Rooted petunia plantlets (Petunia x hybrida L. lines P720, W115 [cv. MD], Blue Spark, and Blue Ray) were obtained from Danziger–Dan Flower Farm (Mishmar Hashiva, Israel). Plants were grown in the greenhouse under a 25/20°C day/night temperature regime and a 12/12 h light/dark photoperiod, with light period starting at 0800 h.

For dynamic headspace analysis (Spitzer-Rimon et al., 2012) of different developmental stages, petunia P720 flowers [ranging from 3.5 cm buds to 3 days postanthesis (DPA)] were collected at 1600 h. For analysis of diel floral emissions, petunia P720 flowers were collected at anthesis stage and headspace was initiated at 0300 h with 4-h intervals for 24 h. Volatiles emitted from detached flowers (three flowers per sample) were collected using an adsorbent trap consisting of a glass tube containing 200 mg Porapak Type Q polymer (80/100 mesh; Alltech) held in place with steel mesh plugs. Trapped volatiles were eluted using 1.5 ml hexane, and 2 μg isobutylbenzene was added to each sample as an internal standard. A calibration plot with increasing isobutylbenzene concentrations was generated to determine the relative amounts of target volatiles within each sample.

To determine the pool sizes of free volatile compounds (aglycones) in the corolla limbs (Spitzer-Rimon et al., 2012), petal tissues (150 mg fresh weight from at least three different flowers) were collected, ground in liquid nitrogen and extracted in 800 μl hexane containing 2 μg isobutylbenzene as the internal standard. Following a 2 h incubation with shaking at 150 rpm, extracts were centrifuged (10,000 g for 10 min) and the supernatant was further centrifuged and evaporated prior to chromatography. Glycosylated scent compounds were extracted by grinding corolla limb tissue (150 mg fresh weight from at least three different flowers) in liquid nitrogen, suspending in 1.2 ml 80% MeOH:double-distilled water (DDW) (v/v) and sonicating for 20 min. Samples were centrifuged (16,000 g, 10 min) and the supernatant was vacuum-dried and then resuspended in 0.9 ml 0.2 M citrate–phosphate buffer, pH 5.4. Viscozyme (150 μl; Sigma Aldrich, ca. 1.2 g/ml) was added to hydrolyze and release the sugar-bound aroma compounds. Following overnight incubation at 37°C with the enzyme, samples were extracted in 800 μl hexane containing 2 μg isobutylbenzene as the internal standard, and prepared for GC–MS analysis as described above for aglycone determination. To avoid possible contamination of the preparation with aglycones prior to hydrolysis, we calibrated the protocol for glycoside extractions by extending the evaporation of the samples while heating to promote emission of volatiles. Extraction of samples with an organic solvent after this process, prior to enzymatic hydrolysis of the conjugates, showed no detectable levels of volatiles. For the assessment of diel accumulation of scent compounds (internal pools of aglycones and glycosides), flowers of petunia line P720 were collected between anthesis at 1600 h and 1 DPA at 2400 h, at 8-h intervals for the first day and every 4 h thereafter.

GC–MS analysis (1 μl sample) was performed using a device composed of a Pal autosampler (CTC Analytic), a TRACE GC 2000 equipped with an Rtx-5SIL mass spectrometer fused-silica capillary column (Restek; i.d. 0.25 μm, 30 m × 0.25 mm) and a TRACE DSQ quadruple mass spectrometer (ThermoFinnigan). Helium was used as the carrier gas at a flow rate of 1 ml/min. The injector temperature was set to 220°C (splitless mode) and the interface to 240°C, and the ion source was adjusted to 200°C. The analysis was performed under the following temperature program: 2 min of isothermal heating at 40°C followed by a 7°C/min oven temperature ramp to 250°C then 2 min of isothermal heating. The system was equilibrated for 1 min at 70°C before injection of the next sample. Mass spectra were recorded at 3.15 scan/s with a scanning range of 40–450 mass-to-charge ratio and electron energy of 70 eV. Compounds were tentatively identified (>95% match) based on NIST/EPA/NIH Mass Spectral Library data version NIST 05 (software version 2.0d) using the XCALIBUR v1.3 (ThermoFinnigan). Further identification of major compounds was based on comparison of mass spectra and retention times with those of authentic standards (Sigma-Aldrich) analyzed under similar conditions.

