Harvesting at the Right Time: Maturity and its Effects on the Aromatic Characteristics of Cabernet Sauvignon Wine

The aim of this paper was to investigate how maturity affects the aroma characteristics of Cabernet Sauvignon wine. A series of four Vitis vinifera cv. Cabernet Sauvignon wines were produced from grapes of different harvest dates. The berries of sequential harvest treatments showed an increase in total soluble solids and anthocyanin and a decrease in titratable acidity. Berry shriveling was observed as berry weight decreased. In the wines, anthocyanin, dry extract, alcoholic strength, and pH were enhanced with the sequential harvest, whereas polyphenol and tannin were decreased. The concentrations of volatile compounds in sequential harvests were found to be at higher levels. Isopentanol, phenylethyl alcohol, ethyl acetate, ethyl lactate, benzaldehyde, citronellol, and linalool significantly increased when harvest was delayed by one or two weeks. Through a principal component analysis, the volatile compounds and phenols characterizing each harvest date were clearly differentiated. These results suggest that sequential harvest may be an optional strategy for winemakers to produce high-quality wine.


Introduction
Aroma is one of the most important organoleptic characteristics for consumers and is a key attribute shaping wine styles. It is essential in the highly competitive market and the food industry [1,2]. Unique combinations of volatiles and the differences in their concentrations provide aromatically different and characteristic wines [3]. Among the several hundred volatile compounds identified in wines, the most important families are alcohols, esters, aldehydes, ketones, acids, terpenoids, norisoprenoids, pyrazines, and thiols [4].
The volatile profile of a wine is related not only to the fermentation process [5] but to the maturity of the grape as well. Some studies have focused on the evolution of volatile compounds in wines at different grape maturity levels [2,[6][7][8]. Cordonnier and Bayonne emphasized the importance of harvesting during the correct stage; harvesting too early could result in a pronounced unpleasant grassy character, whereas harvesting too late could result in a loss of aroma [9]. The harvest date can have a direct influence on the final wine's 3-isobutyl-2-methoxypyrazine (IBMP) concentrations; IBMP concentrations in wine from consecutive harvests significantly decrease [10,11]. Wines produced from Titratable acidity is expressed in g/L of tartaric acid. The total anthocyanin, total polyphenol, and total tannin concentrations are expressed in mg/g cyanidin-3-mono-glucoside, mg/g gallic acid equivalence, and mg/g (+)-catechin per gram of dry berry skin, respectively. The four harvest dates were September 25 (CK) and October 3 (T1), October 10 (T2), and October 17 (T3), 2016, with CK being the normal harvest date at the vineyard.
These changes may be attributed to water removal from the grape with sequential harvest. Increased transpiration and decreased phloem can result in berry shriveling concomitant with increased TSS concentrations [22][23][24]. In this study, the grape juice TSS increased and titratable acidity decreased during the sequential harvest period in accordance with other reports about the tendency of glycolic and acid during the grape ripening process [2,25]. Anthocyanin accumulations in the grape berry are correlated with increased sugar accumulation, and increased grape anthocyanins can affect wine color and aging capacity [14].

Enological Parameters
The enological parameters are shown in Table 2. Wine titratable acidity decreased during the sequential harvesting and was associated with an increase in pH. The pH did not exceed 4.1 by the final sampling date. The alcoholic strength increased from 11.86% vol (CK) to 12.46% vol (T2, harvest date was October 10), and the dry extract increased from 27.41 g/L to 29.49 g/L throughout the sequential harvesting period. There were no clear trends in the reducing sugar and volatile acidity of the wines with grape maturity. The results are expressed as the mean values ± standard deviation of the triplicate samples. Reducing sugar is expressed in g/L glucose, titratable acidity in g/L tartaric acid, and volatile acidity in g/L acetic acid. Total anthocyanins, total polyphenols and total tannins are expressed in mg/L cyanidin-3-mono-glucoside, mg/L gallic acid, and mg/L ( + )-catechin, respectively. An increase in the anthocyanin level was observed in the sequential harvesting treatments. The change in wine tannin concentrations showed a close relationship with the trend of the grape tannin. Work by Ristic demonstrated a strong relationship between the grape skin tannin concentration and wine tannin concentrations [26]. Moreover, the negative correlation between the change in the anthocyanins and tannins may be attributed to the capacity of anthocyanins to bind tannin under the vinification condition [27]. Interestingly, the change in wine polyphenols was attributed to the grape polyphenols; the T1 treatment had the highest polyphenols, and the wine and grape polyphenol concentrations decreased with harvesting period prolongation. Lower anthocyanin, polyphenol, and tannin levels as a result of berry shriveling during sequential harvesting were previously reported in Shiraz wines [28].

