Differences in the Aroma Profiles of Seedless-treated and Nontreated ‘Shine Muscat’ Grape Berries Decrease with Ripening

Abstract

In the production of seedless table grapes, it is conventional to use plant growth regulators including gibberellins. Little is known about the differences in aroma volatiles between seedless-treated grapes and nontreated (seeded) grapes. Therefore, in this study, the aroma volatile profiles of seedless-treated and nontreated ‘Shine Muscat’ grape berries during ripening were compared using gas chromatography-mass spectrometry. Measurements of volatiles during ripening showed 202 peaks in the seedless-treated and nontreated whole grape berries. According to two-way analysis of variance, the number of volatiles with differences between seedless-treated and nontreated berries and/or between ripening stages was 123, whereas those with no differences between treatments and between ripening stages was 79. Two-way hierarchical clustering analysis for the 123 volatiles showed that seedless-treated berries at the early ripening stages were separated from the other berries, and the seedless-treated and nontreated berries at the post-ripening stage were classified into the same cluster. At the early ripening stage, more lipoxygenase-pathway volatiles were produced in the seedless-treated berries than in the non-treated ones. Linalool compounds increased in both seedless-treated and nontreated berries with ripening. Gene expression profile comparisons using principal component analysis of RNA-sequencing data showed that the seedless-treated berries ripened earlier than the nontreated berries at the early ripening stage. The number of differentially expressed genes in the seedless-treated berries decreased during ripening. Using weighted gene co-expression network analyses, 12 modules and 24 modules were detected in berry skin and flesh, respectively. The correlation analysis revealed that 33 volatiles correlated with four modules in the skin and 50 volatiles correlated with nine modules in the flesh. Most of the volatiles correlated with these modules were those that showed significant differences between treatments and/or ripening stages by two-way analysis of variance. The differences in the aroma volatile profiles between seedless-treated and nontreated berries decreased as harvest was delayed, suggesting that delaying the harvest may make it possible to bring the aroma of seedless-treated ‘Shine Muscat’ berries closer to the original aroma of the seeded berries.

Introduction

Grapes (Vitis L.) are the fourth most produced fruit crop worldwide (https://www.fao.org/faostat/en/#data/QCL, March 28, 2024). Grape seedlessness is an important trait for table use and processing. Seedless table grapes are preferred by consumers owing to their ease of consumption. Seedless grape cultivars with parthenocarpies or stenospermocarpies have been used since the earliest times (Costantini et al., 2021). Recently, plant growth regulators (PGRs) have been used to produce seedless grapes. Originally seeded grape cultivars were rendered seedless by gibberellic acid (GA) treatment. Streptomycin (SM) treatment prior to GA treatment improves the seedlessness efficiency (Pommer et al., 1996). Because seedless grape berries are small, additional GA and cytokinin (CTK) treatments are applied to GA-treated grapes to increase berry size (Retamales et al., 1994; Reynolds et al., 1992). However, the GA and CTK treatments were found to affect fruit traits other than seedlessness and berry size. These traits included an increase in tannin content and astringency of grape berries (Maoz et al., 2014). Tyagi et al. (2021) reported that CTK treatment affected the volatile profiles and phenylpropanoid pathway, resulting in reduced anthocyanin synthesis in berry skin during grape berry ripening.

Aroma is a trait of significant interest in grapes. Approximately 500 volatile compounds have been identified in grapes and wine (Schreier, 1979). Lin et al. (2019) classified grape aromas into six different categories along the biosynthetic pathway: terpenoids, methoxypyrazines, furan derivatives, lipoxygenase (LOX)-pathway products, volatile sulfur compounds, and phenylpropanoids. Terpenoids are further classified into monoterpenes, sesquiterpenes, and norisoprenoids, which are generally perceived as having fruity scents in sensory evaluations. Among the terpenoids, monoterpenes, including geraniol, nerol, and linalool, are especially important to the characteristic aroma of Muscat grapes, such as Vitis vinifera, cv. Muscat Blanc and cv. Muscat of Alexandria (Mateo and Jiménez, 2000; Rapp, 1998).

‘Shine Muscat’ (Vitis labruscana Bailey × V. vinifera L.) bred in the National Agriculture and Food Research Organization of Japan, is a table grape cultivar with yellow-green skinned berries (Yamada et al., 2008), and its production is rapidly on the increase in Asian countries. The grape has an intense muscat aroma (Matsumoto and Ikoma, 2016) and contains a large amount of linalool compared with that in other table grapes (Sasaki et al., 2020). Wu et al. (2020) analyzed the free and bound volatile compounds during ‘Shine Muscat’ grape berry development without PGR treatment and found that free geraniol, nerol and linalool increased during ripening. Choi et al. (2022) reported that only eight compounds, (E)-2-hexenol, hexanol, (Z)-3-hexenal, (E)-2-hexenal, hexanal, (E)-linalool oxide, (E)-damascenone, and linalool, were active odorants in the GA- and CTK-treated ‘Shine Muscat’ berries by measuring odor activity values. Zha et al. (2022) examined the effect of covering vines by colored shade nets on the aroma components of ‘Shine Muscat’ berries and found that covering reduced the production of some aroma volatiles, including linalool and α-terpineol, but increased C6 compound content.