Petunia flowers (line P720) were collected 1 DPA at 1500 h. Detached corollas were excised 1.5 cm below the limbs and placed in a 150-μl solution containing 30 mM ²H₅-Phe stable isotope (98% purity, Cambridge Isotope Laboratories). After 20-min incubation, corollas were washed twice with DDW and then transferred to water for the rest of the experiment. Limb samples (150 mg) were collected at different time points (0.5, 2, 4, 6, and 9 h after transfer to water) to determine the internal pool of glycosylated scent compounds (labeled and non-labeled). Detection of conjugated compounds derived from labeled and non-labeled ²H₅-Phe was performed as described by Boatright et al. (2004). Authentic standards were all purchased from Sigma Aldrich.

Transient overexpression of pBINPLUS carrying CaMV 35S-driven AnBGL1 coding for Aspergillus niger β-glucosidase 1, with or without lytic vacuole-targeting determinants (AnBGL1vac or AnBGL1cyt, respectively) cloned at the N-terminus (Wei et al., 2004), was performed in petunia line P720 flowers as described previously (Spitzer-Rimon et al., 2012). Plasmids harboring the BGL1 constructs were sequenced to confirm their validity for further use. Petunia buds 1 day before anthesis were subjected to vacuum infiltration with a mix of Agrobacterium tumefaciens strain AGLO carrying pRCS2-35S:GFP and bacteria carrying pBINPLUS-35S:AnBGL1vac or bacteria carrying pBINPLUS-35S:AnBGL1cyt in a 1:1 (v/v) mixture. As a control, buds were inoculated with AGLO carrying only pRCS2-35S:GFP. Three days postinfiltration (at 1500 h), open flowers were examined using a fluorescence binocular microscope (FLOUIII, Leica Microsystems) under white and UV light with GFP filters. GFP-positive tissues were excised and glycosylated scent compounds were extracted and analyzed by GC–MS. For β-glucosidase activity assay, tissues (300 mg) were ground in liquid nitrogen and extracted in 1.2 ml 50 mM phosphate buffer pH 4.3, containing 10 mM EDTA and 4 mM dithiothreitol. Total protein level was measured using the Bradford method (Bradford, 1976). Enzyme activity was quantified as in Wei et al. (2004) using pNPG (Sigma Aldrich) as the substrate. Activity was calculated as the amount (in mol) of liberated p-nitrophenyl (pNP) per mg protein per 1 min of reaction time.