Qualitative and Quantitative Analyses of Volatile Compounds
A total of 47 free volatile compounds was identified and quantified in all wines, including 15 higher alcohols, 16 esters, 6 fatty acids, 6 aldehydes and ketones, and 4 terpenes and norisoprenoids. Tables 3 and 4 show the quantitative results and odor activity values (OAVs) of these compounds. The odor descriptors and thresholds were obtained from the literature [29][30][31][32][33][34]. The most abundant volatile compounds were higher alcohols and esters, especially the higher alcohols, which made up 86%-89% of the aroma concentrations. The trace compounds were terpenes and norisoprenoids.  Note: a Aroma series 1 = fruity, 2 = floral, 3 = herbaceous (or vegetal), 4 = nutty, 5 = caramel, 6 = earthy, 7 = chemical, 8 = fatty, 9 = roasted. b The volatile compound concentrations were less than the limit of quantification (LQ). Different letters in the same row means significant differences according to Duncan test (p < 0.05).

Higher Alcohols
Fifteen higher alcohols were detected in this study. As shown in Table 3, higher alcohols were the largest group in terms of the aromatic compound concentrations identified in all the wine samples. They accounted for > 86% of the total volatile compounds; however, their concentrations were much lower than their thresholds, with only 1-octen-3-ol, isopentanol, and phenylethyl alcohol above the threshold (Table 4). Sequential harvesting treatments presented significantly more higher alcohols, in line with the result of a previous study [7]. Higher alcohols can contribute to a positive effect on wine aroma when they are present at less than 400 mg/L [30]. The higher alcohol concentrations in all the wine samples in this study were above 500 mg/L, which might explain the lack of desirable complexity in the aroma of wines from this wine region.
Sequential harvesting treatments presented significantly higher alcohol levels due to the presence of isopentanol and phenylethyl alcohol. As their concentrations exceeded the threshold, the floral, chemical aromas from the wines were influenced. The compound 1-octen-3-ol is a well-known compound associated with a fresh mushroom odor in grapes and wines [37] and showed no notable fluctuation among all wines. Other higher alcohols, such as 4-methyl-1-pentanol, 3-methyl-1-pentanol, 2-heptanol, 1-octanol, 1-nonanol, 2-nonanol, and 1-decanol, were far lower than their threshold levels. In addition, sequential harvesting treatment can enhance the 1-heptanol concentration, which can impart a fruity aroma to the wines.