‘Shine Muscat’ grapevines are mainly cultivated as seedless grapes in Japan using GA and CTK treatments, possibly in combination with SM treatment. Primarily, we compared the flavor of seedless-treated and nontreated (seeded) ‘Shine Muscat’ grape berries at the optimal ripening stages using a brief sensory evaluation and noticed a difference in flavor, and particularly aroma. The nontreated grapes were perceived as having a rich aroma, while the seedless-treated grapes were perceived as flat. Because the main purpose of PGR treatments is to make grapes seedless and enlarged, knowledge on the effects of these treatments on aroma profiles is currently very limited. Therefore, in this study, we used gas chromatography-mass spectrometry (GC-MS) to measure and compare the aroma volatile profiles of seedless-treated and nontreated ‘Shine Muscat’ grape berries. In addition, RNA sequencing (RNA-seq) and weighted gene co-expression network analyses (WGCNA) were conducted to identify the specific modules correlated with the volatile profiles in ‘Shine Muscat’ berry skin and flesh during ripening.

Materials and Methods

Plant material

Two ‘Shine Muscat’ (Vitis labruscana Bailey × V. vinifera L.) vines grown at the farm of the University of Tokyo (35°7'N, 139°5'E) under rain shelters throughout the year were used. The vines grafted onto 5BB were six years old in 2020 and were grown under conventional orchard practices. One grapevine was assigned to the seedless-treated group, and the other tree was assigned to the nontreated control group.

Seedless treatment

One week before full bloom (13 May), flower clusters in the seedless treatment trees were treated with SM (200 ppm). On 20 May, the first GA treatment (25 ppm GA₃, Kyowa Hakko Bio, Tokyo, Japan) was applied at full bloom, and the second GA treatment (25 ppm) was administered on June 8. The second GA treatment solution was mixed with CTK (5 ppm forchlorfenuron, Sumitomo Chemical, Tokyo, Japan) for berry enlargement.

Berry sampling

In early July, the berries were thinned to 60 berries per cluster, and grape berries reached the véraison stage in late June. Seedless-treated berry sampling was done on 6 Aug (stage 1, 11 weeks after flowering [WAF]), 13 Aug (stage 2, 12 WAF), 20 Aug (stage 3, 13 WAF), 7 Sep (stage 4, 17 WAF), and 28 Sep (stage 5, 20 WAF). For nontreated berries, sampling was done on 8 Aug (stage 1, 11 WAF), 13 Aug (stage 2, 12 WAF), 24 Aug (stage 3, 13 WAF), 16 Sep (stage 4, 18 WAF), and 28 Sep (stage 5, 20 WAF). In both treatments, stage 4 was the optimal harvesting period. Eighteen clusters in the seedless-treated and nontreated vines were divided into three groups (six clusters per group). Four berries per cluster were collected from each sample. Among the total 24 berries per group, six berries were weighed, and soluble solid content (SSC) was measured on the day of sampling. The SSC was measured using a reflector (PAL-1, ATAGO, Tokyo, Japan), and volatile compounds and RNA were extracted from the remaining 18 berries. After adding 5% (w/w) NaCl to a portion of 18 berries, including skin and flesh, the berries were squeezed tightly with gloved hands to obtain 10–15 mL of juice. The juice, filtered through double gauze, was aliquoted in 1.5 mL microtubes of 0.6 mL each and stored at −30°C until GC-MS analysis. The skin and flesh were collected from the remaining berries and stored at −80°C until RNA extraction.

Analysis of volatile compounds

Volatile organic compound analysis using GC-MS equipped with thermal desorption (TD) was performed according to Tanaka et al. (2018). The silica monolithic trap material (MonoTrap^TM RGPS TD, GL Science Co., Ltd., Tokyo, Japan) was soaked in 0.6 mL of thawed juice and shaken for 30 min at 25°C. The trap material was then rinsed with purified water, and the attached water was carefully removed using a cellulose tissue. An internal standard method was not employed in this study because the juice volume was constant.

A thermal desorption unit (TDU; GERSTEL, Mülheim an der Ruhr, Germany) with programmable temperature vaporization, and a cooled injection system (CIS4; GERSTEL), were used to introduce the trapped volatiles into the analytical column. The TDU was programmed from 40°C (held for 0.5 min) at 720°C·min⁻¹ to 240°C (held for 5 min) with a 50 mL·min⁻¹ desorption flow in a splitless mode. The transfer lines for TDU and CIS4 were set at 300°C. Desorbed volatiles were focused on a Tenax TA packed liner in the CIS4 at −40°C using liquid N₂. For the desorption, the CIS4 was programmed from −40 to 240°C (held for 10 min) to inject the volatiles into the analytical column.

For GC-MS analysis, DB-HeavyWax (length: 60 m, inner diameter: 0.25 mm, film thickness: 0.25 μm; Agilent Technologies Inc., Palo Alto, CA, USA) was used as the analytical column. At the end of the column, an inactivated fused silica tube connected to the column (0.57 m in length, 0.10 mm inside diameter) was connected with capillary flow technology plates and inserted into the mass spectrometer. Solvent vent mode injection at 50-mL vent flow for 2 min was used for the injection. The constant column flow rate of helium was set to 1.1 mL·min⁻¹. The oven temperature was maintained at 40°C for 1 min and then increased to 240°C at a rate of 4°C·min⁻¹. This temperature was maintained for 8 min. The mass spectrometer scanned m/z = 30–400, and the ionization voltage was 70 eV.

