Increase the Content of Ester Compounds in Blueberry Wine Fermentation with the Ester-Producing Yeast: Candida glabrata, Pichia anomala, and Wickerhamomyces anomalus

The co-fermentation of Saccharomyces cerevisiae and ester-producing yeasts is considered to be an effective way to improve the flavor and quality of fruit wine. In this study, three kinds of ester-producing yeasts (Candida glabrata NCUF308.1, Pichia anomala NCUF306.1, and Wickerhamomyces anomalus NCUF307.1) and S. cerevisiae NCUF309.2 were used to simulate blueberry wine co-fermentation at different ratios. The results showed that, compared with S. cerevisiae NCUF309.2 fermentation (S), the population of S. cerevisiae NCUF309.2 in co-fermentation samples decreased to varying degrees, and the content of ethanol also decreased. The results also showed that the co-fermentation of C. glabrata NCUF308.1 and S. cerevisiae NCUF309.2 at the ratio of 1:1 (CS1), co-fermentation of P. anomala NCUF306.1 and S. cerevisiae NCUF309.2 at the ratio of 5:1 (PS5), and co-fermentation of W. anomalus NCUF307.1 and S. cerevisiae NCUF309.2 at the ratio of 5:1 (WS5) could significantly increase the content of ester compounds (p < 0.05), which was 3.29, 4.75, and 3.04 times that of the S sample, respectively. Among them, the sample of CS1 was characterized by phenethyl acetate and isoamyl acetate, while the samples of CS5 and PS5 were characterized by propyl octanoate and ethyl decanoate, and the sample of WS5 was characterized by 3-methylbutyl hexanoate. However, the contents of odor active compounds were higher in the CS1 sample. Therefore, the samples of CS1 had the potential to create the distinctive flavor of blueberry wine.


Introduction
Blueberries are known to taste both sour and sweet and have a mellow bouquet. Upon consumption, blueberries can not only supplement various vitamins and trace elements, but also enhance immune function and anti-aging, especially having an obvious protective effect on cardiovascular disease and renal failure [1]. However, blueberries are extremely susceptible to decay during storage and processing; thus, blueberry wine provides an alternative method to preserve nutrition and extend sales time [2,3]. However, the current strain used for blueberry wine is generally just Saccharomyces cerevisiae, which makes the flavor of blueberry wine relatively simple and cannot meet the demand of current consumers for blueberry wine with various flavors.
At present, ester-producing yeasts received significant attention for their role in both ethanol yield reduction and flavor enhancement in wine [4,5]. Studies found that esterproducing yeasts, belonging to the genera of Hanseniaspora, Candida, Metschnikowia, Pichia, Wickeramomyces, and Torulaspora, had some desired oenological features that were absent in S. cerevisiae, which could make up for the deficiency of pure fermentation of S. cerevisiae [6][7][8][9][10]. Ester-producing yeasts can change the fermentation environment, flavor, and quality of wine

Determination of Growth of Yeasts
Samples were taken every 24 h during the simulated fermentation of blueberry wine and diluted onto Wallerstein laboratory nutrient (WLN) medium (Hopebiol, Qingdao, China), then incubated at 28 °C for 2 d. Different yeast colonies on the WLN plates presented different morphologies and colors. Therefore, colonies on the WLN plate were counted according to the differences in colony morphologies, and the yeast populations were expressed as CFU/mL.

Determination of the Content of Reducing Sugar and Ethanol
Samples were taken every 24 h during the simulated fermentation of blueberry wine, and the contents of ethanol and reducing sugar were determined via alcoholmeter after distillation and Fehling's solution, according to the method of Yin et al. [25]. The contents of reducing sugar and ethanol were expressed as g/L and vol%, respectively.