Esters
Esters, as the most important odorants in wines, impart abundant floral and tropical fruity aromas [38]. Sixteen esters were detected in the wine samples, including 8 ethyl esters, 4 acetates esters, and 4 other esters ( Table 3). The T1 and T2 treatments significantly enhanced the ester concentration.
Ethyl esters are important esters in wines. Most ethyl esters have quite low thresholds (Table 3). In this study, ethyl lactate and ethyl hexanoate were the main ethyl esters in terms of concentration, and they can impart a pleasant fruity aroma. Ethyl hexanoate and ethyl octanoate might contribute to the wine aroma directly due to their relatively high OAVs (Table 4). Compared with the CK wines, the ethyl hexanoate, ethyl heptanoate, ethyl lactate, ethyl laurate, and ethyl phenylacetate levels were significantly enhanced under the T2 treatment.
Ethyl acetate and isoamyl acetate were the main acetate esters, and the concentrations all exceeded their thresholds. The T1 and T2 treatments increased the two acetate ester concentrations. Ethyl acetate was the most abundant ester in this fraction, generating ethereal fruity aromas in the wines. A continuous increase in ethyl acetate concentration in Cabernet Sauvignon wines with grapes maturity was observed by Bindon [7]. Unlike those of ethyl acetate and isoamyl acetate, the concentrations of hexyl acetate and 2-phenethyl acetate were below their thresholds. The T2 treatment increased the 2-phenethyl acetate concentration. Notably, the isoamyl acetate and 2-phenethyl acetate concentrations in the T3 wine were inferior to those of the other treatments; isoamyl acetate and 2-phenethyl acetate can contribute to the pleasant fruity and floral aromas of the wine. This effect maybe be associated with berry over-ripening.
In addition, four other esters, methyl octanoate, methyl salicylate, butyl butanoate, and isoamyl hexanoate, were also detected in all wines ( Table 3). The thresholds of methyl salicylate and butyl butanoate were unavailable in the literature, so their real impact on the wine aroma is unknown. Methyl salicylate and isoamyl hexanoate were present at levels below their thresholds (Table 3).

Fatty Acids
Six fatty acids were detected in the wine samples ( Table 3). The fatty acid concentrations increased with sequential harvesting. T2 and T3 wines presented high concentration of all fatty acids (> 8mg/L). Butanoic acid, hexanoic acid, octanoic acid, heptanoic acid n-decanoic acid, and 2-methyl-propanoic acid all belong to the C 6 -C 10 fatty acids, which are important in aromatic compound balance. C 6 -C 10 fatty acids are related to negative flavors, an unpleasant fatty odor, and even a rancid smell in wine when present at higher concentrations (>20 mg/L); however, they provide the smell of cheese and cream at concentrations of 4 to 10 mg/L [39]. As seen in Table 3, all wines had an appropriate fatty acid content, less than 10 mg/L, which can contribute a pleasant fatty smell. The 2-Methyl-propanoic acid was markedly the most abundant fatty acid, which had an OAV above 1 ( Table 4). The concentration of butanoic acid increased during sequential harvesting, being present at a higher level, but its threshold was unavailable in the literature. Among the other fatty acids detected in our work, only octanoic acid was slightly above its threshold (Table 4).

Aldehydes and Ketones
Five aldehydes (hexanal, nonanal, decanal, benzaldehyde, and benzeneacetaldehyde) and one ketone (acetoin) were identified in this study (Table 3). They can be reduced to the corresponding alcohols during the fermentation process. Compared to in the CK, the total aldehyde and ketone contents in sequential harvest wines increased. Among this group of compounds, hexanal was the only compound with a concentration exceeding its threshold, and an increase of hexanal was observed in the T2 wines. Benzaldehyde showed the highest fraction; however, the perception threshold of benzaldehydes was very high compared to its concentration. The only ketone in the wine samples was acetoin, and no notable difference was found during sequential harvesting.

Terpenes and Norisoprenoids
Terpenes and norisoprenoids are generally associated with floral, sweet fruit and citric aromas. Norisoprenoids are trace compounds in wine, while their olfactory threshold is very low-between 0.05 and 0.09 µg/L. They usually have an odor activity [38]. Three terpenes and one norisoprenoid were detected in the wine samples, including citronellol, linalool, geraniol, and β-damascenone, with relatively low thresholds. Although wines have a low content of these compounds, they may contribute to the wine aroma directly ( Table 3). The citronellol and linalool concentrations tended to increase, whereas the geraniol concentration did not significantly change with sequential harvesting. As seen in Table 3, citronellol levels significantly increased under sequential harvesting treatments. Linalool and β-damascenone were the only two of these compounds with OAVs above 1 (Table 4). Due to having the highest OAV, β-damascenone had a significant contribution to the wine aroma. A significant decrease was observed in β-damascenone during sequential harvesting.