MassHunter (ver. B.09.00, Agilent Technologies) was used to analyze the data. Deconvolution followed by peak identification was carried out using MassHunter-Unknowns Analysis software combined with Aroma Office 2D software ver. 7.0 (Gerstel KK, Tokyo, Japan). After deconvolution, a mass spectra search was carried out using the in-house library (built from authentic compounds on a DB-HeavyWax column with the same specifications as in the above analysis) and NIST20 with a match score threshold of 80. The retention index match was examined using the in-house library and Aroma Office 2D database with 10 and 20 allowable limits of difference, respectively. Each optimal quantifier ion (m/z) for integration of peak intensity was manually chosen by considering the mass spectra the near peaks (Table 1). A target list comprising the component retention time and m/z for integration was introduced into the MassHunter Quantitative Analysis software, and each peak was integrated. The resulting intensity of each component itself was used for data analysis.

Table 1

Volatile compounds in ‘Shine Muscat’ grapes.

Table 1

Continued

RNA-seq analysis

Total RNA was manually extracted from grape berry skin and flesh, respectively, at stages 1, 3, and 4 according to the method described by Reid et al. (2006), and was purified using a NucleoSpin RNA Plant kit (MACHEREY-NAGEL, Düren, Germany). For all time points and treatments, RNA was extracted from a pool of three biological replicates. The total RNA was quantified using a Nanodrop Lite Spectrophotometer (Thermo Scientific, Waltham, MA, USA). The quality was determined using a 2100 Bioanalyzer (Agilent Technologies), and the RNA-seq library was prepared using Lasy-Seq v1.1 (https://sites.google.com/view/lasy-seq/, March 28, 2024) (Kamitani et al., 2019). The library was sequenced using the Illumina HiSeqX platform to generate 150-nt long paired-end reads.

RNA-seq data processing

The generated reads were subjected to quality control using the fastp 0.20.0 program (Chen et al., 2018: https://github.com/OpenGene/fastp, March 28, 2024). The cleaning procedure included trimming ambiguous reads and adapters at either the 5' or 3' ends of the reads and filtering reads with default parameters. Cleaned reads were aligned to the grape reference genome assembly (Vitisvinifera.12X.dna_sm.toplevel.fa) from EnsemblPlants (https://ftp.ensemblgenomes.ebi.ac.uk/pub/plants/release-50, March 28, 2024) using Hisat2-2.2.1 (Kim et al., 2019) with default parameters. To obtain expression data, mapped reads were counted for each gene using featureCounts (Liao et al., 2014) with default parameters using Vitis_vinifera.12X.50. gtf from Ensembl Plants as a gene annotation file.

Differential expression analysis

Differential gene expression in the seedless treatment relative to the control was determined at stages 1, 3, and 4 using the DESeq2 program on iDEP.91 (Ge et al., 2018: http://bioinformatics.sdstate.edu/idep90, March 28, 2024). Significantly differentially expressed genes (DEG) were determined at a false discovery rate-adjusted P-value (q-value) < 0.05, and log2 fold change > |1|. The terms up- or down-regulated were used to refer to the expression values of the seedless-treated samples relative to the mock expression values. The datasets generated for this study are available under the NCBI SRA accession (GSE247972 and GSM7903968–GSM7904003).

WGCNA protocol

WGCNA was performed using R software with the WGCNA (1.72.1) package (Langfelder and Horvath, 2008). The expression data for all samples were normalized using the RUVg function with in silico empirical negative controls in the RUVseq (1.28.0) package in R (Risso et al., 2014) with default parameters. DEGs from the skin and flesh samples at each stage of berry development were collected as datasets for network analysis. In network analysis, we used the pickSoftThreshold function to decide β with the options: networkType = “unsigned”, RsquaredCut = 0.8. The parameters for the option to detect the module were as follows: minClusterSize = 30 in the cutreeDynamic function and cutHeight = 0.20 in the mergeCloseModules function. Default parameters were used for the other parameters. To identify the modules that were significantly associated with volatile compounds, we performed a correlation analysis between module eigengenes and volatile compounds. Gene Ontology (GO) enrichment analyses for the modules were conducted using PlantRegMap (Tian et al., 2020) and the Kyoto Encyclopedia of Genes and Genomes Orthology-Based Annotation System (KOBAS) (Bu et al., 2021), respectively. Transcription factor (TF) genes in a module were detected using the PlantTFDB database (Tian et al., 2020). A dendrogram of the TF gene expression levels was generated using R with the ComplexHeatmap (2.10.0) package (Gu et al., 2016; Gu, 2022) using the robust_dist function.

Statistical analysis

In the measurement of fruit weight and SSC, statistical differences between treatments at the respective time points were determined using the Student’s t-test. Statistical differences between seedless treatments and stages were determined using a two-way analysis of variance (ANOVA). A two-way hierarchical clustering analysis was conducted using Ward’s method (JMP13; SAS Institute Japan, Tokyo, Japan). Multivariate analysis was performed after normalizing each volatile and gene expression data.

Results

Change in fruit weight and SSC content

SM treatment was conducted prior to GA treatment to reduce the production of seeded berries in the seedless-treated clusters. However, in the seedless-treated samples, the proportion of berries with seeds larger than 8 mm averaged 1 or 2 out of 24 (data not shown). One seeded berry was included in the seedless-treated samples at stage 1 (1 of 24), but later berries without seeds were chosen as the other seedless-treated samples. All nontreated berries contained seeds. Therefore, the effects of both the seedless treatment with PGRs and the absence of seeds acted on fruit traits.