Analysis of Ester Compounds
The ester compounds were analyzed by headspace solid-phase microextraction-gas chromatography-mass spectrometry (HS-SPME-GC-MS). A total of 10 mL sample was placed into a 20 mL headspace bottle, followed by the addition of 1.0 g sodium chloride and 10 μL 4-methyl-2-pentanol as internal standard with 10 μg/L. Then, the mixture was placed in a water bath at 45 °C for 30 min. Meanwhile, the ester compounds were collected by a divinyl-benzene carboxen on polydimethylsiloxane-coated (DVB/CAR/PDMS) fiber (Supelco Co., Bellefonte, PA, USA) for 40 min. After extraction, the fiber was inserted into the injection port of the gas chromatography-mass spectrometry (GC-MS) for desorption at 250 °C for 5 min.
The extracts were analyzed using GC-MS (Agilent Technologies 8860/5977B, Palo Alto, CA, USA) equipped with an HP-5MS column (30 m × 0.25 mm × 0.25 μm, J & W Scientific, Folsom, CA, USA). The condition of HS-SPME-GC-MS was based on the method described by Shi et al., with a little modification [26]. The conditions of GC were as follows: 35 °C for 3 min at an increment of 3 °C/min to 160 °C and maintaining a constant temperature for 2 min, then 8 °C/min to 230 °C and maintaining for 3 min. The inlet temperature was set at 250 °C. The conditions of MS were as follows: the ion source was electron ionization (EI), the ion source temperature was 250 °C, electron impact ionization mode was at 70 eV, the scanning range was 33-450 m/z, and the scanning mode was full

Determination of the Content of Reducing Sugar and Ethanol
Samples were taken every 24 h during the simulated fermentation of blueberry wine, and the contents of ethanol and reducing sugar were determined via alcoholmeter after distillation and Fehling's solution, according to the method of Yin et al. [25]. The contents of reducing sugar and ethanol were expressed as g/L and vol%, respectively.

Analysis of Ester Compounds
The ester compounds were analyzed by headspace solid-phase microextraction-gas chromatography-mass spectrometry (HS-SPME-GC-MS). A total of 10 mL sample was placed into a 20 mL headspace bottle, followed by the addition of 1.0 g sodium chloride and 10 µL 4-methyl-2-pentanol as internal standard with 10 µg/L. Then, the mixture was placed in a water bath at 45 • C for 30 min. Meanwhile, the ester compounds were collected by a divinyl-benzene carboxen on polydimethylsiloxane-coated (DVB/CAR/PDMS) fiber (Supelco Co., Bellefonte, PA, USA) for 40 min. After extraction, the fiber was inserted into the injection port of the gas chromatography-mass spectrometry (GC-MS) for desorption at 250 • C for 5 min.
The extracts were analyzed using GC-MS (Agilent Technologies 8860/5977B, Palo Alto, CA, USA) equipped with an HP-5MS column (30 m × 0.25 mm × 0.25 µm, J & W Scientific, Folsom, CA, USA). The condition of HS-SPME-GC-MS was based on the method described by Shi et al., with a little modification [26]. The conditions of GC were as follows: 35 • C for 3 min at an increment of 3 • C/min to 160 • C and maintaining a constant temperature for 2 min, then 8 • C/min to 230 • C and maintaining for 3 min. The inlet temperature was set at 250 • C. The conditions of MS were as follows: the ion source was electron ionization (EI), the ion source temperature was 250 • C, electron impact ionization mode was at 70 eV, the scanning range was 33-450 m/z, and the scanning mode was full scan. The qualitative and quantitative analyses were carried out by the method of Ulrich et al. [27].

Data Analysis
All the experiments were performed in triplicates. All the data were processed and analyzed by Excel 2013 and Origin 9.0. Any significant differences among variances were determined by the one-way analysis of variance (ANOVA), with Duncan test of the SPSS 19.0 software (p < 0.05, there was a significant difference between the data). Simultaneously, SIMCA software (Version 14.1) was used for principal component analysis (PCA), and R software (Version 3.6.1) was used for hierarchical cluster analysis (HCA) of ester compounds.

Growth of Yeasts during the Simulated Fermentation of Blueberry Wine
As presented in Figure 2A, the growth of S. cerevisiae NCUF309.2 in the S sample was slow in the first 3 d and rapid in 3-6 d, reaching the highest population of 3.43 × 10 8 CFU/mL on 6 d. However, in the co-fermentation samples, ester-producing yeasts could adapt quickly to the fermentation environment and reach a growth peak in the first 3 d (Figure 2). Anyway, the addition of ester-producing yeasts could promote the growth of S. cerevisiae NCUF309.2 in the first 3 d, especially in the CS1, PS1, PS5, WS1, and WS5 samples (Table S1).