Odor Activity Values (OAVs) and Aroma Profiles
According to the report by Cai, the aromatic compounds were grouped into nine aroma series based on similar odor descriptors (Table 3) [40]. The total OAVs ( OAV) of each series were calculated (Figure 1). The analysis of the aroma series indicated that the main aroma profiles of Cabernet Sauvignon wines in this study were fruity and floral aromas ( OAV > 160). The earth and herbaceous series had a relatively low contribution to the overall wine aroma ( OAV < 5).
The fruity and floral series were the major aroma series. Further, β-damascenone, ethyl hexanoate, isoamyl acetate, ethyl octanoate, ethyl acetate, and phenylethyl alcohol were the main contributors to the fruity and floral series, as the OAVs of these compounds all exceeded 1 (Table 4). T2 wines presented the fruitiest aromas, while T1 and T3 wines had relatively little fruity and floral aromas due to the concentrations of β-damascenone. For the sequential harvesting treatments, an increase was found in other aroma series (caramel, chemical, and fatty) compared with CK. The caramel, chemical, and fatty series were dominated by ispentanol. In addition, the herbaceous series was mainly based on the C 6 compounds, including 1-hexanol, (E)-3-hexen-1-ol, (Z)-3-hexen-1-ol, and hexanal. In this study, only hexanal exceeded its threshold and was the main contributor to this series.
In addition, some volatile compounds might be present at sub-threshold concentrations; their potential contribution to the wine aroma because of additive effects should not be excluded.

Principal Component Analysis (PCA)
To interpret the influence of sequential harvesting on the wine compound profiles, the effective data were also processed using principal component analysis (PCA). Figure 2 shows the compound loadings ( Figure 2B) and the wine sample distributions (Figure 2A) across the first two principle components (PCs). The first two PCs explained 64.55% of the total variance, comprising 45.37% from PC1 and 19.18% from PC2, and it was clear that PC1 was the major discriminator in explaining the variance among the wine samples. Most of the volatile compounds, anthocyanin, and tannin contribute to the PC1 loading, which suggests that these compounds are more easily influenced by sequential harvesting. As shown in Figure 2A, CK and T1 present negative scores in PC1, and T2 and T3 lie in the first and fourth quadrants, respectively, which showed that there were major differences between CK and T2 or T3 in volatile compounds related to PC1. Only a few volatile compounds and polyphenol were observed to make contributions to the PC2 loading, and CK and T1 showed significant differences in compounds related to PC2. According to the PCA, the volatile compounds and phenols characterizing each harvest date were clearly differentiated. It could be deduced that sequential harvesting can enhance the volatile compound and anthocyanin concentrations in wine. In addition, when combined with the enological parameters, T2 was superior to T3 in avoiding excessive pH and reducing sugars in the wines.

Principal Component Analysis (PCA)
To interpret the influence of sequential harvesting on the wine compound profiles, the effective data were also processed using principal component analysis (PCA). Figure 2 shows the compound loadings ( Figure 2B) and the wine sample distributions (Figure 2A) across the first two principle components (PCs). The first two PCs explained 64.55% of the total variance, comprising 45.37% from PC1 and 19.18% from PC2, and it was clear that PC1 was the major discriminator in explaining the variance among the wine samples. Most of the volatile compounds, anthocyanin, and tannin contribute to the PC1 loading, which suggests that these compounds are more easily influenced by sequential harvesting. As shown in Figure 2A, CK and T1 present negative scores in PC1, and T2 and T3 lie in the first and fourth quadrants, respectively, which showed that there were major differences between CK and T2 or T3 in volatile compounds related to PC1. Only a few volatile compounds and polyphenol were observed to make contributions to the PC2 loading, and CK and T1 showed significant differences in compounds related to PC2. According to the PCA, the volatile compounds and phenols characterizing each harvest date were clearly differentiated. It could be deduced that sequential harvesting can enhance the volatile compound and anthocyanin concentrations in wine. In addition, when combined with the enological parameters, T2 was superior to T3 in avoiding excessive pH and reducing sugars in the wines.  Table 3. According to the results of the fruit maturity monitoring to determine the normal maturity of Cabernet Sauvignon, a first sampling was conducted, followed by a sampling every 7 days for a total of 4 times. The four harvest dates were September 25 (CK) and October 3 (T1), October 10 (T2), and October 17 (T3), 2016, with CK being the normal harvest date at the vineyard. To obtain representative samples, the experimental region was divided into four regions randomly arranged as a sample block. Each region was divided into three blocks, and 120 vines were planted in each block.