From stages 1 to 5, the weights of the seedless-treated grape berries were higher than those of the nontreated berries because of the effect of GA and CTK treatments on fruit enlargement (Fig. 1). The average weights at the optimal stage for harvesting (stage 4) were 15.5 g and 10.7 g in the seedless-treated and nontreated berries, respectively. SSC in the seedless-treated berries was higher than that in the nontreated berries until stage 3 (Fig. 2). At stage 4, the SSC in the seedless-treated and nontreated berries was 19.3 and 19.8 °Brix, respectively, and there was no significant difference between the two groups. The harvest criterion for SSC in ‘Shine Muscat’ is 18 °Brix in Japan, so both study groups generally met this criterion. At stage 5, the SSC in the nontreated berries (21.3 °Brix) was higher than that in the seedless-treated berries (20.4 °Brix).

Fig. 1

Changes in fruit weight of seedless-treated and nontreated ‘Shine Muscat’ berries. The fruit weights of seedless-treated (SL) and nontreated (NT) berries are shown in black and grey, respectively. The sampling dates for the seedless-treated and nontreated berries are indicated in the upper and lower rows, respectively. Stages 1–5 are described as S1–S5. Samples at the stage enclosed by squares were used for RNA sequencing analysis. * and ** indicate significant differences of P < 0.05, and P < 0.01, respectively, using a two-way analysis of variance (ANOVA). Data on the weight per berry are shown as boxplots. The bottom and top of the box indicate the first and third quartiles, respectively; the inside line indicates the median; the whiskers are located at 1.5 [interquartile range (IQR)] above and below the box; any data point further than this distance is considered an outlier and is indicated by a dot.

Fig. 2

Changes in soluble solids content (SSC) of seedless-treated and nontreated ‘Shine Muscat’ berries. The SSCs of seedless-treated (SL) and nontreated (NT) berries are shown in black and grey, respectively. The sampling dates for the seedless-treated and nontreated berries are indicated in the upper and lower rows, respectively. Stages 1–5 are described as S1–S5. Samples at the stage enclosed by squares were used for RNA sequencing analysis. NS, *, and ** indicate non-significant or significant differences of P < 0.05, and P < 0.01, respectively, using two-way analysis of variance (ANOVA). The SSC data per berry are shown as boxplots. The bottom and top of the box indicate the first and third quartiles, respectively; the inside line indicates the median; the whiskers are located at 1.5 [interquartile range (IQR)] above and below the box; any data point further than this distance is considered an outlier and is indicated by a dot.

Volatile compound profiles

We analyzed 202 common peaks detected in the seedless-treated and nontreated ‘Shine Muscat’ grape berries using GC-MS (Table 1). Among these, 149 volatile compounds were identified. Additional information on the identification of volatile compounds is provided in Table S1. ‘Shine Muscat’ grape berries contained terpenoids, furan derivatives, lipoxygenase pathway products, and volatile sulfur compounds, but did not contain methoxypyrazines or phenylpropanoids. A two-way ANOVA was conducted to examine the effect of the seedless treatment and sampling stage on volatile production (peak intensities) in the berries (Table S1). The 78 and 96 volatiles showed significant differences according to the treatment and ripening stage, respectively. Among the 202 volatiles, the number of volatiles with differences due to the treatment and/or ripening stage was 123 (60.9%). The number of compounds that showed no difference according to the treatment and stage was 79 (39.1%).

Two-way hierarchical clustering analysis was performed to determine the peak intensities of the 123 volatiles (Fig. 3). In the dendrogram, the volatiles were classified into two clusters, A and B. Cluster A1 (20 volatiles) included compounds with high peak intensities in both the seedless-treated and nontreated berries at stages 4 and/or 5, such as linalool-related compounds, farnesene, and (E)-2-hexenal. Cluster B1 (71 volatiles) contained compounds with high peak intensities in seedless-treated grape berries, mainly at stages 1, 2, and 3. Cluster B1 contained geranial, L-α-terpineol, benzaldehyde and related compounds, and LOX-pathway volatiles, including nonanoic acid and (2E,4E)-nonadienal.

Fig. 3

Two-way hierarchical clustering analysis of significant differences between volatiles in ‘Shine Muscat’ grape berries according to treatment or ripening stage. The peak intensities of the 123 volatiles normalized from −1 to +1 are shown as a heatmap. Mean peak intensities are listed in Table S1. Stages 1–5 are described as S1–S5. The stages of seedless-treated (SL) and nontreated (NT) berries are shown in blue and red, respectively.

The dendrogram of the seedless-treated berries at stages 1, 2, and 3 (cluster D) was clearly different from the others (cluster C). The seedless-treated berries at stage 4 were classified into the same cluster as the nontreated berries at stages 3 and 4 (cluster C2). Seedless-treated and nontreated berries at stage 5 were classified into cluster C3. These results indicated that the volatile profiles of seedless-treated berries at stages 1, 2, and 3 were specific or abnormal. However, as the stages progressed, the difference in the volatile profiles between the seedless-treated and nontreated grape berries decreased.