Data Analysis
All the experiments were performed in triplicates. All the data were processed and analyzed by Excel 2013 and Origin 9.0. Any significant differences among variances were determined by the one-way analysis of variance (ANOVA), with Duncan test of the SPSS 19.0 software (p < 0.05, there was a significant difference between the data). Simultaneously, SIMCA software (Version 14.1) was used for principal component analysis (PCA), and R software (Version 3.6.1) was used for hierarchical cluster analysis (HCA) of ester compounds.

Growth of Yeasts during the Simulated Fermentation of Blueberry Wine
As presented in Figure 2A, the growth of S. cerevisiae NCUF309.2 in the S sample was slow in the first 3 d and rapid in 3-6 d, reaching the highest population of 3.43 × 10 8 CFU/mL on 6 d. However, in the co-fermentation samples, ester-producing yeasts could adapt quickly to the fermentation environment and reach a growth peak in the first 3 d ( Figure 2). Anyway, the addition of ester-producing yeasts could promote the growth of S. cerevisiae NCUF309.2 in the first 3 d, especially in the CS1, PS1, PS5, WS1, and WS5 samples (Table S1). In the co-fermentation samples of CS, the growth of S. cerevisiae NCUF309.2 was obviously affected in the CS5 and CS10 samples ( Figure 2B). Similarly, in the co-fermentation samples of PS ( Figure 2C), the population of S. cerevisiae NCUF309.2 gradually decreased with the increase of P. anomala NCUF306.1 inoculation ratio. The reason for these phenomena was that ester-producing yeast competed with S. cerevisiae NCUF309.2 for nutrients, and the competition increased with the inoculation ratio of ester-producing yeast [28]. Anyway, we also found that the growth of W. anomalus NCUF307.1 was immediately inhibited by S. cerevisiae NCUF309.2, with a sharp decrease in population during fermentation. The growth inhibition phenomenon of ester-producing yeasts was also reported in other studies [29,30]. For example, even with a high inoculation of Torulaspora delbrueckii (twenty times more than S. cerevisiae) in wine fermentation, S. cerevisiae still had a negative impact on the viability of T. delbrueckii [31]. This could be attributed to cell-cell contact that induced the death of ester-producing yeasts [32].
These results indicated that different ester-producing yeasts and their inoculation ratios had different effects on the growth of S. cerevisiae NCUF309.2, thereby its ability to produce ethanol could be affected. Therefore, the fermentation characteristics during the simulated fermentation of blueberry wine were also studied.

The Content of Reducing Sugar and Ethanol during the Simulated Fermentation of Blueberry Wine
As presented in Figure 3A, the content of reducing sugar in the S sample decreased rapidly during the first 5 d of fermentation, but the content of ethanol did not change too much during the first 3 d. After 3 d, the ethanol content increased sharply. Finally, the contents of reducing sugar and ethanol were 25.26 g/L and 12.52% (vol), respectively. These results indicated that the reducing sugar was mainly used for the growth of yeasts during the first 3 d of fermentation, and then was gradually converted to ethanol after 3 d [33].  In the co-fermentation samples, the content of reducing sugar was also consumed rapidly in the early stage of fermentation. Although the change of ethanol content in the co-fermentation samples was similar to that in the sample of S, the ethanol content in the co-fermentation samples was lower than that in the S sample ( Figure 3 and Figure 4), indicating that the use of ester-producing yeasts could reduce the ethanol production. In the co-fermentation samples of CS, the consumption of reducing sugar significantly increased  In the co-fermentation samples, the content of reducing sugar was also consumed rapidly in the early stage of fermentation. Although the change of ethanol content in the co-fermentation samples was similar to that in the sample of S, the ethanol content in the co-fermentation samples was lower than that in the S sample (Figures 3 and 4), indicating that the use of ester-producing yeasts could reduce the ethanol production. In the cofermentation samples of CS, the consumption of reducing sugar significantly increased in the CS1 sample, and the contents of ethanol decreased with the increases of the C. glabrata NCUF308.1 inoculation ratios, which were 11.33%, 9.83%, and 9.46%, respectively. Anyway, the contents of ethanol also decreased with the increases of P. anomala NCUF306.1 inoculation ratios, which were 12.22%, 11.56%, and 9.81%, respectively. Belda et al. also found that the final ethanol content was associated with the amount of ester-producing yeast [34]. This could be related to the decrease of the population of S. cerevisiae by esterproducing yeast in the samples, which affected the conversion of reducing sugars to ethanol [35]. However, the effects of W. anomalus NCUF307.1 on the final ethanol contents were not obvious, which were 12.01%, 11.95%, and 11.79%, respectively. This was accordant with the result of Ye et al., who found that S. cerevisiae showed the highest fermentability in the pure fermentation of S. cerevisiae with the production of 12.13% ethanol, while the ethanol contents ranged from 11.75% to 12.04% in all mixed fermentations with S. cerevisiae and W. anomalus [18].