Protoplasts were isolated from petunia petals (line Blue Ray) as described by Faraco et al. (2014). Protoplast viability was assayed by incubating cells with 50 μg/ml fluorescein diacetate solution (Sigma Aldrich) in a 1:1 mixture (v/v) for 5 min prior to observation under the microscope. Imaging of fluorescent cells was performed by fully motorized epifluorescence inverted microscope (Olympus-IX8 Cell-R) with a 12-bit Orca-AG CCD camera using a GFP filter. Intact vacuoles were isolated from protoplasts as described previously (Robert et al., 2007) with some modifications. Briefly, protoplasts were osmotically and thermally lysed, and the vacuole fraction was separated from the lysate using a Ficoll step gradient in 15-ml tubes. A typical gradient consisted of 5.5 ml lysed protoplasts in 10% (w/v) Ficoll overlain by 4 ml of 5% Ficoll solution and 1.5 ml vacuole buffer (containing 0.25 M mannitol, 5 mM sodium phosphate pH 7.5, and 2 mM EDTA pH 8) layered over that. The tubes were centrifuged for 45 min at 1,500 g. Vacuoles were located between the 0 and 5% Ficoll layers. Vacuole purity was evaluated using light microscopy and western blot analysis (see details below). Anthocyanin content of protoplasts and vacuoles was measured as described by Spitzer-Rimon et al. (2010). Extraction of glycosides from protoplasts and vacuoles and their quantitation by GC–MS were performed as above. To extract proteins for western blot analysis, aliquots of protoplasts and vacuoles were first precipitated in 10% (w/v) trichloroacetic acid, washed with 100% acetone and resuspended in extraction buffer (0.1 M Tris–HCl pH 8, 150 mM dithiothreitol, 2% v/v SDS, 2 mM phenylmethylsulfonyl fluoride, and 5% v/v β-mercaptoethanol). Extracts were further treated as in Vishnevetsky et al. (1999). Briefly, samples were centrifuged and the supernatants were mixed with equal volumes of Tris-saturated phenol (pH 8). After centrifugation, the phenolic phases were washed with the extraction buffer, and 5 volumes of MeOH–ammonium sulfate (0.1 M) were added. Samples were washed three more times with MeOH–ammonium sulfate (0.1 M) and then with 80% acetone. The pellets were resuspended in NuPAGE LDS Sample Buffer (Thermo Fisher Scientific), and the tubes were incubated for 5 min at 55°C prior to electrophoresis. The samples were analyzed on NuPAGE 4–12% Bis–Tris gels (Thermo Fisher Scientific) with the Precision Plus Protein™ All Blue Standards marker (Bio-Rad Laboratories). The volumes of loaded samples from protoplasts and vacuoles were normalized so that they reflected an identical amount of cells/protoplasts. Samples were transferred to iBlot nitrocellulose membranes for western blot analyses and following incubation with skimmed milk-based blocking solution, membranes were reacted overnight at 18°C with the primary antibodies: epsilon subunit of tonoplast V-ATPase and cFBPase (Agrisera, diluted 1:200) (Dima et al., 2015; de Michele et al., 2016), PAL, diluted 1:1,000 (Howles et al., 1996), and LHCII, diluted 1:500 (Vainstein and Sharon, 1993). Following three 10-min washes in Tween–Tris buffered saline, the membranes were incubated for 1 h at room temperature in blocking buffer containing a 1:10,000 dilution of secondary goat anti-rabbit antibody coupled to horseradish peroxidase. After three washes, the membranes were incubated for 1 min with SuperSignal West Pico Chemiluminescent Substrate (Thermo Fisher Scientific) and developed by autoradiography using Image Quant Las 500 (GE Healthcare Bio-Sciences AB).

Significance of differences between treatments were calculated by Student's t-test, ^*P ≤ 0.05, using the SigmaPlot software v13.0 (SYSTAT).

To ascertain the prevalence of glycosylated phenylpropanoid volatiles in petunia flowers, their levels were measured using gas chromatography–mass spectrometry (GC–MS) in four Petunia x hybrida cultivars with different genetic backgrounds: Blue Ray, W115 [Mitchell Diploid (MD)], P720 and Blue Spark. These cultivars are all fragrant (Figure S1), but have different flower shapes, colors, and sizes. Glycosylated and free (non-glycosylated, aglycone) scent compounds were extracted at 1600 h—the onset of scent production—from flowers at anthesis. In all cultivars, the glycosylated fraction of the volatiles, i.e., those with glycosylation-compatible hydroxyl groups (benzyl alcohol, phenylethyl alcohol, eugenol, isoeugenol, and vanillin) was detected as the predominant one, ranging from ~72 to 90% of the combined pool of accumulated (aglycone + glycoside) compounds (Figure 1).

To detail the accumulation of glycosylated phenylpropanoid scent compounds, we measured their levels, as well as those of aglycones (accumulated and emitted) in line P720 flowers at different stages of development, at 1800 h. This line was chosen for further examination due to its prolific flowering and in-depth-characterized patterns of diel/developmental scent production (Spitzer-Rimon et al., 2012; Cna'ani et al., 2015; Ravid et al., 2017). As expected (Figure 2A), the level of aglycones increased in parallel to flower development from young bud (3.5 cm) to mature flower. Levels of volatiles emitted from petunia flowers increased similarly until 2DPAand decreased thereafter, as previously reported (Figure 2G and Figure S2; Spitzer-Rimon et al., 2010). Free phenylethyl alcohol, eugenol, and isoeugenol volatiles were below detection levels in young buds (Figures 2C–E). Similar to the aglycones' accumulation and emission, levels of glycosylated scent compounds increased in parallel to flower development (Figure 2A). The total level of glycosides increased up to 3 DPA, reaching 352-fold that in young buds (3.5 cm stage) and 17-fold that at anthesis. Overall, the fraction of glycosides out of the combined pool was 87.5% in flowers 3 DPA. At this stage, with the exception of vanillin (Figure 2F), the proportion of each individual glycoside of a scent compound out of its combined pool (glycosides + aglycones) was highest (Figures 2B–E). Among the individual glycosides, phenylethyl alcohol glycosides showed the most prominent increase, with a 397-fold change from 3.5 cm buds and a 38.5-fold change from anthesis to 3 DPA (Figure 2C). In young buds, glycosylated phenylethyl alcohol and isoeugenol were detected, while the aglycone fraction was not evident (Figures 2C,E). The opposite trend was apparent with vanillin: only aglycones and no glycosides were detected at early stages of flower development (Figure 2F). No emission of volatiles could be detected in young flower buds (Figure S2).