The changes in the peak intensities of the major volatiles are shown in Figures S1 and S2. The peak intensities of terpenes, including linalool-related compounds, α-terpinene and (E)-2-hexenal, increased toward stage 5 in both seedless-treated and nontreated berries, and there were no significant differences between the treatments (Fig. S1; Table S1). The compounds in Figure S2 (geranial, hexanal, (2E,4Z)-decadienal, 2,6-dimethylocta-3,7-dien-2,6-diol, pentanol, and 2-phenylfuran) differed significantly between the treatments (Table S1). In seedless-treated berries, the peak intensities of these compounds peaked during the early ripening stages and then decreased as ripening progressed.

RNA-seq analysis

RNA-seq of seedless-treated and nontreated ‘Shine Muscat’ berries was conducted at three time points (stages 1, 3, and 4). We chose these three points to detect differences in gene expression during the process of reaching optimum ripening stages. RNA was extracted from the berry skin and flesh samples, respectively, and analyzed. The total number of reads before filtering was 870 M, with an average of 24 M reads per sample. Of these, an average of 3 M reads adhered to the genome. Moreover, iDep91 (Ge et al., 2018) was used for principal component analysis (PCA) and DEG detection.

The results of PCA for the gene expression profiles of seedless-treated and nontreated berries are shown in Figure 4 and Table S2. Principal component (PC)1 (36%) reflected the sampled tissues (berry skin or flesh) and sampling stages to a certain degree, whereas PC2 (19%) reflected the sampling stages (Fig. 4A). PC3 contributed 5% of the total variance and seemed to reflect the effect of the seed treatment (Fig. 4B). In the skin and flesh of the nontreated berries, the plots of stages 1 and 3 were separated, whereas in the skin and flesh of the seedless-treated berries, they were not only close to each other, but also to the plot of the nontreated berries at stage 3 (Fig. 4A, B).

Fig. 4

Principal component analysis (PCA) score plot for RNA-sequencing (RNA-seq) data. RNA was extracted from the skin and flesh of ‘Shine Muscat’ berries at stages 1, 3, and 4 (S1, S3, and S4), and subjected to RNA-seq analysis. The ripening stages of seedless-treated (SL) and nontreated (NT) berries are shown in black and grey, respectively. (A) Score plot of principal component (PC)1 and PC2, and (B) score plot of PC1 and PC3. The arrows indicate that the locations of the S1 and S3 plots are far apart. The PCA loading scores are listed in Table S2.

DEG analysis

DEGs were detected by comparing the seedless-treated and nontreated samples (Fig. S3; Table S3). The number of DEGs in the seedless-treated grape berries (P < 0.05, fold-change > 2) was higher at the early ripening stages (stages 1 and 3) than at the ripening stage (stage 4) in both berry skin and flesh and decreased as ripening progressed. At stage 1, the numbers of downregulated DEGs in the skin and flesh of seedless-treated berries were 903 and 992, respectively, whereas the numbers of upregulated DEGs were 772 and 572, respectively. At stage 4, the numbers of downregulated DEGs in the skin and flesh of seedless-treated berries were 120 and 147, respectively, whereas the numbers of upregulated DEGs were 173 and 205, respectively.

The results of DEG comparisons among the ripening stages are shown in Figure S4. In both the skin and flesh of the seedless-treated berries, the number of DEGs with expression up- or down-regulated in all three stages was small. The numbers of up- or down-regulated DEGs in common with stages 1 and 3 were larger than those common to stages 3 and 4, indicating that the gene expression profiles in early (stages 1 and 3) and optimum (stage 4) stages were more different in berry skin and flesh, respectively.

WGCNA outcomes

WGCNA detected 12 modules and 24 modules in the skin and flesh of ‘Shine Muscat’ berries, respectively (Fig. S5). To identify which of these modules functioned, correlation analysis between the modules and volatile profiles was performed. The results of the analysis showed that 33 volatiles were correlated with 4 modules in the skin (Table 2) and 50 volatiles were correlated with 9 modules in the flesh (Table 3). Most of the volatiles correlated with these modules showed significant differences between treatments and/or ripening stages by two-way ANOVA (Tables 2, 3, and S1). We found two modules, “midnightblue” and “brown” in the berry skin and flesh, respectively. The midnight blue and brown modules were significantly correlated with the accumulation of 22 and 26 volatile compounds (r > 0.9), as listed in Tables 1 (no. 6–27) and 2 (no. 3–29), respectively. Of the 22 compounds that correlated with the midnight blue modules of the skin, 17 were common to the brown modules of the flesh and showed significant differences between treatments.

Table 2

List of volatile compounds with high correlation to the modules detected in the ‘Shine Muscat’ berry skin.

Table 3

List of volatile compounds with high correlation to the modules detected in the ‘Shine Muscat’ berry flesh.

GO analysis of the midnight blue module of berry skin revealed significant enrichment of stress-responsive genes and organelle-related genes (Table S4). KOBAS pathway analysis for the midnight blue module detected protein processing in the endoplasmic reticulum, endocytosis, and plant hormone signal transduction (Table S5). These results indicate that seedless treatment may cause metabolic changes in the cells of seedless-treated berries and affect the metabolism of intracellular organelles. PlantTFDB predicted 32 TF genes in the midnight blue module and the expression patterns of these TF genes were classified by hierarchical clustering analysis (Fig. 5). The dendrogram showed that the 32 genes were grouped into two clusters, A and B. Cluster A contained genes for which expression in the seedless-treated berries from stages 1 to 4 was higher than that in the nontreated berries. In contrast, cluster B contained genes for which expression in the seedless-treated berries from stages 1 to 4 was lower than that in the nontreated berries.