Accumulation of Ester Compounds during the Simulated Fermentation of Blueberry Wine
Our results show that, compared with the pure fermentation of S. cerevisiae NCUF309.2, the co-fermentation of S. cerevisiae NCUF309.2 and C. glabrata NCUF308.1 or P. anomala NCUF306.1 could significantly increase the contents of ester compounds (p < 0.05), which were 3.29, 5.57, 2.34, 1.50, 4.75, and 2.41 times that of S sample, respectively ( Figure 5). This could be related to species belonging to the genera of Candida and Pichia, with a strong capacity to produce ester compounds [36]. Anyway, in the WS samples, the contents of ester compounds significantly increased only in the sample of WS5, which were 3.04 times that of the S sample. The production of ester compounds mainly occurred in the middle and late stages of fermentation. However, in the samples of WS1 and WS10, the population of W. anomalus NCUF307.1 in these stages was rare or even absent, resulting in a low content of ester compounds in the samples.

Accumulation of Ester Compounds during the Simulated Fermentation of Blueberry Wine
Our results show that, compared with the pure fermentation of S. cerevisiae NCUF309.2, the co-fermentation of S. cerevisiae NCUF309.2 and C. glabrata NCUF308.1 or P. anomala NCUF306.1 could significantly increase the contents of ester compounds (p < 0.05), which were 3.29, 5.57, 2.34, 1.50, 4.75, and 2.41 times that of S sample, respectively ( Figure 5). This could be related to species belonging to the genera of Candida and Pichia, with a strong capacity to produce ester compounds [36]. Anyway, in the WS samples, the contents of ester compounds significantly increased only in the sample of WS5, which were 3.04 times that of the S sample. The production of ester compounds mainly occurred in the middle and late stages of fermentation. However, in the samples of WS1 and WS10, the population of W. anomalus NCUF307.1 in these stages was rare or even absent, resulting in a low content of ester compounds in the samples.
Our results show that, compared with the pure fermentation of S. cerevisiae NCUF309.2, the co-fermentation of S. cerevisiae NCUF309.2 and C. glabrata NCUF308.1 or P. anomala NCUF306.1 could significantly increase the contents of ester compounds (p < 0.05), which were 3.29, 5.57, 2.34, 1.50, 4.75, and 2.41 times that of S sample, respectively ( Figure 5). This could be related to species belonging to the genera of Candida and Pichia, with a strong capacity to produce ester compounds [36]. Anyway, in the WS samples, the contents of ester compounds significantly increased only in the sample of WS5, which were 3.04 times that of the S sample. The production of ester compounds mainly occurred in the middle and late stages of fermentation. However, in the samples of WS1 and WS10, the population of W. anomalus NCUF307.1 in these stages was rare or even absent, resulting in a low content of ester compounds in the samples. Ester compounds can be further categorized into acetate esters, ethyl esters, and other esters [37]. Both acetate esters and ethyl esters are of key importance in the whole wine flavor, impressing a positive contribution by distinct sensory notes: sweet-fruity, grape-like odor, and sweet-balsamic [38]. In the co-fermentation samples of CS, the contents of acetate esters and ethyl esters both significantly increased, and the content of ethyl esters was higher in the CS5 sample, while the content of acetate esters was higher in the CS1 sample ( Figure 5). Additionally, in the co-fermentation samples of PS, the content of ethyl esters was higher in the PS5 sample, while the content of acetate esters was higher in the PS10 sample. These results indicated that the ratio of ester-producing yeast co-fermentation with S. cerevisiae had an effect on the formation of ester compounds during fermentation. Lee et al. also found that W. saturnus and S. cerevisiae could produce a higher content of acetate esters at the ratio of 10:1, while they could produce more ethyl esters at the ratio of 1:1 and 1:10 [39]. In the co-fermentation samples of WS, the contents of acetate esters and ethyl esters both significantly increased in only the WS5 sample. Anyway, the contents of other esters were also significantly increased in the co-fermentation samples (except the samples of PS1 and WS10), and the highest content was in the WS5 sample.
As for acetate esters, ethyl acetate, phenethyl acetate, and isoamyl acetate were the main acetate esters (Table 1). Additionally, their contents both increased in the cofermentation samples of CS and PS, while both increased in the co-fermentation samples of WS5 and WS10. Hu et al. also found the contents of ethyl acetate, isoamyl acetate, and phenethyl acetate increased in the co-fermentation samples, compared with those of S. cerevisiae pure fermentation [40]. Ethyl acetate could improve the aroma complexity of fruit wine, and phenethyl acetate and isoamyl acetate were recognized as important flavor compounds to contribute to the fruity flavor of fruit wine [41,42]. These results suggested that co-fermentation could improve the acetate contents, especially the contents of ethyl acetate, isoamyl acetate, and phenethyl acetates, compared with those of S. cerevisiae pure fermentation, which would be favorable to the flavor of blueberry wine. As for ethyl esters, medium-chain fatty acid ethyl esters were the main ethyl esters, such as ethyl hexanoate, ethyl octanoate, and ethyl decanoate. Ethyl octanoate has a fruity flavor and is found in many fruits, while ethyl decanoate imparts a fruity and brandy-like flavor to wine [43]. Additionally, the contents of these compounds both significantly increased in the co-fermentation samples of CS and PS, while their contents only significantly increased in the WS5 sample. These results were consistent with the result of a previous study, i.e., that the co-fermentation of T. delbrueckii and S. cerevisiae had higher contents of ethyl hexanoate, ethyl octanoate, and ethyl decanoate than those of S. cerevisiae pure fermentation [44]. In addition, the results also showed that the content of ethyl laurate also increased in the co-fermentation samples of CS and PS. As for other esters, the content of 3-methylbutyl hexanoate increased in the co-fermentation samples of WS, and the content of 3-methylbutyl decanoate increased in the co-fermentation samples of CS. All of these compounds could give the wine more complex flavor characteristics.
These results indicated that different ratios and strains of ester-producing yeast cofermentation with S. cerevisiae could lead to wines with a wide range of ester compounds, which was similar to the results of Comitini et al. [45].