Sample Collection and Analysis of the General Index
To obtain a representative sample, for each sampling, 500 berries were randomly selected from each sample block. Grape berries were manually collected from both the inside and outside of the vine canopies included in the experiment. All samples were placed in a foam box with ice and immediately placed in a −40 °C cryogenic refrigerator after transport to the laboratory.
A total of 100 berries were randomly collected and measured for berry weight and berry length. Then, these berries were manually crushed to obtain must, and the supernatant of the must was used to measure the total soluble solids (TSS) and titratable acidity. The TSS were measured using a TD-45 digital refractometer (TOP, Zhejiang, China), and the titratable acidity was determined using sodium hydroxide titration with 0.05 M NaOH to pH 8.2. Phenolic extraction was performed using 200 berry skins. The skins were ground under liquid nitrogen protection, then the powder of the skins was lyophilized using an FD5-series Vacuum Freeze Drying Plant (GOLD-SIM, USA). Polyphenols, tannins, and anthocyanins were extracted and measured as described in a previous article [41].

Small-Scale Wine Making
For each replicate, approximately 80 kg of grapes was manually harvested in each harvesting period. Wines were produced by utilizing the same vinification process as in previous articles [29,42].  Table 3. According to the results of the fruit maturity monitoring to determine the normal maturity of Cabernet Sauvignon, a first sampling was conducted, followed by a sampling every 7 days for a total of 4 times. The four harvest dates were September 25 (CK) and October 3 (T1), October 10 (T2), and October 17 (T3), 2016, with CK being the normal harvest date at the vineyard. To obtain representative samples, the experimental region was divided into four regions randomly arranged as a sample block. Each region was divided into three blocks, and 120 vines were planted in each block.

Sample Collection and Analysis of the General Index
To obtain a representative sample, for each sampling, 500 berries were randomly selected from each sample block. Grape berries were manually collected from both the inside and outside of the vine canopies included in the experiment. All samples were placed in a foam box with ice and immediately placed in a −40 • C cryogenic refrigerator after transport to the laboratory.
A total of 100 berries were randomly collected and measured for berry weight and berry length. Then, these berries were manually crushed to obtain must, and the supernatant of the must was used to measure the total soluble solids (TSS) and titratable acidity. The TSS were measured using a TD-45 digital refractometer (TOP, Zhejiang, China), and the titratable acidity was determined using sodium hydroxide titration with 0.05 M NaOH to pH 8.2. Phenolic extraction was performed using 200 berry skins. The skins were ground under liquid nitrogen protection, then the powder of the skins was lyophilized using an FD5-series Vacuum Freeze Drying Plant (GOLD-SIM, USA). Polyphenols, tannins, and anthocyanins were extracted and measured as described in a previous article [41].