Fig. 5

Dendrogram of transcription factor (TF) gene expression levels in the midnight blue module. The expression levels of the 32 predicted TF genes in the midnight blue module of the ‘Shine Muscat’ berry skin at stages 1, 3, and 4 (S1, S3, and S4) were normalized from −2 to +2. SL and NT denote seedless-treated and nontreated berries, respectively. For each sample, three replicates of expression data are presented. The abbreviation after the gene name is TF, deduced from the motif using PlantTFDB.

Discussion

‘Shine Muscat’ is a table grape cultivar with an intense Muscat flavor (Matsumoto and Ikoma, 2016) that is commercially produced as seedless grapes through seedless treatment with PGRs in Japan. We compared the volatile compounds of seedless-treated and nontreated ‘Shine Muscat’ berries because the aroma was perceived to be different between the seedless-treated and nontreated berries at suitable ripeness stages. A total of 202 volatile compounds were detected in ‘Shine Muscat’ berries using GC-MS (Table 1) and 149 volatiles were identified. Wu et al. (2020) identified 75 volatiles in nontreated ‘Shine Muscat’ berries and Choi et al. (2022) identified 44 volatiles in PGR-treated ‘Shine Muscat’ berries using solid-phase microextraction. In this study, a larger number of components was detected than previously reported using MonoTrap as an absorbent. In these previous reports, the amounts of volatiles in seedless-treated and nontreated ‘Shine Muscat’ berries showed an overall increase towards ripening. In contrast, the current study showed an increase in some volatiles in seedless-treated berries at the early ripening stages (cluster B1 in Fig. 3). This may be because the type of PGRs applied varied with each report.

Ribéreau-Gayon et al. (1975) and Doligez et al. (2006) reported that muscat aroma was characterized by linalool, geraniol, nerol, α-terpineol, and linalool oxides. In this study, geraniol and nerol were not detected in ‘Shine Muscat’ berries, although geranial was detected (Table S1). Matsumoto and Ikoma (2016) focused on six aroma volatiles (linalool, hexanal, (E)-2-hexenal, hexanol, (Z)-3-hexeol, and nerol) in the skin of ‘Shine Muscat’ berries and five aroma volatiles, except nerol, in the flesh. The nerol content was found to be low, even in the skin of ‘Shine Muscat’ berries, compared with that in the other five compounds. The concentration of nerol may have been below the detectable level in our study because whole berries, including the skin and flesh, were used for volatile extraction.

At the early-ripening stages, the seedless-treated ‘Shine Muscat’ berries emitted volatiles that were originally produced only in small amounts (cluster B1 in Fig. 3). These volatiles contained terpenoids, furan derivatives, LOX pathway products, and volatile sulfur compounds, suggesting that seedless treatment with PGRs affected multiple biosynthetic pathways. In particular, LOX pathway products, including nonanoic acid and (2E,4E)-nonadienal, were produced in large amounts in seedless-treated berries at early ripening stages. The production of volatiles in the LOX pathway is induced by biotic and abiotic stressors (Schwab et al., 2008; Singh et al., 2022). The production of such unusual LOX-pathway volatiles may indicate that the berries subjected to PGR treatment were under stress conditions during the early growth period. This result was consistent with the midnight blue module containing stress-related genes based on GO analysis (Table S4). In contrast, some volatiles, such as linalool-related compounds, increased during ripening, regardless of treatment (Fig. S1).

In cluster C2 in Figure 3, the seedless-treated berries at stage 4 were separated from the nontreated berries at stages 3 and 4. This result supported the experiential perception that the difference in flavor of the seedless-treated and nontreated berries at the optimum ripening stage was due to the difference in aroma. The dendrogram of treatments and stages showed that the volatile profiles of the seedless-treated berries at stages 1, 2, and 3 differed from those of the nontreated berries, and the difference between treatments decreased at stage 5 (Fig. 3). The trend of decreasing differences in volatiles between seedless-treated and nontreated grape berries as ripening progressed was similar to that reported by Tyagi et al. (2021) using the seedless grape cv. ‘Sugarsixteen’. These results suggest that delaying harvest may make allow the aroma of seedless-treated ‘Shine Muscat’ berries to be closer to the original aroma of the seeded berries.

The concentration of the aroma volatiles in the ‘Shine Muscat’ berry skin were higher than that in the berry flesh (Matsumoto and Ikoma, 2016), indicating that the regulatory mechanisms of volatile compound production may partially differ between the skin and flesh. The fact that PCA of RNA-seq data showed a strong separation of gene expression according to the sampled tissues with PC1 (Fig. 4) may be related to this result. In the skin and flesh of the seedless-treated berries, the plots of stages 1 and 3 were close to the plot of the nontreated berries at stage 3, which was undergoing a normal growth process (Fig. 4). These gene expression profiles suggest that seedless-treated grapes are precocious because of the application of GA and CTK for fruit enlargement.

WGCNA showed that the volatiles correlated with midnight blue (skin) and brown (flesh) modules were significantly different between treatments and/or between ripening stages (Tables 2, 3, and S1). Seventeen volatiles were common to both modules, with significant differences between treatments, indicating that some volatiles for which profiles were altered by the seedless treatment belonged to specific modules. Cheng et al. (2015) identified the DEGs encoding TFs following GA treatment in ‘Kyoho’ inflorescences and showed that the expressions of some TF genes were rapidly up-regulated 1 h after GA treatment and then down-regulated 24 h later. Nishiyama et al. (2022) found that expression of cell cycle-related genes was down-regulated in the ovules six days after GA treatment of ‘Shine Muscat’ clusters at full bloom. In this study, the seedless treatment had an effect on gene expression for more than two months after treatment (Fig. 5). It is possible that PGR treatment had a long-term effect on gene expression, or that gene expression was altered by the aroma compounds, which were altered by the PGR treatment.