To characterize the diel accumulation pattern of glycosylated scent compounds and to assess its relevance to the rhythmic nature of scent production, volatile compounds accumulated in and emitted from flowers of petunia line P720 were sampled between anthesis and 1 DPA. Glycosides and corresponding aglycones were sampled in 8-h intervals at anthesis and 4-h during the following day. Headspace analysis was conducted with 4-h intervals for 24 h. GC–MS analyses of the aglycone fraction of phenylpropanoid scent compounds in the pool and headspace (Figures 3A,G, and Figure S3) revealed a typical oscillating pattern during the day, with maximum amplitude levels at around midnight and low points at midday (Oyama-Okubo et al., 2005; Fenske et al., 2015). The peak level of aglycones in the pool at 2400 h 1 DPA was 2-fold higher than at anthesis. As with scent emission (Figure 2G and Figure S3), between the end of the dark period and 1200 h there was a decrease, whereas from 1600 h onward there was a sharp increase in aglycones' internal pool levels. GC–MS analyses of the collective pool of glycosylated scent compounds revealed a different accumulation pattern: following anthesis, total pool size continued to increase from the morning hours throughout the time frame of the experiment. Highest levels of glycosides each day were detected during the dark period, with a 2.2-fold increase 1 DPA vs. anthesis at 2400 h (Figure 3A). The ratio between total glycosylated and free forms of volatiles was highest during the day (1200 to 1600 h), when accumulated aglycones and emission levels were lowest. At those time points, the glycosylated forms accounted for as much as ~70 and 80% of the combined pool, respectively. The lowest ratios were measured at midnight, when the level of free volatiles accumulated in the petals, and of those emitted, peaked. Analyses of individual floral scent compounds revealed that the diel accumulation pattern of total glycosides is largely determined by the levels of benzyl alcohol, phenylethyl alcohol, and vanillin, which together constituted up to 90% of all glycosides within the internal pool of petunia flowers (Figures 3B,C,F). Glycosides of eugenol and isoeugenol, comprising 10–25% of total glycosides (depending on the time of day), accumulated in a pattern that was somewhat similar to their corresponding aglycones, i.e., lower levels around midday and higher levels at night (Figures 3D,E).

Based on the observation that the levels of glycosylated benzyl alcohol, phenylethyl alcohol, and vanillin do not decrease during flower development or on a diel basis (Figures 2, 3), we tested whether the glycosylated pools are dynamic, i.e., that they are not simply sequestered for storage but rather also catabolized during scent production. To monitor changes in glycosides, labeling experiments using deuterium ring-labeled Phe (²H₅-Phe) were performed. This stable isotope has been used to elucidate the pathways leading to benzenoid formation in petunia flowers (Boatright et al., 2004). Flowers of petunia line P720, collected 1 DPA at 1500 h, were fed for 20 min with a solution containing ²H₅-Phe, and changes in the levels of labeled glycosides were analyzed by GC–MS over the following 9 h. The time frame of the experiment was chosen to reflect both the onset and peak in scent production. Preliminary experiments had shown that a short feeding time (20 min) is sufficient to label almost 99% of the pool of phenylacetaldehyde, the first volatile metabolite directly derived from Phe (Figure S4; Farhi et al., 2010). The detection of labeled phenyethyl alcohol, benzyl alcohol, vanillin, and isoeugenol glycosides, and the discrimination between labeled and non-labeled compounds are presented in Figures S5A–D. Labeled eugenol glycosides were not detected in our experimental setup, given that the level of eugenol glycoside is <1% of those of other detectable glycosides and its production necessitates many additional enzymatic steps as compared to, for example, phenyethyl alcohol. Levels of non-labeled glycosides of all scent compounds increased, as expected, during the course of the experiment (Figure 4). Levels of labeled phenyethyl alcohol and isoeugenol glycosides peaked 2 h after feeding with ²H₅-Phe; the former then sharply decreased to undetectable levels, while the latter decreased gradually (Figures 4A,B). Levels of labeled benzyl alcohol and vanillin glycosides also increased after feeding and then decreased 6 h after labeling (Figures 4C,D).