Small-Scale Wine Making
For each replicate, approximately 80 kg of grapes was manually harvested in each harvesting period. Wines were produced by utilizing the same vinification process as in previous articles [29,42]. Briefly, three 80 kg replicates of grapes harvested at every period were directly crushed. The fermentation was conducted in stainless steel fermenters (50 L), in which 80 mL of 6% sulfurous acid was added to achieve 60 mg/L SO 2 . The volume of must was 80% of the fermenter volumes. A total of 1.6 g pectinase (Lallzyme Ex) was added to achieve 20 mg/L and mixed by hand. After maceration of the musts for 12 h, 200 mg/L of dried active yeast (Saccharomyces cerevisiae strain RC212, Lavlin, France) with 5% sugar water activated at 37 • C for 30 min was added to the musts according to commercial specifications. The temperature and specific gravity were monitored daily after the fermentation process was started. Alcoholic fermentation was conducted at 25-27 • C. The wines were separated from the pomace when the specific gravity decreased to 1.000 and fermentation (specific gravity 0.992-0.996) was continued to dryness (reducing sugars <2 g/L). At the end of alcoholic fermentation, 60 mg/L sulfurous acid was added to the wine to maintain 50 mg/L SO 2 . Finally, finished wines did not undergo malolactic fermentation and were bottled for analysis after two months. Enological parameters, such as alcoholic strength, reducing sugars, titratable acidity, and volatile acidity, were analyzed; for the procedures used in each index, we referred to the International Organization of Vine and Wine (OIV) standard [43].

Headspace Solid-Phase Microextraction (HS-SPME)
Volatile compounds in all the wine samples were extracted using headspace solid-phase microextraction (HS-SPME). A 5 mL wine sample and 1 g of NaCl were placed in a 15 mL sample vial, which contained a magnetic stirrer (1 cm) and 10 µL internal standard 4-methyl-2-pentylalcohol (1.0018 g/L, Sigma-Aldrich, Milwaukee, WI, USA). The vial was tightly capped with a polytetrafluoroethylene (PTFE) -silicon septum, heated at 40 • C for 30 min on a heating platform, and agitated at 400 rpm. The solid-phase microextraction (SPME) (50/30-µm DVB/Carboxen/PDMS, Supelco, Bellefonte, PA, USA), preconditioned according to the manufacturer's instructions, was then inserted into the headspace, where extraction was allowed to occur for 30 min with continued heating and agitation via a magnetic stirrer. The volatiles from the fiber were subsequently desorbed by injecting the fiber into the gas chromatography (GC) injector for 8 min [44].

GC-MS Analysis
Gas chromatographic analyses were performed with an Agilent gas chromatograph model 7890 equipped using an Agilent 5975 mass spectrometer and 7683 automatic sampler (Agilent, Santa Clara, CA, USA). Samples were separated on an HP-INNOWAX capillary column (60 m × 0.25 mm × 0.25 µm, J &W Scientific, Folsom, CA, USA). The carrier gas was helium (purity > 99.999%) at 1 mL/min. The temperature in the injection port was 250 • C. Samples were injected by placing the SPME fiber at the GC inlet for 8 min in the splitless mode. The oven temperature program was as follows: 50 • C for 1 min, then increased to 220 • C at a rate of 3 • C/min and held at 220 • C for 5 min. The mass detector conditions were as follows: electron impact mode (MS/EI) at 70 eV, mass scanning range m/z 20 to 350 U, ionic source temperature 230 • C. The mass spectrometry interface temperature was 280 • C. The mass spectrophotometer was operated in the selective ion mode under autotune conditions, and the area of each peak was determined using the ChemStation software F.01.01.2317 (Agilent Technologies, Inc. Santa Clara, CA, USA) [44].