In this study, ‘Shine Muscat’ grapes were treated with SM, in addition to GA and CTK, to produce seedless berries. Therefore, the differences in the volatile compound profiles and gene expression between the seedless-treated and nontreated grape berries were due to the effects of the PGR treatment and the absence of seeds. Tyagi et al (2021) reported that CTK, but not GA, treatment changed the composition of phenylpropanoid compounds using seedless ‘Sugarsixteen’, indicating that only PGRs may cause changes to secondary metabolism. Further research is necessary to clarify whether the differences in the aroma volatile profiles of ‘Shine Muscat’ grape berries are caused by PGR treatment and/or the absence of seeds.

Literature Cited

•  Bu,  D.,  H.  Luo,  P.  Huo,  Z.  Wang,  S.  Zhang,  Z.  He,  Y.  Wu,  L.  Zhao,  J.  Liu,  J.  Guo,  S.  Fang,  W.  Cao,  L.  Yi,  Y.  Zhao and  L.  Kong. 2021. KOBAS-i: intelligent prioritization and exploratory visualization of biological functions for gene enrichment analysis. Nucleic Acids Res. 49: W317–W325.
•  Chen,  S.,  Y.  Zhou,  Y.  Chen and  J.  Gu. 2018. fastp: an ultra-fast all-in-one FASTQ preprocessor. Bioinformatics 34: i884–i890.
•  Cheng,  C.,  C.  Jiao,  S. D.  Singer,  M.  Gao,  X.  Xu,  Y.  Zhou,  Z.  Li,  Z.  Fei,  Y.  Wang and  X.  Wang. 2015. Gibberellin-induced changes in the transcriptome of grapevine (Vitis labrusca × V. vinifera) cv. Kyoho flowers. BMC Genomics 16: 128. DOI: 10.1186/s12864-015-1324-8.
•  Choi,  K. O.,  Y. Y.  Hur,  S. J.  Park,  D. H.  Lee,  S. J.  Kim and  D. J.  Im. 2022. Relationships between instrumental and sensory quality indices of Shine Muscat grapes with different harvesting times. Foods 11: 2482. DOI: 10.3390/foods11162482.
•  Costantini,  L.,  P.  Moreno-Sanz,  C. C.  Nwafor,  S.  Lorenzi,  A.  Marrano,  F.  Cristofolini,  E.  Gottardini,  S.  Raimondi,  P.  Ruffa,  I.  Gribaudo,  A.  Schneider and  M. S.  Grando. 2021. Somatic variants for seed and fruit set in grapevine. BMC Plant Biol. 21: 135. DOI: 10.1186/s12870-021-02865-2.
•  Doligez,  A.,  E.  Audiot,  R.  Baumes and  P.  This. 2006. QTLs for muscat flavor and monoterpenic odorant content in grapevine (Vitis vinifera L.). Mol. Breed. 18: 109–125.
•  Ge,  S. X.,  E. W.  Son and  R.  Yao. 2018. iDEP: an integrated web application for differential expression and pathway analysis of RNA-Seq data. BMC Bioinformatics 19: 534. DOI: 10.1186/s12859-018-2486-6.
•  Gu,  Z. 2022. Complex heatmap visualization. iMeta 1: e43. DOI: 10.1002/imt2.43.
•  Gu,  Z.,  R.  Eils and  M.  Schlesner. 2016. Complex heatmaps reveal patterns and correlations in multidimensional genomic data. Bioinformatics 32: 2847–2849.
•  Kamitani,  M.,  M.  Kashima,  A.  Tezuka and  A. J.  Nagano. 2019. Lasy-Seq: a high-throughput library preparation method for RNA-Seq and its application in the analysis of plant responses to fluctuating temperatures. Sci. Rep. 9: 7091. DOI: 10.1038/s41598-019-43600-0.
•  Kim,  D.,  J. M.  Paggi,  C.  Park,  C.  Bennett and  S. L.  Salzberg. 2019. Graph-based genome alignment and genotyping with HISAT2 and HISAT-genotype. Nat. Biotechnol. 37: 907–915.
•  Langfelder,  P. and  S.  Horvath. 2008. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 9: 559. DOI: 10.1186/1471-2105-9-559.
•  Liao,  Y.,  G. K.  Smyth and  W.  Shi. 2014. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 30: 923–930.
•  Lin,  J.,  M.  Massonnet and  D.  Cantu. 2019. The genetic basis of grape and wine aroma. Hortic. Res. 6: 81. DOI: 10.1038/s41438-019-0163-1.
•  Maoz,  I.,  A.  Bahar,  T.  Kaplunov,  Y.  Zutchi,  A.  Daus,  S.  Lurie and  A.  Lichter. 2014. Effect of the cytokinin forchlorfenuron on tannin content of Thompson Seedless table grapes. Am. J. Enol. Vitic. 65: 230–237.
•  Mateo,  J. J. and  M.  Jiménez. 2000. Monoterpenes in grape juice and wines. J. Chromatogr. A 881: 557–567.
