ANOVA and t-tests, 24 genes showed significantly different expression (P < 0.05) among the analyzed plants during the four phenological stages (12 at anthesis, nine at veraison, four at pre-veraison and three at maturity) (Figure 4).

The simultaneous analysis of splicing variants of ten genes did not highlight relevant differences in most cases. However, for VIT_04s0079g00680 (PSY) and VIT_12s0057g01200 (AACT) only one splicing variant was expressed throughout berry development. A differential profile was also observed when both isoforms were expressed at least in one stage, as for VIT_06s0009g00770 (AAO3) and VIT_18s0001g10500 (ABA 8′-HX) (Figure S5, Table S3).

The expression data derived from the TaqMan card assay were validated with sybr-green real-time PCR on a set of relevant genes belonging to the MEP-, MVA-, carotenoid-, sesquiterpene-, and monoterpene biosynthetic pathways (Table S4). A strong correlation between the two sets of measurements was observed, with Pearson correlation coefficient (R) values approximately or >0.9 for seven genes and between 0.8 and 0.7 for three genes (Figure S4). Difference in expression levels between transformed and WT lines was the most notable at anthesis, and therefore a further expression analysis on a set of modulated genes (VvDXS1, VvAACT, VvHMGS, VvHMGR3, VvOciS, VvFPPS, VvValS) was repeated by qPCR on flowers collected in triplicate in 2016. The significant differences found in 2014 were confirmed in 2016 (data not shown).

The first indications of the role that VvDXS1 plays in the genetic control of monoterpene biosynthesis derived from the analysis of segregating progenies (Battilana et al., 2009; Duchêne et al., 2009) and were subsequently confirmed by a genetic association study based on berry taste (Emanuelli et al., 2010). In the present work a grapevine germplasm collection was considered, for which VvDXS1 genotype and terpene content (as assessed by chemical methods) were analyzed in parallel. This analysis clearly highlighted that sequence variation of VvDXS1 has a very high predictive value for the accumulation of monoterpenes and, as a novel finding, also of sesquiterpenes. The strength of such genetic effect is even more evident if considering that the investigated plants have been grown under field conditions and have been influenced by uncontrolled environmental factors.

A metabolic engineering approach was adopted for studying the regulatory role of VvDXS1 on terpenoid metabolism in grapevine, given its relevance for the quality of grapes and wines. Besides Chardonnay and Brachetto, which are frequently used cultivars amenable to genetic transformation (Iocco et al., 2001; Dalla Costa et al., 2009; Dhekney et al., 2011; Perrone et al., 2012), the microvine genotype was also employed. Several genetic experiments have used the microvine (Luchaire et al., 2017) since it was presented as a grapevine model system for functional genomic studies (Chaïb et al., 2010), however no studies based on engineered microvine plants have been published so far. The present work was focused on the microvine transformation with different VvDXS1 alleles. The underlying goal was to investigate two potential perturbation mechanisms, the former based on a significant over-expression of the wild-type allele and the latter on the ex-novo expression of an enzyme with increased catalytic efficiency from the mutated allele. Here, we obtained a total of 13 transformed lines (nine microvines, two Chardonnay and two Brachetto) carrying the mutated or non-mutated form of VvDXS1 gene with integrations ranging from a single copy to multiple copies (Figure 1, Figures S1, S2).

A high level of transgene expression was measured in the lines Mi-N4, C-M5, C-M6, and B-N1. Conversely, a low expression rate was observed in the other lines, likely due to multi-copy insertions that may produce silencing effects or to the “position effect,” which is related to the genomic region where the T-DNA copies are integrated. Both causes of epigenetic silencing are well documented in literature (Matzke and Matzke, 1998; De Buck et al., 2007; Tang et al., 2007) and nowadays they represent two of the major concerns in plant genetic engineering (Rajeevkumar et al., 2015).

The period required by 04C023V0006 microvine plants for flowering in the greenhouse (on average 197 days in Table 1) was four times longer than that reported by Chaïb et al. (2010) for two different microvine genotypes. Alternatively, the time needed to reach fruit maturity (berry sugar content >18 °Brix) was consistent with that observed by Chaïb (i.e., 16 weeks post anthesis). Comparing microvine with the conventional grapevine genotypes, we confirmed that the time necessary for flowering is dramatically reduced while fruit ripening times are very similar (data not shown). Moreover, the VvDXS1 insertion did not affect the phenology or the morphology of the microvine plants (Table 1, Table S2).

The spatio-temporal analysis of the VvDXS1 transcript in microvine confirmed that this gene is not expressed in a steady-state level in different grapevine organs and during fruit ripening but it is highly modulated as previously observed in Moscato Bianco (Battilana et al., 2011). Moreover, the VvDXS1 expression trend in the transformed lines Mi-M1 and Mi-N4 was similar to that in the WT plant (Figure 6, Figure S4). The modulation of VvDXS1 was even more evident when the expression profile of the endogenous gene and transgene were simultaneously analyzed in Mi-M1 and Mi-N4 lines (Figure 3, Figure S3). Under these conditions, the transgene proved to be controlled by the same organ-specific and developmental signals as the endogenous counterpart. In particular, a significantly higher expression was detected in the leaf with respect to the other tissues, an outcome which may be explained by the strong requirement of chlorophyll and carotenoids during leaf maturation (Estévez et al., 2000). Regarding the berry developmental stages, a significant increase was assessed after veraison, confirming the findings of Battilana et al. (2011). Such data may indicate additional levels of gene expression regulation occurring post-transcriptionally or with a feedback mechanism, as reported for other plant species (Hemmerlin et al., 2012).