Principal Component Analysis (PCA) of Ester Compounds during the Simulated Fermentation of Blueberry Wine
PCA was used to reveal the correlation and separation of ester compounds in different samples ( Figure 6). The variance of 64.8% was explained by 29 different ester compounds, and PC1 and PC2 accounted for 38.8% and 26.0% of the variance, respectively. The PCA results showed that the co-fermentation samples of PS1, WS1, and WS10 were close to S sample, suggesting that they had similar ester compound profiles. In these samples, the effects of ester-producing yeast were not obvious, and the characteristics of ester compounds were not prominent. The co-fermentation samples of CS1, CS5, PS5, and WS5 were separated from the S sample and had their own characteristic ester compounds. For example, the sample of CS1 was characterized by phenethyl acetate and isoamyl acetate, the samples of CS5 and PS5 were characterized by propyl octanoate and ethyl decanoate, and the sample of WS5 was characterized by 3-methylbutyl hexanoate, which provided the potential for creating a distinctive flavor of blueberry wine.

Principal Component Analysis (PCA) of Ester Compounds during the Simulated Fermentation of Blueberry Wine
PCA was used to reveal the correlation and separation of ester compounds in different samples ( Figure 6). The variance of 64.8% was explained by 29 different ester compounds, and PC1 and PC2 accounted for 38.8% and 26.0% of the variance, respectively. The PCA results showed that the co-fermentation samples of PS1, WS1, and WS10 were close to S sample, suggesting that they had similar ester compound profiles. In these samples, the effects of ester-producing yeast were not obvious, and the characteristics of ester compounds were not prominent. The co-fermentation samples of CS1, CS5, PS5, and WS5 were separated from the S sample and had their own characteristic ester compounds. For example, the sample of CS1 was characterized by phenethyl acetate and isoamyl acetate, the samples of CS5 and PS5 were characterized by propyl octanoate and ethyl decanoate, and the sample of WS5 was characterized by 3-methylbutyl hexanoate, which provided the potential for creating a distinctive flavor of blueberry wine.