•  Matsumoto,  H. and  Y.  Ikoma. 2016. Effect of postharvest temperature on the muscat flavor and aroma volatile content in the berries of ‘Shine Mascut’ (Vitis Labruscana Bailey × V. vinifera L.). Postharvest Biol. Technol. 112: 256–265.
•  Nishiyama,  S.,  D.  Yoshimura,  A.  Sato and  K.  Yonemori. 2022. Characterization of tissue-specific transcriptomic responses to seedlessness induction by gibberellin in table grape. Hort. J. 19: 157–168.
•  Pommer,  C. V.,  E. J. P.  Pires,  M. M.  Terra and  I. R. S.  Passos. 1996. Streptomycin-induced seedlessness in the grape cultivar Rubi (Italia red). Am. J. Enol. Vitic. 47: 340–342.
•  Rapp,  A. 1998. Volatile flavour of wine: correlation between instrumental analysis and sensory perception. Nahrung 42: 351–363.
•  Reid,  K. E.,  N.  Olsson,  J.  Schlosser,  F.  Peng and  S. T.  Lund. 2006. An optimized grapevine RNA isolation procedure and statistical determination of reference genes for real-time RT-PCR during berry development. BMC Plant Biol. 6: 27. DOI: 10.1186/1471-2229-6-27.
•  Retamales,  J.,  F.  Bangerth,  T.  Copper and  R.  Callejas. 1994. Effects of CPPU and GA₃ on fruit quality of Sultania table grape. Acta Hortic. 394: 149–157.
•  Reynolds,  A. G.,  D. A.  Wardle,  C.  Zurowski and  N. E.  Looney. 1992. Phenylureas CPPU and thidiazuron affect yield components, fruit composition, and storage potential of four seedless grape selections. J. Amer. Soc. Hort. Sci. 117: 85–89.
•  Ribéreau-Gayon,  P.,  J. N.  Boidron and  A.  Terrier. 1975. Aroma of Muscat grape varieties. J. Agric. Food Chem. 23: 1042–1047.
•  Risso,  D.,  J.  Ngai,  T. P.  Speed and  S.  Dudoit. 2014. Normalization of RNA-seq data using factor analysis of control genes or samples. Nat. Biotechnol. 32: 896–902.
•  Sasaki,  T.,  S.  Ando,  T.  Miyazawa,  D.  Yamauchi,  H.  Take,  Y.  Yamazaki and  T.  Enomoto. 2020. Characterisation of ‘Ruby Roman’ table grapes (Vitis Labruscana Bailey) by sensory evaluation and analysis of aroma and taste compounds. Food Sci. Technol. Res. 26: 423–434.
•  Schreier,  P. 1979. Flavor composition of wines: a review. Crit. Rev. Food Sci. Nutr. 12: 59–111.
•  Schwab,  W.,  R.  Davidovich-Rikanati and  E.  Lewinsohn. 2008. Biosynthesis of plant-derived flavor compounds. Plant J. 54: 712–732.
•  Singh,  P.,  Y.  Arif,  E.  Miszczuk,  A.  Bajguz and  S.  Hayat. 2022. Specific roles of lipoxygenases in development and responses to stress in plants. Plants 11: 979. DOI: 10.3390/plants11070979.
•  Tanaka,  F.,  M.  Tatsuki,  K.  Matsubara,  K.  Okazaki,  M.  Yoshimura and  S.  Kasai. 2018. Methyl ester generation associated with flesh browning in ‘Fuji’ apples after long storage under repressed ethylene function. Postharvest Biol. Technol. 145: 53–60.
•  Tian,  F.,  D. C.  Yang,  Y. Q.  Meng,  J. P.  Jin and  G.  Gao. 2020. PlantRegMap: charting functional regulatory maps in plants. Nucleic Acids Res. 48: D1104–D1113.
•  Tyagi,  K.,  I.  Maoz,  B.  Kochanek,  N.  Sela,  L.  Lerno,  S. E.  Ebeler and  A.  Lichter. 2021. Cytokinin but not gibberellin application had major impact on the phenylpropanoid pathway in grape. Hortic. Res. 8: 51. DOI: 10.1038/s41438-021-00488-0.
•  Wu,  Y.,  W.  Zhang,  S.  Song,  W.  Xu,  C.  Zhang,  C.  Ma,  L.  Wang and  S.  Wang. 2020. Evolution of volatile compounds during the development of Muscat grape ‘Shine Muscat’ (Vitis labrusca × V. vinifera). Food Chem. 309: 125778. DOI: 10.1016/j.foodchem.2019.125778.
•  Yamada,  M.,  H.  Yamane,  A.  Sato,  N.  Hirakawa,  H.  Iwanami,  K.  Yoshinaga,  T.  Ozawa,  N.  Mitani,  M.  Shiraishi,  M.  Yoshioka,  I.  Nakajima,  M.  Nakano and  R.  Nakaune. 2008. New grape cultivar ‘Shine Muscat’. Bull. Natl. Inst. Fruit Tree Sci. 7: 21–38 (In Japanese with English abstract).
•  Zha,  Q.,  J.  Wu,  X.  Xi,  Y.  He,  X.  Yin and  A.  Jiang. 2022. Effects of colored shade nets on grapes and leaves of Shine Muscat grown under greenhouse conditions. Am. J. Enol. Vitic. 73: 39–47.