Volatile Compound Identification and Quantification
A synthetic wine matrix was prepared in distilled water containing 13% ethanol (v/v), 2 g/L glucose, and 5 g/L citric acid. The pH was adjusted to 3.8 using a 5 M NaOH solution. All aromatic compound standards with purity greater than 99% were purchased from Sigma-Aldrich (Milwaukee, WI, USA). The volatile compounds stock solution was prepared and dissolved in the synthetic wine matrix. The standard solution was successively diluted to fifteen levels, and 10 µL internal standard (4-methyl-2-pentylalcohol, 1.0018 g/L) was added. Afterwards, the known concentrations of the standard volatile compounds were extracted and analyzed under the same conditions as the wine samples.
The volatile compound identification and quantification methods were based on a previous work [44]. Volatile compounds were identified by a comparison of Kováts' retention indices based on the even n-alkanes (C7-C24) (Supelco, Bellefonte, PA, USA) of the reference standard, and mass spectra matching to the standard National Institute of Standards and Technology Library (NIST11) ( [45]. Comparison of retention indices to those reported in the literature was used without available external standards. For quantification, all the calibration curves had regression coefficients greater than 95%; the detailed quantification information is listed in Table 5. The volatile compound concentrations were calculated from the quantitative ion peak areas with regard to the internal standard. The compounds without established calibration curves were quantified according to the standards with the same functional group or similar numbers of carbon atoms. Table 5. Calibration parameters (Chemical Abstracts Service Number (CASN), Retention Indices (RI), Identification (ID), Manufactures, Purity, Internal standards, Quantitative ion), calibration curves' linear correlation coefficients (R 2 ), and range for the quantitative analysis of volatile aroma compounds in wine using gas chromatography-mass spectrometry (GC-MS)-solid-phase microextraction (SPME).  Note: a The retention index (RI) was calculated on the HP-INNOWAX capillary column. b In identification of the compounds, "A" means those identified by mass spectrum and RI agree with standards, "B" means those tentatively identified by mass spectrum agree with the mass spectral database and RI agrees with literature.

Odor Activity Values (OAVs) and Aroma Series
The odor activity value (OAV) is commonly used to evaluate the contribution of a volatile compound to a wine's characteristic aroma [46,47]. The OAVs were calculated using the equation OAV = c/t, where c is the concentration (in µg/L) of each compound in the wine sample and t is the odor threshold value (in µg/L) of the compound in water/ethanol solution [46]. The perception threshold was found in the literature (Table 3).
To simulate the aroma profile according to the wine volatile composition, the volatile compounds were grouped based on similar odor descriptors. Then, the sum of the OAVs ( OAV) was calculated, which simulated the wine aroma profile. In this study, the volatile compounds were grouped into nine aroma series, namely, fruity, floral, herbaceous, nutty, caramel, earthy, chemical, fatty, and roasted. The aroma series division followed one performed in a previous study [40]. Due to the complexity of aroma characteristics, some volatile compounds may be included in several aroma series.

Statistical Analysis
Statistical data processing was performed using the software Statistical Product and Service Solutions (SPSS 20.0) for Windows (IBM, Armonk, NY, USA). Statistical analyses of the data were performed using one-way analysis of variance (ANOVA) and Duncan's test at the p < 0.05 level. To obtain an overview of the different wine samples, the data of volatile compounds and phenols were subjected to principle component analysis (PCA) to visualize all information in the data set. All plots were prepared using Origin 2016 (OriginLab Corporation, Northampton, MA, USA).

Conclusions
This work investigated the effect of sequential harvesting on the volatile profiles of Cabernet Sauvignon wines. We found that sequential harvesting treatments enhanced the TSS and anthocyanin levels in the berries, while the titratable acid levels decreased. Berry shriveling was observed with over-ripening and resulted in decreased berry weight. Meanwhile, enological parameters and wine volatile compounds were influenced by the sequential harvesting treatments. Volatile profiles, anthocyanins, dry extracts, alcoholic strength, and pH increased under sequential harvesting. The volatile profiles in the sequential harvesting treatments were found to be more abundant, especially in the T1 and T2 wines; isopentanol, phenylethyl alcohol, ethyl acetate, ethyl lactate, benzaldehyde, citronellol, and linalool all showed higher levels. The fruity aroma was enhanced, while the level of β-damascenone was lessened in the T2 wines. The PCA indicated that sequential harvesting could enhance the concentrations of volatile compounds in wine. In conclusion, sequential harvesting is an optional strategy for winemakers to avoid the restrictions of winery equipment when berries reach maturity in large areas growing a single wine grape cultivar.