Volatile Odor Active Ester Compounds during the Simulated Fermentation of Blueberry Wine
The odor activity value (OAV) is a common index used to evaluate the contribution of volatile compounds of wine to the actual flavor, which can be calculated by the ratio of the content of the volatile compound to its odor

Volatile Odor Active Ester Compounds during the Simulated Fermentation of Blueberry Wine
The odor activity value (OAV) is a common index used to evaluate the contribution of volatile compounds of wine to the actual flavor, which can be calculated by the ratio of the content of the volatile compound to its odor threshold [46]. It is generally believed that an OAV greater than 1 indicates that it contributes to the odor, and a larger OAV indicates a greater individual contribution of the compound [47]. The ester compounds with OAV greater than 1 are shown in Table 2. In the S sample, the OAVs of phenethyl acetate and isoamyl acetate were lower than 1; however, the OAVs of phenethyl acetate and isoamyl acetate were greater than 1 in the CS1, CS5, CS10, PS5, PS10, WS5, and WS10 samples, indicating that co-fermentation directly promoted the contribution of phenethyl acetate and isoamyl acetate to the fruity and floral flavors. Although the content of isoamyl acetate was lower than that of phenethyl acetate, its threshold was lower, leading to a higher OAV than phenethyl acetate. The OAV of phenethyl acetate and isoamyl acetate were both the highest in the sample of CS1, which was 9.68 and 32.33 times that of the S sample, suggesting that they could contribute to more floral and fruity flavors in the sample of CS1.
As for ethyl esters, ethyl octanoate, with a high content and low threshold (5 µg/L), had the greatest influence on flavor of fruit wine and could give wine flavors of pineapple, pear, and floral [48]. The OAV of ethyl hexanoate with the same low threshold (5 µg/L) was second only to ethyl octanoate and could give wines with green apple and purple apple flavors. The OAVs of ethyl octanoate and ethyl hexanoate were the highest in the CS1 sample, which were 1350.05 and 116.43. In addition, ethyl decanoate also had significant effects on the overall flavor and could directly promote the fruity and floral flavors of samples [49]. Anyway, the OAV of ethyl laurate was greater than 1 in only the samples of CS1, CS5, PS1, and PS5, thus directly giving these samples sweet, waxiness, and floral flavors.

Conclusions
In this study, S. cerevisiae NCUF309.2 was fermented with three different ester-producing yeasts (C. glabrata NCUF308.1, P. anomala NCUF306.1, and W. anomalus NCUF307.1) under different inoculation ratios to simulate the co-fermentation of blueberry wine. The growth of the yeasts, fermentation characteristics, and the accumulation of ester compounds during the fermentation were studied. The results show that, compared with the S sample, the presence of P. anomala NCUF306.1 and C. glabrata NCUF308.1 could affect the growth of S. cerevisiae NCUF309.2 and decrease the population of S. cerevisiae NCUF309.2, leading to a decrease in the content of ethanol. However, the growth of S. cerevisiae NCUF309.2 was not affected in the sample of WS5, and the presence of W. anomalus NCUF307.1 had little effect on ethanol content. The results also showed that, compared with the S sample, the co-fermentation of S. cerevisiae NCUF309.2 with ester-producing yeast could produce wine with distinct flavor characteristics (Figure 6), especially in the samples of CS1, PS5, and WS5. Additionally, the contents of ester compounds in these samples were significantly increased. Among them, the CS1 sample had the highest contents of odor active compounds, including phenethyl acetate, isoamyl acetate, ethyl octanoate, and ethyl hexanoate, which could directly endow fruit wine with more floral and fruity flavor, compared with other samples. In conclusion, the co-fermentation of C. glabrata NCUF308.1 and S. cerevisiae NCUF309.2 at the ratio of 1:1 was a feasible method for improving the aroma diversity of blueberry wine and meet the requirements of consumers.

Data Availability Statement:
The datasets generated for this study are available on request to the corresponding author.