Flavonoids, such as, anthocyanins, proanthocyanidins, and monolignol-derived lignin monomers, as well as coumarins, hydroxycinnamates, and salicylic acid accumulate as glycosides in the central vacuole (Dean et al., 2005; Chanoca et al., 2015; Dima et al., 2015; Le Roy et al., 2016). To evaluate the vacuole as the subcellular accumulation site of glycosylated volatile phenylpropanoids, we took an indirect approach because methods for determining intracellular localization of small molecules are not yet widely available (Hicks and Raikhel, 2009). We transiently expressed vacuole- or cytosol-targeted versions of β-glucosidase from A. niger (pBINPLUS-35S:AnBGL1vac; Wei et al., 2004) in 4 cm petunia buds to test whether it would lead to a reduction in glycosylated scent compounds accumulating in petals. Green fluorescent protein (GFP) was expressed with or without β-glucosidase to facilitate the identification of inoculated tissues (Figure 5A). Transient expression of AnBGL1vac led to a ca. 2.4-fold increase in β-glucosidase activity compared to control flowers expressing GFP alone, as measured by p-nitrophenyl glucopyranoside (pNPG) substrate assay of petal extracts 3 days after inoculation (Figure 5B). GC–MS analyses of the 35S:AnBGL1vac-expressing flowers revealed the accumulation of significantly less glycosylated scent compounds compared to control flowers (Figure 5C) with a concomitant increase in levels of the corresponding aglycones (Figure S6). It should be noted that isoeugenol and eugenol aglycones were not detected in either treatment, probably due to their low levels at this time of day. This trend was evident with each individual compound, namely benzyl alcohol, phenyethyl alcohol, vanillin, eugenol, and isoeugenol (Figures 5D–H). Inoculation of petunia flowers with 35S:AnBGL1 targeted to the cytosol (pBINPLUS-35S:AnBGL1cyt) did not affect the levels of glycosylated scent compounds (Figure 5I).

To further corroborate the results implicating the vacuole as the storage site for glycosides of phenylpropanoid volatiles, we isolated intact vacuoles for GC–MS analysis. About 90% of the protoplasts used for the isolation of vacuoles were found to be viable, as evaluated by fluorescein diacetate staining (Figure 6B). Intact pure vacuoles (based on light microscopy observation, Figure 6A, inset and Figure 6C) were isolated from these protoplasts with an efficiency of 10–15% based on protoplast vs. vacuole numbers or their anthocyanin contents. Purity of the vacuole isolation was further confirmed by western blot analysis with antibodies against the epsilon subunit of tonoplast V-type H⁺ATPase (V-ATPase, tonoplast), cytosolic fructose-1,6 bisphosphatase (cFBPase, cytosol), phenylalanine ammonia lyase (PAL, endomembrane/cytosol), and major light-harvesting chlorophyll a/b protein (LHCII, plastid; Vainstein and Sharon, 1993; Howles et al., 1996; Dima et al., 2015; de Michele et al., 2016). V-ATPase signal was detected in protein extracts of the vacuoles, whereas there was no signal from antibodies against markers of other cellular compartments. Proteins extracted from protoplasts cross-reacted, as expected, with all of the aforementioned antibodies (Figure 6D). GC–MS analysis of the vacuole fraction revealed pools of benzyl alcohol and phenylethyl alcohol glycosides. These were the dominant metabolites in the pool of glycosides of phenylpropanoid floral scent compounds (amounting to ca. 80%, Figure 2), and were the only ones that could also be detected in the protoplast fraction. Most of protoplasts' benzyl alcohol and phenylethyl alcohol glycosides were detected in vacuoles. Normalized to anthocyanin levels, 86.7% of the former and 67.9% of the latter were found in vacuoles (Figure 6E).

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.