The TaqMan array, which was developed ad hoc to harbor as many as possible truly expressed genes, allowed to evaluate the transcription profile of a number of terpenoid pathway genes in the same technical conditions. Four important stages in the grape reproductive life were considered (Figure 4).

The microvine system has allowed to evaluate the metabolic profile of fruits engineered for an important regulatory gene of terpenoid metabolism. We detected significant differences between transformed and control grapes in the accumulation of monoterpenes at harvest time. In particular, the total monoterpene content was 1.7- and 4.4-fold in Mi-N4 and Mi-M1 lines with respect to WT microvines, with ratios ranging from 1.3 to 1.9 (Mi-N4) and from 3.2 to 4.7 (Mi-M1) for free and glycosidically bound monoterpenes, respectively (Figure 6). These values are similar, albeit often higher in the case of Mi-M1, to those reported in previous experiments assessing the enhancement of isoprenoid compounds upon DXS overexpression in other plants (Lois et al., 2000; Estévez et al., 2001; Enfissi et al., 2005; Carretero-Paulet et al., 2006; Morris et al., 2006; Muñoz-Bertomeu et al., 2006; Peebles et al., 2011; Vaccaro et al., 2014; Wright et al., 2014; Shi et al., 2016; Simpson et al., 2016; Zhou et al., 2016; Jadaun et al., 2017).

Even if we cannot exclude an effect of the gai mutation on terpene biosynthesis (Hong et al., 2012; Murcia et al., 2017), our findings conform with the idea that DXS1 ectopic expression can raise the metabolic flux through the MEP pathway, thereby improving the formation of isoprenoids. Comparing Mi-N4 with WT plants, a close relationship may be observed between the increase in DXS1 total transcripts (2.2-fold) and its end products (1.7-fold increase in total monoterpene content) (Figure 6). Oppositely, in the case of Mi-M1 line, a weak expression of the DXS1 mutated allele resulted in an important gain in total monoterpenes (4.4-fold increase) (Figure 6). This outcome is in line with the results of Battilana et al. (2011) who found improved catalytic performances of the muscat DXS1 enzyme in comparison with the neutral form resulting in enhanced monoterpene biosynthesis.

The effect played by VvDXS1 on sesquiterpene content in the FEM aromatic core collection could not be confirmed in the microvine lines analyzed here, as they did not accumulate sesquiterpenes in ripe berry skins. A possible explanation is that the microvines have been grown in a greenhouse (according to the extant restrictions on GMOs in Italy), whereas the germplasm collection is planted in open field. Although the investigated microvines proved to be a convenient model system to detect differences in terpenoid pathway genes and monoterpene content among lines, it is evident that such system maintained in the greenhouse does not reproduce exactly the environmental conditions present in the vineyard (especially the exposure to light). Moreover, additional genes might play a limiting role in terpene production in the microvine genetic background. Both facts might also explain the lower monoterpene content of the transformed microvines (Figure 6) compared to the monoterpene content of the germplasm accessions with the muscat mutation (Figure 5A). An effect of VvDXS1 on sesquiterpene biosynthesis cannot be excluded in flowers, which represent the only tissue where a significant modulation of sesquiterpene pathway genes was observed (Figure 4). However, the quantification of sesquiterpenes in flowers was out of the aim of the present study.

The metabolic engineering approach adopted in the present work provided new insights into the functional effect of VvDXS1 alleles on terpenoid metabolism in grapevine. In order to optimize this basic approach from a biotechnological point of view, one should keep in mind some important aspects. First, DXS1 is post-transcriptionally regulated both at the level of gene expression (e.g., the DXS down-regulation by PSY activity and carotenoid synthesis) and of protein abundance/activity (e.g., the feedback inhibition of DXS1 activity by IPP and DMAPP), which is especially important in vivo to rapidly link the pathway with environmental and physiological challenges or metabolic fluctuations (Lois et al., 2000; Banerjee et al., 2013; Hemmerlin, 2013; Ghirardo et al., 2014; Pokhilko et al., 2015; Rodríguez-Concepción and Boronat, 2015). Secondly, the MEP pathway flux may be diverted via cross-talk with the MVA route (Hemmerlin et al., 2012; Pazouki and Niinemets, 2016) or via export of intermediates like methylerythritol cyclodiphoshate (MecPP), as observed in transformed Arabidopsis plants overexpressing DXS (Xiao et al., 2012; Wright et al., 2014; González-Cabanelas et al., 2015), and hydroxymethylbutenyl diphosphate (HMbPP) (Ward et al., 2011). Finally, other limiting steps in the MEP-pathway or in upstream and downstream pathways may exist, which includes the need for coordination between plant development and secondary metabolite production in order to not compete for carbon sources (Estévez et al., 2001; Enfissi et al., 2005; Muñoz-Bertomeu et al., 2006; Rodríguez-Concepción and Boronat, 2015; Shi et al., 2016; Zeng et al., 2016). In this regard, the analysis of additional metabolites (e.g., carotenoids and chlorophylls) could point out potentially competing reactions. The involvement of multiple key enzymes might also explain the lack of linalool in the microvines under investigation; in particular, terpene synthases or other genes in the confidence interval of the previously identified linalool-specific QTL on chromosome 10 (Battilana et al., 2009; Emanuelli et al., 2011; Costantini et al., 2017) might play a limiting role in linalool production in combination with VvDXS1, as highlighted in the grapevine germplasm (Figures 5C,D) and reported in other plants (Zeng et al., 2016).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.